From d1e879ec600f9b3bdd253167533959facfefb17b Mon Sep 17 00:00:00 2001 From: Miri Korenblit Date: Sun, 16 Feb 2025 11:42:40 +0200 Subject: [PATCH] wifi: iwlwifi: add iwlmld sub-driver iwlwifi is the driver of all Intel wifi devices since 2008. Since then, the hardware has changed a lot, but the firmware API has changed even more. The need to keep one driver that supports all those different APIs led us to introduce a new architecture circa 2012 which allowed us to keep the same interface to the hardware (DMAs, Tx queues, etc...) with a new layer to implement the mid-layer between mac80211 and the firmware. The first component is called the 'transport' and the latter is called 'operation_mode' a.k.a op_mode. In 2013 we took advantage of the new architecture to introduce iwlmvm which allowed us to implement the, then, new firmware API. This op_mode supports 7260 and up, those devices supports support at least VHT. Since then, wifi evolved and so did the firmware. It became much bigger and took a lot of functionality from the driver. It became increasingly hard to keep the same op_mode for the newest devices and we experienced frequent regressions on older devices. In order to avoid those regressions and keep the code maintainable, we decided it was about time to start a new op_mode. iwlmld is a new op_mode that supports BE200 or newer if the firmware being used is 97.ucode or newer. If the user has an older devices or BE200 with .96.ucode, iwlmvm will be loaded. Of course, this op_mode selection is seamless. All the features supported in iwlmvm are supported in iwlmld besides a few seldom used use cases: injection and Hotspot 2.0. Those are under work. A few points about the implementation: * iwlmld doesn't have any mutexes, it relies on the wiphy_lock * iwlmld is more "resource oriented": stations, links and interfaces are allocated and freed only after all the relevant flows are completed. * Firmware notifications' sizes are validated in a more structured way. We would love to see this new op_mode merged in 6.15. The firmware for this new driver (.97.ucode) is not yet publicly available but it'll be sent very soon. People eager to get an early version of this firmware can contact Emmanuel at: emmanuel.grumbach@intel.com I've listed the people who directly contributed code, but many others from various teams have contributed in other ways. Co-developed-by: Johannes Berg Signed-off-by: Johannes Berg Co-developed-by: Avraham Stern Signed-off-by: Avraham Stern Co-developed-by: Daniel Gabay Signed-off-by: Daniel Gabay Co-developed-by: Emmanuel Grumbach Signed-off-by: Emmanuel Grumbach Co-developed-by: Anjaneyulu Signed-off-by: Anjaneyulu Co-developed-by: Yedidya Benshimol Signed-off-by: Yedidya Benshimol Co-developed-by: Benjamin Berg Signed-off-by: Benjamin Berg Co-developed-by: Shaul Triebitz Signed-off-by: Shaul Triebitz Signed-off-by: Miri Korenblit Link: https://lore.kernel.org/linux-wireless/20250216094321.537988-1-miriam.rachel.korenblit@intel.com/ [fix Kconfig, fix api/phy.h includes, SPDX tag and coding style issues, duplicated includes per 0-day robot] Signed-off-by: Johannes Berg --- drivers/net/wireless/intel/iwlwifi/Kconfig | 15 +- drivers/net/wireless/intel/iwlwifi/Makefile | 5 +- drivers/net/wireless/intel/iwlwifi/cfg/bz.c | 5 + .../net/wireless/intel/iwlwifi/fw/api/phy.h | 4 +- .../net/wireless/intel/iwlwifi/iwl-config.h | 6 +- drivers/net/wireless/intel/iwlwifi/iwl-drv.c | 28 +- .../net/wireless/intel/iwlwifi/mld/Makefile | 16 + drivers/net/wireless/intel/iwlwifi/mld/agg.c | 670 +++++ drivers/net/wireless/intel/iwlwifi/mld/agg.h | 127 + drivers/net/wireless/intel/iwlwifi/mld/ap.c | 344 +++ drivers/net/wireless/intel/iwlwifi/mld/ap.h | 45 + drivers/net/wireless/intel/iwlwifi/mld/coex.c | 40 + drivers/net/wireless/intel/iwlwifi/mld/coex.h | 15 + .../wireless/intel/iwlwifi/mld/constants.h | 88 + drivers/net/wireless/intel/iwlwifi/mld/d3.c | 1998 ++++++++++++ drivers/net/wireless/intel/iwlwifi/mld/d3.h | 51 + .../net/wireless/intel/iwlwifi/mld/debugfs.c | 823 +++++ .../net/wireless/intel/iwlwifi/mld/debugfs.h | 200 ++ .../intel/iwlwifi/mld/ftm-initiator.c | 451 +++ .../intel/iwlwifi/mld/ftm-initiator.h | 15 + drivers/net/wireless/intel/iwlwifi/mld/fw.c | 536 ++++ drivers/net/wireless/intel/iwlwifi/mld/hcmd.h | 54 + .../net/wireless/intel/iwlwifi/mld/iface.c | 671 +++++ .../net/wireless/intel/iwlwifi/mld/iface.h | 234 ++ drivers/net/wireless/intel/iwlwifi/mld/key.c | 358 +++ drivers/net/wireless/intel/iwlwifi/mld/key.h | 39 + drivers/net/wireless/intel/iwlwifi/mld/led.c | 100 + drivers/net/wireless/intel/iwlwifi/mld/led.h | 29 + drivers/net/wireless/intel/iwlwifi/mld/link.c | 1216 ++++++++ drivers/net/wireless/intel/iwlwifi/mld/link.h | 153 + .../wireless/intel/iwlwifi/mld/low_latency.c | 335 +++ .../wireless/intel/iwlwifi/mld/low_latency.h | 68 + .../net/wireless/intel/iwlwifi/mld/mac80211.c | 2671 +++++++++++++++++ .../net/wireless/intel/iwlwifi/mld/mac80211.h | 13 + drivers/net/wireless/intel/iwlwifi/mld/mcc.c | 329 ++ drivers/net/wireless/intel/iwlwifi/mld/mcc.h | 17 + drivers/net/wireless/intel/iwlwifi/mld/mld.c | 707 +++++ drivers/net/wireless/intel/iwlwifi/mld/mld.h | 584 ++++ drivers/net/wireless/intel/iwlwifi/mld/mlo.c | 994 ++++++ drivers/net/wireless/intel/iwlwifi/mld/mlo.h | 157 + .../net/wireless/intel/iwlwifi/mld/notif.c | 758 +++++ .../net/wireless/intel/iwlwifi/mld/notif.h | 35 + drivers/net/wireless/intel/iwlwifi/mld/phy.c | 116 + drivers/net/wireless/intel/iwlwifi/mld/phy.h | 49 + .../net/wireless/intel/iwlwifi/mld/power.c | 395 +++ .../net/wireless/intel/iwlwifi/mld/power.h | 33 + drivers/net/wireless/intel/iwlwifi/mld/ptp.c | 321 ++ drivers/net/wireless/intel/iwlwifi/mld/ptp.h | 45 + .../wireless/intel/iwlwifi/mld/regulatory.c | 393 +++ .../wireless/intel/iwlwifi/mld/regulatory.h | 23 + drivers/net/wireless/intel/iwlwifi/mld/roc.c | 224 ++ drivers/net/wireless/intel/iwlwifi/mld/roc.h | 20 + drivers/net/wireless/intel/iwlwifi/mld/rx.c | 2060 +++++++++++++ drivers/net/wireless/intel/iwlwifi/mld/rx.h | 72 + drivers/net/wireless/intel/iwlwifi/mld/scan.c | 2006 +++++++++++++ drivers/net/wireless/intel/iwlwifi/mld/scan.h | 135 + .../intel/iwlwifi/mld/session-protect.c | 222 ++ .../intel/iwlwifi/mld/session-protect.h | 102 + drivers/net/wireless/intel/iwlwifi/mld/sta.c | 1265 ++++++++ drivers/net/wireless/intel/iwlwifi/mld/sta.h | 266 ++ .../net/wireless/intel/iwlwifi/mld/stats.c | 509 ++++ .../net/wireless/intel/iwlwifi/mld/stats.h | 22 + .../wireless/intel/iwlwifi/mld/tests/Makefile | 5 + .../wireless/intel/iwlwifi/mld/tests/agg.c | 663 ++++ .../wireless/intel/iwlwifi/mld/tests/hcmd.c | 62 + .../intel/iwlwifi/mld/tests/link-selection.c | 171 ++ .../wireless/intel/iwlwifi/mld/tests/link.c | 103 + .../wireless/intel/iwlwifi/mld/tests/module.c | 11 + .../net/wireless/intel/iwlwifi/mld/tests/rx.c | 353 +++ .../wireless/intel/iwlwifi/mld/tests/utils.c | 475 +++ .../wireless/intel/iwlwifi/mld/tests/utils.h | 124 + .../net/wireless/intel/iwlwifi/mld/thermal.c | 438 +++ .../net/wireless/intel/iwlwifi/mld/thermal.h | 36 + .../wireless/intel/iwlwifi/mld/time_sync.c | 240 ++ .../wireless/intel/iwlwifi/mld/time_sync.h | 26 + drivers/net/wireless/intel/iwlwifi/mld/tlc.c | 717 +++++ drivers/net/wireless/intel/iwlwifi/mld/tlc.h | 23 + drivers/net/wireless/intel/iwlwifi/mld/tx.c | 1374 +++++++++ drivers/net/wireless/intel/iwlwifi/mld/tx.h | 77 + drivers/net/wireless/intel/iwlwifi/pcie/drv.c | 9 +- 80 files changed, 28178 insertions(+), 11 deletions(-) create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/Makefile create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/agg.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/agg.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/ap.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/ap.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/coex.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/coex.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/constants.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/d3.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/d3.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/debugfs.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/debugfs.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/ftm-initiator.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/ftm-initiator.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/fw.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/hcmd.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/iface.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/iface.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/key.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/key.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/led.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/led.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/link.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/link.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/low_latency.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/low_latency.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/mac80211.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/mac80211.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/mcc.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/mcc.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/mld.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/mld.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/mlo.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/mlo.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/notif.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/notif.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/phy.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/phy.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/power.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/power.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/ptp.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/ptp.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/regulatory.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/regulatory.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/roc.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/roc.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/rx.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/rx.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/scan.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/scan.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/session-protect.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/session-protect.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/sta.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/sta.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/stats.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/stats.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/tests/Makefile create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/tests/agg.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/tests/hcmd.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/tests/link-selection.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/tests/link.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/tests/module.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/tests/rx.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/tests/utils.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/tests/utils.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/thermal.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/thermal.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/time_sync.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/time_sync.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/tlc.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/tlc.h create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/tx.c create mode 100644 drivers/net/wireless/intel/iwlwifi/mld/tx.h diff --git a/drivers/net/wireless/intel/iwlwifi/Kconfig b/drivers/net/wireless/intel/iwlwifi/Kconfig index 4b04865fc2c9e..82f577da1a8b5 100644 --- a/drivers/net/wireless/intel/iwlwifi/Kconfig +++ b/drivers/net/wireless/intel/iwlwifi/Kconfig @@ -81,14 +81,25 @@ config IWLMVM of the devices that use this firmware is available here: https://wireless.wiki.kernel.org/en/users/drivers/iwlwifi#firmware +config IWLMLD + tristate "Intel Wireless WiFi MLD Firmware support" + select WANT_DEV_COREDUMP + depends on MAC80211 + depends on PTP_1588_CLOCK_OPTIONAL + help + This is the driver that supports firmwares of MLD capable devices. + The list of the devices that use this firmware is available here: + https://wireless.wiki.kernel.org/en/users/drivers/iwlwifi#firmware + # don't call it _MODULE -- will confuse Kconfig/fixdep/... config IWLWIFI_OPMODE_MODULAR bool default y if IWLDVM=m default y if IWLMVM=m + default y if IWLMLD=m -comment "WARNING: iwlwifi is useless without IWLDVM or IWLMVM" - depends on IWLDVM=n && IWLMVM=n +comment "WARNING: iwlwifi is useless without IWLDVM or IWLMVM or IWLMLD" + depends on IWLDVM=n && IWLMVM=n && IWLMLD=n menu "Debugging Options" diff --git a/drivers/net/wireless/intel/iwlwifi/Makefile b/drivers/net/wireless/intel/iwlwifi/Makefile index 19c4ce6f2465f..9546ceeaf5e37 100644 --- a/drivers/net/wireless/intel/iwlwifi/Makefile +++ b/drivers/net/wireless/intel/iwlwifi/Makefile @@ -12,7 +12,8 @@ iwlwifi-objs += pcie/ctxt-info.o pcie/ctxt-info-gen3.o iwlwifi-objs += pcie/trans-gen2.o pcie/tx-gen2.o iwlwifi-$(CONFIG_IWLDVM) += cfg/1000.o cfg/2000.o cfg/5000.o cfg/6000.o iwlwifi-$(CONFIG_IWLMVM) += cfg/7000.o cfg/8000.o cfg/9000.o cfg/22000.o -iwlwifi-$(CONFIG_IWLMVM) += cfg/ax210.o cfg/bz.o cfg/sc.o cfg/dr.o +iwlwifi-$(CONFIG_IWLMVM) += cfg/ax210.o +iwlwifi-$(CONFIG_IWLMLD) += cfg/bz.o cfg/sc.o cfg/dr.o iwlwifi-objs += iwl-dbg-tlv.o iwlwifi-objs += iwl-trans.o @@ -20,6 +21,7 @@ iwlwifi-objs += fw/img.o fw/notif-wait.o fw/rs.o iwlwifi-objs += fw/dbg.o fw/pnvm.o fw/dump.o iwlwifi-objs += fw/regulatory.o iwlwifi-$(CONFIG_IWLMVM) += fw/paging.o fw/smem.o fw/init.o +iwlwifi-$(CONFIG_IWLMLD) += fw/smem.o fw/init.o iwlwifi-$(CONFIG_ACPI) += fw/acpi.o iwlwifi-$(CONFIG_EFI) += fw/uefi.o iwlwifi-$(CONFIG_IWLWIFI_DEBUGFS) += fw/debugfs.o @@ -33,6 +35,7 @@ ccflags-y += -I$(src) obj-$(CONFIG_IWLDVM) += dvm/ obj-$(CONFIG_IWLMVM) += mvm/ obj-$(CONFIG_IWLMEI) += mei/ +obj-$(CONFIG_IWLMLD) += mld/ obj-$(CONFIG_IWLWIFI_KUNIT_TESTS) += tests/ diff --git a/drivers/net/wireless/intel/iwlwifi/cfg/bz.c b/drivers/net/wireless/intel/iwlwifi/cfg/bz.c index 64a5bbb45c836..36e8e3a991dcb 100644 --- a/drivers/net/wireless/intel/iwlwifi/cfg/bz.c +++ b/drivers/net/wireless/intel/iwlwifi/cfg/bz.c @@ -38,6 +38,11 @@ #define IWL_BZ_A_HR_B_MODULE_FIRMWARE(api) \ IWL_BZ_A_HR_B_FW_PRE "-" __stringify(api) ".ucode" +#if !IS_ENABLED(CONFIG_IWLMVM) +const char iwl_ax211_name[] = "Intel(R) Wi-Fi 6E AX211 160MHz"; +const char iwl_ax201_name[] = "Intel(R) Wi-Fi 6 AX201 160MHz"; +#endif + static const struct iwl_base_params iwl_bz_base_params = { .eeprom_size = OTP_LOW_IMAGE_SIZE_32K, .num_of_queues = 512, diff --git a/drivers/net/wireless/intel/iwlwifi/fw/api/phy.h b/drivers/net/wireless/intel/iwlwifi/fw/api/phy.h index eb8961b51cb01..f63b25b03b7e4 100644 --- a/drivers/net/wireless/intel/iwlwifi/fw/api/phy.h +++ b/drivers/net/wireless/intel/iwlwifi/fw/api/phy.h @@ -1,11 +1,13 @@ /* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ /* - * Copyright (C) 2012-2014, 2019-2022, 2024 Intel Corporation + * Copyright (C) 2012-2014, 2019-2022, 2024-2025 Intel Corporation * Copyright (C) 2013-2015 Intel Mobile Communications GmbH * Copyright (C) 2016-2017 Intel Deutschland GmbH */ #ifndef __iwl_fw_api_phy_h__ #define __iwl_fw_api_phy_h__ +#include +#include /** * enum iwl_phy_ops_subcmd_ids - PHY group commands diff --git a/drivers/net/wireless/intel/iwlwifi/iwl-config.h b/drivers/net/wireless/intel/iwlwifi/iwl-config.h index 5fc9ce2b350e9..7e4864c00780b 100644 --- a/drivers/net/wireless/intel/iwlwifi/iwl-config.h +++ b/drivers/net/wireless/intel/iwlwifi/iwl-config.h @@ -653,6 +653,10 @@ extern const struct iwl_cfg iwl_cfg_ma; extern const struct iwl_cfg iwl_cfg_so_a0_hr_a0; extern const struct iwl_cfg iwl_cfg_quz_a0_hr_b0; +#endif /* CONFIG_IWLMVM */ + +#if IS_ENABLED(CONFIG_IWLMLD) +extern const struct iwl_ht_params iwl_bz_ht_params; extern const struct iwl_ht_params iwl_bz_ht_params; @@ -664,6 +668,6 @@ extern const struct iwl_cfg iwl_cfg_sc2; extern const struct iwl_cfg iwl_cfg_sc2f; extern const struct iwl_cfg iwl_cfg_dr; extern const struct iwl_cfg iwl_cfg_br; -#endif /* CONFIG_IWLMVM */ +#endif /* CONFIG_IWLMLD */ #endif /* __IWL_CONFIG_H__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/iwl-drv.c b/drivers/net/wireless/intel/iwlwifi/iwl-drv.c index d3a65f33097cb..7606780c59deb 100644 --- a/drivers/net/wireless/intel/iwlwifi/iwl-drv.c +++ b/drivers/net/wireless/intel/iwlwifi/iwl-drv.c @@ -75,6 +75,9 @@ struct iwl_drv { enum { DVM_OP_MODE, MVM_OP_MODE, +#if IS_ENABLED(CONFIG_IWLMLD) + MLD_OP_MODE, +#endif }; /* Protects the table contents, i.e. the ops pointer & drv list */ @@ -86,6 +89,9 @@ static struct iwlwifi_opmode_table { } iwlwifi_opmode_table[] = { /* ops set when driver is initialized */ [DVM_OP_MODE] = { .name = "iwldvm", .ops = NULL }, [MVM_OP_MODE] = { .name = "iwlmvm", .ops = NULL }, +#if IS_ENABLED(CONFIG_IWLMLD) + [MLD_OP_MODE] = { .name = "iwlmld", .ops = NULL }, +#endif }; #define IWL_DEFAULT_SCAN_CHANNELS 40 @@ -316,6 +322,7 @@ struct iwl_firmware_pieces { size_t dbg_trigger_tlv_len[FW_DBG_TRIGGER_MAX]; struct iwl_fw_dbg_mem_seg_tlv *dbg_mem_tlv; size_t n_mem_tlv; + u32 major; }; static void alloc_sec_data(struct iwl_firmware_pieces *pieces, @@ -957,19 +964,19 @@ static int iwl_parse_tlv_firmware(struct iwl_drv *drv, break; case IWL_UCODE_TLV_FW_VERSION: { const __le32 *ptr = (const void *)tlv_data; - u32 major, minor; + u32 minor; u8 local_comp; if (tlv_len != sizeof(u32) * 3) goto invalid_tlv_len; - major = le32_to_cpup(ptr++); + pieces->major = le32_to_cpup(ptr++); minor = le32_to_cpup(ptr++); local_comp = le32_to_cpup(ptr); snprintf(drv->fw.fw_version, sizeof(drv->fw.fw_version), - "%u.%08x.%u %s", major, minor, + "%u.%08x.%u %s", pieces->major, minor, local_comp, iwl_reduced_fw_name(drv)); break; } @@ -1468,6 +1475,8 @@ static void _iwl_op_mode_stop(struct iwl_drv *drv) } } +#define IWL_MLD_SUPPORTED_FW_VERSION 97 + /* * iwl_req_fw_callback - callback when firmware was loaded * @@ -1720,6 +1729,19 @@ static void iwl_req_fw_callback(const struct firmware *ucode_raw, void *context) break; } +#if IS_ENABLED(CONFIG_IWLMLD) + if (pieces->major >= IWL_MLD_SUPPORTED_FW_VERSION) + op = &iwlwifi_opmode_table[MLD_OP_MODE]; + else +#else + if (pieces->major >= IWL_MLD_SUPPORTED_FW_VERSION) { + IWL_ERR(drv, + "IWLMLD needs to be compiled to support this firmware\n"); + mutex_unlock(&iwlwifi_opmode_table_mtx); + goto out_unbind; + } +#endif + IWL_INFO(drv, "loaded firmware version %s op_mode %s\n", drv->fw.fw_version, op->name); diff --git a/drivers/net/wireless/intel/iwlwifi/mld/Makefile b/drivers/net/wireless/intel/iwlwifi/mld/Makefile new file mode 100644 index 0000000000000..ece66e7a9be41 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/Makefile @@ -0,0 +1,16 @@ +# SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +obj-$(CONFIG_IWLMLD) += iwlmld.o +obj-$(CONFIG_IWLWIFI_KUNIT_TESTS) += tests/ + +iwlmld-y += mld.o notif.o mac80211.o fw.o power.o iface.o link.o rx.o mcc.o session-protect.o phy.o +iwlmld-y += scan.o sta.o tx.o coex.o tlc.o agg.o key.o regulatory.o ap.o thermal.o roc.o stats.o +iwlmld-y += low_latency.o mlo.o ptp.o time_sync.o ftm-initiator.o +iwlmld-$(CONFIG_IWLWIFI_DEBUGFS) += debugfs.o +iwlmld-$(CONFIG_IWLWIFI_LEDS) += led.o +iwlmld-$(CONFIG_PM_SLEEP) += d3.o + +# non-upstream things +iwlmld-$(CONFIG_IWL_VENDOR_CMDS) += vendor-cmd.o +iwlmld-$(CONFIG_IWLMVM_AX_SOFTAP_TESTMODE) += ax-softap-testmode.o + +subdir-ccflags-y += -I$(src)/../ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/agg.c b/drivers/net/wireless/intel/iwlwifi/mld/agg.c new file mode 100644 index 0000000000000..db9e0f04f4b77 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/agg.c @@ -0,0 +1,670 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024-2025 Intel Corporation + */ +#include "agg.h" +#include "sta.h" +#include "hcmd.h" + +static void +iwl_mld_reorder_release_frames(struct iwl_mld *mld, struct ieee80211_sta *sta, + struct napi_struct *napi, + struct iwl_mld_baid_data *baid_data, + struct iwl_mld_reorder_buffer *reorder_buf, + u16 nssn) +{ + struct iwl_mld_reorder_buf_entry *entries = + &baid_data->entries[reorder_buf->queue * + baid_data->entries_per_queue]; + u16 ssn = reorder_buf->head_sn; + + while (ieee80211_sn_less(ssn, nssn)) { + int index = ssn % baid_data->buf_size; + struct sk_buff_head *skb_list = &entries[index].frames; + struct sk_buff *skb; + + ssn = ieee80211_sn_inc(ssn); + + /* Empty the list. Will have more than one frame for A-MSDU. + * Empty list is valid as well since nssn indicates frames were + * received. + */ + while ((skb = __skb_dequeue(skb_list))) { + iwl_mld_pass_packet_to_mac80211(mld, napi, skb, + reorder_buf->queue, + sta); + reorder_buf->num_stored--; + } + } + reorder_buf->head_sn = nssn; +} + +static void iwl_mld_release_frames_from_notif(struct iwl_mld *mld, + struct napi_struct *napi, + u8 baid, u16 nssn, int queue) +{ + struct iwl_mld_reorder_buffer *reorder_buf; + struct iwl_mld_baid_data *ba_data; + struct ieee80211_link_sta *link_sta; + u32 sta_id; + + IWL_DEBUG_HT(mld, "Frame release notification for BAID %u, NSSN %d\n", + baid, nssn); + + if (WARN_ON_ONCE(baid == IWL_RX_REORDER_DATA_INVALID_BAID || + baid >= ARRAY_SIZE(mld->fw_id_to_ba))) + return; + + rcu_read_lock(); + + ba_data = rcu_dereference(mld->fw_id_to_ba[baid]); + if (!ba_data) { + IWL_DEBUG_HT(mld, "BAID %d not found in map\n", baid); + goto out_unlock; + } + + /* pick any STA ID to find the pointer */ + sta_id = ffs(ba_data->sta_mask) - 1; + link_sta = rcu_dereference(mld->fw_id_to_link_sta[sta_id]); + if (WARN_ON_ONCE(IS_ERR_OR_NULL(link_sta) || !link_sta->sta)) + goto out_unlock; + + reorder_buf = &ba_data->reorder_buf[queue]; + + iwl_mld_reorder_release_frames(mld, link_sta->sta, napi, ba_data, + reorder_buf, nssn); +out_unlock: + rcu_read_unlock(); +} + +void iwl_mld_handle_frame_release_notif(struct iwl_mld *mld, + struct napi_struct *napi, + struct iwl_rx_packet *pkt, int queue) +{ + struct iwl_frame_release *release = (void *)pkt->data; + u32 pkt_len = iwl_rx_packet_payload_len(pkt); + + if (IWL_FW_CHECK(mld, pkt_len < sizeof(*release), + "Unexpected frame release notif size %u (expected %zu)\n", + pkt_len, sizeof(*release))) + return; + + iwl_mld_release_frames_from_notif(mld, napi, release->baid, + le16_to_cpu(release->nssn), + queue); +} + +void iwl_mld_handle_bar_frame_release_notif(struct iwl_mld *mld, + struct napi_struct *napi, + struct iwl_rx_packet *pkt, + int queue) +{ + struct iwl_bar_frame_release *release = (void *)pkt->data; + struct iwl_mld_baid_data *baid_data; + unsigned int baid, nssn, sta_id, tid; + u32 pkt_len = iwl_rx_packet_payload_len(pkt); + + if (IWL_FW_CHECK(mld, pkt_len < sizeof(*release), + "Unexpected frame release notif size %u (expected %zu)\n", + pkt_len, sizeof(*release))) + return; + + baid = le32_get_bits(release->ba_info, + IWL_BAR_FRAME_RELEASE_BAID_MASK); + nssn = le32_get_bits(release->ba_info, + IWL_BAR_FRAME_RELEASE_NSSN_MASK); + sta_id = le32_get_bits(release->sta_tid, + IWL_BAR_FRAME_RELEASE_STA_MASK); + tid = le32_get_bits(release->sta_tid, + IWL_BAR_FRAME_RELEASE_TID_MASK); + + if (IWL_FW_CHECK(mld, baid >= ARRAY_SIZE(mld->fw_id_to_ba), + "BAR release: invalid BAID (%x)\n", baid)) + return; + + rcu_read_lock(); + baid_data = rcu_dereference(mld->fw_id_to_ba[baid]); + if (!IWL_FW_CHECK(mld, !baid_data, + "Got valid BAID %d but not allocated, invalid BAR release!\n", + baid)) + goto out_unlock; + + if (IWL_FW_CHECK(mld, tid != baid_data->tid || + sta_id > mld->fw->ucode_capa.num_stations || + !(baid_data->sta_mask & BIT(sta_id)), + "BAID 0x%x is mapped to sta_mask:0x%x tid:%d, but BAR release received for sta:%d tid:%d\n", + baid, baid_data->sta_mask, baid_data->tid, sta_id, + tid)) + goto out_unlock; + + IWL_DEBUG_DROP(mld, "Received a BAR, expect packet loss: nssn %d\n", + nssn); + + iwl_mld_release_frames_from_notif(mld, napi, baid, nssn, queue); +out_unlock: + rcu_read_unlock(); +} + +void iwl_mld_del_ba(struct iwl_mld *mld, int queue, + struct iwl_mld_delba_data *data) +{ + struct iwl_mld_baid_data *ba_data; + struct iwl_mld_reorder_buffer *reorder_buf; + struct ieee80211_link_sta *link_sta; + u8 baid = data->baid; + u32 sta_id; + + if (WARN_ONCE(baid >= IWL_MAX_BAID, "invalid BAID: %x\n", baid)) + return; + + rcu_read_lock(); + + ba_data = rcu_dereference(mld->fw_id_to_ba[baid]); + if (WARN_ON_ONCE(!ba_data)) + goto out_unlock; + + /* pick any STA ID to find the pointer */ + sta_id = ffs(ba_data->sta_mask) - 1; + link_sta = rcu_dereference(mld->fw_id_to_link_sta[sta_id]); + if (WARN_ON_ONCE(IS_ERR_OR_NULL(link_sta) || !link_sta->sta)) + goto out_unlock; + + reorder_buf = &ba_data->reorder_buf[queue]; + + /* release all frames that are in the reorder buffer to the stack */ + iwl_mld_reorder_release_frames(mld, link_sta->sta, NULL, + ba_data, reorder_buf, + ieee80211_sn_add(reorder_buf->head_sn, + ba_data->buf_size)); +out_unlock: + rcu_read_unlock(); +} + +/* Returns true if the MPDU was buffered\dropped, false if it should be passed + * to upper layer. + */ +enum iwl_mld_reorder_result +iwl_mld_reorder(struct iwl_mld *mld, struct napi_struct *napi, + int queue, struct ieee80211_sta *sta, + struct sk_buff *skb, struct iwl_rx_mpdu_desc *desc) +{ + struct ieee80211_hdr *hdr = (void *)skb_mac_header(skb); + struct iwl_mld_baid_data *baid_data; + struct iwl_mld_reorder_buffer *buffer; + struct iwl_mld_reorder_buf_entry *entries; + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(sta); + struct iwl_mld_link_sta *mld_link_sta; + u32 reorder = le32_to_cpu(desc->reorder_data); + bool amsdu, last_subframe, is_old_sn, is_dup; + u8 tid = ieee80211_get_tid(hdr); + u8 baid; + u16 nssn, sn; + u32 sta_mask = 0; + int index; + u8 link_id; + + baid = u32_get_bits(reorder, IWL_RX_MPDU_REORDER_BAID_MASK); + + /* This also covers the case of receiving a Block Ack Request + * outside a BA session; we'll pass it to mac80211 and that + * then sends a delBA action frame. + * This also covers pure monitor mode, in which case we won't + * have any BA sessions. + */ + if (baid == IWL_RX_REORDER_DATA_INVALID_BAID) + return IWL_MLD_PASS_SKB; + + /* no sta yet */ + if (WARN_ONCE(!sta, + "Got valid BAID without a valid station assigned\n")) + return IWL_MLD_PASS_SKB; + + /* not a data packet */ + if (!ieee80211_is_data_qos(hdr->frame_control) || + is_multicast_ether_addr(hdr->addr1)) + return IWL_MLD_PASS_SKB; + + if (unlikely(!ieee80211_is_data_present(hdr->frame_control))) + return IWL_MLD_PASS_SKB; + + baid_data = rcu_dereference(mld->fw_id_to_ba[baid]); + if (!baid_data) { + IWL_DEBUG_HT(mld, + "Got valid BAID but no baid allocated, bypass re-ordering (BAID=%d reorder=0x%x)\n", + baid, reorder); + return IWL_MLD_PASS_SKB; + } + + for_each_mld_link_sta(mld_sta, mld_link_sta, link_id) + sta_mask |= BIT(mld_link_sta->fw_id); + + /* verify the BAID is correctly mapped to the sta and tid */ + if (IWL_FW_CHECK(mld, + tid != baid_data->tid || + !(sta_mask & baid_data->sta_mask), + "BAID 0x%x is mapped to sta_mask:0x%x tid:%d, but was received for sta_mask:0x%x tid:%d\n", + baid, baid_data->sta_mask, baid_data->tid, + sta_mask, tid)) + return IWL_MLD_PASS_SKB; + + buffer = &baid_data->reorder_buf[queue]; + entries = &baid_data->entries[queue * baid_data->entries_per_queue]; + + is_old_sn = !!(reorder & IWL_RX_MPDU_REORDER_BA_OLD_SN); + + if (!buffer->valid && is_old_sn) + return IWL_MLD_PASS_SKB; + + buffer->valid = true; + + is_dup = !!(desc->status & cpu_to_le32(IWL_RX_MPDU_STATUS_DUPLICATE)); + + /* drop any duplicated or outdated packets */ + if (is_dup || is_old_sn) + return IWL_MLD_DROP_SKB; + + sn = u32_get_bits(reorder, IWL_RX_MPDU_REORDER_SN_MASK); + nssn = u32_get_bits(reorder, IWL_RX_MPDU_REORDER_NSSN_MASK); + amsdu = desc->mac_flags2 & IWL_RX_MPDU_MFLG2_AMSDU; + last_subframe = desc->amsdu_info & IWL_RX_MPDU_AMSDU_LAST_SUBFRAME; + + /* release immediately if allowed by nssn and no stored frames */ + if (!buffer->num_stored && ieee80211_sn_less(sn, nssn)) { + if (!amsdu || last_subframe) + buffer->head_sn = nssn; + return IWL_MLD_PASS_SKB; + } + + /* release immediately if there are no stored frames, and the sn is + * equal to the head. + * This can happen due to reorder timer, where NSSN is behind head_sn. + * When we released everything, and we got the next frame in the + * sequence, according to the NSSN we can't release immediately, + * while technically there is no hole and we can move forward. + */ + if (!buffer->num_stored && sn == buffer->head_sn) { + if (!amsdu || last_subframe) + buffer->head_sn = ieee80211_sn_inc(buffer->head_sn); + return IWL_MLD_PASS_SKB; + } + + /* put in reorder buffer */ + index = sn % baid_data->buf_size; + __skb_queue_tail(&entries[index].frames, skb); + buffer->num_stored++; + + /* We cannot trust NSSN for AMSDU sub-frames that are not the last. The + * reason is that NSSN advances on the first sub-frame, and may cause + * the reorder buffer to advance before all the sub-frames arrive. + * + * Example: reorder buffer contains SN 0 & 2, and we receive AMSDU with + * SN 1. NSSN for first sub frame will be 3 with the result of driver + * releasing SN 0,1, 2. When sub-frame 1 arrives - reorder buffer is + * already ahead and it will be dropped. + * If the last sub-frame is not on this queue - we will get frame + * release notification with up to date NSSN. + */ + if (!amsdu || last_subframe) + iwl_mld_reorder_release_frames(mld, sta, napi, baid_data, + buffer, nssn); + + return IWL_MLD_BUFFERED_SKB; +} +EXPORT_SYMBOL_IF_IWLWIFI_KUNIT(iwl_mld_reorder); + +static void iwl_mld_rx_agg_session_expired(struct timer_list *t) +{ + struct iwl_mld_baid_data *data = + from_timer(data, t, session_timer); + struct iwl_mld_baid_data __rcu **rcu_ptr = data->rcu_ptr; + struct iwl_mld_baid_data *ba_data; + struct ieee80211_link_sta *link_sta; + struct iwl_mld_sta *mld_sta; + unsigned long timeout; + unsigned int sta_id; + + rcu_read_lock(); + + ba_data = rcu_dereference(*rcu_ptr); + if (WARN_ON(!ba_data)) + goto unlock; + + if (WARN_ON(!ba_data->timeout)) + goto unlock; + + timeout = ba_data->last_rx_timestamp + + TU_TO_JIFFIES(ba_data->timeout * 2); + if (time_is_after_jiffies(timeout)) { + mod_timer(&ba_data->session_timer, timeout); + goto unlock; + } + + /* timer expired, pick any STA ID to find the pointer */ + sta_id = ffs(ba_data->sta_mask) - 1; + link_sta = rcu_dereference(ba_data->mld->fw_id_to_link_sta[sta_id]); + + /* sta should be valid unless the following happens: + * The firmware asserts which triggers a reconfig flow, but + * the reconfig fails before we set the pointer to sta into + * the fw_id_to_link_sta pointer table. mac80211 can't stop + * A-MPDU and hence the timer continues to run. Then, the + * timer expires and sta is NULL. + */ + if (IS_ERR_OR_NULL(link_sta) || WARN_ON(!link_sta->sta)) + goto unlock; + + mld_sta = iwl_mld_sta_from_mac80211(link_sta->sta); + ieee80211_rx_ba_timer_expired(mld_sta->vif, link_sta->sta->addr, + ba_data->tid); +unlock: + rcu_read_unlock(); +} + +static int +iwl_mld_stop_ba_in_fw(struct iwl_mld *mld, struct ieee80211_sta *sta, int tid) +{ + struct iwl_rx_baid_cfg_cmd cmd = { + .action = cpu_to_le32(IWL_RX_BAID_ACTION_REMOVE), + .remove.sta_id_mask = + cpu_to_le32(iwl_mld_fw_sta_id_mask(mld, sta)), + .remove.tid = cpu_to_le32(tid), + + }; + int ret; + + ret = iwl_mld_send_cmd_pdu(mld, + WIDE_ID(DATA_PATH_GROUP, + RX_BAID_ALLOCATION_CONFIG_CMD), + &cmd); + if (ret) + return ret; + + IWL_DEBUG_HT(mld, "RX BA Session stopped in fw\n"); + + return ret; +} + +static int +iwl_mld_start_ba_in_fw(struct iwl_mld *mld, struct ieee80211_sta *sta, + int tid, u16 ssn, u16 buf_size) +{ + struct iwl_rx_baid_cfg_cmd cmd = { + .action = cpu_to_le32(IWL_RX_BAID_ACTION_ADD), + .alloc.sta_id_mask = + cpu_to_le32(iwl_mld_fw_sta_id_mask(mld, sta)), + .alloc.tid = tid, + .alloc.ssn = cpu_to_le16(ssn), + .alloc.win_size = cpu_to_le16(buf_size), + }; + struct iwl_host_cmd hcmd = { + .id = WIDE_ID(DATA_PATH_GROUP, RX_BAID_ALLOCATION_CONFIG_CMD), + .flags = CMD_WANT_SKB, + .len[0] = sizeof(cmd), + .data[0] = &cmd, + }; + struct iwl_rx_baid_cfg_resp *resp; + struct iwl_rx_packet *pkt; + u32 resp_len; + int ret, baid; + + BUILD_BUG_ON(sizeof(*resp) != sizeof(baid)); + + ret = iwl_mld_send_cmd(mld, &hcmd); + if (ret) + return ret; + + pkt = hcmd.resp_pkt; + + resp_len = iwl_rx_packet_payload_len(pkt); + if (IWL_FW_CHECK(mld, resp_len != sizeof(*resp), + "BAID_ALLOC_CMD: unexpected response length %d\n", + resp_len)) { + ret = -EIO; + goto out; + } + + IWL_DEBUG_HT(mld, "RX BA Session started in fw\n"); + + resp = (void *)pkt->data; + baid = le32_to_cpu(resp->baid); + + if (IWL_FW_CHECK(mld, baid < 0 || baid >= ARRAY_SIZE(mld->fw_id_to_ba), + "BAID_ALLOC_CMD: invalid BAID response %d\n", baid)) { + ret = -EINVAL; + goto out; + } + + ret = baid; +out: + iwl_free_resp(&hcmd); + return ret; +} + +static void iwl_mld_init_reorder_buffer(struct iwl_mld *mld, + struct iwl_mld_baid_data *data, + u16 ssn) +{ + for (int i = 0; i < mld->trans->num_rx_queues; i++) { + struct iwl_mld_reorder_buffer *reorder_buf = + &data->reorder_buf[i]; + struct iwl_mld_reorder_buf_entry *entries = + &data->entries[i * data->entries_per_queue]; + + reorder_buf->head_sn = ssn; + reorder_buf->queue = i; + + for (int j = 0; j < data->buf_size; j++) + __skb_queue_head_init(&entries[j].frames); + } +} + +static void iwl_mld_free_reorder_buffer(struct iwl_mld *mld, + struct iwl_mld_baid_data *data) +{ + struct iwl_mld_delba_data delba_data = { + .baid = data->baid, + }; + + iwl_mld_sync_rx_queues(mld, IWL_MLD_RXQ_NOTIF_DEL_BA, + &delba_data, sizeof(delba_data)); + + for (int i = 0; i < mld->trans->num_rx_queues; i++) { + struct iwl_mld_reorder_buffer *reorder_buf = + &data->reorder_buf[i]; + struct iwl_mld_reorder_buf_entry *entries = + &data->entries[i * data->entries_per_queue]; + + if (likely(!reorder_buf->num_stored)) + continue; + + /* This shouldn't happen in regular DELBA since the RX queues + * sync internal DELBA notification should trigger a release + * of all frames in the reorder buffer. + */ + WARN_ON(1); + + for (int j = 0; j < data->buf_size; j++) + __skb_queue_purge(&entries[j].frames); + } +} + +int iwl_mld_ampdu_rx_start(struct iwl_mld *mld, struct ieee80211_sta *sta, + int tid, u16 ssn, u16 buf_size, u16 timeout) +{ + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(sta); + struct iwl_mld_baid_data *baid_data = NULL; + u32 reorder_buf_size = buf_size * sizeof(baid_data->entries[0]); + int ret, baid; + u32 sta_mask; + + lockdep_assert_wiphy(mld->wiphy); + + if (mld->num_rx_ba_sessions >= IWL_MAX_BAID) { + IWL_DEBUG_HT(mld, + "Max num of RX BA sessions reached; blocking new session\n"); + return -ENOSPC; + } + + sta_mask = iwl_mld_fw_sta_id_mask(mld, sta); + if (WARN_ON(!sta_mask)) + return -EINVAL; + + /* sparse doesn't like the __align() so don't check */ +#ifndef __CHECKER__ + /* The division below will be OK if either the cache line size + * can be divided by the entry size (ALIGN will round up) or if + * the entry size can be divided by the cache line size, in which + * case the ALIGN() will do nothing. + */ + BUILD_BUG_ON(SMP_CACHE_BYTES % sizeof(baid_data->entries[0]) && + sizeof(baid_data->entries[0]) % SMP_CACHE_BYTES); +#endif + + /* Upward align the reorder buffer size to fill an entire cache + * line for each queue, to avoid sharing cache lines between + * different queues. + */ + reorder_buf_size = ALIGN(reorder_buf_size, SMP_CACHE_BYTES); + + /* Allocate here so if allocation fails we can bail out early + * before starting the BA session in the firmware + */ + baid_data = kzalloc(sizeof(*baid_data) + + mld->trans->num_rx_queues * reorder_buf_size, + GFP_KERNEL); + if (!baid_data) + return -ENOMEM; + + /* This division is why we need the above BUILD_BUG_ON(), + * if that doesn't hold then this will not be right. + */ + baid_data->entries_per_queue = + reorder_buf_size / sizeof(baid_data->entries[0]); + + baid = iwl_mld_start_ba_in_fw(mld, sta, tid, ssn, buf_size); + if (baid < 0) { + ret = baid; + goto out_free; + } + + mld->num_rx_ba_sessions++; + mld_sta->tid_to_baid[tid] = baid; + + baid_data->baid = baid; + baid_data->mld = mld; + baid_data->tid = tid; + baid_data->buf_size = buf_size; + baid_data->sta_mask = sta_mask; + baid_data->timeout = timeout; + baid_data->last_rx_timestamp = jiffies; + baid_data->rcu_ptr = &mld->fw_id_to_ba[baid]; + + iwl_mld_init_reorder_buffer(mld, baid_data, ssn); + + timer_setup(&baid_data->session_timer, iwl_mld_rx_agg_session_expired, + 0); + if (timeout) + mod_timer(&baid_data->session_timer, + TU_TO_EXP_TIME(timeout * 2)); + + IWL_DEBUG_HT(mld, "STA mask=0x%x (tid=%d) is assigned to BAID %d\n", + baid_data->sta_mask, tid, baid); + + /* protect the BA data with RCU to cover a case where our + * internal RX sync mechanism will timeout (not that it's + * supposed to happen) and we will free the session data while + * RX is being processed in parallel + */ + WARN_ON(rcu_access_pointer(mld->fw_id_to_ba[baid])); + rcu_assign_pointer(mld->fw_id_to_ba[baid], baid_data); + + return 0; + +out_free: + kfree(baid_data); + return ret; +} + +int iwl_mld_ampdu_rx_stop(struct iwl_mld *mld, struct ieee80211_sta *sta, + int tid) +{ + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(sta); + int baid = mld_sta->tid_to_baid[tid]; + struct iwl_mld_baid_data *baid_data; + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + /* during firmware restart, do not send the command as the firmware no + * longer recognizes the session. instead, only clear the driver BA + * session data. + */ + if (!mld->fw_status.in_hw_restart) { + ret = iwl_mld_stop_ba_in_fw(mld, sta, tid); + if (ret) + return ret; + } + + if (!WARN_ON(mld->num_rx_ba_sessions == 0)) + mld->num_rx_ba_sessions--; + + baid_data = wiphy_dereference(mld->wiphy, mld->fw_id_to_ba[baid]); + if (WARN_ON(!baid_data)) + return -EINVAL; + + if (timer_pending(&baid_data->session_timer)) + timer_shutdown_sync(&baid_data->session_timer); + + iwl_mld_free_reorder_buffer(mld, baid_data); + + RCU_INIT_POINTER(mld->fw_id_to_ba[baid], NULL); + kfree_rcu(baid_data, rcu_head); + + IWL_DEBUG_HT(mld, "BAID %d is free\n", baid); + + return 0; +} + +int iwl_mld_update_sta_baids(struct iwl_mld *mld, + u32 old_sta_mask, + u32 new_sta_mask) +{ + struct iwl_rx_baid_cfg_cmd cmd = { + .action = cpu_to_le32(IWL_RX_BAID_ACTION_MODIFY), + .modify.old_sta_id_mask = cpu_to_le32(old_sta_mask), + .modify.new_sta_id_mask = cpu_to_le32(new_sta_mask), + }; + u32 cmd_id = WIDE_ID(DATA_PATH_GROUP, RX_BAID_ALLOCATION_CONFIG_CMD); + int baid; + + /* mac80211 will remove sessions later, but we ignore all that */ + if (mld->fw_status.in_hw_restart) + return 0; + + BUILD_BUG_ON(sizeof(struct iwl_rx_baid_cfg_resp) != sizeof(baid)); + + for (baid = 0; baid < ARRAY_SIZE(mld->fw_id_to_ba); baid++) { + struct iwl_mld_baid_data *data; + int ret; + + data = wiphy_dereference(mld->wiphy, mld->fw_id_to_ba[baid]); + if (!data) + continue; + + if (!(data->sta_mask & old_sta_mask)) + continue; + + WARN_ONCE(data->sta_mask != old_sta_mask, + "BAID data for %d corrupted - expected 0x%x found 0x%x\n", + baid, old_sta_mask, data->sta_mask); + + cmd.modify.tid = cpu_to_le32(data->tid); + + ret = iwl_mld_send_cmd_pdu(mld, cmd_id, &cmd); + if (ret) + return ret; + data->sta_mask = new_sta_mask; + } + + return 0; +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/agg.h b/drivers/net/wireless/intel/iwlwifi/mld/agg.h new file mode 100644 index 0000000000000..651c80d1c7cda --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/agg.h @@ -0,0 +1,127 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024 Intel Corporation + */ +#ifndef __iwl_agg_h__ +#define __iwl_agg_h__ + +#include "mld.h" +#include "fw/api/rx.h" + +/** + * struct iwl_mld_reorder_buffer - per ra/tid/queue reorder buffer + * @head_sn: reorder window head sequence number + * @num_stored: number of MPDUs stored in the buffer + * @queue: queue of this reorder buffer + * @valid: true if reordering is valid for this queue + */ +struct iwl_mld_reorder_buffer { + u16 head_sn; + u16 num_stored; + int queue; + bool valid; +} ____cacheline_aligned_in_smp; + +/** + * struct iwl_mld_reorder_buf_entry - reorder buffer entry per-queue/per-seqno + * @frames: list of skbs stored. a list is necessary because in an A-MSDU, + * all sub-frames share the same sequence number, so they are stored + * together in the same list. + */ +struct iwl_mld_reorder_buf_entry { + struct sk_buff_head frames; +} +#ifndef __CHECKER__ +/* sparse doesn't like this construct: "bad integer constant expression" */ +__aligned(roundup_pow_of_two(sizeof(struct sk_buff_head))) +#endif +; + +/** + * struct iwl_mld_baid_data - Block Ack session data + * @rcu_head: RCU head for freeing this data + * @sta_mask: station mask for the BAID + * @tid: tid of the session + * @baid: baid of the session + * @buf_size: the reorder buffer size as set by the last ADDBA request + * @entries_per_queue: number of buffers per queue, this actually gets + * aligned up to avoid cache line sharing between queues + * @timeout: the timeout value specified in the ADDBA request. + * @last_rx_timestamp: timestamp of the last received packet (in jiffies). This + * value is updated only when the configured @timeout has passed since + * the last update to minimize cache bouncing between RX queues. + * @session_timer: timer is set to expire after 2 * @timeout (since we want + * to minimize the cache bouncing by updating @last_rx_timestamp only once + * after @timeout has passed). If no packets are received within this + * period, it informs mac80211 to initiate delBA flow, terminating the + * BA session. + * @rcu_ptr: BA data RCU protected access + * @mld: mld pointer, needed for timer context + * @reorder_buf: reorder buffer, allocated per queue + * @entries: data + */ +struct iwl_mld_baid_data { + struct rcu_head rcu_head; + u32 sta_mask; + u8 tid; + u8 baid; + u16 buf_size; + u16 entries_per_queue; + u16 timeout; + struct timer_list session_timer; + unsigned long last_rx_timestamp; + struct iwl_mld_baid_data __rcu **rcu_ptr; + struct iwl_mld *mld; + struct iwl_mld_reorder_buffer reorder_buf[IWL_MAX_RX_HW_QUEUES]; + struct iwl_mld_reorder_buf_entry entries[] ____cacheline_aligned_in_smp; +}; + +/** + * struct iwl_mld_delba_data - RX queue sync data for %IWL_MLD_RXQ_NOTIF_DEL_BA + * + * @baid: Block Ack id, used to identify the BA session to be removed + */ +struct iwl_mld_delba_data { + u32 baid; +} __packed; + +/** + * enum iwl_mld_reorder_result - Possible return values for iwl_mld_reorder() + * indicating how the caller should handle the skb based on the result. + * + * @IWL_MLD_PASS_SKB: skb should be passed to upper layer. + * @IWL_MLD_BUFFERED_SKB: skb has been buffered, don't pass it to upper layer. + * @IWL_MLD_DROP_SKB: skb should be dropped and freed by the caller. + */ +enum iwl_mld_reorder_result { + IWL_MLD_PASS_SKB, + IWL_MLD_BUFFERED_SKB, + IWL_MLD_DROP_SKB +}; + +int iwl_mld_ampdu_rx_start(struct iwl_mld *mld, struct ieee80211_sta *sta, + int tid, u16 ssn, u16 buf_size, u16 timeout); +int iwl_mld_ampdu_rx_stop(struct iwl_mld *mld, struct ieee80211_sta *sta, + int tid); + +enum iwl_mld_reorder_result +iwl_mld_reorder(struct iwl_mld *mld, struct napi_struct *napi, + int queue, struct ieee80211_sta *sta, + struct sk_buff *skb, struct iwl_rx_mpdu_desc *desc); + +void iwl_mld_handle_frame_release_notif(struct iwl_mld *mld, + struct napi_struct *napi, + struct iwl_rx_packet *pkt, int queue); +void iwl_mld_handle_bar_frame_release_notif(struct iwl_mld *mld, + struct napi_struct *napi, + struct iwl_rx_packet *pkt, + int queue); + +void iwl_mld_del_ba(struct iwl_mld *mld, int queue, + struct iwl_mld_delba_data *data); + +int iwl_mld_update_sta_baids(struct iwl_mld *mld, + u32 old_sta_mask, + u32 new_sta_mask); + +#endif /* __iwl_agg_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/ap.c b/drivers/net/wireless/intel/iwlwifi/mld/ap.c new file mode 100644 index 0000000000000..571eabd0b511f --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/ap.c @@ -0,0 +1,344 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024 Intel Corporation + */ +#include + +#include + +#include "ap.h" +#include "hcmd.h" +#include "tx.h" +#include "power.h" +#include "key.h" +#include "iwl-utils.h" + +#include "fw/api/sta.h" + +void iwl_mld_set_tim_idx(struct iwl_mld *mld, __le32 *tim_index, + u8 *beacon, u32 frame_size) +{ + u32 tim_idx; + struct ieee80211_mgmt *mgmt = (void *)beacon; + + /* The index is relative to frame start but we start looking at the + * variable-length part of the beacon. + */ + tim_idx = mgmt->u.beacon.variable - beacon; + + /* Parse variable-length elements of beacon to find WLAN_EID_TIM */ + while ((tim_idx < (frame_size - 2)) && + (beacon[tim_idx] != WLAN_EID_TIM)) + tim_idx += beacon[tim_idx + 1] + 2; + + /* If TIM field was found, set variables */ + if ((tim_idx < (frame_size - 1)) && beacon[tim_idx] == WLAN_EID_TIM) + *tim_index = cpu_to_le32(tim_idx); + else + IWL_WARN(mld, "Unable to find TIM Element in beacon\n"); +} + +u8 iwl_mld_get_rate_flags(struct iwl_mld *mld, + struct ieee80211_tx_info *info, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link, + enum nl80211_band band) +{ + u32 legacy = link->beacon_tx_rate.control[band].legacy; + u32 rate_idx, rate_flags = 0, fw_rate; + + /* if beacon rate was configured try using it */ + if (hweight32(legacy) == 1) { + u32 rate = ffs(legacy) - 1; + struct ieee80211_supported_band *sband = + mld->hw->wiphy->bands[band]; + + rate_idx = sband->bitrates[rate].hw_value; + } else { + rate_idx = iwl_mld_get_lowest_rate(mld, info, vif); + } + + if (rate_idx <= IWL_LAST_CCK_RATE) + rate_flags = IWL_MAC_BEACON_CCK; + + /* Legacy rates are indexed as follows: + * 0 - 3 for CCK and 0 - 7 for OFDM. + */ + fw_rate = (rate_idx >= IWL_FIRST_OFDM_RATE ? + rate_idx - IWL_FIRST_OFDM_RATE : rate_idx); + + return fw_rate | rate_flags; +} + +int iwl_mld_send_beacon_template_cmd(struct iwl_mld *mld, + struct sk_buff *beacon, + struct iwl_mac_beacon_cmd *cmd) +{ + struct iwl_host_cmd hcmd = { + .id = BEACON_TEMPLATE_CMD, + }; + + hcmd.len[0] = sizeof(*cmd); + hcmd.data[0] = cmd; + + hcmd.len[1] = beacon->len; + hcmd.data[1] = beacon->data; + hcmd.dataflags[1] = IWL_HCMD_DFL_DUP; + + return iwl_mld_send_cmd(mld, &hcmd); +} + +static int iwl_mld_fill_beacon_template_cmd(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct sk_buff *beacon, + struct iwl_mac_beacon_cmd *cmd, + struct ieee80211_bss_conf *link) +{ + struct iwl_mld_link *mld_link = iwl_mld_link_from_mac80211(link); + struct ieee80211_tx_info *info = IEEE80211_SKB_CB(beacon); + struct ieee80211_chanctx_conf *ctx; + bool enable_fils; + u16 flags = 0; + + lockdep_assert_wiphy(mld->wiphy); + + if (WARN_ON(!mld_link)) + return -EINVAL; + + cmd->link_id = cpu_to_le32(mld_link->fw_id); + + ctx = wiphy_dereference(mld->wiphy, link->chanctx_conf); + if (WARN_ON(!ctx || !ctx->def.chan)) + return -EINVAL; + + enable_fils = cfg80211_channel_is_psc(ctx->def.chan) || + (ctx->def.chan->band == NL80211_BAND_6GHZ && + ctx->def.width >= NL80211_CHAN_WIDTH_80); + + if (enable_fils) { + flags |= IWL_MAC_BEACON_FILS; + cmd->short_ssid = cpu_to_le32(~crc32_le(~0, vif->cfg.ssid, + vif->cfg.ssid_len)); + } + + cmd->byte_cnt = cpu_to_le16((u16)beacon->len); + + flags |= iwl_mld_get_rate_flags(mld, info, vif, link, + ctx->def.chan->band); + + cmd->flags = cpu_to_le16(flags); + + if (vif->type == NL80211_IFTYPE_AP) { + iwl_mld_set_tim_idx(mld, &cmd->tim_idx, + beacon->data, beacon->len); + + cmd->btwt_offset = + cpu_to_le32(iwl_find_ie_offset(beacon->data, + WLAN_EID_S1G_TWT, + beacon->len)); + } + + cmd->csa_offset = + cpu_to_le32(iwl_find_ie_offset(beacon->data, + WLAN_EID_CHANNEL_SWITCH, + beacon->len)); + cmd->ecsa_offset = + cpu_to_le32(iwl_find_ie_offset(beacon->data, + WLAN_EID_EXT_CHANSWITCH_ANN, + beacon->len)); + + return 0; +} + +/* The beacon template for the AP/GO/IBSS has changed and needs update */ +int iwl_mld_update_beacon_template(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link_conf) +{ + struct iwl_mac_beacon_cmd cmd = {}; + struct sk_buff *beacon; + int ret; +#ifdef CONFIG_IWLWIFI_DEBUGFS + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); +#endif + + WARN_ON(vif->type != NL80211_IFTYPE_AP && + vif->type != NL80211_IFTYPE_ADHOC); + + if (IWL_MLD_NON_TRANSMITTING_AP) + return 0; + +#ifdef CONFIG_IWLWIFI_DEBUGFS + if (mld_vif->beacon_inject_active) { + IWL_DEBUG_INFO(mld, + "Can't update template, beacon injection's active\n"); + return -EBUSY; + } + +#endif + beacon = ieee80211_beacon_get_template(mld->hw, vif, NULL, + link_conf->link_id); + if (!beacon) + return -ENOMEM; + + ret = iwl_mld_fill_beacon_template_cmd(mld, vif, beacon, &cmd, + link_conf); + + if (!ret) + ret = iwl_mld_send_beacon_template_cmd(mld, beacon, &cmd); + + dev_kfree_skb(beacon); + + return ret; +} + +void iwl_mld_free_ap_early_key(struct iwl_mld *mld, + struct ieee80211_key_conf *key, + struct iwl_mld_vif *mld_vif) +{ + struct iwl_mld_link *link; + + if (WARN_ON(key->link_id < 0)) + return; + + link = iwl_mld_link_dereference_check(mld_vif, key->link_id); + if (WARN_ON(!link)) + return; + + for (int i = 0; i < ARRAY_SIZE(link->ap_early_keys); i++) { + if (link->ap_early_keys[i] != key) + continue; + /* Those weren't sent to FW, so should be marked as INVALID */ + if (WARN_ON(key->hw_key_idx != STA_KEY_IDX_INVALID)) + key->hw_key_idx = STA_KEY_IDX_INVALID; + link->ap_early_keys[i] = NULL; + } +} + +int iwl_mld_store_ap_early_key(struct iwl_mld *mld, + struct ieee80211_key_conf *key, + struct iwl_mld_vif *mld_vif) +{ + struct iwl_mld_link *link; + + if (WARN_ON(key->link_id < 0)) + return -EINVAL; + + link = iwl_mld_link_dereference_check(mld_vif, key->link_id); + if (WARN_ON(!link)) + return -EINVAL; + + for (int i = 0; i < ARRAY_SIZE(link->ap_early_keys); i++) { + if (!link->ap_early_keys[i]) { + link->ap_early_keys[i] = key; + return 0; + } + } + + return -ENOSPC; +} + +static int iwl_mld_send_ap_early_keys(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link) +{ + struct iwl_mld_link *mld_link = iwl_mld_link_from_mac80211(link); + int ret = 0; + + if (WARN_ON(!link)) + return -EINVAL; + + for (int i = 0; i < ARRAY_SIZE(mld_link->ap_early_keys); i++) { + struct ieee80211_key_conf *key = mld_link->ap_early_keys[i]; + + if (!key) + continue; + + mld_link->ap_early_keys[i] = NULL; + + ret = iwl_mld_add_key(mld, vif, NULL, key); + if (ret) + break; + } + return ret; +} + +int iwl_mld_start_ap_ibss(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + int ret; + + if (vif->type == NL80211_IFTYPE_AP) + iwl_mld_send_ap_tx_power_constraint_cmd(mld, vif, link); + + ret = iwl_mld_update_beacon_template(mld, vif, link); + if (ret) + return ret; + + /* the link should be already activated when assigning chan context, + * and LINK_CONTEXT_MODIFY_EHT_PARAMS is deprecated + */ + ret = iwl_mld_change_link_in_fw(mld, link, + LINK_CONTEXT_MODIFY_ALL & + ~(LINK_CONTEXT_MODIFY_ACTIVE | + LINK_CONTEXT_MODIFY_EHT_PARAMS)); + if (ret) + return ret; + + ret = iwl_mld_add_mcast_sta(mld, vif, link); + if (ret) + return ret; + + ret = iwl_mld_add_bcast_sta(mld, vif, link); + if (ret) + goto rm_mcast; + + /* Those keys were configured by the upper layers before starting the + * AP. Now that it is started and the bcast and mcast sta were added to + * the FW, we can add the keys too. + */ + ret = iwl_mld_send_ap_early_keys(mld, vif, link); + if (ret) + goto rm_bcast; + + if (ieee80211_vif_type_p2p(vif) == NL80211_IFTYPE_AP) + iwl_mld_vif_update_low_latency(mld, vif, true, + LOW_LATENCY_VIF_TYPE); + + mld_vif->ap_ibss_active = true; + + if (vif->p2p && mld->p2p_device_vif) + return iwl_mld_mac_fw_action(mld, mld->p2p_device_vif, + FW_CTXT_ACTION_MODIFY); + + return 0; +rm_bcast: + iwl_mld_remove_bcast_sta(mld, vif, link); +rm_mcast: + iwl_mld_remove_mcast_sta(mld, vif, link); + return ret; +} + +void iwl_mld_stop_ap_ibss(struct ieee80211_hw *hw, struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + + mld_vif->ap_ibss_active = false; + + if (vif->p2p && mld->p2p_device_vif) + iwl_mld_mac_fw_action(mld, mld->p2p_device_vif, + FW_CTXT_ACTION_MODIFY); + + if (ieee80211_vif_type_p2p(vif) == NL80211_IFTYPE_AP) + iwl_mld_vif_update_low_latency(mld, vif, false, + LOW_LATENCY_VIF_TYPE); + + iwl_mld_remove_bcast_sta(mld, vif, link); + + iwl_mld_remove_mcast_sta(mld, vif, link); +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/ap.h b/drivers/net/wireless/intel/iwlwifi/mld/ap.h new file mode 100644 index 0000000000000..4a6f52b9552d7 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/ap.h @@ -0,0 +1,45 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024 Intel Corporation + */ +#ifndef __iwl_ap_h__ +#define __iwl_ap_h__ + +#include "mld.h" +#include "iface.h" + +#include "fw/api/tx.h" + +int iwl_mld_update_beacon_template(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link_conf); + +int iwl_mld_start_ap_ibss(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link); + +void iwl_mld_stop_ap_ibss(struct ieee80211_hw *hw, struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link); + +int iwl_mld_store_ap_early_key(struct iwl_mld *mld, + struct ieee80211_key_conf *key, + struct iwl_mld_vif *mld_vif); + +void iwl_mld_free_ap_early_key(struct iwl_mld *mld, + struct ieee80211_key_conf *key, + struct iwl_mld_vif *mld_vif); + +u8 iwl_mld_get_rate_flags(struct iwl_mld *mld, + struct ieee80211_tx_info *info, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link, + enum nl80211_band band); + +void iwl_mld_set_tim_idx(struct iwl_mld *mld, __le32 *tim_index, + u8 *beacon, u32 frame_size); + +int iwl_mld_send_beacon_template_cmd(struct iwl_mld *mld, + struct sk_buff *beacon, + struct iwl_mac_beacon_cmd *cmd); + +#endif /* __iwl_ap_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/coex.c b/drivers/net/wireless/intel/iwlwifi/mld/coex.c new file mode 100644 index 0000000000000..5f262bd43f21d --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/coex.c @@ -0,0 +1,40 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024-2025 Intel Corporation + */ + +#include "fw/api/coex.h" + +#include "coex.h" +#include "mld.h" +#include "hcmd.h" +#include "mlo.h" + +int iwl_mld_send_bt_init_conf(struct iwl_mld *mld) +{ + struct iwl_bt_coex_cmd cmd = { + .mode = cpu_to_le32(BT_COEX_NW), + .enabled_modules = cpu_to_le32(BT_COEX_MPLUT_ENABLED | + BT_COEX_HIGH_BAND_RET), + }; + + return iwl_mld_send_cmd_pdu(mld, BT_CONFIG, &cmd); +} + +void iwl_mld_handle_bt_coex_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + const struct iwl_bt_coex_profile_notif *notif = (void *)pkt->data; + const struct iwl_bt_coex_profile_notif zero_notif = {}; + /* zeroed structure means that BT is OFF */ + bool bt_is_active = memcmp(notif, &zero_notif, sizeof(*notif)); + + if (bt_is_active == mld->bt_is_active) + return; + + IWL_DEBUG_INFO(mld, "BT was turned %s\n", bt_is_active ? "ON" : "OFF"); + + mld->bt_is_active = bt_is_active; + + iwl_mld_emlsr_check_bt(mld); +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/coex.h b/drivers/net/wireless/intel/iwlwifi/mld/coex.h new file mode 100644 index 0000000000000..a77c5dc9613c0 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/coex.h @@ -0,0 +1,15 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024 Intel Corporation + */ +#ifndef __iwl_mld_coex_h__ +#define __iwl_mld_coex_h__ + +#include "mld.h" + +int iwl_mld_send_bt_init_conf(struct iwl_mld *mld); + +void iwl_mld_handle_bt_coex_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt); + +#endif /* __iwl_mld_coex_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/constants.h b/drivers/net/wireless/intel/iwlwifi/mld/constants.h new file mode 100644 index 0000000000000..9a24996014b8d --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/constants.h @@ -0,0 +1,88 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024-2025 Intel Corporation + */ +#ifndef __iwl_mld_constants_h__ +#define __iwl_mld_constants_h__ + +#define IWL_MLD_MISSED_BEACONS_SINCE_RX_THOLD 4 +#define IWL_MLD_MISSED_BEACONS_THRESHOLD 8 +#define IWL_MLD_MISSED_BEACONS_THRESHOLD_LONG 19 +#define IWL_MLD_BCN_LOSS_EXIT_ESR_THRESH_2_LINKS 5 +#define IWL_MLD_BCN_LOSS_EXIT_ESR_THRESH 15 +#define IWL_MLD_BCN_LOSS_EXIT_ESR_THRESH_BSS_PARAM_CHANGED 11 +#define IWL_MLD_LOW_RSSI_MLO_SCAN_THRESH -72 + +#define IWL_MLD_DEFAULT_PS_TX_DATA_TIMEOUT (100 * USEC_PER_MSEC) +#define IWL_MLD_DEFAULT_PS_RX_DATA_TIMEOUT (100 * USEC_PER_MSEC) +#define IWL_MLD_WOWLAN_PS_TX_DATA_TIMEOUT (10 * USEC_PER_MSEC) +#define IWL_MLD_WOWLAN_PS_RX_DATA_TIMEOUT (10 * USEC_PER_MSEC) +#define IWL_MLD_SHORT_PS_TX_DATA_TIMEOUT (2 * 1024) /* defined in TU */ +#define IWL_MLD_SHORT_PS_RX_DATA_TIMEOUT (40 * 1024) /* defined in TU */ + +#define IWL_MLD_UAPSD_RX_DATA_TIMEOUT (50 * USEC_PER_MSEC) +#define IWL_MLD_UAPSD_TX_DATA_TIMEOUT (50 * USEC_PER_MSEC) + +#define IWL_MLD_PS_SNOOZE_INTERVAL 25 +#define IWL_MLD_PS_SNOOZE_INTERVAL 25 +#define IWL_MLD_PS_SNOOZE_WINDOW 50 + +#define IWL_MLD_PS_SNOOZE_HEAVY_TX_THLD_PACKETS 30 +#define IWL_MLD_PS_SNOOZE_HEAVY_RX_THLD_PACKETS 20 + +#define IWL_MLD_PS_HEAVY_TX_THLD_PERCENT 50 +#define IWL_MLD_PS_HEAVY_RX_THLD_PERCENT 50 +#define IWL_MLD_PS_HEAVY_TX_THLD_PACKETS 20 +#define IWL_MLD_PS_HEAVY_RX_THLD_PACKETS 8 + +#define IWL_MLD_TRIGGER_LINK_SEL_TIME_SEC 30 +#define IWL_MLD_SCAN_EXPIRE_TIME_SEC 20 + +#define IWL_MLD_TPT_COUNT_WINDOW (5 * HZ) + +/* OMI reduced BW thresholds (channel load percentage) */ +#define IWL_MLD_OMI_ENTER_CHAN_LOAD 10 +#define IWL_MLD_OMI_EXIT_CHAN_LOAD_160 20 +#define IWL_MLD_OMI_EXIT_CHAN_LOAD_320 30 +/* time (in milliseconds) to let AP "settle" the OMI */ +#define IWL_MLD_OMI_AP_SETTLE_DELAY 27 +/* time (in milliseconds) to not enter OMI reduced BW after leaving */ +#define IWL_MLD_OMI_EXIT_PROTECTION 5000 + +#define IWL_MLD_DIS_RANDOM_FW_ID false +#define IWL_MLD_D3_DEBUG false +#define IWL_MLD_NON_TRANSMITTING_AP false +#define IWL_MLD_6GHZ_PASSIVE_SCAN_TIMEOUT 3000 /* in seconds */ +#define IWL_MLD_6GHZ_PASSIVE_SCAN_ASSOC_TIMEOUT 60 /* in seconds */ +#define IWL_MLD_CONN_LISTEN_INTERVAL 10 +#define IWL_MLD_ADAPTIVE_DWELL_NUM_APS_OVERRIDE 0 +#define IWL_MLD_AUTO_EML_ENABLE true + +#define IWL_MLD_HIGH_RSSI_THRESH_20MHZ -67 +#define IWL_MLD_LOW_RSSI_THRESH_20MHZ -72 +#define IWL_MLD_HIGH_RSSI_THRESH_40MHZ -64 +#define IWL_MLD_LOW_RSSI_THRESH_40MHZ -72 +#define IWL_MLD_HIGH_RSSI_THRESH_80MHZ -61 +#define IWL_MLD_LOW_RSSI_THRESH_80MHZ -72 +#define IWL_MLD_HIGH_RSSI_THRESH_160MHZ -58 +#define IWL_MLD_LOW_RSSI_THRESH_160MHZ -72 + +#define IWL_MLD_ENTER_EMLSR_TPT_THRESH 400 +#define IWL_MLD_CHAN_LOAD_THRESH 2 /* in percentage */ + +#define IWL_MLD_FTM_INITIATOR_ALGO IWL_TOF_ALGO_TYPE_MAX_LIKE +#define IWL_MLD_FTM_INITIATOR_DYNACK true +#define IWL_MLD_FTM_LMR_FEEDBACK_TERMINATE false +#define IWL_MLD_FTM_TEST_INCORRECT_SAC false +#define IWL_MLD_FTM_R2I_MAX_REP 7 +#define IWL_MLD_FTM_I2R_MAX_REP 7 +#define IWL_MLD_FTM_R2I_MAX_STS 1 +#define IWL_MLD_FTM_I2R_MAX_STS 1 +#define IWL_MLD_FTM_R2I_MAX_TOTAL_LTF 3 +#define IWL_MLD_FTM_I2R_MAX_TOTAL_LTF 3 +#define IWL_MLD_FTM_RESP_NDP_SUPPORT true +#define IWL_MLD_FTM_RESP_LMR_FEEDBACK_SUPPORT true +#define IWL_MLD_FTM_NON_TB_MIN_TIME_BETWEEN_MSR 7 +#define IWL_MLD_FTM_NON_TB_MAX_TIME_BETWEEN_MSR 1000 + +#endif /* __iwl_mld_constants_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/d3.c b/drivers/net/wireless/intel/iwlwifi/mld/d3.c new file mode 100644 index 0000000000000..5a7207accd867 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/d3.c @@ -0,0 +1,1998 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024-2025 Intel Corporation + */ +#include "mld.h" + +#include "d3.h" +#include "power.h" +#include "hcmd.h" +#include "iface.h" +#include "mcc.h" +#include "sta.h" +#include "mlo.h" + +#include "fw/api/d3.h" +#include "fw/api/offload.h" +#include "fw/api/sta.h" +#include "fw/dbg.h" + +#include +#include +#include + +/** + * enum iwl_mld_d3_notif - d3 notifications + * @IWL_D3_NOTIF_WOWLAN_INFO: WOWLAN_INFO_NOTIF is expected/was received + * @IWL_D3_NOTIF_WOWLAN_WAKE_PKT: WOWLAN_WAKE_PKT_NOTIF is expected/was received + * @IWL_D3_NOTIF_PROT_OFFLOAD: PROT_OFFLOAD_NOTIF is expected/was received + * @IWL_D3_ND_MATCH_INFO: OFFLOAD_MATCH_INFO_NOTIF is expected/was received + * @IWL_D3_NOTIF_D3_END_NOTIF: D3_END_NOTIF is expected/was received + */ +enum iwl_mld_d3_notif { + IWL_D3_NOTIF_WOWLAN_INFO = BIT(0), + IWL_D3_NOTIF_WOWLAN_WAKE_PKT = BIT(1), + IWL_D3_NOTIF_PROT_OFFLOAD = BIT(2), + IWL_D3_ND_MATCH_INFO = BIT(3), + IWL_D3_NOTIF_D3_END_NOTIF = BIT(4) +}; + +struct iwl_mld_resume_key_iter_data { + struct iwl_mld *mld; + struct iwl_mld_wowlan_status *wowlan_status; + u32 num_keys, gtk_cipher, igtk_cipher, bigtk_cipher; + bool unhandled_cipher; +}; + +struct iwl_mld_suspend_key_iter_data { + struct iwl_wowlan_rsc_tsc_params_cmd *rsc; + bool have_rsc; + int gtks; + int found_gtk_idx[4]; + __le32 gtk_cipher; + __le32 igtk_cipher; + __le32 bigtk_cipher; +}; + +struct iwl_mld_mcast_key_data { + u8 key[WOWLAN_KEY_MAX_SIZE]; + u8 len; + u8 flags; + u8 id; + union { + struct { + struct ieee80211_key_seq aes_seq[IWL_MAX_TID_COUNT]; + struct ieee80211_key_seq tkip_seq[IWL_MAX_TID_COUNT]; + } gtk; + struct { + struct ieee80211_key_seq cmac_gmac_seq; + } igtk_bigtk; + }; + +}; + +/** + * struct iwl_mld_wowlan_status - contains wowlan status data from + * all wowlan notifications + * @wakeup_reasons: wakeup reasons, see &enum iwl_wowlan_wakeup_reason + * @replay_ctr: GTK rekey replay counter + * @pattern_number: number of the matched patterns on packets + * @last_qos_seq: QoS sequence counter of offloaded tid + * @num_of_gtk_rekeys: number of GTK rekeys during D3 + * @tid_offloaded_tx: tid used by the firmware to transmit data packets + * while in wowlan + * @wake_packet: wakeup packet received + * @wake_packet_length: wake packet length + * @wake_packet_bufsize: wake packet bufsize + * @gtk: data of the last two used gtk's by the FW upon resume + * @igtk: data of the last used igtk by the FW upon resume + * @bigtk: data of the last two used gtk's by the FW upon resume + * @ptk: last seq numbers per tid passed by the FW, + * holds both in tkip and aes formats + */ +struct iwl_mld_wowlan_status { + u32 wakeup_reasons; + u64 replay_ctr; + u16 pattern_number; + u16 last_qos_seq; + u32 num_of_gtk_rekeys; + u8 tid_offloaded_tx; + u8 *wake_packet; + u32 wake_packet_length; + u32 wake_packet_bufsize; + struct iwl_mld_mcast_key_data gtk[WOWLAN_GTK_KEYS_NUM]; + struct iwl_mld_mcast_key_data igtk; + struct iwl_mld_mcast_key_data bigtk[WOWLAN_BIGTK_KEYS_NUM]; + struct { + struct ieee80211_key_seq aes_seq[IWL_MAX_TID_COUNT]; + struct ieee80211_key_seq tkip_seq[IWL_MAX_TID_COUNT]; + + } ptk; +}; + +#define NETDETECT_QUERY_BUF_LEN \ + (sizeof(struct iwl_scan_offload_profile_match) * \ + IWL_SCAN_MAX_PROFILES_V2) + +/** + * struct iwl_mld_netdetect_res - contains netdetect results from + * match_info_notif + * @matched_profiles: bitmap of matched profiles, referencing the + * matches passed in the scan offload request + * @matches: array of match information, one for each match + */ +struct iwl_mld_netdetect_res { + u32 matched_profiles; + u8 matches[NETDETECT_QUERY_BUF_LEN]; +}; + +/** + * struct iwl_mld_resume_data - d3 resume flow data + * @notifs_expected: bitmap of expected notifications from fw, + * see &enum iwl_mld_d3_notif + * @notifs_received: bitmap of received notifications from fw, + * see &enum iwl_mld_d3_notif + * @d3_end_flags: bitmap of flags from d3_end_notif + * @notif_handling_err: error handling one of the resume notifications + * @wowlan_status: wowlan status data from all wowlan notifications + * @netdetect_res: contains netdetect results from match_info_notif + */ +struct iwl_mld_resume_data { + u32 notifs_expected; + u32 notifs_received; + u32 d3_end_flags; + bool notif_handling_err; + struct iwl_mld_wowlan_status *wowlan_status; + struct iwl_mld_netdetect_res *netdetect_res; +}; + +#define IWL_WOWLAN_WAKEUP_REASON_HAS_WAKEUP_PKT \ + (IWL_WOWLAN_WAKEUP_BY_MAGIC_PACKET | \ + IWL_WOWLAN_WAKEUP_BY_PATTERN | \ + IWL_WAKEUP_BY_PATTERN_IPV4_TCP_SYN |\ + IWL_WAKEUP_BY_PATTERN_IPV4_TCP_SYN_WILDCARD |\ + IWL_WAKEUP_BY_PATTERN_IPV6_TCP_SYN |\ + IWL_WAKEUP_BY_PATTERN_IPV6_TCP_SYN_WILDCARD) + +#define IWL_WOWLAN_OFFLOAD_TID 0 + +void iwl_mld_set_rekey_data(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct cfg80211_gtk_rekey_data *data) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mld_wowlan_data *wowlan_data = &mld_vif->wowlan_data; + + lockdep_assert_wiphy(mld->wiphy); + + wowlan_data->rekey_data.kek_len = data->kek_len; + wowlan_data->rekey_data.kck_len = data->kck_len; + memcpy(wowlan_data->rekey_data.kek, data->kek, data->kek_len); + memcpy(wowlan_data->rekey_data.kck, data->kck, data->kck_len); + wowlan_data->rekey_data.akm = data->akm & 0xFF; + wowlan_data->rekey_data.replay_ctr = + cpu_to_le64(be64_to_cpup((const __be64 *)data->replay_ctr)); + wowlan_data->rekey_data.valid = true; +} + +#if IS_ENABLED(CONFIG_IPV6) +void iwl_mld_ipv6_addr_change(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct inet6_dev *idev) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mld_wowlan_data *wowlan_data = &mld_vif->wowlan_data; + struct inet6_ifaddr *ifa; + int idx = 0; + + memset(wowlan_data->tentative_addrs, 0, + sizeof(wowlan_data->tentative_addrs)); + + read_lock_bh(&idev->lock); + list_for_each_entry(ifa, &idev->addr_list, if_list) { + wowlan_data->target_ipv6_addrs[idx] = ifa->addr; + if (ifa->flags & IFA_F_TENTATIVE) + __set_bit(idx, wowlan_data->tentative_addrs); + idx++; + if (idx >= IWL_PROTO_OFFLOAD_NUM_IPV6_ADDRS_MAX) + break; + } + read_unlock_bh(&idev->lock); + + wowlan_data->num_target_ipv6_addrs = idx; +} +#endif + +enum rt_status { + FW_ALIVE, + FW_NEEDS_RESET, + FW_ERROR, +}; + +static enum rt_status iwl_mld_check_err_tables(struct iwl_mld *mld, + struct ieee80211_vif *vif) +{ + u32 err_id; + + /* check for lmac1 error */ + if (iwl_fwrt_read_err_table(mld->trans, + mld->trans->dbg.lmac_error_event_table[0], + &err_id)) { + if (err_id == RF_KILL_INDICATOR_FOR_WOWLAN && vif) { + struct cfg80211_wowlan_wakeup wakeup = { + .rfkill_release = true, + }; + ieee80211_report_wowlan_wakeup(vif, &wakeup, + GFP_KERNEL); + + return FW_NEEDS_RESET; + } + return FW_ERROR; + } + + /* check if we have lmac2 set and check for error */ + if (iwl_fwrt_read_err_table(mld->trans, + mld->trans->dbg.lmac_error_event_table[1], + NULL)) + return FW_ERROR; + + /* check for umac error */ + if (iwl_fwrt_read_err_table(mld->trans, + mld->trans->dbg.umac_error_event_table, + NULL)) + return FW_ERROR; + + return FW_ALIVE; +} + +static bool iwl_mld_fw_needs_restart(struct iwl_mld *mld, + struct ieee80211_vif *vif) +{ + enum rt_status rt_status = iwl_mld_check_err_tables(mld, vif); + + if (rt_status == FW_ALIVE) + return false; + + if (rt_status == FW_ERROR) { + IWL_ERR(mld, "FW Error occurred during suspend\n"); + iwl_fwrt_dump_error_logs(&mld->fwrt); + iwl_dbg_tlv_time_point(&mld->fwrt, + IWL_FW_INI_TIME_POINT_FW_ASSERT, NULL); + } + + return true; +} + +static int +iwl_mld_netdetect_config(struct iwl_mld *mld, + struct ieee80211_vif *vif, + const struct cfg80211_wowlan *wowlan) +{ + int ret; + struct cfg80211_sched_scan_request *netdetect_cfg = + wowlan->nd_config; + struct ieee80211_scan_ies ies = {}; + + ret = iwl_mld_scan_stop(mld, IWL_MLD_SCAN_SCHED, true); + if (ret) + return ret; + + ret = iwl_mld_sched_scan_start(mld, vif, netdetect_cfg, &ies, + IWL_MLD_SCAN_NETDETECT); + return ret; +} + +static void +iwl_mld_le64_to_tkip_seq(__le64 le_pn, struct ieee80211_key_seq *seq) +{ + u64 pn = le64_to_cpu(le_pn); + + seq->tkip.iv16 = (u16)pn; + seq->tkip.iv32 = (u32)(pn >> 16); +} + +static void +iwl_mld_le64_to_aes_seq(__le64 le_pn, struct ieee80211_key_seq *seq) +{ + u64 pn = le64_to_cpu(le_pn); + + seq->ccmp.pn[0] = pn >> 40; + seq->ccmp.pn[1] = pn >> 32; + seq->ccmp.pn[2] = pn >> 24; + seq->ccmp.pn[3] = pn >> 16; + seq->ccmp.pn[4] = pn >> 8; + seq->ccmp.pn[5] = pn; +} + +static void +iwl_mld_convert_gtk_resume_seq(struct iwl_mld_mcast_key_data *gtk_data, + const struct iwl_wowlan_all_rsc_tsc_v5 *sc, + int rsc_idx) +{ + struct ieee80211_key_seq *aes_seq = gtk_data->gtk.aes_seq; + struct ieee80211_key_seq *tkip_seq = gtk_data->gtk.tkip_seq; + + if (rsc_idx >= ARRAY_SIZE(sc->mcast_rsc)) + return; + + /* We store both the TKIP and AES representations coming from the + * FW because we decode the data from there before we iterate + * the keys and know which type is used. + */ + for (int tid = 0; tid < IWL_MAX_TID_COUNT; tid++) { + iwl_mld_le64_to_tkip_seq(sc->mcast_rsc[rsc_idx][tid], + &tkip_seq[tid]); + iwl_mld_le64_to_aes_seq(sc->mcast_rsc[rsc_idx][tid], + &aes_seq[tid]); + } +} + +static void +iwl_mld_convert_gtk_resume_data(struct iwl_mld *mld, + struct iwl_mld_wowlan_status *wowlan_status, + const struct iwl_wowlan_gtk_status_v3 *gtk_data, + const struct iwl_wowlan_all_rsc_tsc_v5 *sc) +{ + int status_idx = 0; + + BUILD_BUG_ON(sizeof(wowlan_status->gtk[0].key) < + sizeof(gtk_data[0].key)); + BUILD_BUG_ON(ARRAY_SIZE(wowlan_status->gtk) < WOWLAN_GTK_KEYS_NUM); + + for (int notif_idx = 0; notif_idx < ARRAY_SIZE(wowlan_status->gtk); + notif_idx++) { + int rsc_idx; + + if (!(gtk_data[notif_idx].key_len)) + continue; + + wowlan_status->gtk[status_idx].len = + gtk_data[notif_idx].key_len; + wowlan_status->gtk[status_idx].flags = + gtk_data[notif_idx].key_flags; + wowlan_status->gtk[status_idx].id = + wowlan_status->gtk[status_idx].flags & + IWL_WOWLAN_GTK_IDX_MASK; + memcpy(wowlan_status->gtk[status_idx].key, + gtk_data[notif_idx].key, + sizeof(gtk_data[notif_idx].key)); + + /* The rsc for both gtk keys are stored in gtk[0]->sc->mcast_rsc + * The gtk ids can be any two numbers between 0 and 3, + * the id_map maps between the key id and the index in sc->mcast + */ + rsc_idx = + sc->mcast_key_id_map[wowlan_status->gtk[status_idx].id]; + iwl_mld_convert_gtk_resume_seq(&wowlan_status->gtk[status_idx], + sc, rsc_idx); + + /* if it's as long as the TKIP encryption key, copy MIC key */ + if (wowlan_status->gtk[status_idx].len == + NL80211_TKIP_DATA_OFFSET_TX_MIC_KEY) + memcpy(wowlan_status->gtk[status_idx].key + + NL80211_TKIP_DATA_OFFSET_RX_MIC_KEY, + gtk_data[notif_idx].tkip_mic_key, + sizeof(gtk_data[notif_idx].tkip_mic_key)); + status_idx++; + } +} + +static void +iwl_mld_convert_ptk_resume_seq(struct iwl_mld *mld, + struct iwl_mld_wowlan_status *wowlan_status, + const struct iwl_wowlan_all_rsc_tsc_v5 *sc) +{ + struct ieee80211_key_seq *aes_seq = wowlan_status->ptk.aes_seq; + struct ieee80211_key_seq *tkip_seq = wowlan_status->ptk.tkip_seq; + + BUILD_BUG_ON(ARRAY_SIZE(sc->ucast_rsc) != IWL_MAX_TID_COUNT); + + for (int tid = 0; tid < IWL_MAX_TID_COUNT; tid++) { + iwl_mld_le64_to_aes_seq(sc->ucast_rsc[tid], &aes_seq[tid]); + iwl_mld_le64_to_tkip_seq(sc->ucast_rsc[tid], &tkip_seq[tid]); + } +} + +static void +iwl_mld_convert_mcast_ipn(struct iwl_mld_mcast_key_data *key_status, + const struct iwl_wowlan_igtk_status *key) +{ + struct ieee80211_key_seq *seq = + &key_status->igtk_bigtk.cmac_gmac_seq; + u8 ipn_len = ARRAY_SIZE(key->ipn); + + BUILD_BUG_ON(ipn_len != ARRAY_SIZE(seq->aes_gmac.pn)); + BUILD_BUG_ON(ipn_len != ARRAY_SIZE(seq->aes_cmac.pn)); + BUILD_BUG_ON(offsetof(struct ieee80211_key_seq, aes_gmac) != + offsetof(struct ieee80211_key_seq, aes_cmac)); + + /* mac80211 expects big endian for memcmp() to work, convert. + * We don't have the key cipher yet so copy to both to cmac and gmac + */ + for (int i = 0; i < ipn_len; i++) { + seq->aes_gmac.pn[i] = key->ipn[ipn_len - i - 1]; + seq->aes_cmac.pn[i] = key->ipn[ipn_len - i - 1]; + } +} + +static void +iwl_mld_convert_igtk_resume_data(struct iwl_mld_wowlan_status *wowlan_status, + const struct iwl_wowlan_igtk_status *igtk) +{ + BUILD_BUG_ON(sizeof(wowlan_status->igtk.key) < sizeof(igtk->key)); + + if (!igtk->key_len) + return; + + wowlan_status->igtk.len = igtk->key_len; + wowlan_status->igtk.flags = igtk->key_flags; + wowlan_status->igtk.id = + u32_get_bits(igtk->key_flags, + IWL_WOWLAN_IGTK_BIGTK_IDX_MASK) + + WOWLAN_IGTK_MIN_INDEX; + + memcpy(wowlan_status->igtk.key, igtk->key, sizeof(igtk->key)); + iwl_mld_convert_mcast_ipn(&wowlan_status->igtk, igtk); +} + +static void +iwl_mld_convert_bigtk_resume_data(struct iwl_mld_wowlan_status *wowlan_status, + const struct iwl_wowlan_igtk_status *bigtk) +{ + int status_idx = 0; + + BUILD_BUG_ON(ARRAY_SIZE(wowlan_status->bigtk) < WOWLAN_BIGTK_KEYS_NUM); + + for (int notif_idx = 0; notif_idx < WOWLAN_BIGTK_KEYS_NUM; + notif_idx++) { + if (!bigtk[notif_idx].key_len) + continue; + + wowlan_status->bigtk[status_idx].len = bigtk[notif_idx].key_len; + wowlan_status->bigtk[status_idx].flags = + bigtk[notif_idx].key_flags; + wowlan_status->bigtk[status_idx].id = + u32_get_bits(bigtk[notif_idx].key_flags, + IWL_WOWLAN_IGTK_BIGTK_IDX_MASK) + + WOWLAN_BIGTK_MIN_INDEX; + + BUILD_BUG_ON(sizeof(wowlan_status->bigtk[status_idx].key) < + sizeof(bigtk[notif_idx].key)); + memcpy(wowlan_status->bigtk[status_idx].key, + bigtk[notif_idx].key, sizeof(bigtk[notif_idx].key)); + iwl_mld_convert_mcast_ipn(&wowlan_status->bigtk[status_idx], + &bigtk[notif_idx]); + status_idx++; + } +} + +static bool +iwl_mld_handle_wowlan_info_notif(struct iwl_mld *mld, + struct iwl_mld_wowlan_status *wowlan_status, + struct iwl_rx_packet *pkt) +{ + const struct iwl_wowlan_info_notif *notif = (void *)pkt->data; + u32 expected_len, len = iwl_rx_packet_payload_len(pkt); + + expected_len = sizeof(*notif); + + if (IWL_FW_CHECK(mld, len < expected_len, + "Invalid wowlan_info_notif (expected=%ud got=%ud)\n", + expected_len, len)) + return true; + + if (IWL_FW_CHECK(mld, notif->tid_offloaded_tx != IWL_WOWLAN_OFFLOAD_TID, + "Invalid tid_offloaded_tx %d\n", + wowlan_status->tid_offloaded_tx)) + return true; + + iwl_mld_convert_gtk_resume_data(mld, wowlan_status, notif->gtk, + ¬if->gtk[0].sc); + iwl_mld_convert_ptk_resume_seq(mld, wowlan_status, ¬if->gtk[0].sc); + /* only one igtk is passed by FW */ + iwl_mld_convert_igtk_resume_data(wowlan_status, ¬if->igtk[0]); + iwl_mld_convert_bigtk_resume_data(wowlan_status, notif->bigtk); + + wowlan_status->replay_ctr = le64_to_cpu(notif->replay_ctr); + wowlan_status->pattern_number = le16_to_cpu(notif->pattern_number); + + wowlan_status->tid_offloaded_tx = notif->tid_offloaded_tx; + wowlan_status->last_qos_seq = le16_to_cpu(notif->qos_seq_ctr); + wowlan_status->num_of_gtk_rekeys = + le32_to_cpu(notif->num_of_gtk_rekeys); + wowlan_status->wakeup_reasons = le32_to_cpu(notif->wakeup_reasons); + return false; + /* TODO: mlo_links (task=MLO)*/ +} + +static bool +iwl_mld_handle_wake_pkt_notif(struct iwl_mld *mld, + struct iwl_mld_wowlan_status *wowlan_status, + struct iwl_rx_packet *pkt) +{ + const struct iwl_wowlan_wake_pkt_notif *notif = (void *)pkt->data; + u32 actual_size, len = iwl_rx_packet_payload_len(pkt); + u32 expected_size = le32_to_cpu(notif->wake_packet_length); + + if (IWL_FW_CHECK(mld, len < sizeof(*notif), + "Invalid WoWLAN wake packet notification (expected size=%zu got=%u)\n", + sizeof(*notif), len)) + return true; + + if (IWL_FW_CHECK(mld, !(wowlan_status->wakeup_reasons & + IWL_WOWLAN_WAKEUP_REASON_HAS_WAKEUP_PKT), + "Got wake packet but wakeup reason is %x\n", + wowlan_status->wakeup_reasons)) + return true; + + actual_size = len - offsetof(struct iwl_wowlan_wake_pkt_notif, + wake_packet); + + /* actual_size got the padding from the notification, remove it. */ + if (expected_size < actual_size) + actual_size = expected_size; + wowlan_status->wake_packet = kmemdup(notif->wake_packet, actual_size, + GFP_ATOMIC); + if (!wowlan_status->wake_packet) + return true; + + wowlan_status->wake_packet_length = expected_size; + wowlan_status->wake_packet_bufsize = actual_size; + + return false; +} + +static void +iwl_mld_set_wake_packet(struct iwl_mld *mld, + struct ieee80211_vif *vif, + const struct iwl_mld_wowlan_status *wowlan_status, + struct cfg80211_wowlan_wakeup *wakeup, + struct sk_buff **_pkt) +{ + int pkt_bufsize = wowlan_status->wake_packet_bufsize; + int expected_pktlen = wowlan_status->wake_packet_length; + const u8 *pktdata = wowlan_status->wake_packet; + const struct ieee80211_hdr *hdr = (const void *)pktdata; + int truncated = expected_pktlen - pkt_bufsize; + + if (ieee80211_is_data(hdr->frame_control)) { + int hdrlen = ieee80211_hdrlen(hdr->frame_control); + int ivlen = 0, icvlen = 4; /* also FCS */ + + struct sk_buff *pkt = alloc_skb(pkt_bufsize, GFP_KERNEL); + *_pkt = pkt; + if (!pkt) + return; + + skb_put_data(pkt, pktdata, hdrlen); + pktdata += hdrlen; + pkt_bufsize -= hdrlen; + + /* if truncated, FCS/ICV is (partially) gone */ + if (truncated >= icvlen) { + truncated -= icvlen; + icvlen = 0; + } else { + icvlen -= truncated; + truncated = 0; + } + + pkt_bufsize -= ivlen + icvlen; + pktdata += ivlen; + + skb_put_data(pkt, pktdata, pkt_bufsize); + + if (ieee80211_data_to_8023(pkt, vif->addr, vif->type)) + return; + wakeup->packet = pkt->data; + wakeup->packet_present_len = pkt->len; + wakeup->packet_len = pkt->len - truncated; + wakeup->packet_80211 = false; + } else { + int fcslen = 4; + + if (truncated >= 4) { + truncated -= 4; + fcslen = 0; + } else { + fcslen -= truncated; + truncated = 0; + } + pkt_bufsize -= fcslen; + wakeup->packet = wowlan_status->wake_packet; + wakeup->packet_present_len = pkt_bufsize; + wakeup->packet_len = expected_pktlen - truncated; + wakeup->packet_80211 = true; + } +} + +static void +iwl_mld_report_wowlan_wakeup(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct iwl_mld_wowlan_status *wowlan_status) +{ + struct sk_buff *pkt = NULL; + struct cfg80211_wowlan_wakeup wakeup = { + .pattern_idx = -1, + }; + u32 reasons = wowlan_status->wakeup_reasons; + + if (reasons == IWL_WOWLAN_WAKEUP_BY_NON_WIRELESS) { + ieee80211_report_wowlan_wakeup(vif, NULL, GFP_KERNEL); + return; + } + + pm_wakeup_event(mld->dev, 0); + + if (reasons & IWL_WOWLAN_WAKEUP_BY_MAGIC_PACKET) + wakeup.magic_pkt = true; + + if (reasons & IWL_WOWLAN_WAKEUP_BY_PATTERN) + wakeup.pattern_idx = + wowlan_status->pattern_number; + + if (reasons & (IWL_WOWLAN_WAKEUP_BY_DISCONNECTION_ON_MISSED_BEACON | + IWL_WOWLAN_WAKEUP_BY_DISCONNECTION_ON_DEAUTH | + IWL_WOWLAN_WAKEUP_BY_GTK_REKEY_FAILURE)) + wakeup.disconnect = true; + + if (reasons & IWL_WOWLAN_WAKEUP_BY_GTK_REKEY_FAILURE) + wakeup.gtk_rekey_failure = true; + + if (reasons & IWL_WOWLAN_WAKEUP_BY_RFKILL_DEASSERTED) + wakeup.rfkill_release = true; + + if (reasons & IWL_WOWLAN_WAKEUP_BY_EAPOL_REQUEST) + wakeup.eap_identity_req = true; + + if (reasons & IWL_WOWLAN_WAKEUP_BY_FOUR_WAY_HANDSHAKE) + wakeup.four_way_handshake = true; + + if (reasons & IWL_WOWLAN_WAKEUP_BY_REM_WAKE_LINK_LOSS) + wakeup.tcp_connlost = true; + + if (reasons & IWL_WOWLAN_WAKEUP_BY_REM_WAKE_SIGNATURE_TABLE) + wakeup.tcp_nomoretokens = true; + + if (reasons & IWL_WOWLAN_WAKEUP_BY_REM_WAKE_WAKEUP_PACKET) + wakeup.tcp_match = true; + + if (reasons & IWL_WAKEUP_BY_11W_UNPROTECTED_DEAUTH_OR_DISASSOC) + wakeup.unprot_deauth_disassoc = true; + + if (wowlan_status->wake_packet) + iwl_mld_set_wake_packet(mld, vif, wowlan_status, &wakeup, &pkt); + + ieee80211_report_wowlan_wakeup(vif, &wakeup, GFP_KERNEL); + kfree_skb(pkt); +} + +static void +iwl_mld_set_key_rx_seq_tids(struct ieee80211_key_conf *key, + struct ieee80211_key_seq *seq) +{ + int tid; + + for (tid = 0; tid < IWL_MAX_TID_COUNT; tid++) + ieee80211_set_key_rx_seq(key, tid, &seq[tid]); +} + +static void +iwl_mld_set_key_rx_seq(struct ieee80211_key_conf *key, + struct iwl_mld_mcast_key_data *key_data) +{ + switch (key->cipher) { + case WLAN_CIPHER_SUITE_CCMP: + case WLAN_CIPHER_SUITE_GCMP: + case WLAN_CIPHER_SUITE_GCMP_256: + iwl_mld_set_key_rx_seq_tids(key, + key_data->gtk.aes_seq); + break; + case WLAN_CIPHER_SUITE_TKIP: + iwl_mld_set_key_rx_seq_tids(key, + key_data->gtk.tkip_seq); + break; + case WLAN_CIPHER_SUITE_BIP_GMAC_128: + case WLAN_CIPHER_SUITE_BIP_GMAC_256: + case WLAN_CIPHER_SUITE_BIP_CMAC_256: + case WLAN_CIPHER_SUITE_AES_CMAC: + /* igtk/bigtk ciphers*/ + ieee80211_set_key_rx_seq(key, 0, + &key_data->igtk_bigtk.cmac_gmac_seq); + break; + default: + WARN_ON(1); + } +} + +static void +iwl_mld_d3_update_mcast_key(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct iwl_mld_wowlan_status *wowlan_status, + struct ieee80211_key_conf *key, + struct iwl_mld_mcast_key_data *key_data) +{ + if (key->keyidx != key_data->id && + (key->keyidx < 4 || key->keyidx > 5)) { + IWL_ERR(mld, + "Unexpected keyId mismatch. Old keyId:%d, New keyId:%d\n", + key->keyidx, key_data->id); + return; + } + + /* All installed keys are sent by the FW, even weren't + * rekeyed during D3. + * We remove an existing key if it has the same index as + * a new key and a rekey has occurred during d3 + */ + if (wowlan_status->num_of_gtk_rekeys && key_data->len) { + if (key->keyidx == 4 || key->keyidx == 5) { + struct iwl_mld_vif *mld_vif = + iwl_mld_vif_from_mac80211(vif); + struct iwl_mld_link *mld_link; + int link_id = vif->active_links ? + __ffs(vif->active_links) : 0; + + mld_link = iwl_mld_link_dereference_check(mld_vif, + link_id); + if (WARN_ON(!mld_link)) + return; + + if (mld_link->igtk == key) + mld_link->igtk = NULL; + mld->num_igtks--; + } + + ieee80211_remove_key(key); + return; + } + + iwl_mld_set_key_rx_seq(key, key_data); +} + +static void +iwl_mld_update_ptk_rx_seq(struct iwl_mld *mld, + struct iwl_mld_wowlan_status *wowlan_status, + struct ieee80211_sta *sta, + struct ieee80211_key_conf *key, + bool is_tkip) +{ + struct iwl_mld_sta *mld_sta = + iwl_mld_sta_from_mac80211(sta); + struct iwl_mld_ptk_pn *mld_ptk_pn = + wiphy_dereference(mld->wiphy, + mld_sta->ptk_pn[key->keyidx]); + + iwl_mld_set_key_rx_seq_tids(key, is_tkip ? + wowlan_status->ptk.tkip_seq : + wowlan_status->ptk.aes_seq); + if (is_tkip) + return; + + if (WARN_ON(!mld_ptk_pn)) + return; + + for (int tid = 0; tid < IWL_MAX_TID_COUNT; tid++) { + for (int i = 1; i < mld->trans->num_rx_queues; i++) + memcpy(mld_ptk_pn->q[i].pn[tid], + wowlan_status->ptk.aes_seq[tid].ccmp.pn, + IEEE80211_CCMP_PN_LEN); + } +} + +static void +iwl_mld_resume_keys_iter(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + struct ieee80211_key_conf *key, + void *_data) +{ + struct iwl_mld_resume_key_iter_data *data = _data; + struct iwl_mld_wowlan_status *wowlan_status = data->wowlan_status; + u8 status_idx; + + /* TODO: check key link id (task=MLO) */ + if (data->unhandled_cipher) + return; + + switch (key->cipher) { + case WLAN_CIPHER_SUITE_WEP40: + case WLAN_CIPHER_SUITE_WEP104: + /* ignore WEP completely, nothing to do */ + return; + case WLAN_CIPHER_SUITE_CCMP: + case WLAN_CIPHER_SUITE_GCMP: + case WLAN_CIPHER_SUITE_GCMP_256: + case WLAN_CIPHER_SUITE_TKIP: + if (sta) { + iwl_mld_update_ptk_rx_seq(data->mld, wowlan_status, + sta, key, + key->cipher == + WLAN_CIPHER_SUITE_TKIP); + return; + } + + if (WARN_ON(data->gtk_cipher && + data->gtk_cipher != key->cipher)) + return; + + data->gtk_cipher = key->cipher; + status_idx = key->keyidx == wowlan_status->gtk[1].id; + iwl_mld_d3_update_mcast_key(data->mld, vif, wowlan_status, key, + &wowlan_status->gtk[status_idx]); + break; + case WLAN_CIPHER_SUITE_BIP_GMAC_128: + case WLAN_CIPHER_SUITE_BIP_GMAC_256: + case WLAN_CIPHER_SUITE_BIP_CMAC_256: + case WLAN_CIPHER_SUITE_AES_CMAC: + if (key->keyidx == 4 || key->keyidx == 5) { + if (WARN_ON(data->igtk_cipher && + data->igtk_cipher != key->cipher)) + return; + + data->igtk_cipher = key->cipher; + iwl_mld_d3_update_mcast_key(data->mld, vif, + wowlan_status, + key, &wowlan_status->igtk); + } + if (key->keyidx == 6 || key->keyidx == 7) { + if (WARN_ON(data->bigtk_cipher && + data->bigtk_cipher != key->cipher)) + return; + + data->bigtk_cipher = key->cipher; + status_idx = key->keyidx == wowlan_status->bigtk[1].id; + iwl_mld_d3_update_mcast_key(data->mld, vif, + wowlan_status, key, + &wowlan_status->bigtk[status_idx]); + } + break; + default: + data->unhandled_cipher = true; + return; + } + data->num_keys++; +} + +static bool +iwl_mld_add_mcast_rekey(struct ieee80211_vif *vif, + struct iwl_mld *mld, + struct iwl_mld_mcast_key_data *key_data, + struct ieee80211_bss_conf *link_conf, + u32 cipher) +{ + struct ieee80211_key_conf *key_config; + struct { + struct ieee80211_key_conf conf; + u8 key[WOWLAN_KEY_MAX_SIZE]; + } conf = { + .conf.cipher = cipher, + .conf.keyidx = key_data->id, + }; + int link_id = vif->active_links ? __ffs(vif->active_links) : -1; + + BUILD_BUG_ON(WLAN_KEY_LEN_CCMP != WLAN_KEY_LEN_GCMP); + BUILD_BUG_ON(sizeof(conf.key) < WLAN_KEY_LEN_CCMP); + BUILD_BUG_ON(sizeof(conf.key) < WLAN_KEY_LEN_GCMP_256); + BUILD_BUG_ON(sizeof(conf.key) < WLAN_KEY_LEN_TKIP); + BUILD_BUG_ON(sizeof(conf.key) < WLAN_KEY_LEN_BIP_GMAC_128); + BUILD_BUG_ON(sizeof(conf.key) < WLAN_KEY_LEN_BIP_GMAC_256); + BUILD_BUG_ON(sizeof(conf.key) < WLAN_KEY_LEN_AES_CMAC); + BUILD_BUG_ON(sizeof(conf.key) < sizeof(key_data->key)); + + if (!key_data->len) + return true; + + switch (cipher) { + case WLAN_CIPHER_SUITE_CCMP: + case WLAN_CIPHER_SUITE_GCMP: + conf.conf.keylen = WLAN_KEY_LEN_CCMP; + break; + case WLAN_CIPHER_SUITE_GCMP_256: + conf.conf.keylen = WLAN_KEY_LEN_GCMP_256; + break; + case WLAN_CIPHER_SUITE_TKIP: + conf.conf.keylen = WLAN_KEY_LEN_TKIP; + break; + case WLAN_CIPHER_SUITE_BIP_GMAC_128: + conf.conf.keylen = WLAN_KEY_LEN_BIP_GMAC_128; + break; + case WLAN_CIPHER_SUITE_BIP_GMAC_256: + conf.conf.keylen = WLAN_KEY_LEN_BIP_GMAC_256; + break; + case WLAN_CIPHER_SUITE_AES_CMAC: + conf.conf.keylen = WLAN_KEY_LEN_AES_CMAC; + break; + case WLAN_CIPHER_SUITE_BIP_CMAC_256: + conf.conf.keylen = WLAN_KEY_LEN_BIP_CMAC_256; + break; + default: + WARN_ON(1); + } + + memcpy(conf.conf.key, key_data->key, conf.conf.keylen); + key_config = ieee80211_gtk_rekey_add(vif, &conf.conf, link_id); + if (IS_ERR(key_config)) + return false; + + iwl_mld_set_key_rx_seq(key_config, key_data); + + /* The FW holds only one igtk so we keep track of the valid one */ + if (key_config->keyidx == 4 || key_config->keyidx == 5) { + struct iwl_mld_link *mld_link = + iwl_mld_link_from_mac80211(link_conf); + mld_link->igtk = key_config; + mld->num_igtks++; + } + return true; +} + +static bool +iwl_mld_add_all_rekeys(struct ieee80211_vif *vif, + struct iwl_mld_wowlan_status *wowlan_status, + struct iwl_mld_resume_key_iter_data *key_iter_data, + struct ieee80211_bss_conf *link_conf) +{ + int i; + + for (i = 0; i < ARRAY_SIZE(wowlan_status->gtk); i++) + if (!iwl_mld_add_mcast_rekey(vif, key_iter_data->mld, + &wowlan_status->gtk[i], + link_conf, + key_iter_data->gtk_cipher)) + return false; + + if (!iwl_mld_add_mcast_rekey(vif, key_iter_data->mld, + &wowlan_status->igtk, + link_conf, key_iter_data->igtk_cipher)) + return false; + + for (i = 0; i < ARRAY_SIZE(wowlan_status->bigtk); i++) + if (!iwl_mld_add_mcast_rekey(vif, key_iter_data->mld, + &wowlan_status->bigtk[i], + link_conf, + key_iter_data->bigtk_cipher)) + return false; + + return true; +} + +static bool +iwl_mld_update_sec_keys(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct iwl_mld_wowlan_status *wowlan_status) +{ + int link_id = vif->active_links ? __ffs(vif->active_links) : 0; + struct ieee80211_bss_conf *link_conf = + link_conf_dereference_protected(vif, link_id); + __be64 replay_ctr = cpu_to_be64(wowlan_status->replay_ctr); + struct iwl_mld_resume_key_iter_data key_iter_data = { + .mld = mld, + .wowlan_status = wowlan_status, + }; + + if (WARN_ON(!link_conf)) + return false; + + ieee80211_iter_keys(mld->hw, vif, iwl_mld_resume_keys_iter, + &key_iter_data); + + if (key_iter_data.unhandled_cipher) + return false; + + IWL_DEBUG_WOWLAN(mld, + "Number of installed keys: %d, Number of rekeys: %d\n", + key_iter_data.num_keys, + wowlan_status->num_of_gtk_rekeys); + + if (!key_iter_data.num_keys || !wowlan_status->num_of_gtk_rekeys) + return true; + + iwl_mld_add_all_rekeys(vif, wowlan_status, &key_iter_data, + link_conf); + + ieee80211_gtk_rekey_notify(vif, link_conf->bssid, + (void *)&replay_ctr, GFP_KERNEL); + /* TODO: MLO rekey (task=MLO) */ + return true; +} + +static bool +iwl_mld_process_wowlan_status(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct iwl_mld_wowlan_status *wowlan_status) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct ieee80211_sta *ap_sta = mld_vif->ap_sta; + struct iwl_mld_txq *mld_txq; + + iwl_mld_report_wowlan_wakeup(mld, vif, wowlan_status); + + if (WARN_ON(!ap_sta)) + return false; + + mld_txq = + iwl_mld_txq_from_mac80211(ap_sta->txq[wowlan_status->tid_offloaded_tx]); + + /* Update the pointers of the Tx queue that may have moved during + * suspend if the firmware sent frames. + * The firmware stores last-used value, we store next value. + */ + WARN_ON(!mld_txq->status.allocated); + iwl_trans_set_q_ptrs(mld->trans, mld_txq->fw_id, + (wowlan_status->last_qos_seq + + 0x10) >> 4); + + if (!iwl_mld_update_sec_keys(mld, vif, wowlan_status)) + return false; + + if (wowlan_status->wakeup_reasons & + (IWL_WOWLAN_WAKEUP_BY_DISCONNECTION_ON_MISSED_BEACON | + IWL_WOWLAN_WAKEUP_BY_DISCONNECTION_ON_DEAUTH | + IWL_WOWLAN_WAKEUP_BY_GTK_REKEY_FAILURE)) + return false; + + return true; +} + +static bool +iwl_mld_netdetect_match_info_handler(struct iwl_mld *mld, + struct iwl_mld_resume_data *resume_data, + struct iwl_rx_packet *pkt) +{ + struct iwl_mld_netdetect_res *results = resume_data->netdetect_res; + const struct iwl_scan_offload_match_info *notif = (void *)pkt->data; + u32 len = iwl_rx_packet_payload_len(pkt); + + if (IWL_FW_CHECK(mld, !mld->netdetect, + "Got scan match info notif when mld->netdetect==%d\n", + mld->netdetect)) + return true; + + if (IWL_FW_CHECK(mld, len < sizeof(*notif), + "Invalid scan offload match notif of length: %d\n", + len)) + return true; + + if (IWL_FW_CHECK(mld, resume_data->wowlan_status->wakeup_reasons != + IWL_WOWLAN_WAKEUP_BY_NON_WIRELESS, + "Ignore scan match info: unexpected wakeup reason (expected=0x%x got=0x%x)\n", + IWL_WOWLAN_WAKEUP_BY_NON_WIRELESS, + resume_data->wowlan_status->wakeup_reasons)) + return true; + + results->matched_profiles = le32_to_cpu(notif->matched_profiles); + IWL_DEBUG_WOWLAN(mld, "number of matched profiles=%u\n", + results->matched_profiles); + + if (results->matched_profiles) + memcpy(results->matches, notif->matches, + NETDETECT_QUERY_BUF_LEN); + + /* No scan should be active at this point */ + mld->scan.status = 0; + memset(mld->scan.uid_status, 0, sizeof(mld->scan.uid_status)); + return false; +} + +static void +iwl_mld_set_netdetect_info(struct iwl_mld *mld, + const struct cfg80211_sched_scan_request *netdetect_cfg, + struct cfg80211_wowlan_nd_info *netdetect_info, + struct iwl_mld_netdetect_res *netdetect_res, + unsigned long matched_profiles) +{ + int i; + + for_each_set_bit(i, &matched_profiles, netdetect_cfg->n_match_sets) { + struct cfg80211_wowlan_nd_match *match; + int idx, j, n_channels = 0; + struct iwl_scan_offload_profile_match *matches = + (void *)netdetect_res->matches; + + for (int k = 0; k < SCAN_OFFLOAD_MATCHING_CHANNELS_LEN; k++) + n_channels += + hweight8(matches[i].matching_channels[k]); + match = kzalloc(struct_size(match, channels, n_channels), + GFP_KERNEL); + if (!match) + return; + + netdetect_info->matches[netdetect_info->n_matches++] = match; + + /* We inverted the order of the SSIDs in the scan + * request, so invert the index here. + */ + idx = netdetect_cfg->n_match_sets - i - 1; + match->ssid.ssid_len = + netdetect_cfg->match_sets[idx].ssid.ssid_len; + memcpy(match->ssid.ssid, + netdetect_cfg->match_sets[idx].ssid.ssid, + match->ssid.ssid_len); + + if (netdetect_cfg->n_channels < n_channels) + continue; + + for_each_set_bit(j, + (unsigned long *)&matches[i].matching_channels[0], + sizeof(matches[i].matching_channels)) + match->channels[match->n_channels++] = + netdetect_cfg->channels[j]->center_freq; + } +} + +static void +iwl_mld_process_netdetect_res(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct iwl_mld_resume_data *resume_data) +{ + struct cfg80211_wowlan_nd_info *netdetect_info = NULL; + const struct cfg80211_sched_scan_request *netdetect_cfg; + struct cfg80211_wowlan_wakeup wakeup = { + .pattern_idx = -1, + }; + struct cfg80211_wowlan_wakeup *wakeup_report = &wakeup; + unsigned long matched_profiles; + u32 wakeup_reasons; + int n_matches; + + lockdep_assert_wiphy(mld->wiphy); + + if (WARN_ON(!mld->wiphy->wowlan_config || + !mld->wiphy->wowlan_config->nd_config)) { + IWL_DEBUG_WOWLAN(mld, + "Netdetect isn't configured on resume flow\n"); + goto out; + } + + netdetect_cfg = mld->wiphy->wowlan_config->nd_config; + wakeup_reasons = resume_data->wowlan_status->wakeup_reasons; + + if (wakeup_reasons & IWL_WOWLAN_WAKEUP_BY_RFKILL_DEASSERTED) + wakeup.rfkill_release = true; + + if (wakeup_reasons != IWL_WOWLAN_WAKEUP_BY_NON_WIRELESS) + goto out; + + if (!resume_data->netdetect_res->matched_profiles) { + IWL_DEBUG_WOWLAN(mld, + "Netdetect results aren't valid\n"); + wakeup_report = NULL; + goto out; + } + + matched_profiles = resume_data->netdetect_res->matched_profiles; + if (!netdetect_cfg->n_match_sets) { + IWL_DEBUG_WOWLAN(mld, + "No netdetect match sets are configured\n"); + goto out; + } + n_matches = hweight_long(matched_profiles); + netdetect_info = kzalloc(struct_size(netdetect_info, matches, + n_matches), GFP_KERNEL); + if (netdetect_info) + iwl_mld_set_netdetect_info(mld, netdetect_cfg, netdetect_info, + resume_data->netdetect_res, + matched_profiles); + + wakeup.net_detect = netdetect_info; + out: + ieee80211_report_wowlan_wakeup(vif, wakeup_report, GFP_KERNEL); + if (netdetect_info) { + for (int i = 0; i < netdetect_info->n_matches; i++) + kfree(netdetect_info->matches[i]); + kfree(netdetect_info); + } +} + +static bool iwl_mld_handle_d3_notif(struct iwl_notif_wait_data *notif_wait, + struct iwl_rx_packet *pkt, void *data) +{ + struct iwl_mld_resume_data *resume_data = data; + struct iwl_mld *mld = + container_of(notif_wait, struct iwl_mld, notif_wait); + + switch (WIDE_ID(pkt->hdr.group_id, pkt->hdr.cmd)) { + case WIDE_ID(PROT_OFFLOAD_GROUP, WOWLAN_INFO_NOTIFICATION): { + if (resume_data->notifs_received & IWL_D3_NOTIF_WOWLAN_INFO) { + IWL_DEBUG_WOWLAN(mld, + "got additional wowlan_info notif\n"); + break; + } + resume_data->notif_handling_err = + iwl_mld_handle_wowlan_info_notif(mld, + resume_data->wowlan_status, + pkt); + resume_data->notifs_received |= IWL_D3_NOTIF_WOWLAN_INFO; + + if (resume_data->wowlan_status->wakeup_reasons & + IWL_WOWLAN_WAKEUP_REASON_HAS_WAKEUP_PKT) + resume_data->notifs_expected |= + IWL_D3_NOTIF_WOWLAN_WAKE_PKT; + break; + } + case WIDE_ID(PROT_OFFLOAD_GROUP, WOWLAN_WAKE_PKT_NOTIFICATION): { + if (resume_data->notifs_received & + IWL_D3_NOTIF_WOWLAN_WAKE_PKT) { + /* We shouldn't get two wake packet notifications */ + IWL_DEBUG_WOWLAN(mld, + "Got additional wowlan wake packet notification\n"); + break; + } + resume_data->notif_handling_err = + iwl_mld_handle_wake_pkt_notif(mld, + resume_data->wowlan_status, + pkt); + resume_data->notifs_received |= IWL_D3_NOTIF_WOWLAN_WAKE_PKT; + break; + } + case WIDE_ID(SCAN_GROUP, OFFLOAD_MATCH_INFO_NOTIF): { + if (resume_data->notifs_received & IWL_D3_ND_MATCH_INFO) { + IWL_ERR(mld, + "Got additional netdetect match info\n"); + break; + } + + resume_data->notif_handling_err = + iwl_mld_netdetect_match_info_handler(mld, resume_data, + pkt); + resume_data->notifs_received |= IWL_D3_ND_MATCH_INFO; + break; + } + case WIDE_ID(PROT_OFFLOAD_GROUP, D3_END_NOTIFICATION): { + struct iwl_d3_end_notif *notif = (void *)pkt->data; + + resume_data->d3_end_flags = le32_to_cpu(notif->flags); + resume_data->notifs_received |= IWL_D3_NOTIF_D3_END_NOTIF; + break; + } + default: + WARN_ON(1); + } + + return resume_data->notifs_received == resume_data->notifs_expected; +} + +#define IWL_MLD_D3_NOTIF_TIMEOUT (HZ / 3) + +static int iwl_mld_wait_d3_notif(struct iwl_mld *mld, + struct iwl_mld_resume_data *resume_data, + bool with_wowlan) +{ + static const u16 wowlan_resume_notif[] = { + WIDE_ID(PROT_OFFLOAD_GROUP, WOWLAN_INFO_NOTIFICATION), + WIDE_ID(PROT_OFFLOAD_GROUP, WOWLAN_WAKE_PKT_NOTIFICATION), + WIDE_ID(SCAN_GROUP, OFFLOAD_MATCH_INFO_NOTIF), + WIDE_ID(PROT_OFFLOAD_GROUP, D3_END_NOTIFICATION) + }; + static const u16 d3_resume_notif[] = { + WIDE_ID(PROT_OFFLOAD_GROUP, D3_END_NOTIFICATION) + }; + struct iwl_notification_wait wait_d3_notif; + enum iwl_d3_status d3_status; + int ret; + + if (with_wowlan) + iwl_init_notification_wait(&mld->notif_wait, &wait_d3_notif, + wowlan_resume_notif, + ARRAY_SIZE(wowlan_resume_notif), + iwl_mld_handle_d3_notif, + resume_data); + else + iwl_init_notification_wait(&mld->notif_wait, &wait_d3_notif, + d3_resume_notif, + ARRAY_SIZE(d3_resume_notif), + iwl_mld_handle_d3_notif, + resume_data); + + ret = iwl_trans_d3_resume(mld->trans, &d3_status, false, false); + if (ret || d3_status != IWL_D3_STATUS_ALIVE) { + if (d3_status != IWL_D3_STATUS_ALIVE) { + IWL_INFO(mld, "Device was reset during suspend\n"); + ret = -ENOENT; + } else { + IWL_ERR(mld, "Transport resume failed\n"); + } + iwl_remove_notification(&mld->notif_wait, &wait_d3_notif); + return ret; + } + + ret = iwl_wait_notification(&mld->notif_wait, &wait_d3_notif, + IWL_MLD_D3_NOTIF_TIMEOUT); + if (ret) + IWL_ERR(mld, "Couldn't get the d3 notif %d\n", ret); + + if (resume_data->notif_handling_err) + ret = -EIO; + + return ret; +} + +int iwl_mld_no_wowlan_suspend(struct iwl_mld *mld) +{ + struct iwl_d3_manager_config d3_cfg_cmd_data = {}; + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + IWL_DEBUG_WOWLAN(mld, "Starting the no wowlan suspend flow\n"); + + iwl_mld_low_latency_stop(mld); + + /* This will happen if iwl_mld_supsend failed with FW error */ + if (mld->trans->state == IWL_TRANS_NO_FW && + test_bit(STATUS_FW_ERROR, &mld->trans->status)) + return -ENODEV; + + ret = iwl_mld_update_device_power(mld, true); + if (ret) { + IWL_ERR(mld, + "d3 suspend: couldn't send power_device %d\n", ret); + goto out; + } + + ret = iwl_mld_send_cmd_pdu(mld, D3_CONFIG_CMD, + &d3_cfg_cmd_data); + if (ret) { + IWL_ERR(mld, + "d3 suspend: couldn't send D3_CONFIG_CMD %d\n", ret); + goto out; + } + + ret = iwl_trans_d3_suspend(mld->trans, false, false); + if (ret) { + IWL_ERR(mld, "d3 suspend: trans_d3_suspend failed %d\n", ret); + } else { + mld->trans->system_pm_mode = IWL_PLAT_PM_MODE_D3; + mld->fw_status.in_d3 = true; + } + + out: + if (ret) { + mld->trans->state = IWL_TRANS_NO_FW; + set_bit(STATUS_FW_ERROR, &mld->trans->status); + } + + return ret; +} + +int iwl_mld_no_wowlan_resume(struct iwl_mld *mld) +{ + struct iwl_mld_resume_data resume_data = { + .notifs_expected = + IWL_D3_NOTIF_D3_END_NOTIF, + }; + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + IWL_DEBUG_WOWLAN(mld, "Starting the no wowlan resume flow\n"); + + mld->trans->system_pm_mode = IWL_PLAT_PM_MODE_DISABLED; + mld->fw_status.in_d3 = false; + iwl_fw_dbg_read_d3_debug_data(&mld->fwrt); + + if (iwl_mld_fw_needs_restart(mld, NULL)) + ret = -ENODEV; + else + ret = iwl_mld_wait_d3_notif(mld, &resume_data, false); + + if (!ret && (resume_data.d3_end_flags & IWL_D0I3_RESET_REQUIRE)) + return -ENODEV; + + if (ret) { + mld->trans->state = IWL_TRANS_NO_FW; + set_bit(STATUS_FW_ERROR, &mld->trans->status); + return ret; + } + iwl_mld_low_latency_restart(mld); + + return iwl_mld_update_device_power(mld, false); +} + +static void +iwl_mld_aes_seq_to_le64_pn(struct ieee80211_key_conf *key, + __le64 *key_rsc) +{ + for (int i = 0; i < IWL_MAX_TID_COUNT; i++) { + struct ieee80211_key_seq seq; + u8 *pn = key->cipher == WLAN_CIPHER_SUITE_CCMP ? seq.ccmp.pn : + seq.gcmp.pn; + + ieee80211_get_key_rx_seq(key, i, &seq); + key_rsc[i] = cpu_to_le64((u64)pn[5] | + ((u64)pn[4] << 8) | + ((u64)pn[3] << 16) | + ((u64)pn[2] << 24) | + ((u64)pn[1] << 32) | + ((u64)pn[0] << 40)); + } +} + +static void +iwl_mld_suspend_set_ucast_pn(struct iwl_mld *mld, struct ieee80211_sta *sta, + struct ieee80211_key_conf *key, __le64 *key_rsc) +{ + struct iwl_mld_sta *mld_sta = + iwl_mld_sta_from_mac80211(sta); + struct iwl_mld_ptk_pn *mld_ptk_pn; + + if (WARN_ON(key->keyidx >= ARRAY_SIZE(mld_sta->ptk_pn))) + return; + + mld_ptk_pn = wiphy_dereference(mld->wiphy, + mld_sta->ptk_pn[key->keyidx]); + if (WARN_ON(!mld_ptk_pn)) + return; + + for (int tid = 0; tid < IWL_MAX_TID_COUNT; tid++) { + struct ieee80211_key_seq seq; + u8 *max_pn = seq.ccmp.pn; + + /* get the PN from mac80211, used on the default queue */ + ieee80211_get_key_rx_seq(key, tid, &seq); + + /* and use the internal data for all queues */ + for (int que = 1; que < mld->trans->num_rx_queues; que++) { + u8 *cur_pn = mld_ptk_pn->q[que].pn[tid]; + + if (memcmp(max_pn, cur_pn, IEEE80211_CCMP_PN_LEN) < 0) + max_pn = cur_pn; + } + key_rsc[tid] = cpu_to_le64((u64)max_pn[5] | + ((u64)max_pn[4] << 8) | + ((u64)max_pn[3] << 16) | + ((u64)max_pn[2] << 24) | + ((u64)max_pn[1] << 32) | + ((u64)max_pn[0] << 40)); + } +} + +static void +iwl_mld_suspend_convert_tkip_ipn(struct ieee80211_key_conf *key, + __le64 *rsc) +{ + struct ieee80211_key_seq seq; + + for (int i = 0; i < IWL_MAX_TID_COUNT; i++) { + ieee80211_get_key_rx_seq(key, i, &seq); + rsc[i] = + cpu_to_le64(((u64)seq.tkip.iv32 << 16) | + seq.tkip.iv16); + } +} + +static void +iwl_mld_suspend_key_data_iter(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + struct ieee80211_key_conf *key, + void *_data) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct iwl_mld_suspend_key_iter_data *data = _data; + __le64 *key_rsc; + __le32 cipher = 0; + + switch (key->cipher) { + case WLAN_CIPHER_SUITE_CCMP: + cipher = cpu_to_le32(STA_KEY_FLG_CCM); + fallthrough; + case WLAN_CIPHER_SUITE_GCMP: + case WLAN_CIPHER_SUITE_GCMP_256: + if (!cipher) + cipher = cpu_to_le32(STA_KEY_FLG_GCMP); + fallthrough; + case WLAN_CIPHER_SUITE_TKIP: + if (!cipher) + cipher = cpu_to_le32(STA_KEY_FLG_TKIP); + if (sta) { + key_rsc = data->rsc->ucast_rsc; + if (key->cipher == WLAN_CIPHER_SUITE_TKIP) + iwl_mld_suspend_convert_tkip_ipn(key, key_rsc); + else + iwl_mld_suspend_set_ucast_pn(mld, sta, key, + key_rsc); + + data->have_rsc = true; + return; + } + /* We're iterating from old to new, there're 4 possible + * gtk ids, and only the last two keys matter + */ + if (WARN_ON(data->gtks >= + ARRAY_SIZE(data->found_gtk_idx))) + return; + + if (WARN_ON(key->keyidx >= + ARRAY_SIZE(data->rsc->mcast_key_id_map))) + return; + data->gtk_cipher = cipher; + data->found_gtk_idx[data->gtks] = key->keyidx; + key_rsc = data->rsc->mcast_rsc[data->gtks % 2]; + data->rsc->mcast_key_id_map[key->keyidx] = + data->gtks % 2; + + if (data->gtks >= 2) { + int prev = data->gtks % 2; + int prev_idx = data->found_gtk_idx[prev]; + + data->rsc->mcast_key_id_map[prev_idx] = + IWL_MCAST_KEY_MAP_INVALID; + } + + if (key->cipher == WLAN_CIPHER_SUITE_TKIP) + iwl_mld_suspend_convert_tkip_ipn(key, key_rsc); + else + iwl_mld_aes_seq_to_le64_pn(key, key_rsc); + + data->gtks++; + data->have_rsc = true; + break; + case WLAN_CIPHER_SUITE_BIP_GMAC_128: + case WLAN_CIPHER_SUITE_BIP_GMAC_256: + cipher = cpu_to_le32(STA_KEY_FLG_GCMP); + fallthrough; + case WLAN_CIPHER_SUITE_BIP_CMAC_256: + case WLAN_CIPHER_SUITE_AES_CMAC: + if (!cipher) + cipher = cpu_to_le32(STA_KEY_FLG_CCM); + if (key->keyidx == 4 || key->keyidx == 5) + data->igtk_cipher = cipher; + + if (key->keyidx == 6 || key->keyidx == 7) + data->bigtk_cipher = cipher; + + break; + } +} + +static int +iwl_mld_send_kek_kck_cmd(struct iwl_mld *mld, + struct iwl_mld_vif *mld_vif, + struct iwl_mld_suspend_key_iter_data data, + int ap_sta_id) +{ + struct iwl_wowlan_kek_kck_material_cmd_v4 kek_kck_cmd = {}; + struct iwl_mld_rekey_data *rekey_data = + &mld_vif->wowlan_data.rekey_data; + + memcpy(kek_kck_cmd.kck, rekey_data->kck, + rekey_data->kck_len); + kek_kck_cmd.kck_len = cpu_to_le16(rekey_data->kck_len); + memcpy(kek_kck_cmd.kek, rekey_data->kek, + rekey_data->kek_len); + kek_kck_cmd.kek_len = cpu_to_le16(rekey_data->kek_len); + kek_kck_cmd.replay_ctr = rekey_data->replay_ctr; + kek_kck_cmd.akm = cpu_to_le32(rekey_data->akm); + kek_kck_cmd.sta_id = cpu_to_le32(ap_sta_id); + kek_kck_cmd.gtk_cipher = data.gtk_cipher; + kek_kck_cmd.igtk_cipher = data.igtk_cipher; + kek_kck_cmd.bigtk_cipher = data.bigtk_cipher; + + IWL_DEBUG_WOWLAN(mld, "setting akm %d\n", + rekey_data->akm); + + return iwl_mld_send_cmd_pdu(mld, WOWLAN_KEK_KCK_MATERIAL, + &kek_kck_cmd); +} + +static int +iwl_mld_suspend_send_security_cmds(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct iwl_mld_vif *mld_vif, + int ap_sta_id) +{ + struct iwl_mld_suspend_key_iter_data data = {}; + int ret; + + data.rsc = kzalloc(sizeof(*data.rsc), GFP_KERNEL); + if (!data.rsc) + return -ENOMEM; + + memset(data.rsc->mcast_key_id_map, IWL_MCAST_KEY_MAP_INVALID, + ARRAY_SIZE(data.rsc->mcast_key_id_map)); + + data.rsc->sta_id = cpu_to_le32(ap_sta_id); + ieee80211_iter_keys(mld->hw, vif, + iwl_mld_suspend_key_data_iter, + &data); + + if (data.have_rsc) + ret = iwl_mld_send_cmd_pdu(mld, WOWLAN_TSC_RSC_PARAM, + data.rsc); + else + ret = 0; + + if (!ret && mld_vif->wowlan_data.rekey_data.valid) + ret = iwl_mld_send_kek_kck_cmd(mld, mld_vif, data, ap_sta_id); + + kfree(data.rsc); + + return ret; +} + +static void +iwl_mld_set_wowlan_config_cmd(struct iwl_mld *mld, + struct cfg80211_wowlan *wowlan, + struct iwl_wowlan_config_cmd *wowlan_config_cmd, + struct ieee80211_sta *ap_sta) +{ + wowlan_config_cmd->is_11n_connection = + ap_sta->deflink.ht_cap.ht_supported; + wowlan_config_cmd->flags = ENABLE_L3_FILTERING | + ENABLE_NBNS_FILTERING | ENABLE_DHCP_FILTERING; + + if (ap_sta->mfp) + wowlan_config_cmd->flags |= IS_11W_ASSOC; + + if (wowlan->disconnect) + wowlan_config_cmd->wakeup_filter |= + cpu_to_le32(IWL_WOWLAN_WAKEUP_BEACON_MISS | + IWL_WOWLAN_WAKEUP_LINK_CHANGE); + if (wowlan->magic_pkt) + wowlan_config_cmd->wakeup_filter |= + cpu_to_le32(IWL_WOWLAN_WAKEUP_MAGIC_PACKET); + if (wowlan->gtk_rekey_failure) + wowlan_config_cmd->wakeup_filter |= + cpu_to_le32(IWL_WOWLAN_WAKEUP_GTK_REKEY_FAIL); + if (wowlan->eap_identity_req) + wowlan_config_cmd->wakeup_filter |= + cpu_to_le32(IWL_WOWLAN_WAKEUP_EAP_IDENT_REQ); + if (wowlan->four_way_handshake) + wowlan_config_cmd->wakeup_filter |= + cpu_to_le32(IWL_WOWLAN_WAKEUP_4WAY_HANDSHAKE); + if (wowlan->n_patterns) + wowlan_config_cmd->wakeup_filter |= + cpu_to_le32(IWL_WOWLAN_WAKEUP_PATTERN_MATCH); + + if (wowlan->rfkill_release) + wowlan_config_cmd->wakeup_filter |= + cpu_to_le32(IWL_WOWLAN_WAKEUP_RF_KILL_DEASSERT); + + if (wowlan->any) { + wowlan_config_cmd->wakeup_filter |= + cpu_to_le32(IWL_WOWLAN_WAKEUP_BEACON_MISS | + IWL_WOWLAN_WAKEUP_LINK_CHANGE | + IWL_WOWLAN_WAKEUP_RX_FRAME | + IWL_WOWLAN_WAKEUP_BCN_FILTERING); + } +} + +static int iwl_mld_send_patterns(struct iwl_mld *mld, + struct cfg80211_wowlan *wowlan, + int ap_sta_id) +{ + struct iwl_wowlan_patterns_cmd *pattern_cmd; + struct iwl_host_cmd cmd = { + .id = WOWLAN_PATTERNS, + .dataflags[0] = IWL_HCMD_DFL_NOCOPY, + }; + int ret; + + if (!wowlan->n_patterns) + return 0; + + cmd.len[0] = struct_size(pattern_cmd, patterns, wowlan->n_patterns); + + pattern_cmd = kzalloc(cmd.len[0], GFP_KERNEL); + if (!pattern_cmd) + return -ENOMEM; + + pattern_cmd->n_patterns = wowlan->n_patterns; + pattern_cmd->sta_id = ap_sta_id; + + for (int i = 0; i < wowlan->n_patterns; i++) { + int mask_len = DIV_ROUND_UP(wowlan->patterns[i].pattern_len, 8); + + pattern_cmd->patterns[i].pattern_type = + WOWLAN_PATTERN_TYPE_BITMASK; + + memcpy(&pattern_cmd->patterns[i].u.bitmask.mask, + wowlan->patterns[i].mask, mask_len); + memcpy(&pattern_cmd->patterns[i].u.bitmask.pattern, + wowlan->patterns[i].pattern, + wowlan->patterns[i].pattern_len); + pattern_cmd->patterns[i].u.bitmask.mask_size = mask_len; + pattern_cmd->patterns[i].u.bitmask.pattern_size = + wowlan->patterns[i].pattern_len; + } + + cmd.data[0] = pattern_cmd; + ret = iwl_mld_send_cmd(mld, &cmd); + kfree(pattern_cmd); + return ret; +} + +static int +iwl_mld_send_proto_offload(struct iwl_mld *mld, + struct ieee80211_vif *vif, + u8 ap_sta_id) +{ + struct iwl_proto_offload_cmd_v4 *cmd __free(kfree); + struct iwl_host_cmd hcmd = { + .id = PROT_OFFLOAD_CONFIG_CMD, + .dataflags[0] = IWL_HCMD_DFL_NOCOPY, + .len[0] = sizeof(*cmd), + }; + u32 enabled = 0; + + cmd = kzalloc(hcmd.len[0], GFP_KERNEL); + +#if IS_ENABLED(CONFIG_IPV6) + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mld_wowlan_data *wowlan_data = &mld_vif->wowlan_data; + struct iwl_ns_config *nsc; + struct iwl_targ_addr *addrs; + int n_nsc, n_addrs; + int i, c; + int num_skipped = 0; + + nsc = cmd->ns_config; + n_nsc = IWL_PROTO_OFFLOAD_NUM_NS_CONFIG_V3L; + addrs = cmd->targ_addrs; + n_addrs = IWL_PROTO_OFFLOAD_NUM_IPV6_ADDRS_V3L; + + /* For each address we have (and that will fit) fill a target + * address struct and combine for NS offload structs with the + * solicited node addresses. + */ + for (i = 0, c = 0; + i < wowlan_data->num_target_ipv6_addrs && + i < n_addrs && c < n_nsc; i++) { + int j; + struct in6_addr solicited_addr; + + /* Because ns is offloaded skip tentative address to avoid + * violating RFC4862. + */ + if (test_bit(i, wowlan_data->tentative_addrs)) { + num_skipped++; + continue; + } + + addrconf_addr_solict_mult(&wowlan_data->target_ipv6_addrs[i], + &solicited_addr); + for (j = 0; j < c; j++) + if (ipv6_addr_cmp(&nsc[j].dest_ipv6_addr, + &solicited_addr) == 0) + break; + if (j == c) + c++; + addrs[i].addr = wowlan_data->target_ipv6_addrs[i]; + addrs[i].config_num = cpu_to_le32(j); + nsc[j].dest_ipv6_addr = solicited_addr; + memcpy(nsc[j].target_mac_addr, vif->addr, ETH_ALEN); + } + + if (wowlan_data->num_target_ipv6_addrs - num_skipped) + enabled |= IWL_D3_PROTO_IPV6_VALID; + + cmd->num_valid_ipv6_addrs = cpu_to_le32(i - num_skipped); + if (enabled & IWL_D3_PROTO_IPV6_VALID) + enabled |= IWL_D3_PROTO_OFFLOAD_NS; +#endif + + if (vif->cfg.arp_addr_cnt) { + enabled |= IWL_D3_PROTO_OFFLOAD_ARP | IWL_D3_PROTO_IPV4_VALID; + cmd->common.host_ipv4_addr = vif->cfg.arp_addr_list[0]; + ether_addr_copy(cmd->common.arp_mac_addr, vif->addr); + } + + enabled |= IWL_D3_PROTO_OFFLOAD_BTM; + cmd->common.enabled = cpu_to_le32(enabled); + cmd->sta_id = cpu_to_le32(ap_sta_id); + hcmd.data[0] = cmd; + return iwl_mld_send_cmd(mld, &hcmd); +} + +static int +iwl_mld_wowlan_config(struct iwl_mld *mld, struct ieee80211_vif *bss_vif, + struct cfg80211_wowlan *wowlan) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(bss_vif); + struct ieee80211_sta *ap_sta = mld_vif->ap_sta; + struct iwl_wowlan_config_cmd wowlan_config_cmd = { + .offloading_tid = IWL_WOWLAN_OFFLOAD_TID, + }; + u32 sta_id_mask; + int ap_sta_id, ret; + int link_id = iwl_mld_get_primary_link(bss_vif); + struct ieee80211_bss_conf *link_conf; + + ret = iwl_mld_block_emlsr_sync(mld, bss_vif, + IWL_MLD_EMLSR_BLOCKED_WOWLAN, link_id); + if (ret) + return ret; + + link_conf = link_conf_dereference_protected(bss_vif, link_id); + + if (WARN_ON(!ap_sta || !link_conf)) + return -EINVAL; + + sta_id_mask = iwl_mld_fw_sta_id_mask(mld, ap_sta); + if (WARN_ON(hweight32(sta_id_mask) != 1)) + return -EINVAL; + + ap_sta_id = __ffs(sta_id_mask); + wowlan_config_cmd.sta_id = ap_sta_id; + + ret = iwl_mld_ensure_queue(mld, + ap_sta->txq[wowlan_config_cmd.offloading_tid]); + if (ret) + return ret; + + iwl_mld_set_wowlan_config_cmd(mld, wowlan, + &wowlan_config_cmd, ap_sta); + ret = iwl_mld_send_cmd_pdu(mld, WOWLAN_CONFIGURATION, + &wowlan_config_cmd); + if (ret) + return ret; + + ret = iwl_mld_suspend_send_security_cmds(mld, bss_vif, mld_vif, + ap_sta_id); + if (ret) + return ret; + + ret = iwl_mld_send_patterns(mld, wowlan, ap_sta_id); + if (ret) + return ret; + + ret = iwl_mld_send_proto_offload(mld, bss_vif, ap_sta_id); + if (ret) + return ret; + + iwl_mld_enable_beacon_filter(mld, link_conf, true); + return iwl_mld_update_mac_power(mld, bss_vif, true); +} + +int iwl_mld_wowlan_suspend(struct iwl_mld *mld, struct cfg80211_wowlan *wowlan) +{ + struct ieee80211_vif *bss_vif; + + lockdep_assert_wiphy(mld->wiphy); + + if (WARN_ON(!wowlan)) + return 1; + + IWL_DEBUG_WOWLAN(mld, "Starting the wowlan suspend flow\n"); + + bss_vif = iwl_mld_get_bss_vif(mld); + if (WARN_ON(!bss_vif)) + return 1; + + if (!bss_vif->cfg.assoc) { + int ret; + /* If we're not associated, this must be netdetect */ + if (WARN_ON(!wowlan->nd_config)) + return 1; + + ret = iwl_mld_netdetect_config(mld, bss_vif, wowlan); + if (!ret) + mld->netdetect = true; + + return ret; + } + + return iwl_mld_wowlan_config(mld, bss_vif, wowlan); +} + +/* Returns 0 on success, 1 if an error occurred in firmware during d3, + * A negative value is expected only in unrecovreable cases. + */ +int iwl_mld_wowlan_resume(struct iwl_mld *mld) +{ + struct ieee80211_vif *bss_vif; + struct ieee80211_bss_conf *link_conf; + struct iwl_mld_netdetect_res netdetect_res; + struct iwl_mld_resume_data resume_data = { + .notifs_expected = + IWL_D3_NOTIF_WOWLAN_INFO | + IWL_D3_NOTIF_D3_END_NOTIF, + .netdetect_res = &netdetect_res, + }; + int link_id; + int ret; + bool fw_err = false; + bool keep_connection; + + lockdep_assert_wiphy(mld->wiphy); + + IWL_DEBUG_WOWLAN(mld, "Starting the wowlan resume flow\n"); + + mld->trans->system_pm_mode = IWL_PLAT_PM_MODE_DISABLED; + if (!mld->fw_status.in_d3) { + IWL_DEBUG_WOWLAN(mld, + "Device_powered_off() was called during wowlan\n"); + goto err; + } + + mld->fw_status.in_d3 = false; + mld->scan.last_start_time_jiffies = jiffies; + + bss_vif = iwl_mld_get_bss_vif(mld); + if (WARN_ON(!bss_vif)) + goto err; + + /* We can't have several links upon wowlan entry, + * this is enforced in the suspend flow. + */ + WARN_ON(hweight16(bss_vif->active_links) > 1); + link_id = bss_vif->active_links ? __ffs(bss_vif->active_links) : 0; + link_conf = link_conf_dereference_protected(bss_vif, link_id); + + if (WARN_ON(!link_conf)) + goto err; + + iwl_fw_dbg_read_d3_debug_data(&mld->fwrt); + + if (iwl_mld_fw_needs_restart(mld, bss_vif)) { + fw_err = true; + goto err; + } + + resume_data.wowlan_status = kzalloc(sizeof(*resume_data.wowlan_status), + GFP_KERNEL); + if (!resume_data.wowlan_status) + return -1; + + if (mld->netdetect) + resume_data.notifs_expected |= IWL_D3_ND_MATCH_INFO; + + ret = iwl_mld_wait_d3_notif(mld, &resume_data, true); + if (ret) { + IWL_ERR(mld, "Couldn't get the d3 notifs %d\n", ret); + fw_err = true; + goto err; + } + + if (resume_data.d3_end_flags & IWL_D0I3_RESET_REQUIRE) { + mld->fw_status.in_hw_restart = true; + goto process_wakeup_results; + } + + iwl_mld_update_changed_regdomain(mld); + iwl_mld_update_mac_power(mld, bss_vif, false); + iwl_mld_enable_beacon_filter(mld, link_conf, false); + iwl_mld_update_device_power(mld, false); + + if (mld->netdetect) + ret = iwl_mld_scan_stop(mld, IWL_MLD_SCAN_NETDETECT, false); + + process_wakeup_results: + if (mld->netdetect) { + iwl_mld_process_netdetect_res(mld, bss_vif, &resume_data); + mld->netdetect = false; + } else { + keep_connection = + iwl_mld_process_wowlan_status(mld, bss_vif, + resume_data.wowlan_status); + + /* EMLSR state will be cleared if the connection is not kept */ + if (keep_connection) + iwl_mld_unblock_emlsr(mld, bss_vif, + IWL_MLD_EMLSR_BLOCKED_WOWLAN); + } + + if (!mld->netdetect && !keep_connection) + ieee80211_resume_disconnect(bss_vif); + + goto out; + + err: + if (fw_err) { + mld->trans->state = IWL_TRANS_NO_FW; + set_bit(STATUS_FW_ERROR, &mld->trans->status); + } + + mld->fw_status.in_hw_restart = true; + ret = 1; + out: + if (resume_data.wowlan_status) { + kfree(resume_data.wowlan_status->wake_packet); + kfree(resume_data.wowlan_status); + } + + return ret; +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/d3.h b/drivers/net/wireless/intel/iwlwifi/mld/d3.h new file mode 100644 index 0000000000000..618d6fb3c7964 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/d3.h @@ -0,0 +1,51 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024 Intel Corporation + */ +#ifndef __iwl_mld_d3_h__ +#define __iwl_mld_d3_h__ + +#include "fw/api/d3.h" + +struct iwl_mld_rekey_data { + bool valid; + u8 kck[NL80211_KCK_EXT_LEN]; + u8 kek[NL80211_KEK_EXT_LEN]; + size_t kck_len; + size_t kek_len; + __le64 replay_ctr; + u32 akm; +}; + +/** + * struct iwl_mld_wowlan_data - data used by the wowlan suspend flow + * + * @target_ipv6_addrs: IPv6 addresses on this interface for offload + * @tentative_addrs: bitmap of tentative IPv6 addresses in @target_ipv6_addrs + * @num_target_ipv6_addrs: number of @target_ipv6_addrs + * @rekey_data: security key data used for rekeying during D3 + */ +struct iwl_mld_wowlan_data { +#if IS_ENABLED(CONFIG_IPV6) + struct in6_addr target_ipv6_addrs[IWL_PROTO_OFFLOAD_NUM_IPV6_ADDRS_MAX]; + unsigned long tentative_addrs[BITS_TO_LONGS(IWL_PROTO_OFFLOAD_NUM_IPV6_ADDRS_MAX)]; + int num_target_ipv6_addrs; +#endif + struct iwl_mld_rekey_data rekey_data; +}; + +int iwl_mld_no_wowlan_resume(struct iwl_mld *mld); +int iwl_mld_no_wowlan_suspend(struct iwl_mld *mld); +int iwl_mld_wowlan_suspend(struct iwl_mld *mld, + struct cfg80211_wowlan *wowlan); +int iwl_mld_wowlan_resume(struct iwl_mld *mld); +void iwl_mld_set_rekey_data(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct cfg80211_gtk_rekey_data *data); +#if IS_ENABLED(CONFIG_IPV6) +void iwl_mld_ipv6_addr_change(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct inet6_dev *idev); +#endif + +#endif /* __iwl_mld_d3_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/debugfs.c b/drivers/net/wireless/intel/iwlwifi/mld/debugfs.c new file mode 100644 index 0000000000000..c759c5c68dc04 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/debugfs.c @@ -0,0 +1,823 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024-2025 Intel Corporation + */ + +#include "mld.h" +#include "debugfs.h" +#include "iwl-io.h" +#include "hcmd.h" +#include "iface.h" +#include "sta.h" +#include "tlc.h" +#include "power.h" +#include "notif.h" +#include "ap.h" +#include "iwl-utils.h" +#ifdef CONFIG_THERMAL +#include "thermal.h" +#endif + +#include "fw/api/rs.h" +#include "fw/api/dhc.h" +#include "fw/api/rfi.h" + +#define MLD_DEBUGFS_READ_FILE_OPS(name, bufsz) \ + _MLD_DEBUGFS_READ_FILE_OPS(name, bufsz, struct iwl_mld) + +#define MLD_DEBUGFS_ADD_FILE_ALIAS(alias, name, parent, mode) \ + debugfs_create_file(alias, mode, parent, mld, \ + &iwl_dbgfs_##name##_ops) +#define MLD_DEBUGFS_ADD_FILE(name, parent, mode) \ + MLD_DEBUGFS_ADD_FILE_ALIAS(#name, name, parent, mode) + +static bool iwl_mld_dbgfs_fw_cmd_disabled(struct iwl_mld *mld) +{ +#ifdef CONFIG_PM_SLEEP + return !mld->fw_status.running || mld->fw_status.in_d3; +#else + return !mld->fw_status.running; +#endif /* CONFIG_PM_SLEEP */ +} + +static ssize_t iwl_dbgfs_fw_dbg_clear_write(struct iwl_mld *mld, + char *buf, size_t count) +{ + /* If the firmware is not running, silently succeed since there is + * no data to clear. + */ + if (iwl_mld_dbgfs_fw_cmd_disabled(mld)) + return 0; + + iwl_fw_dbg_clear_monitor_buf(&mld->fwrt); + + return count; +} + +static ssize_t iwl_dbgfs_fw_nmi_write(struct iwl_mld *mld, char *buf, + size_t count) +{ + if (iwl_mld_dbgfs_fw_cmd_disabled(mld)) + return -EIO; + + IWL_ERR(mld, "Triggering an NMI from debugfs\n"); + + if (count == 6 && !strcmp(buf, "nolog\n")) + mld->fw_status.do_not_dump_once = true; + + iwl_force_nmi(mld->trans); + + return count; +} + +static ssize_t iwl_dbgfs_fw_restart_write(struct iwl_mld *mld, char *buf, + size_t count) +{ + int __maybe_unused ret; + + if (!iwlwifi_mod_params.fw_restart) + return -EPERM; + + if (iwl_mld_dbgfs_fw_cmd_disabled(mld)) + return -EIO; + + if (count == 6 && !strcmp(buf, "nolog\n")) { + mld->fw_status.do_not_dump_once = true; + set_bit(STATUS_SUPPRESS_CMD_ERROR_ONCE, &mld->trans->status); + } + + /* take the return value to make compiler happy - it will + * fail anyway + */ + ret = iwl_mld_send_cmd_empty(mld, WIDE_ID(LONG_GROUP, REPLY_ERROR)); + + return count; +} + +static ssize_t iwl_dbgfs_send_echo_cmd_write(struct iwl_mld *mld, char *buf, + size_t count) +{ + if (iwl_mld_dbgfs_fw_cmd_disabled(mld)) + return -EIO; + + return iwl_mld_send_cmd_empty(mld, ECHO_CMD) ?: count; +} + +struct iwl_mld_sniffer_apply { + struct iwl_mld *mld; + const u8 *bssid; + u16 aid; +}; + +static bool iwl_mld_sniffer_apply(struct iwl_notif_wait_data *notif_data, + struct iwl_rx_packet *pkt, void *data) +{ + struct iwl_mld_sniffer_apply *apply = data; + + apply->mld->monitor.cur_aid = cpu_to_le16(apply->aid); + memcpy(apply->mld->monitor.cur_bssid, apply->bssid, + sizeof(apply->mld->monitor.cur_bssid)); + + return true; +} + +static ssize_t +iwl_dbgfs_he_sniffer_params_write(struct iwl_mld *mld, char *buf, + size_t count) +{ + struct iwl_notification_wait wait; + struct iwl_he_monitor_cmd he_mon_cmd = {}; + struct iwl_mld_sniffer_apply apply = { + .mld = mld, + }; + u16 wait_cmds[] = { + WIDE_ID(DATA_PATH_GROUP, HE_AIR_SNIFFER_CONFIG_CMD), + }; + u32 aid; + int ret; + + if (iwl_mld_dbgfs_fw_cmd_disabled(mld)) + return -EIO; + + if (!mld->monitor.on) + return -ENODEV; + + ret = sscanf(buf, "%x %2hhx:%2hhx:%2hhx:%2hhx:%2hhx:%2hhx", &aid, + &he_mon_cmd.bssid[0], &he_mon_cmd.bssid[1], + &he_mon_cmd.bssid[2], &he_mon_cmd.bssid[3], + &he_mon_cmd.bssid[4], &he_mon_cmd.bssid[5]); + if (ret != 7) + return -EINVAL; + + he_mon_cmd.aid = cpu_to_le16(aid); + + apply.aid = aid; + apply.bssid = (void *)he_mon_cmd.bssid; + + /* Use the notification waiter to get our function triggered + * in sequence with other RX. This ensures that frames we get + * on the RX queue _before_ the new configuration is applied + * still have mld->cur_aid pointing to the old AID, and that + * frames on the RX queue _after_ the firmware processed the + * new configuration (and sent the response, synchronously) + * get mld->cur_aid correctly set to the new AID. + */ + iwl_init_notification_wait(&mld->notif_wait, &wait, + wait_cmds, ARRAY_SIZE(wait_cmds), + iwl_mld_sniffer_apply, &apply); + + ret = iwl_mld_send_cmd_pdu(mld, + WIDE_ID(DATA_PATH_GROUP, + HE_AIR_SNIFFER_CONFIG_CMD), + &he_mon_cmd); + + /* no need to really wait, we already did anyway */ + iwl_remove_notification(&mld->notif_wait, &wait); + + return ret ?: count; +} + +static ssize_t +iwl_dbgfs_he_sniffer_params_read(struct iwl_mld *mld, size_t count, + char *buf) +{ + return scnprintf(buf, count, + "%d %02hhx:%02hhx:%02hhx:%02hhx:%02hhx:%02hhx\n", + le16_to_cpu(mld->monitor.cur_aid), + mld->monitor.cur_bssid[0], mld->monitor.cur_bssid[1], + mld->monitor.cur_bssid[2], mld->monitor.cur_bssid[3], + mld->monitor.cur_bssid[4], mld->monitor.cur_bssid[5]); +} + +/* The size computation is as follows: + * each number needs at most 3 characters, number of rows is the size of + * the table; So, need 5 chars for the "freq: " part and each tuple afterwards + * needs 6 characters for numbers and 5 for the punctuation around. 32 bytes + * for feature support message. + */ +#define IWL_RFI_DDR_BUF_SIZE (IWL_RFI_DDR_LUT_INSTALLED_SIZE *\ + (5 + IWL_RFI_DDR_LUT_ENTRY_CHANNELS_NUM *\ + (6 + 5)) + 32) +#define IWL_RFI_DLVR_BUF_SIZE (IWL_RFI_DLVR_LUT_INSTALLED_SIZE *\ + (5 + IWL_RFI_DLVR_LUT_ENTRY_CHANNELS_NUM *\ + (6 + 5)) + 32) +#define IWL_RFI_DESENSE_BUF_SIZE IWL_RFI_DDR_BUF_SIZE + +/* Extra 32 for "DDR and DLVR table" message */ +#define IWL_RFI_BUF_SIZE (IWL_RFI_DDR_BUF_SIZE + IWL_RFI_DLVR_BUF_SIZE +\ + IWL_RFI_DESENSE_BUF_SIZE + 32) + +WIPHY_DEBUGFS_WRITE_FILE_OPS_MLD(fw_nmi, 10); +WIPHY_DEBUGFS_WRITE_FILE_OPS_MLD(fw_restart, 10); +WIPHY_DEBUGFS_READ_WRITE_FILE_OPS_MLD(he_sniffer_params, 32); +WIPHY_DEBUGFS_WRITE_FILE_OPS_MLD(fw_dbg_clear, 10); +WIPHY_DEBUGFS_WRITE_FILE_OPS_MLD(send_echo_cmd, 8); + +static ssize_t iwl_dbgfs_wifi_6e_enable_read(struct iwl_mld *mld, + size_t count, u8 *buf) +{ + int err; + u32 value; + + err = iwl_bios_get_dsm(&mld->fwrt, DSM_FUNC_ENABLE_6E, &value); + if (err) + return err; + + return scnprintf(buf, count, "0x%08x\n", value); +} + +MLD_DEBUGFS_READ_FILE_OPS(wifi_6e_enable, 64); + +static ssize_t iwl_dbgfs_inject_packet_write(struct iwl_mld *mld, + char *buf, size_t count) +{ + struct iwl_op_mode *opmode = container_of((void *)mld, + struct iwl_op_mode, + op_mode_specific); + struct iwl_rx_cmd_buffer rxb = {}; + struct iwl_rx_packet *pkt; + int n_bytes = count / 2; + int ret = -EINVAL; + + if (iwl_mld_dbgfs_fw_cmd_disabled(mld)) + return -EIO; + + rxb._page = alloc_pages(GFP_KERNEL, 0); + if (!rxb._page) + return -ENOMEM; + pkt = rxb_addr(&rxb); + + ret = hex2bin(page_address(rxb._page), buf, n_bytes); + if (ret) + goto out; + + /* avoid invalid memory access and malformed packet */ + if (n_bytes < sizeof(*pkt) || + n_bytes != sizeof(*pkt) + iwl_rx_packet_payload_len(pkt)) + goto out; + + local_bh_disable(); + iwl_mld_rx(opmode, NULL, &rxb); + local_bh_enable(); + ret = 0; + +out: + iwl_free_rxb(&rxb); + + return ret ?: count; +} + +WIPHY_DEBUGFS_WRITE_FILE_OPS_MLD(inject_packet, 512); + +#ifdef CONFIG_THERMAL + +static ssize_t iwl_dbgfs_stop_ctdp_write(struct iwl_mld *mld, + char *buf, size_t count) +{ + if (iwl_mld_dbgfs_fw_cmd_disabled(mld)) + return -EIO; + + return iwl_mld_config_ctdp(mld, mld->cooling_dev.cur_state, + CTDP_CMD_OPERATION_STOP) ? : count; +} + +WIPHY_DEBUGFS_WRITE_FILE_OPS_MLD(stop_ctdp, 8); + +static ssize_t iwl_dbgfs_start_ctdp_write(struct iwl_mld *mld, + char *buf, size_t count) +{ + if (iwl_mld_dbgfs_fw_cmd_disabled(mld)) + return -EIO; + + return iwl_mld_config_ctdp(mld, mld->cooling_dev.cur_state, + CTDP_CMD_OPERATION_START) ? : count; +} + +WIPHY_DEBUGFS_WRITE_FILE_OPS_MLD(start_ctdp, 8); + +#endif /* CONFIG_THERMAL */ + +void +iwl_mld_add_debugfs_files(struct iwl_mld *mld, struct dentry *debugfs_dir) +{ + /* Add debugfs files here */ + + MLD_DEBUGFS_ADD_FILE(fw_nmi, debugfs_dir, 0200); + MLD_DEBUGFS_ADD_FILE(fw_restart, debugfs_dir, 0200); + MLD_DEBUGFS_ADD_FILE(wifi_6e_enable, debugfs_dir, 0400); + MLD_DEBUGFS_ADD_FILE(he_sniffer_params, debugfs_dir, 0600); + MLD_DEBUGFS_ADD_FILE(fw_dbg_clear, debugfs_dir, 0200); + MLD_DEBUGFS_ADD_FILE(send_echo_cmd, debugfs_dir, 0200); +#ifdef CONFIG_THERMAL + MLD_DEBUGFS_ADD_FILE(start_ctdp, debugfs_dir, 0200); + MLD_DEBUGFS_ADD_FILE(stop_ctdp, debugfs_dir, 0200); +#endif + MLD_DEBUGFS_ADD_FILE(inject_packet, debugfs_dir, 0200); + + /* Create a symlink with mac80211. It will be removed when mac80211 + * exits (before the opmode exits which removes the target.) + */ + if (!IS_ERR(debugfs_dir)) { + char buf[100]; + + snprintf(buf, 100, "../../%pd2", debugfs_dir->d_parent); + debugfs_create_symlink("iwlwifi", mld->wiphy->debugfsdir, + buf); + } +} + +#define VIF_DEBUGFS_WRITE_FILE_OPS(name, bufsz) \ + WIPHY_DEBUGFS_WRITE_FILE_OPS(vif_##name, bufsz, vif) + +#define VIF_DEBUGFS_READ_WRITE_FILE_OPS(name, bufsz) \ + IEEE80211_WIPHY_DEBUGFS_READ_WRITE_FILE_OPS(vif_##name, bufsz, vif) \ + +#define VIF_DEBUGFS_ADD_FILE_ALIAS(alias, name, parent, mode) \ + debugfs_create_file(alias, mode, parent, vif, \ + &iwl_dbgfs_vif_##name##_ops) +#define VIF_DEBUGFS_ADD_FILE(name, parent, mode) \ + VIF_DEBUGFS_ADD_FILE_ALIAS(#name, name, parent, mode) + +static ssize_t iwl_dbgfs_vif_bf_params_write(struct iwl_mld *mld, char *buf, + size_t count, void *data) +{ + struct ieee80211_vif *vif = data; + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + int link_id = vif->active_links ? __ffs(vif->active_links) : 0; + struct ieee80211_bss_conf *link_conf; + int val; + + if (!strncmp("bf_enable_beacon_filter=", buf, 24)) { + if (sscanf(buf + 24, "%d", &val) != 1) + return -EINVAL; + } else { + return -EINVAL; + } + + if (val != 0 && val != 1) + return -EINVAL; + + link_conf = link_conf_dereference_protected(vif, link_id); + if (WARN_ON(!link_conf)) + return -ENODEV; + + if (iwl_mld_dbgfs_fw_cmd_disabled(mld)) + return -EIO; + + mld_vif->disable_bf = !val; + + if (val) + return iwl_mld_enable_beacon_filter(mld, link_conf, + false) ?: count; + else + return iwl_mld_disable_beacon_filter(mld, vif) ?: count; +} + +static ssize_t iwl_dbgfs_vif_pm_params_write(struct iwl_mld *mld, + char *buf, + size_t count, void *data) +{ + struct ieee80211_vif *vif = data; + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + int val; + + if (!strncmp("use_ps_poll=", buf, 12)) { + if (sscanf(buf + 12, "%d", &val) != 1) + return -EINVAL; + } else { + return -EINVAL; + } + + if (iwl_mld_dbgfs_fw_cmd_disabled(mld)) + return -EIO; + + mld_vif->use_ps_poll = val; + + return iwl_mld_update_mac_power(mld, vif, false) ?: count; +} + +static ssize_t iwl_dbgfs_vif_low_latency_write(struct iwl_mld *mld, + char *buf, size_t count, + void *data) +{ + struct ieee80211_vif *vif = data; + u8 value; + int ret; + + ret = kstrtou8(buf, 0, &value); + if (ret) + return ret; + + if (value > 1) + return -EINVAL; + + iwl_mld_vif_update_low_latency(mld, vif, value, LOW_LATENCY_DEBUGFS); + + return count; +} + +static ssize_t iwl_dbgfs_vif_low_latency_read(struct ieee80211_vif *vif, + size_t count, char *buf) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + char format[] = "traffic=%d\ndbgfs=%d\nvif_type=%d\nactual=%d\n"; + u8 ll_causes; + + if (WARN_ON(count < sizeof(format))) + return -EINVAL; + + ll_causes = READ_ONCE(mld_vif->low_latency_causes); + + /* all values in format are boolean so the size of format is enough + * for holding the result string + */ + return scnprintf(buf, count, format, + !!(ll_causes & LOW_LATENCY_TRAFFIC), + !!(ll_causes & LOW_LATENCY_DEBUGFS), + !!(ll_causes & LOW_LATENCY_VIF_TYPE), + !!(ll_causes)); +} + +VIF_DEBUGFS_WRITE_FILE_OPS(pm_params, 32); +VIF_DEBUGFS_WRITE_FILE_OPS(bf_params, 32); +VIF_DEBUGFS_READ_WRITE_FILE_OPS(low_latency, 45); + +static int +_iwl_dbgfs_inject_beacon_ie(struct iwl_mld *mld, struct ieee80211_vif *vif, + char *bin, ssize_t len, + bool restore) +{ + struct iwl_mld_vif *mld_vif; + struct iwl_mld_link *mld_link; + struct iwl_mac_beacon_cmd beacon_cmd = {}; + int n_bytes = len / 2; + + /* Element len should be represented by u8 */ + if (n_bytes >= U8_MAX) + return -EINVAL; + + if (iwl_mld_dbgfs_fw_cmd_disabled(mld)) + return -EIO; + + if (!vif) + return -EINVAL; + + mld_vif = iwl_mld_vif_from_mac80211(vif); + mld_vif->beacon_inject_active = true; + mld->hw->extra_beacon_tailroom = n_bytes; + + for_each_mld_vif_valid_link(mld_vif, mld_link) { + u32 offset; + struct ieee80211_tx_info *info; + struct ieee80211_bss_conf *link_conf = + link_conf_dereference_protected(vif, link_id); + struct ieee80211_chanctx_conf *ctx = + wiphy_dereference(mld->wiphy, link_conf->chanctx_conf); + struct sk_buff *beacon = + ieee80211_beacon_get_template(mld->hw, vif, + NULL, link_id); + + if (!beacon) + return -EINVAL; + + if (!restore && (WARN_ON(!n_bytes || !bin) || + hex2bin(skb_put_zero(beacon, n_bytes), + bin, n_bytes))) { + dev_kfree_skb(beacon); + return -EINVAL; + } + + info = IEEE80211_SKB_CB(beacon); + + beacon_cmd.flags = + cpu_to_le16(iwl_mld_get_rate_flags(mld, info, vif, + link_conf, + ctx->def.chan->band)); + beacon_cmd.byte_cnt = cpu_to_le16((u16)beacon->len); + beacon_cmd.link_id = + cpu_to_le32(mld_link->fw_id); + + iwl_mld_set_tim_idx(mld, &beacon_cmd.tim_idx, + beacon->data, beacon->len); + + offset = iwl_find_ie_offset(beacon->data, + WLAN_EID_S1G_TWT, + beacon->len); + + beacon_cmd.btwt_offset = cpu_to_le32(offset); + + iwl_mld_send_beacon_template_cmd(mld, beacon, &beacon_cmd); + dev_kfree_skb(beacon); + } + + if (restore) + mld_vif->beacon_inject_active = false; + + return 0; +} + +static ssize_t +iwl_dbgfs_vif_inject_beacon_ie_write(struct iwl_mld *mld, + char *buf, size_t count, + void *data) +{ + struct ieee80211_vif *vif = data; + int ret = _iwl_dbgfs_inject_beacon_ie(mld, vif, buf, + count, false); + + mld->hw->extra_beacon_tailroom = 0; + return ret ?: count; +} + +VIF_DEBUGFS_WRITE_FILE_OPS(inject_beacon_ie, 512); + +static ssize_t +iwl_dbgfs_vif_inject_beacon_ie_restore_write(struct iwl_mld *mld, + char *buf, + size_t count, + void *data) +{ + struct ieee80211_vif *vif = data; + int ret = _iwl_dbgfs_inject_beacon_ie(mld, vif, NULL, + 0, true); + + mld->hw->extra_beacon_tailroom = 0; + return ret ?: count; +} + +VIF_DEBUGFS_WRITE_FILE_OPS(inject_beacon_ie_restore, 512); + +static ssize_t +iwl_dbgfs_vif_twt_setup_write(struct iwl_mld *mld, char *buf, size_t count, + void *data) +{ + struct iwl_host_cmd hcmd = { + .id = WIDE_ID(IWL_ALWAYS_LONG_GROUP, DEBUG_HOST_COMMAND), + }; + struct ieee80211_vif *vif = data; + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_dhc_twt_operation *dhc_twt_cmd; + struct iwl_dhc_cmd *cmd __free(kfree); + u64 target_wake_time; + u32 twt_operation, interval_exp, interval_mantissa, min_wake_duration; + u8 trigger, flow_type, flow_id, protection, tenth_param; + u8 twt_request = 1, broadcast = 0; + int ret; + + if (iwl_mld_dbgfs_fw_cmd_disabled(mld)) + return -EIO; + + ret = sscanf(buf, "%u %llu %u %u %u %hhu %hhu %hhu %hhu %hhu", + &twt_operation, &target_wake_time, &interval_exp, + &interval_mantissa, &min_wake_duration, &trigger, + &flow_type, &flow_id, &protection, &tenth_param); + + /* the new twt_request parameter is optional for station */ + if ((ret != 9 && ret != 10) || + (ret == 10 && vif->type != NL80211_IFTYPE_STATION && + tenth_param == 1)) + return -EINVAL; + + /* The 10th parameter: + * In STA mode - the TWT type (broadcast or individual) + * In AP mode - the role (0 responder, 2 unsolicited) + */ + if (ret == 10) { + if (vif->type == NL80211_IFTYPE_STATION) + broadcast = tenth_param; + else + twt_request = tenth_param; + } + + cmd = kzalloc(sizeof(*cmd) + sizeof(*dhc_twt_cmd), GFP_KERNEL); + if (!cmd) + return -ENOMEM; + + dhc_twt_cmd = (void *)cmd->data; + dhc_twt_cmd->mac_id = cpu_to_le32(mld_vif->fw_id); + dhc_twt_cmd->twt_operation = cpu_to_le32(twt_operation); + dhc_twt_cmd->target_wake_time = cpu_to_le64(target_wake_time); + dhc_twt_cmd->interval_exp = cpu_to_le32(interval_exp); + dhc_twt_cmd->interval_mantissa = cpu_to_le32(interval_mantissa); + dhc_twt_cmd->min_wake_duration = cpu_to_le32(min_wake_duration); + dhc_twt_cmd->trigger = trigger; + dhc_twt_cmd->flow_type = flow_type; + dhc_twt_cmd->flow_id = flow_id; + dhc_twt_cmd->protection = protection; + dhc_twt_cmd->twt_request = twt_request; + dhc_twt_cmd->negotiation_type = broadcast ? 3 : 0; + + cmd->length = cpu_to_le32(sizeof(*dhc_twt_cmd) >> 2); + cmd->index_and_mask = + cpu_to_le32(DHC_TABLE_INTEGRATION | DHC_TARGET_UMAC | + DHC_INT_UMAC_TWT_OPERATION); + + hcmd.len[0] = sizeof(*cmd) + sizeof(*dhc_twt_cmd); + hcmd.data[0] = cmd; + + ret = iwl_mld_send_cmd(mld, &hcmd); + + return ret ?: count; +} + +VIF_DEBUGFS_WRITE_FILE_OPS(twt_setup, 256); + +static ssize_t +iwl_dbgfs_vif_twt_operation_write(struct iwl_mld *mld, char *buf, size_t count, + void *data) +{ + struct ieee80211_vif *vif = data; + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_twt_operation_cmd twt_cmd = {}; + int link_id = vif->active_links ? __ffs(vif->active_links) : 0; + struct iwl_mld_link *mld_link = iwl_mld_link_dereference_check(mld_vif, + link_id); + int ret; + + if (WARN_ON(!mld_link)) + return -ENODEV; + + if (iwl_mld_dbgfs_fw_cmd_disabled(mld)) + return -EIO; + + if (hweight16(vif->active_links) > 1) + return -EOPNOTSUPP; + + ret = sscanf(buf, + "%u %llu %u %u %u %hhu %hhu %hhu %hhu %hhu %hhu %hhu %hhu %hhu %hhu %hhu %hhu %hhu %hhu %hhu %hhu", + &twt_cmd.twt_operation, &twt_cmd.target_wake_time, + &twt_cmd.interval_exponent, &twt_cmd.interval_mantissa, + &twt_cmd.minimum_wake_duration, &twt_cmd.trigger, + &twt_cmd.flow_type, &twt_cmd.flow_id, + &twt_cmd.twt_protection, &twt_cmd.ndp_paging_indicator, + &twt_cmd.responder_pm_mode, &twt_cmd.negotiation_type, + &twt_cmd.twt_request, &twt_cmd.implicit, + &twt_cmd.twt_group_assignment, &twt_cmd.twt_channel, + &twt_cmd.restricted_info_present, &twt_cmd.dl_bitmap_valid, + &twt_cmd.ul_bitmap_valid, &twt_cmd.dl_tid_bitmap, + &twt_cmd.ul_tid_bitmap); + + if (ret != 21) + return -EINVAL; + + twt_cmd.link_id = cpu_to_le32(mld_link->fw_id); + + ret = iwl_mld_send_cmd_pdu(mld, + WIDE_ID(MAC_CONF_GROUP, TWT_OPERATION_CMD), + &twt_cmd); + return ret ?: count; +} + +VIF_DEBUGFS_WRITE_FILE_OPS(twt_operation, 256); + +void iwl_mld_add_vif_debugfs(struct ieee80211_hw *hw, + struct ieee80211_vif *vif) +{ + struct dentry *mld_vif_dbgfs = + debugfs_create_dir("iwlmld", vif->debugfs_dir); + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + char target[3 * 3 + 11 + (NL80211_WIPHY_NAME_MAXLEN + 1) + + (7 + IFNAMSIZ + 1) + 6 + 1]; + char name[7 + IFNAMSIZ + 1]; + + /* Create symlink for convenience pointing to interface specific + * debugfs entries for the driver. For example, under + * /sys/kernel/debug/iwlwifi/0000\:02\:00.0/iwlmld/ + * find + * netdev:wlan0 -> ../../../ieee80211/phy0/netdev:wlan0/iwlmld/ + */ + snprintf(name, sizeof(name), "%pd", vif->debugfs_dir); + snprintf(target, sizeof(target), "../../../%pd3/iwlmld", + vif->debugfs_dir); + mld_vif->dbgfs_slink = + debugfs_create_symlink(name, mld->debugfs_dir, target); + + if (iwlmld_mod_params.power_scheme != IWL_POWER_SCHEME_CAM && + vif->type == NL80211_IFTYPE_STATION) { + VIF_DEBUGFS_ADD_FILE(pm_params, mld_vif_dbgfs, 0200); + VIF_DEBUGFS_ADD_FILE(bf_params, mld_vif_dbgfs, 0200); + } + + if (vif->type == NL80211_IFTYPE_AP) { + VIF_DEBUGFS_ADD_FILE(inject_beacon_ie, mld_vif_dbgfs, 0200); + VIF_DEBUGFS_ADD_FILE(inject_beacon_ie_restore, + mld_vif_dbgfs, 0200); + } + + VIF_DEBUGFS_ADD_FILE(low_latency, mld_vif_dbgfs, 0600); + VIF_DEBUGFS_ADD_FILE(twt_setup, mld_vif_dbgfs, 0200); + VIF_DEBUGFS_ADD_FILE(twt_operation, mld_vif_dbgfs, 0200); +} + +#define LINK_DEBUGFS_WRITE_FILE_OPS(name, bufsz) \ + WIPHY_DEBUGFS_WRITE_FILE_OPS(link_##name, bufsz, bss_conf) + +#define LINK_DEBUGFS_ADD_FILE_ALIAS(alias, name, parent, mode) \ + debugfs_create_file(alias, mode, parent, link_conf, \ + &iwl_dbgfs_link_##name##_ops) +#define LINK_DEBUGFS_ADD_FILE(name, parent, mode) \ + LINK_DEBUGFS_ADD_FILE_ALIAS(#name, name, parent, mode) + +void iwl_mld_add_link_debugfs(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link_conf, + struct dentry *dir) +{ + struct dentry *mld_link_dir; + + mld_link_dir = debugfs_lookup("iwlmld", dir); + + /* For non-MLO vifs, the dir of deflink is the same as the vif's one. + * so if iwlmld dir already exists, this means that this is deflink. + * If not, this is a per-link dir of a MLO vif, add in it the iwlmld + * dir. + */ + if (!mld_link_dir) + mld_link_dir = debugfs_create_dir("iwlmld", dir); +} + +static ssize_t iwl_dbgfs_fixed_rate_write(struct iwl_mld *mld, char *buf, + size_t count, void *data) +{ + struct ieee80211_link_sta *link_sta = data; + struct iwl_mld_link_sta *mld_link_sta; + u32 rate; + u32 partial = false; + char pretty_rate[100]; + int ret; + u8 fw_sta_id; + + mld_link_sta = iwl_mld_link_sta_from_mac80211(link_sta); + if (WARN_ON(!mld_link_sta)) + return -EINVAL; + + fw_sta_id = mld_link_sta->fw_id; + + if (sscanf(buf, "%i %i", &rate, &partial) == 0) + return -EINVAL; + + if (iwl_mld_dbgfs_fw_cmd_disabled(mld)) + return -EIO; + + ret = iwl_mld_send_tlc_dhc(mld, fw_sta_id, + partial ? IWL_TLC_DEBUG_PARTIAL_FIXED_RATE : + IWL_TLC_DEBUG_FIXED_RATE, + rate); + + rs_pretty_print_rate(pretty_rate, sizeof(pretty_rate), rate); + + IWL_DEBUG_RATE(mld, "sta_id %d rate %s partial: %d, ret:%d\n", + fw_sta_id, pretty_rate, partial, ret); + + return ret ? : count; +} + +static ssize_t iwl_dbgfs_tlc_dhc_write(struct iwl_mld *mld, char *buf, + size_t count, void *data) +{ + struct ieee80211_link_sta *link_sta = data; + struct iwl_mld_link_sta *mld_link_sta; + u32 type, value; + int ret; + u8 fw_sta_id; + + mld_link_sta = iwl_mld_link_sta_from_mac80211(link_sta); + if (WARN_ON(!mld_link_sta)) + return -EINVAL; + + fw_sta_id = mld_link_sta->fw_id; + + if (sscanf(buf, "%i %i", &type, &value) != 2) { + IWL_DEBUG_RATE(mld, "usage \n"); + return -EINVAL; + } + + if (iwl_mld_dbgfs_fw_cmd_disabled(mld)) + return -EIO; + + ret = iwl_mld_send_tlc_dhc(mld, fw_sta_id, type, value); + + return ret ? : count; +} + +#define LINK_STA_DEBUGFS_ADD_FILE_ALIAS(alias, name, parent, mode) \ + debugfs_create_file(alias, mode, parent, link_sta, \ + &iwl_dbgfs_##name##_ops) +#define LINK_STA_DEBUGFS_ADD_FILE(name, parent, mode) \ + LINK_STA_DEBUGFS_ADD_FILE_ALIAS(#name, name, parent, mode) + +#define LINK_STA_WIPHY_DEBUGFS_WRITE_OPS(name, bufsz) \ + WIPHY_DEBUGFS_WRITE_FILE_OPS(name, bufsz, link_sta) + +LINK_STA_WIPHY_DEBUGFS_WRITE_OPS(tlc_dhc, 64); +LINK_STA_WIPHY_DEBUGFS_WRITE_OPS(fixed_rate, 64); + +void iwl_mld_add_link_sta_debugfs(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_link_sta *link_sta, + struct dentry *dir) +{ + LINK_STA_DEBUGFS_ADD_FILE(fixed_rate, dir, 0200); + LINK_STA_DEBUGFS_ADD_FILE(tlc_dhc, dir, 0200); +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/debugfs.h b/drivers/net/wireless/intel/iwlwifi/mld/debugfs.h new file mode 100644 index 0000000000000..04c536db8b166 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/debugfs.h @@ -0,0 +1,200 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024 Intel Corporation + */ +#include "iface.h" +#include "sta.h" + +#define MLD_DEBUGFS_OPEN_WRAPPER(name, buflen, argtype) \ +struct dbgfs_##name##_data { \ + argtype *arg; \ + bool read_done; \ + ssize_t rlen; \ + char buf[buflen]; \ +}; \ +static int _iwl_dbgfs_##name##_open(struct inode *inode, \ + struct file *file) \ +{ \ + struct dbgfs_##name##_data *data; \ + \ + if ((file->f_flags & O_ACCMODE) == O_RDWR) \ + return -EOPNOTSUPP; \ + \ + data = kzalloc(sizeof(*data), GFP_KERNEL); \ + if (!data) \ + return -ENOMEM; \ + \ + data->read_done = false; \ + data->arg = inode->i_private; \ + file->private_data = data; \ + \ + return 0; \ +} + +#define MLD_DEBUGFS_READ_WRAPPER(name) \ +static ssize_t _iwl_dbgfs_##name##_read(struct file *file, \ + char __user *user_buf, \ + size_t count, loff_t *ppos) \ +{ \ + struct dbgfs_##name##_data *data = file->private_data; \ + \ + if (!data->read_done) { \ + data->read_done = true; \ + data->rlen = iwl_dbgfs_##name##_read(data->arg, \ + sizeof(data->buf),\ + data->buf); \ + } \ + \ + if (data->rlen < 0) \ + return data->rlen; \ + return simple_read_from_buffer(user_buf, count, ppos, \ + data->buf, data->rlen); \ +} + +static int _iwl_dbgfs_release(struct inode *inode, struct file *file) +{ + kfree(file->private_data); + return 0; +} + +#define _MLD_DEBUGFS_READ_FILE_OPS(name, buflen, argtype) \ +MLD_DEBUGFS_OPEN_WRAPPER(name, buflen, argtype) \ +MLD_DEBUGFS_READ_WRAPPER(name) \ +static const struct file_operations iwl_dbgfs_##name##_ops = { \ + .read = _iwl_dbgfs_##name##_read, \ + .open = _iwl_dbgfs_##name##_open, \ + .llseek = generic_file_llseek, \ + .release = _iwl_dbgfs_release, \ +} + +#define WIPHY_DEBUGFS_HANDLER_WRAPPER(name) \ +static ssize_t iwl_dbgfs_##name##_write_handler(struct wiphy *wiphy, \ + struct file *file, char *buf, \ + size_t count, void *data) \ +{ \ + struct ieee80211_hw *hw = wiphy_to_ieee80211_hw(wiphy); \ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); \ + return iwl_dbgfs_##name##_write(mld, buf, count, data); \ +} + +static inline struct iwl_mld * +iwl_mld_from_link_sta(struct ieee80211_link_sta *link_sta) +{ + struct ieee80211_vif *vif = + iwl_mld_sta_from_mac80211(link_sta->sta)->vif; + return iwl_mld_vif_from_mac80211(vif)->mld; +} + +static inline struct iwl_mld * +iwl_mld_from_bss_conf(struct ieee80211_bss_conf *link) +{ + return iwl_mld_vif_from_mac80211(link->vif)->mld; +} + +static inline struct iwl_mld *iwl_mld_from_vif(struct ieee80211_vif *vif) +{ + return iwl_mld_vif_from_mac80211(vif)->mld; +} + +#define WIPHY_DEBUGFS_WRITE_WRAPPER(name, bufsz, objtype) \ +WIPHY_DEBUGFS_HANDLER_WRAPPER(name) \ +static ssize_t __iwl_dbgfs_##name##_write(struct file *file, \ + const char __user *user_buf, \ + size_t count, loff_t *ppos) \ +{ \ + struct ieee80211_##objtype *arg = file->private_data; \ + struct iwl_mld *mld = iwl_mld_from_##objtype(arg); \ + char buf[bufsz] = {}; \ + \ + return wiphy_locked_debugfs_write(mld->wiphy, file, \ + buf, sizeof(buf), \ + user_buf, count, \ + iwl_dbgfs_##name##_write_handler, \ + arg); \ +} + +#define WIPHY_DEBUGFS_WRITE_FILE_OPS(name, bufsz, objtype) \ + WIPHY_DEBUGFS_WRITE_WRAPPER(name, bufsz, objtype) \ + static const struct file_operations iwl_dbgfs_##name##_ops = { \ + .write = __iwl_dbgfs_##name##_write, \ + .open = simple_open, \ + .llseek = generic_file_llseek, \ + } + +#define WIPHY_DEBUGFS_HANDLER_WRAPPER_MLD(name) \ +static ssize_t iwl_dbgfs_##name##_write_handler(struct wiphy *wiphy, \ + struct file *file, char *buf, \ + size_t count, void *data) \ +{ \ + struct ieee80211_hw *hw = wiphy_to_ieee80211_hw(wiphy); \ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); \ + return iwl_dbgfs_##name##_write(mld, buf, count); \ +} + +#define WIPHY_DEBUGFS_WRITE_WRAPPER_MLD(name) \ +WIPHY_DEBUGFS_HANDLER_WRAPPER_MLD(name) \ +static ssize_t __iwl_dbgfs_##name##_write(struct file *file, \ + const char __user *user_buf, \ + size_t count, loff_t *ppos) \ +{ \ + struct dbgfs_##name##_data *data = file->private_data; \ + struct iwl_mld *mld = data->arg; \ + \ + return wiphy_locked_debugfs_write(mld->wiphy, file, \ + data->buf, sizeof(data->buf), \ + user_buf, count, \ + iwl_dbgfs_##name##_write_handler, \ + NULL); \ +} + +#define WIPHY_DEBUGFS_WRITE_FILE_OPS_MLD(name, bufsz) \ + MLD_DEBUGFS_OPEN_WRAPPER(name, bufsz, struct iwl_mld) \ + WIPHY_DEBUGFS_WRITE_WRAPPER_MLD(name) \ + static const struct file_operations iwl_dbgfs_##name##_ops = { \ + .write = __iwl_dbgfs_##name##_write, \ + .open = _iwl_dbgfs_##name##_open, \ + .llseek = generic_file_llseek, \ + .release = _iwl_dbgfs_release, \ + } + +#define WIPHY_DEBUGFS_READ_WRITE_FILE_OPS_MLD(name, bufsz) \ + MLD_DEBUGFS_OPEN_WRAPPER(name, bufsz, struct iwl_mld) \ + WIPHY_DEBUGFS_WRITE_WRAPPER_MLD(name) \ + MLD_DEBUGFS_READ_WRAPPER(name) \ + static const struct file_operations iwl_dbgfs_##name##_ops = { \ + .write = __iwl_dbgfs_##name##_write, \ + .read = _iwl_dbgfs_##name##_read, \ + .open = _iwl_dbgfs_##name##_open, \ + .llseek = generic_file_llseek, \ + .release = _iwl_dbgfs_release, \ + } + +#define WIPHY_DEBUGFS_WRITE_WRAPPER_IEEE80211(name, bufsz, objtype) \ +WIPHY_DEBUGFS_HANDLER_WRAPPER(name) \ +static ssize_t _iwl_dbgfs_##name##_write(struct file *file, \ + const char __user *user_buf, \ + size_t count, loff_t *ppos) \ +{ \ + struct dbgfs_##name##_data *data = file->private_data; \ + struct ieee80211_##objtype *arg = data->arg; \ + struct iwl_mld *mld = iwl_mld_from_##objtype(arg); \ + char buf[bufsz] = {}; \ + \ + return wiphy_locked_debugfs_write(mld->wiphy, file, \ + buf, sizeof(buf), \ + user_buf, count, \ + iwl_dbgfs_##name##_write_handler, \ + arg); \ +} + +#define IEEE80211_WIPHY_DEBUGFS_READ_WRITE_FILE_OPS(name, bufsz, objtype) \ + MLD_DEBUGFS_OPEN_WRAPPER(name, bufsz, struct ieee80211_##objtype) \ + WIPHY_DEBUGFS_WRITE_WRAPPER_IEEE80211(name, bufsz, objtype) \ + MLD_DEBUGFS_READ_WRAPPER(name) \ + static const struct file_operations iwl_dbgfs_##name##_ops = { \ + .write = _iwl_dbgfs_##name##_write, \ + .read = _iwl_dbgfs_##name##_read, \ + .open = _iwl_dbgfs_##name##_open, \ + .llseek = generic_file_llseek, \ + .release = _iwl_dbgfs_release, \ + } diff --git a/drivers/net/wireless/intel/iwlwifi/mld/ftm-initiator.c b/drivers/net/wireless/intel/iwlwifi/mld/ftm-initiator.c new file mode 100644 index 0000000000000..f77ba21a174dd --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/ftm-initiator.c @@ -0,0 +1,451 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2025 Intel Corporation + */ +#include +#include +#include +#include "mld.h" +#include "iface.h" +#include "phy.h" +#include "iwl-io.h" +#include "iwl-prph.h" +#include "constants.h" +#include "fw/api/location.h" +#include "ftm-initiator.h" + +static void iwl_mld_ftm_cmd_common(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct iwl_tof_range_req_cmd *cmd, + struct cfg80211_pmsr_request *req) +{ + int i; + + cmd->initiator_flags = + cpu_to_le32(IWL_TOF_INITIATOR_FLAGS_MACADDR_RANDOM | + IWL_TOF_INITIATOR_FLAGS_NON_ASAP_SUPPORT); + cmd->request_id = req->cookie; + cmd->num_of_ap = req->n_peers; + + /* Use a large value for "no timeout". Don't use the maximum value + * because of fw limitations. + */ + if (req->timeout) + cmd->req_timeout_ms = cpu_to_le32(min(req->timeout, 0xfffff)); + else + cmd->req_timeout_ms = cpu_to_le32(0xfffff); + + memcpy(cmd->macaddr_template, req->mac_addr, ETH_ALEN); + for (i = 0; i < ETH_ALEN; i++) + cmd->macaddr_mask[i] = ~req->mac_addr_mask[i]; + + if (vif->cfg.assoc) { + memcpy(cmd->range_req_bssid, vif->bss_conf.bssid, ETH_ALEN); + + /* AP's TSF is only relevant if associated */ + for (i = 0; i < req->n_peers; i++) { + if (req->peers[i].report_ap_tsf) { + struct iwl_mld_vif *mld_vif = + iwl_mld_vif_from_mac80211(vif); + + cmd->tsf_mac_id = cpu_to_le32(mld_vif->fw_id); + return; + } + } + } else { + eth_broadcast_addr(cmd->range_req_bssid); + } + + /* Don't report AP's TSF */ + cmd->tsf_mac_id = cpu_to_le32(0xff); +} + +static int +iwl_mld_ftm_set_target_chandef(struct iwl_mld *mld, + struct cfg80211_pmsr_request_peer *peer, + struct iwl_tof_range_req_ap_entry *target) +{ + u32 freq = peer->chandef.chan->center_freq; + + target->channel_num = ieee80211_frequency_to_channel(freq); + + switch (peer->chandef.width) { + case NL80211_CHAN_WIDTH_20_NOHT: + target->format_bw = IWL_LOCATION_FRAME_FORMAT_LEGACY; + target->format_bw |= IWL_LOCATION_BW_20MHZ << LOCATION_BW_POS; + break; + case NL80211_CHAN_WIDTH_20: + target->format_bw = IWL_LOCATION_FRAME_FORMAT_HT; + target->format_bw |= IWL_LOCATION_BW_20MHZ << LOCATION_BW_POS; + break; + case NL80211_CHAN_WIDTH_40: + target->format_bw = IWL_LOCATION_FRAME_FORMAT_HT; + target->format_bw |= IWL_LOCATION_BW_40MHZ << LOCATION_BW_POS; + break; + case NL80211_CHAN_WIDTH_80: + target->format_bw = IWL_LOCATION_FRAME_FORMAT_VHT; + target->format_bw |= IWL_LOCATION_BW_80MHZ << LOCATION_BW_POS; + break; + case NL80211_CHAN_WIDTH_160: + target->format_bw = IWL_LOCATION_FRAME_FORMAT_HE; + target->format_bw |= IWL_LOCATION_BW_160MHZ << LOCATION_BW_POS; + break; + default: + IWL_ERR(mld, "Unsupported BW in FTM request (%d)\n", + peer->chandef.width); + return -EINVAL; +} + + /* non EDCA based measurement must use HE preamble */ + if (peer->ftm.trigger_based || peer->ftm.non_trigger_based) + target->format_bw |= IWL_LOCATION_FRAME_FORMAT_HE; + + target->ctrl_ch_position = + (peer->chandef.width > NL80211_CHAN_WIDTH_20) ? + iwl_mld_get_fw_ctrl_pos(&peer->chandef) : 0; + + target->band = iwl_mld_nl80211_band_to_fw(peer->chandef.chan->band); + return 0; +} + +#define FTM_SET_FLAG(flag) (target->initiator_ap_flags |= \ + cpu_to_le32(IWL_INITIATOR_AP_FLAGS_##flag)) + +static void +iwl_mld_ftm_set_target_flags(struct iwl_mld *mld, + struct cfg80211_pmsr_request_peer *peer, + struct iwl_tof_range_req_ap_entry *target) +{ + target->initiator_ap_flags = cpu_to_le32(0); + + if (peer->ftm.asap) + FTM_SET_FLAG(ASAP); + + if (peer->ftm.request_lci) + FTM_SET_FLAG(LCI_REQUEST); + + if (peer->ftm.request_civicloc) + FTM_SET_FLAG(CIVIC_REQUEST); + + if (IWL_MLD_FTM_INITIATOR_DYNACK) + FTM_SET_FLAG(DYN_ACK); + + if (IWL_MLD_FTM_INITIATOR_ALGO == IWL_TOF_ALGO_TYPE_LINEAR_REG) + FTM_SET_FLAG(ALGO_LR); + else if (IWL_MLD_FTM_INITIATOR_ALGO == IWL_TOF_ALGO_TYPE_FFT) + FTM_SET_FLAG(ALGO_FFT); + + if (peer->ftm.trigger_based) + FTM_SET_FLAG(TB); + else if (peer->ftm.non_trigger_based) + FTM_SET_FLAG(NON_TB); + + if ((peer->ftm.trigger_based || peer->ftm.non_trigger_based) && + peer->ftm.lmr_feedback) + FTM_SET_FLAG(LMR_FEEDBACK); +} + +static void iwl_mld_ftm_set_sta(struct iwl_mld *mld, struct ieee80211_vif *vif, + struct cfg80211_pmsr_request_peer *peer, + struct iwl_tof_range_req_ap_entry *target) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + u32 sta_id_mask; + + target->sta_id = IWL_INVALID_STA; + + /* TODO: add ftm_unprotected debugfs support */ + + if (!vif->cfg.assoc || !mld_vif->ap_sta) + return; + + sta_id_mask = iwl_mld_fw_sta_id_mask(mld, mld_vif->ap_sta); + if (WARN_ON(hweight32(sta_id_mask) != 1)) + return; + + target->sta_id = __ffs(sta_id_mask); + + if (mld_vif->ap_sta->mfp && + (peer->ftm.trigger_based || peer->ftm.non_trigger_based)) + FTM_SET_FLAG(PMF); +} + +static int +iwl_mld_ftm_set_target(struct iwl_mld *mld, struct ieee80211_vif *vif, + struct cfg80211_pmsr_request_peer *peer, + struct iwl_tof_range_req_ap_entry *target) +{ + u32 i2r_max_sts; + int ret; + + ret = iwl_mld_ftm_set_target_chandef(mld, peer, target); + if (ret) + return ret; + + memcpy(target->bssid, peer->addr, ETH_ALEN); + target->burst_period = cpu_to_le16(peer->ftm.burst_period); + target->samples_per_burst = peer->ftm.ftms_per_burst; + target->num_of_bursts = peer->ftm.num_bursts_exp; + iwl_mld_ftm_set_target_flags(mld, peer, target); + iwl_mld_ftm_set_sta(mld, vif, peer, target); + + /* TODO: add secured ranging support */ + + i2r_max_sts = IWL_MLD_FTM_I2R_MAX_STS > 1 ? 1 : + IWL_MLD_FTM_I2R_MAX_STS; + + target->r2i_ndp_params = IWL_MLD_FTM_R2I_MAX_REP | + (IWL_MLD_FTM_R2I_MAX_STS << IWL_LOCATION_MAX_STS_POS) | + (IWL_MLD_FTM_R2I_MAX_TOTAL_LTF << IWL_LOCATION_TOTAL_LTF_POS); + target->i2r_ndp_params = IWL_MLD_FTM_I2R_MAX_REP | + (i2r_max_sts << IWL_LOCATION_MAX_STS_POS) | + (IWL_MLD_FTM_I2R_MAX_TOTAL_LTF << IWL_LOCATION_TOTAL_LTF_POS); + + if (peer->ftm.non_trigger_based) { + target->min_time_between_msr = + cpu_to_le16(IWL_MLD_FTM_NON_TB_MIN_TIME_BETWEEN_MSR); + target->burst_period = + cpu_to_le16(IWL_MLD_FTM_NON_TB_MAX_TIME_BETWEEN_MSR); + } else { + target->min_time_between_msr = cpu_to_le16(0); + } + + /* TODO: Beacon interval is currently unknown, so use the common value + * of 100 TUs. + */ + target->beacon_interval = cpu_to_le16(100); + + return 0; +} + +int iwl_mld_ftm_start(struct iwl_mld *mld, struct ieee80211_vif *vif, + struct cfg80211_pmsr_request *req) +{ + struct iwl_tof_range_req_cmd cmd; + struct iwl_host_cmd hcmd = { + .id = WIDE_ID(LOCATION_GROUP, TOF_RANGE_REQ_CMD), + .dataflags[0] = IWL_HCMD_DFL_DUP, + .data[0] = &cmd, + .len[0] = sizeof(cmd), + }; + u8 i; + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + if (mld->ftm_initiator.req) + return -EBUSY; + + if (req->n_peers > ARRAY_SIZE(cmd.ap)) + return -EINVAL; + + memset(&cmd, 0, sizeof(cmd)); + + iwl_mld_ftm_cmd_common(mld, vif, (void *)&cmd, req); + + for (i = 0; i < cmd.num_of_ap; i++) { + struct cfg80211_pmsr_request_peer *peer = &req->peers[i]; + struct iwl_tof_range_req_ap_entry *target = &cmd.ap[i]; + + ret = iwl_mld_ftm_set_target(mld, vif, peer, target); + if (ret) + return ret; + } + + /* TODO: get the status from the response*/ + ret = iwl_mld_send_cmd(mld, &hcmd); + if (!ret) { + mld->ftm_initiator.req = req; + mld->ftm_initiator.req_wdev = ieee80211_vif_to_wdev(vif); + } + + return ret; +} + +static void iwl_mld_ftm_reset(struct iwl_mld *mld) +{ + lockdep_assert_wiphy(mld->wiphy); + + mld->ftm_initiator.req = NULL; + mld->ftm_initiator.req_wdev = NULL; + memset(mld->ftm_initiator.responses, 0, + sizeof(mld->ftm_initiator.responses)); +} + +static int iwl_mld_ftm_range_resp_valid(struct iwl_mld *mld, u8 request_id, + u8 num_of_aps) +{ + if (IWL_FW_CHECK(mld, request_id != (u8)mld->ftm_initiator.req->cookie, + "Request ID mismatch, got %u, active %u\n", + request_id, (u8)mld->ftm_initiator.req->cookie)) + return -EINVAL; + + if (IWL_FW_CHECK(mld, num_of_aps > mld->ftm_initiator.req->n_peers || + num_of_aps > IWL_TOF_MAX_APS, + "FTM range response: invalid num of APs (%u)\n", + num_of_aps)) + return -EINVAL; + + return 0; +} + +static int iwl_mld_ftm_find_peer(struct cfg80211_pmsr_request *req, + const u8 *addr) +{ + for (int i = 0; i < req->n_peers; i++) { + struct cfg80211_pmsr_request_peer *peer = &req->peers[i]; + + if (ether_addr_equal_unaligned(peer->addr, addr)) + return i; + } + + return -ENOENT; +} + +static void iwl_mld_debug_range_resp(struct iwl_mld *mld, u8 index, + struct cfg80211_pmsr_result *res) +{ + s64 rtt_avg = div_s64(res->ftm.rtt_avg * 100, 6666); + + IWL_DEBUG_INFO(mld, "entry %d\n", index); + IWL_DEBUG_INFO(mld, "\tstatus: %d\n", res->status); + IWL_DEBUG_INFO(mld, "\tBSSID: %pM\n", res->addr); + IWL_DEBUG_INFO(mld, "\thost time: %llu\n", res->host_time); + IWL_DEBUG_INFO(mld, "\tburst index: %d\n", res->ftm.burst_index); + IWL_DEBUG_INFO(mld, "\tsuccess num: %u\n", res->ftm.num_ftmr_successes); + IWL_DEBUG_INFO(mld, "\trssi: %d\n", res->ftm.rssi_avg); + IWL_DEBUG_INFO(mld, "\trssi spread: %d\n", res->ftm.rssi_spread); + IWL_DEBUG_INFO(mld, "\trtt: %lld\n", res->ftm.rtt_avg); + IWL_DEBUG_INFO(mld, "\trtt var: %llu\n", res->ftm.rtt_variance); + IWL_DEBUG_INFO(mld, "\trtt spread: %llu\n", res->ftm.rtt_spread); + IWL_DEBUG_INFO(mld, "\tdistance: %lld\n", rtt_avg); +} + +void iwl_mld_handle_ftm_resp_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + struct iwl_tof_range_rsp_ntfy *fw_resp = (void *)pkt->data; + u8 num_of_aps, last_in_batch; + + if (IWL_FW_CHECK(mld, !mld->ftm_initiator.req, + "FTM response without a pending request\n")) + return; + + if (iwl_mld_ftm_range_resp_valid(mld, fw_resp->request_id, + fw_resp->num_of_aps)) + return; + + num_of_aps = fw_resp->num_of_aps; + last_in_batch = fw_resp->last_report; + + IWL_DEBUG_INFO(mld, "Range response received\n"); + IWL_DEBUG_INFO(mld, "request id: %llu, num of entries: %u\n", + mld->ftm_initiator.req->cookie, num_of_aps); + + for (int i = 0; i < num_of_aps; i++) { + struct cfg80211_pmsr_result result = {}; + struct iwl_tof_range_rsp_ap_entry_ntfy *fw_ap; + int peer_idx; + + fw_ap = &fw_resp->ap[i]; + result.final = fw_ap->last_burst; + result.ap_tsf = le32_to_cpu(fw_ap->start_tsf); + result.ap_tsf_valid = 1; + + peer_idx = iwl_mld_ftm_find_peer(mld->ftm_initiator.req, + fw_ap->bssid); + if (peer_idx < 0) { + IWL_WARN(mld, + "Unknown address (%pM, target #%d) in FTM response\n", + fw_ap->bssid, i); + continue; + } + + switch (fw_ap->measure_status) { + case IWL_TOF_ENTRY_SUCCESS: + result.status = NL80211_PMSR_STATUS_SUCCESS; + break; + case IWL_TOF_ENTRY_TIMING_MEASURE_TIMEOUT: + result.status = NL80211_PMSR_STATUS_TIMEOUT; + break; + case IWL_TOF_ENTRY_NO_RESPONSE: + result.status = NL80211_PMSR_STATUS_FAILURE; + result.ftm.failure_reason = + NL80211_PMSR_FTM_FAILURE_NO_RESPONSE; + break; + case IWL_TOF_ENTRY_REQUEST_REJECTED: + result.status = NL80211_PMSR_STATUS_FAILURE; + result.ftm.failure_reason = + NL80211_PMSR_FTM_FAILURE_PEER_BUSY; + result.ftm.busy_retry_time = fw_ap->refusal_period; + break; + default: + result.status = NL80211_PMSR_STATUS_FAILURE; + result.ftm.failure_reason = + NL80211_PMSR_FTM_FAILURE_UNSPECIFIED; + break; + } + memcpy(result.addr, fw_ap->bssid, ETH_ALEN); + + /* TODO: convert the timestamp from the result to systime */ + result.host_time = ktime_get_boottime_ns(); + + result.type = NL80211_PMSR_TYPE_FTM; + result.ftm.burst_index = mld->ftm_initiator.responses[peer_idx]; + mld->ftm_initiator.responses[peer_idx]++; + result.ftm.rssi_avg = fw_ap->rssi; + result.ftm.rssi_avg_valid = 1; + result.ftm.rssi_spread = fw_ap->rssi_spread; + result.ftm.rssi_spread_valid = 1; + result.ftm.rtt_avg = (s32)le32_to_cpu(fw_ap->rtt); + result.ftm.rtt_avg_valid = 1; + result.ftm.rtt_variance = le32_to_cpu(fw_ap->rtt_variance); + result.ftm.rtt_variance_valid = 1; + result.ftm.rtt_spread = le32_to_cpu(fw_ap->rtt_spread); + result.ftm.rtt_spread_valid = 1; + + cfg80211_pmsr_report(mld->ftm_initiator.req_wdev, + mld->ftm_initiator.req, + &result, GFP_KERNEL); + + if (fw_has_api(&mld->fw->ucode_capa, + IWL_UCODE_TLV_API_FTM_RTT_ACCURACY)) + IWL_DEBUG_INFO(mld, "RTT confidence: %u\n", + fw_ap->rttConfidence); + + iwl_mld_debug_range_resp(mld, i, &result); + } + + if (last_in_batch) { + cfg80211_pmsr_complete(mld->ftm_initiator.req_wdev, + mld->ftm_initiator.req, + GFP_KERNEL); + iwl_mld_ftm_reset(mld); + } +} + +void iwl_mld_ftm_restart_cleanup(struct iwl_mld *mld) +{ + struct cfg80211_pmsr_result result = { + .status = NL80211_PMSR_STATUS_FAILURE, + .final = 1, + .host_time = ktime_get_boottime_ns(), + .type = NL80211_PMSR_TYPE_FTM, + }; + + if (!mld->ftm_initiator.req) + return; + + for (int i = 0; i < mld->ftm_initiator.req->n_peers; i++) { + memcpy(result.addr, mld->ftm_initiator.req->peers[i].addr, + ETH_ALEN); + + cfg80211_pmsr_report(mld->ftm_initiator.req_wdev, + mld->ftm_initiator.req, + &result, GFP_KERNEL); + } + + cfg80211_pmsr_complete(mld->ftm_initiator.req_wdev, + mld->ftm_initiator.req, GFP_KERNEL); + iwl_mld_ftm_reset(mld); +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/ftm-initiator.h b/drivers/net/wireless/intel/iwlwifi/mld/ftm-initiator.h new file mode 100644 index 0000000000000..e98fac34beba4 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/ftm-initiator.h @@ -0,0 +1,15 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2025 Intel Corporation + */ +#ifndef __iwl_mld_ftm_initiator_h__ +#define __iwl_mld_ftm_initiator_h__ + +int iwl_mld_ftm_start(struct iwl_mld *mld, struct ieee80211_vif *vif, + struct cfg80211_pmsr_request *req); + +void iwl_mld_handle_ftm_resp_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt); +void iwl_mld_ftm_restart_cleanup(struct iwl_mld *mld); + +#endif /* __iwl_mld_ftm_initiator_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/fw.c b/drivers/net/wireless/intel/iwlwifi/mld/fw.c new file mode 100644 index 0000000000000..ec93af362cfc8 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/fw.c @@ -0,0 +1,536 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024-2025 Intel Corporation + */ + +#include "mld.h" + +#include "fw/api/alive.h" +#include "fw/api/scan.h" +#include "fw/api/rx.h" +#include "fw/dbg.h" +#include "fw/pnvm.h" +#include "hcmd.h" +#include "iwl-nvm-parse.h" +#include "power.h" +#include "mcc.h" +#include "led.h" +#include "coex.h" +#include "regulatory.h" +#include "thermal.h" + +static int iwl_mld_send_tx_ant_cfg(struct iwl_mld *mld) +{ + struct iwl_tx_ant_cfg_cmd cmd; + + lockdep_assert_wiphy(mld->wiphy); + + cmd.valid = cpu_to_le32(iwl_mld_get_valid_tx_ant(mld)); + + IWL_DEBUG_FW(mld, "select valid tx ant: %u\n", cmd.valid); + + return iwl_mld_send_cmd_pdu(mld, TX_ANT_CONFIGURATION_CMD, &cmd); +} + +static int iwl_mld_send_rss_cfg_cmd(struct iwl_mld *mld) +{ + struct iwl_rss_config_cmd cmd = { + .flags = cpu_to_le32(IWL_RSS_ENABLE), + .hash_mask = BIT(IWL_RSS_HASH_TYPE_IPV4_TCP) | + BIT(IWL_RSS_HASH_TYPE_IPV4_UDP) | + BIT(IWL_RSS_HASH_TYPE_IPV4_PAYLOAD) | + BIT(IWL_RSS_HASH_TYPE_IPV6_TCP) | + BIT(IWL_RSS_HASH_TYPE_IPV6_UDP) | + BIT(IWL_RSS_HASH_TYPE_IPV6_PAYLOAD), + }; + + lockdep_assert_wiphy(mld->wiphy); + + /* Do not direct RSS traffic to Q 0 which is our fallback queue */ + for (int i = 0; i < ARRAY_SIZE(cmd.indirection_table); i++) + cmd.indirection_table[i] = + 1 + (i % (mld->trans->num_rx_queues - 1)); + netdev_rss_key_fill(cmd.secret_key, sizeof(cmd.secret_key)); + + return iwl_mld_send_cmd_pdu(mld, RSS_CONFIG_CMD, &cmd); +} + +static int iwl_mld_config_scan(struct iwl_mld *mld) +{ + struct iwl_scan_config cmd = { + .tx_chains = cpu_to_le32(iwl_mld_get_valid_tx_ant(mld)), + .rx_chains = cpu_to_le32(iwl_mld_get_valid_rx_ant(mld)) + }; + + return iwl_mld_send_cmd_pdu(mld, WIDE_ID(LONG_GROUP, SCAN_CFG_CMD), + &cmd); +} + +static void iwl_mld_alive_imr_data(struct iwl_trans *trans, + const struct iwl_imr_alive_info *imr_info) +{ + struct iwl_imr_data *imr_data = &trans->dbg.imr_data; + + imr_data->imr_enable = le32_to_cpu(imr_info->enabled); + imr_data->imr_size = le32_to_cpu(imr_info->size); + imr_data->imr2sram_remainbyte = imr_data->imr_size; + imr_data->imr_base_addr = imr_info->base_addr; + imr_data->imr_curr_addr = le64_to_cpu(imr_data->imr_base_addr); + + if (imr_data->imr_enable) + return; + + for (int i = 0; i < ARRAY_SIZE(trans->dbg.active_regions); i++) { + struct iwl_fw_ini_region_tlv *reg; + + if (!trans->dbg.active_regions[i]) + continue; + + reg = (void *)trans->dbg.active_regions[i]->data; + + /* We have only one DRAM IMR region, so we + * can break as soon as we find the first + * one. + */ + if (reg->type == IWL_FW_INI_REGION_DRAM_IMR) { + trans->dbg.unsupported_region_msk |= BIT(i); + break; + } + } +} + +static bool iwl_alive_fn(struct iwl_notif_wait_data *notif_wait, + struct iwl_rx_packet *pkt, void *data) +{ + unsigned int pkt_len = iwl_rx_packet_payload_len(pkt); + struct iwl_mld *mld = + container_of(notif_wait, struct iwl_mld, notif_wait); + struct iwl_trans *trans = mld->trans; + u32 version = iwl_fw_lookup_notif_ver(mld->fw, LEGACY_GROUP, + UCODE_ALIVE_NTFY, 0); + struct iwl_alive_ntf_v6 *palive; + bool *alive_valid = data; + struct iwl_umac_alive *umac; + struct iwl_lmac_alive *lmac1; + struct iwl_lmac_alive *lmac2 = NULL; + u32 lmac_error_event_table; + u32 umac_error_table; + u16 status; + + if (version < 6 || version > 7 || pkt_len != sizeof(*palive)) + return false; + + palive = (void *)pkt->data; + + iwl_mld_alive_imr_data(trans, &palive->imr); + + umac = &palive->umac_data; + lmac1 = &palive->lmac_data[0]; + lmac2 = &palive->lmac_data[1]; + status = le16_to_cpu(palive->status); + + trans->sku_id[0] = le32_to_cpu(palive->sku_id.data[0]); + trans->sku_id[1] = le32_to_cpu(palive->sku_id.data[1]); + trans->sku_id[2] = le32_to_cpu(palive->sku_id.data[2]); + + IWL_DEBUG_FW(mld, "Got sku_id: 0x0%x 0x0%x 0x0%x\n", + trans->sku_id[0], trans->sku_id[1], trans->sku_id[2]); + + lmac_error_event_table = + le32_to_cpu(lmac1->dbg_ptrs.error_event_table_ptr); + iwl_fw_lmac1_set_alive_err_table(trans, lmac_error_event_table); + + if (lmac2) + trans->dbg.lmac_error_event_table[1] = + le32_to_cpu(lmac2->dbg_ptrs.error_event_table_ptr); + + umac_error_table = le32_to_cpu(umac->dbg_ptrs.error_info_addr) & + ~FW_ADDR_CACHE_CONTROL; + + if (umac_error_table >= trans->cfg->min_umac_error_event_table) + iwl_fw_umac_set_alive_err_table(trans, umac_error_table); + else + IWL_ERR(mld, "Not valid error log pointer 0x%08X\n", + umac_error_table); + + *alive_valid = status == IWL_ALIVE_STATUS_OK; + + IWL_DEBUG_FW(mld, + "Alive ucode status 0x%04x revision 0x%01X 0x%01X\n", + status, lmac1->ver_type, lmac1->ver_subtype); + + if (lmac2) + IWL_DEBUG_FW(mld, "Alive ucode CDB\n"); + + IWL_DEBUG_FW(mld, + "UMAC version: Major - 0x%x, Minor - 0x%x\n", + le32_to_cpu(umac->umac_major), + le32_to_cpu(umac->umac_minor)); + + if (version >= 7) + IWL_DEBUG_FW(mld, "FW alive flags 0x%x\n", + le16_to_cpu(palive->flags)); + + iwl_fwrt_update_fw_versions(&mld->fwrt, lmac1, umac); + + return true; +} + +#define MLD_ALIVE_TIMEOUT (2 * HZ) +#define MLD_INIT_COMPLETE_TIMEOUT (2 * HZ) + +static void iwl_mld_print_alive_notif_timeout(struct iwl_mld *mld) +{ + struct iwl_trans *trans = mld->trans; + struct iwl_pc_data *pc_data; + u8 count; + + IWL_ERR(mld, + "SecBoot CPU1 Status: 0x%x, CPU2 Status: 0x%x\n", + iwl_read_umac_prph(trans, UMAG_SB_CPU_1_STATUS), + iwl_read_umac_prph(trans, + UMAG_SB_CPU_2_STATUS)); +#define IWL_FW_PRINT_REG_INFO(reg_name) \ + IWL_ERR(mld, #reg_name ": 0x%x\n", iwl_read_umac_prph(trans, reg_name)) + + IWL_FW_PRINT_REG_INFO(WFPM_LMAC1_PD_NOTIFICATION); + + IWL_FW_PRINT_REG_INFO(HPM_SECONDARY_DEVICE_STATE); + + /* print OTP info */ + IWL_FW_PRINT_REG_INFO(WFPM_MAC_OTP_CFG7_ADDR); + IWL_FW_PRINT_REG_INFO(WFPM_MAC_OTP_CFG7_DATA); +#undef IWL_FW_PRINT_REG_INFO + + pc_data = trans->dbg.pc_data; + for (count = 0; count < trans->dbg.num_pc; count++, pc_data++) + IWL_ERR(mld, "%s: 0x%x\n", pc_data->pc_name, + pc_data->pc_address); +} + +static int iwl_mld_load_fw_wait_alive(struct iwl_mld *mld) +{ + const struct fw_img *fw = + iwl_get_ucode_image(mld->fw, IWL_UCODE_REGULAR); + static const u16 alive_cmd[] = { UCODE_ALIVE_NTFY }; + struct iwl_notification_wait alive_wait; + bool alive_valid = false; + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + iwl_init_notification_wait(&mld->notif_wait, &alive_wait, + alive_cmd, ARRAY_SIZE(alive_cmd), + iwl_alive_fn, &alive_valid); + + iwl_dbg_tlv_time_point(&mld->fwrt, IWL_FW_INI_TIME_POINT_EARLY, NULL); + + ret = iwl_trans_start_fw(mld->trans, fw, true); + if (ret) { + iwl_remove_notification(&mld->notif_wait, &alive_wait); + return ret; + } + + ret = iwl_wait_notification(&mld->notif_wait, &alive_wait, + MLD_ALIVE_TIMEOUT); + + if (ret) { + if (ret == -ETIMEDOUT) + iwl_fw_dbg_error_collect(&mld->fwrt, + FW_DBG_TRIGGER_ALIVE_TIMEOUT); + iwl_mld_print_alive_notif_timeout(mld); + goto alive_failure; + } + + if (!alive_valid) { + IWL_ERR(mld, "Loaded firmware is not valid!\n"); + ret = -EIO; + goto alive_failure; + } + + iwl_trans_fw_alive(mld->trans, 0); + + return 0; + +alive_failure: + iwl_trans_stop_device(mld->trans); + return ret; +} + +int iwl_mld_run_fw_init_sequence(struct iwl_mld *mld) +{ + struct iwl_notification_wait init_wait; + struct iwl_init_extended_cfg_cmd init_cfg = {}; + static const u16 init_complete[] = { + INIT_COMPLETE_NOTIF, + }; + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + ret = iwl_mld_load_fw_wait_alive(mld); + if (ret) + return ret; + + mld->trans->step_urm = + !!(iwl_read_umac_prph(mld->trans, CNVI_PMU_STEP_FLOW) & + CNVI_PMU_STEP_FLOW_FORCE_URM); + + ret = iwl_pnvm_load(mld->trans, &mld->notif_wait, + &mld->fw->ucode_capa); + if (ret) { + IWL_ERR(mld, "Timeout waiting for PNVM load %d\n", ret); + goto init_failure; + } + + iwl_dbg_tlv_time_point(&mld->fwrt, IWL_FW_INI_TIME_POINT_AFTER_ALIVE, + NULL); + + iwl_init_notification_wait(&mld->notif_wait, + &init_wait, + init_complete, + ARRAY_SIZE(init_complete), + NULL, NULL); + + ret = iwl_mld_send_cmd_pdu(mld, + WIDE_ID(SYSTEM_GROUP, INIT_EXTENDED_CFG_CMD), + &init_cfg); + if (ret) { + IWL_ERR(mld, "Failed to send init config command: %d\n", ret); + iwl_remove_notification(&mld->notif_wait, &init_wait); + goto init_failure; + } + + ret = iwl_wait_notification(&mld->notif_wait, &init_wait, + MLD_INIT_COMPLETE_TIMEOUT); + if (ret) { + IWL_ERR(mld, "Failed to get INIT_COMPLETE %d\n", ret); + goto init_failure; + } + + if (!mld->nvm_data) { + mld->nvm_data = iwl_get_nvm(mld->trans, mld->fw, 0, 0); + if (IS_ERR(mld->nvm_data)) { + ret = PTR_ERR(mld->nvm_data); + mld->nvm_data = NULL; + IWL_ERR(mld, "Failed to read NVM: %d\n", ret); + goto init_failure; + } + } + + return 0; + +init_failure: + iwl_trans_stop_device(mld->trans); + return ret; +} + +int iwl_mld_load_fw(struct iwl_mld *mld) +{ + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + ret = iwl_trans_start_hw(mld->trans); + if (ret) + return ret; + + ret = iwl_mld_run_fw_init_sequence(mld); + if (ret) + return ret; + + mld->fw_status.running = true; + + return 0; +} + +void iwl_mld_stop_fw(struct iwl_mld *mld) +{ + lockdep_assert_wiphy(mld->wiphy); + + iwl_abort_notification_waits(&mld->notif_wait); + + iwl_fw_dbg_stop_sync(&mld->fwrt); + + iwl_trans_stop_device(mld->trans); + + mld->fw_status.running = false; +} + +static void iwl_mld_restart_disconnect_iter(void *data, u8 *mac, + struct ieee80211_vif *vif) +{ + if (vif->type == NL80211_IFTYPE_STATION) + ieee80211_hw_restart_disconnect(vif); +} + +void iwl_mld_send_recovery_cmd(struct iwl_mld *mld, u32 flags) +{ + u32 error_log_size = mld->fw->ucode_capa.error_log_size; + struct iwl_fw_error_recovery_cmd recovery_cmd = { + .flags = cpu_to_le32(flags), + }; + struct iwl_host_cmd cmd = { + .id = WIDE_ID(SYSTEM_GROUP, FW_ERROR_RECOVERY_CMD), + .flags = CMD_WANT_SKB, + .data = {&recovery_cmd, }, + .len = {sizeof(recovery_cmd), }, + }; + int ret; + + /* no error log was defined in TLV */ + if (!error_log_size) + return; + + if (flags & ERROR_RECOVERY_UPDATE_DB) { + /* no buf was allocated upon NIC error */ + if (!mld->error_recovery_buf) + return; + + cmd.data[1] = mld->error_recovery_buf; + cmd.len[1] = error_log_size; + cmd.dataflags[1] = IWL_HCMD_DFL_NOCOPY; + recovery_cmd.buf_size = cpu_to_le32(error_log_size); + } + + ret = iwl_mld_send_cmd(mld, &cmd); + + /* we no longer need the recovery buffer */ + kfree(mld->error_recovery_buf); + mld->error_recovery_buf = NULL; + + if (ret) { + IWL_ERR(mld, "Failed to send recovery cmd %d\n", ret); + return; + } + + if (flags & ERROR_RECOVERY_UPDATE_DB) { + struct iwl_rx_packet *pkt = cmd.resp_pkt; + u32 pkt_len = iwl_rx_packet_payload_len(pkt); + u32 resp; + + if (IWL_FW_CHECK(mld, pkt_len != sizeof(resp), + "Unexpected recovery cmd response size %u (expected %zu)\n", + pkt_len, sizeof(resp))) + goto out; + + resp = le32_to_cpup((__le32 *)cmd.resp_pkt->data); + if (!resp) + goto out; + + IWL_ERR(mld, + "Failed to send recovery cmd blob was invalid %d\n", + resp); + + ieee80211_iterate_interfaces(mld->hw, 0, + iwl_mld_restart_disconnect_iter, + NULL); + } + +out: + iwl_free_resp(&cmd); +} + +static int iwl_mld_config_fw(struct iwl_mld *mld) +{ + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + iwl_fw_disable_dbg_asserts(&mld->fwrt); + iwl_get_shared_mem_conf(&mld->fwrt); + + ret = iwl_mld_send_tx_ant_cfg(mld); + if (ret) + return ret; + + ret = iwl_mld_send_bt_init_conf(mld); + if (ret) + return ret; + + ret = iwl_set_soc_latency(&mld->fwrt); + if (ret) + return ret; + + iwl_mld_configure_lari(mld); + + ret = iwl_mld_config_temp_report_ths(mld); + if (ret) + return ret; + +#ifdef CONFIG_THERMAL + ret = iwl_mld_config_ctdp(mld, mld->cooling_dev.cur_state, + CTDP_CMD_OPERATION_START); + if (ret) + return ret; +#endif + + ret = iwl_configure_rxq(&mld->fwrt); + if (ret) + return ret; + + ret = iwl_mld_send_rss_cfg_cmd(mld); + if (ret) + return ret; + + ret = iwl_mld_config_scan(mld); + if (ret) + return ret; + + ret = iwl_mld_update_device_power(mld, false); + if (ret) + return ret; + + ret = iwl_mld_init_mcc(mld); + if (ret) + return ret; + + if (mld->fw_status.in_hw_restart) { + iwl_mld_send_recovery_cmd(mld, ERROR_RECOVERY_UPDATE_DB); + iwl_mld_time_sync_fw_config(mld); + } + + iwl_mld_led_config_fw(mld); + + ret = iwl_mld_init_ppag(mld); + if (ret) + return ret; + + ret = iwl_mld_init_sar(mld); + if (ret) + return ret; + + ret = iwl_mld_init_sgom(mld); + if (ret) + return ret; + + iwl_mld_init_tas(mld); + iwl_mld_init_uats(mld); + + return 0; +} + +int iwl_mld_start_fw(struct iwl_mld *mld) +{ + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + ret = iwl_mld_load_fw(mld); + if (IWL_FW_CHECK(mld, ret, "Failed to start firmware %d\n", ret)) { + iwl_fw_dbg_error_collect(&mld->fwrt, FW_DBG_TRIGGER_DRIVER); + goto error; + } + + IWL_DEBUG_INFO(mld, "uCode started.\n"); + + ret = iwl_mld_config_fw(mld); + if (ret) + goto error; + + return 0; + +error: + iwl_mld_stop_fw(mld); + return ret; +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/hcmd.h b/drivers/net/wireless/intel/iwlwifi/mld/hcmd.h new file mode 100644 index 0000000000000..773bc1b093921 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/hcmd.h @@ -0,0 +1,54 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024 Intel Corporation + */ +#ifndef __iwl_mld_hcmd_h__ +#define __iwl_mld_hcmd_h__ + +static inline int iwl_mld_send_cmd(struct iwl_mld *mld, struct iwl_host_cmd *cmd) +{ + /* No commands, including the d3 related commands, should be sent + * after entering d3 + */ + if (WARN_ON(mld->fw_status.in_d3)) + return -EIO; + + if (!(cmd->flags & CMD_ASYNC)) + lockdep_assert_wiphy(mld->wiphy); + + /* Devices that need to shutdown immediately on rfkill are not + * supported, so we can send all the cmds in rfkill + */ + cmd->flags |= CMD_SEND_IN_RFKILL; + + return iwl_trans_send_cmd(mld->trans, cmd); +} + +static inline int +__iwl_mld_send_cmd_with_flags_pdu(struct iwl_mld *mld, u32 id, + u32 flags, const void *data, u16 len) +{ + struct iwl_host_cmd cmd = { + .id = id, + .len = { data ? len : 0, }, + .data = { data, }, + .flags = flags, + }; + + return iwl_mld_send_cmd(mld, &cmd); +} + +#define _iwl_mld_send_cmd_with_flags_pdu(mld, id, flags, data, len, \ + ignored...) \ + __iwl_mld_send_cmd_with_flags_pdu(mld, id, flags, data, len) +#define iwl_mld_send_cmd_with_flags_pdu(mld, id, flags, data, len...) \ + _iwl_mld_send_cmd_with_flags_pdu(mld, id, flags, data, ##len, \ + sizeof(*(data))) + +#define iwl_mld_send_cmd_pdu(mld, id, ...) \ + iwl_mld_send_cmd_with_flags_pdu(mld, id, 0, __VA_ARGS__) + +#define iwl_mld_send_cmd_empty(mld, id) \ + iwl_mld_send_cmd_with_flags_pdu(mld, id, 0, NULL, 0) + +#endif /* __iwl_mld_hcmd_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/iface.c b/drivers/net/wireless/intel/iwlwifi/mld/iface.c new file mode 100644 index 0000000000000..691062b444619 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/iface.c @@ -0,0 +1,671 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024-2025 Intel Corporation + */ +#include + +#include "iface.h" +#include "hcmd.h" +#include "key.h" +#include "mlo.h" +#include "mac80211.h" + +#include "fw/api/context.h" +#include "fw/api/mac.h" +#include "fw/api/time-event.h" +#include "fw/api/datapath.h" + +/* Cleanup function for struct iwl_mld_vif, will be called in restart */ +void iwl_mld_cleanup_vif(void *data, u8 *mac, struct ieee80211_vif *vif) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mld *mld = mld_vif->mld; + struct iwl_mld_link *link; + + /* EMLSR is turned back on during recovery */ + vif->driver_flags &= ~IEEE80211_VIF_EML_ACTIVE; + + mld_vif->roc_activity = ROC_NUM_ACTIVITIES; + + for_each_mld_vif_valid_link(mld_vif, link) { + iwl_mld_cleanup_link(mld_vif->mld, link); + + /* Correctly allocated primary link in non-MLO mode */ + if (!ieee80211_vif_is_mld(vif) && + link_id == 0 && link == &mld_vif->deflink) + continue; + + if (vif->active_links & BIT(link_id)) + continue; + + /* Should not happen as link removal should always succeed */ + WARN_ON(1); + if (link != &mld_vif->deflink) + kfree_rcu(link, rcu_head); + RCU_INIT_POINTER(mld_vif->link[link_id], NULL); + } + + ieee80211_iter_keys(mld->hw, vif, iwl_mld_cleanup_keys_iter, NULL); + + CLEANUP_STRUCT(mld_vif); +} + +static int iwl_mld_send_mac_cmd(struct iwl_mld *mld, + struct iwl_mac_config_cmd *cmd) +{ + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + ret = iwl_mld_send_cmd_pdu(mld, + WIDE_ID(MAC_CONF_GROUP, MAC_CONFIG_CMD), + cmd); + if (ret) + IWL_ERR(mld, "Failed to send MAC_CONFIG_CMD ret = %d\n", ret); + + return ret; +} + +int iwl_mld_mac80211_iftype_to_fw(const struct ieee80211_vif *vif) +{ + switch (vif->type) { + case NL80211_IFTYPE_STATION: + return vif->p2p ? FW_MAC_TYPE_P2P_STA : FW_MAC_TYPE_BSS_STA; + case NL80211_IFTYPE_AP: + return FW_MAC_TYPE_GO; + case NL80211_IFTYPE_MONITOR: + return FW_MAC_TYPE_LISTENER; + case NL80211_IFTYPE_P2P_DEVICE: + return FW_MAC_TYPE_P2P_DEVICE; + case NL80211_IFTYPE_ADHOC: + return FW_MAC_TYPE_IBSS; + default: + WARN_ON_ONCE(1); + } + return FW_MAC_TYPE_BSS_STA; +} + +static bool iwl_mld_is_nic_ack_enabled(struct iwl_mld *mld, + struct ieee80211_vif *vif) +{ + const struct ieee80211_supported_band *sband; + const struct ieee80211_sta_he_cap *own_he_cap; + + lockdep_assert_wiphy(mld->wiphy); + + /* This capability is the same for all bands, + * so take it from one of them. + */ + sband = mld->hw->wiphy->bands[NL80211_BAND_2GHZ]; + own_he_cap = ieee80211_get_he_iftype_cap_vif(sband, vif); + + return own_he_cap && (own_he_cap->he_cap_elem.mac_cap_info[2] & + IEEE80211_HE_MAC_CAP2_ACK_EN); +} + +/* fill the common part for all interface types */ +static void iwl_mld_mac_cmd_fill_common(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct iwl_mac_config_cmd *cmd, + u32 action) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct ieee80211_bss_conf *link_conf; + unsigned int link_id; + + lockdep_assert_wiphy(mld->wiphy); + + cmd->id_and_color = cpu_to_le32(mld_vif->fw_id); + cmd->action = cpu_to_le32(action); + + cmd->mac_type = + cpu_to_le32(iwl_mld_mac80211_iftype_to_fw(vif)); + + memcpy(cmd->local_mld_addr, vif->addr, ETH_ALEN); + + if (iwlwifi_mod_params.disable_11ax) + return; + + cmd->nic_not_ack_enabled = + cpu_to_le32(!iwl_mld_is_nic_ack_enabled(mld, vif)); + + /* If we have MLO enabled, then the firmware needs to enable + * address translation for the station(s) we add. That depends + * on having EHT enabled in firmware, which in turn depends on + * mac80211 in the code below. + * However, mac80211 doesn't enable HE/EHT until it has parsed + * the association response successfully, so just skip all that + * and enable both when we have MLO. + */ + if (ieee80211_vif_is_mld(vif)) { + if (vif->type == NL80211_IFTYPE_AP) + cmd->he_ap_support = cpu_to_le16(1); + else + cmd->he_support = cpu_to_le16(1); + + cmd->eht_support = cpu_to_le32(1); + return; + } + + for_each_vif_active_link(vif, link_conf, link_id) { + if (!link_conf->he_support) + continue; + + if (vif->type == NL80211_IFTYPE_AP) + cmd->he_ap_support = cpu_to_le16(1); + else + cmd->he_support = cpu_to_le16(1); + + /* EHT, if supported, was already set above */ + break; + } +} + +static void iwl_mld_fill_mac_cmd_sta(struct iwl_mld *mld, + struct ieee80211_vif *vif, u32 action, + struct iwl_mac_config_cmd *cmd) +{ + struct ieee80211_bss_conf *link; + u32 twt_policy = 0; + int link_id; + + lockdep_assert_wiphy(mld->wiphy); + + WARN_ON(vif->type != NL80211_IFTYPE_STATION); + + /* We always want to hear MCAST frames, if we're not authorized yet, + * we'll drop them. + */ + cmd->filter_flags |= cpu_to_le32(MAC_CFG_FILTER_ACCEPT_GRP); + + /* Adding a MAC ctxt with is_assoc set is not allowed in fw + * (and shouldn't happen) + */ + if (vif->cfg.assoc && action != FW_CTXT_ACTION_ADD) { + cmd->client.is_assoc = 1; + + if (!iwl_mld_vif_from_mac80211(vif)->authorized) + cmd->client.data_policy |= + cpu_to_le16(COEX_HIGH_PRIORITY_ENABLE); + } else { + /* Allow beacons to pass through as long as we are not + * associated + */ + cmd->filter_flags |= cpu_to_le32(MAC_CFG_FILTER_ACCEPT_BEACON); + } + + cmd->client.assoc_id = cpu_to_le16(vif->cfg.aid); + + if (ieee80211_vif_is_mld(vif)) { + u16 esr_transition_timeout = + u16_get_bits(vif->cfg.eml_cap, + IEEE80211_EML_CAP_TRANSITION_TIMEOUT); + + cmd->client.esr_transition_timeout = + min_t(u16, IEEE80211_EML_CAP_TRANSITION_TIMEOUT_128TU, + esr_transition_timeout); + cmd->client.medium_sync_delay = + cpu_to_le16(vif->cfg.eml_med_sync_delay); + } + + for_each_vif_active_link(vif, link, link_id) { + if (!link->he_support) + continue; + + if (link->twt_requester) + twt_policy |= TWT_SUPPORTED; + if (link->twt_protected) + twt_policy |= PROTECTED_TWT_SUPPORTED; + if (link->twt_broadcast) + twt_policy |= BROADCAST_TWT_SUPPORTED; + } + + if (!iwlwifi_mod_params.disable_11ax) + cmd->client.data_policy |= cpu_to_le16(twt_policy); + + if (vif->probe_req_reg && vif->cfg.assoc && vif->p2p) + cmd->filter_flags |= + cpu_to_le32(MAC_CFG_FILTER_ACCEPT_PROBE_REQ); + + if (vif->p2p) + cmd->client.ctwin = + cpu_to_le32(vif->bss_conf.p2p_noa_attr.oppps_ctwindow & + IEEE80211_P2P_OPPPS_CTWINDOW_MASK); +} + +static void iwl_mld_fill_mac_cmd_ap(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct iwl_mac_config_cmd *cmd) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + + lockdep_assert_wiphy(mld->wiphy); + + WARN_ON(vif->type != NL80211_IFTYPE_AP); + + cmd->filter_flags |= cpu_to_le32(MAC_CFG_FILTER_ACCEPT_PROBE_REQ); + + /* in AP mode, pass beacons from other APs (needed for ht protection). + * When there're no any associated station, which means that we are not + * TXing anyway, don't ask FW to pass beacons to prevent unnecessary + * wake-ups. + */ + if (mld_vif->num_associated_stas) + cmd->filter_flags |= cpu_to_le32(MAC_CFG_FILTER_ACCEPT_BEACON); +} + +static void iwl_mld_go_iterator(void *_data, u8 *mac, struct ieee80211_vif *vif) +{ + bool *go_active = _data; + + if (ieee80211_vif_type_p2p(vif) == NL80211_IFTYPE_P2P_GO && + iwl_mld_vif_from_mac80211(vif)->ap_ibss_active) + *go_active = true; +} + +static bool iwl_mld_p2p_dev_has_extended_disc(struct iwl_mld *mld) +{ + bool go_active = false; + + /* This flag should be set to true when the P2P Device is + * discoverable and there is at least a P2P GO. Setting + * this flag will allow the P2P Device to be discoverable on other + * channels in addition to its listen channel. + * Note that this flag should not be set in other cases as it opens the + * Rx filters on all MAC and increases the number of interrupts. + */ + ieee80211_iterate_active_interfaces(mld->hw, + IEEE80211_IFACE_ITER_RESUME_ALL, + iwl_mld_go_iterator, &go_active); + + return go_active; +} + +static void iwl_mld_fill_mac_cmd_p2p_dev(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct iwl_mac_config_cmd *cmd) +{ + bool ext_disc = iwl_mld_p2p_dev_has_extended_disc(mld); + + lockdep_assert_wiphy(mld->wiphy); + + /* Override the filter flags to accept all management frames. This is + * needed to support both P2P device discovery using probe requests and + * P2P service discovery using action frames + */ + cmd->filter_flags = cpu_to_le32(MAC_CFG_FILTER_ACCEPT_CONTROL_AND_MGMT); + + if (ext_disc) + cmd->p2p_dev.is_disc_extended = cpu_to_le32(1); +} + +static void iwl_mld_fill_mac_cmd_ibss(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct iwl_mac_config_cmd *cmd) +{ + lockdep_assert_wiphy(mld->wiphy); + + WARN_ON(vif->type != NL80211_IFTYPE_ADHOC); + + cmd->filter_flags |= cpu_to_le32(MAC_CFG_FILTER_ACCEPT_BEACON | + MAC_CFG_FILTER_ACCEPT_PROBE_REQ | + MAC_CFG_FILTER_ACCEPT_GRP); +} + +static int +iwl_mld_rm_mac_from_fw(struct iwl_mld *mld, struct ieee80211_vif *vif) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mac_config_cmd cmd = { + .action = cpu_to_le32(FW_CTXT_ACTION_REMOVE), + .id_and_color = cpu_to_le32(mld_vif->fw_id), + }; + + return iwl_mld_send_mac_cmd(mld, &cmd); +} + +int iwl_mld_mac_fw_action(struct iwl_mld *mld, struct ieee80211_vif *vif, + u32 action) +{ + struct iwl_mac_config_cmd cmd = {}; + + lockdep_assert_wiphy(mld->wiphy); + + if (action == FW_CTXT_ACTION_REMOVE) + return iwl_mld_rm_mac_from_fw(mld, vif); + + iwl_mld_mac_cmd_fill_common(mld, vif, &cmd, action); + + switch (vif->type) { + case NL80211_IFTYPE_STATION: + iwl_mld_fill_mac_cmd_sta(mld, vif, action, &cmd); + break; + case NL80211_IFTYPE_AP: + iwl_mld_fill_mac_cmd_ap(mld, vif, &cmd); + break; + case NL80211_IFTYPE_MONITOR: + cmd.filter_flags = + cpu_to_le32(MAC_CFG_FILTER_PROMISC | + MAC_CFG_FILTER_ACCEPT_CONTROL_AND_MGMT | + MAC_CFG_FILTER_ACCEPT_BEACON | + MAC_CFG_FILTER_ACCEPT_PROBE_REQ | + MAC_CFG_FILTER_ACCEPT_GRP); + break; + case NL80211_IFTYPE_P2P_DEVICE: + iwl_mld_fill_mac_cmd_p2p_dev(mld, vif, &cmd); + break; + case NL80211_IFTYPE_ADHOC: + iwl_mld_fill_mac_cmd_ibss(mld, vif, &cmd); + break; + default: + WARN(1, "not supported yet\n"); + return -EOPNOTSUPP; + } + + return iwl_mld_send_mac_cmd(mld, &cmd); +} + +IWL_MLD_ALLOC_FN(vif, vif) + +/* Constructor function for struct iwl_mld_vif */ +static int +iwl_mld_init_vif(struct iwl_mld *mld, struct ieee80211_vif *vif) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + mld_vif->mld = mld; + mld_vif->roc_activity = ROC_NUM_ACTIVITIES; + + ret = iwl_mld_allocate_vif_fw_id(mld, &mld_vif->fw_id, vif); + if (ret) + return ret; + + if (!mld->fw_status.in_hw_restart) { + wiphy_work_init(&mld_vif->emlsr.unblock_tpt_wk, + iwl_mld_emlsr_unblock_tpt_wk); + wiphy_delayed_work_init(&mld_vif->emlsr.check_tpt_wk, + iwl_mld_emlsr_check_tpt); + wiphy_delayed_work_init(&mld_vif->emlsr.prevent_done_wk, + iwl_mld_emlsr_prevent_done_wk); + wiphy_delayed_work_init(&mld_vif->emlsr.tmp_non_bss_done_wk, + iwl_mld_emlsr_tmp_non_bss_done_wk); + } + + return 0; +} + +int iwl_mld_add_vif(struct iwl_mld *mld, struct ieee80211_vif *vif) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + ret = iwl_mld_init_vif(mld, vif); + if (ret) + return ret; + + ret = iwl_mld_mac_fw_action(mld, vif, FW_CTXT_ACTION_ADD); + if (ret) + RCU_INIT_POINTER(mld->fw_id_to_vif[mld_vif->fw_id], NULL); + + return ret; +} + +int iwl_mld_rm_vif(struct iwl_mld *mld, struct ieee80211_vif *vif) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + ret = iwl_mld_mac_fw_action(mld, vif, FW_CTXT_ACTION_REMOVE); + + if (WARN_ON(mld_vif->fw_id >= ARRAY_SIZE(mld->fw_id_to_vif))) + return -EINVAL; + + RCU_INIT_POINTER(mld->fw_id_to_vif[mld_vif->fw_id], NULL); + + iwl_mld_cancel_notifications_of_object(mld, IWL_MLD_OBJECT_TYPE_VIF, + mld_vif->fw_id); + + return ret; +} + +void iwl_mld_set_vif_associated(struct iwl_mld *mld, + struct ieee80211_vif *vif) +{ + struct ieee80211_bss_conf *link; + unsigned int link_id; + + for_each_vif_active_link(vif, link, link_id) { + if (iwl_mld_link_set_associated(mld, vif, link)) + IWL_ERR(mld, "failed to update link %d\n", link_id); + } + + iwl_mld_recalc_multicast_filter(mld); +} + +static void iwl_mld_get_fw_id_bss_bitmap_iter(void *_data, u8 *mac, + struct ieee80211_vif *vif) +{ + u8 *fw_id_bitmap = _data; + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + + if (ieee80211_vif_type_p2p(vif) != NL80211_IFTYPE_STATION) + return; + + *fw_id_bitmap |= BIT(mld_vif->fw_id); +} + +u8 iwl_mld_get_fw_bss_vifs_ids(struct iwl_mld *mld) +{ + u8 fw_id_bitmap = 0; + + ieee80211_iterate_interfaces(mld->hw, + IEEE80211_IFACE_SKIP_SDATA_NOT_IN_DRIVER, + iwl_mld_get_fw_id_bss_bitmap_iter, + &fw_id_bitmap); + + return fw_id_bitmap; +} + +void iwl_mld_handle_probe_resp_data_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + const struct iwl_probe_resp_data_notif *notif = (void *)pkt->data; + struct iwl_probe_resp_data *old_data, *new_data; + struct ieee80211_vif *vif; + struct iwl_mld_link *mld_link; + + IWL_DEBUG_INFO(mld, "Probe response data notif: noa %d, csa %d\n", + notif->noa_active, notif->csa_counter); + + if (IWL_FW_CHECK(mld, le32_to_cpu(notif->mac_id) >= + ARRAY_SIZE(mld->fw_id_to_vif), + "mac id is invalid: %d\n", + le32_to_cpu(notif->mac_id))) + return; + + vif = wiphy_dereference(mld->wiphy, + mld->fw_id_to_vif[le32_to_cpu(notif->mac_id)]); + + /* the firmware gives us the mac_id (and not the link_id), mac80211 + * gets a vif and not a link, bottom line, this flow is not MLD ready + * yet. + */ + if (WARN_ON(!vif) || ieee80211_vif_is_mld(vif)) + return; + + if (notif->csa_counter != IWL_PROBE_RESP_DATA_NO_CSA && + notif->csa_counter >= 1) + ieee80211_beacon_set_cntdwn(vif, notif->csa_counter); + + if (!vif->p2p) + return; + + mld_link = &iwl_mld_vif_from_mac80211(vif)->deflink; + + new_data = kzalloc(sizeof(*new_data), GFP_KERNEL); + if (!new_data) + return; + + memcpy(&new_data->notif, notif, sizeof(new_data->notif)); + + /* noa_attr contains 1 reserved byte, need to substruct it */ + new_data->noa_len = sizeof(struct ieee80211_vendor_ie) + + sizeof(new_data->notif.noa_attr) - 1; + + /* + * If it's a one time NoA, only one descriptor is needed, + * adjust the length according to len_low. + */ + if (new_data->notif.noa_attr.len_low == + sizeof(struct ieee80211_p2p_noa_desc) + 2) + new_data->noa_len -= sizeof(struct ieee80211_p2p_noa_desc); + + old_data = wiphy_dereference(mld->wiphy, mld_link->probe_resp_data); + rcu_assign_pointer(mld_link->probe_resp_data, new_data); + + if (old_data) + kfree_rcu(old_data, rcu_head); +} + +void iwl_mld_handle_uapsd_misbehaving_ap_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + struct iwl_uapsd_misbehaving_ap_notif *notif = (void *)pkt->data; + struct ieee80211_vif *vif; + + if (IWL_FW_CHECK(mld, notif->mac_id >= ARRAY_SIZE(mld->fw_id_to_vif), + "mac id is invalid: %d\n", notif->mac_id)) + return; + + vif = wiphy_dereference(mld->wiphy, mld->fw_id_to_vif[notif->mac_id]); + + if (WARN_ON(!vif) || ieee80211_vif_is_mld(vif)) + return; + + IWL_WARN(mld, "uapsd misbehaving AP: %pM\n", vif->bss_conf.bssid); +} + +void iwl_mld_handle_datapath_monitor_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + struct iwl_datapath_monitor_notif *notif = (void *)pkt->data; + struct ieee80211_bss_conf *link; + struct ieee80211_supported_band *sband; + const struct ieee80211_sta_he_cap *he_cap; + struct ieee80211_vif *vif; + struct iwl_mld_vif *mld_vif; + + if (notif->type != cpu_to_le32(IWL_DP_MON_NOTIF_TYPE_EXT_CCA)) + return; + + link = iwl_mld_fw_id_to_link_conf(mld, notif->link_id); + if (WARN_ON(!link)) + return; + + vif = link->vif; + if (WARN_ON(!vif) || vif->type != NL80211_IFTYPE_STATION || + !vif->cfg.assoc) + return; + + if (!link->chanreq.oper.chan || + link->chanreq.oper.chan->band != NL80211_BAND_2GHZ || + link->chanreq.oper.width < NL80211_CHAN_WIDTH_40) + return; + + mld_vif = iwl_mld_vif_from_mac80211(vif); + + /* this shouldn't happen *again*, ignore it */ + if (mld_vif->cca_40mhz_workaround != CCA_40_MHZ_WA_NONE) + return; + + mld_vif->cca_40mhz_workaround = CCA_40_MHZ_WA_RECONNECT; + + /* + * This capability manipulation isn't really ideal, but it's the + * easiest choice - otherwise we'd have to do some major changes + * in mac80211 to support this, which isn't worth it. This does + * mean that userspace may have outdated information, but that's + * actually not an issue at all. + */ + sband = mld->wiphy->bands[NL80211_BAND_2GHZ]; + + WARN_ON(!sband->ht_cap.ht_supported); + WARN_ON(!(sband->ht_cap.cap & IEEE80211_HT_CAP_SUP_WIDTH_20_40)); + sband->ht_cap.cap &= ~IEEE80211_HT_CAP_SUP_WIDTH_20_40; + + he_cap = ieee80211_get_he_iftype_cap_vif(sband, vif); + + if (he_cap) { + /* we know that ours is writable */ + struct ieee80211_sta_he_cap *he = (void *)(uintptr_t)he_cap; + + WARN_ON(!he->has_he); + WARN_ON(!(he->he_cap_elem.phy_cap_info[0] & + IEEE80211_HE_PHY_CAP0_CHANNEL_WIDTH_SET_40MHZ_IN_2G)); + he->he_cap_elem.phy_cap_info[0] &= + ~IEEE80211_HE_PHY_CAP0_CHANNEL_WIDTH_SET_40MHZ_IN_2G; + } + + ieee80211_disconnect(vif, true); +} + +void iwl_mld_reset_cca_40mhz_workaround(struct iwl_mld *mld, + struct ieee80211_vif *vif) +{ + struct ieee80211_supported_band *sband; + const struct ieee80211_sta_he_cap *he_cap; + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + + if (vif->type != NL80211_IFTYPE_STATION) + return; + + if (mld_vif->cca_40mhz_workaround == CCA_40_MHZ_WA_NONE) + return; + + /* Now we are just reconnecting with the new capabilities, + * but remember to reset the capabilities when we disconnect for real + */ + if (mld_vif->cca_40mhz_workaround == CCA_40_MHZ_WA_RECONNECT) { + mld_vif->cca_40mhz_workaround = CCA_40_MHZ_WA_RESET; + return; + } + + /* Now cca_40mhz_workaround == CCA_40_MHZ_WA_RESET */ + + sband = mld->wiphy->bands[NL80211_BAND_2GHZ]; + + sband->ht_cap.cap |= IEEE80211_HT_CAP_SUP_WIDTH_20_40; + + he_cap = ieee80211_get_he_iftype_cap_vif(sband, vif); + + if (he_cap) { + /* we know that ours is writable */ + struct ieee80211_sta_he_cap *he = (void *)(uintptr_t)he_cap; + + he->he_cap_elem.phy_cap_info[0] |= + IEEE80211_HE_PHY_CAP0_CHANNEL_WIDTH_SET_40MHZ_IN_2G; + } + + mld_vif->cca_40mhz_workaround = CCA_40_MHZ_WA_NONE; +} + +struct ieee80211_vif *iwl_mld_get_bss_vif(struct iwl_mld *mld) +{ + unsigned long fw_id_bitmap = iwl_mld_get_fw_bss_vifs_ids(mld); + int fw_id; + + if (hweight8(fw_id_bitmap) != 1) + return NULL; + + fw_id = __ffs(fw_id_bitmap); + + return wiphy_dereference(mld->wiphy, + mld->fw_id_to_vif[fw_id]); +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/iface.h b/drivers/net/wireless/intel/iwlwifi/mld/iface.h new file mode 100644 index 0000000000000..2c928113f6802 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/iface.h @@ -0,0 +1,234 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024-2025 Intel Corporation + */ +#ifndef __iwl_mld_iface_h__ +#define __iwl_mld_iface_h__ + +#include + +#include "link.h" +#include "session-protect.h" +#include "d3.h" + +enum iwl_mld_cca_40mhz_wa_status { + CCA_40_MHZ_WA_NONE, + CCA_40_MHZ_WA_RESET, + CCA_40_MHZ_WA_RECONNECT, +}; + +/** + * enum iwl_mld_emlsr_blocked - defines reasons for which EMLSR is blocked + * + * These blocks are applied/stored per-VIF. + * + * @IWL_MLD_EMLSR_BLOCKED_PREVENTION: Prevent repeated EMLSR enter/exit + * @IWL_MLD_EMLSR_BLOCKED_WOWLAN: WOWLAN is preventing EMLSR + * @IWL_MLD_EMLSR_BLOCKED_FW: FW did not recommend MLO + * @IWL_MLD_EMLSR_BLOCKED_ROC: remain-on-channel is preventing EMLSR + * @IWL_MLD_EMLSR_BLOCKED_NON_BSS: An active non-BSS interface's link is + * preventing EMLSR + * @IWL_MLD_EMLSR_BLOCKED_TMP_NON_BSS: An expected active non-BSS interface's + * link is preventing EMLSR. This is a temporary blocking that is set when + * there is an indication that a non-BSS interface is to be added. + * @IWL_MLD_EMLSR_BLOCKED_TPT: throughput is too low to make EMLSR worthwhile + */ +enum iwl_mld_emlsr_blocked { + IWL_MLD_EMLSR_BLOCKED_PREVENTION = 0x1, + IWL_MLD_EMLSR_BLOCKED_WOWLAN = 0x2, + IWL_MLD_EMLSR_BLOCKED_FW = 0x4, + IWL_MLD_EMLSR_BLOCKED_ROC = 0x8, + IWL_MLD_EMLSR_BLOCKED_NON_BSS = 0x10, + IWL_MLD_EMLSR_BLOCKED_TMP_NON_BSS = 0x20, + IWL_MLD_EMLSR_BLOCKED_TPT = 0x40, +}; + +/** + * enum iwl_mld_emlsr_exit - defines reasons for exiting EMLSR + * + * Reasons to exit EMLSR may be either link specific or even specific to a + * combination of links. + * + * @IWL_MLD_EMLSR_EXIT_BLOCK: Exit due to a block reason being set + * @IWL_MLD_EMLSR_EXIT_MISSED_BEACON: Exit due to missed beacons + * @IWL_MLD_EMLSR_EXIT_FAIL_ENTRY: FW failed to enter EMLSR + * @IWL_MLD_EMLSR_EXIT_CSA: EMLSR prevented due to channel switch on link + * @IWL_MLD_EMLSR_EXIT_EQUAL_BAND: EMLSR prevented as both links share the band + * @IWL_MLD_EMLSR_EXIT_BANDWIDTH: Bandwidths of primary and secondary links are + * not equal + * @IWL_MLD_EMLSR_EXIT_LOW_RSSI: Link RSSI is unsuitable for EMLSR + * @IWL_MLD_EMLSR_EXIT_LINK_USAGE: Exit EMLSR due to low TPT on secondary link + * @IWL_MLD_EMLSR_EXIT_BT_COEX: Exit EMLSR due to BT coexistence + * @IWL_MLD_EMLSR_EXIT_CHAN_LOAD: Exit EMLSR because the primary channel is not + * loaded enough to justify EMLSR. + * @IWL_MLD_EMLSR_EXIT_RFI: Exit EMLSR due to RFI + */ +enum iwl_mld_emlsr_exit { + IWL_MLD_EMLSR_EXIT_BLOCK = 0x1, + IWL_MLD_EMLSR_EXIT_MISSED_BEACON = 0x2, + IWL_MLD_EMLSR_EXIT_FAIL_ENTRY = 0x4, + IWL_MLD_EMLSR_EXIT_CSA = 0x8, + IWL_MLD_EMLSR_EXIT_EQUAL_BAND = 0x10, + IWL_MLD_EMLSR_EXIT_BANDWIDTH = 0x20, + IWL_MLD_EMLSR_EXIT_LOW_RSSI = 0x40, + IWL_MLD_EMLSR_EXIT_LINK_USAGE = 0x80, + IWL_MLD_EMLSR_EXIT_BT_COEX = 0x100, + IWL_MLD_EMLSR_EXIT_CHAN_LOAD = 0x200, + IWL_MLD_EMLSR_EXIT_RFI = 0x400, +}; + +/** + * struct iwl_mld_emlsr - per-VIF data about EMLSR operation + * + * @primary: The current primary link + * @selected_primary: Primary link as selected during the last link selection + * @selected_links: Links as selected during the last link selection + * @blocked_reasons: Reasons preventing EMLSR from being enabled + * @last_exit_reason: Reason for the last EMLSR exit + * @last_exit_ts: Time of the last EMLSR exit (if @last_exit_reason is non-zero) + * @exit_repeat_count: Number of times EMLSR was exited for the same reason + * @unblock_tpt_wk: Unblock EMLSR because the throughput limit was reached + * @check_tpt_wk: a worker to check if IWL_MLD_EMLSR_BLOCKED_TPT should be + * added, for example if there is no longer enough traffic. + * @prevent_done_wk: Worker to remove %IWL_MLD_EMLSR_BLOCKED_PREVENTION + * @tmp_non_bss_done_wk: Worker to remove %IWL_MLD_EMLSR_BLOCKED_TMP_NON_BSS + */ +struct iwl_mld_emlsr { + struct_group(zeroed_on_not_authorized, + u8 primary; + + u8 selected_primary; + u16 selected_links; + + enum iwl_mld_emlsr_blocked blocked_reasons; + + enum iwl_mld_emlsr_exit last_exit_reason; + unsigned long last_exit_ts; + u8 exit_repeat_count; + ); + + struct wiphy_work unblock_tpt_wk; + struct wiphy_delayed_work check_tpt_wk; + + struct wiphy_delayed_work prevent_done_wk; + struct wiphy_delayed_work tmp_non_bss_done_wk; +}; + +/** + * struct iwl_mld_vif - virtual interface (MAC context) configuration parameters + * + * @fw_id: fw id of the mac context. + * @session_protect: session protection parameters + * @ap_sta: pointer to AP sta, for easier access to it. + * Relevant only for STA vifs. + * @authorized: indicates the AP station was set to authorized + * @bigtks: BIGTKs of the AP, for beacon protection. + * Only valid for STA. (FIXME: needs to be per link) + * @num_associated_stas: number of associated STAs. Relevant only for AP mode. + * @ap_ibss_active: whether the AP/IBSS was started + * @roc_activity: the id of the roc_activity running. Relevant for p2p device + * only. Set to %ROC_NUM_ACTIVITIES when not in use. + * @cca_40mhz_workaround: When we are connected in 2.4 GHz and 40 MHz, and the + * environment is too loaded, we work around this by reconnecting to the + * same AP with 20 MHz. This manages the status of the workaround. + * @beacon_inject_active: indicates an active debugfs beacon ie injection + * @low_latency_causes: bit flags, indicating the causes for low-latency, + * see @iwl_mld_low_latency_cause. + * @mld: pointer to the mld structure. + * @deflink: default link data, for use in non-MLO, + * @link: reference to link data for each valid link, for use in MLO. + * @emlsr: information related to EMLSR + * @wowlan_data: data used by the wowlan suspend flow + * @use_ps_poll: use ps_poll frames + * @disable_bf: disable beacon filter + * @dbgfs_slink: debugfs symlink for this interface + */ +struct iwl_mld_vif { + /* Add here fields that need clean up on restart */ + struct_group(zeroed_on_hw_restart, + u8 fw_id; + struct iwl_mld_session_protect session_protect; + struct ieee80211_sta *ap_sta; + bool authorized; + struct ieee80211_key_conf __rcu *bigtks[2]; + u8 num_associated_stas; + bool ap_ibss_active; + u32 roc_activity; + enum iwl_mld_cca_40mhz_wa_status cca_40mhz_workaround; +#ifdef CONFIG_IWLWIFI_DEBUGFS + bool beacon_inject_active; +#endif + u8 low_latency_causes; + ); + /* And here fields that survive a fw restart */ + struct iwl_mld *mld; + struct iwl_mld_link deflink; + struct iwl_mld_link __rcu *link[IEEE80211_MLD_MAX_NUM_LINKS]; + + struct iwl_mld_emlsr emlsr; + +#if CONFIG_PM_SLEEP + struct iwl_mld_wowlan_data wowlan_data; +#endif +#ifdef CONFIG_IWLWIFI_DEBUGFS + bool use_ps_poll; + bool disable_bf; + struct dentry *dbgfs_slink; +#endif +}; + +static inline struct iwl_mld_vif * +iwl_mld_vif_from_mac80211(struct ieee80211_vif *vif) +{ + return (void *)vif->drv_priv; +} + +#define iwl_mld_link_dereference_check(mld_vif, link_id) \ + rcu_dereference_check((mld_vif)->link[link_id], \ + lockdep_is_held(&mld_vif->mld->wiphy->mtx)) + +#define for_each_mld_vif_valid_link(mld_vif, mld_link) \ + for (int link_id = 0; link_id < ARRAY_SIZE((mld_vif)->link); \ + link_id++) \ + if ((mld_link = iwl_mld_link_dereference_check(mld_vif, link_id))) + +/* Retrieve pointer to mld link from mac80211 structures */ +static inline struct iwl_mld_link * +iwl_mld_link_from_mac80211(struct ieee80211_bss_conf *bss_conf) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(bss_conf->vif); + + return iwl_mld_link_dereference_check(mld_vif, bss_conf->link_id); +} + +int iwl_mld_mac80211_iftype_to_fw(const struct ieee80211_vif *vif); + +/* Cleanup function for struct iwl_mld_vif, will be called in restart */ +void iwl_mld_cleanup_vif(void *data, u8 *mac, struct ieee80211_vif *vif); +int iwl_mld_mac_fw_action(struct iwl_mld *mld, struct ieee80211_vif *vif, + u32 action); +int iwl_mld_add_vif(struct iwl_mld *mld, struct ieee80211_vif *vif); +int iwl_mld_rm_vif(struct iwl_mld *mld, struct ieee80211_vif *vif); +void iwl_mld_set_vif_associated(struct iwl_mld *mld, + struct ieee80211_vif *vif); +u8 iwl_mld_get_fw_bss_vifs_ids(struct iwl_mld *mld); +void iwl_mld_handle_probe_resp_data_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt); + +void iwl_mld_handle_datapath_monitor_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt); + +void iwl_mld_handle_uapsd_misbehaving_ap_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt); + +void iwl_mld_reset_cca_40mhz_workaround(struct iwl_mld *mld, + struct ieee80211_vif *vif); + +static inline bool iwl_mld_vif_low_latency(const struct iwl_mld_vif *mld_vif) +{ + return !!mld_vif->low_latency_causes; +} + +struct ieee80211_vif *iwl_mld_get_bss_vif(struct iwl_mld *mld); + +#endif /* __iwl_mld_iface_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/key.c b/drivers/net/wireless/intel/iwlwifi/mld/key.c new file mode 100644 index 0000000000000..0eff13e5ffd57 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/key.c @@ -0,0 +1,358 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024 Intel Corporation + */ +#include "key.h" +#include "iface.h" +#include "sta.h" +#include "fw/api/datapath.h" + +static u32 iwl_mld_get_key_flags(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + struct ieee80211_key_conf *key) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + bool pairwise = key->flags & IEEE80211_KEY_FLAG_PAIRWISE; + bool igtk = key->keyidx == 4 || key->keyidx == 5; + u32 flags = 0; + + if (!pairwise) + flags |= IWL_SEC_KEY_FLAG_MCAST_KEY; + + switch (key->cipher) { + case WLAN_CIPHER_SUITE_TKIP: + flags |= IWL_SEC_KEY_FLAG_CIPHER_TKIP; + break; + case WLAN_CIPHER_SUITE_AES_CMAC: + case WLAN_CIPHER_SUITE_CCMP: + flags |= IWL_SEC_KEY_FLAG_CIPHER_CCMP; + break; + case WLAN_CIPHER_SUITE_GCMP_256: + case WLAN_CIPHER_SUITE_BIP_GMAC_256: + flags |= IWL_SEC_KEY_FLAG_KEY_SIZE; + fallthrough; + case WLAN_CIPHER_SUITE_GCMP: + case WLAN_CIPHER_SUITE_BIP_GMAC_128: + flags |= IWL_SEC_KEY_FLAG_CIPHER_GCMP; + break; + } + + if (!sta && vif->type == NL80211_IFTYPE_STATION) + sta = mld_vif->ap_sta; + + /* If we are installing an iGTK (in AP or STA mode), we need to tell + * the firmware this key will en/decrypt MGMT frames. + * Same goes if we are installing a pairwise key for an MFP station. + * In case we're installing a groupwise key (which is not an iGTK), + * then, we will not use this key for MGMT frames. + */ + if ((sta && sta->mfp && pairwise) || igtk) + flags |= IWL_SEC_KEY_FLAG_MFP; + + if (key->flags & IEEE80211_KEY_FLAG_SPP_AMSDU) + flags |= IWL_SEC_KEY_FLAG_SPP_AMSDU; + + return flags; +} + +static u32 iwl_mld_get_key_sta_mask(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + struct ieee80211_key_conf *key) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct ieee80211_link_sta *link_sta; + int sta_id; + + lockdep_assert_wiphy(mld->wiphy); + + /* AP group keys are per link and should be on the mcast/bcast STA */ + if (vif->type == NL80211_IFTYPE_AP && + !(key->flags & IEEE80211_KEY_FLAG_PAIRWISE)) { + struct iwl_mld_link *link = NULL; + + if (key->link_id >= 0) + link = iwl_mld_link_dereference_check(mld_vif, + key->link_id); + + if (WARN_ON(!link)) + return 0; + + /* In this stage we should have both the bcast and mcast STAs */ + if (WARN_ON(link->bcast_sta.sta_id == IWL_INVALID_STA || + link->mcast_sta.sta_id == IWL_INVALID_STA)) + return 0; + + /* IGTK/BIGTK to bcast STA */ + if (key->keyidx >= 4) + return BIT(link->bcast_sta.sta_id); + + /* GTK for data to mcast STA */ + return BIT(link->mcast_sta.sta_id); + } + + /* for client mode use the AP STA also for group keys */ + if (!sta && vif->type == NL80211_IFTYPE_STATION) + sta = mld_vif->ap_sta; + + /* STA should be non-NULL now */ + if (WARN_ON(!sta)) + return 0; + + /* Key is not per-link, get the full sta mask */ + if (key->link_id < 0) + return iwl_mld_fw_sta_id_mask(mld, sta); + + /* The link_sta shouldn't be NULL now, but this is checked in + * iwl_mld_fw_sta_id_mask + */ + link_sta = link_sta_dereference_check(sta, key->link_id); + + sta_id = iwl_mld_fw_sta_id_from_link_sta(mld, link_sta); + if (sta_id < 0) + return 0; + + return BIT(sta_id); +} + +static int iwl_mld_add_key_to_fw(struct iwl_mld *mld, u32 sta_mask, + u32 key_flags, struct ieee80211_key_conf *key) +{ + struct iwl_sec_key_cmd cmd = { + .action = cpu_to_le32(FW_CTXT_ACTION_ADD), + .u.add.sta_mask = cpu_to_le32(sta_mask), + .u.add.key_id = cpu_to_le32(key->keyidx), + .u.add.key_flags = cpu_to_le32(key_flags), + .u.add.tx_seq = cpu_to_le64(atomic64_read(&key->tx_pn)), + }; + bool tkip = key->cipher == WLAN_CIPHER_SUITE_TKIP; + int max_key_len = sizeof(cmd.u.add.key); + + if (WARN_ON(!sta_mask)) + return -EINVAL; + + if (WARN_ON(key->keylen > max_key_len)) + return -EINVAL; + + memcpy(cmd.u.add.key, key->key, key->keylen); + + if (tkip) { + memcpy(cmd.u.add.tkip_mic_rx_key, + key->key + NL80211_TKIP_DATA_OFFSET_RX_MIC_KEY, + 8); + memcpy(cmd.u.add.tkip_mic_tx_key, + key->key + NL80211_TKIP_DATA_OFFSET_TX_MIC_KEY, + 8); + } + + return iwl_mld_send_cmd_pdu(mld, WIDE_ID(DATA_PATH_GROUP, SEC_KEY_CMD), + &cmd); +} + +static void iwl_mld_remove_key_from_fw(struct iwl_mld *mld, u32 sta_mask, + u32 key_flags, u32 keyidx) +{ + struct iwl_sec_key_cmd cmd = { + .action = cpu_to_le32(FW_CTXT_ACTION_REMOVE), + .u.remove.sta_mask = cpu_to_le32(sta_mask), + .u.remove.key_id = cpu_to_le32(keyidx), + .u.remove.key_flags = cpu_to_le32(key_flags), + }; + + if (WARN_ON(!sta_mask)) + return; + + iwl_mld_send_cmd_pdu(mld, WIDE_ID(DATA_PATH_GROUP, SEC_KEY_CMD), &cmd); +} + +void iwl_mld_remove_key(struct iwl_mld *mld, struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + struct ieee80211_key_conf *key) +{ + u32 sta_mask = iwl_mld_get_key_sta_mask(mld, vif, sta, key); + u32 key_flags = iwl_mld_get_key_flags(mld, vif, sta, key); + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + + lockdep_assert_wiphy(mld->wiphy); + + if (!sta_mask) + return; + + if (key->keyidx == 4 || key->keyidx == 5) { + struct iwl_mld_link *mld_link; + unsigned int link_id = 0; + + /* set to -1 for non-MLO right now */ + if (key->link_id >= 0) + link_id = key->link_id; + + mld_link = iwl_mld_link_dereference_check(mld_vif, link_id); + if (WARN_ON(!mld_link)) + return; + + if (mld_link->igtk == key) + mld_link->igtk = NULL; + + mld->num_igtks--; + } + + iwl_mld_remove_key_from_fw(mld, sta_mask, key_flags, key->keyidx); + + /* no longer in HW */ + key->hw_key_idx = STA_KEY_IDX_INVALID; +} + +int iwl_mld_add_key(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + struct ieee80211_key_conf *key) +{ + u32 sta_mask = iwl_mld_get_key_sta_mask(mld, vif, sta, key); + u32 key_flags = iwl_mld_get_key_flags(mld, vif, sta, key); + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mld_link *mld_link = NULL; + bool igtk = key->keyidx == 4 || key->keyidx == 5; + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + if (!sta_mask) + return -EINVAL; + + if (igtk) { + if (mld->num_igtks == IWL_MAX_NUM_IGTKS) + return -EOPNOTSUPP; + + u8 link_id = 0; + + /* set to -1 for non-MLO right now */ + if (key->link_id >= 0) + link_id = key->link_id; + + mld_link = iwl_mld_link_dereference_check(mld_vif, link_id); + + if (WARN_ON(!mld_link)) + return -EINVAL; + + if (mld_link->igtk) { + IWL_DEBUG_MAC80211(mld, "remove old IGTK %d\n", + mld_link->igtk->keyidx); + iwl_mld_remove_key(mld, vif, sta, mld_link->igtk); + } + + WARN_ON(mld_link->igtk); + } + + ret = iwl_mld_add_key_to_fw(mld, sta_mask, key_flags, key); + if (ret) + return ret; + + if (mld_link) { + mld_link->igtk = key; + mld->num_igtks++; + } + + /* We don't really need this, but need it to be not invalid, + * so we will know if the key is in fw. + */ + key->hw_key_idx = 0; + + return 0; +} + +struct remove_ap_keys_iter_data { + u8 link_id; + struct ieee80211_sta *sta; +}; + +static void iwl_mld_remove_ap_keys_iter(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + struct ieee80211_key_conf *key, + void *_data) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct remove_ap_keys_iter_data *data = _data; + + if (key->hw_key_idx == STA_KEY_IDX_INVALID) + return; + + /* All the pairwise keys should have been removed by now */ + if (WARN_ON(sta)) + return; + + if (key->link_id >= 0 && key->link_id != data->link_id) + return; + + iwl_mld_remove_key(mld, vif, data->sta, key); +} + +void iwl_mld_remove_ap_keys(struct iwl_mld *mld, struct ieee80211_vif *vif, + struct ieee80211_sta *sta, unsigned int link_id) +{ + struct remove_ap_keys_iter_data iter_data = { + .link_id = link_id, + .sta = sta, + }; + + if (WARN_ON_ONCE(vif->type != NL80211_IFTYPE_STATION)) + return; + + ieee80211_iter_keys(mld->hw, vif, + iwl_mld_remove_ap_keys_iter, + &iter_data); +} + +struct iwl_mvm_sta_key_update_data { + struct ieee80211_sta *sta; + u32 old_sta_mask; + u32 new_sta_mask; + int err; +}; + +static void iwl_mld_update_sta_key_iter(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + struct ieee80211_key_conf *key, + void *_data) +{ + struct iwl_mvm_sta_key_update_data *data = _data; + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct iwl_sec_key_cmd cmd = { + .action = cpu_to_le32(FW_CTXT_ACTION_MODIFY), + .u.modify.old_sta_mask = cpu_to_le32(data->old_sta_mask), + .u.modify.new_sta_mask = cpu_to_le32(data->new_sta_mask), + .u.modify.key_id = cpu_to_le32(key->keyidx), + .u.modify.key_flags = + cpu_to_le32(iwl_mld_get_key_flags(mld, vif, sta, key)), + }; + int err; + + /* only need to do this for pairwise keys (link_id == -1) */ + if (sta != data->sta || key->link_id >= 0) + return; + + err = iwl_mld_send_cmd_pdu(mld, WIDE_ID(DATA_PATH_GROUP, SEC_KEY_CMD), + &cmd); + + if (err) + data->err = err; +} + +int iwl_mld_update_sta_keys(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + u32 old_sta_mask, + u32 new_sta_mask) +{ + struct iwl_mvm_sta_key_update_data data = { + .sta = sta, + .old_sta_mask = old_sta_mask, + .new_sta_mask = new_sta_mask, + }; + + ieee80211_iter_keys(mld->hw, vif, iwl_mld_update_sta_key_iter, + &data); + return data.err; +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/key.h b/drivers/net/wireless/intel/iwlwifi/mld/key.h new file mode 100644 index 0000000000000..a68ea48913be6 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/key.h @@ -0,0 +1,39 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024 Intel Corporation + */ +#ifndef __iwl_mld_key_h__ +#define __iwl_mld_key_h__ + +#include "mld.h" +#include +#include "fw/api/sta.h" + +void iwl_mld_remove_key(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + struct ieee80211_key_conf *key); +int iwl_mld_add_key(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + struct ieee80211_key_conf *key); +void iwl_mld_remove_ap_keys(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + unsigned int link_id); + +int iwl_mld_update_sta_keys(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + u32 old_sta_mask, + u32 new_sta_mask); + +static inline void +iwl_mld_cleanup_keys_iter(struct ieee80211_hw *hw, struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + struct ieee80211_key_conf *key, void *data) +{ + key->hw_key_idx = STA_KEY_IDX_INVALID; +} + +#endif /* __iwl_mld_key_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/led.c b/drivers/net/wireless/intel/iwlwifi/mld/led.c new file mode 100644 index 0000000000000..a37b32cbc6e65 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/led.c @@ -0,0 +1,100 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024 Intel Corporation + */ +#include +#include + +#include "fw/api/led.h" +#include "mld.h" +#include "led.h" +#include "hcmd.h" + +static void iwl_mld_send_led_fw_cmd(struct iwl_mld *mld, bool on) +{ + struct iwl_led_cmd led_cmd = { + .status = cpu_to_le32(on), + }; + int err; + + if (WARN_ON(!mld->fw_status.running)) + return; + + err = iwl_mld_send_cmd_with_flags_pdu(mld, WIDE_ID(LONG_GROUP, + LEDS_CMD), + CMD_ASYNC, &led_cmd); + + if (err) + IWL_WARN(mld, "LED command failed: %d\n", err); +} + +static void iwl_led_brightness_set(struct led_classdev *led_cdev, + enum led_brightness brightness) +{ + struct iwl_mld *mld = container_of(led_cdev, struct iwl_mld, led); + + if (!mld->fw_status.running) + return; + + iwl_mld_send_led_fw_cmd(mld, brightness > 0); +} + +int iwl_mld_leds_init(struct iwl_mld *mld) +{ + int mode = iwlwifi_mod_params.led_mode; + int ret; + + switch (mode) { + case IWL_LED_BLINK: + IWL_ERR(mld, "Blink led mode not supported, used default\n"); + fallthrough; + case IWL_LED_DEFAULT: + case IWL_LED_RF_STATE: + mode = IWL_LED_RF_STATE; + break; + case IWL_LED_DISABLE: + IWL_INFO(mld, "Led disabled\n"); + return 0; + default: + return -EINVAL; + } + + mld->led.name = kasprintf(GFP_KERNEL, "%s-led", + wiphy_name(mld->hw->wiphy)); + if (!mld->led.name) + return -ENOMEM; + + mld->led.brightness_set = iwl_led_brightness_set; + mld->led.max_brightness = 1; + + if (mode == IWL_LED_RF_STATE) + mld->led.default_trigger = + ieee80211_get_radio_led_name(mld->hw); + + ret = led_classdev_register(mld->trans->dev, &mld->led); + if (ret) { + kfree(mld->led.name); + mld->led.name = NULL; + IWL_INFO(mld, "Failed to enable led\n"); + } + + return ret; +} + +void iwl_mld_led_config_fw(struct iwl_mld *mld) +{ + if (!mld->led.name) + return; + + iwl_mld_send_led_fw_cmd(mld, mld->led.brightness > 0); +} + +void iwl_mld_leds_exit(struct iwl_mld *mld) +{ + if (!mld->led.name) + return; + + led_classdev_unregister(&mld->led); + kfree(mld->led.name); + mld->led.name = NULL; +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/led.h b/drivers/net/wireless/intel/iwlwifi/mld/led.h new file mode 100644 index 0000000000000..0954e214e73d6 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/led.h @@ -0,0 +1,29 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024 Intel Corporation + */ +#ifndef __iwl_mld_led_h__ +#define __iwl_mld_led_h__ + +#include "mld.h" + +#ifdef CONFIG_IWLWIFI_LEDS +int iwl_mld_leds_init(struct iwl_mld *mld); +void iwl_mld_leds_exit(struct iwl_mld *mld); +void iwl_mld_led_config_fw(struct iwl_mld *mld); +#else +static inline int iwl_mld_leds_init(struct iwl_mld *mld) +{ + return 0; +} + +static inline void iwl_mld_leds_exit(struct iwl_mld *mld) +{ +} + +static inline void iwl_mld_led_config_fw(struct iwl_mld *mld) +{ +} +#endif + +#endif /* __iwl_mld_led_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/link.c b/drivers/net/wireless/intel/iwlwifi/mld/link.c new file mode 100644 index 0000000000000..64ebafc35c9bf --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/link.c @@ -0,0 +1,1216 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024-2025 Intel Corporation + */ + +#include "constants.h" +#include "link.h" +#include "iface.h" +#include "mlo.h" +#include "hcmd.h" +#include "phy.h" +#include "fw/api/rs.h" +#include "fw/api/txq.h" +#include "fw/api/mac.h" + +#include "fw/api/context.h" +#include "fw/dbg.h" + +static int iwl_mld_send_link_cmd(struct iwl_mld *mld, + struct iwl_link_config_cmd *cmd, + enum iwl_ctxt_action action) +{ + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + cmd->action = cpu_to_le32(action); + ret = iwl_mld_send_cmd_pdu(mld, + WIDE_ID(MAC_CONF_GROUP, LINK_CONFIG_CMD), + cmd); + if (ret) + IWL_ERR(mld, "Failed to send LINK_CONFIG_CMD (action:%d): %d\n", + action, ret); + return ret; +} + +static int iwl_mld_add_link_to_fw(struct iwl_mld *mld, + struct ieee80211_bss_conf *link_conf) +{ + struct ieee80211_vif *vif = link_conf->vif; + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mld_link *link = iwl_mld_link_from_mac80211(link_conf); + struct iwl_link_config_cmd cmd = {}; + + lockdep_assert_wiphy(mld->wiphy); + + if (WARN_ON(!link)) + return -EINVAL; + + cmd.link_id = cpu_to_le32(link->fw_id); + cmd.mac_id = cpu_to_le32(mld_vif->fw_id); + cmd.spec_link_id = link_conf->link_id; + cmd.phy_id = cpu_to_le32(FW_CTXT_ID_INVALID); + + ether_addr_copy(cmd.local_link_addr, link_conf->addr); + + if (vif->type == NL80211_IFTYPE_ADHOC && link_conf->bssid) + ether_addr_copy(cmd.ibss_bssid_addr, link_conf->bssid); + + return iwl_mld_send_link_cmd(mld, &cmd, FW_CTXT_ACTION_ADD); +} + +/* Get the basic rates of the used band and add the mandatory ones */ +static void iwl_mld_fill_rates(struct iwl_mld *mld, + struct ieee80211_bss_conf *link, + struct ieee80211_chanctx_conf *chan_ctx, + __le32 *cck_rates, __le32 *ofdm_rates) +{ + struct cfg80211_chan_def *chandef = + iwl_mld_get_chandef_from_chanctx(chan_ctx); + struct ieee80211_supported_band *sband = + mld->hw->wiphy->bands[chandef->chan->band]; + unsigned long basic = link->basic_rates; + int lowest_present_ofdm = 100; + int lowest_present_cck = 100; + u32 cck = 0; + u32 ofdm = 0; + int i; + + for_each_set_bit(i, &basic, BITS_PER_LONG) { + int hw = sband->bitrates[i].hw_value; + + if (hw >= IWL_FIRST_OFDM_RATE) { + ofdm |= BIT(hw - IWL_FIRST_OFDM_RATE); + if (lowest_present_ofdm > hw) + lowest_present_ofdm = hw; + } else { + BUILD_BUG_ON(IWL_FIRST_CCK_RATE != 0); + + cck |= BIT(hw); + if (lowest_present_cck > hw) + lowest_present_cck = hw; + } + } + + /* Now we've got the basic rates as bitmaps in the ofdm and cck + * variables. This isn't sufficient though, as there might not + * be all the right rates in the bitmap. E.g. if the only basic + * rates are 5.5 Mbps and 11 Mbps, we still need to add 1 Mbps + * and 6 Mbps because the 802.11-2007 standard says in 9.6: + * + * [...] a STA responding to a received frame shall transmit + * its Control Response frame [...] at the highest rate in the + * BSSBasicRateSet parameter that is less than or equal to the + * rate of the immediately previous frame in the frame exchange + * sequence ([...]) and that is of the same modulation class + * ([...]) as the received frame. If no rate contained in the + * BSSBasicRateSet parameter meets these conditions, then the + * control frame sent in response to a received frame shall be + * transmitted at the highest mandatory rate of the PHY that is + * less than or equal to the rate of the received frame, and + * that is of the same modulation class as the received frame. + * + * As a consequence, we need to add all mandatory rates that are + * lower than all of the basic rates to these bitmaps. + */ + + if (lowest_present_ofdm > IWL_RATE_24M_INDEX) + ofdm |= IWL_RATE_BIT_MSK(24) >> IWL_FIRST_OFDM_RATE; + if (lowest_present_ofdm > IWL_RATE_12M_INDEX) + ofdm |= IWL_RATE_BIT_MSK(12) >> IWL_FIRST_OFDM_RATE; + /* 6M already there or needed so always add */ + ofdm |= IWL_RATE_BIT_MSK(6) >> IWL_FIRST_OFDM_RATE; + + /* CCK is a bit more complex with DSSS vs. HR/DSSS vs. ERP. + * Note, however: + * - if no CCK rates are basic, it must be ERP since there must + * be some basic rates at all, so they're OFDM => ERP PHY + * (or we're in 5 GHz, and the cck bitmap will never be used) + * - if 11M is a basic rate, it must be ERP as well, so add 5.5M + * - if 5.5M is basic, 1M and 2M are mandatory + * - if 2M is basic, 1M is mandatory + * - if 1M is basic, that's the only valid ACK rate. + * As a consequence, it's not as complicated as it sounds, just add + * any lower rates to the ACK rate bitmap. + */ + if (lowest_present_cck > IWL_RATE_11M_INDEX) + cck |= IWL_RATE_BIT_MSK(11) >> IWL_FIRST_CCK_RATE; + if (lowest_present_cck > IWL_RATE_5M_INDEX) + cck |= IWL_RATE_BIT_MSK(5) >> IWL_FIRST_CCK_RATE; + if (lowest_present_cck > IWL_RATE_2M_INDEX) + cck |= IWL_RATE_BIT_MSK(2) >> IWL_FIRST_CCK_RATE; + /* 1M already there or needed so always add */ + cck |= IWL_RATE_BIT_MSK(1) >> IWL_FIRST_CCK_RATE; + + *cck_rates = cpu_to_le32((u32)cck); + *ofdm_rates = cpu_to_le32((u32)ofdm); +} + +static void iwl_mld_fill_protection_flags(struct iwl_mld *mld, + struct ieee80211_bss_conf *link, + __le32 *protection_flags) +{ + u8 protection_mode = link->ht_operation_mode & + IEEE80211_HT_OP_MODE_PROTECTION; + u8 ht_flag = LINK_PROT_FLG_HT_PROT | LINK_PROT_FLG_FAT_PROT; + + IWL_DEBUG_RATE(mld, "HT protection mode: %d\n", protection_mode); + + if (link->use_cts_prot) + *protection_flags |= cpu_to_le32(LINK_PROT_FLG_TGG_PROTECT); + + /* See section 9.23.3.1 of IEEE 80211-2012. + * Nongreenfield HT STAs Present is not supported. + */ + switch (protection_mode) { + case IEEE80211_HT_OP_MODE_PROTECTION_NONE: + break; + case IEEE80211_HT_OP_MODE_PROTECTION_NONMEMBER: + case IEEE80211_HT_OP_MODE_PROTECTION_NONHT_MIXED: + *protection_flags |= cpu_to_le32(ht_flag); + break; + case IEEE80211_HT_OP_MODE_PROTECTION_20MHZ: + /* Protect when channel wider than 20MHz */ + if (link->chanreq.oper.width > NL80211_CHAN_WIDTH_20) + *protection_flags |= cpu_to_le32(ht_flag); + break; + } +} + +static u8 iwl_mld_mac80211_ac_to_fw_ac(enum ieee80211_ac_numbers ac) +{ + static const u8 mac80211_ac_to_fw[] = { + AC_VO, + AC_VI, + AC_BE, + AC_BK + }; + + return mac80211_ac_to_fw[ac]; +} + +static void iwl_mld_fill_qos_params(struct ieee80211_bss_conf *link, + struct iwl_ac_qos *ac, __le32 *qos_flags) +{ + struct iwl_mld_link *mld_link = iwl_mld_link_from_mac80211(link); + + /* no need to check mld_link since it is done in the caller */ + + for (int mac_ac = 0; mac_ac < IEEE80211_NUM_ACS; mac_ac++) { + u8 txf = iwl_mld_mac80211_ac_to_fw_tx_fifo(mac_ac); + u8 fw_ac = iwl_mld_mac80211_ac_to_fw_ac(mac_ac); + + ac[fw_ac].cw_min = + cpu_to_le16(mld_link->queue_params[mac_ac].cw_min); + ac[fw_ac].cw_max = + cpu_to_le16(mld_link->queue_params[mac_ac].cw_max); + ac[fw_ac].edca_txop = + cpu_to_le16(mld_link->queue_params[mac_ac].txop * 32); + ac[fw_ac].aifsn = mld_link->queue_params[mac_ac].aifs; + ac[fw_ac].fifos_mask = BIT(txf); + } + + if (link->qos) + *qos_flags |= cpu_to_le32(MAC_QOS_FLG_UPDATE_EDCA); + + if (link->chanreq.oper.width != NL80211_CHAN_WIDTH_20_NOHT) + *qos_flags |= cpu_to_le32(MAC_QOS_FLG_TGN); +} + +static bool iwl_mld_fill_mu_edca(struct iwl_mld *mld, + const struct iwl_mld_link *mld_link, + struct iwl_he_backoff_conf *trig_based_txf) +{ + for (int mac_ac = 0; mac_ac < IEEE80211_NUM_ACS; mac_ac++) { + const struct ieee80211_he_mu_edca_param_ac_rec *mu_edca = + &mld_link->queue_params[mac_ac].mu_edca_param_rec; + u8 fw_ac = iwl_mld_mac80211_ac_to_fw_ac(mac_ac); + + if (!mld_link->queue_params[mac_ac].mu_edca) + return false; + + trig_based_txf[fw_ac].cwmin = + cpu_to_le16(mu_edca->ecw_min_max & 0xf); + trig_based_txf[fw_ac].cwmax = + cpu_to_le16((mu_edca->ecw_min_max & 0xf0) >> 4); + trig_based_txf[fw_ac].aifsn = + cpu_to_le16(mu_edca->aifsn & 0xf); + trig_based_txf[fw_ac].mu_time = + cpu_to_le16(mu_edca->mu_edca_timer); + } + return true; +} + +static u8 iwl_mld_sta_rx_bw_to_fw(enum ieee80211_sta_rx_bandwidth bw) +{ + switch (bw) { + default: /* potential future values not supported by this hw/driver */ + case IEEE80211_STA_RX_BW_20: + return IWL_LINK_MODIFY_BW_20; + case IEEE80211_STA_RX_BW_40: + return IWL_LINK_MODIFY_BW_40; + case IEEE80211_STA_RX_BW_80: + return IWL_LINK_MODIFY_BW_80; + case IEEE80211_STA_RX_BW_160: + return IWL_LINK_MODIFY_BW_160; + case IEEE80211_STA_RX_BW_320: + return IWL_LINK_MODIFY_BW_320; + } +} + +static int _iwl_mld_change_link_in_fw(struct iwl_mld *mld, + struct ieee80211_bss_conf *link, + enum ieee80211_sta_rx_bandwidth bw, + u32 changes) +{ + struct iwl_mld_link *mld_link = iwl_mld_link_from_mac80211(link); + struct ieee80211_vif *vif = link->vif; + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct ieee80211_chanctx_conf *chan_ctx; + struct iwl_link_config_cmd cmd = {}; + u32 flags = 0; + + lockdep_assert_wiphy(mld->wiphy); + + if (WARN_ON(!mld_link)) + return -EINVAL; + + cmd.link_id = cpu_to_le32(mld_link->fw_id); + cmd.spec_link_id = link->link_id; + cmd.mac_id = cpu_to_le32(mld_vif->fw_id); + + chan_ctx = wiphy_dereference(mld->wiphy, mld_link->chan_ctx); + + cmd.phy_id = cpu_to_le32(chan_ctx ? + iwl_mld_phy_from_mac80211(chan_ctx)->fw_id : + FW_CTXT_ID_INVALID); + + ether_addr_copy(cmd.local_link_addr, link->addr); + + cmd.active = cpu_to_le32(mld_link->active); + + if ((changes & LINK_CONTEXT_MODIFY_ACTIVE) && !mld_link->active && + mld_link->silent_deactivation) { + /* We are de-activating a link that is having CSA with + * immediate quiet in EMLSR. Tell the firmware not to send any + * frame. + */ + cmd.block_tx = 1; + mld_link->silent_deactivation = false; + } + + if (vif->type == NL80211_IFTYPE_ADHOC && link->bssid) + ether_addr_copy(cmd.ibss_bssid_addr, link->bssid); + + /* Channel context is needed to get the rates */ + if (chan_ctx) + iwl_mld_fill_rates(mld, link, chan_ctx, &cmd.cck_rates, + &cmd.ofdm_rates); + + cmd.cck_short_preamble = cpu_to_le32(link->use_short_preamble); + cmd.short_slot = cpu_to_le32(link->use_short_slot); + + iwl_mld_fill_protection_flags(mld, link, &cmd.protection_flags); + + iwl_mld_fill_qos_params(link, cmd.ac, &cmd.qos_flags); + + cmd.bi = cpu_to_le32(link->beacon_int); + cmd.dtim_interval = cpu_to_le32(link->beacon_int * link->dtim_period); + + if (changes & LINK_CONTEXT_MODIFY_BANDWIDTH) + cmd.modify_bandwidth = iwl_mld_sta_rx_bw_to_fw(bw); + + /* Configure HE parameters only if HE is supported, and only after + * the parameters are set in mac80211 (meaning after assoc) + */ + if (!link->he_support || iwlwifi_mod_params.disable_11ax || + (vif->type == NL80211_IFTYPE_STATION && !vif->cfg.assoc)) { + changes &= ~LINK_CONTEXT_MODIFY_HE_PARAMS; + goto send_cmd; + } + + /* ap_sta may be NULL if we're disconnecting */ + if (mld_vif->ap_sta) { + struct ieee80211_link_sta *link_sta = + link_sta_dereference_check(mld_vif->ap_sta, + link->link_id); + + if (!WARN_ON(!link_sta) && link_sta->he_cap.has_he && + link_sta->he_cap.he_cap_elem.mac_cap_info[5] & + IEEE80211_HE_MAC_CAP5_OM_CTRL_UL_MU_DATA_DIS_RX) + cmd.ul_mu_data_disable = 1; + } + + cmd.htc_trig_based_pkt_ext = link->htc_trig_based_pkt_ext; + + if (link->uora_exists) { + cmd.rand_alloc_ecwmin = link->uora_ocw_range & 0x7; + cmd.rand_alloc_ecwmax = (link->uora_ocw_range >> 3) & 0x7; + } + + if (iwl_mld_fill_mu_edca(mld, mld_link, cmd.trig_based_txf)) + flags |= LINK_FLG_MU_EDCA_CW; + + cmd.bss_color = link->he_bss_color.color; + + if (!link->he_bss_color.enabled) + flags |= LINK_FLG_BSS_COLOR_DIS; + + cmd.frame_time_rts_th = cpu_to_le16(link->frame_time_rts_th); + + /* Block 26-tone RU OFDMA transmissions */ + if (mld_link->he_ru_2mhz_block) + flags |= LINK_FLG_RU_2MHZ_BLOCK; + + if (link->nontransmitted) { + ether_addr_copy(cmd.ref_bssid_addr, link->transmitter_bssid); + cmd.bssid_index = link->bssid_index; + } + + /* The only EHT parameter is puncturing, and starting from PHY cmd + * version 6 - it is sent there. For older versions of the PHY cmd, + * puncturing is not needed at all. + */ + if (WARN_ON(changes & LINK_CONTEXT_MODIFY_EHT_PARAMS)) + changes &= ~LINK_CONTEXT_MODIFY_EHT_PARAMS; + +send_cmd: + cmd.modify_mask = cpu_to_le32(changes); + cmd.flags = cpu_to_le32(flags); + + return iwl_mld_send_link_cmd(mld, &cmd, FW_CTXT_ACTION_MODIFY); +} + +int iwl_mld_change_link_in_fw(struct iwl_mld *mld, + struct ieee80211_bss_conf *link, + u32 changes) +{ + if (WARN_ON(changes & LINK_CONTEXT_MODIFY_BANDWIDTH)) + changes &= ~LINK_CONTEXT_MODIFY_BANDWIDTH; + + return _iwl_mld_change_link_in_fw(mld, link, 0, changes); +} + +int iwl_mld_change_link_omi_bw(struct iwl_mld *mld, + struct ieee80211_bss_conf *link, + enum ieee80211_sta_rx_bandwidth bw) +{ + return _iwl_mld_change_link_in_fw(mld, link, bw, + LINK_CONTEXT_MODIFY_BANDWIDTH); +} + +int iwl_mld_activate_link(struct iwl_mld *mld, + struct ieee80211_bss_conf *link) +{ + struct iwl_mld_link *mld_link = iwl_mld_link_from_mac80211(link); + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + if (WARN_ON(!mld_link || mld_link->active)) + return -EINVAL; + + mld_link->rx_omi.exit_ts = jiffies; + mld_link->active = true; + + ret = iwl_mld_change_link_in_fw(mld, link, + LINK_CONTEXT_MODIFY_ACTIVE); + if (ret) + mld_link->active = false; + + return ret; +} + +void iwl_mld_deactivate_link(struct iwl_mld *mld, + struct ieee80211_bss_conf *link) +{ + struct iwl_mld_link *mld_link = iwl_mld_link_from_mac80211(link); + struct iwl_probe_resp_data *probe_data; + + lockdep_assert_wiphy(mld->wiphy); + + if (WARN_ON(!mld_link || !mld_link->active)) + return; + + iwl_mld_cancel_session_protection(mld, link->vif, link->link_id); + + /* If we deactivate the link, we will probably remove it, or switch + * channel. In both cases, the CSA or Notice of Absence information is + * now irrelevant. Remove the data here. + */ + probe_data = wiphy_dereference(mld->wiphy, mld_link->probe_resp_data); + RCU_INIT_POINTER(mld_link->probe_resp_data, NULL); + if (probe_data) + kfree_rcu(probe_data, rcu_head); + + mld_link->active = false; + + iwl_mld_change_link_in_fw(mld, link, LINK_CONTEXT_MODIFY_ACTIVE); + + /* Now that the link is not active in FW, we don't expect any new + * notifications for it. Cancel the ones that are already pending + */ + iwl_mld_cancel_notifications_of_object(mld, IWL_MLD_OBJECT_TYPE_LINK, + mld_link->fw_id); +} + +static int +iwl_mld_rm_link_from_fw(struct iwl_mld *mld, struct ieee80211_bss_conf *link) +{ + struct iwl_mld_link *mld_link = iwl_mld_link_from_mac80211(link); + struct iwl_link_config_cmd cmd = {}; + + lockdep_assert_wiphy(mld->wiphy); + + if (WARN_ON(!mld_link)) + return -EINVAL; + + cmd.link_id = cpu_to_le32(mld_link->fw_id); + cmd.spec_link_id = link->link_id; + cmd.phy_id = cpu_to_le32(FW_CTXT_ID_INVALID); + + return iwl_mld_send_link_cmd(mld, &cmd, FW_CTXT_ACTION_REMOVE); +} + +static void iwl_mld_omi_bw_update(struct iwl_mld *mld, + struct ieee80211_bss_conf *link_conf, + struct iwl_mld_link *mld_link, + struct ieee80211_link_sta *link_sta, + enum ieee80211_sta_rx_bandwidth bw, + bool ap_update) +{ + enum ieee80211_sta_rx_bandwidth apply_bw; + + mld_link->rx_omi.desired_bw = bw; + + /* Can't update OMI while already in progress, desired_bw was + * set so on FW notification the worker will see the change + * and apply new the new desired bw. + */ + if (mld_link->rx_omi.bw_in_progress) + return; + + if (bw == IEEE80211_STA_RX_BW_MAX) + apply_bw = ieee80211_chan_width_to_rx_bw(link_conf->chanreq.oper.width); + else + apply_bw = bw; + + if (!ap_update) { + /* The update isn't due to AP tracking after leaving OMI, + * where the AP could increase BW and then we must tell + * it that we can do the increased BW as well, if we did + * update the chandef. + * In this case, if we want MAX, then we will need to send + * a new OMI to the AP if it increases its own bandwidth as + * we can (due to internal and FW limitations, and being + * worried the AP might break) only send to what we're doing + * at the moment. In this case, set last_max_bw; otherwise + * if we really want to decrease our bandwidth set it to 0 + * to indicate no updates are needed if the AP changes. + */ + if (bw != IEEE80211_STA_RX_BW_MAX) + mld_link->rx_omi.last_max_bw = apply_bw; + else + mld_link->rx_omi.last_max_bw = 0; + } else { + /* Otherwise, if we're already trying to do maximum and + * the AP is changing, set last_max_bw to the new max the + * AP is using, we'll only get to this code path if the + * new bandwidth of the AP is bigger than what we sent it + * previously. This avoids repeatedly sending updates if + * it changes bandwidth, only doing it once on an increase. + */ + mld_link->rx_omi.last_max_bw = apply_bw; + } + + if (ieee80211_prepare_rx_omi_bw(link_sta, bw)) { + mld_link->rx_omi.bw_in_progress = apply_bw; + iwl_mld_change_link_omi_bw(mld, link_conf, apply_bw); + } +} + +static void iwl_mld_omi_bw_finished_work(struct wiphy *wiphy, + struct wiphy_work *work) +{ + struct ieee80211_hw *hw = wiphy_to_ieee80211_hw(wiphy); + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct iwl_mld_link *mld_link = + container_of(work, typeof(*mld_link), rx_omi.finished_work.work); + enum ieee80211_sta_rx_bandwidth desired_bw, switched_to_bw; + struct ieee80211_vif *vif = mld_link->vif; + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct ieee80211_bss_conf *link_conf; + struct ieee80211_link_sta *link_sta; + + if (!mld_vif->ap_sta) + return; + + link_sta = wiphy_dereference(mld->wiphy, + mld_vif->ap_sta->link[mld_link->link_id]); + if (WARN_ON_ONCE(!link_sta)) + return; + + link_conf = link_conf_dereference_protected(vif, link_sta->link_id); + if (WARN_ON_ONCE(!link_conf)) + return; + + if (WARN_ON(!mld_link->rx_omi.bw_in_progress)) + return; + + desired_bw = mld_link->rx_omi.desired_bw; + switched_to_bw = mld_link->rx_omi.bw_in_progress; + + ieee80211_finalize_rx_omi_bw(link_sta); + mld_link->rx_omi.bw_in_progress = 0; + + if (desired_bw != switched_to_bw) + iwl_mld_omi_bw_update(mld, link_conf, mld_link, link_sta, + desired_bw, false); +} + +static struct ieee80211_vif * +iwl_mld_get_omi_bw_reduction_pointers(struct iwl_mld *mld, + struct ieee80211_link_sta **link_sta, + struct iwl_mld_link **mld_link) +{ + struct iwl_mld_vif *mld_vif; + struct ieee80211_vif *vif; + int n_link_stas = 0; + + *link_sta = NULL; + + if (mld->trans->trans_cfg->device_family < IWL_DEVICE_FAMILY_SC) + return NULL; + + vif = iwl_mld_get_bss_vif(mld); + if (!vif) + return NULL; + + for (int i = 0; i < ARRAY_SIZE(mld->fw_id_to_link_sta); i++) { + struct ieee80211_link_sta *tmp; + + tmp = wiphy_dereference(mld->wiphy, mld->fw_id_to_link_sta[i]); + if (IS_ERR_OR_NULL(tmp)) + continue; + + n_link_stas++; + *link_sta = tmp; + } + + /* can't do anything if we have TDLS peers or EMLSR */ + if (n_link_stas != 1) + return NULL; + + mld_vif = iwl_mld_vif_from_mac80211(vif); + *mld_link = iwl_mld_link_dereference_check(mld_vif, + (*link_sta)->link_id); + if (WARN_ON(!*mld_link)) + return NULL; + + return vif; +} + +void iwl_mld_omi_ap_changed_bw(struct iwl_mld *mld, + struct ieee80211_bss_conf *link_conf, + enum ieee80211_sta_rx_bandwidth bw) +{ + struct ieee80211_link_sta *link_sta; + struct iwl_mld_link *mld_link; + struct ieee80211_vif *vif; + + vif = iwl_mld_get_omi_bw_reduction_pointers(mld, &link_sta, &mld_link); + if (!vif) + return; + + if (WARN_ON(link_conf->vif != vif)) + return; + + /* This is 0 if we requested an OMI BW reduction and don't want to + * be sending an OMI when the AP's bandwidth changes. + */ + if (!mld_link->rx_omi.last_max_bw) + return; + + /* We only need to tell the AP if it increases BW over what we last + * told it we were using, if it reduces then our last OMI to it will + * not get used anyway (e.g. we said we want 160 but it's doing 80.) + */ + if (bw < mld_link->rx_omi.last_max_bw) + return; + + iwl_mld_omi_bw_update(mld, link_conf, mld_link, link_sta, bw, true); +} + +void iwl_mld_handle_omi_status_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + struct ieee80211_link_sta *link_sta; + struct iwl_mld_link *mld_link; + struct ieee80211_vif *vif; + + vif = iwl_mld_get_omi_bw_reduction_pointers(mld, &link_sta, &mld_link); + if (IWL_FW_CHECK(mld, !vif, "unexpected OMI notification\n")) + return; + + if (IWL_FW_CHECK(mld, !mld_link->rx_omi.bw_in_progress, + "OMI notification when not requested\n")) + return; + + wiphy_delayed_work_queue(mld->hw->wiphy, + &mld_link->rx_omi.finished_work, + msecs_to_jiffies(IWL_MLD_OMI_AP_SETTLE_DELAY)); +} + +void iwl_mld_leave_omi_bw_reduction(struct iwl_mld *mld) +{ + struct ieee80211_bss_conf *link_conf; + struct ieee80211_link_sta *link_sta; + struct iwl_mld_link *mld_link; + struct ieee80211_vif *vif; + + vif = iwl_mld_get_omi_bw_reduction_pointers(mld, &link_sta, &mld_link); + if (!vif) + return; + + link_conf = link_conf_dereference_protected(vif, link_sta->link_id); + if (WARN_ON_ONCE(!link_conf)) + return; + + if (!link_conf->he_support) + return; + + mld_link->rx_omi.exit_ts = jiffies; + + iwl_mld_omi_bw_update(mld, link_conf, mld_link, link_sta, + IEEE80211_STA_RX_BW_MAX, false); +} + +void iwl_mld_check_omi_bw_reduction(struct iwl_mld *mld) +{ + enum ieee80211_sta_rx_bandwidth bw = IEEE80211_STA_RX_BW_MAX; + struct ieee80211_chanctx_conf *chanctx; + struct ieee80211_bss_conf *link_conf; + struct ieee80211_link_sta *link_sta; + struct cfg80211_chan_def chandef; + struct iwl_mld_link *mld_link; + struct iwl_mld_vif *mld_vif; + struct ieee80211_vif *vif; + struct iwl_mld_phy *phy; + u16 punctured; + int exit_thr; + + /* not allowed in CAM mode */ + if (iwlmld_mod_params.power_scheme == IWL_POWER_SCHEME_CAM) + return; + + /* must have one BSS connection (no P2P), no TDLS, nor EMLSR */ + vif = iwl_mld_get_omi_bw_reduction_pointers(mld, &link_sta, &mld_link); + if (!vif) + return; + + link_conf = link_conf_dereference_protected(vif, link_sta->link_id); + if (WARN_ON_ONCE(!link_conf)) + return; + + if (!link_conf->he_support) + return; + + chanctx = wiphy_dereference(mld->wiphy, mld_link->chan_ctx); + if (WARN_ON(!chanctx)) + return; + + mld_vif = iwl_mld_vif_from_mac80211(vif); + if (!mld_vif->authorized) + goto apply; + + /* must not be in low-latency mode */ + if (iwl_mld_vif_low_latency(mld_vif)) + goto apply; + + chandef = link_conf->chanreq.oper; + + switch (chandef.width) { + case NL80211_CHAN_WIDTH_320: + exit_thr = IWL_MLD_OMI_EXIT_CHAN_LOAD_320; + break; + case NL80211_CHAN_WIDTH_160: + exit_thr = IWL_MLD_OMI_EXIT_CHAN_LOAD_160; + break; + default: + /* since we reduce to 80 MHz, must have more to start with */ + goto apply; + } + + /* not to be done if primary 80 MHz is punctured */ + if (cfg80211_chandef_primary(&chandef, NL80211_CHAN_WIDTH_80, + &punctured) < 0 || + punctured != 0) + goto apply; + + phy = iwl_mld_phy_from_mac80211(chanctx); + + if (phy->channel_load_by_us > exit_thr) { + /* send OMI for max bandwidth */ + goto apply; + } + + if (phy->channel_load_by_us > IWL_MLD_OMI_ENTER_CHAN_LOAD) { + /* no changes between enter/exit thresholds */ + return; + } + + if (time_is_before_jiffies(mld_link->rx_omi.exit_ts + + msecs_to_jiffies(IWL_MLD_OMI_EXIT_PROTECTION))) + return; + + /* reduce bandwidth to 80 MHz to save power */ + bw = IEEE80211_STA_RX_BW_80; +apply: + iwl_mld_omi_bw_update(mld, link_conf, mld_link, link_sta, bw, false); +} + +IWL_MLD_ALLOC_FN(link, bss_conf) + +/* Constructor function for struct iwl_mld_link */ +static int +iwl_mld_init_link(struct iwl_mld *mld, struct ieee80211_bss_conf *link, + struct iwl_mld_link *mld_link) +{ + mld_link->vif = link->vif; + mld_link->link_id = link->link_id; + + iwl_mld_init_internal_sta(&mld_link->bcast_sta); + iwl_mld_init_internal_sta(&mld_link->mcast_sta); + iwl_mld_init_internal_sta(&mld_link->aux_sta); + + wiphy_delayed_work_init(&mld_link->rx_omi.finished_work, + iwl_mld_omi_bw_finished_work); + + return iwl_mld_allocate_link_fw_id(mld, &mld_link->fw_id, link); +} + +/* Initializes the link structure, maps fw id to the ieee80211_bss_conf, and + * adds a link to the fw + */ +int iwl_mld_add_link(struct iwl_mld *mld, + struct ieee80211_bss_conf *bss_conf) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(bss_conf->vif); + struct iwl_mld_link *link = iwl_mld_link_from_mac80211(bss_conf); + bool is_deflink = bss_conf == &bss_conf->vif->bss_conf; + int ret; + + if (!link) { + if (is_deflink) + link = &mld_vif->deflink; + else + link = kzalloc(sizeof(*link), GFP_KERNEL); + } else { + WARN_ON(!mld->fw_status.in_hw_restart); + } + + ret = iwl_mld_init_link(mld, bss_conf, link); + if (ret) + goto free; + + rcu_assign_pointer(mld_vif->link[bss_conf->link_id], link); + + ret = iwl_mld_add_link_to_fw(mld, bss_conf); + if (ret) { + RCU_INIT_POINTER(mld->fw_id_to_bss_conf[link->fw_id], NULL); + RCU_INIT_POINTER(mld_vif->link[bss_conf->link_id], NULL); + goto free; + } + + return ret; + +free: + if (!is_deflink) + kfree(link); + return ret; +} + +/* Remove link from fw, unmap the bss_conf, and destroy the link structure */ +int iwl_mld_remove_link(struct iwl_mld *mld, + struct ieee80211_bss_conf *bss_conf) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(bss_conf->vif); + struct iwl_mld_link *link = iwl_mld_link_from_mac80211(bss_conf); + bool is_deflink = link == &mld_vif->deflink; + int ret; + + if (WARN_ON(!link || link->active)) + return -EINVAL; + + ret = iwl_mld_rm_link_from_fw(mld, bss_conf); + /* Continue cleanup on failure */ + + if (!is_deflink) + kfree_rcu(link, rcu_head); + + RCU_INIT_POINTER(mld_vif->link[bss_conf->link_id], NULL); + + wiphy_delayed_work_cancel(mld->wiphy, &link->rx_omi.finished_work); + + if (WARN_ON(link->fw_id >= mld->fw->ucode_capa.num_links)) + return -EINVAL; + + RCU_INIT_POINTER(mld->fw_id_to_bss_conf[link->fw_id], NULL); + + return ret; +} + +void iwl_mld_handle_missed_beacon_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + const struct iwl_missed_beacons_notif *notif = (const void *)pkt->data; + union iwl_dbg_tlv_tp_data tp_data = { .fw_pkt = pkt }; + u32 link_id = le32_to_cpu(notif->link_id); + u32 missed_bcon = le32_to_cpu(notif->consec_missed_beacons); + u32 missed_bcon_since_rx = + le32_to_cpu(notif->consec_missed_beacons_since_last_rx); + u32 scnd_lnk_bcn_lost = + le32_to_cpu(notif->consec_missed_beacons_other_link); + struct ieee80211_bss_conf *link_conf = + iwl_mld_fw_id_to_link_conf(mld, link_id); + u32 bss_param_ch_cnt_link_id; + struct ieee80211_vif *vif; + + if (WARN_ON(!link_conf)) + return; + + vif = link_conf->vif; + bss_param_ch_cnt_link_id = link_conf->bss_param_ch_cnt_link_id; + + IWL_DEBUG_INFO(mld, + "missed bcn link_id=%u, %u consecutive=%u\n", + link_id, missed_bcon, missed_bcon_since_rx); + + if (WARN_ON(!vif)) + return; + + mld->trans->dbg.dump_file_name_ext_valid = true; + snprintf(mld->trans->dbg.dump_file_name_ext, IWL_FW_INI_MAX_NAME, + "LinkId_%d_MacType_%d", link_id, + iwl_mld_mac80211_iftype_to_fw(vif)); + + iwl_dbg_tlv_time_point(&mld->fwrt, + IWL_FW_INI_TIME_POINT_MISSED_BEACONS, &tp_data); + + if (missed_bcon >= IWL_MLD_MISSED_BEACONS_THRESHOLD_LONG) { + if (missed_bcon_since_rx >= + IWL_MLD_MISSED_BEACONS_SINCE_RX_THOLD) { + ieee80211_connection_loss(vif); + return; + } + IWL_WARN(mld, + "missed beacons exceeds threshold, but receiving data. Stay connected, Expect bugs.\n"); + return; + } + + if (missed_bcon_since_rx > IWL_MLD_MISSED_BEACONS_THRESHOLD) { + ieee80211_cqm_beacon_loss_notify(vif, GFP_ATOMIC); + + /* try to switch links, no-op if we don't have MLO */ + iwl_mld_trigger_link_selection(mld, vif); + } + + /* no more logic if we're not in EMLSR */ + if (hweight16(vif->active_links) <= 1) + return; + + /* We are processing a notification before link activation */ + if (le32_to_cpu(notif->other_link_id) == FW_CTXT_ID_INVALID) + return; + + /* Exit EMLSR if we lost more than + * IWL_MLD_MISSED_BEACONS_EXIT_ESR_THRESH beacons on boths links + * OR more than IWL_MLD_BCN_LOSS_EXIT_ESR_THRESH on current link. + * OR more than IWL_MLD_BCN_LOSS_EXIT_ESR_THRESH_BSS_PARAM_CHANGED + * on current link and the link's bss_param_ch_count has changed on + * the other link's beacon. + */ + if ((missed_bcon >= IWL_MLD_BCN_LOSS_EXIT_ESR_THRESH_2_LINKS && + scnd_lnk_bcn_lost >= IWL_MLD_BCN_LOSS_EXIT_ESR_THRESH_2_LINKS) || + missed_bcon >= IWL_MLD_BCN_LOSS_EXIT_ESR_THRESH || + (bss_param_ch_cnt_link_id != link_id && + missed_bcon >= + IWL_MLD_BCN_LOSS_EXIT_ESR_THRESH_BSS_PARAM_CHANGED)) { + iwl_mld_exit_emlsr(mld, vif, IWL_MLD_EMLSR_EXIT_MISSED_BEACON, + iwl_mld_get_primary_link(vif)); + } +} +EXPORT_SYMBOL_IF_IWLWIFI_KUNIT(iwl_mld_handle_missed_beacon_notif); + +bool iwl_mld_cancel_missed_beacon_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt, + u32 removed_link_id) +{ + struct iwl_missed_beacons_notif *notif = (void *)pkt->data; + + if (le32_to_cpu(notif->other_link_id) == removed_link_id) { + /* Second link is being removed. Don't cancel the notification, + * but mark second link as invalid. + */ + notif->other_link_id = cpu_to_le32(FW_CTXT_ID_INVALID); + } + + /* If the primary link is removed, cancel the notification */ + return le32_to_cpu(notif->link_id) == removed_link_id; +} + +int iwl_mld_link_set_associated(struct iwl_mld *mld, struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link) +{ + return iwl_mld_change_link_in_fw(mld, link, LINK_CONTEXT_MODIFY_ALL & + ~(LINK_CONTEXT_MODIFY_ACTIVE | + LINK_CONTEXT_MODIFY_EHT_PARAMS)); +} + +struct iwl_mld_rssi_to_grade { + s8 rssi[2]; + u16 grade; +}; + +#define RSSI_TO_GRADE_LINE(_lb, _hb_uhb, _grade) \ + { \ + .rssi = {_lb, _hb_uhb}, \ + .grade = _grade \ + } + +/* + * This array must be sorted by increasing RSSI for proper functionality. + * The grades are actually estimated throughput, represented as fixed-point + * with a scale factor of 1/10. + */ +static const struct iwl_mld_rssi_to_grade rssi_to_grade_map[] = { + RSSI_TO_GRADE_LINE(-85, -89, 172), + RSSI_TO_GRADE_LINE(-83, -86, 344), + RSSI_TO_GRADE_LINE(-82, -85, 516), + RSSI_TO_GRADE_LINE(-80, -83, 688), + RSSI_TO_GRADE_LINE(-77, -79, 1032), + RSSI_TO_GRADE_LINE(-73, -76, 1376), + RSSI_TO_GRADE_LINE(-70, -74, 1548), + RSSI_TO_GRADE_LINE(-69, -72, 1720), + RSSI_TO_GRADE_LINE(-65, -68, 2064), + RSSI_TO_GRADE_LINE(-61, -66, 2294), + RSSI_TO_GRADE_LINE(-58, -61, 2580), + RSSI_TO_GRADE_LINE(-55, -58, 2868), + RSSI_TO_GRADE_LINE(-46, -55, 3098), + RSSI_TO_GRADE_LINE(-43, -54, 3442) +}; + +#define MAX_GRADE (rssi_to_grade_map[ARRAY_SIZE(rssi_to_grade_map) - 1].grade) + +#define DEFAULT_CHAN_LOAD_2GHZ 30 +#define DEFAULT_CHAN_LOAD_5GHZ 15 +#define DEFAULT_CHAN_LOAD_6GHZ 0 + +/* Factors calculation is done with fixed-point with a scaling factor of 1/256 */ +#define SCALE_FACTOR 256 +#define MAX_CHAN_LOAD 256 + +static unsigned int +iwl_mld_get_n_subchannels(const struct ieee80211_bss_conf *link_conf) +{ + enum nl80211_chan_width chan_width = + link_conf->chanreq.oper.width; + int mhz = nl80211_chan_width_to_mhz(chan_width); + unsigned int n_subchannels; + + if (WARN_ONCE(mhz < 20 || mhz > 320, + "Invalid channel width : (%d)\n", mhz)) + return 1; + + /* total number of subchannels */ + n_subchannels = mhz / 20; + + /* No puncturing if less than 80 MHz */ + if (mhz >= 80) + n_subchannels -= hweight16(link_conf->chanreq.oper.punctured); + + return n_subchannels; +} + +static int +iwl_mld_get_chan_load_from_element(struct iwl_mld *mld, + struct ieee80211_bss_conf *link_conf) +{ + struct ieee80211_vif *vif = link_conf->vif; + const struct cfg80211_bss_ies *ies; + const struct element *bss_load_elem = NULL; + const struct ieee80211_bss_load_elem *bss_load; + + guard(rcu)(); + + if (ieee80211_vif_link_active(vif, link_conf->link_id)) + ies = rcu_dereference(link_conf->bss->beacon_ies); + else + ies = rcu_dereference(link_conf->bss->ies); + + if (ies) + bss_load_elem = cfg80211_find_elem(WLAN_EID_QBSS_LOAD, + ies->data, ies->len); + + if (!bss_load_elem || + bss_load_elem->datalen != sizeof(*bss_load)) + return -EINVAL; + + bss_load = (const void *)bss_load_elem->data; + + return bss_load->channel_util; +} + +static unsigned int +iwl_mld_get_chan_load_by_us(struct iwl_mld *mld, + struct ieee80211_bss_conf *link_conf, + bool expect_active_link) +{ + struct iwl_mld_link *mld_link = iwl_mld_link_from_mac80211(link_conf); + struct ieee80211_chanctx_conf *chan_ctx; + struct iwl_mld_phy *phy; + + if (!mld_link || !mld_link->active) { + WARN_ON(expect_active_link); + return 0; + } + + if (WARN_ONCE(!rcu_access_pointer(mld_link->chan_ctx), + "Active link (%u) without channel ctxt assigned!\n", + link_conf->link_id)) + return 0; + + chan_ctx = wiphy_dereference(mld->wiphy, mld_link->chan_ctx); + phy = iwl_mld_phy_from_mac80211(chan_ctx); + + return phy->channel_load_by_us; +} + +/* Returns error if the channel utilization element is invalid/unavailable */ +int iwl_mld_get_chan_load_by_others(struct iwl_mld *mld, + struct ieee80211_bss_conf *link_conf, + bool expect_active_link) +{ + int chan_load; + unsigned int chan_load_by_us; + + /* get overall load */ + chan_load = iwl_mld_get_chan_load_from_element(mld, link_conf); + if (chan_load < 0) + return chan_load; + + chan_load_by_us = iwl_mld_get_chan_load_by_us(mld, link_conf, + expect_active_link); + + /* channel load by us is given in percentage */ + chan_load_by_us = + NORMALIZE_PERCENT_TO_255(chan_load_by_us); + + /* Use only values that firmware sends that can possibly be valid */ + if (chan_load_by_us <= chan_load) + chan_load -= chan_load_by_us; + + return chan_load; +} + +static unsigned int +iwl_mld_get_default_chan_load(struct ieee80211_bss_conf *link_conf) +{ + enum nl80211_band band = link_conf->chanreq.oper.chan->band; + + switch (band) { + case NL80211_BAND_2GHZ: + return DEFAULT_CHAN_LOAD_2GHZ; + case NL80211_BAND_5GHZ: + return DEFAULT_CHAN_LOAD_5GHZ; + case NL80211_BAND_6GHZ: + return DEFAULT_CHAN_LOAD_6GHZ; + default: + WARN_ON(1); + return 0; + } +} + +unsigned int iwl_mld_get_chan_load(struct iwl_mld *mld, + struct ieee80211_bss_conf *link_conf) +{ + int chan_load; + + chan_load = iwl_mld_get_chan_load_by_others(mld, link_conf, false); + if (chan_load >= 0) + return chan_load; + + /* No information from the element, take the defaults */ + chan_load = iwl_mld_get_default_chan_load(link_conf); + + /* The defaults are given in percentage */ + return NORMALIZE_PERCENT_TO_255(chan_load); +} + +static unsigned int +iwl_mld_get_avail_chan_load(struct iwl_mld *mld, + struct ieee80211_bss_conf *link_conf) +{ + return MAX_CHAN_LOAD - iwl_mld_get_chan_load(mld, link_conf); +} + +/* This function calculates the grade of a link. Returns 0 in error case */ +unsigned int iwl_mld_get_link_grade(struct iwl_mld *mld, + struct ieee80211_bss_conf *link_conf) +{ + enum nl80211_band band; + int rssi_idx; + s32 link_rssi; + unsigned int grade = MAX_GRADE; + + if (WARN_ON_ONCE(!link_conf)) + return 0; + + band = link_conf->chanreq.oper.chan->band; + if (WARN_ONCE(band != NL80211_BAND_2GHZ && + band != NL80211_BAND_5GHZ && + band != NL80211_BAND_6GHZ, + "Invalid band (%u)\n", band)) + return 0; + + link_rssi = MBM_TO_DBM(link_conf->bss->signal); + /* + * For 6 GHz the RSSI of the beacons is lower than + * the RSSI of the data. + */ + if (band == NL80211_BAND_6GHZ && link_rssi) + link_rssi += 4; + + rssi_idx = band == NL80211_BAND_2GHZ ? 0 : 1; + + /* No valid RSSI - take the lowest grade */ + if (!link_rssi) + link_rssi = rssi_to_grade_map[0].rssi[rssi_idx]; + + IWL_DEBUG_EHT(mld, + "Calculating grade of link %d: band = %d, bandwidth = %d, punctured subchannels =0x%x RSSI = %d\n", + link_conf->link_id, band, + link_conf->chanreq.oper.width, + link_conf->chanreq.oper.punctured, link_rssi); + + /* Get grade based on RSSI */ + for (int i = 0; i < ARRAY_SIZE(rssi_to_grade_map); i++) { + const struct iwl_mld_rssi_to_grade *line = + &rssi_to_grade_map[i]; + + if (link_rssi > line->rssi[rssi_idx]) + continue; + grade = line->grade; + break; + } + + /* Apply the channel load and puncturing factors */ + grade = grade * iwl_mld_get_avail_chan_load(mld, link_conf) / SCALE_FACTOR; + grade = grade * iwl_mld_get_n_subchannels(link_conf); + + IWL_DEBUG_EHT(mld, "Link %d's grade: %d\n", link_conf->link_id, grade); + + return grade; +} +EXPORT_SYMBOL_IF_IWLWIFI_KUNIT(iwl_mld_get_link_grade); diff --git a/drivers/net/wireless/intel/iwlwifi/mld/link.h b/drivers/net/wireless/intel/iwlwifi/mld/link.h new file mode 100644 index 0000000000000..3a4f3ac990d7e --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/link.h @@ -0,0 +1,153 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024-2025 Intel Corporation + */ +#ifndef __iwl_mld_link_h__ +#define __iwl_mld_link_h__ + +#include + +#include "mld.h" +#include "sta.h" + +/** + * struct iwl_probe_resp_data - data for NoA/CSA updates + * @rcu_head: used for freeing the data on update + * @notif: notification data + * @noa_len: length of NoA attribute, calculated from the notification + */ +struct iwl_probe_resp_data { + struct rcu_head rcu_head; + struct iwl_probe_resp_data_notif notif; + int noa_len; +}; + +/** + * struct iwl_mld_link - link configuration parameters + * + * @rcu_head: RCU head for freeing this data. + * @fw_id: the fw id of the link. + * @active: if the link is active or not. + * @queue_params: QoS data from mac80211. This is updated with a call to + * drv_conf_tx per each AC, and then notified once with BSS_CHANGED_QOS. + * So we store it here and then send one link cmd for all the ACs. + * @chan_ctx: pointer to the channel context assigned to the link. If a link + * has an assigned channel context it means that it is active. + * @he_ru_2mhz_block: 26-tone RU OFDMA transmissions should be blocked. + * @igtk: fw can only have one IGTK at a time, whereas mac80211 can have two. + * This tracks the one IGTK that currently exists in FW. + * @vif: the vif this link belongs to + * @bcast_sta: station used for broadcast packets. Used in AP, GO and IBSS. + * @mcast_sta: station used for multicast packets. Used in AP, GO and IBSS. + * @aux_sta: station used for remain on channel. Used in P2P device. + * @link_id: over the air link ID + * @ap_early_keys: The firmware cannot install keys before bcast/mcast STAs, + * but higher layers work differently, so we store the keys here for + * later installation. + * @silent_deactivation: next deactivation needs to be silent. + * @probe_resp_data: data from FW notification to store NOA related data to be + * inserted into probe response. + * @rx_omi: data for BW reduction with OMI + * @rx_omi.bw_in_progress: update is in progress (indicates target BW) + * @rx_omi.exit_ts: timestamp of last exit + * @rx_omi.finished_work: work for the delayed reaction to the firmware saying + * the change was applied, and for then applying a new mode if it was + * updated while waiting for firmware/AP settle delay. + * @rx_omi.desired_bw: desired bandwidth + * @rx_omi.last_max_bw: last maximum BW used by firmware, for AP BW changes + */ +struct iwl_mld_link { + struct rcu_head rcu_head; + + /* Add here fields that need clean up on restart */ + struct_group(zeroed_on_hw_restart, + u8 fw_id; + bool active; + struct ieee80211_tx_queue_params queue_params[IEEE80211_NUM_ACS]; + struct ieee80211_chanctx_conf __rcu *chan_ctx; + bool he_ru_2mhz_block; + struct ieee80211_key_conf *igtk; + ); + /* And here fields that survive a fw restart */ + struct ieee80211_vif *vif; + struct iwl_mld_int_sta bcast_sta; + struct iwl_mld_int_sta mcast_sta; + struct iwl_mld_int_sta aux_sta; + u8 link_id; + + struct { + struct wiphy_delayed_work finished_work; + unsigned long exit_ts; + enum ieee80211_sta_rx_bandwidth bw_in_progress, + desired_bw, + last_max_bw; + } rx_omi; + + /* we can only have 2 GTK + 2 IGTK + 2 BIGTK active at a time */ + struct ieee80211_key_conf *ap_early_keys[6]; + bool silent_deactivation; + struct iwl_probe_resp_data __rcu *probe_resp_data; +}; + +/* Cleanup function for struct iwl_mld_phy, will be called in restart */ +static inline void +iwl_mld_cleanup_link(struct iwl_mld *mld, struct iwl_mld_link *link) +{ + struct iwl_probe_resp_data *probe_data; + + probe_data = wiphy_dereference(mld->wiphy, link->probe_resp_data); + RCU_INIT_POINTER(link->probe_resp_data, NULL); + if (probe_data) + kfree_rcu(probe_data, rcu_head); + + CLEANUP_STRUCT(link); + if (link->bcast_sta.sta_id != IWL_INVALID_STA) + iwl_mld_free_internal_sta(mld, &link->bcast_sta); + if (link->mcast_sta.sta_id != IWL_INVALID_STA) + iwl_mld_free_internal_sta(mld, &link->mcast_sta); + if (link->aux_sta.sta_id != IWL_INVALID_STA) + iwl_mld_free_internal_sta(mld, &link->aux_sta); +} + +/* Convert a percentage from [0,100] to [0,255] */ +#define NORMALIZE_PERCENT_TO_255(percentage) ((percentage) * 256 / 100) + +int iwl_mld_add_link(struct iwl_mld *mld, + struct ieee80211_bss_conf *bss_conf); +int iwl_mld_remove_link(struct iwl_mld *mld, + struct ieee80211_bss_conf *bss_conf); +int iwl_mld_activate_link(struct iwl_mld *mld, + struct ieee80211_bss_conf *link); +void iwl_mld_deactivate_link(struct iwl_mld *mld, + struct ieee80211_bss_conf *link); +int iwl_mld_change_link_omi_bw(struct iwl_mld *mld, + struct ieee80211_bss_conf *link, + enum ieee80211_sta_rx_bandwidth bw); +int iwl_mld_change_link_in_fw(struct iwl_mld *mld, + struct ieee80211_bss_conf *link, u32 changes); +void iwl_mld_handle_missed_beacon_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt); +bool iwl_mld_cancel_missed_beacon_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt, + u32 removed_link_id); +int iwl_mld_link_set_associated(struct iwl_mld *mld, struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link); + +unsigned int iwl_mld_get_link_grade(struct iwl_mld *mld, + struct ieee80211_bss_conf *link_conf); + +unsigned int iwl_mld_get_chan_load(struct iwl_mld *mld, + struct ieee80211_bss_conf *link_conf); + +int iwl_mld_get_chan_load_by_others(struct iwl_mld *mld, + struct ieee80211_bss_conf *link_conf, + bool expect_active_link); +void iwl_mld_handle_omi_status_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt); +void iwl_mld_leave_omi_bw_reduction(struct iwl_mld *mld); +void iwl_mld_check_omi_bw_reduction(struct iwl_mld *mld); +void iwl_mld_omi_ap_changed_bw(struct iwl_mld *mld, + struct ieee80211_bss_conf *link_conf, + enum ieee80211_sta_rx_bandwidth bw); + +#endif /* __iwl_mld_link_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/low_latency.c b/drivers/net/wireless/intel/iwlwifi/mld/low_latency.c new file mode 100644 index 0000000000000..439fc10a4a412 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/low_latency.c @@ -0,0 +1,335 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024-2025 Intel Corporation + */ +#include "mld.h" +#include "iface.h" +#include "low_latency.h" +#include "hcmd.h" +#include "power.h" + +#define MLD_LL_WK_INTERVAL_MSEC 500 +#define MLD_LL_PERIOD (HZ * MLD_LL_WK_INTERVAL_MSEC / 1000) +#define MLD_LL_ACTIVE_WK_PERIOD (HZ * 10) + +/* packets/MLD_LL_WK_PERIOD seconds */ +#define MLD_LL_ENABLE_THRESH 100 + +static bool iwl_mld_calc_low_latency(struct iwl_mld *mld, + unsigned long timestamp) +{ + struct iwl_mld_low_latency *ll = &mld->low_latency; + bool global_low_latency = false; + u8 num_rx_q = mld->trans->num_rx_queues; + + for (int mac_id = 0; mac_id < NUM_MAC_INDEX_DRIVER; mac_id++) { + u32 total_vo_vi_pkts = 0; + bool ll_period_expired; + + /* If it's not initialized yet, it means we have not yet + * received/transmitted any vo/vi packet on this MAC. + */ + if (!ll->window_start[mac_id]) + continue; + + ll_period_expired = + time_after(timestamp, ll->window_start[mac_id] + + MLD_LL_ACTIVE_WK_PERIOD); + + if (ll_period_expired) + ll->window_start[mac_id] = timestamp; + + for (int q = 0; q < num_rx_q; q++) { + struct iwl_mld_low_latency_packets_counters *counters = + &mld->low_latency.pkts_counters[q]; + + spin_lock_bh(&counters->lock); + + total_vo_vi_pkts += counters->vo_vi[mac_id]; + + if (ll_period_expired) + counters->vo_vi[mac_id] = 0; + + spin_unlock_bh(&counters->lock); + } + + /* enable immediately with enough packets but defer + * disabling only if the low-latency period expired and + * below threshold. + */ + if (total_vo_vi_pkts > MLD_LL_ENABLE_THRESH) + mld->low_latency.result[mac_id] = true; + else if (ll_period_expired) + mld->low_latency.result[mac_id] = false; + + global_low_latency |= mld->low_latency.result[mac_id]; + } + + return global_low_latency; +} + +static void iwl_mld_low_latency_iter(void *_data, u8 *mac, + struct ieee80211_vif *vif) +{ + struct iwl_mld *mld = _data; + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + bool prev = mld_vif->low_latency_causes & LOW_LATENCY_TRAFFIC; + bool low_latency; + + if (WARN_ON(mld_vif->fw_id >= ARRAY_SIZE(mld->low_latency.result))) + return; + + low_latency = mld->low_latency.result[mld_vif->fw_id]; + + if (prev != low_latency) + iwl_mld_vif_update_low_latency(mld, vif, low_latency, + LOW_LATENCY_TRAFFIC); +} + +static void iwl_mld_low_latency_wk(struct wiphy *wiphy, struct wiphy_work *wk) +{ + struct iwl_mld *mld = container_of(wk, struct iwl_mld, + low_latency.work.work); + unsigned long timestamp = jiffies; + bool low_latency_active; + + if (mld->fw_status.in_hw_restart) + return; + + /* It is assumed that the work was scheduled only after checking + * at least MLD_LL_PERIOD has passed since the last update. + */ + + low_latency_active = iwl_mld_calc_low_latency(mld, timestamp); + + /* Update the timestamp now after the low-latency calculation */ + mld->low_latency.timestamp = timestamp; + + /* If low-latency is active we need to force re-evaluation after + * 10 seconds, so that we can disable low-latency when + * the low-latency traffic ends. + * + * Otherwise, we don't need to run the work because there is nothing to + * disable. + * + * Note that this has no impact on the regular scheduling of the + * updates triggered by traffic - those happen whenever the + * MLD_LL_PERIOD timeout expire. + */ + if (low_latency_active) + wiphy_delayed_work_queue(mld->wiphy, &mld->low_latency.work, + MLD_LL_ACTIVE_WK_PERIOD); + + ieee80211_iterate_active_interfaces(mld->hw, + IEEE80211_IFACE_ITER_NORMAL, + iwl_mld_low_latency_iter, mld); +} + +int iwl_mld_low_latency_init(struct iwl_mld *mld) +{ + struct iwl_mld_low_latency *ll = &mld->low_latency; + unsigned long ts = jiffies; + + ll->pkts_counters = kcalloc(mld->trans->num_rx_queues, + sizeof(*ll->pkts_counters), GFP_KERNEL); + if (!ll->pkts_counters) + return -ENOMEM; + + for (int q = 0; q < mld->trans->num_rx_queues; q++) + spin_lock_init(&ll->pkts_counters[q].lock); + + wiphy_delayed_work_init(&ll->work, iwl_mld_low_latency_wk); + + ll->timestamp = ts; + + /* The low-latency window_start will be initialized per-MAC on + * the first vo/vi packet received/transmitted. + */ + + return 0; +} + +void iwl_mld_low_latency_free(struct iwl_mld *mld) +{ + struct iwl_mld_low_latency *ll = &mld->low_latency; + + kfree(ll->pkts_counters); + ll->pkts_counters = NULL; +} + +void iwl_mld_low_latency_restart_cleanup(struct iwl_mld *mld) +{ + struct iwl_mld_low_latency *ll = &mld->low_latency; + + ll->timestamp = jiffies; + + memset(ll->window_start, 0, sizeof(ll->window_start)); + memset(ll->result, 0, sizeof(ll->result)); + + for (int q = 0; q < mld->trans->num_rx_queues; q++) + memset(ll->pkts_counters[q].vo_vi, 0, + sizeof(ll->pkts_counters[q].vo_vi)); +} + +static int iwl_mld_send_low_latency_cmd(struct iwl_mld *mld, bool low_latency, + u16 mac_id) +{ + struct iwl_mac_low_latency_cmd cmd = { + .mac_id = cpu_to_le32(mac_id) + }; + u16 cmd_id = WIDE_ID(MAC_CONF_GROUP, LOW_LATENCY_CMD); + int ret; + + if (low_latency) { + /* Currently we don't care about the direction */ + cmd.low_latency_rx = 1; + cmd.low_latency_tx = 1; + } + + ret = iwl_mld_send_cmd_pdu(mld, cmd_id, &cmd); + if (ret) + IWL_ERR(mld, "Failed to send low latency command\n"); + + return ret; +} + +static void iwl_mld_vif_set_low_latency(struct iwl_mld_vif *mld_vif, bool set, + enum iwl_mld_low_latency_cause cause) +{ + if (set) + mld_vif->low_latency_causes |= cause; + else + mld_vif->low_latency_causes &= ~cause; +} + +void iwl_mld_vif_update_low_latency(struct iwl_mld *mld, + struct ieee80211_vif *vif, + bool low_latency, + enum iwl_mld_low_latency_cause cause) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + bool prev; + + prev = iwl_mld_vif_low_latency(mld_vif); + iwl_mld_vif_set_low_latency(mld_vif, low_latency, cause); + + low_latency = iwl_mld_vif_low_latency(mld_vif); + if (low_latency == prev) + return; + + if (iwl_mld_send_low_latency_cmd(mld, low_latency, mld_vif->fw_id)) { + /* revert to previous low-latency state */ + iwl_mld_vif_set_low_latency(mld_vif, prev, cause); + return; + } + + if (low_latency) + iwl_mld_leave_omi_bw_reduction(mld); + + if (ieee80211_vif_type_p2p(vif) != NL80211_IFTYPE_P2P_CLIENT) + return; + + iwl_mld_update_mac_power(mld, vif, false); +} + +static bool iwl_mld_is_vo_vi_pkt(struct ieee80211_hdr *hdr) +{ + u8 tid; + static const u8 tid_to_mac80211_ac[] = { + IEEE80211_AC_BE, + IEEE80211_AC_BK, + IEEE80211_AC_BK, + IEEE80211_AC_BE, + IEEE80211_AC_VI, + IEEE80211_AC_VI, + IEEE80211_AC_VO, + IEEE80211_AC_VO, + }; + + if (!hdr || !ieee80211_is_data_qos(hdr->frame_control)) + return false; + + tid = ieee80211_get_tid(hdr); + if (tid >= IWL_MAX_TID_COUNT) + return false; + + return tid_to_mac80211_ac[tid] < IEEE80211_AC_VI; +} + +void iwl_mld_low_latency_update_counters(struct iwl_mld *mld, + struct ieee80211_hdr *hdr, + struct ieee80211_sta *sta, + u8 queue) +{ + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(sta); + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(mld_sta->vif); + struct iwl_mld_low_latency_packets_counters *counters; + unsigned long ts = jiffies ? jiffies : 1; + u8 fw_id = mld_vif->fw_id; + + /* we should have failed op mode init if NULL */ + if (WARN_ON_ONCE(!mld->low_latency.pkts_counters)) + return; + + if (WARN_ON_ONCE(fw_id >= ARRAY_SIZE(counters->vo_vi) || + queue >= mld->trans->num_rx_queues)) + return; + + if (mld->low_latency.stopped) + return; + + if (!iwl_mld_is_vo_vi_pkt(hdr)) + return; + + counters = &mld->low_latency.pkts_counters[queue]; + + spin_lock_bh(&counters->lock); + counters->vo_vi[fw_id]++; + spin_unlock_bh(&counters->lock); + + /* Initialize the window_start on the first vo/vi packet */ + if (!mld->low_latency.window_start[fw_id]) + mld->low_latency.window_start[fw_id] = ts; + + if (time_is_before_jiffies(mld->low_latency.timestamp + MLD_LL_PERIOD)) + wiphy_delayed_work_queue(mld->wiphy, &mld->low_latency.work, + 0); +} + +void iwl_mld_low_latency_stop(struct iwl_mld *mld) +{ + lockdep_assert_wiphy(mld->wiphy); + + mld->low_latency.stopped = true; + + wiphy_delayed_work_cancel(mld->wiphy, &mld->low_latency.work); +} + +void iwl_mld_low_latency_restart(struct iwl_mld *mld) +{ + struct iwl_mld_low_latency *ll = &mld->low_latency; + bool low_latency = false; + unsigned long ts = jiffies; + + lockdep_assert_wiphy(mld->wiphy); + + ll->timestamp = ts; + mld->low_latency.stopped = false; + + for (int mac = 0; mac < NUM_MAC_INDEX_DRIVER; mac++) { + ll->window_start[mac] = 0; + low_latency |= ll->result[mac]; + + for (int q = 0; q < mld->trans->num_rx_queues; q++) { + spin_lock_bh(&ll->pkts_counters[q].lock); + ll->pkts_counters[q].vo_vi[mac] = 0; + spin_unlock_bh(&ll->pkts_counters[q].lock); + } + } + + /* if low latency is active, force re-evaluation to cover the case of + * no traffic. + */ + if (low_latency) + wiphy_delayed_work_queue(mld->wiphy, &ll->work, MLD_LL_PERIOD); +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/low_latency.h b/drivers/net/wireless/intel/iwlwifi/mld/low_latency.h new file mode 100644 index 0000000000000..f59684d235af1 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/low_latency.h @@ -0,0 +1,68 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024 Intel Corporation + */ +#ifndef __iwl_mld_low_latency_h__ +#define __iwl_mld_low_latency_h__ + +/** + * struct iwl_mld_low_latency_packets_counters - Packets counters + * @lock: synchronize the counting in data path against the worker + * @vo_vi: per-mac, counts the number of TX and RX voice and video packets + */ +struct iwl_mld_low_latency_packets_counters { + spinlock_t lock; + u32 vo_vi[NUM_MAC_INDEX_DRIVER]; +} ____cacheline_aligned_in_smp; + +/** + * enum iwl_mld_low_latency_cause - low-latency set causes + * + * @LOW_LATENCY_TRAFFIC: indicates low-latency traffic was detected + * @LOW_LATENCY_DEBUGFS: low-latency mode set from debugfs + * @LOW_LATENCY_VIF_TYPE: low-latency mode set because of vif type (AP) + */ +enum iwl_mld_low_latency_cause { + LOW_LATENCY_TRAFFIC = BIT(0), + LOW_LATENCY_DEBUGFS = BIT(1), + LOW_LATENCY_VIF_TYPE = BIT(2), +}; + +/** + * struct iwl_mld_low_latency - Manage low-latency detection and activation. + * @work: this work is used to detect low-latency by monitoring the number of + * voice and video packets transmitted in a period of time. If the + * threshold is reached, low-latency is activated. When active, + * it is deactivated if the threshold is not reached within a + * 10-second period. + * @timestamp: timestamp of the last update. + * @window_start: per-mac, timestamp of the start of the current window. when + * the window is over, the counters are reset. + * @pkts_counters: per-queue array voice/video packet counters + * @result: per-mac latest low-latency result + * @stopped: if true, ignore the requests to update the counters + */ +struct iwl_mld_low_latency { + struct wiphy_delayed_work work; + unsigned long timestamp; + unsigned long window_start[NUM_MAC_INDEX_DRIVER]; + struct iwl_mld_low_latency_packets_counters *pkts_counters; + bool result[NUM_MAC_INDEX_DRIVER]; + bool stopped; +}; + +int iwl_mld_low_latency_init(struct iwl_mld *mld); +void iwl_mld_low_latency_free(struct iwl_mld *mld); +void iwl_mld_low_latency_restart_cleanup(struct iwl_mld *mld); +void iwl_mld_vif_update_low_latency(struct iwl_mld *mld, + struct ieee80211_vif *vif, + bool low_latency, + enum iwl_mld_low_latency_cause cause); +void iwl_mld_low_latency_update_counters(struct iwl_mld *mld, + struct ieee80211_hdr *hdr, + struct ieee80211_sta *sta, + u8 queue); +void iwl_mld_low_latency_stop(struct iwl_mld *mld); +void iwl_mld_low_latency_restart(struct iwl_mld *mld); + +#endif /* __iwl_mld_low_latency_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/mac80211.c b/drivers/net/wireless/intel/iwlwifi/mld/mac80211.c new file mode 100644 index 0000000000000..d73bd0179f7dc --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/mac80211.c @@ -0,0 +1,2671 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024-2025 Intel Corporation + */ + +#include +#include + +#include "mld.h" +#include "mac80211.h" +#include "phy.h" +#include "iface.h" +#include "power.h" +#include "sta.h" +#include "agg.h" +#include "scan.h" +#include "d3.h" +#include "tlc.h" +#include "key.h" +#include "ap.h" +#include "tx.h" +#include "roc.h" +#include "mlo.h" +#include "stats.h" +#include "ftm-initiator.h" +#include "low_latency.h" +#include "fw/api/scan.h" +#include "fw/api/context.h" +#include "fw/api/filter.h" +#include "fw/api/sta.h" +#include "fw/api/tdls.h" +#ifdef CONFIG_PM_SLEEP +#include "fw/api/d3.h" +#endif /* CONFIG_PM_SLEEP */ +#include "iwl-trans.h" + +#define IWL_MLD_LIMITS(ap) \ + { \ + .max = 2, \ + .types = BIT(NL80211_IFTYPE_STATION), \ + }, \ + { \ + .max = 1, \ + .types = ap | \ + BIT(NL80211_IFTYPE_P2P_CLIENT) | \ + BIT(NL80211_IFTYPE_P2P_GO), \ + }, \ + { \ + .max = 1, \ + .types = BIT(NL80211_IFTYPE_P2P_DEVICE), \ + } + +static const struct ieee80211_iface_limit iwl_mld_limits[] = { + IWL_MLD_LIMITS(0) +}; + +static const struct ieee80211_iface_limit iwl_mld_limits_ap[] = { + IWL_MLD_LIMITS(BIT(NL80211_IFTYPE_AP)) +}; + +static const struct ieee80211_iface_combination +iwl_mld_iface_combinations[] = { + { + .num_different_channels = 2, + .max_interfaces = 4, + .limits = iwl_mld_limits, + .n_limits = ARRAY_SIZE(iwl_mld_limits), + }, + { + .num_different_channels = 1, + .max_interfaces = 4, + .limits = iwl_mld_limits_ap, + .n_limits = ARRAY_SIZE(iwl_mld_limits_ap), + }, +}; + +static const u8 if_types_ext_capa_sta[] = { + [0] = WLAN_EXT_CAPA1_EXT_CHANNEL_SWITCHING, + [2] = WLAN_EXT_CAPA3_MULTI_BSSID_SUPPORT, + [7] = WLAN_EXT_CAPA8_OPMODE_NOTIF | + WLAN_EXT_CAPA8_MAX_MSDU_IN_AMSDU_LSB, + [8] = WLAN_EXT_CAPA9_MAX_MSDU_IN_AMSDU_MSB, + [9] = WLAN_EXT_CAPA10_TWT_REQUESTER_SUPPORT, +}; + +#define IWL_MLD_EMLSR_CAPA (IEEE80211_EML_CAP_EMLSR_SUPP | \ + IEEE80211_EML_CAP_EMLSR_PADDING_DELAY_32US << \ + __bf_shf(IEEE80211_EML_CAP_EMLSR_PADDING_DELAY) | \ + IEEE80211_EML_CAP_EMLSR_TRANSITION_DELAY_64US << \ + __bf_shf(IEEE80211_EML_CAP_EMLSR_TRANSITION_DELAY)) +#define IWL_MLD_CAPA_OPS (FIELD_PREP_CONST( \ + IEEE80211_MLD_CAP_OP_TID_TO_LINK_MAP_NEG_SUPP, \ + IEEE80211_MLD_CAP_OP_TID_TO_LINK_MAP_NEG_SUPP_SAME) | \ + IEEE80211_MLD_CAP_OP_LINK_RECONF_SUPPORT) + +static const struct wiphy_iftype_ext_capab iftypes_ext_capa[] = { + { + .iftype = NL80211_IFTYPE_STATION, + .extended_capabilities = if_types_ext_capa_sta, + .extended_capabilities_mask = if_types_ext_capa_sta, + .extended_capabilities_len = sizeof(if_types_ext_capa_sta), + /* relevant only if EHT is supported */ + .eml_capabilities = IWL_MLD_EMLSR_CAPA, + .mld_capa_and_ops = IWL_MLD_CAPA_OPS, + }, +}; + +static void iwl_mld_hw_set_addresses(struct iwl_mld *mld) +{ + struct wiphy *wiphy = mld->wiphy; + int num_addrs = 1; + + /* Extract MAC address */ + memcpy(mld->addresses[0].addr, mld->nvm_data->hw_addr, ETH_ALEN); + wiphy->addresses = mld->addresses; + wiphy->n_addresses = 1; + + /* Extract additional MAC addresses if available */ + if (mld->nvm_data->n_hw_addrs > 1) + num_addrs = min(mld->nvm_data->n_hw_addrs, + IWL_MLD_MAX_ADDRESSES); + + for (int i = 1; i < num_addrs; i++) { + memcpy(mld->addresses[i].addr, + mld->addresses[i - 1].addr, + ETH_ALEN); + mld->addresses[i].addr[ETH_ALEN - 1]++; + wiphy->n_addresses++; + } +} + +static void iwl_mld_hw_set_channels(struct iwl_mld *mld) +{ + struct wiphy *wiphy = mld->wiphy; + struct ieee80211_supported_band *bands = mld->nvm_data->bands; + + wiphy->bands[NL80211_BAND_2GHZ] = &bands[NL80211_BAND_2GHZ]; + wiphy->bands[NL80211_BAND_5GHZ] = &bands[NL80211_BAND_5GHZ]; + + if (bands[NL80211_BAND_6GHZ].n_channels) + wiphy->bands[NL80211_BAND_6GHZ] = &bands[NL80211_BAND_6GHZ]; +} + +static void iwl_mld_hw_set_security(struct iwl_mld *mld) +{ + struct ieee80211_hw *hw = mld->hw; + static const u32 mld_ciphers[] = { + WLAN_CIPHER_SUITE_WEP40, + WLAN_CIPHER_SUITE_WEP104, + WLAN_CIPHER_SUITE_TKIP, + WLAN_CIPHER_SUITE_CCMP, + WLAN_CIPHER_SUITE_GCMP, + WLAN_CIPHER_SUITE_GCMP_256, + WLAN_CIPHER_SUITE_AES_CMAC, + WLAN_CIPHER_SUITE_BIP_GMAC_128, + WLAN_CIPHER_SUITE_BIP_GMAC_256 + }; + + hw->wiphy->n_cipher_suites = ARRAY_SIZE(mld_ciphers); + hw->wiphy->cipher_suites = mld_ciphers; + + ieee80211_hw_set(hw, MFP_CAPABLE); + wiphy_ext_feature_set(hw->wiphy, + NL80211_EXT_FEATURE_BEACON_PROTECTION); +} + +static void iwl_mld_hw_set_regulatory(struct iwl_mld *mld) +{ + struct wiphy *wiphy = mld->wiphy; + + wiphy->regulatory_flags |= REGULATORY_WIPHY_SELF_MANAGED; + wiphy->regulatory_flags |= REGULATORY_ENABLE_RELAX_NO_IR; +} + +static void iwl_mld_hw_set_antennas(struct iwl_mld *mld) +{ + struct wiphy *wiphy = mld->wiphy; + + wiphy->available_antennas_tx = iwl_mld_get_valid_tx_ant(mld); + wiphy->available_antennas_rx = iwl_mld_get_valid_rx_ant(mld); +} + +static void iwl_mld_hw_set_pm(struct iwl_mld *mld) +{ +#ifdef CONFIG_PM_SLEEP + struct wiphy *wiphy = mld->wiphy; + + if (!device_can_wakeup(mld->trans->dev)) + return; + + mld->wowlan.flags |= WIPHY_WOWLAN_MAGIC_PKT | + WIPHY_WOWLAN_DISCONNECT | + WIPHY_WOWLAN_EAP_IDENTITY_REQ | + WIPHY_WOWLAN_RFKILL_RELEASE | + WIPHY_WOWLAN_NET_DETECT | + WIPHY_WOWLAN_SUPPORTS_GTK_REKEY | + WIPHY_WOWLAN_GTK_REKEY_FAILURE | + WIPHY_WOWLAN_4WAY_HANDSHAKE; + + mld->wowlan.n_patterns = IWL_WOWLAN_MAX_PATTERNS; + mld->wowlan.pattern_min_len = IWL_WOWLAN_MIN_PATTERN_LEN; + mld->wowlan.pattern_max_len = IWL_WOWLAN_MAX_PATTERN_LEN; + mld->wowlan.max_nd_match_sets = IWL_SCAN_MAX_PROFILES_V2; + + wiphy->wowlan = &mld->wowlan; +#endif /* CONFIG_PM_SLEEP */ +} + +static void iwl_mac_hw_set_radiotap(struct iwl_mld *mld) +{ + struct ieee80211_hw *hw = mld->hw; + + hw->radiotap_mcs_details |= IEEE80211_RADIOTAP_MCS_HAVE_FEC | + IEEE80211_RADIOTAP_MCS_HAVE_STBC; + + hw->radiotap_vht_details |= IEEE80211_RADIOTAP_VHT_KNOWN_STBC | + IEEE80211_RADIOTAP_VHT_KNOWN_BEAMFORMED; + + hw->radiotap_timestamp.units_pos = + IEEE80211_RADIOTAP_TIMESTAMP_UNIT_US | + IEEE80211_RADIOTAP_TIMESTAMP_SPOS_PLCP_SIG_ACQ; + + /* this is the case for CCK frames, it's better (only 8) for OFDM */ + hw->radiotap_timestamp.accuracy = 22; +} + +static void iwl_mac_hw_set_flags(struct iwl_mld *mld) +{ + struct ieee80211_hw *hw = mld->hw; + + ieee80211_hw_set(hw, USES_RSS); + ieee80211_hw_set(hw, HANDLES_QUIET_CSA); + ieee80211_hw_set(hw, AP_LINK_PS); + ieee80211_hw_set(hw, SIGNAL_DBM); + ieee80211_hw_set(hw, SPECTRUM_MGMT); + ieee80211_hw_set(hw, REPORTS_TX_ACK_STATUS); + ieee80211_hw_set(hw, WANT_MONITOR_VIF); + ieee80211_hw_set(hw, SUPPORTS_PS); + ieee80211_hw_set(hw, SUPPORTS_DYNAMIC_PS); + ieee80211_hw_set(hw, AMPDU_AGGREGATION); + ieee80211_hw_set(hw, CONNECTION_MONITOR); + ieee80211_hw_set(hw, CHANCTX_STA_CSA); + ieee80211_hw_set(hw, SUPPORT_FAST_XMIT); + ieee80211_hw_set(hw, SUPPORTS_CLONED_SKBS); + ieee80211_hw_set(hw, NEEDS_UNIQUE_STA_ADDR); + ieee80211_hw_set(hw, SUPPORTS_VHT_EXT_NSS_BW); + ieee80211_hw_set(hw, BUFF_MMPDU_TXQ); + ieee80211_hw_set(hw, STA_MMPDU_TXQ); + ieee80211_hw_set(hw, TX_AMSDU); + ieee80211_hw_set(hw, TX_FRAG_LIST); + ieee80211_hw_set(hw, TX_AMPDU_SETUP_IN_HW); + ieee80211_hw_set(hw, HAS_RATE_CONTROL); + ieee80211_hw_set(hw, SUPPORTS_REORDERING_BUFFER); + ieee80211_hw_set(hw, DISALLOW_PUNCTURING_5GHZ); + ieee80211_hw_set(hw, SINGLE_SCAN_ON_ALL_BANDS); + ieee80211_hw_set(hw, SUPPORTS_AMSDU_IN_AMPDU); + ieee80211_hw_set(hw, TDLS_WIDER_BW); +} + +static void iwl_mac_hw_set_wiphy(struct iwl_mld *mld) +{ + struct ieee80211_hw *hw = mld->hw; + struct wiphy *wiphy = hw->wiphy; + const struct iwl_ucode_capabilities *ucode_capa = &mld->fw->ucode_capa; + + snprintf(wiphy->fw_version, + sizeof(wiphy->fw_version), + "%.31s", mld->fw->fw_version); + + wiphy->interface_modes = BIT(NL80211_IFTYPE_STATION) | + BIT(NL80211_IFTYPE_P2P_CLIENT) | + BIT(NL80211_IFTYPE_AP) | + BIT(NL80211_IFTYPE_P2P_GO) | + BIT(NL80211_IFTYPE_P2P_DEVICE) | + BIT(NL80211_IFTYPE_ADHOC); + + wiphy->features |= NL80211_FEATURE_SCHED_SCAN_RANDOM_MAC_ADDR | + NL80211_FEATURE_SCAN_RANDOM_MAC_ADDR | + NL80211_FEATURE_ND_RANDOM_MAC_ADDR | + NL80211_FEATURE_HT_IBSS | + NL80211_FEATURE_P2P_GO_CTWIN | + NL80211_FEATURE_LOW_PRIORITY_SCAN | + NL80211_FEATURE_P2P_GO_OPPPS | + NL80211_FEATURE_AP_MODE_CHAN_WIDTH_CHANGE | + NL80211_FEATURE_SUPPORTS_WMM_ADMISSION | + NL80211_FEATURE_TX_POWER_INSERTION | + NL80211_FEATURE_DS_PARAM_SET_IE_IN_PROBES; + + wiphy->flags |= WIPHY_FLAG_IBSS_RSN | + WIPHY_FLAG_AP_UAPSD | + WIPHY_FLAG_HAS_CHANNEL_SWITCH | + WIPHY_FLAG_SPLIT_SCAN_6GHZ | + WIPHY_FLAG_SUPPORTS_TDLS | + WIPHY_FLAG_SUPPORTS_EXT_KEK_KCK; + + if (mld->nvm_data->sku_cap_11be_enable && + !iwlwifi_mod_params.disable_11ax && + !iwlwifi_mod_params.disable_11be) + wiphy->flags |= WIPHY_FLAG_SUPPORTS_MLO; + + /* the firmware uses u8 for num of iterations, but 0xff is saved for + * infinite loop, so the maximum number of iterations is actually 254. + */ + wiphy->max_sched_scan_plan_iterations = 254; + wiphy->max_sched_scan_ie_len = iwl_mld_scan_max_template_size(); + wiphy->max_scan_ie_len = iwl_mld_scan_max_template_size(); + wiphy->max_sched_scan_ssids = PROBE_OPTION_MAX; + wiphy->max_scan_ssids = PROBE_OPTION_MAX; + wiphy->max_sched_scan_plans = IWL_MAX_SCHED_SCAN_PLANS; + wiphy->max_sched_scan_reqs = 1; + wiphy->max_sched_scan_plan_interval = U16_MAX; + wiphy->max_match_sets = IWL_SCAN_MAX_PROFILES_V2; + + wiphy->max_remain_on_channel_duration = 10000; + + wiphy->hw_version = mld->trans->hw_id; + + wiphy->hw_timestamp_max_peers = 1; + + wiphy->iface_combinations = iwl_mld_iface_combinations; + wiphy->n_iface_combinations = ARRAY_SIZE(iwl_mld_iface_combinations); + + wiphy_ext_feature_set(wiphy, NL80211_EXT_FEATURE_VHT_IBSS); + wiphy_ext_feature_set(wiphy, NL80211_EXT_FEATURE_DFS_CONCURRENT); + wiphy_ext_feature_set(wiphy, NL80211_EXT_FEATURE_BEACON_RATE_LEGACY); + wiphy_ext_feature_set(wiphy, NL80211_EXT_FEATURE_SCAN_START_TIME); + wiphy_ext_feature_set(wiphy, NL80211_EXT_FEATURE_BSS_PARENT_TSF); + wiphy_ext_feature_set(wiphy, NL80211_EXT_FEATURE_SCAN_MIN_PREQ_CONTENT); + wiphy_ext_feature_set(wiphy, NL80211_EXT_FEATURE_ACCEPT_BCAST_PROBE_RESP); + wiphy_ext_feature_set(wiphy, NL80211_EXT_FEATURE_FILS_MAX_CHANNEL_TIME); + wiphy_ext_feature_set(wiphy, NL80211_EXT_FEATURE_OCE_PROBE_REQ_HIGH_TX_RATE); + wiphy_ext_feature_set(wiphy, NL80211_EXT_FEATURE_MU_MIMO_AIR_SNIFFER); + wiphy_ext_feature_set(wiphy, NL80211_EXT_FEATURE_SPP_AMSDU_SUPPORT); + + if (fw_has_capa(ucode_capa, IWL_UCODE_TLV_CAPA_PROTECTED_TWT)) + wiphy_ext_feature_set(wiphy, NL80211_EXT_FEATURE_PROTECTED_TWT); + + wiphy->iftype_ext_capab = NULL; + wiphy->num_iftype_ext_capab = 0; + + if (!iwlwifi_mod_params.disable_11ax) { + wiphy->iftype_ext_capab = iftypes_ext_capa; + wiphy->num_iftype_ext_capab = ARRAY_SIZE(iftypes_ext_capa); + + ieee80211_hw_set(hw, SUPPORTS_MULTI_BSSID); + ieee80211_hw_set(hw, SUPPORTS_ONLY_HE_MULTI_BSSID); + } + + if (iwlmld_mod_params.power_scheme != IWL_POWER_SCHEME_CAM) + wiphy->flags |= WIPHY_FLAG_PS_ON_BY_DEFAULT; + else + wiphy->flags &= ~WIPHY_FLAG_PS_ON_BY_DEFAULT; +} + +static void iwl_mac_hw_set_misc(struct iwl_mld *mld) +{ + struct ieee80211_hw *hw = mld->hw; + + hw->queues = IEEE80211_NUM_ACS; + + hw->netdev_features = NETIF_F_HIGHDMA | NETIF_F_SG; + hw->netdev_features |= mld->cfg->features; + + hw->max_tx_fragments = mld->trans->max_skb_frags; + hw->max_listen_interval = IWL_MLD_CONN_LISTEN_INTERVAL; + + hw->uapsd_max_sp_len = IEEE80211_WMM_IE_STA_QOSINFO_SP_ALL; + hw->uapsd_queues = IEEE80211_WMM_IE_STA_QOSINFO_AC_VO | + IEEE80211_WMM_IE_STA_QOSINFO_AC_VI | + IEEE80211_WMM_IE_STA_QOSINFO_AC_BK | + IEEE80211_WMM_IE_STA_QOSINFO_AC_BE; + + hw->chanctx_data_size = sizeof(struct iwl_mld_phy); + hw->vif_data_size = sizeof(struct iwl_mld_vif); + hw->sta_data_size = sizeof(struct iwl_mld_sta); + hw->txq_data_size = sizeof(struct iwl_mld_txq); + + /* TODO: Remove this division when IEEE80211_MAX_AMPDU_BUF_EHT size + * is supported. + * Note: ensure that IWL_DEFAULT_QUEUE_SIZE_EHT is updated accordingly. + */ + hw->max_rx_aggregation_subframes = IEEE80211_MAX_AMPDU_BUF_EHT / 2; +} + +static int iwl_mld_hw_verify_preconditions(struct iwl_mld *mld) +{ + /* 11ax is expected to be enabled for all supported devices */ + if (WARN_ON(!mld->nvm_data->sku_cap_11ax_enable)) + return -EINVAL; + + /* LAR is expected to be enabled for all supported devices */ + if (WARN_ON(!mld->nvm_data->lar_enabled)) + return -EINVAL; + + /* All supported devices are currently using version 3 of the cmd. + * Since version 3, IWL_SCAN_MAX_PROFILES_V2 shall be used where + * necessary. + */ + if (WARN_ON(iwl_fw_lookup_cmd_ver(mld->fw, + SCAN_OFFLOAD_UPDATE_PROFILES_CMD, + IWL_FW_CMD_VER_UNKNOWN) != 3)) + return -EINVAL; + + return 0; +} + +int iwl_mld_register_hw(struct iwl_mld *mld) +{ + /* verify once essential preconditions required for setting + * the hw capabilities + */ + if (iwl_mld_hw_verify_preconditions(mld)) + return -EINVAL; + + iwl_mld_hw_set_addresses(mld); + iwl_mld_hw_set_channels(mld); + iwl_mld_hw_set_security(mld); + iwl_mld_hw_set_regulatory(mld); + iwl_mld_hw_set_pm(mld); + iwl_mld_hw_set_antennas(mld); + iwl_mac_hw_set_radiotap(mld); + iwl_mac_hw_set_flags(mld); + iwl_mac_hw_set_wiphy(mld); + iwl_mac_hw_set_misc(mld); + + SET_IEEE80211_DEV(mld->hw, mld->trans->dev); + + return ieee80211_register_hw(mld->hw); +} + +static void +iwl_mld_mac80211_tx(struct ieee80211_hw *hw, + struct ieee80211_tx_control *control, struct sk_buff *skb) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct ieee80211_sta *sta = control->sta; + struct ieee80211_tx_info *info = IEEE80211_SKB_CB(skb); + struct ieee80211_hdr *hdr = (void *)skb->data; + u32 link_id = u32_get_bits(info->control.flags, + IEEE80211_TX_CTRL_MLO_LINK); + + /* In AP mode, mgmt frames are sent on the bcast station, + * so the FW can't translate the MLD addr to the link addr. Do it here + */ + if (ieee80211_is_mgmt(hdr->frame_control) && sta && + link_id != IEEE80211_LINK_UNSPECIFIED && + !ieee80211_is_probe_resp(hdr->frame_control)) { + /* translate MLD addresses to LINK addresses */ + struct ieee80211_link_sta *link_sta = + rcu_dereference(sta->link[link_id]); + struct ieee80211_bss_conf *link_conf = + rcu_dereference(info->control.vif->link_conf[link_id]); + struct ieee80211_mgmt *mgmt; + + if (WARN_ON(!link_sta || !link_conf)) { + ieee80211_free_txskb(hw, skb); + return; + } + + mgmt = (void *)hdr; + memcpy(mgmt->da, link_sta->addr, ETH_ALEN); + memcpy(mgmt->sa, link_conf->addr, ETH_ALEN); + memcpy(mgmt->bssid, link_conf->bssid, ETH_ALEN); + } + + iwl_mld_tx_skb(mld, skb, NULL); +} + +static void +iwl_mld_restart_cleanup(struct iwl_mld *mld) +{ + iwl_cleanup_mld(mld); + + ieee80211_iterate_interfaces(mld->hw, IEEE80211_IFACE_ITER_ACTIVE, + iwl_mld_cleanup_vif, NULL); + + ieee80211_iterate_stations_atomic(mld->hw, + iwl_mld_cleanup_sta, NULL); + + iwl_mld_ftm_restart_cleanup(mld); +} + +static +int iwl_mld_mac80211_start(struct ieee80211_hw *hw) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + int ret; + bool in_d3 = false; + + lockdep_assert_wiphy(mld->wiphy); + +#ifdef CONFIG_PM_SLEEP + /* Unless the host goes into hibernate the FW always stays on and + * the d3_resume flow is used. When wowlan is configured, mac80211 + * would call it's resume callback and the wowlan_resume flow + * would be used. + */ + + in_d3 = mld->fw_status.in_d3; + if (in_d3) { + /* mac80211 already cleaned up the state, no need for cleanup */ + ret = iwl_mld_no_wowlan_resume(mld); + if (ret) + iwl_mld_stop_fw(mld); + } +#endif /* CONFIG_PM_SLEEP */ + + if (mld->fw_status.in_hw_restart) { + iwl_mld_stop_fw(mld); + iwl_mld_restart_cleanup(mld); + } + + if (!in_d3 || ret) { + ret = iwl_mld_start_fw(mld); + if (ret) + goto error; + } + + mld->scan.last_start_time_jiffies = jiffies; + + iwl_dbg_tlv_time_point(&mld->fwrt, IWL_FW_INI_TIME_POINT_POST_INIT, + NULL); + iwl_dbg_tlv_time_point(&mld->fwrt, IWL_FW_INI_TIME_POINT_PERIODIC, + NULL); + + return 0; + +error: + /* If we failed to restart the hw, there is nothing useful + * we can do but indicate we are no longer in restart. + */ + mld->fw_status.in_hw_restart = false; + + return ret; +} + +static +void iwl_mld_mac80211_stop(struct ieee80211_hw *hw, bool suspend) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + + lockdep_assert_wiphy(mld->wiphy); + + wiphy_work_cancel(mld->wiphy, &mld->add_txqs_wk); + + /* if the suspend flow fails the fw is in error. Stop it here, and it + * will be started upon wakeup + */ + if (!suspend || iwl_mld_no_wowlan_suspend(mld)) + iwl_mld_stop_fw(mld); + + /* HW is stopped, no more coming RX. OTOH, the worker can't run as the + * wiphy lock is held. Cancel it in case it was scheduled just before + * we stopped the HW. + */ + wiphy_work_cancel(mld->wiphy, &mld->async_handlers_wk); + + /* Empty out the list, as the worker won't do that */ + iwl_mld_purge_async_handlers_list(mld); + + /* Clear in_hw_restart flag when stopping the hw, as mac80211 won't + * execute the restart. + */ + mld->fw_status.in_hw_restart = false; + + /* We shouldn't have any UIDs still set. Loop over all the UIDs to + * make sure there's nothing left there and warn if any is found. + */ + for (int i = 0; i < ARRAY_SIZE(mld->scan.uid_status); i++) + if (WARN_ONCE(mld->scan.uid_status[i], + "UMAC scan UID %d status was not cleaned (0x%x 0x%x)\n", + i, mld->scan.uid_status[i], mld->scan.status)) + mld->scan.uid_status[i] = 0; +} + +static +int iwl_mld_mac80211_config(struct ieee80211_hw *hw, u32 changed) +{ + return 0; +} + +static +int iwl_mld_mac80211_add_interface(struct ieee80211_hw *hw, + struct ieee80211_vif *vif) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + /* Construct mld_vif, add it to fw, and map its ID to ieee80211_vif */ + ret = iwl_mld_add_vif(mld, vif); + if (ret) + return ret; + + /* + * Add the default link, but not if this is an MLD vif as that implies + * the HW is restarting and it will be configured by change_vif_links. + */ + if (!ieee80211_vif_is_mld(vif)) + ret = iwl_mld_add_link(mld, &vif->bss_conf); + if (ret) + goto err; + + if (vif->type == NL80211_IFTYPE_STATION) { + vif->driver_flags |= IEEE80211_VIF_REMOVE_AP_AFTER_DISASSOC; + if (!vif->p2p) + vif->driver_flags |= IEEE80211_VIF_BEACON_FILTER | + IEEE80211_VIF_SUPPORTS_CQM_RSSI; + } + + if (vif->p2p || iwl_fw_lookup_cmd_ver(mld->fw, PHY_CONTEXT_CMD, 0) < 5) + vif->driver_flags |= IEEE80211_VIF_IGNORE_OFDMA_WIDER_BW; + + /* + * For an MLD vif (in restart) we may not have a link; delay the call + * the initial change_vif_links. + */ + if (vif->type == NL80211_IFTYPE_STATION && + !ieee80211_vif_is_mld(vif)) + iwl_mld_update_mac_power(mld, vif, false); + + if (vif->type == NL80211_IFTYPE_MONITOR) { + mld->monitor.on = true; + ieee80211_hw_set(mld->hw, RX_INCLUDES_FCS); + } + + if (vif->type == NL80211_IFTYPE_P2P_DEVICE) + mld->p2p_device_vif = vif; + + return 0; + +err: + iwl_mld_rm_vif(mld, vif); + return ret; +} + +static +void iwl_mld_mac80211_remove_interface(struct ieee80211_hw *hw, + struct ieee80211_vif *vif) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + + lockdep_assert_wiphy(mld->wiphy); + + if (ieee80211_vif_type_p2p(vif) == NL80211_IFTYPE_STATION) + vif->driver_flags &= ~(IEEE80211_VIF_BEACON_FILTER | + IEEE80211_VIF_SUPPORTS_CQM_RSSI); + + if (vif->type == NL80211_IFTYPE_MONITOR) { + __clear_bit(IEEE80211_HW_RX_INCLUDES_FCS, mld->hw->flags); + mld->monitor.on = false; + } + + if (vif->type == NL80211_IFTYPE_P2P_DEVICE) + mld->p2p_device_vif = NULL; + + iwl_mld_remove_link(mld, &vif->bss_conf); + +#ifdef CONFIG_IWLWIFI_DEBUGFS + debugfs_remove(iwl_mld_vif_from_mac80211(vif)->dbgfs_slink); +#endif + + iwl_mld_rm_vif(mld, vif); +} + +struct iwl_mld_mc_iter_data { + struct iwl_mld *mld; + int port_id; +}; + +static void iwl_mld_mc_iface_iterator(void *data, u8 *mac, + struct ieee80211_vif *vif) +{ + struct iwl_mld_mc_iter_data *mc_data = data; + struct iwl_mld *mld = mc_data->mld; + struct iwl_mcast_filter_cmd *cmd = mld->mcast_filter_cmd; + struct iwl_host_cmd hcmd = { + .id = MCAST_FILTER_CMD, + .dataflags[0] = IWL_HCMD_DFL_NOCOPY, + }; + int ret, len; + + /* If we don't have free ports, mcast frames will be dropped */ + if (WARN_ON_ONCE(mc_data->port_id >= MAX_PORT_ID_NUM)) + return; + + if (vif->type != NL80211_IFTYPE_STATION || !vif->cfg.assoc) + return; + + cmd->port_id = mc_data->port_id++; + ether_addr_copy(cmd->bssid, vif->bss_conf.bssid); + len = roundup(sizeof(*cmd) + cmd->count * ETH_ALEN, 4); + + hcmd.len[0] = len; + hcmd.data[0] = cmd; + + ret = iwl_mld_send_cmd(mld, &hcmd); + if (ret) + IWL_ERR(mld, "mcast filter cmd error. ret=%d\n", ret); +} + +void iwl_mld_recalc_multicast_filter(struct iwl_mld *mld) +{ + struct iwl_mld_mc_iter_data iter_data = { + .mld = mld, + }; + + if (WARN_ON_ONCE(!mld->mcast_filter_cmd)) + return; + + ieee80211_iterate_active_interfaces(mld->hw, + IEEE80211_IFACE_ITER_NORMAL, + iwl_mld_mc_iface_iterator, + &iter_data); +} + +static u64 +iwl_mld_mac80211_prepare_multicast(struct ieee80211_hw *hw, + struct netdev_hw_addr_list *mc_list) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct iwl_mcast_filter_cmd *cmd; + struct netdev_hw_addr *addr; + int addr_count = netdev_hw_addr_list_count(mc_list); + bool pass_all = addr_count > MAX_MCAST_FILTERING_ADDRESSES; + int len; + + if (pass_all) + addr_count = 0; + + /* len must be a multiple of 4 */ + len = roundup(sizeof(*cmd) + addr_count * ETH_ALEN, 4); + cmd = kzalloc(len, GFP_ATOMIC); + if (!cmd) + return 0; + + if (pass_all) { + cmd->pass_all = 1; + goto out; + } + + netdev_hw_addr_list_for_each(addr, mc_list) { + IWL_DEBUG_MAC80211(mld, "mcast addr (%d): %pM\n", + cmd->count, addr->addr); + ether_addr_copy(&cmd->addr_list[cmd->count * ETH_ALEN], + addr->addr); + cmd->count++; + } + +out: + return (u64)(unsigned long)cmd; +} + +static +void iwl_mld_mac80211_configure_filter(struct ieee80211_hw *hw, + unsigned int changed_flags, + unsigned int *total_flags, + u64 multicast) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct iwl_mcast_filter_cmd *cmd = (void *)(unsigned long)multicast; + + /* Replace previous configuration */ + kfree(mld->mcast_filter_cmd); + mld->mcast_filter_cmd = cmd; + + if (!cmd) + goto out; + + if (changed_flags & FIF_ALLMULTI) + cmd->pass_all = !!(*total_flags & FIF_ALLMULTI); + + if (cmd->pass_all) + cmd->count = 0; + + iwl_mld_recalc_multicast_filter(mld); +out: + *total_flags = 0; +} + +static +void iwl_mld_mac80211_wake_tx_queue(struct ieee80211_hw *hw, + struct ieee80211_txq *txq) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct iwl_mld_txq *mld_txq = iwl_mld_txq_from_mac80211(txq); + + if (likely(mld_txq->status.allocated) || !txq->sta) { + iwl_mld_tx_from_txq(mld, txq); + return; + } + + /* We don't support TSPEC tids. %IEEE80211_NUM_TIDS is for mgmt */ + if (txq->tid != IEEE80211_NUM_TIDS && txq->tid >= IWL_MAX_TID_COUNT) { + IWL_DEBUG_MAC80211(mld, "TID %d is not supported\n", txq->tid); + return; + } + + /* The worker will handle any packets we leave on the txq now */ + + spin_lock_bh(&mld->add_txqs_lock); + /* The list is being deleted only after the queue is fully allocated. */ + if (list_empty(&mld_txq->list) && + /* recheck under lock, otherwise it can be added twice */ + !mld_txq->status.allocated) { + list_add_tail(&mld_txq->list, &mld->txqs_to_add); + wiphy_work_queue(mld->wiphy, &mld->add_txqs_wk); + } + spin_unlock_bh(&mld->add_txqs_lock); +} + +static void iwl_mld_teardown_tdls_peers(struct iwl_mld *mld) +{ + lockdep_assert_wiphy(mld->wiphy); + + for (int i = 0; i < mld->fw->ucode_capa.num_stations; i++) { + struct ieee80211_link_sta *link_sta; + struct iwl_mld_sta *mld_sta; + + link_sta = wiphy_dereference(mld->wiphy, + mld->fw_id_to_link_sta[i]); + if (IS_ERR_OR_NULL(link_sta)) + continue; + + if (!link_sta->sta->tdls) + continue; + + mld_sta = iwl_mld_sta_from_mac80211(link_sta->sta); + + ieee80211_tdls_oper_request(mld_sta->vif, link_sta->addr, + NL80211_TDLS_TEARDOWN, + WLAN_REASON_TDLS_TEARDOWN_UNSPECIFIED, + GFP_KERNEL); + } +} + +static +int iwl_mld_add_chanctx(struct ieee80211_hw *hw, + struct ieee80211_chanctx_conf *ctx) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct iwl_mld_phy *phy = iwl_mld_phy_from_mac80211(ctx); + int fw_id = iwl_mld_allocate_fw_phy_id(mld); + int ret; + + if (fw_id < 0) + return fw_id; + + phy->fw_id = fw_id; + phy->chandef = *iwl_mld_get_chandef_from_chanctx(ctx); + + ret = iwl_mld_phy_fw_action(mld, ctx, FW_CTXT_ACTION_ADD); + if (ret) { + mld->used_phy_ids &= ~BIT(phy->fw_id); + return ret; + } + + if (hweight8(mld->used_phy_ids) > 1) + iwl_mld_teardown_tdls_peers(mld); + + return 0; +} + +static +void iwl_mld_remove_chanctx(struct ieee80211_hw *hw, + struct ieee80211_chanctx_conf *ctx) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct iwl_mld_phy *phy = iwl_mld_phy_from_mac80211(ctx); + + iwl_mld_phy_fw_action(mld, ctx, FW_CTXT_ACTION_REMOVE); + mld->used_phy_ids &= ~BIT(phy->fw_id); +} + +static +void iwl_mld_change_chanctx(struct ieee80211_hw *hw, + struct ieee80211_chanctx_conf *ctx, u32 changed) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct iwl_mld_phy *phy = iwl_mld_phy_from_mac80211(ctx); + struct cfg80211_chan_def *chandef = + iwl_mld_get_chandef_from_chanctx(ctx); + + /* We don't care about these */ + if (!(changed & ~(IEEE80211_CHANCTX_CHANGE_RX_CHAINS | + IEEE80211_CHANCTX_CHANGE_RADAR | + IEEE80211_CHANCTX_CHANGE_CHANNEL))) + return; + + /* Check if a FW update is required */ + + if (changed & IEEE80211_CHANCTX_CHANGE_AP) + goto update; + + if (chandef->chan == phy->chandef.chan && + chandef->center_freq1 == phy->chandef.center_freq1 && + chandef->punctured == phy->chandef.punctured) { + /* Check if we are toggling between HT and non-HT, no-op */ + if (phy->chandef.width == chandef->width || + (phy->chandef.width <= NL80211_CHAN_WIDTH_20 && + chandef->width <= NL80211_CHAN_WIDTH_20)) + return; + } +update: + phy->chandef = *chandef; + + iwl_mld_phy_fw_action(mld, ctx, FW_CTXT_ACTION_MODIFY); +} + +static u8 +iwl_mld_chandef_get_primary_80(struct cfg80211_chan_def *chandef) +{ + int data_start; + int control_start; + int bw; + + if (chandef->width == NL80211_CHAN_WIDTH_320) + bw = 320; + else if (chandef->width == NL80211_CHAN_WIDTH_160) + bw = 160; + else + return 0; + + /* data is bw wide so the start is half the width */ + data_start = chandef->center_freq1 - bw / 2; + /* control is 20Mhz width */ + control_start = chandef->chan->center_freq - 10; + + return (control_start - data_start) / 80; +} + +static bool iwl_mld_can_activate_link(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mld_sta *mld_sta; + struct iwl_mld_link_sta *link_sta; + + /* In association, we activate the assoc link before adding the STA. */ + if (!mld_vif->ap_sta || !vif->cfg.assoc) + return true; + + mld_sta = iwl_mld_sta_from_mac80211(mld_vif->ap_sta); + + /* When switching links, we need to wait with the activation until the + * STA was added to the FW. It'll be activated in + * iwl_mld_update_link_stas + */ + link_sta = wiphy_dereference(mld->wiphy, mld_sta->link[link->link_id]); + + /* In restart we can have a link_sta that doesn't exist in FW yet */ + return link_sta && link_sta->in_fw; +} + +static +int iwl_mld_assign_vif_chanctx(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link, + struct ieee80211_chanctx_conf *ctx) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct iwl_mld_link *mld_link = iwl_mld_link_from_mac80211(link); + unsigned int n_active = iwl_mld_count_active_links(mld, vif); + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + if (WARN_ON(!mld_link)) + return -EINVAL; + + /* if the assigned one was not counted yet, count it now */ + if (!rcu_access_pointer(mld_link->chan_ctx)) { + n_active++; + + /* Track addition of non-BSS link */ + if (ieee80211_vif_type_p2p(vif) != NL80211_IFTYPE_STATION) { + ret = iwl_mld_emlsr_check_non_bss_block(mld, 1); + if (ret) + return ret; + } + } + + /* for AP, mac parameters such as HE support are updated at this stage. */ + if (vif->type == NL80211_IFTYPE_AP) { + ret = iwl_mld_mac_fw_action(mld, vif, FW_CTXT_ACTION_MODIFY); + + if (ret) { + IWL_ERR(mld, "failed to update MAC %pM\n", vif->addr); + return -EINVAL; + } + } + + rcu_assign_pointer(mld_link->chan_ctx, ctx); + + if (n_active > 1) { + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + + iwl_mld_leave_omi_bw_reduction(mld); + + /* Indicate to mac80211 that EML is enabled */ + vif->driver_flags |= IEEE80211_VIF_EML_ACTIVE; + + if (vif->active_links & BIT(mld_vif->emlsr.selected_links)) + mld_vif->emlsr.primary = mld_vif->emlsr.selected_primary; + else + mld_vif->emlsr.primary = __ffs(vif->active_links); + + iwl_dbg_tlv_time_point(&mld->fwrt, IWL_FW_INI_TIME_ESR_LINK_UP, + NULL); + } + + /* First send the link command with the phy context ID. + * Now that we have the phy, we know the band so also the rates + */ + ret = iwl_mld_change_link_in_fw(mld, link, + LINK_CONTEXT_MODIFY_RATES_INFO); + if (ret) + goto err; + + /* TODO: Initialize rate control for the AP station, since we might be + * doing a link switch here - we cannot initialize it before since + * this needs the phy context assigned (and in FW?), and we cannot + * do it later because it needs to be initialized as soon as we're + * able to TX on the link, i.e. when active. (task=link-switch) + */ + + /* Now activate the link */ + if (iwl_mld_can_activate_link(mld, vif, link)) { + ret = iwl_mld_activate_link(mld, link); + if (ret) + goto err; + } + + if (vif->type == NL80211_IFTYPE_STATION) + iwl_mld_send_ap_tx_power_constraint_cmd(mld, vif, link); + + if (vif->type == NL80211_IFTYPE_MONITOR) { + /* TODO: task=sniffer add sniffer station */ + mld->monitor.p80 = + iwl_mld_chandef_get_primary_80(&vif->bss_conf.chanreq.oper); + } + + return 0; +err: + RCU_INIT_POINTER(mld_link->chan_ctx, NULL); + return ret; +} + +static +void iwl_mld_unassign_vif_chanctx(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link, + struct ieee80211_chanctx_conf *ctx) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mld_link *mld_link = iwl_mld_link_from_mac80211(link); + unsigned int n_active = iwl_mld_count_active_links(mld, vif); + + if (WARN_ON(!mld_link)) + return; + + /* Track removal of non-BSS link */ + if (ieee80211_vif_type_p2p(vif) != NL80211_IFTYPE_STATION) + iwl_mld_emlsr_check_non_bss_block(mld, -1); + + iwl_mld_deactivate_link(mld, link); + + /* TODO: task=sniffer remove sniffer station */ + + if (n_active > 1) { + /* Indicate to mac80211 that EML is disabled */ + vif->driver_flags &= ~IEEE80211_VIF_EML_ACTIVE; + + iwl_dbg_tlv_time_point(&mld->fwrt, + IWL_FW_INI_TIME_ESR_LINK_DOWN, + NULL); + } + + RCU_INIT_POINTER(mld_link->chan_ctx, NULL); + + /* in the non-MLO case, remove/re-add the link to clean up FW state. + * In MLO, it'll be done in drv_change_vif_link + */ + if (!ieee80211_vif_is_mld(vif) && !mld_vif->ap_sta && + !WARN_ON_ONCE(vif->cfg.assoc) && + vif->type != NL80211_IFTYPE_AP && !mld->fw_status.in_hw_restart) { + iwl_mld_remove_link(mld, link); + iwl_mld_add_link(mld, link); + } +} + +static +int iwl_mld_mac80211_set_rts_threshold(struct ieee80211_hw *hw, u32 value) +{ + return 0; +} + +static void +iwl_mld_link_info_changed_ap_ibss(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link, + u64 changes) +{ + u32 link_changes = 0; + + if (changes & BSS_CHANGED_ERP_SLOT) + link_changes |= LINK_CONTEXT_MODIFY_RATES_INFO; + + if (changes & (BSS_CHANGED_ERP_CTS_PROT | BSS_CHANGED_HT)) + link_changes |= LINK_CONTEXT_MODIFY_PROTECT_FLAGS; + + if (changes & (BSS_CHANGED_QOS | BSS_CHANGED_BANDWIDTH)) + link_changes |= LINK_CONTEXT_MODIFY_QOS_PARAMS; + + if (changes & BSS_CHANGED_HE_BSS_COLOR) + link_changes |= LINK_CONTEXT_MODIFY_HE_PARAMS; + + if (link_changes) + iwl_mld_change_link_in_fw(mld, link, link_changes); + + if (changes & BSS_CHANGED_BEACON) + iwl_mld_update_beacon_template(mld, vif, link); +} + +static +u32 iwl_mld_link_changed_mapping(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link_conf, + u64 changes) +{ + u32 link_changes = 0; + bool has_he, has_eht; + + if (changes & BSS_CHANGED_QOS && vif->cfg.assoc && link_conf->qos) + link_changes |= LINK_CONTEXT_MODIFY_QOS_PARAMS; + + if (changes & (BSS_CHANGED_ERP_PREAMBLE | BSS_CHANGED_BASIC_RATES | + BSS_CHANGED_ERP_SLOT)) + link_changes |= LINK_CONTEXT_MODIFY_RATES_INFO; + + if (changes & (BSS_CHANGED_HT | BSS_CHANGED_ERP_CTS_PROT)) + link_changes |= LINK_CONTEXT_MODIFY_PROTECT_FLAGS; + + /* TODO: task=MLO check mac80211's HE flags and if command is needed + * every time there's a link change. Currently used flags are + * BSS_CHANGED_HE_OBSS_PD and BSS_CHANGED_HE_BSS_COLOR. + */ + has_he = link_conf->he_support && !iwlwifi_mod_params.disable_11ax; + has_eht = link_conf->eht_support && !iwlwifi_mod_params.disable_11be; + + if (vif->cfg.assoc && (has_he || has_eht)) { + IWL_DEBUG_MAC80211(mld, "Associated in HE mode\n"); + link_changes |= LINK_CONTEXT_MODIFY_HE_PARAMS; + } + + return link_changes; +} + +static void +iwl_mld_mac80211_link_info_changed_sta(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link_conf, + u64 changes) +{ + u32 link_changes = iwl_mld_link_changed_mapping(mld, vif, link_conf, + changes); + + if (link_changes) + iwl_mld_change_link_in_fw(mld, link_conf, link_changes); + + if (changes & BSS_CHANGED_TPE) + iwl_mld_send_ap_tx_power_constraint_cmd(mld, vif, link_conf); + + if (changes & BSS_CHANGED_BEACON_INFO) + iwl_mld_update_mac_power(mld, vif, false); + + /* The firmware will wait quite a while after association before it + * starts filtering the beacons. We can safely enable beacon filtering + * upon CQM configuration, even if we didn't get a beacon yet. + */ + if (changes & (BSS_CHANGED_CQM | BSS_CHANGED_BEACON_INFO)) + iwl_mld_enable_beacon_filter(mld, link_conf, false); + + /* If we have used OMI before to reduce bandwidth to 80 MHz and then + * increased to 160 MHz again, and then the AP changes to 320 MHz, it + * will think that we're limited to 160 MHz right now. Update it by + * requesting a new OMI bandwidth. + */ + if (changes & BSS_CHANGED_BANDWIDTH) { + enum ieee80211_sta_rx_bandwidth bw; + + bw = ieee80211_chan_width_to_rx_bw(link_conf->chanreq.oper.width); + + iwl_mld_omi_ap_changed_bw(mld, link_conf, bw); + } + + if (changes & BSS_CHANGED_BANDWIDTH) + iwl_mld_emlsr_check_equal_bw(mld, vif, link_conf); +} + +static int iwl_mld_update_mu_groups(struct iwl_mld *mld, + struct ieee80211_bss_conf *link_conf) +{ + struct iwl_mu_group_mgmt_cmd cmd = {}; + + BUILD_BUG_ON(sizeof(cmd.membership_status) != + sizeof(link_conf->mu_group.membership)); + BUILD_BUG_ON(sizeof(cmd.user_position) != + sizeof(link_conf->mu_group.position)); + + memcpy(cmd.membership_status, link_conf->mu_group.membership, + WLAN_MEMBERSHIP_LEN); + memcpy(cmd.user_position, link_conf->mu_group.position, + WLAN_USER_POSITION_LEN); + + return iwl_mld_send_cmd_pdu(mld, + WIDE_ID(DATA_PATH_GROUP, + UPDATE_MU_GROUPS_CMD), + &cmd); +} + +static void +iwl_mld_mac80211_link_info_changed(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link_conf, + u64 changes) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + + switch (vif->type) { + case NL80211_IFTYPE_STATION: + iwl_mld_mac80211_link_info_changed_sta(mld, vif, link_conf, + changes); + break; + case NL80211_IFTYPE_AP: + case NL80211_IFTYPE_ADHOC: + iwl_mld_link_info_changed_ap_ibss(mld, vif, link_conf, + changes); + break; + case NL80211_IFTYPE_MONITOR: + /* The firmware tracks this on its own in STATION mode, but + * obviously not in sniffer mode. + */ + if (changes & BSS_CHANGED_MU_GROUPS) + iwl_mld_update_mu_groups(mld, link_conf); + break; + default: + /* shouldn't happen */ + WARN_ON_ONCE(1); + } + + /* We now know our BSSID, we can configure the MAC context with + * eht_support if needed. + */ + if (changes & BSS_CHANGED_BSSID) + iwl_mld_mac_fw_action(mld, vif, FW_CTXT_ACTION_MODIFY); + + if (changes & BSS_CHANGED_TXPOWER) + iwl_mld_set_tx_power(mld, link_conf, link_conf->txpower); +} + +static +void iwl_mld_mac80211_vif_cfg_changed(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + u64 changes) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + if (vif->type != NL80211_IFTYPE_STATION) + return; + + if (changes & BSS_CHANGED_ASSOC) { + ret = iwl_mld_mac_fw_action(mld, vif, FW_CTXT_ACTION_MODIFY); + if (ret) + IWL_ERR(mld, "failed to update context\n"); + + if (vif->cfg.assoc) { + /* Clear statistics to get clean beacon counter, and + * ask for periodic statistics, as they are needed for + * link selection and RX OMI decisions. + */ + iwl_mld_clear_stats_in_fw(mld); + iwl_mld_request_periodic_fw_stats(mld, true); + + iwl_mld_set_vif_associated(mld, vif); + } else { + iwl_mld_request_periodic_fw_stats(mld, false); + } + } + + if (changes & BSS_CHANGED_PS) { + /* Send both device-level and MAC-level power commands since the + * firmware checks the POWER_TABLE_CMD's POWER_SAVE_EN bit to + * determine SMPS mode. + */ + iwl_mld_update_device_power(mld, false); + iwl_mld_update_mac_power(mld, vif, false); + } + + /* TODO: task=MLO BSS_CHANGED_MLD_VALID_LINKS/CHANGED_MLD_TTLM */ +} + +static int +iwl_mld_mac80211_hw_scan(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_scan_request *hw_req) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + + if (WARN_ON(!hw_req->req.n_channels || + hw_req->req.n_channels > + mld->fw->ucode_capa.n_scan_channels)) + return -EINVAL; + + return iwl_mld_regular_scan_start(mld, vif, &hw_req->req, &hw_req->ies); +} + +static void +iwl_mld_mac80211_cancel_hw_scan(struct ieee80211_hw *hw, + struct ieee80211_vif *vif) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + + /* Due to a race condition, it's possible that mac80211 asks + * us to stop a hw_scan when it's already stopped. This can + * happen, for instance, if we stopped the scan ourselves, + * called ieee80211_scan_completed() and the userspace called + * cancel scan before ieee80211_scan_work() could run. + * To handle that, simply return if the scan is not running. + */ + if (mld->scan.status & IWL_MLD_SCAN_REGULAR) + iwl_mld_scan_stop(mld, IWL_MLD_SCAN_REGULAR, true); +} + +static int +iwl_mld_mac80211_sched_scan_start(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct cfg80211_sched_scan_request *req, + struct ieee80211_scan_ies *ies) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + + return iwl_mld_sched_scan_start(mld, vif, req, ies, IWL_MLD_SCAN_SCHED); +} + +static int +iwl_mld_mac80211_sched_scan_stop(struct ieee80211_hw *hw, + struct ieee80211_vif *vif) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + + /* Due to a race condition, it's possible that mac80211 asks + * us to stop a sched_scan when it's already stopped. This + * can happen, for instance, if we stopped the scan ourselves, + * called ieee80211_sched_scan_stopped() and the userspace called + * stop sched scan before ieee80211_sched_scan_stopped_work() + * could run. To handle this, simply return if the scan is + * not running. + */ + if (!(mld->scan.status & IWL_MLD_SCAN_SCHED)) + return 0; + + return iwl_mld_scan_stop(mld, IWL_MLD_SCAN_SCHED, false); +} + +static void +iwl_mld_restart_complete_vif(void *data, u8 *mac, struct ieee80211_vif *vif) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct ieee80211_bss_conf *link_conf; + struct iwl_mld *mld = data; + int link_id; + + for_each_vif_active_link(vif, link_conf, link_id) { + enum ieee80211_sta_rx_bandwidth bw; + struct iwl_mld_link *mld_link; + + mld_link = wiphy_dereference(mld->wiphy, + mld_vif->link[link_id]); + + if (WARN_ON_ONCE(!mld_link)) + continue; + + bw = mld_link->rx_omi.bw_in_progress; + if (bw) + iwl_mld_change_link_omi_bw(mld, link_conf, bw); + } +} + +static void +iwl_mld_mac80211_reconfig_complete(struct ieee80211_hw *hw, + enum ieee80211_reconfig_type reconfig_type) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + + switch (reconfig_type) { + case IEEE80211_RECONFIG_TYPE_RESTART: + mld->fw_status.in_hw_restart = false; + iwl_mld_send_recovery_cmd(mld, ERROR_RECOVERY_END_OF_RECOVERY); + + ieee80211_iterate_interfaces(mld->hw, + IEEE80211_IFACE_ITER_NORMAL, + iwl_mld_restart_complete_vif, mld); + + iwl_trans_finish_sw_reset(mld->trans); + /* no need to lock, adding in parallel would schedule too */ + if (!list_empty(&mld->txqs_to_add)) + wiphy_work_queue(mld->wiphy, &mld->add_txqs_wk); + + IWL_INFO(mld, "restart completed\n"); + break; + case IEEE80211_RECONFIG_TYPE_SUSPEND: + break; + } +} + +static +void iwl_mld_mac80211_mgd_prepare_tx(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_prep_tx_info *info) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + u32 duration = IWL_MLD_SESSION_PROTECTION_ASSOC_TIME_MS; + + /* After a successful association the connection is etalibeshed + * and we can rely on the quota to send the disassociation frame. + */ + if (info->was_assoc) + return; + + if (info->duration > duration) + duration = info->duration; + + iwl_mld_schedule_session_protection(mld, vif, duration, + IWL_MLD_SESSION_PROTECTION_MIN_TIME_MS, + info->link_id); +} + +static +void iwl_mld_mac_mgd_complete_tx(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_prep_tx_info *info) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + + /* Successful authentication is the only case that requires to let + * the session protection go. We'll need it for the upcoming + * association. For all the other cases, we need to cancel the session + * protection. + * After successful association the connection is established and + * further mgd tx can rely on the quota. + */ + if (info->success && info->subtype == IEEE80211_STYPE_AUTH) + return; + + /* The firmware will be on medium after we configure the vif as + * associated. Removing the session protection allows the firmware + * to stop being on medium. In order to ensure the continuity of our + * presence on medium, we need first to configure the vif as associated + * and only then, remove the session protection. + * Currently, mac80211 calls vif_cfg_changed() first and then, + * drv_mgd_complete_tx(). Ensure that this assumption stays true by + * a warning. + */ + WARN_ON(info->success && + (info->subtype == IEEE80211_STYPE_ASSOC_REQ || + info->subtype == IEEE80211_STYPE_REASSOC_REQ) && + !vif->cfg.assoc); + + iwl_mld_cancel_session_protection(mld, vif, info->link_id); +} + +static int +iwl_mld_mac80211_conf_tx(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + unsigned int link_id, u16 ac, + const struct ieee80211_tx_queue_params *params) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mld_link *link; + + lockdep_assert_wiphy(mld->wiphy); + + link = iwl_mld_link_dereference_check(mld_vif, link_id); + if (!link) + return -EINVAL; + + link->queue_params[ac] = *params; + + /* No need to update right away, we'll get BSS_CHANGED_QOS + * The exception is P2P_DEVICE interface which needs immediate update. + */ + if (vif->type == NL80211_IFTYPE_P2P_DEVICE) + iwl_mld_change_link_in_fw(mld, &vif->bss_conf, + LINK_CONTEXT_MODIFY_QOS_PARAMS); + + return 0; +} + +static void iwl_mld_set_uapsd(struct iwl_mld *mld, struct ieee80211_vif *vif) +{ + vif->driver_flags &= ~IEEE80211_VIF_SUPPORTS_UAPSD; + + if (vif->type != NL80211_IFTYPE_STATION) + return; + + if (vif->p2p && + !(iwlwifi_mod_params.uapsd_disable & IWL_DISABLE_UAPSD_P2P_CLIENT)) + vif->driver_flags |= IEEE80211_VIF_SUPPORTS_UAPSD; + + if (!vif->p2p && + !(iwlwifi_mod_params.uapsd_disable & IWL_DISABLE_UAPSD_BSS)) + vif->driver_flags |= IEEE80211_VIF_SUPPORTS_UAPSD; +} + +int iwl_mld_tdls_sta_count(struct iwl_mld *mld) +{ + int count = 0; + + lockdep_assert_wiphy(mld->wiphy); + + for (int i = 0; i < mld->fw->ucode_capa.num_stations; i++) { + struct ieee80211_link_sta *link_sta; + + link_sta = wiphy_dereference(mld->wiphy, + mld->fw_id_to_link_sta[i]); + if (IS_ERR_OR_NULL(link_sta)) + continue; + + if (!link_sta->sta->tdls) + continue; + + count++; + } + + return count; +} + +static void iwl_mld_check_he_obss_narrow_bw_ru_iter(struct wiphy *wiphy, + struct cfg80211_bss *bss, + void *_data) +{ + bool *tolerated = _data; + const struct cfg80211_bss_ies *ies; + const struct element *elem; + + rcu_read_lock(); + ies = rcu_dereference(bss->ies); + elem = cfg80211_find_elem(WLAN_EID_EXT_CAPABILITY, ies->data, + ies->len); + + if (!elem || elem->datalen < 10 || + !(elem->data[10] & + WLAN_EXT_CAPA10_OBSS_NARROW_BW_RU_TOLERANCE_SUPPORT)) { + *tolerated = false; + } + rcu_read_unlock(); +} + +static void +iwl_mld_check_he_obss_narrow_bw_ru(struct iwl_mld *mld, + struct iwl_mld_link *mld_link, + struct ieee80211_bss_conf *link_conf) +{ + bool tolerated = true; + + if (WARN_ON_ONCE(!link_conf->chanreq.oper.chan)) + return; + + if (!(link_conf->chanreq.oper.chan->flags & IEEE80211_CHAN_RADAR)) { + mld_link->he_ru_2mhz_block = false; + return; + } + + cfg80211_bss_iter(mld->wiphy, &link_conf->chanreq.oper, + iwl_mld_check_he_obss_narrow_bw_ru_iter, &tolerated); + + /* If there is at least one AP on radar channel that cannot + * tolerate 26-tone RU UL OFDMA transmissions using HE TB PPDU. + */ + mld_link->he_ru_2mhz_block = !tolerated; +} + +static void iwl_mld_link_set_2mhz_block(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta) +{ + struct ieee80211_link_sta *link_sta; + unsigned int link_id; + + for_each_sta_active_link(vif, sta, link_sta, link_id) { + struct ieee80211_bss_conf *link_conf = + link_conf_dereference_protected(vif, link_id); + struct iwl_mld_link *mld_link = + iwl_mld_link_from_mac80211(link_conf); + + if (WARN_ON(!link_conf || !mld_link)) + continue; + + if (link_sta->he_cap.has_he) + iwl_mld_check_he_obss_narrow_bw_ru(mld, mld_link, + link_conf); + } +} + +static int iwl_mld_move_sta_state_up(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + enum ieee80211_sta_state old_state, + enum ieee80211_sta_state new_state) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + int tdls_count = 0; + int ret; + + if (old_state == IEEE80211_STA_NOTEXIST && + new_state == IEEE80211_STA_NONE) { + if (sta->tdls) { + if (vif->p2p || hweight8(mld->used_phy_ids) != 1) + return -EBUSY; + + tdls_count = iwl_mld_tdls_sta_count(mld); + if (tdls_count >= IWL_TDLS_STA_COUNT) + return -EBUSY; + } + + /* + * If this is the first STA (i.e. the AP) it won't do + * anything, otherwise must leave for any new STA on + * any other interface, or for TDLS, etc. + * Need to call this _before_ adding the STA so it can + * look up the one STA to use to ask mac80211 to leave + * OMI; in the unlikely event that adding the new STA + * then fails we'll just re-enter OMI later (via the + * statistics notification handling.) + */ + iwl_mld_leave_omi_bw_reduction(mld); + + ret = iwl_mld_add_sta(mld, sta, vif, STATION_TYPE_PEER); + if (ret) + return ret; + + /* just added first TDLS STA, so disable PM */ + if (sta->tdls && tdls_count == 0) + iwl_mld_update_mac_power(mld, vif, false); + + if (vif->type == NL80211_IFTYPE_STATION && !sta->tdls) + mld_vif->ap_sta = sta; + + /* Initialize TLC here already - this really tells + * the firmware only what the supported legacy rates are + * (may be) since it's initialized already from what the + * AP advertised in the beacon/probe response. This will + * allow the firmware to send auth/assoc frames with one + * of the supported rates already, rather than having to + * use a mandatory rate. + * If we're the AP, we'll just assume mandatory rates at + * this point, but we know nothing about the STA anyway. + */ + iwl_mld_config_tlc(mld, vif, sta); + + return ret; + } else if (old_state == IEEE80211_STA_NONE && + new_state == IEEE80211_STA_AUTH) { + iwl_mld_set_uapsd(mld, vif); + return 0; + } else if (old_state == IEEE80211_STA_AUTH && + new_state == IEEE80211_STA_ASSOC) { + ret = iwl_mld_update_all_link_stations(mld, sta); + + if (vif->type == NL80211_IFTYPE_STATION) + iwl_mld_link_set_2mhz_block(mld, vif, sta); + /* Now the link_sta's capabilities are set, update the FW */ + iwl_mld_config_tlc(mld, vif, sta); + + if (vif->type == NL80211_IFTYPE_AP) { + /* Update MAC_CFG_FILTER_ACCEPT_BEACON if at least + * one sta is associated + */ + if (++mld_vif->num_associated_stas == 1) + ret = iwl_mld_mac_fw_action(mld, vif, + FW_CTXT_ACTION_MODIFY); + } + + return ret; + } else if (old_state == IEEE80211_STA_ASSOC && + new_state == IEEE80211_STA_AUTHORIZED) { + ret = 0; + + if (!sta->tdls) { + mld_vif->authorized = true; + + /* Ensure any block due to a non-BSS link is synced */ + iwl_mld_emlsr_check_non_bss_block(mld, 0); + + /* Block EMLSR until a certain throughput it reached */ + if (!mld->fw_status.in_hw_restart && + IWL_MLD_ENTER_EMLSR_TPT_THRESH > 0) + iwl_mld_block_emlsr(mld_vif->mld, vif, + IWL_MLD_EMLSR_BLOCKED_TPT, + 0); + + /* Wait for the FW to send a recommendation */ + iwl_mld_block_emlsr(mld, vif, + IWL_MLD_EMLSR_BLOCKED_FW, 0); + + /* clear COEX_HIGH_PRIORITY_ENABLE */ + ret = iwl_mld_mac_fw_action(mld, vif, + FW_CTXT_ACTION_MODIFY); + if (ret) + return ret; + } + + /* MFP is set by default before the station is authorized. + * Clear it here in case it's not used. + */ + if (!sta->mfp) + ret = iwl_mld_update_all_link_stations(mld, sta); + + /* We can use wide bandwidth now, not only 20 MHz */ + iwl_mld_config_tlc(mld, vif, sta); + + return ret; + } else { + return -EINVAL; + } +} + +static int iwl_mld_move_sta_state_down(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + enum ieee80211_sta_state old_state, + enum ieee80211_sta_state new_state) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + + if (old_state == IEEE80211_STA_AUTHORIZED && + new_state == IEEE80211_STA_ASSOC) { + if (!sta->tdls) { + mld_vif->authorized = false; + + memset(&mld_vif->emlsr.zeroed_on_not_authorized, 0, + sizeof(mld_vif->emlsr.zeroed_on_not_authorized)); + + wiphy_delayed_work_cancel(mld->wiphy, + &mld_vif->emlsr.prevent_done_wk); + wiphy_delayed_work_cancel(mld->wiphy, + &mld_vif->emlsr.tmp_non_bss_done_wk); + wiphy_work_cancel(mld->wiphy, &mld_vif->emlsr.unblock_tpt_wk); + wiphy_delayed_work_cancel(mld->wiphy, + &mld_vif->emlsr.check_tpt_wk); + + iwl_mld_reset_cca_40mhz_workaround(mld, vif); + } + + /* once we move into assoc state, need to update the FW to + * stop using wide bandwidth + */ + iwl_mld_config_tlc(mld, vif, sta); + } else if (old_state == IEEE80211_STA_ASSOC && + new_state == IEEE80211_STA_AUTH) { + if (vif->type == NL80211_IFTYPE_AP && + !WARN_ON(!mld_vif->num_associated_stas)) { + /* Update MAC_CFG_FILTER_ACCEPT_BEACON if the last sta + * is disassociating + */ + if (--mld_vif->num_associated_stas == 0) + iwl_mld_mac_fw_action(mld, vif, + FW_CTXT_ACTION_MODIFY); + } + } else if (old_state == IEEE80211_STA_AUTH && + new_state == IEEE80211_STA_NONE) { + /* nothing */ + } else if (old_state == IEEE80211_STA_NONE && + new_state == IEEE80211_STA_NOTEXIST) { + iwl_mld_remove_sta(mld, sta); + + if (sta->tdls && iwl_mld_tdls_sta_count(mld) == 0) { + /* just removed last TDLS STA, so enable PM */ + iwl_mld_update_mac_power(mld, vif, false); + } + } else { + return -EINVAL; + } + return 0; +} + +static int iwl_mld_mac80211_sta_state(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + enum ieee80211_sta_state old_state, + enum ieee80211_sta_state new_state) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(sta); + + IWL_DEBUG_MAC80211(mld, "station %pM state change %d->%d\n", + sta->addr, old_state, new_state); + + mld_sta->sta_state = new_state; + + if (old_state < new_state) + return iwl_mld_move_sta_state_up(mld, vif, sta, old_state, + new_state); + else + return iwl_mld_move_sta_state_down(mld, vif, sta, old_state, + new_state); +} + +static void iwl_mld_mac80211_flush(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + u32 queues, bool drop) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + + /* Make sure we're done with the deferred traffic before flushing */ + iwl_mld_add_txq_list(mld); + + for (int i = 0; i < mld->fw->ucode_capa.num_stations; i++) { + struct ieee80211_link_sta *link_sta = + wiphy_dereference(mld->wiphy, + mld->fw_id_to_link_sta[i]); + + if (IS_ERR_OR_NULL(link_sta)) + continue; + + /* Check that the sta belongs to the given vif */ + if (vif && vif != iwl_mld_sta_from_mac80211(link_sta->sta)->vif) + continue; + + if (drop) + iwl_mld_flush_sta_txqs(mld, link_sta->sta); + else + iwl_mld_wait_sta_txqs_empty(mld, link_sta->sta); + } +} + +static void iwl_mld_mac80211_flush_sta(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + + iwl_mld_flush_sta_txqs(mld, sta); +} + +static int +iwl_mld_mac80211_ampdu_action(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_ampdu_params *params) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct ieee80211_sta *sta = params->sta; + enum ieee80211_ampdu_mlme_action action = params->action; + u16 tid = params->tid; + u16 ssn = params->ssn; + u16 buf_size = params->buf_size; + u16 timeout = params->timeout; + int ret; + + IWL_DEBUG_HT(mld, "A-MPDU action on addr %pM tid: %d action: %d\n", + sta->addr, tid, action); + + switch (action) { + case IEEE80211_AMPDU_RX_START: + ret = iwl_mld_ampdu_rx_start(mld, sta, tid, ssn, buf_size, + timeout); + break; + case IEEE80211_AMPDU_RX_STOP: + ret = iwl_mld_ampdu_rx_stop(mld, sta, tid); + break; + default: + /* The mac80211 TX_AMPDU_SETUP_IN_HW flag is set for all + * devices, since all support TX A-MPDU offload in hardware. + * Therefore, no TX action should be requested here. + */ + WARN_ON_ONCE(1); + return -EINVAL; + } + + return ret; +} + +static bool iwl_mld_can_hw_csum(struct sk_buff *skb) +{ + u8 protocol = ip_hdr(skb)->protocol; + + return protocol == IPPROTO_TCP || protocol == IPPROTO_UDP; +} + +static bool iwl_mld_mac80211_can_aggregate(struct ieee80211_hw *hw, + struct sk_buff *head, + struct sk_buff *skb) +{ + if (!IS_ENABLED(CONFIG_INET)) + return false; + + /* For now don't aggregate IPv6 in AMSDU */ + if (skb->protocol != htons(ETH_P_IP)) + return false; + + /* Allow aggregation only if both frames have the same HW csum offload + * capability, ensuring consistent HW or SW csum handling in A-MSDU. + */ + return iwl_mld_can_hw_csum(skb) == iwl_mld_can_hw_csum(head); +} + +static void iwl_mld_mac80211_sync_rx_queues(struct ieee80211_hw *hw) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + + iwl_mld_sync_rx_queues(mld, IWL_MLD_RXQ_EMPTY, NULL, 0); +} + +static void iwl_mld_sta_rc_update(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_link_sta *link_sta, + u32 changed) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + + if (changed & (IEEE80211_RC_BW_CHANGED | + IEEE80211_RC_SUPP_RATES_CHANGED | + IEEE80211_RC_NSS_CHANGED)) { + struct ieee80211_bss_conf *link = + link_conf_dereference_check(vif, link_sta->link_id); + + if (WARN_ON(!link)) + return; + + iwl_mld_config_tlc_link(mld, vif, link, link_sta); + } +} + +static void iwl_mld_set_wakeup(struct ieee80211_hw *hw, bool enabled) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + + device_set_wakeup_enable(mld->trans->dev, enabled); +} + +/* Returns 0 on success. 1 if failed to suspend with wowlan: + * If the circumstances didn't satisfy the conditions for suspension + * with wowlan, mac80211 would use the no_wowlan flow. + * If an error had occurred we update the trans status and state here + * and the result will be stopping the FW. + */ +static int +iwl_mld_suspend(struct ieee80211_hw *hw, struct cfg80211_wowlan *wowlan) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + int ret; + + iwl_fw_runtime_suspend(&mld->fwrt); + + ret = iwl_mld_wowlan_suspend(mld, wowlan); + if (ret) { + if (ret < 0) { + mld->trans->state = IWL_TRANS_NO_FW; + set_bit(STATUS_FW_ERROR, &mld->trans->status); + } + return 1; + } + + if (iwl_mld_no_wowlan_suspend(mld)) + return 1; + + return 0; +} + +static int iwl_mld_resume(struct ieee80211_hw *hw) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + int ret; + + ret = iwl_mld_wowlan_resume(mld); + if (ret) + return ret; + + iwl_fw_runtime_resume(&mld->fwrt); + + iwl_mld_low_latency_restart(mld); + + return 0; +} + +static int iwl_mld_alloc_ptk_pn(struct iwl_mld *mld, + struct iwl_mld_sta *mld_sta, + struct ieee80211_key_conf *key, + struct iwl_mld_ptk_pn **ptk_pn) +{ + u8 num_rx_queues = mld->trans->num_rx_queues; + int keyidx = key->keyidx; + struct ieee80211_key_seq seq; + + if (WARN_ON(keyidx >= ARRAY_SIZE(mld_sta->ptk_pn))) + return -EINVAL; + + WARN_ON(rcu_access_pointer(mld_sta->ptk_pn[keyidx])); + *ptk_pn = kzalloc(struct_size(*ptk_pn, q, num_rx_queues), + GFP_KERNEL); + if (!*ptk_pn) + return -ENOMEM; + + for (u8 tid = 0; tid < IWL_MAX_TID_COUNT; tid++) { + ieee80211_get_key_rx_seq(key, tid, &seq); + for (u8 q = 0; q < num_rx_queues; q++) + memcpy((*ptk_pn)->q[q].pn[tid], seq.ccmp.pn, + IEEE80211_CCMP_PN_LEN); + } + + rcu_assign_pointer(mld_sta->ptk_pn[keyidx], *ptk_pn); + + return 0; +} + +static int iwl_mld_set_key_add(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + struct ieee80211_key_conf *key) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mld_sta *mld_sta = + sta ? iwl_mld_sta_from_mac80211(sta) : NULL; + struct iwl_mld_ptk_pn *ptk_pn = NULL; + int keyidx = key->keyidx; + int ret; + + /* Will be set to 0 if added successfully */ + key->hw_key_idx = STA_KEY_IDX_INVALID; + + switch (key->cipher) { + case WLAN_CIPHER_SUITE_WEP40: + case WLAN_CIPHER_SUITE_WEP104: + IWL_DEBUG_MAC80211(mld, "Use SW encryption for WEP\n"); + return -EOPNOTSUPP; + case WLAN_CIPHER_SUITE_TKIP: + if (vif->type == NL80211_IFTYPE_STATION) { + key->flags |= IEEE80211_KEY_FLAG_PUT_MIC_SPACE; + break; + } + IWL_DEBUG_MAC80211(mld, "Use SW encryption for TKIP\n"); + return -EOPNOTSUPP; + case WLAN_CIPHER_SUITE_CCMP: + case WLAN_CIPHER_SUITE_GCMP: + case WLAN_CIPHER_SUITE_GCMP_256: + case WLAN_CIPHER_SUITE_AES_CMAC: + case WLAN_CIPHER_SUITE_BIP_GMAC_128: + case WLAN_CIPHER_SUITE_BIP_GMAC_256: + break; + default: + return -EOPNOTSUPP; + } + + if (vif->type == NL80211_IFTYPE_STATION && + (keyidx == 6 || keyidx == 7)) + rcu_assign_pointer(mld_vif->bigtks[keyidx - 6], key); + + /* After exiting from RFKILL, hostapd configures GTK/ITGK before the + * AP is started, but those keys can't be sent to the FW before the + * MCAST/BCAST STAs are added to it (which happens upon AP start). + * Store it here to be sent later when the AP is started. + */ + if ((vif->type == NL80211_IFTYPE_ADHOC || + vif->type == NL80211_IFTYPE_AP) && !sta && + !mld_vif->ap_ibss_active) + return iwl_mld_store_ap_early_key(mld, key, mld_vif); + + if (!mld->fw_status.in_hw_restart && mld_sta && + key->flags & IEEE80211_KEY_FLAG_PAIRWISE && + (key->cipher == WLAN_CIPHER_SUITE_CCMP || + key->cipher == WLAN_CIPHER_SUITE_GCMP || + key->cipher == WLAN_CIPHER_SUITE_GCMP_256)) { + ret = iwl_mld_alloc_ptk_pn(mld, mld_sta, key, &ptk_pn); + if (ret) + return ret; + } + + IWL_DEBUG_MAC80211(mld, "set hwcrypto key (sta:%pM, id:%d)\n", + sta ? sta->addr : NULL, keyidx); + + ret = iwl_mld_add_key(mld, vif, sta, key); + if (ret) { + IWL_WARN(mld, "set key failed (%d)\n", ret); + if (ptk_pn) { + RCU_INIT_POINTER(mld_sta->ptk_pn[keyidx], NULL); + kfree(ptk_pn); + } + + return -EOPNOTSUPP; + } + + return 0; +} + +static void iwl_mld_set_key_remove(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + struct ieee80211_key_conf *key) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mld_sta *mld_sta = + sta ? iwl_mld_sta_from_mac80211(sta) : NULL; + int keyidx = key->keyidx; + + if (vif->type == NL80211_IFTYPE_STATION && + (keyidx == 6 || keyidx == 7)) + RCU_INIT_POINTER(mld_vif->bigtks[keyidx - 6], NULL); + + if (mld_sta && key->flags & IEEE80211_KEY_FLAG_PAIRWISE && + (key->cipher == WLAN_CIPHER_SUITE_CCMP || + key->cipher == WLAN_CIPHER_SUITE_GCMP || + key->cipher == WLAN_CIPHER_SUITE_GCMP_256)) { + struct iwl_mld_ptk_pn *ptk_pn; + + if (WARN_ON(keyidx >= ARRAY_SIZE(mld_sta->ptk_pn))) + return; + + ptk_pn = wiphy_dereference(mld->wiphy, + mld_sta->ptk_pn[keyidx]); + RCU_INIT_POINTER(mld_sta->ptk_pn[keyidx], NULL); + if (!WARN_ON(!ptk_pn)) + kfree_rcu(ptk_pn, rcu_head); + } + + /* if this key was stored to be added later to the FW - free it here */ + if (!(key->flags & IEEE80211_KEY_FLAG_PAIRWISE)) + iwl_mld_free_ap_early_key(mld, key, mld_vif); + + /* We already removed it */ + if (key->hw_key_idx == STA_KEY_IDX_INVALID) + return; + + IWL_DEBUG_MAC80211(mld, "disable hwcrypto key\n"); + + iwl_mld_remove_key(mld, vif, sta, key); +} + +static int iwl_mld_mac80211_set_key(struct ieee80211_hw *hw, + enum set_key_cmd cmd, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + struct ieee80211_key_conf *key) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + int ret; + + switch (cmd) { + case SET_KEY: + ret = iwl_mld_set_key_add(mld, vif, sta, key); + if (ret) + ret = -EOPNOTSUPP; + break; + case DISABLE_KEY: + iwl_mld_set_key_remove(mld, vif, sta, key); + ret = 0; + break; + default: + ret = -EINVAL; + break; + } + + return ret; +} + +static int +iwl_mld_pre_channel_switch(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_channel_switch *chsw) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mld_link *mld_link = + iwl_mld_link_dereference_check(mld_vif, chsw->link_id); + u8 primary; + int selected; + + lockdep_assert_wiphy(mld->wiphy); + + if (WARN_ON(!mld_link)) + return -EINVAL; + + IWL_DEBUG_MAC80211(mld, "pre CSA to freq %d\n", + chsw->chandef.center_freq1); + + if (!iwl_mld_emlsr_active(vif)) + return 0; + + primary = iwl_mld_get_primary_link(vif); + + /* stay on the primary link unless it is undergoing a CSA with quiet */ + if (chsw->link_id == primary && chsw->block_tx) + selected = iwl_mld_get_other_link(vif, primary); + else + selected = primary; + + /* Remember to tell the firmware that this link can't tx + * Note that this logic seems to be unrelated to emlsr, but it + * really is needed only when emlsr is active. When we have a + * single link, the firmware will handle all this on its own. + * In multi-link scenarios, we can learn about the CSA from + * another link and this logic is too complex for the firmware + * to track. + * Since we want to de-activate the link that got a CSA with mode=1, + * we need to tell the firmware not to send any frame on that link + * as the firmware may not be aware that link is under a CSA + * with mode=1 (no Tx allowed). + */ + mld_link->silent_deactivation = chsw->block_tx; + iwl_mld_exit_emlsr(mld, vif, IWL_MLD_EMLSR_EXIT_CSA, selected); + + return 0; +} + +static void +iwl_mld_channel_switch(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_channel_switch *chsw) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + + /* By implementing this operation, we prevent mac80211 from + * starting its own channel switch timer, so that we can call + * ieee80211_chswitch_done() ourselves at the right time + * (Upon receiving the channel_switch_start notification from the fw) + */ + IWL_DEBUG_MAC80211(mld, + "dummy channel switch op\n"); +} + +static int +iwl_mld_post_channel_switch(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link_conf) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct iwl_mld_link *mld_link = iwl_mld_link_from_mac80211(link_conf); + + lockdep_assert_wiphy(mld->wiphy); + + WARN_ON(mld_link->silent_deactivation); + + return 0; +} + +static void +iwl_mld_abort_channel_switch(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link_conf) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct iwl_mld_link *mld_link = iwl_mld_link_from_mac80211(link_conf); + + IWL_DEBUG_MAC80211(mld, + "abort channel switch op\n"); + mld_link->silent_deactivation = false; +} + +static int +iwl_mld_switch_vif_chanctx_swap(struct ieee80211_hw *hw, + struct ieee80211_vif_chanctx_switch *vifs) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + int ret; + + iwl_mld_unassign_vif_chanctx(hw, vifs[0].vif, vifs[0].link_conf, + vifs[0].old_ctx); + iwl_mld_remove_chanctx(hw, vifs[0].old_ctx); + + ret = iwl_mld_add_chanctx(hw, vifs[0].new_ctx); + if (ret) { + IWL_ERR(mld, "failed to add new_ctx during channel switch\n"); + goto out_reassign; + } + + ret = iwl_mld_assign_vif_chanctx(hw, vifs[0].vif, vifs[0].link_conf, + vifs[0].new_ctx); + if (ret) { + IWL_ERR(mld, + "failed to assign new_ctx during channel switch\n"); + goto out_remove; + } + + return 0; + + out_remove: + iwl_mld_remove_chanctx(hw, vifs[0].new_ctx); + out_reassign: + if (iwl_mld_add_chanctx(hw, vifs[0].old_ctx)) { + IWL_ERR(mld, "failed to add old_ctx after failure\n"); + return ret; + } + + if (iwl_mld_assign_vif_chanctx(hw, vifs[0].vif, vifs[0].link_conf, + vifs[0].old_ctx)) + IWL_ERR(mld, "failed to reassign old_ctx after failure\n"); + + return ret; +} + +static int +iwl_mld_switch_vif_chanctx_reassign(struct ieee80211_hw *hw, + struct ieee80211_vif_chanctx_switch *vifs) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + int ret; + + iwl_mld_unassign_vif_chanctx(hw, vifs[0].vif, vifs[0].link_conf, + vifs[0].old_ctx); + ret = iwl_mld_assign_vif_chanctx(hw, vifs[0].vif, vifs[0].link_conf, + vifs[0].new_ctx); + if (ret) { + IWL_ERR(mld, + "failed to assign new_ctx during channel switch\n"); + goto out_reassign; + } + + return 0; + +out_reassign: + if (iwl_mld_assign_vif_chanctx(hw, vifs[0].vif, vifs[0].link_conf, + vifs[0].old_ctx)) + IWL_ERR(mld, "failed to reassign old_ctx after failure\n"); + + return ret; +} + +static int +iwl_mld_switch_vif_chanctx(struct ieee80211_hw *hw, + struct ieee80211_vif_chanctx_switch *vifs, + int n_vifs, + enum ieee80211_chanctx_switch_mode mode) +{ + int ret; + + /* we only support a single-vif right now */ + if (n_vifs > 1) + return -EOPNOTSUPP; + + switch (mode) { + case CHANCTX_SWMODE_SWAP_CONTEXTS: + ret = iwl_mld_switch_vif_chanctx_swap(hw, vifs); + break; + case CHANCTX_SWMODE_REASSIGN_VIF: + ret = iwl_mld_switch_vif_chanctx_reassign(hw, vifs); + break; + default: + ret = -EOPNOTSUPP; + break; + } + + return ret; +} + +static void iwl_mld_sta_pre_rcu_remove(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(sta); + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mld_link_sta *mld_link_sta; + u8 link_id; + + lockdep_assert_wiphy(mld->wiphy); + + /* This is called before mac80211 does RCU synchronisation, + * so here we already invalidate our internal RCU-protected + * station pointer. The rest of the code will thus no longer + * be able to find the station this way, and we don't rely + * on further RCU synchronisation after the sta_state() + * callback deleted the station. + */ + for_each_mld_link_sta(mld_sta, mld_link_sta, link_id) + RCU_INIT_POINTER(mld->fw_id_to_link_sta[mld_link_sta->fw_id], + NULL); + + if (sta == mld_vif->ap_sta) + mld_vif->ap_sta = NULL; +} + +static void +iwl_mld_mac80211_mgd_protect_tdls_discover(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + unsigned int link_id) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct ieee80211_bss_conf *link_conf; + u32 duration; + int ret; + + link_conf = wiphy_dereference(hw->wiphy, vif->link_conf[link_id]); + if (WARN_ON_ONCE(!link_conf)) + return; + + /* Protect the session to hear the TDLS setup response on the channel */ + + duration = 2 * link_conf->dtim_period * link_conf->beacon_int; + + ret = iwl_mld_start_session_protection(mld, vif, duration, duration, + link_id, HZ / 5); + if (ret) + IWL_ERR(mld, + "Failed to start session protection for TDLS: %d\n", + ret); +} + +static bool iwl_mld_can_activate_links(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + u16 desired_links) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + int n_links = hweight16(desired_links); + + /* Check if HW supports the wanted number of links */ + return n_links <= iwl_mld_max_active_links(mld, vif); +} + +static int +iwl_mld_change_vif_links(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + u16 old_links, u16 new_links, + struct ieee80211_bss_conf *old[IEEE80211_MLD_MAX_NUM_LINKS]) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct ieee80211_bss_conf *link_conf; + u16 removed = old_links & ~new_links; + u16 added = new_links & ~old_links; + int err; + + lockdep_assert_wiphy(mld->wiphy); + + /* + * No bits designate non-MLO mode. We can handle MLO exit/enter by + * simply mapping that to link ID zero internally. + * Note that mac80211 does such a non-MLO to MLO switch during restart + * if it was in MLO before. In that case, we do not have a link to + * remove. + */ + if (old_links == 0 && !mld->fw_status.in_hw_restart) + removed |= BIT(0); + + if (new_links == 0) + added |= BIT(0); + + for (int i = 0; i < IEEE80211_MLD_MAX_NUM_LINKS; i++) { + if (removed & BIT(i)) { + link_conf = old[i]; + + err = iwl_mld_remove_link(mld, link_conf); + if (err) + return err; + } + } + + for (int i = 0; i < IEEE80211_MLD_MAX_NUM_LINKS; i++) { + if (added & BIT(i)) { + link_conf = link_conf_dereference_protected(vif, i); + if (WARN_ON(!link_conf)) + return -EINVAL; + + err = iwl_mld_add_link(mld, link_conf); + if (err) + goto remove_added_links; + } + } + + /* + * Ensure we always have a valid primary_link. When using multiple + * links the proper value is set in assign_vif_chanctx. + */ + mld_vif->emlsr.primary = new_links ? __ffs(new_links) : 0; + + /* + * Special MLO restart case. We did not have a link when the interface + * was added, so do the power configuration now. + */ + if (old_links == 0 && mld->fw_status.in_hw_restart) + iwl_mld_update_mac_power(mld, vif, false); + + return 0; + +remove_added_links: + for (int i = 0; i < IEEE80211_MLD_MAX_NUM_LINKS; i++) { + if (!(added & BIT(i))) + continue; + + link_conf = link_conf_dereference_protected(vif, i); + if (!link_conf || !iwl_mld_link_from_mac80211(link_conf)) + continue; + + iwl_mld_remove_link(mld, link_conf); + } + + return err; +} + +static int iwl_mld_change_sta_links(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + u16 old_links, u16 new_links) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + + return iwl_mld_update_link_stas(mld, vif, sta, old_links, new_links); +} + +static int iwl_mld_mac80211_join_ibss(struct ieee80211_hw *hw, + struct ieee80211_vif *vif) +{ + return iwl_mld_start_ap_ibss(hw, vif, &vif->bss_conf); +} + +static void iwl_mld_mac80211_leave_ibss(struct ieee80211_hw *hw, + struct ieee80211_vif *vif) +{ + return iwl_mld_stop_ap_ibss(hw, vif, &vif->bss_conf); +} + +static int iwl_mld_mac80211_tx_last_beacon(struct ieee80211_hw *hw) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + + return mld->ibss_manager; +} + +#define IWL_MLD_EMLSR_BLOCKED_TMP_NON_BSS_TIMEOUT (5 * HZ) + +static void iwl_mld_vif_iter_emlsr_block_tmp_non_bss(void *_data, u8 *mac, + struct ieee80211_vif *vif) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + int ret; + + if (!iwl_mld_vif_has_emlsr_cap(vif)) + return; + + ret = iwl_mld_block_emlsr_sync(mld_vif->mld, vif, + IWL_MLD_EMLSR_BLOCKED_TMP_NON_BSS, + iwl_mld_get_primary_link(vif)); + if (ret) + return; + + wiphy_delayed_work_queue(mld_vif->mld->wiphy, + &mld_vif->emlsr.tmp_non_bss_done_wk, + IWL_MLD_EMLSR_BLOCKED_TMP_NON_BSS_TIMEOUT); +} + +static void iwl_mld_prep_add_interface(struct ieee80211_hw *hw, + enum nl80211_iftype type) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + + IWL_DEBUG_MAC80211(mld, "prep_add_interface: type=%u\n", type); + + if (!(type == NL80211_IFTYPE_AP || + type == NL80211_IFTYPE_P2P_GO || + type == NL80211_IFTYPE_P2P_CLIENT)) + return; + + ieee80211_iterate_active_interfaces_mtx(mld->hw, + IEEE80211_IFACE_ITER_NORMAL, + iwl_mld_vif_iter_emlsr_block_tmp_non_bss, + NULL); +} + +static int iwl_mld_set_hw_timestamp(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct cfg80211_set_hw_timestamp *hwts) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + u32 protocols = 0; + + /* HW timestamping is only supported for a specific station */ + if (!hwts->macaddr) + return -EOPNOTSUPP; + + if (hwts->enable) + protocols = + IWL_TIME_SYNC_PROTOCOL_TM | IWL_TIME_SYNC_PROTOCOL_FTM; + + return iwl_mld_time_sync_config(mld, hwts->macaddr, protocols); +} + +static int iwl_mld_start_pmsr(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct cfg80211_pmsr_request *request) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + + return iwl_mld_ftm_start(mld, vif, request); +} + +const struct ieee80211_ops iwl_mld_hw_ops = { + .tx = iwl_mld_mac80211_tx, + .start = iwl_mld_mac80211_start, + .stop = iwl_mld_mac80211_stop, + .config = iwl_mld_mac80211_config, + .add_interface = iwl_mld_mac80211_add_interface, + .remove_interface = iwl_mld_mac80211_remove_interface, + .conf_tx = iwl_mld_mac80211_conf_tx, + .prepare_multicast = iwl_mld_mac80211_prepare_multicast, + .configure_filter = iwl_mld_mac80211_configure_filter, + .reconfig_complete = iwl_mld_mac80211_reconfig_complete, + .wake_tx_queue = iwl_mld_mac80211_wake_tx_queue, + .add_chanctx = iwl_mld_add_chanctx, + .remove_chanctx = iwl_mld_remove_chanctx, + .change_chanctx = iwl_mld_change_chanctx, + .assign_vif_chanctx = iwl_mld_assign_vif_chanctx, + .unassign_vif_chanctx = iwl_mld_unassign_vif_chanctx, + .set_rts_threshold = iwl_mld_mac80211_set_rts_threshold, + .link_info_changed = iwl_mld_mac80211_link_info_changed, + .vif_cfg_changed = iwl_mld_mac80211_vif_cfg_changed, + .set_key = iwl_mld_mac80211_set_key, + .hw_scan = iwl_mld_mac80211_hw_scan, + .cancel_hw_scan = iwl_mld_mac80211_cancel_hw_scan, + .sched_scan_start = iwl_mld_mac80211_sched_scan_start, + .sched_scan_stop = iwl_mld_mac80211_sched_scan_stop, + .mgd_prepare_tx = iwl_mld_mac80211_mgd_prepare_tx, + .mgd_complete_tx = iwl_mld_mac_mgd_complete_tx, + .sta_state = iwl_mld_mac80211_sta_state, + .sta_statistics = iwl_mld_mac80211_sta_statistics, + .flush = iwl_mld_mac80211_flush, + .flush_sta = iwl_mld_mac80211_flush_sta, + .ampdu_action = iwl_mld_mac80211_ampdu_action, + .can_aggregate_in_amsdu = iwl_mld_mac80211_can_aggregate, + .sync_rx_queues = iwl_mld_mac80211_sync_rx_queues, + .link_sta_rc_update = iwl_mld_sta_rc_update, + .start_ap = iwl_mld_start_ap_ibss, + .stop_ap = iwl_mld_stop_ap_ibss, + .pre_channel_switch = iwl_mld_pre_channel_switch, + .channel_switch = iwl_mld_channel_switch, + .post_channel_switch = iwl_mld_post_channel_switch, + .abort_channel_switch = iwl_mld_abort_channel_switch, + .switch_vif_chanctx = iwl_mld_switch_vif_chanctx, + .sta_pre_rcu_remove = iwl_mld_sta_pre_rcu_remove, + .remain_on_channel = iwl_mld_start_roc, + .cancel_remain_on_channel = iwl_mld_cancel_roc, + .can_activate_links = iwl_mld_can_activate_links, + .change_vif_links = iwl_mld_change_vif_links, + .change_sta_links = iwl_mld_change_sta_links, +#ifdef CONFIG_PM_SLEEP + .suspend = iwl_mld_suspend, + .resume = iwl_mld_resume, + .set_wakeup = iwl_mld_set_wakeup, + .set_rekey_data = iwl_mld_set_rekey_data, +#if IS_ENABLED(CONFIG_IPV6) + .ipv6_addr_change = iwl_mld_ipv6_addr_change, +#endif /* IS_ENABLED(CONFIG_IPV6) */ +#endif /* CONFIG_PM_SLEEP */ +#ifdef CONFIG_IWLWIFI_DEBUGFS + .vif_add_debugfs = iwl_mld_add_vif_debugfs, + .link_add_debugfs = iwl_mld_add_link_debugfs, + .link_sta_add_debugfs = iwl_mld_add_link_sta_debugfs, +#endif + .mgd_protect_tdls_discover = iwl_mld_mac80211_mgd_protect_tdls_discover, + .join_ibss = iwl_mld_mac80211_join_ibss, + .leave_ibss = iwl_mld_mac80211_leave_ibss, + .tx_last_beacon = iwl_mld_mac80211_tx_last_beacon, + .prep_add_interface = iwl_mld_prep_add_interface, + .set_hw_timestamp = iwl_mld_set_hw_timestamp, + .start_pmsr = iwl_mld_start_pmsr, +}; diff --git a/drivers/net/wireless/intel/iwlwifi/mld/mac80211.h b/drivers/net/wireless/intel/iwlwifi/mld/mac80211.h new file mode 100644 index 0000000000000..aad04d7b2617d --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/mac80211.h @@ -0,0 +1,13 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024 Intel Corporation + */ +#ifndef __iwl_mld_mac80211_h__ +#define __iwl_mld_mac80211_h__ + +#include "mld.h" + +int iwl_mld_register_hw(struct iwl_mld *mld); +void iwl_mld_recalc_multicast_filter(struct iwl_mld *mld); + +#endif /* __iwl_mld_mac80211_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/mcc.c b/drivers/net/wireless/intel/iwlwifi/mld/mcc.c new file mode 100644 index 0000000000000..daca14e208bde --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/mcc.c @@ -0,0 +1,329 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024 Intel Corporation + */ + +#include +#include + +#include +#include + +#include "mld.h" +#include "hcmd.h" +#include "mcc.h" + +/* It is the caller's responsibility to free the pointer returned here */ +static struct iwl_mcc_update_resp_v8 * +iwl_mld_parse_mcc_update_resp_v8(const struct iwl_rx_packet *pkt) +{ + const struct iwl_mcc_update_resp_v8 *mcc_resp_v8 = (const void *)pkt->data; + int n_channels = __le32_to_cpu(mcc_resp_v8->n_channels); + struct iwl_mcc_update_resp_v8 *resp_cp; + int notif_len = struct_size(resp_cp, channels, n_channels); + + if (iwl_rx_packet_payload_len(pkt) != notif_len) + return ERR_PTR(-EINVAL); + + resp_cp = kmemdup(mcc_resp_v8, notif_len, GFP_KERNEL); + if (!resp_cp) + return ERR_PTR(-ENOMEM); + + return resp_cp; +} + +/* It is the caller's responsibility to free the pointer returned here */ +static struct iwl_mcc_update_resp_v8 * +iwl_mld_parse_mcc_update_resp_v5_v6(const struct iwl_rx_packet *pkt) +{ + const struct iwl_mcc_update_resp_v4 *mcc_resp_v4 = (const void *)pkt->data; + struct iwl_mcc_update_resp_v8 *resp_cp; + int n_channels = __le32_to_cpu(mcc_resp_v4->n_channels); + int resp_len; + + if (iwl_rx_packet_payload_len(pkt) != + struct_size(mcc_resp_v4, channels, n_channels)) + return ERR_PTR(-EINVAL); + + resp_len = struct_size(resp_cp, channels, n_channels); + resp_cp = kzalloc(resp_len, GFP_KERNEL); + if (!resp_cp) + return ERR_PTR(-ENOMEM); + + resp_cp->status = mcc_resp_v4->status; + resp_cp->mcc = mcc_resp_v4->mcc; + resp_cp->cap = cpu_to_le32(le16_to_cpu(mcc_resp_v4->cap)); + resp_cp->source_id = mcc_resp_v4->source_id; + resp_cp->geo_info = mcc_resp_v4->geo_info; + resp_cp->n_channels = mcc_resp_v4->n_channels; + memcpy(resp_cp->channels, mcc_resp_v4->channels, + n_channels * sizeof(__le32)); + + return resp_cp; +} + +/* It is the caller's responsibility to free the pointer returned here */ +static struct iwl_mcc_update_resp_v8 * +iwl_mld_update_mcc(struct iwl_mld *mld, const char *alpha2, + enum iwl_mcc_source src_id) +{ + int resp_ver = iwl_fw_lookup_notif_ver(mld->fw, LONG_GROUP, + MCC_UPDATE_CMD, 0); + struct iwl_mcc_update_cmd mcc_update_cmd = { + .mcc = cpu_to_le16(alpha2[0] << 8 | alpha2[1]), + .source_id = (u8)src_id, + }; + struct iwl_mcc_update_resp_v8 *resp_cp; + struct iwl_rx_packet *pkt; + struct iwl_host_cmd cmd = { + .id = MCC_UPDATE_CMD, + .flags = CMD_WANT_SKB, + .data = { &mcc_update_cmd }, + .len[0] = sizeof(mcc_update_cmd), + }; + int ret; + u16 mcc; + + IWL_DEBUG_LAR(mld, "send MCC update to FW with '%c%c' src = %d\n", + alpha2[0], alpha2[1], src_id); + + ret = iwl_mld_send_cmd(mld, &cmd); + if (ret) + return ERR_PTR(ret); + + pkt = cmd.resp_pkt; + + /* For Wifi-7 radios, we get version 8 + * For Wifi-6E radios, we get version 6 + * For Wifi-6 radios, we get version 5, but 5, 6, and 4 are compatible. + */ + switch (resp_ver) { + case 5: + case 6: + resp_cp = iwl_mld_parse_mcc_update_resp_v5_v6(pkt); + break; + case 8: + resp_cp = iwl_mld_parse_mcc_update_resp_v8(pkt); + break; + default: + IWL_FW_CHECK_FAILED(mld, "Unknown MCC_UPDATE_CMD version %d\n", resp_ver); + resp_cp = ERR_PTR(-EINVAL); + } + + if (IS_ERR(resp_cp)) + goto exit; + + mcc = le16_to_cpu(resp_cp->mcc); + + IWL_FW_CHECK(mld, !mcc, "mcc can't be 0: %d\n", mcc); + + IWL_DEBUG_LAR(mld, + "MCC response status: 0x%x. new MCC: 0x%x ('%c%c')\n", + le32_to_cpu(resp_cp->status), mcc, mcc >> 8, mcc & 0xff); + +exit: + iwl_free_resp(&cmd); + return resp_cp; +} + +/* It is the caller's responsibility to free the pointer returned here */ +struct ieee80211_regdomain * +iwl_mld_get_regdomain(struct iwl_mld *mld, + const char *alpha2, + enum iwl_mcc_source src_id, + bool *changed) +{ + struct ieee80211_regdomain *regd = NULL; + struct iwl_mcc_update_resp_v8 *resp; + u8 resp_ver = iwl_fw_lookup_notif_ver(mld->fw, IWL_ALWAYS_LONG_GROUP, + MCC_UPDATE_CMD, 0); + + IWL_DEBUG_LAR(mld, "Getting regdomain data for %s from FW\n", alpha2); + + lockdep_assert_wiphy(mld->wiphy); + + resp = iwl_mld_update_mcc(mld, alpha2, src_id); + if (IS_ERR(resp)) { + IWL_DEBUG_LAR(mld, "Could not get update from FW %ld\n", + PTR_ERR(resp)); + resp = NULL; + goto out; + } + + if (changed) { + u32 status = le32_to_cpu(resp->status); + + *changed = (status == MCC_RESP_NEW_CHAN_PROFILE || + status == MCC_RESP_ILLEGAL); + } + IWL_DEBUG_LAR(mld, "MCC update response version: %d\n", resp_ver); + + regd = iwl_parse_nvm_mcc_info(mld->trans->dev, mld->cfg, + __le32_to_cpu(resp->n_channels), + resp->channels, + __le16_to_cpu(resp->mcc), + __le16_to_cpu(resp->geo_info), + le32_to_cpu(resp->cap), resp_ver); + + if (IS_ERR(regd)) { + IWL_DEBUG_LAR(mld, "Could not get parse update from FW %ld\n", + PTR_ERR(regd)); + goto out; + } + + IWL_DEBUG_LAR(mld, "setting alpha2 from FW to %s (0x%x, 0x%x) src=%d\n", + regd->alpha2, regd->alpha2[0], + regd->alpha2[1], resp->source_id); + + mld->mcc_src = resp->source_id; + + if (!iwl_puncturing_is_allowed_in_bios(mld->bios_enable_puncturing, + le16_to_cpu(resp->mcc))) + ieee80211_hw_set(mld->hw, DISALLOW_PUNCTURING); + else + __clear_bit(IEEE80211_HW_DISALLOW_PUNCTURING, mld->hw->flags); + +out: + kfree(resp); + return regd; +} + +/* It is the caller's responsibility to free the pointer returned here */ +static struct ieee80211_regdomain * +iwl_mld_get_current_regdomain(struct iwl_mld *mld, + bool *changed) +{ + return iwl_mld_get_regdomain(mld, "ZZ", + MCC_SOURCE_GET_CURRENT, changed); +} + +void iwl_mld_update_changed_regdomain(struct iwl_mld *mld) +{ + struct ieee80211_regdomain *regd; + bool changed; + + regd = iwl_mld_get_current_regdomain(mld, &changed); + + if (IS_ERR_OR_NULL(regd)) + return; + + if (changed) + regulatory_set_wiphy_regd(mld->wiphy, regd); + kfree(regd); +} + +static int iwl_mld_apply_last_mcc(struct iwl_mld *mld, + const char *alpha2) +{ + struct ieee80211_regdomain *regd; + u32 used_src; + bool changed; + int ret; + + /* save the last source in case we overwrite it below */ + used_src = mld->mcc_src; + + /* Notify the firmware we support wifi location updates */ + regd = iwl_mld_get_current_regdomain(mld, NULL); + if (!IS_ERR_OR_NULL(regd)) + kfree(regd); + + /* Now set our last stored MCC and source */ + regd = iwl_mld_get_regdomain(mld, alpha2, used_src, + &changed); + if (IS_ERR_OR_NULL(regd)) + return -EIO; + + /* update cfg80211 if the regdomain was changed */ + if (changed) + ret = regulatory_set_wiphy_regd_sync(mld->wiphy, regd); + else + ret = 0; + + kfree(regd); + return ret; +} + +int iwl_mld_init_mcc(struct iwl_mld *mld) +{ + const struct ieee80211_regdomain *r; + struct ieee80211_regdomain *regd; + char mcc[3]; + int retval; + + /* try to replay the last set MCC to FW */ + r = wiphy_dereference(mld->wiphy, mld->wiphy->regd); + + if (r) + return iwl_mld_apply_last_mcc(mld, r->alpha2); + + regd = iwl_mld_get_current_regdomain(mld, NULL); + if (IS_ERR_OR_NULL(regd)) + return -EIO; + + if (!iwl_bios_get_mcc(&mld->fwrt, mcc)) { + kfree(regd); + regd = iwl_mld_get_regdomain(mld, mcc, MCC_SOURCE_BIOS, NULL); + if (IS_ERR_OR_NULL(regd)) + return -EIO; + } + + retval = regulatory_set_wiphy_regd_sync(mld->wiphy, regd); + + kfree(regd); + return retval; +} + +static void iwl_mld_find_assoc_vif_iterator(void *data, u8 *mac, + struct ieee80211_vif *vif) +{ + bool *assoc = data; + + if (vif->type == NL80211_IFTYPE_STATION && + vif->cfg.assoc) + *assoc = true; +} + +static bool iwl_mld_is_a_vif_assoc(struct iwl_mld *mld) +{ + bool assoc = false; + + ieee80211_iterate_active_interfaces_atomic(mld->hw, + IEEE80211_IFACE_ITER_NORMAL, + iwl_mld_find_assoc_vif_iterator, + &assoc); + return assoc; +} + +void iwl_mld_handle_update_mcc(struct iwl_mld *mld, struct iwl_rx_packet *pkt) +{ + struct iwl_mcc_chub_notif *notif = (void *)pkt->data; + enum iwl_mcc_source src; + char mcc[3]; + struct ieee80211_regdomain *regd; + bool changed; + + lockdep_assert_wiphy(mld->wiphy); + + if (iwl_mld_is_a_vif_assoc(mld) && + notif->source_id == MCC_SOURCE_WIFI) { + IWL_DEBUG_LAR(mld, "Ignore mcc update while associated\n"); + return; + } + + mcc[0] = le16_to_cpu(notif->mcc) >> 8; + mcc[1] = le16_to_cpu(notif->mcc) & 0xff; + mcc[2] = '\0'; + src = notif->source_id; + + IWL_DEBUG_LAR(mld, + "RX: received chub update mcc cmd (mcc '%s' src %d)\n", + mcc, src); + regd = iwl_mld_get_regdomain(mld, mcc, src, &changed); + if (IS_ERR_OR_NULL(regd)) + return; + + if (changed) + regulatory_set_wiphy_regd(mld->hw->wiphy, regd); + kfree(regd); +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/mcc.h b/drivers/net/wireless/intel/iwlwifi/mld/mcc.h new file mode 100644 index 0000000000000..2b31e5b5e2edf --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/mcc.h @@ -0,0 +1,17 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024 Intel Corporation + */ +#ifndef __iwl_mld_mcc_h__ +#define __iwl_mld_mcc_h__ + +int iwl_mld_init_mcc(struct iwl_mld *mld); +void iwl_mld_handle_update_mcc(struct iwl_mld *mld, struct iwl_rx_packet *pkt); +void iwl_mld_update_changed_regdomain(struct iwl_mld *mld); +struct ieee80211_regdomain * +iwl_mld_get_regdomain(struct iwl_mld *mld, + const char *alpha2, + enum iwl_mcc_source src_id, + bool *changed); + +#endif /* __iwl_mld_mcc_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/mld.c b/drivers/net/wireless/intel/iwlwifi/mld/mld.c new file mode 100644 index 0000000000000..adc3d23aa019f --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/mld.c @@ -0,0 +1,707 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024-2025 Intel Corporation + */ + +#include + +#include "fw/api/rx.h" +#include "fw/api/datapath.h" +#include "fw/api/commands.h" +#include "fw/api/offload.h" +#include "fw/api/coex.h" +#include "fw/dbg.h" +#include "fw/uefi.h" + +#include "mld.h" +#include "mlo.h" +#include "mac80211.h" +#include "led.h" +#include "scan.h" +#include "tx.h" +#include "sta.h" +#include "regulatory.h" +#include "thermal.h" +#include "low_latency.h" +#include "hcmd.h" +#include "fw/api/location.h" + +#define DRV_DESCRIPTION "Intel(R) MLD wireless driver for Linux" +MODULE_DESCRIPTION(DRV_DESCRIPTION); +MODULE_LICENSE("GPL"); +MODULE_IMPORT_NS("IWLWIFI"); + +static const struct iwl_op_mode_ops iwl_mld_ops; + +static int __init iwl_mld_init(void) +{ + int ret = iwl_opmode_register("iwlmld", &iwl_mld_ops); + + if (ret) + pr_err("Unable to register MLD op_mode: %d\n", ret); + + return ret; +} +module_init(iwl_mld_init); + +static void __exit iwl_mld_exit(void) +{ + iwl_opmode_deregister("iwlmld"); +} +module_exit(iwl_mld_exit); + +VISIBLE_IF_IWLWIFI_KUNIT +void iwl_construct_mld(struct iwl_mld *mld, struct iwl_trans *trans, + const struct iwl_cfg *cfg, const struct iwl_fw *fw, + struct ieee80211_hw *hw, struct dentry *dbgfs_dir) +{ + mld->dev = trans->dev; + mld->trans = trans; + mld->cfg = cfg; + mld->fw = fw; + mld->hw = hw; + mld->wiphy = hw->wiphy; + mld->debugfs_dir = dbgfs_dir; + + iwl_notification_wait_init(&mld->notif_wait); + + /* Setup async RX handling */ + spin_lock_init(&mld->async_handlers_lock); + INIT_LIST_HEAD(&mld->async_handlers_list); + wiphy_work_init(&mld->async_handlers_wk, + iwl_mld_async_handlers_wk); + + /* Dynamic Queue Allocation */ + spin_lock_init(&mld->add_txqs_lock); + INIT_LIST_HEAD(&mld->txqs_to_add); + wiphy_work_init(&mld->add_txqs_wk, iwl_mld_add_txqs_wk); + + /* Setup RX queues sync wait queue */ + init_waitqueue_head(&mld->rxq_sync.waitq); +} +EXPORT_SYMBOL_IF_IWLWIFI_KUNIT(iwl_construct_mld); + +static void __acquires(&mld->wiphy->mtx) +iwl_mld_fwrt_dump_start(void *ctx) +{ + struct iwl_mld *mld = ctx; + + wiphy_lock(mld->wiphy); +} + +static void __releases(&mld->wiphy->mtx) +iwl_mld_fwrt_dump_end(void *ctx) +{ + struct iwl_mld *mld = ctx; + + wiphy_unlock(mld->wiphy); +} + +static bool iwl_mld_d3_debug_enable(void *ctx) +{ + return IWL_MLD_D3_DEBUG; +} + +static int iwl_mld_fwrt_send_hcmd(void *ctx, struct iwl_host_cmd *host_cmd) +{ + struct iwl_mld *mld = (struct iwl_mld *)ctx; + int ret; + + wiphy_lock(mld->wiphy); + ret = iwl_mld_send_cmd(mld, host_cmd); + wiphy_unlock(mld->wiphy); + + return ret; +} + +static const struct iwl_fw_runtime_ops iwl_mld_fwrt_ops = { + .dump_start = iwl_mld_fwrt_dump_start, + .dump_end = iwl_mld_fwrt_dump_end, + .send_hcmd = iwl_mld_fwrt_send_hcmd, + .d3_debug_enable = iwl_mld_d3_debug_enable, +}; + +static void +iwl_mld_construct_fw_runtime(struct iwl_mld *mld, struct iwl_trans *trans, + const struct iwl_fw *fw, + struct dentry *debugfs_dir) +{ + iwl_fw_runtime_init(&mld->fwrt, trans, fw, &iwl_mld_fwrt_ops, mld, + NULL, NULL, debugfs_dir); + + iwl_fw_set_current_image(&mld->fwrt, IWL_UCODE_REGULAR); +} + +/* Please keep this array *SORTED* by hex value. + * Access is done through binary search + */ +static const struct iwl_hcmd_names iwl_mld_legacy_names[] = { + HCMD_NAME(UCODE_ALIVE_NTFY), + HCMD_NAME(INIT_COMPLETE_NOTIF), + HCMD_NAME(PHY_CONTEXT_CMD), + HCMD_NAME(SCAN_CFG_CMD), + HCMD_NAME(SCAN_REQ_UMAC), + HCMD_NAME(SCAN_ABORT_UMAC), + HCMD_NAME(SCAN_COMPLETE_UMAC), + HCMD_NAME(TX_CMD), + HCMD_NAME(TXPATH_FLUSH), + HCMD_NAME(LEDS_CMD), + HCMD_NAME(WNM_80211V_TIMING_MEASUREMENT_NOTIFICATION), + HCMD_NAME(WNM_80211V_TIMING_MEASUREMENT_CONFIRM_NOTIFICATION), + HCMD_NAME(SCAN_OFFLOAD_UPDATE_PROFILES_CMD), + HCMD_NAME(POWER_TABLE_CMD), + HCMD_NAME(PSM_UAPSD_AP_MISBEHAVING_NOTIFICATION), + HCMD_NAME(BEACON_NOTIFICATION), + HCMD_NAME(BEACON_TEMPLATE_CMD), + HCMD_NAME(TX_ANT_CONFIGURATION_CMD), + HCMD_NAME(REDUCE_TX_POWER_CMD), + HCMD_NAME(MISSED_BEACONS_NOTIFICATION), + HCMD_NAME(MAC_PM_POWER_TABLE), + HCMD_NAME(MFUART_LOAD_NOTIFICATION), + HCMD_NAME(RSS_CONFIG_CMD), + HCMD_NAME(SCAN_ITERATION_COMPLETE_UMAC), + HCMD_NAME(REPLY_RX_MPDU_CMD), + HCMD_NAME(BA_NOTIF), + HCMD_NAME(MCC_UPDATE_CMD), + HCMD_NAME(MCC_CHUB_UPDATE_CMD), + HCMD_NAME(MCAST_FILTER_CMD), + HCMD_NAME(REPLY_BEACON_FILTERING_CMD), + HCMD_NAME(PROT_OFFLOAD_CONFIG_CMD), + HCMD_NAME(MATCH_FOUND_NOTIFICATION), + HCMD_NAME(WOWLAN_PATTERNS), + HCMD_NAME(WOWLAN_CONFIGURATION), + HCMD_NAME(WOWLAN_TSC_RSC_PARAM), + HCMD_NAME(WOWLAN_KEK_KCK_MATERIAL), + HCMD_NAME(DEBUG_HOST_COMMAND), + HCMD_NAME(LDBG_CONFIG_CMD), +}; + +/* Please keep this array *SORTED* by hex value. + * Access is done through binary search + */ +static const struct iwl_hcmd_names iwl_mld_system_names[] = { + HCMD_NAME(SHARED_MEM_CFG_CMD), + HCMD_NAME(SOC_CONFIGURATION_CMD), + HCMD_NAME(INIT_EXTENDED_CFG_CMD), + HCMD_NAME(FW_ERROR_RECOVERY_CMD), + HCMD_NAME(RFI_GET_FREQ_TABLE_CMD), + HCMD_NAME(SYSTEM_STATISTICS_CMD), + HCMD_NAME(SYSTEM_STATISTICS_END_NOTIF), +}; + +/* Please keep this array *SORTED* by hex value. + * Access is done through binary search + */ +static const struct iwl_hcmd_names iwl_mld_reg_and_nvm_names[] = { + HCMD_NAME(LARI_CONFIG_CHANGE), + HCMD_NAME(NVM_GET_INFO), + HCMD_NAME(TAS_CONFIG), + HCMD_NAME(SAR_OFFSET_MAPPING_TABLE_CMD), + HCMD_NAME(MCC_ALLOWED_AP_TYPE_CMD), +}; + +/* Please keep this array *SORTED* by hex value. + * Access is done through binary search + */ +static const struct iwl_hcmd_names iwl_mld_debug_names[] = { + HCMD_NAME(HOST_EVENT_CFG), + HCMD_NAME(DBGC_SUSPEND_RESUME), +}; + +/* Please keep this array *SORTED* by hex value. + * Access is done through binary search + */ +static const struct iwl_hcmd_names iwl_mld_mac_conf_names[] = { + HCMD_NAME(LOW_LATENCY_CMD), + HCMD_NAME(SESSION_PROTECTION_CMD), + HCMD_NAME(MAC_CONFIG_CMD), + HCMD_NAME(LINK_CONFIG_CMD), + HCMD_NAME(STA_CONFIG_CMD), + HCMD_NAME(AUX_STA_CMD), + HCMD_NAME(STA_REMOVE_CMD), + HCMD_NAME(ROC_CMD), + HCMD_NAME(MISSED_BEACONS_NOTIF), + HCMD_NAME(EMLSR_TRANS_FAIL_NOTIF), + HCMD_NAME(ROC_NOTIF), + HCMD_NAME(CHANNEL_SWITCH_ERROR_NOTIF), + HCMD_NAME(SESSION_PROTECTION_NOTIF), + HCMD_NAME(PROBE_RESPONSE_DATA_NOTIF), + HCMD_NAME(CHANNEL_SWITCH_START_NOTIF), +}; + +/* Please keep this array *SORTED* by hex value. + * Access is done through binary search + */ +static const struct iwl_hcmd_names iwl_mld_data_path_names[] = { + HCMD_NAME(TRIGGER_RX_QUEUES_NOTIF_CMD), + HCMD_NAME(WNM_PLATFORM_PTM_REQUEST_CMD), + HCMD_NAME(WNM_80211V_TIMING_MEASUREMENT_CONFIG_CMD), + HCMD_NAME(RFH_QUEUE_CONFIG_CMD), + HCMD_NAME(TLC_MNG_CONFIG_CMD), + HCMD_NAME(RX_BAID_ALLOCATION_CONFIG_CMD), + HCMD_NAME(SCD_QUEUE_CONFIG_CMD), + HCMD_NAME(OMI_SEND_STATUS_NOTIF), + HCMD_NAME(ESR_MODE_NOTIF), + HCMD_NAME(MONITOR_NOTIF), + HCMD_NAME(TLC_MNG_UPDATE_NOTIF), + HCMD_NAME(MU_GROUP_MGMT_NOTIF), +}; + +/* Please keep this array *SORTED* by hex value. + * Access is done through binary search + */ +static const struct iwl_hcmd_names iwl_mld_location_names[] = { + HCMD_NAME(TOF_RANGE_REQ_CMD), + HCMD_NAME(TOF_RANGE_RESPONSE_NOTIF), +}; + +/* Please keep this array *SORTED* by hex value. + * Access is done through binary search + */ +static const struct iwl_hcmd_names iwl_mld_phy_names[] = { + HCMD_NAME(CMD_DTS_MEASUREMENT_TRIGGER_WIDE), + HCMD_NAME(CTDP_CONFIG_CMD), + HCMD_NAME(TEMP_REPORTING_THRESHOLDS_CMD), + HCMD_NAME(PER_CHAIN_LIMIT_OFFSET_CMD), + HCMD_NAME(CT_KILL_NOTIFICATION), + HCMD_NAME(DTS_MEASUREMENT_NOTIF_WIDE), +}; + +/* Please keep this array *SORTED* by hex value. + * Access is done through binary search + */ +static const struct iwl_hcmd_names iwl_mld_statistics_names[] = { + HCMD_NAME(STATISTICS_OPER_NOTIF), + HCMD_NAME(STATISTICS_OPER_PART1_NOTIF), +}; + +/* Please keep this array *SORTED* by hex value. + * Access is done through binary search + */ +static const struct iwl_hcmd_names iwl_mld_prot_offload_names[] = { + HCMD_NAME(STORED_BEACON_NTF), +}; + +/* Please keep this array *SORTED* by hex value. + * Access is done through binary search + */ +static const struct iwl_hcmd_names iwl_mld_coex_names[] = { + HCMD_NAME(PROFILE_NOTIF), +}; + +VISIBLE_IF_IWLWIFI_KUNIT +const struct iwl_hcmd_arr iwl_mld_groups[] = { + [LEGACY_GROUP] = HCMD_ARR(iwl_mld_legacy_names), + [LONG_GROUP] = HCMD_ARR(iwl_mld_legacy_names), + [SYSTEM_GROUP] = HCMD_ARR(iwl_mld_system_names), + [MAC_CONF_GROUP] = HCMD_ARR(iwl_mld_mac_conf_names), + [DATA_PATH_GROUP] = HCMD_ARR(iwl_mld_data_path_names), + [LOCATION_GROUP] = HCMD_ARR(iwl_mld_location_names), + [REGULATORY_AND_NVM_GROUP] = HCMD_ARR(iwl_mld_reg_and_nvm_names), + [DEBUG_GROUP] = HCMD_ARR(iwl_mld_debug_names), + [PHY_OPS_GROUP] = HCMD_ARR(iwl_mld_phy_names), + [STATISTICS_GROUP] = HCMD_ARR(iwl_mld_statistics_names), + [PROT_OFFLOAD_GROUP] = HCMD_ARR(iwl_mld_prot_offload_names), + [BT_COEX_GROUP] = HCMD_ARR(iwl_mld_coex_names), +}; +EXPORT_SYMBOL_IF_IWLWIFI_KUNIT(iwl_mld_groups); + +#if IS_ENABLED(CONFIG_IWLWIFI_KUNIT_TESTS) +const unsigned int global_iwl_mld_goups_size = ARRAY_SIZE(iwl_mld_groups); +EXPORT_SYMBOL_IF_IWLWIFI_KUNIT(global_iwl_mld_goups_size); +#endif + +static void +iwl_mld_configure_trans(struct iwl_op_mode *op_mode) +{ + const struct iwl_mld *mld = IWL_OP_MODE_GET_MLD(op_mode); + static const u8 no_reclaim_cmds[] = {TX_CMD}; + struct iwl_trans_config trans_cfg = { + .op_mode = op_mode, + /* Rx is not supported yet, but add it to avoid warnings */ + .rx_buf_size = iwl_amsdu_size_to_rxb_size(), + .command_groups = iwl_mld_groups, + .command_groups_size = ARRAY_SIZE(iwl_mld_groups), + .fw_reset_handshake = true, + .queue_alloc_cmd_ver = + iwl_fw_lookup_cmd_ver(mld->fw, + WIDE_ID(DATA_PATH_GROUP, + SCD_QUEUE_CONFIG_CMD), + 0), + .no_reclaim_cmds = no_reclaim_cmds, + .n_no_reclaim_cmds = ARRAY_SIZE(no_reclaim_cmds), + .cb_data_offs = offsetof(struct ieee80211_tx_info, + driver_data[2]), + }; + struct iwl_trans *trans = mld->trans; + + trans->rx_mpdu_cmd = REPLY_RX_MPDU_CMD; + trans->iml = mld->fw->iml; + trans->iml_len = mld->fw->iml_len; + trans->wide_cmd_header = true; + + iwl_trans_configure(trans, &trans_cfg); +} + +/* + ***************************************************** + * op mode ops functions + ***************************************************** + */ + +#define NUM_FW_LOAD_RETRIES 3 +static struct iwl_op_mode * +iwl_op_mode_mld_start(struct iwl_trans *trans, const struct iwl_cfg *cfg, + const struct iwl_fw *fw, struct dentry *dbgfs_dir) +{ + struct ieee80211_hw *hw; + struct iwl_op_mode *op_mode; + struct iwl_mld *mld; + u32 eckv_value; + int ret; + + /* Allocate and initialize a new hardware device */ + hw = ieee80211_alloc_hw(sizeof(struct iwl_op_mode) + + sizeof(struct iwl_mld), + &iwl_mld_hw_ops); + if (!hw) + return ERR_PTR(-ENOMEM); + + op_mode = hw->priv; + + op_mode->ops = &iwl_mld_ops; + + mld = IWL_OP_MODE_GET_MLD(op_mode); + + iwl_construct_mld(mld, trans, cfg, fw, hw, dbgfs_dir); + + iwl_mld_construct_fw_runtime(mld, trans, fw, dbgfs_dir); + + iwl_mld_get_bios_tables(mld); + iwl_uefi_get_sgom_table(trans, &mld->fwrt); + iwl_uefi_get_step_table(trans); + if (iwl_bios_get_eckv(&mld->fwrt, &eckv_value)) + IWL_DEBUG_RADIO(mld, "ECKV table doesn't exist in BIOS\n"); + else + trans->ext_32khz_clock_valid = !!eckv_value; + iwl_bios_setup_step(trans, &mld->fwrt); + mld->bios_enable_puncturing = iwl_uefi_get_puncturing(&mld->fwrt); + + /* Configure transport layer with the opmode specific params */ + iwl_mld_configure_trans(op_mode); + + /* Needed for sending commands */ + wiphy_lock(mld->wiphy); + + for (int i = 0; i < NUM_FW_LOAD_RETRIES; i++) { + ret = iwl_mld_load_fw(mld); + if (!ret) + break; + } + + if (ret) { + wiphy_unlock(mld->wiphy); + iwl_fw_flush_dumps(&mld->fwrt); + goto free_hw; + } + + iwl_mld_stop_fw(mld); + + wiphy_unlock(mld->wiphy); + + ret = iwl_mld_leds_init(mld); + if (ret) + goto free_nvm; + + ret = iwl_mld_alloc_scan_cmd(mld); + if (ret) + goto leds_exit; + + ret = iwl_mld_low_latency_init(mld); + if (ret) + goto free_scan_cmd; + + ret = iwl_mld_register_hw(mld); + if (ret) + goto low_latency_free; + + iwl_mld_toggle_tx_ant(mld, &mld->mgmt_tx_ant); + + iwl_mld_add_debugfs_files(mld, dbgfs_dir); + iwl_mld_thermal_initialize(mld); + + iwl_mld_ptp_init(mld); + + return op_mode; + +low_latency_free: + iwl_mld_low_latency_free(mld); +free_scan_cmd: + kfree(mld->scan.cmd); +leds_exit: + iwl_mld_leds_exit(mld); +free_nvm: + kfree(mld->nvm_data); +free_hw: + ieee80211_free_hw(mld->hw); + return ERR_PTR(ret); +} + +static void +iwl_op_mode_mld_stop(struct iwl_op_mode *op_mode) +{ + struct iwl_mld *mld = IWL_OP_MODE_GET_MLD(op_mode); + + iwl_mld_ptp_remove(mld); + iwl_mld_leds_exit(mld); + + wiphy_lock(mld->wiphy); + iwl_mld_thermal_exit(mld); + iwl_mld_low_latency_stop(mld); + iwl_mld_deinit_time_sync(mld); + wiphy_unlock(mld->wiphy); + + ieee80211_unregister_hw(mld->hw); + + iwl_fw_runtime_free(&mld->fwrt); + iwl_mld_low_latency_free(mld); + + iwl_trans_op_mode_leave(mld->trans); + + kfree(mld->nvm_data); + kfree(mld->scan.cmd); + kfree(mld->error_recovery_buf); + kfree(mld->mcast_filter_cmd); + + ieee80211_free_hw(mld->hw); +} + +static void iwl_mld_queue_state_change(struct iwl_op_mode *op_mode, + int hw_queue, bool queue_full) +{ + struct iwl_mld *mld = IWL_OP_MODE_GET_MLD(op_mode); + struct ieee80211_txq *txq; + struct iwl_mld_sta *mld_sta; + struct iwl_mld_txq *mld_txq; + + rcu_read_lock(); + + txq = rcu_dereference(mld->fw_id_to_txq[hw_queue]); + if (!txq) { + rcu_read_unlock(); + + if (queue_full) { + /* An internal queue is not expected to become full */ + IWL_WARN(mld, + "Internal hw_queue %d is full! stopping all queues\n", + hw_queue); + /* Stop all queues, as an internal queue is not + * mapped to a mac80211 one + */ + ieee80211_stop_queues(mld->hw); + } else { + ieee80211_wake_queues(mld->hw); + } + + return; + } + + mld_txq = iwl_mld_txq_from_mac80211(txq); + mld_sta = txq->sta ? iwl_mld_sta_from_mac80211(txq->sta) : NULL; + + mld_txq->status.stop_full = queue_full; + + if (!queue_full && mld_sta && + mld_sta->sta_state != IEEE80211_STA_NOTEXIST) { + local_bh_disable(); + iwl_mld_tx_from_txq(mld, txq); + local_bh_enable(); + } + + rcu_read_unlock(); +} + +static void +iwl_mld_queue_full(struct iwl_op_mode *op_mode, int hw_queue) +{ + iwl_mld_queue_state_change(op_mode, hw_queue, true); +} + +static void +iwl_mld_queue_not_full(struct iwl_op_mode *op_mode, int hw_queue) +{ + iwl_mld_queue_state_change(op_mode, hw_queue, false); +} + +static bool +iwl_mld_set_hw_rfkill_state(struct iwl_op_mode *op_mode, bool state) +{ + struct iwl_mld *mld = IWL_OP_MODE_GET_MLD(op_mode); + + iwl_mld_set_hwkill(mld, state); + + return false; +} + +static void +iwl_mld_free_skb(struct iwl_op_mode *op_mode, struct sk_buff *skb) +{ + struct iwl_mld *mld = IWL_OP_MODE_GET_MLD(op_mode); + struct ieee80211_tx_info *info = IEEE80211_SKB_CB(skb); + + iwl_trans_free_tx_cmd(mld->trans, info->driver_data[1]); + ieee80211_free_txskb(mld->hw, skb); +} + +static void iwl_mld_read_error_recovery_buffer(struct iwl_mld *mld) +{ + u32 src_size = mld->fw->ucode_capa.error_log_size; + u32 src_addr = mld->fw->ucode_capa.error_log_addr; + u8 *recovery_buf; + int ret; + + /* no recovery buffer size defined in a TLV */ + if (!src_size) + return; + + recovery_buf = kzalloc(src_size, GFP_ATOMIC); + if (!recovery_buf) + return; + + ret = iwl_trans_read_mem_bytes(mld->trans, src_addr, + recovery_buf, src_size); + if (ret) { + IWL_ERR(mld, "Failed to read error recovery buffer (%d)\n", + ret); + kfree(recovery_buf); + return; + } + + mld->error_recovery_buf = recovery_buf; +} + +static void iwl_mld_restart_nic(struct iwl_mld *mld) +{ + iwl_mld_read_error_recovery_buffer(mld); + + mld->fwrt.trans->dbg.restart_required = false; + + ieee80211_restart_hw(mld->hw); +} + +static void +iwl_mld_nic_error(struct iwl_op_mode *op_mode, + enum iwl_fw_error_type type) +{ + struct iwl_mld *mld = IWL_OP_MODE_GET_MLD(op_mode); + bool trans_dead = test_bit(STATUS_TRANS_DEAD, &mld->trans->status); + + if (type == IWL_ERR_TYPE_CMD_QUEUE_FULL) + IWL_ERR(mld, "Command queue full!\n"); + else if (!trans_dead && !mld->fw_status.do_not_dump_once) + iwl_fwrt_dump_error_logs(&mld->fwrt); + + mld->fw_status.do_not_dump_once = false; + + /* It is necessary to abort any os scan here because mac80211 requires + * having the scan cleared before restarting. + * We'll reset the scan_status to NONE in restart cleanup in + * the next drv_start() call from mac80211. If ieee80211_hw_restart + * isn't called scan status will stay busy. + */ + iwl_mld_report_scan_aborted(mld); + + /* + * This should be first thing before trying to collect any + * data to avoid endless loops if any HW error happens while + * collecting debug data. + * It might not actually be true that we'll restart, but the + * setting doesn't matter if we're going to be unbound either. + */ + if (type != IWL_ERR_TYPE_RESET_HS_TIMEOUT) + mld->fw_status.in_hw_restart = true; +} + +static void iwl_mld_dump_error(struct iwl_op_mode *op_mode, + struct iwl_fw_error_dump_mode *mode) +{ + struct iwl_mld *mld = IWL_OP_MODE_GET_MLD(op_mode); + + /* if we come in from opmode we have the mutex held */ + if (mode->context == IWL_ERR_CONTEXT_FROM_OPMODE) { + lockdep_assert_wiphy(mld->wiphy); + iwl_fw_error_collect(&mld->fwrt); + } else { + wiphy_lock(mld->wiphy); + if (mode->context != IWL_ERR_CONTEXT_ABORT) + iwl_fw_error_collect(&mld->fwrt); + wiphy_unlock(mld->wiphy); + } +} + +static bool iwl_mld_sw_reset(struct iwl_op_mode *op_mode, + enum iwl_fw_error_type type) +{ + struct iwl_mld *mld = IWL_OP_MODE_GET_MLD(op_mode); + + /* Do restart only in the following conditions are met: + * - we consider the FW as running + * - The trigger that brought us here is defined as one that requires + * a restart (in the debug TLVs) + */ + if (!mld->fw_status.running || !mld->fwrt.trans->dbg.restart_required) + return false; + + iwl_mld_restart_nic(mld); + return true; +} + +static void +iwl_mld_time_point(struct iwl_op_mode *op_mode, + enum iwl_fw_ini_time_point tp_id, + union iwl_dbg_tlv_tp_data *tp_data) +{ + struct iwl_mld *mld = IWL_OP_MODE_GET_MLD(op_mode); + + iwl_dbg_tlv_time_point(&mld->fwrt, tp_id, tp_data); +} + +#ifdef CONFIG_PM_SLEEP +static void iwl_mld_device_powered_off(struct iwl_op_mode *op_mode) +{ + struct iwl_mld *mld = IWL_OP_MODE_GET_MLD(op_mode); + + wiphy_lock(mld->wiphy); + mld->trans->system_pm_mode = IWL_PLAT_PM_MODE_DISABLED; + iwl_mld_stop_fw(mld); + mld->fw_status.in_d3 = false; + wiphy_unlock(mld->wiphy); +} +#else +static void iwl_mld_device_powered_off(struct iwl_op_mode *op_mode) +{} +#endif + +static const struct iwl_op_mode_ops iwl_mld_ops = { + .start = iwl_op_mode_mld_start, + .stop = iwl_op_mode_mld_stop, + .rx = iwl_mld_rx, + .rx_rss = iwl_mld_rx_rss, + .queue_full = iwl_mld_queue_full, + .queue_not_full = iwl_mld_queue_not_full, + .hw_rf_kill = iwl_mld_set_hw_rfkill_state, + .free_skb = iwl_mld_free_skb, + .nic_error = iwl_mld_nic_error, + .dump_error = iwl_mld_dump_error, + .sw_reset = iwl_mld_sw_reset, + .time_point = iwl_mld_time_point, + .device_powered_off = pm_sleep_ptr(iwl_mld_device_powered_off), +}; + +struct iwl_mld_mod_params iwlmld_mod_params = { + .power_scheme = IWL_POWER_SCHEME_BPS, +}; + +module_param_named(power_scheme, iwlmld_mod_params.power_scheme, int, 0444); +MODULE_PARM_DESC(power_scheme, + "power management scheme: 1-active, 2-balanced, default: 2"); diff --git a/drivers/net/wireless/intel/iwlwifi/mld/mld.h b/drivers/net/wireless/intel/iwlwifi/mld/mld.h new file mode 100644 index 0000000000000..6eda6081c8b41 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/mld.h @@ -0,0 +1,584 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024-2025 Intel Corporation + */ +#ifndef __iwl_mld_h__ +#define __iwl_mld_h__ + +#include +#include + +#include "iwl-trans.h" +#include "iwl-op-mode.h" +#include "fw/runtime.h" +#include "fw/notif-wait.h" +#include "fw/api/commands.h" +#include "fw/api/scan.h" +#include "fw/api/mac-cfg.h" +#include "fw/api/mac.h" +#include "fw/api/phy-ctxt.h" +#include "fw/api/datapath.h" +#include "fw/api/rx.h" +#include "fw/api/rs.h" +#include "fw/api/context.h" +#include "fw/api/coex.h" +#include "fw/api/location.h" + +#include "fw/dbg.h" + +#include "notif.h" +#include "scan.h" +#include "rx.h" +#include "thermal.h" +#include "low_latency.h" +#include "constants.h" +#include "ptp.h" +#include "time_sync.h" + +/** + * DOC: Introduction + * + * iwlmld is an operation mode (a.k.a. op_mode) for Intel wireless devices. + * It is used for devices that ship after 2024 which typically support + * the WiFi-7 features. MLD stands for multi-link device. Note that there are + * devices that do not support WiFi-7 or even WiFi 6E and yet use iwlmld, but + * the firmware APIs used in this driver are WiFi-7 compatible. + * + * In the architecture of iwlwifi, an op_mode is a layer that translates + * mac80211's APIs into commands for the firmware and, of course, notifications + * from the firmware to mac80211's APIs. An op_mode must implement the + * interface defined in iwl-op-mode.h to interact with the transport layer + * which allows to send and receive data to the device, start the hardware, + * etc... + */ + +/** + * DOC: Locking policy + * + * iwlmld has a very simple locking policy: it doesn't have any mutexes. It + * relies on cfg80211's wiphy->mtx and takes the lock when needed. All the + * control flows originating from mac80211 already acquired the lock, so that + * part is trivial, but also notifications that are received from the firmware + * and handled asynchronously are handled only after having taken the lock. + * This is described in notif.c. + * There are spin_locks needed to synchronize with the data path, around the + * allocation of the queues, for example. + */ + +/** + * DOC: Debugfs + * + * iwlmld adds its share of debugfs hooks and its handlers are synchronized + * with the wiphy_lock using wiphy_locked_debugfs. This avoids races against + * resources deletion while the debugfs hook is being used. + */ + +/** + * DOC: Main resources + * + * iwlmld is designed with the life cycle of the resource in mind. The + * resources are: + * + * - struct iwl_mld (matches mac80211's struct ieee80211_hw) + * + * - struct iwl_mld_vif (matches macu80211's struct ieee80211_vif) + * iwl_mld_vif contains an array of pointers to struct iwl_mld_link + * which describe the links for this vif. + * + * - struct iwl_mld_sta (matches mac80211's struct ieee80211_sta) + * iwl_mld_sta contains an array of points to struct iwl_mld_link_sta + * which describes the link stations for this station + * + * Each object has properties that can survive a firmware reset or not. + * Asynchronous firmware notifications can declare themselves as dependent on a + * certain instance of those resources and that means that the notifications + * will be cancelled once the instance is destroyed. + */ + +#define IWL_MLD_MAX_ADDRESSES 5 + +/** + * struct iwl_mld - MLD op mode + * + * @fw_id_to_bss_conf: maps a fw id of a link to the corresponding + * ieee80211_bss_conf. + * @fw_id_to_vif: maps a fw id of a MAC context to the corresponding + * ieee80211_vif. Mapping is valid only when the MAC exists in the fw. + * @fw_id_to_txq: maps a fw id of a txq to the corresponding + * ieee80211_txq. + * @used_phy_ids: a bitmap of the phy IDs used. If a bit is set, it means + * that the index of this bit is already used as a PHY id. + * @num_igtks: the number if iGTKs that were sent to the FW. + * @monitor: monitor related data + * @monitor.on: does a monitor vif exist (singleton hence bool) + * @monitor.ampdu_ref: the id of the A-MPDU for sniffer + * @monitor.ampdu_toggle: the state of the previous packet to track A-MPDU + * @monitor.cur_aid: current association id tracked by the sniffer + * @monitor.cur_bssid: current bssid tracked by the sniffer + * @monitor.p80: primary channel position relative to he whole bandwidth, in + * steps of 80 MHz + * @fw_id_to_link_sta: maps a fw id of a sta to the corresponding + * ieee80211_link_sta. This is not cleaned up on restart since we want to + * preserve the fw sta ids during a restart (for SN/PN restoring). + * FW ids of internal stations will be mapped to ERR_PTR, and will be + * re-allocated during a restart, so make sure to free it in restart + * cleanup using iwl_mld_free_internal_sta + * @netdetect: indicates the FW is in suspend mode with netdetect configured + * @p2p_device_vif: points to the p2p device vif if exists + * @bt_is_active: indicates that BT is active + * @dev: pointer to device struct. For printing purposes + * @trans: pointer to the transport layer + * @cfg: pointer to the device configuration + * @fw: a pointer to the fw object + * @hw: pointer to the hw object. + * @wiphy: a pointer to the wiphy struct, for easier access to it. + * @nvm_data: pointer to the nvm_data that includes all our capabilities + * @fwrt: fw runtime data + * @debugfs_dir: debugfs directory + * @notif_wait: notification wait related data. + * @async_handlers_list: a list of all async RX handlers. When a notifciation + * with an async handler is received, it is added to this list. + * When &async_handlers_wk runs - it runs these handlers one by one. + * @async_handlers_lock: a lock for &async_handlers_list. Sync + * &async_handlers_wk and RX notifcation path. + * @async_handlers_wk: A work to run all async RX handlers from + * &async_handlers_list. + * @ct_kill_exit_wk: worker to exit thermal kill + * @fw_status: bitmap of fw status bits + * @running: true if the firmware is running + * @do_not_dump_once: true if firmware dump must be prevented once + * @in_d3: indicates FW is in suspend mode and should be resumed + * @in_hw_restart: indicates that we are currently in restart flow. + * rather than restarted. Should be unset upon restart. + * @radio_kill: bitmap of radio kill status + * @radio_kill.hw: radio is killed by hw switch + * @radio_kill.ct: radio is killed because the device it too hot + * @addresses: device MAC addresses. + * @scan: instance of the scan object + * @wowlan: WoWLAN support data. + * @led: the led device + * @mcc_src: the source id of the MCC, comes from the firmware + * @bios_enable_puncturing: is puncturing enabled by bios + * @fw_id_to_ba: maps a fw (BA) id to a corresponding Block Ack session data. + * @num_rx_ba_sessions: tracks the number of active Rx Block Ack (BA) sessions. + * the driver ensures that new BA sessions are blocked once the maximum + * supported by the firmware is reached, preventing firmware asserts. + * @rxq_sync: manages RX queue sync state + * @txqs_to_add: a list of &ieee80211_txq's to allocate in &add_txqs_wk + * @add_txqs_wk: a worker to allocate txqs. + * @add_txqs_lock: to lock the &txqs_to_add list. + * @error_recovery_buf: pointer to the recovery buffer that will be read + * from firmware upon fw/hw error and sent back to the firmware in + * reconfig flow (after NIC reset). + * @mcast_filter_cmd: pointer to the multicast filter command. + * @mgmt_tx_ant: stores the last TX antenna index; used for setting + * TX rate_n_flags for non-STA mgmt frames (toggles on every TX failure). + * @low_latency: low-latency manager. + * @tzone: thermal zone device's data + * @cooling_dev: cooling device's related data + * @ibss_manager: in IBSS mode (only one vif can be active), indicates what + * firmware indicated about having transmitted the last beacon, i.e. + * being IBSS manager for that time and needing to respond to probe + * requests + * @ptp_data: data of the PTP clock + * @time_sync: time sync data. + * @ftm_initiator: FTM initiator data + */ +struct iwl_mld { + /* Add here fields that need clean up on restart */ + struct_group(zeroed_on_hw_restart, + struct ieee80211_bss_conf __rcu *fw_id_to_bss_conf[IWL_FW_MAX_LINK_ID + 1]; + struct ieee80211_vif __rcu *fw_id_to_vif[NUM_MAC_INDEX_DRIVER]; + struct ieee80211_txq __rcu *fw_id_to_txq[IWL_MAX_TVQM_QUEUES]; + u8 used_phy_ids: NUM_PHY_CTX; + u8 num_igtks; + struct { + bool on; + u32 ampdu_ref; + bool ampdu_toggle; + u8 p80; +#ifdef CONFIG_IWLWIFI_DEBUGFS + __le16 cur_aid; + u8 cur_bssid[ETH_ALEN]; +#endif + } monitor; +#ifdef CONFIG_PM_SLEEP + bool netdetect; +#endif /* CONFIG_PM_SLEEP */ + struct ieee80211_vif *p2p_device_vif; + bool bt_is_active; + ); + struct ieee80211_link_sta __rcu *fw_id_to_link_sta[IWL_STATION_COUNT_MAX]; + /* And here fields that survive a fw restart */ + struct device *dev; + struct iwl_trans *trans; + const struct iwl_cfg *cfg; + const struct iwl_fw *fw; + struct ieee80211_hw *hw; + struct wiphy *wiphy; + struct iwl_nvm_data *nvm_data; + struct iwl_fw_runtime fwrt; + struct dentry *debugfs_dir; + struct iwl_notif_wait_data notif_wait; + struct list_head async_handlers_list; + spinlock_t async_handlers_lock; + struct wiphy_work async_handlers_wk; + struct wiphy_delayed_work ct_kill_exit_wk; + + struct { + u32 running:1, + do_not_dump_once:1, +#ifdef CONFIG_PM_SLEEP + in_d3:1, +#endif + in_hw_restart:1; + + } fw_status; + + struct { + u32 hw:1, + ct:1; + } radio_kill; + + struct mac_address addresses[IWL_MLD_MAX_ADDRESSES]; + struct iwl_mld_scan scan; +#ifdef CONFIG_PM_SLEEP + struct wiphy_wowlan_support wowlan; +#endif /* CONFIG_PM_SLEEP */ +#ifdef CONFIG_IWLWIFI_LEDS + struct led_classdev led; +#endif + enum iwl_mcc_source mcc_src; + bool bios_enable_puncturing; + + struct iwl_mld_baid_data __rcu *fw_id_to_ba[IWL_MAX_BAID]; + u8 num_rx_ba_sessions; + + struct iwl_mld_rx_queues_sync rxq_sync; + + struct list_head txqs_to_add; + struct wiphy_work add_txqs_wk; + spinlock_t add_txqs_lock; + + u8 *error_recovery_buf; + struct iwl_mcast_filter_cmd *mcast_filter_cmd; + + u8 mgmt_tx_ant; + + struct iwl_mld_low_latency low_latency; + + bool ibss_manager; +#ifdef CONFIG_THERMAL + struct thermal_zone_device *tzone; + struct iwl_mld_cooling_device cooling_dev; +#endif + + struct ptp_data ptp_data; + + struct iwl_mld_time_sync_data __rcu *time_sync; + + struct { + struct cfg80211_pmsr_request *req; + struct wireless_dev *req_wdev; + int responses[IWL_TOF_MAX_APS]; + } ftm_initiator; +}; + +/* memset the part of the struct that requires cleanup on restart */ +#define CLEANUP_STRUCT(_ptr) \ + memset((void *)&(_ptr)->zeroed_on_hw_restart, 0, \ + sizeof((_ptr)->zeroed_on_hw_restart)) + +/* Cleanup function for struct iwl_mld_vif, will be called in restart */ +static inline void +iwl_cleanup_mld(struct iwl_mld *mld) +{ + CLEANUP_STRUCT(mld); + CLEANUP_STRUCT(&mld->scan); + + mld->fw_status.in_d3 = false; + + iwl_mld_low_latency_restart_cleanup(mld); + + /* Empty the list of async notification handlers so we won't process + * notifications from the dead fw after the reconfig flow. + */ + iwl_mld_purge_async_handlers_list(mld); +} + +enum iwl_power_scheme { + IWL_POWER_SCHEME_CAM = 1, + IWL_POWER_SCHEME_BPS, +}; + +/** + * struct iwl_mld_mod_params - module parameters for iwlmld + * @power_scheme: one of enum iwl_power_scheme + */ +struct iwl_mld_mod_params { + int power_scheme; +}; + +extern struct iwl_mld_mod_params iwlmld_mod_params; + +/* Extract MLD priv from op_mode */ +#define IWL_OP_MODE_GET_MLD(_iwl_op_mode) \ + ((struct iwl_mld *)(_iwl_op_mode)->op_mode_specific) + +#define IWL_MAC80211_GET_MLD(_hw) \ + IWL_OP_MODE_GET_MLD((struct iwl_op_mode *)((_hw)->priv)) + +#ifdef CONFIG_IWLWIFI_DEBUGFS +void +iwl_mld_add_debugfs_files(struct iwl_mld *mld, struct dentry *debugfs_dir); +#else +static inline void +iwl_mld_add_debugfs_files(struct iwl_mld *mld, struct dentry *debugfs_dir) +{} +#endif + +int iwl_mld_run_fw_init_sequence(struct iwl_mld *mld); +int iwl_mld_load_fw(struct iwl_mld *mld); +void iwl_mld_stop_fw(struct iwl_mld *mld); +int iwl_mld_start_fw(struct iwl_mld *mld); +void iwl_mld_send_recovery_cmd(struct iwl_mld *mld, u32 flags); + +static inline void iwl_mld_set_ctkill(struct iwl_mld *mld, bool state) +{ + mld->radio_kill.ct = state; + + wiphy_rfkill_set_hw_state(mld->wiphy, + mld->radio_kill.hw || mld->radio_kill.ct); +} + +static inline void iwl_mld_set_hwkill(struct iwl_mld *mld, bool state) +{ + mld->radio_kill.hw = state; + + wiphy_rfkill_set_hw_state(mld->wiphy, + mld->radio_kill.hw || mld->radio_kill.ct); +} + +static inline u8 iwl_mld_get_valid_tx_ant(const struct iwl_mld *mld) +{ + u8 tx_ant = mld->fw->valid_tx_ant; + + if (mld->nvm_data && mld->nvm_data->valid_tx_ant) + tx_ant &= mld->nvm_data->valid_tx_ant; + + return tx_ant; +} + +static inline u8 iwl_mld_get_valid_rx_ant(const struct iwl_mld *mld) +{ + u8 rx_ant = mld->fw->valid_rx_ant; + + if (mld->nvm_data && mld->nvm_data->valid_rx_ant) + rx_ant &= mld->nvm_data->valid_rx_ant; + + return rx_ant; +} + +static inline u8 iwl_mld_nl80211_band_to_fw(enum nl80211_band band) +{ + switch (band) { + case NL80211_BAND_2GHZ: + return PHY_BAND_24; + case NL80211_BAND_5GHZ: + return PHY_BAND_5; + case NL80211_BAND_6GHZ: + return PHY_BAND_6; + default: + WARN_ONCE(1, "Unsupported band (%u)\n", band); + return PHY_BAND_5; + } +} + +static inline u8 iwl_mld_phy_band_to_nl80211(u8 phy_band) +{ + switch (phy_band) { + case PHY_BAND_24: + return NL80211_BAND_2GHZ; + case PHY_BAND_5: + return NL80211_BAND_5GHZ; + case PHY_BAND_6: + return NL80211_BAND_6GHZ; + default: + WARN_ONCE(1, "Unsupported phy band (%u)\n", phy_band); + return NL80211_BAND_5GHZ; + } +} + +static inline int +iwl_mld_legacy_hw_idx_to_mac80211_idx(u32 rate_n_flags, + enum nl80211_band band) +{ + int format = rate_n_flags & RATE_MCS_MOD_TYPE_MSK; + int rate = rate_n_flags & RATE_LEGACY_RATE_MSK; + bool is_lb = band == NL80211_BAND_2GHZ; + + if (format == RATE_MCS_LEGACY_OFDM_MSK) + return is_lb ? rate + IWL_FIRST_OFDM_RATE : rate; + + /* CCK is not allowed in 5 GHz */ + return is_lb ? rate : -1; +} + +extern const struct ieee80211_ops iwl_mld_hw_ops; + +/** + * enum iwl_rx_handler_context: context for Rx handler + * @RX_HANDLER_SYNC: this means that it will be called in the Rx path + * which can't acquire the wiphy->mutex. + * @RX_HANDLER_ASYNC: If the handler needs to hold wiphy->mutex + * (and only in this case!), it should be set as ASYNC. In that case, + * it will be called from a worker with wiphy->mutex held. + */ +enum iwl_rx_handler_context { + RX_HANDLER_SYNC, + RX_HANDLER_ASYNC, +}; + +/** + * struct iwl_rx_handler: handler for FW notification + * @val_fn: input validation function. + * @sizes: an array that mapps a version to the expected size. + * @fn: the function is called when notification is handled + * @cmd_id: command id + * @n_sizes: number of elements in &sizes. + * @context: see &iwl_rx_handler_context + * @obj_type: the type of the object that this handler is related to. + * See &iwl_mld_object_type. Use IWL_MLD_OBJECT_TYPE_NONE if not related. + * @cancel: function to cancel the notification. valid only if obj_type is not + * IWL_MLD_OBJECT_TYPE_NONE. + */ +struct iwl_rx_handler { + union { + bool (*val_fn)(struct iwl_mld *mld, struct iwl_rx_packet *pkt); + const struct iwl_notif_struct_size *sizes; + }; + void (*fn)(struct iwl_mld *mld, struct iwl_rx_packet *pkt); + u16 cmd_id; + u8 n_sizes; + u8 context; + enum iwl_mld_object_type obj_type; + bool (*cancel)(struct iwl_mld *mld, struct iwl_rx_packet *pkt, + u32 obj_id); +}; + +/** + * struct iwl_notif_struct_size: map a notif ver to the expected size + * + * @size: the size to expect + * @ver: the version of the notification + */ +struct iwl_notif_struct_size { + u32 size:24, ver:8; +}; + +#if IS_ENABLED(CONFIG_IWLWIFI_KUNIT_TESTS) +extern const struct iwl_hcmd_arr iwl_mld_groups[]; +extern const unsigned int global_iwl_mld_goups_size; +extern const struct iwl_rx_handler iwl_mld_rx_handlers[]; +extern const unsigned int iwl_mld_rx_handlers_num; + +bool +iwl_mld_is_dup(struct iwl_mld *mld, struct ieee80211_sta *sta, + struct ieee80211_hdr *hdr, + const struct iwl_rx_mpdu_desc *mpdu_desc, + struct ieee80211_rx_status *rx_status, int queue); + +void iwl_construct_mld(struct iwl_mld *mld, struct iwl_trans *trans, + const struct iwl_cfg *cfg, const struct iwl_fw *fw, + struct ieee80211_hw *hw, struct dentry *dbgfs_dir); +#endif + +#define IWL_MLD_INVALID_FW_ID 0xff + +#define IWL_MLD_ALLOC_FN(_type, _mac80211_type) \ +static int \ +iwl_mld_allocate_##_type##_fw_id(struct iwl_mld *mld, \ + u8 *fw_id, \ + struct ieee80211_##_mac80211_type *mac80211_ptr) \ +{ \ + u8 rand = IWL_MLD_DIS_RANDOM_FW_ID ? 0 : get_random_u8(); \ + u8 arr_sz = ARRAY_SIZE(mld->fw_id_to_##_mac80211_type); \ + if (__builtin_types_compatible_p(typeof(*mac80211_ptr), \ + struct ieee80211_link_sta)) \ + arr_sz = mld->fw->ucode_capa.num_stations; \ + if (__builtin_types_compatible_p(typeof(*mac80211_ptr), \ + struct ieee80211_bss_conf)) \ + arr_sz = mld->fw->ucode_capa.num_links; \ + for (int i = 0; i < arr_sz; i++) { \ + u8 idx = (i + rand) % arr_sz; \ + if (rcu_access_pointer(mld->fw_id_to_##_mac80211_type[idx])) \ + continue; \ + IWL_DEBUG_INFO(mld, "Allocated at index %d / %d\n", idx, arr_sz); \ + *fw_id = idx; \ + rcu_assign_pointer(mld->fw_id_to_##_mac80211_type[idx], mac80211_ptr); \ + return 0; \ + } \ + return -ENOSPC; \ +} + +static inline struct ieee80211_bss_conf * +iwl_mld_fw_id_to_link_conf(struct iwl_mld *mld, u8 fw_link_id) +{ + if (IWL_FW_CHECK(mld, fw_link_id >= mld->fw->ucode_capa.num_links, + "Invalid fw_link_id: %d\n", fw_link_id)) + return NULL; + + return wiphy_dereference(mld->wiphy, + mld->fw_id_to_bss_conf[fw_link_id]); +} + +#define MSEC_TO_TU(_msec) ((_msec) * 1000 / 1024) + +void iwl_mld_add_vif_debugfs(struct ieee80211_hw *hw, + struct ieee80211_vif *vif); +void iwl_mld_add_link_debugfs(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link_conf, + struct dentry *dir); +void iwl_mld_add_link_sta_debugfs(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_link_sta *link_sta, + struct dentry *dir); + +/* Utilities */ + +static inline u8 iwl_mld_mac80211_ac_to_fw_tx_fifo(enum ieee80211_ac_numbers ac) +{ + static const u8 mac80211_ac_to_fw_tx_fifo[] = { + IWL_BZ_EDCA_TX_FIFO_VO, + IWL_BZ_EDCA_TX_FIFO_VI, + IWL_BZ_EDCA_TX_FIFO_BE, + IWL_BZ_EDCA_TX_FIFO_BK, + IWL_BZ_TRIG_TX_FIFO_VO, + IWL_BZ_TRIG_TX_FIFO_VI, + IWL_BZ_TRIG_TX_FIFO_BE, + IWL_BZ_TRIG_TX_FIFO_BK, + }; + return mac80211_ac_to_fw_tx_fifo[ac]; +} + +static inline u32 +iwl_mld_get_lmac_id(struct iwl_mld *mld, enum nl80211_band band) +{ + if (!fw_has_capa(&mld->fw->ucode_capa, + IWL_UCODE_TLV_CAPA_CDB_SUPPORT) || + band == NL80211_BAND_2GHZ) + return IWL_LMAC_24G_INDEX; + return IWL_LMAC_5G_INDEX; +} + +/* Check if we had an error, but reconfig flow didn't start yet */ +static inline bool iwl_mld_error_before_recovery(struct iwl_mld *mld) +{ + return mld->fw_status.in_hw_restart && + !iwl_trans_fw_running(mld->trans); +} + +int iwl_mld_tdls_sta_count(struct iwl_mld *mld); + +#endif /* __iwl_mld_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/mlo.c b/drivers/net/wireless/intel/iwlwifi/mld/mlo.c new file mode 100644 index 0000000000000..99c8501129b8f --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/mlo.c @@ -0,0 +1,994 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024-2025 Intel Corporation + */ +#include "mlo.h" + +/* Block reasons helper */ +#define HANDLE_EMLSR_BLOCKED_REASONS(HOW) \ + HOW(PREVENTION) \ + HOW(WOWLAN) \ + HOW(FW) \ + HOW(ROC) \ + HOW(NON_BSS) \ + HOW(TMP_NON_BSS) \ + HOW(TPT) + +static const char * +iwl_mld_get_emlsr_blocked_string(enum iwl_mld_emlsr_blocked blocked) +{ + /* Using switch without "default" will warn about missing entries */ + switch (blocked) { +#define REASON_CASE(x) case IWL_MLD_EMLSR_BLOCKED_##x: return #x; + HANDLE_EMLSR_BLOCKED_REASONS(REASON_CASE) +#undef REASON_CASE + } + + return "ERROR"; +} + +static void iwl_mld_print_emlsr_blocked(struct iwl_mld *mld, u32 mask) +{ +#define NAME_FMT(x) "%s" +#define NAME_PR(x) (mask & IWL_MLD_EMLSR_BLOCKED_##x) ? "[" #x "]" : "", + IWL_DEBUG_INFO(mld, + "EMLSR blocked = " HANDLE_EMLSR_BLOCKED_REASONS(NAME_FMT) + " (0x%x)\n", + HANDLE_EMLSR_BLOCKED_REASONS(NAME_PR) + mask); +#undef NAME_FMT +#undef NAME_PR +} + +/* Exit reasons helper */ +#define HANDLE_EMLSR_EXIT_REASONS(HOW) \ + HOW(BLOCK) \ + HOW(MISSED_BEACON) \ + HOW(FAIL_ENTRY) \ + HOW(CSA) \ + HOW(EQUAL_BAND) \ + HOW(BANDWIDTH) \ + HOW(LOW_RSSI) \ + HOW(LINK_USAGE) \ + HOW(BT_COEX) \ + HOW(CHAN_LOAD) \ + HOW(RFI) + +static const char * +iwl_mld_get_emlsr_exit_string(enum iwl_mld_emlsr_exit exit) +{ + /* Using switch without "default" will warn about missing entries */ + switch (exit) { +#define REASON_CASE(x) case IWL_MLD_EMLSR_EXIT_##x: return #x; + HANDLE_EMLSR_EXIT_REASONS(REASON_CASE) +#undef REASON_CASE + } + + return "ERROR"; +} + +static void iwl_mld_print_emlsr_exit(struct iwl_mld *mld, u32 mask) +{ +#define NAME_FMT(x) "%s" +#define NAME_PR(x) (mask & IWL_MLD_EMLSR_EXIT_##x) ? "[" #x "]" : "", + IWL_DEBUG_INFO(mld, + "EMLSR exit = " HANDLE_EMLSR_EXIT_REASONS(NAME_FMT) + " (0x%x)\n", + HANDLE_EMLSR_EXIT_REASONS(NAME_PR) + mask); +#undef NAME_FMT +#undef NAME_PR +} + +void iwl_mld_emlsr_prevent_done_wk(struct wiphy *wiphy, struct wiphy_work *wk) +{ + struct iwl_mld_vif *mld_vif = container_of(wk, struct iwl_mld_vif, + emlsr.prevent_done_wk.work); + struct ieee80211_vif *vif = + container_of((void *)mld_vif, struct ieee80211_vif, drv_priv); + + if (WARN_ON(!(mld_vif->emlsr.blocked_reasons & + IWL_MLD_EMLSR_BLOCKED_PREVENTION))) + return; + + iwl_mld_unblock_emlsr(mld_vif->mld, vif, + IWL_MLD_EMLSR_BLOCKED_PREVENTION); +} + +void iwl_mld_emlsr_tmp_non_bss_done_wk(struct wiphy *wiphy, + struct wiphy_work *wk) +{ + struct iwl_mld_vif *mld_vif = container_of(wk, struct iwl_mld_vif, + emlsr.prevent_done_wk.work); + struct ieee80211_vif *vif = + container_of((void *)mld_vif, struct ieee80211_vif, drv_priv); + + if (WARN_ON(!(mld_vif->emlsr.blocked_reasons & + IWL_MLD_EMLSR_BLOCKED_TMP_NON_BSS))) + return; + + iwl_mld_unblock_emlsr(mld_vif->mld, vif, + IWL_MLD_EMLSR_BLOCKED_TMP_NON_BSS); +} + +#define IWL_MLD_TRIGGER_LINK_SEL_TIME (HZ * IWL_MLD_TRIGGER_LINK_SEL_TIME_SEC) +#define IWL_MLD_SCAN_EXPIRE_TIME (HZ * IWL_MLD_SCAN_EXPIRE_TIME_SEC) + +/* Exit reasons that can cause longer EMLSR prevention */ +#define IWL_MLD_PREVENT_EMLSR_REASONS (IWL_MLD_EMLSR_EXIT_MISSED_BEACON | \ + IWL_MLD_EMLSR_EXIT_LINK_USAGE) +#define IWL_MLD_PREVENT_EMLSR_TIMEOUT (HZ * 400) + +#define IWL_MLD_EMLSR_PREVENT_SHORT (HZ * 300) +#define IWL_MLD_EMLSR_PREVENT_LONG (HZ * 600) + +static void iwl_mld_check_emlsr_prevention(struct iwl_mld *mld, + struct iwl_mld_vif *mld_vif, + enum iwl_mld_emlsr_exit reason) +{ + unsigned long delay; + + /* + * Reset the counter if more than 400 seconds have passed between one + * exit and the other, or if we exited due to a different reason. + * Will also reset the counter after the long prevention is done. + */ + if (time_after(jiffies, mld_vif->emlsr.last_exit_ts + + IWL_MLD_PREVENT_EMLSR_TIMEOUT) || + mld_vif->emlsr.last_exit_reason != reason) + mld_vif->emlsr.exit_repeat_count = 0; + + mld_vif->emlsr.last_exit_reason = reason; + mld_vif->emlsr.last_exit_ts = jiffies; + mld_vif->emlsr.exit_repeat_count++; + + /* + * Do not add a prevention when the reason was a block. For a block, + * EMLSR will be enabled again on unblock. + */ + if (reason == IWL_MLD_EMLSR_EXIT_BLOCK) + return; + + /* Set prevention for a minimum of 30 seconds */ + mld_vif->emlsr.blocked_reasons |= IWL_MLD_EMLSR_BLOCKED_PREVENTION; + delay = IWL_MLD_TRIGGER_LINK_SEL_TIME; + + /* Handle repeats for reasons that can cause long prevention */ + if (mld_vif->emlsr.exit_repeat_count > 1 && + reason & IWL_MLD_PREVENT_EMLSR_REASONS) { + if (mld_vif->emlsr.exit_repeat_count == 2) + delay = IWL_MLD_EMLSR_PREVENT_SHORT; + else + delay = IWL_MLD_EMLSR_PREVENT_LONG; + + /* + * The timeouts are chosen so that this will not happen, i.e. + * IWL_MLD_EMLSR_PREVENT_LONG > IWL_MLD_PREVENT_EMLSR_TIMEOUT + */ + WARN_ON(mld_vif->emlsr.exit_repeat_count > 3); + } + + IWL_DEBUG_INFO(mld, + "Preventing EMLSR for %ld seconds due to %u exits with the reason = %s (0x%x)\n", + delay / HZ, mld_vif->emlsr.exit_repeat_count, + iwl_mld_get_emlsr_exit_string(reason), reason); + + wiphy_delayed_work_queue(mld->wiphy, + &mld_vif->emlsr.prevent_done_wk, delay); +} + +static int _iwl_mld_exit_emlsr(struct iwl_mld *mld, struct ieee80211_vif *vif, + enum iwl_mld_emlsr_exit exit, u8 link_to_keep, + bool sync) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + u16 new_active_links; + int ret = 0; + + lockdep_assert_wiphy(mld->wiphy); + + /* On entry failure need to exit anyway, even if entered from debugfs */ + if (exit != IWL_MLD_EMLSR_EXIT_FAIL_ENTRY && !IWL_MLD_AUTO_EML_ENABLE) + return 0; + + /* Ignore exit request if EMLSR is not active */ + if (!iwl_mld_emlsr_active(vif)) + return 0; + + if (WARN_ON(!ieee80211_vif_is_mld(vif) || !mld_vif->authorized)) + return 0; + + if (WARN_ON(!(vif->active_links & BIT(link_to_keep)))) + link_to_keep = __ffs(vif->active_links); + + new_active_links = BIT(link_to_keep); + IWL_DEBUG_INFO(mld, + "Exiting EMLSR. reason = %s (0x%x). Current active links=0x%x, new active links = 0x%x\n", + iwl_mld_get_emlsr_exit_string(exit), exit, + vif->active_links, new_active_links); + + if (sync) + ret = ieee80211_set_active_links(vif, new_active_links); + else + ieee80211_set_active_links_async(vif, new_active_links); + + /* Update latest exit reason and check EMLSR prevention */ + iwl_mld_check_emlsr_prevention(mld, mld_vif, exit); + + return ret; +} + +void iwl_mld_exit_emlsr(struct iwl_mld *mld, struct ieee80211_vif *vif, + enum iwl_mld_emlsr_exit exit, u8 link_to_keep) +{ + _iwl_mld_exit_emlsr(mld, vif, exit, link_to_keep, false); +} + +static int _iwl_mld_emlsr_block(struct iwl_mld *mld, struct ieee80211_vif *vif, + enum iwl_mld_emlsr_blocked reason, + u8 link_to_keep, bool sync) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + + lockdep_assert_wiphy(mld->wiphy); + + if (!IWL_MLD_AUTO_EML_ENABLE || !iwl_mld_vif_has_emlsr_cap(vif)) + return 0; + + if (mld_vif->emlsr.blocked_reasons & reason) + return 0; + + mld_vif->emlsr.blocked_reasons |= reason; + + IWL_DEBUG_INFO(mld, + "Blocking EMLSR mode. reason = %s (0x%x)\n", + iwl_mld_get_emlsr_blocked_string(reason), reason); + iwl_mld_print_emlsr_blocked(mld, mld_vif->emlsr.blocked_reasons); + + if (reason == IWL_MLD_EMLSR_BLOCKED_TPT) + wiphy_delayed_work_cancel(mld_vif->mld->wiphy, + &mld_vif->emlsr.check_tpt_wk); + + return _iwl_mld_exit_emlsr(mld, vif, IWL_MLD_EMLSR_EXIT_BLOCK, + link_to_keep, sync); +} + +void iwl_mld_block_emlsr(struct iwl_mld *mld, struct ieee80211_vif *vif, + enum iwl_mld_emlsr_blocked reason, u8 link_to_keep) +{ + _iwl_mld_emlsr_block(mld, vif, reason, link_to_keep, false); +} + +int iwl_mld_block_emlsr_sync(struct iwl_mld *mld, struct ieee80211_vif *vif, + enum iwl_mld_emlsr_blocked reason, u8 link_to_keep) +{ + return _iwl_mld_emlsr_block(mld, vif, reason, link_to_keep, true); +} + +static void _iwl_mld_select_links(struct iwl_mld *mld, + struct ieee80211_vif *vif); + +void iwl_mld_trigger_link_selection(struct iwl_mld *mld, + struct ieee80211_vif *vif) +{ + bool last_scan_was_recent = + time_before(jiffies, mld->scan.last_mlo_scan_jiffies + + IWL_MLD_SCAN_EXPIRE_TIME); + + if (last_scan_was_recent) { + IWL_DEBUG_EHT(mld, "MLO scan was recent, skip.\n"); + _iwl_mld_select_links(mld, vif); + } else { + IWL_DEBUG_EHT(mld, "Doing link selection after MLO scan\n"); + iwl_mld_int_mlo_scan(mld, vif); + } +} + +void iwl_mld_unblock_emlsr(struct iwl_mld *mld, struct ieee80211_vif *vif, + enum iwl_mld_emlsr_blocked reason) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + + lockdep_assert_wiphy(mld->wiphy); + + if (!IWL_MLD_AUTO_EML_ENABLE || !iwl_mld_vif_has_emlsr_cap(vif)) + return; + + if (!(mld_vif->emlsr.blocked_reasons & reason)) + return; + + mld_vif->emlsr.blocked_reasons &= ~reason; + + IWL_DEBUG_INFO(mld, + "Unblocking EMLSR mode. reason = %s (0x%x)\n", + iwl_mld_get_emlsr_blocked_string(reason), reason); + iwl_mld_print_emlsr_blocked(mld, mld_vif->emlsr.blocked_reasons); + + if (reason == IWL_MLD_EMLSR_BLOCKED_TPT) + wiphy_delayed_work_queue(mld_vif->mld->wiphy, + &mld_vif->emlsr.check_tpt_wk, + round_jiffies_relative(IWL_MLD_TPT_COUNT_WINDOW)); + + if (mld_vif->emlsr.blocked_reasons) + return; + + IWL_DEBUG_INFO(mld, "EMLSR is unblocked\n"); + iwl_mld_trigger_link_selection(mld, vif); +} + +static void +iwl_mld_vif_iter_emlsr_mode_notif(void *data, u8 *mac, + struct ieee80211_vif *vif) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_esr_mode_notif *notif = (void *)data; + + if (!iwl_mld_vif_has_emlsr_cap(vif)) + return; + + switch (le32_to_cpu(notif->action)) { + case ESR_RECOMMEND_ENTER: + iwl_mld_unblock_emlsr(mld_vif->mld, vif, + IWL_MLD_EMLSR_BLOCKED_FW); + break; + case ESR_RECOMMEND_LEAVE: + iwl_mld_block_emlsr(mld_vif->mld, vif, + IWL_MLD_EMLSR_BLOCKED_FW, + iwl_mld_get_primary_link(vif)); + break; + case ESR_FORCE_LEAVE: + default: + /* ESR_FORCE_LEAVE should not happen at this point */ + IWL_WARN(mld_vif->mld, "Unexpected EMLSR notification: %d\n", + le32_to_cpu(notif->action)); + } +} + +void iwl_mld_handle_emlsr_mode_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + ieee80211_iterate_active_interfaces_mtx(mld->hw, + IEEE80211_IFACE_ITER_NORMAL, + iwl_mld_vif_iter_emlsr_mode_notif, + pkt->data); +} + +static void +iwl_mld_vif_iter_disconnect_emlsr(void *data, u8 *mac, + struct ieee80211_vif *vif) +{ + if (!iwl_mld_vif_has_emlsr_cap(vif)) + return; + + ieee80211_connection_loss(vif); +} + +void iwl_mld_handle_emlsr_trans_fail_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + const struct iwl_esr_trans_fail_notif *notif = (const void *)pkt->data; + u32 fw_link_id = le32_to_cpu(notif->link_id); + struct ieee80211_bss_conf *bss_conf = + iwl_mld_fw_id_to_link_conf(mld, fw_link_id); + + IWL_DEBUG_INFO(mld, "Failed to %s EMLSR on link %d (FW: %d), reason %d\n", + le32_to_cpu(notif->activation) ? "enter" : "exit", + bss_conf ? bss_conf->link_id : -1, + le32_to_cpu(notif->link_id), + le32_to_cpu(notif->err_code)); + + if (IWL_FW_CHECK(mld, !bss_conf, + "FW reported failure to %sactivate EMLSR on a non-existing link: %d\n", + le32_to_cpu(notif->activation) ? "" : "de", + fw_link_id)) { + ieee80211_iterate_active_interfaces_mtx( + mld->hw, IEEE80211_IFACE_ITER_NORMAL, + iwl_mld_vif_iter_disconnect_emlsr, NULL); + return; + } + + /* Disconnect if we failed to deactivate a link */ + if (!le32_to_cpu(notif->activation)) { + ieee80211_connection_loss(bss_conf->vif); + return; + } + + /* + * We failed to activate the second link, go back to the link specified + * by the firmware as that is the one that is still valid now. + */ + iwl_mld_exit_emlsr(mld, bss_conf->vif, IWL_MLD_EMLSR_EXIT_FAIL_ENTRY, + bss_conf->link_id); +} + +/* Active non-station link tracking */ +static void iwl_mld_count_non_bss_links(void *_data, u8 *mac, + struct ieee80211_vif *vif) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + int *count = _data; + + if (ieee80211_vif_type_p2p(vif) == NL80211_IFTYPE_STATION) + return; + + *count += iwl_mld_count_active_links(mld_vif->mld, vif); +} + +struct iwl_mld_update_emlsr_block_data { + bool block; + int result; +}; + +static void +iwl_mld_vif_iter_update_emlsr_non_bss_block(void *_data, u8 *mac, + struct ieee80211_vif *vif) +{ + struct iwl_mld_update_emlsr_block_data *data = _data; + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + int ret; + + if (data->block) { + ret = iwl_mld_block_emlsr_sync(mld_vif->mld, vif, + IWL_MLD_EMLSR_BLOCKED_NON_BSS, + iwl_mld_get_primary_link(vif)); + if (ret) + data->result = ret; + } else { + iwl_mld_unblock_emlsr(mld_vif->mld, vif, + IWL_MLD_EMLSR_BLOCKED_NON_BSS); + } +} + +int iwl_mld_emlsr_check_non_bss_block(struct iwl_mld *mld, + int pending_link_changes) +{ + /* An active link of a non-station vif blocks EMLSR. Upon activation + * block EMLSR on the bss vif. Upon deactivation, check if this link + * was the last non-station link active, and if so unblock the bss vif + */ + struct iwl_mld_update_emlsr_block_data block_data = {}; + int count = pending_link_changes; + + /* No need to count if we are activating a non-BSS link */ + if (count <= 0) + ieee80211_iterate_active_interfaces_mtx(mld->hw, + IEEE80211_IFACE_ITER_NORMAL, + iwl_mld_count_non_bss_links, + &count); + + /* + * We could skip updating it if the block change did not change (and + * pending_link_changes is non-zero). + */ + block_data.block = !!count; + + ieee80211_iterate_active_interfaces_mtx(mld->hw, + IEEE80211_IFACE_ITER_NORMAL, + iwl_mld_vif_iter_update_emlsr_non_bss_block, + &block_data); + + return block_data.result; +} + +#define EMLSR_SEC_LINK_MIN_PERC 10 +#define EMLSR_MIN_TX 3000 +#define EMLSR_MIN_RX 400 + +void iwl_mld_emlsr_check_tpt(struct wiphy *wiphy, struct wiphy_work *wk) +{ + struct iwl_mld_vif *mld_vif = container_of(wk, struct iwl_mld_vif, + emlsr.check_tpt_wk.work); + struct ieee80211_vif *vif = + container_of((void *)mld_vif, struct ieee80211_vif, drv_priv); + struct iwl_mld *mld = mld_vif->mld; + struct iwl_mld_sta *mld_sta; + struct iwl_mld_link *sec_link; + unsigned long total_tx = 0, total_rx = 0; + unsigned long sec_link_tx = 0, sec_link_rx = 0; + u8 sec_link_tx_perc, sec_link_rx_perc; + s8 sec_link_id; + + if (!iwl_mld_vif_has_emlsr_cap(vif) || !mld_vif->ap_sta) + return; + + mld_sta = iwl_mld_sta_from_mac80211(mld_vif->ap_sta); + + /* We only count for the AP sta in a MLO connection */ + if (!mld_sta->mpdu_counters) + return; + + /* This wk should only run when the TPT blocker isn't set. + * When the blocker is set, the decision to remove it, as well as + * clearing the counters is done in DP (to avoid having a wk every + * 5 seconds when idle. When the blocker is unset, we are not idle anyway) + */ + if (WARN_ON(mld_vif->emlsr.blocked_reasons & IWL_MLD_EMLSR_BLOCKED_TPT)) + return; + /* + * TPT is unblocked, need to check if the TPT criteria is still met. + * + * If EMLSR is active, then we also need to check the secondar link + * requirements. + */ + if (iwl_mld_emlsr_active(vif)) { + sec_link_id = iwl_mld_get_other_link(vif, iwl_mld_get_primary_link(vif)); + sec_link = iwl_mld_link_dereference_check(mld_vif, sec_link_id); + if (WARN_ON_ONCE(!sec_link)) + return; + /* We need the FW ID here */ + sec_link_id = sec_link->fw_id; + } else { + sec_link_id = -1; + } + + /* Sum up RX and TX MPDUs from the different queues/links */ + for (int q = 0; q < mld->trans->num_rx_queues; q++) { + struct iwl_mld_per_q_mpdu_counter *queue_counter = + &mld_sta->mpdu_counters[q]; + + spin_lock_bh(&queue_counter->lock); + + /* The link IDs that doesn't exist will contain 0 */ + for (int link = 0; + link < ARRAY_SIZE(queue_counter->per_link); + link++) { + total_tx += queue_counter->per_link[link].tx; + total_rx += queue_counter->per_link[link].rx; + } + + if (sec_link_id != -1) { + sec_link_tx += queue_counter->per_link[sec_link_id].tx; + sec_link_rx += queue_counter->per_link[sec_link_id].rx; + } + + memset(queue_counter->per_link, 0, + sizeof(queue_counter->per_link)); + + spin_unlock_bh(&queue_counter->lock); + } + + IWL_DEBUG_INFO(mld, "total Tx MPDUs: %ld. total Rx MPDUs: %ld\n", + total_tx, total_rx); + + /* If we don't have enough MPDUs - exit EMLSR */ + if (total_tx < IWL_MLD_ENTER_EMLSR_TPT_THRESH && + total_rx < IWL_MLD_ENTER_EMLSR_TPT_THRESH) { + iwl_mld_block_emlsr(mld, vif, IWL_MLD_EMLSR_BLOCKED_TPT, + iwl_mld_get_primary_link(vif)); + return; + } + + /* EMLSR is not active */ + if (sec_link_id == -1) + return; + + IWL_DEBUG_INFO(mld, "Secondary Link %d: Tx MPDUs: %ld. Rx MPDUs: %ld\n", + sec_link_id, sec_link_tx, sec_link_rx); + + /* Calculate the percentage of the secondary link TX/RX */ + sec_link_tx_perc = total_tx ? sec_link_tx * 100 / total_tx : 0; + sec_link_rx_perc = total_rx ? sec_link_rx * 100 / total_rx : 0; + + /* + * The TX/RX percentage is checked only if it exceeds the required + * minimum. In addition, RX is checked only if the TX check failed. + */ + if ((total_tx > EMLSR_MIN_TX && + sec_link_tx_perc < EMLSR_SEC_LINK_MIN_PERC) || + (total_rx > EMLSR_MIN_RX && + sec_link_rx_perc < EMLSR_SEC_LINK_MIN_PERC)) { + iwl_mld_exit_emlsr(mld, vif, IWL_MLD_EMLSR_EXIT_LINK_USAGE, + iwl_mld_get_primary_link(vif)); + return; + } + + /* Check again when the next window ends */ + wiphy_delayed_work_queue(mld_vif->mld->wiphy, + &mld_vif->emlsr.check_tpt_wk, + round_jiffies_relative(IWL_MLD_TPT_COUNT_WINDOW)); +} + +void iwl_mld_emlsr_unblock_tpt_wk(struct wiphy *wiphy, struct wiphy_work *wk) +{ + struct iwl_mld_vif *mld_vif = container_of(wk, struct iwl_mld_vif, + emlsr.unblock_tpt_wk); + struct ieee80211_vif *vif = + container_of((void *)mld_vif, struct ieee80211_vif, drv_priv); + + iwl_mld_unblock_emlsr(mld_vif->mld, vif, IWL_MLD_EMLSR_BLOCKED_TPT); +} + +/* + * Link selection + */ +struct iwl_mld_link_sel_data { + u8 link_id; + const struct cfg80211_chan_def *chandef; + s32 signal; + u16 grade; +}; + +s8 iwl_mld_get_emlsr_rssi_thresh(struct iwl_mld *mld, + const struct cfg80211_chan_def *chandef, + bool low) +{ + if (WARN_ON(chandef->chan->band != NL80211_BAND_2GHZ && + chandef->chan->band != NL80211_BAND_5GHZ && + chandef->chan->band != NL80211_BAND_6GHZ)) + return S8_MAX; + +#define RSSI_THRESHOLD(_low, _bw) \ + (_low) ? IWL_MLD_LOW_RSSI_THRESH_##_bw##MHZ \ + : IWL_MLD_HIGH_RSSI_THRESH_##_bw##MHZ + + switch (chandef->width) { + case NL80211_CHAN_WIDTH_20_NOHT: + case NL80211_CHAN_WIDTH_20: + /* 320 MHz has the same thresholds as 20 MHz */ + case NL80211_CHAN_WIDTH_320: + return RSSI_THRESHOLD(low, 20); + case NL80211_CHAN_WIDTH_40: + return RSSI_THRESHOLD(low, 40); + case NL80211_CHAN_WIDTH_80: + return RSSI_THRESHOLD(low, 80); + case NL80211_CHAN_WIDTH_160: + return RSSI_THRESHOLD(low, 160); + default: + WARN_ON(1); + return S8_MAX; + } +#undef RSSI_THRESHOLD +} + +static u32 +iwl_mld_emlsr_disallowed_with_link(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct iwl_mld_link_sel_data *link, + bool primary) +{ + struct wiphy *wiphy = mld->wiphy; + struct ieee80211_bss_conf *conf; + enum iwl_mld_emlsr_exit ret = 0; + + conf = wiphy_dereference(wiphy, vif->link_conf[link->link_id]); + if (WARN_ON_ONCE(!conf)) + return false; + + if (link->chandef->chan->band == NL80211_BAND_2GHZ && mld->bt_is_active) + ret |= IWL_MLD_EMLSR_EXIT_BT_COEX; + + if (link->signal < + iwl_mld_get_emlsr_rssi_thresh(mld, link->chandef, false)) + ret |= IWL_MLD_EMLSR_EXIT_LOW_RSSI; + + if (conf->csa_active) + ret |= IWL_MLD_EMLSR_EXIT_CSA; + + if (ret) { + IWL_DEBUG_INFO(mld, + "Link %d is not allowed for EMLSR as %s\n", + link->link_id, + primary ? "primary" : "secondary"); + iwl_mld_print_emlsr_exit(mld, ret); + } + + return ret; +} + +static u8 +iwl_mld_set_link_sel_data(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct iwl_mld_link_sel_data *data, + unsigned long usable_links, + u8 *best_link_idx) +{ + u8 n_data = 0; + u16 max_grade = 0; + unsigned long link_id; + + /* + * TODO: don't select links that weren't discovered in the last scan + * This requires mac80211 (or cfg80211) changes to forward/track when + * a BSS was last updated. cfg80211 already tracks this information but + * it is not exposed within the kernel. + */ + for_each_set_bit(link_id, &usable_links, IEEE80211_MLD_MAX_NUM_LINKS) { + struct ieee80211_bss_conf *link_conf = + link_conf_dereference_protected(vif, link_id); + + if (WARN_ON_ONCE(!link_conf)) + continue; + + /* Ignore any BSS that was not seen in the last 5 seconds */ + if (ktime_before(link_conf->bss->ts_boottime, + ktime_sub_ns(ktime_get_boottime_ns(), + (u64)5 * NSEC_PER_SEC))) + continue; + + data[n_data].link_id = link_id; + data[n_data].chandef = &link_conf->chanreq.oper; + data[n_data].signal = MBM_TO_DBM(link_conf->bss->signal); + data[n_data].grade = iwl_mld_get_link_grade(mld, link_conf); + + if (n_data == 0 || data[n_data].grade > max_grade) { + max_grade = data[n_data].grade; + *best_link_idx = n_data; + } + n_data++; + } + + return n_data; +} + +static bool +iwl_mld_valid_emlsr_pair(struct ieee80211_vif *vif, + struct iwl_mld_link_sel_data *a, + struct iwl_mld_link_sel_data *b) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mld *mld = mld_vif->mld; + enum iwl_mld_emlsr_exit ret = 0; + + /* Per-link considerations */ + if (iwl_mld_emlsr_disallowed_with_link(mld, vif, a, true) || + iwl_mld_emlsr_disallowed_with_link(mld, vif, b, false)) + return false; + + if (a->chandef->chan->band == b->chandef->chan->band) { + ret |= IWL_MLD_EMLSR_EXIT_EQUAL_BAND; + } else if (a->chandef->width != b->chandef->width) { + /* TODO: task=EMLSR task=statistics + * replace BANDWIDTH exit reason with channel load criteria + */ + ret |= IWL_MLD_EMLSR_EXIT_BANDWIDTH; + } + + if (ret) { + IWL_DEBUG_INFO(mld, + "Links %d and %d are not a valid pair for EMLSR\n", + a->link_id, b->link_id); + IWL_DEBUG_INFO(mld, + "Links bandwidth are: %d and %d\n", + nl80211_chan_width_to_mhz(a->chandef->width), + nl80211_chan_width_to_mhz(b->chandef->width)); + iwl_mld_print_emlsr_exit(mld, ret); + return false; + } + + return true; +} + +/* Calculation is done with fixed-point with a scaling factor of 1/256 */ +#define SCALE_FACTOR 256 + +/* + * Returns the combined grade of two given links. + * Returns 0 if EMLSR is not allowed with these 2 links. + */ +static +unsigned int iwl_mld_get_emlsr_grade(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct iwl_mld_link_sel_data *a, + struct iwl_mld_link_sel_data *b, + u8 *primary_id) +{ + struct ieee80211_bss_conf *primary_conf; + struct wiphy *wiphy = ieee80211_vif_to_wdev(vif)->wiphy; + unsigned int primary_load; + + lockdep_assert_wiphy(wiphy); + + /* a is always primary, b is always secondary */ + if (b->grade > a->grade) + swap(a, b); + + *primary_id = a->link_id; + + if (!iwl_mld_valid_emlsr_pair(vif, a, b)) + return 0; + + primary_conf = wiphy_dereference(wiphy, vif->link_conf[*primary_id]); + + if (WARN_ON_ONCE(!primary_conf)) + return 0; + + primary_load = iwl_mld_get_chan_load(mld, primary_conf); + + /* The more the primary link is loaded, the more worthwhile EMLSR becomes */ + return a->grade + ((b->grade * primary_load) / SCALE_FACTOR); +} + +static void _iwl_mld_select_links(struct iwl_mld *mld, + struct ieee80211_vif *vif) +{ + struct iwl_mld_link_sel_data data[IEEE80211_MLD_MAX_NUM_LINKS]; + struct iwl_mld_link_sel_data *best_link; + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + int max_active_links = iwl_mld_max_active_links(mld, vif); + u16 new_active, usable_links = ieee80211_vif_usable_links(vif); + u8 best_idx, new_primary, n_data; + u16 max_grade; + + lockdep_assert_wiphy(mld->wiphy); + + if (!mld_vif->authorized || hweight16(usable_links) <= 1) + return; + + if (WARN(time_before(mld->scan.last_mlo_scan_jiffies, + jiffies - IWL_MLD_SCAN_EXPIRE_TIME), + "Last MLO scan was too long ago, can't select links\n")) + return; + + /* The logic below is simple and not suited for more than 2 links */ + WARN_ON_ONCE(max_active_links > 2); + + n_data = iwl_mld_set_link_sel_data(mld, vif, data, usable_links, + &best_idx); + + if (WARN(!n_data, "Couldn't find a valid grade for any link!\n")) + return; + + /* Default to selecting the single best link */ + best_link = &data[best_idx]; + new_primary = best_link->link_id; + new_active = BIT(best_link->link_id); + max_grade = best_link->grade; + + /* If EMLSR is not possible, activate the best link */ + if (max_active_links == 1 || n_data == 1 || + !iwl_mld_vif_has_emlsr_cap(vif) || !IWL_MLD_AUTO_EML_ENABLE || + mld_vif->emlsr.blocked_reasons) + goto set_active; + + /* Try to find the best link combination */ + for (u8 a = 0; a < n_data; a++) { + for (u8 b = a + 1; b < n_data; b++) { + u8 best_in_pair; + u16 emlsr_grade = + iwl_mld_get_emlsr_grade(mld, vif, + &data[a], &data[b], + &best_in_pair); + + /* + * Prefer (new) EMLSR combination to prefer EMLSR over + * a single link. + */ + if (emlsr_grade < max_grade) + continue; + + max_grade = emlsr_grade; + new_primary = best_in_pair; + new_active = BIT(data[a].link_id) | + BIT(data[b].link_id); + } + } + +set_active: + IWL_DEBUG_INFO(mld, "Link selection result: 0x%x. Primary = %d\n", + new_active, new_primary); + + mld_vif->emlsr.selected_primary = new_primary; + mld_vif->emlsr.selected_links = new_active; + + ieee80211_set_active_links_async(vif, new_active); +} + +static void iwl_mld_vif_iter_select_links(void *_data, u8 *mac, + struct ieee80211_vif *vif) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mld *mld = mld_vif->mld; + + _iwl_mld_select_links(mld, vif); +} + +void iwl_mld_select_links(struct iwl_mld *mld) +{ + ieee80211_iterate_active_interfaces_mtx(mld->hw, + IEEE80211_IFACE_ITER_NORMAL, + iwl_mld_vif_iter_select_links, + NULL); +} + +void iwl_mld_emlsr_check_equal_bw(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link) +{ + u8 other_link_id = iwl_mld_get_other_link(vif, link->link_id); + struct ieee80211_bss_conf *other_link = + link_conf_dereference_check(vif, other_link_id); + + if (!ieee80211_vif_link_active(vif, link->link_id) || + !iwl_mld_emlsr_active(vif) || + WARN_ON(link->link_id == other_link_id || !other_link)) + return; + + if (link->chanreq.oper.width != other_link->chanreq.oper.width) + iwl_mld_exit_emlsr(mld, vif, IWL_MLD_EMLSR_EXIT_BANDWIDTH, + iwl_mld_get_primary_link(vif)); +} + +static void iwl_mld_emlsr_check_bt_iter(void *_data, u8 *mac, + struct ieee80211_vif *vif) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mld *mld = mld_vif->mld; + struct ieee80211_bss_conf *link; + unsigned int link_id; + + if (!mld->bt_is_active) { + iwl_mld_retry_emlsr(mld, vif); + return; + } + + /* BT is turned ON but we are not in EMLSR, nothing to do */ + if (!iwl_mld_emlsr_active(vif)) + return; + + /* In EMLSR and BT is turned ON */ + + for_each_vif_active_link(vif, link, link_id) { + if (WARN_ON(!link->chanreq.oper.chan)) + continue; + + if (link->chanreq.oper.chan->band == NL80211_BAND_2GHZ) { + iwl_mld_exit_emlsr(mld, vif, IWL_MLD_EMLSR_EXIT_BT_COEX, + iwl_mld_get_primary_link(vif)); + return; + } + } +} + +void iwl_mld_emlsr_check_bt(struct iwl_mld *mld) +{ + ieee80211_iterate_active_interfaces_mtx(mld->hw, + IEEE80211_IFACE_ITER_NORMAL, + iwl_mld_emlsr_check_bt_iter, + NULL); +} + +static void iwl_mld_emlsr_check_chan_load_iter(void *_data, u8 *mac, + struct ieee80211_vif *vif) +{ + struct iwl_mld *mld = (struct iwl_mld *)_data; + struct ieee80211_bss_conf *prim_link; + unsigned int prim_link_id; + int chan_load; + + if (!iwl_mld_emlsr_active(vif)) + return; + + prim_link_id = iwl_mld_get_primary_link(vif); + prim_link = link_conf_dereference_protected(vif, prim_link_id); + if (WARN_ON(!prim_link)) + return; + + chan_load = iwl_mld_get_chan_load_by_others(mld, prim_link, true); + + if (chan_load < 0) + return; + + /* chan_load is in range [0,255] */ + if (chan_load < NORMALIZE_PERCENT_TO_255(IWL_MLD_CHAN_LOAD_THRESH)) + iwl_mld_exit_emlsr(mld, vif, IWL_MLD_EMLSR_EXIT_CHAN_LOAD, + prim_link_id); +} + +void iwl_mld_emlsr_check_chan_load(struct iwl_mld *mld) +{ + ieee80211_iterate_active_interfaces_mtx(mld->hw, + IEEE80211_IFACE_ITER_NORMAL, + iwl_mld_emlsr_check_chan_load_iter, + (void *)(uintptr_t)mld); +} + +void iwl_mld_retry_emlsr(struct iwl_mld *mld, struct ieee80211_vif *vif) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + + if (!iwl_mld_vif_has_emlsr_cap(vif) || iwl_mld_emlsr_active(vif) || + mld_vif->emlsr.blocked_reasons) + return; + + iwl_mld_trigger_link_selection(mld, vif); +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/mlo.h b/drivers/net/wireless/intel/iwlwifi/mld/mlo.h new file mode 100644 index 0000000000000..fd1abe8e6084a --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/mlo.h @@ -0,0 +1,157 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024-2025 Intel Corporation + */ +#ifndef __iwl_mld_mlo_h__ +#define __iwl_mld_mlo_h__ + +#include +#include +#include +#include "iwl-config.h" +#include "iwl-trans.h" +#include "iface.h" + +struct iwl_mld; + +void iwl_mld_emlsr_prevent_done_wk(struct wiphy *wiphy, struct wiphy_work *wk); +void iwl_mld_emlsr_tmp_non_bss_done_wk(struct wiphy *wiphy, + struct wiphy_work *wk); + +static inline bool iwl_mld_emlsr_active(struct ieee80211_vif *vif) +{ + /* Set on phy context activation, so should be a good proxy */ + return !!(vif->driver_flags & IEEE80211_VIF_EML_ACTIVE); +} + +static inline bool iwl_mld_vif_has_emlsr_cap(struct ieee80211_vif *vif) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + + /* We only track/permit EMLSR state once authorized */ + if (!mld_vif->authorized) + return false; + + /* No EMLSR on dual radio devices */ + return ieee80211_vif_type_p2p(vif) == NL80211_IFTYPE_STATION && + ieee80211_vif_is_mld(vif) && + vif->cfg.eml_cap & IEEE80211_EML_CAP_EMLSR_SUPP && + !CSR_HW_RFID_IS_CDB(mld_vif->mld->trans->hw_rf_id); +} + +static inline int +iwl_mld_max_active_links(struct iwl_mld *mld, struct ieee80211_vif *vif) +{ + if (vif->type == NL80211_IFTYPE_AP) + return mld->fw->ucode_capa.num_beacons; + + if (ieee80211_vif_type_p2p(vif) == NL80211_IFTYPE_STATION) + return IWL_FW_MAX_ACTIVE_LINKS_NUM; + + /* For now, do not accept more links on other interface types */ + return 1; +} + +static inline int +iwl_mld_count_active_links(struct iwl_mld *mld, struct ieee80211_vif *vif) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mld_link *mld_link; + int n_active = 0; + + for_each_mld_vif_valid_link(mld_vif, mld_link) { + if (rcu_access_pointer(mld_link->chan_ctx)) + n_active++; + } + + return n_active; +} + +static inline u8 iwl_mld_get_primary_link(struct ieee80211_vif *vif) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + + lockdep_assert_wiphy(mld_vif->mld->wiphy); + + if (!ieee80211_vif_is_mld(vif) || WARN_ON(!vif->active_links)) + return 0; + + /* In AP mode, there is no primary link */ + if (vif->type == NL80211_IFTYPE_AP) + return __ffs(vif->active_links); + + if (iwl_mld_emlsr_active(vif) && + !WARN_ON(!(BIT(mld_vif->emlsr.primary) & vif->active_links))) + return mld_vif->emlsr.primary; + + return __ffs(vif->active_links); +} + +/* + * For non-MLO/single link, this will return the deflink/single active link, + * respectively + */ +static inline u8 iwl_mld_get_other_link(struct ieee80211_vif *vif, u8 link_id) +{ + switch (hweight16(vif->active_links)) { + case 0: + return 0; + default: + WARN_ON(1); + fallthrough; + case 1: + return __ffs(vif->active_links); + case 2: + return __ffs(vif->active_links & ~BIT(link_id)); + } +} + +s8 iwl_mld_get_emlsr_rssi_thresh(struct iwl_mld *mld, + const struct cfg80211_chan_def *chandef, + bool low); + +/* EMLSR block/unblock and exit */ +void iwl_mld_block_emlsr(struct iwl_mld *mld, struct ieee80211_vif *vif, + enum iwl_mld_emlsr_blocked reason, u8 link_to_keep); +int iwl_mld_block_emlsr_sync(struct iwl_mld *mld, struct ieee80211_vif *vif, + enum iwl_mld_emlsr_blocked reason, u8 link_to_keep); +void iwl_mld_unblock_emlsr(struct iwl_mld *mld, struct ieee80211_vif *vif, + enum iwl_mld_emlsr_blocked reason); +void iwl_mld_exit_emlsr(struct iwl_mld *mld, struct ieee80211_vif *vif, + enum iwl_mld_emlsr_exit exit, u8 link_to_keep); + +int iwl_mld_emlsr_check_non_bss_block(struct iwl_mld *mld, + int pending_link_changes); + +void iwl_mld_handle_emlsr_mode_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt); +void iwl_mld_handle_emlsr_trans_fail_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt); + +void iwl_mld_emlsr_check_tpt(struct wiphy *wiphy, struct wiphy_work *wk); +void iwl_mld_emlsr_unblock_tpt_wk(struct wiphy *wiphy, struct wiphy_work *wk); + +void iwl_mld_select_links(struct iwl_mld *mld); + +void iwl_mld_emlsr_check_equal_bw(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link); + +void iwl_mld_emlsr_check_bt(struct iwl_mld *mld); + +void iwl_mld_emlsr_check_chan_load(struct iwl_mld *mld); + +void iwl_mld_trigger_link_selection(struct iwl_mld *mld, + struct ieee80211_vif *vif); + +/** + * iwl_mld_retry_emlsr - Retry entering EMLSR + * @mld: MLD context + * @vif: VIF to retry EMLSR on + * + * Retry entering EMLSR on the given VIF. + * Use this if one of the parameters that can prevent EMLSR has changed. + */ +void iwl_mld_retry_emlsr(struct iwl_mld *mld, struct ieee80211_vif *vif); + +#endif /* __iwl_mld_mlo_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/notif.c b/drivers/net/wireless/intel/iwlwifi/mld/notif.c new file mode 100644 index 0000000000000..06caedf9b0ba1 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/notif.c @@ -0,0 +1,758 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024-2025 Intel Corporation + */ + +#include "mld.h" +#include "notif.h" +#include "scan.h" +#include "iface.h" +#include "mlo.h" +#include "iwl-trans.h" +#include "fw/file.h" +#include "fw/dbg.h" +#include "fw/api/cmdhdr.h" +#include "fw/api/mac-cfg.h" +#include "session-protect.h" +#include "fw/api/time-event.h" +#include "fw/api/tx.h" +#include "fw/api/rs.h" +#include "fw/api/offload.h" +#include "fw/api/stats.h" +#include "fw/api/rfi.h" +#include "fw/api/coex.h" + +#include "mcc.h" +#include "link.h" +#include "tx.h" +#include "rx.h" +#include "tlc.h" +#include "agg.h" +#include "mac80211.h" +#include "thermal.h" +#include "roc.h" +#include "stats.h" +#include "coex.h" +#include "time_sync.h" +#include "ftm-initiator.h" + +/* Please use this in an increasing order of the versions */ +#define CMD_VER_ENTRY(_ver, _struct) \ + { .size = sizeof(struct _struct), .ver = _ver }, +#define CMD_VERSIONS(name, ...) \ + static const struct iwl_notif_struct_size \ + iwl_notif_struct_sizes_##name[] = { __VA_ARGS__ }; + +#define RX_HANDLER_NO_OBJECT(_grp, _cmd, _name, _context) \ + {.cmd_id = WIDE_ID(_grp, _cmd), \ + .context = _context, \ + .fn = iwl_mld_handle_##_name, \ + .sizes = iwl_notif_struct_sizes_##_name, \ + .n_sizes = ARRAY_SIZE(iwl_notif_struct_sizes_##_name), \ + }, + +/* Use this for Rx handlers that do not need notification validation */ +#define RX_HANDLER_NO_VAL(_grp, _cmd, _name, _context) \ + {.cmd_id = WIDE_ID(_grp, _cmd), \ + .context = _context, \ + .fn = iwl_mld_handle_##_name, \ + }, + +#define RX_HANDLER_VAL_FN(_grp, _cmd, _name, _context) \ + { .cmd_id = WIDE_ID(_grp, _cmd), \ + .context = _context, \ + .fn = iwl_mld_handle_##_name, \ + .val_fn = iwl_mld_validate_##_name, \ + }, + +#define DEFINE_SIMPLE_CANCELLATION(name, notif_struct, id_member) \ +static bool iwl_mld_cancel_##name##_notif(struct iwl_mld *mld, \ + struct iwl_rx_packet *pkt, \ + u32 obj_id) \ +{ \ + const struct notif_struct *notif = (const void *)pkt->data; \ + \ + return obj_id == _Generic((notif)->id_member, \ + __le32: le32_to_cpu((notif)->id_member), \ + __le16: le16_to_cpu((notif)->id_member), \ + u8: (notif)->id_member); \ +} + +static bool iwl_mld_always_cancel(struct iwl_mld *mld, + struct iwl_rx_packet *pkt, + u32 obj_id) +{ + return true; +} + +/* Currently only defined for the RX_HANDLER_SIZES options. Use this for + * notifications that belong to a specific object, and that should be + * canceled when the object is removed + */ +#define RX_HANDLER_OF_OBJ(_grp, _cmd, _name, _obj_type) \ + {.cmd_id = WIDE_ID(_grp, _cmd), \ + /* Only async handlers can be canceled */ \ + .context = RX_HANDLER_ASYNC, \ + .fn = iwl_mld_handle_##_name, \ + .sizes = iwl_notif_struct_sizes_##_name, \ + .n_sizes = ARRAY_SIZE(iwl_notif_struct_sizes_##_name), \ + .obj_type = IWL_MLD_OBJECT_TYPE_##_obj_type, \ + .cancel = iwl_mld_cancel_##_name, \ + }, + +#define RX_HANDLER_OF_LINK(_grp, _cmd, _name) \ + RX_HANDLER_OF_OBJ(_grp, _cmd, _name, LINK) \ + +#define RX_HANDLER_OF_VIF(_grp, _cmd, _name) \ + RX_HANDLER_OF_OBJ(_grp, _cmd, _name, VIF) \ + +#define RX_HANDLER_OF_STA(_grp, _cmd, _name) \ + RX_HANDLER_OF_OBJ(_grp, _cmd, _name, STA) \ + +#define RX_HANDLER_OF_ROC(_grp, _cmd, _name) \ + RX_HANDLER_OF_OBJ(_grp, _cmd, _name, ROC) + +#define RX_HANDLER_OF_SCAN(_grp, _cmd, _name) \ + RX_HANDLER_OF_OBJ(_grp, _cmd, _name, SCAN) + +#define RX_HANDLER_OF_FTM_REQ(_grp, _cmd, _name) \ + RX_HANDLER_OF_OBJ(_grp, _cmd, _name, FTM_REQ) + +static void iwl_mld_handle_mfuart_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + struct iwl_mfuart_load_notif *mfuart_notif = (void *)pkt->data; + + IWL_DEBUG_INFO(mld, + "MFUART: installed ver: 0x%08x, external ver: 0x%08x\n", + le32_to_cpu(mfuart_notif->installed_ver), + le32_to_cpu(mfuart_notif->external_ver)); + IWL_DEBUG_INFO(mld, + "MFUART: status: 0x%08x, duration: 0x%08x image size: 0x%08x\n", + le32_to_cpu(mfuart_notif->status), + le32_to_cpu(mfuart_notif->duration), + le32_to_cpu(mfuart_notif->image_size)); +} + +static void iwl_mld_mu_mimo_iface_iterator(void *_data, u8 *mac, + struct ieee80211_vif *vif) +{ + struct ieee80211_bss_conf *bss_conf = &vif->bss_conf; + unsigned int link_id = 0; + + if (WARN(hweight16(vif->active_links) > 1, + "no support for this notif while in EMLSR 0x%x\n", + vif->active_links)) + return; + + if (ieee80211_vif_is_mld(vif)) { + link_id = __ffs(vif->active_links); + bss_conf = link_conf_dereference_check(vif, link_id); + } + + if (!WARN_ON(!bss_conf) && bss_conf->mu_mimo_owner) { + const struct iwl_mu_group_mgmt_notif *notif = _data; + + BUILD_BUG_ON(sizeof(notif->membership_status) != + WLAN_MEMBERSHIP_LEN); + BUILD_BUG_ON(sizeof(notif->user_position) != + WLAN_USER_POSITION_LEN); + + /* MU-MIMO Group Id action frame is little endian. We treat + * the data received from firmware as if it came from the + * action frame, so no conversion is needed. + */ + ieee80211_update_mu_groups(vif, link_id, + (u8 *)¬if->membership_status, + (u8 *)¬if->user_position); + } +} + +/* This handler is called in SYNC mode because it needs to be serialized with + * Rx as specified in ieee80211_update_mu_groups()'s documentation. + */ +static void iwl_mld_handle_mu_mimo_grp_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + struct iwl_mu_group_mgmt_notif *notif = (void *)pkt->data; + + ieee80211_iterate_active_interfaces_atomic(mld->hw, + IEEE80211_IFACE_ITER_NORMAL, + iwl_mld_mu_mimo_iface_iterator, + notif); +} + +static void +iwl_mld_handle_stored_beacon_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + unsigned int pkt_len = iwl_rx_packet_payload_len(pkt); + struct iwl_stored_beacon_notif *sb = (void *)pkt->data; + struct ieee80211_rx_status rx_status = {}; + struct sk_buff *skb; + u32 size = le32_to_cpu(sb->common.byte_count); + + if (size == 0) + return; + + if (pkt_len < struct_size(sb, data, size)) + return; + + skb = alloc_skb(size, GFP_ATOMIC); + if (!skb) { + IWL_ERR(mld, "alloc_skb failed\n"); + return; + } + + /* update rx_status according to the notification's metadata */ + rx_status.mactime = le64_to_cpu(sb->common.tsf); + /* TSF as indicated by the firmware is at INA time */ + rx_status.flag |= RX_FLAG_MACTIME_PLCP_START; + rx_status.device_timestamp = le32_to_cpu(sb->common.system_time); + rx_status.band = + iwl_mld_phy_band_to_nl80211(le16_to_cpu(sb->common.band)); + rx_status.freq = + ieee80211_channel_to_frequency(le16_to_cpu(sb->common.channel), + rx_status.band); + + /* copy the data */ + skb_put_data(skb, sb->data, size); + memcpy(IEEE80211_SKB_RXCB(skb), &rx_status, sizeof(rx_status)); + + /* pass it as regular rx to mac80211 */ + ieee80211_rx_napi(mld->hw, NULL, skb, NULL); +} + +static void +iwl_mld_handle_channel_switch_start_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + struct iwl_channel_switch_start_notif *notif = (void *)pkt->data; + u32 link_id = le32_to_cpu(notif->link_id); + struct ieee80211_bss_conf *link_conf = + iwl_mld_fw_id_to_link_conf(mld, link_id); + struct ieee80211_vif *vif; + + if (WARN_ON(!link_conf)) + return; + + vif = link_conf->vif; + + IWL_DEBUG_INFO(mld, + "CSA Start Notification with vif type: %d, link_id: %d\n", + vif->type, + link_conf->link_id); + + switch (vif->type) { + case NL80211_IFTYPE_AP: + /* We don't support canceling a CSA as it was advertised + * by the AP itself + */ + if (!link_conf->csa_active) + return; + + ieee80211_csa_finish(vif, link_conf->link_id); + break; + case NL80211_IFTYPE_STATION: + if (!link_conf->csa_active) { + /* Either unexpected cs notif or mac80211 chose to + * ignore, for example in channel switch to same channel + */ + struct iwl_cancel_channel_switch_cmd cmd = { + .id = cpu_to_le32(link_id), + }; + + if (iwl_mld_send_cmd_pdu(mld, + WIDE_ID(MAC_CONF_GROUP, + CANCEL_CHANNEL_SWITCH_CMD), + &cmd)) + IWL_ERR(mld, + "Failed to cancel the channel switch\n"); + return; + } + + ieee80211_chswitch_done(vif, true, link_conf->link_id); + break; + + default: + WARN(1, "CSA on invalid vif type: %d", vif->type); + } +} + +static void +iwl_mld_handle_channel_switch_error_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + struct iwl_channel_switch_error_notif *notif = (void *)pkt->data; + struct ieee80211_bss_conf *link_conf; + struct ieee80211_vif *vif; + u32 link_id = le32_to_cpu(notif->link_id); + u32 csa_err_mask = le32_to_cpu(notif->csa_err_mask); + + link_conf = iwl_mld_fw_id_to_link_conf(mld, link_id); + if (WARN_ON(!link_conf)) + return; + + vif = link_conf->vif; + + IWL_DEBUG_INFO(mld, "FW reports CSA error: id=%u, csa_err_mask=%u\n", + link_id, csa_err_mask); + + if (csa_err_mask & (CS_ERR_COUNT_ERROR | + CS_ERR_LONG_DELAY_AFTER_CS | + CS_ERR_TX_BLOCK_TIMER_EXPIRED)) + ieee80211_channel_switch_disconnect(vif); +} + +static void iwl_mld_handle_beacon_notification(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + struct iwl_extended_beacon_notif *beacon = (void *)pkt->data; + + mld->ibss_manager = !!beacon->ibss_mgr_status; +} + +/** + * DOC: Notification versioning + * + * The firmware's notifications change from time to time. In order to + * differentiate between different versions of the same notification, the + * firmware advertises the version of each notification. + * Here are listed all the notifications that are supported. Several versions + * of the same notification can be allowed at the same time: + * + * CMD_VERSION(my_multi_version_notif, + * CMD_VER_ENTRY(1, iwl_my_multi_version_notif_ver1) + * CMD_VER_ENTRY(2, iwl_my_multi_version_notif_ver2) + * + * etc... + * + * The driver will enforce that the notification coming from the firmware + * has its version listed here and it'll also enforce that the firmware sent + * at least enough bytes to cover the structure listed in the CMD_VER_ENTRY. + */ + +CMD_VERSIONS(scan_complete_notif, + CMD_VER_ENTRY(1, iwl_umac_scan_complete)) +CMD_VERSIONS(scan_iter_complete_notif, + CMD_VER_ENTRY(2, iwl_umac_scan_iter_complete_notif)) +CMD_VERSIONS(mfuart_notif, + CMD_VER_ENTRY(2, iwl_mfuart_load_notif)) +CMD_VERSIONS(update_mcc, + CMD_VER_ENTRY(1, iwl_mcc_chub_notif)) +CMD_VERSIONS(session_prot_notif, + CMD_VER_ENTRY(3, iwl_session_prot_notif)) +CMD_VERSIONS(missed_beacon_notif, + CMD_VER_ENTRY(5, iwl_missed_beacons_notif)) +CMD_VERSIONS(tx_resp_notif, + CMD_VER_ENTRY(8, iwl_tx_resp)) +CMD_VERSIONS(compressed_ba_notif, + CMD_VER_ENTRY(5, iwl_compressed_ba_notif)) +CMD_VERSIONS(tlc_notif, + CMD_VER_ENTRY(3, iwl_tlc_update_notif)) +CMD_VERSIONS(mu_mimo_grp_notif, + CMD_VER_ENTRY(1, iwl_mu_group_mgmt_notif)) +CMD_VERSIONS(channel_switch_start_notif, + CMD_VER_ENTRY(3, iwl_channel_switch_start_notif)) +CMD_VERSIONS(channel_switch_error_notif, + CMD_VER_ENTRY(2, iwl_channel_switch_error_notif)) +CMD_VERSIONS(ct_kill_notif, + CMD_VER_ENTRY(2, ct_kill_notif)) +CMD_VERSIONS(temp_notif, + CMD_VER_ENTRY(2, iwl_dts_measurement_notif)) +CMD_VERSIONS(stored_beacon_notif, + CMD_VER_ENTRY(4, iwl_stored_beacon_notif)) +CMD_VERSIONS(roc_notif, + CMD_VER_ENTRY(1, iwl_roc_notif)) +CMD_VERSIONS(probe_resp_data_notif, + CMD_VER_ENTRY(1, iwl_probe_resp_data_notif)) +CMD_VERSIONS(datapath_monitor_notif, + CMD_VER_ENTRY(1, iwl_datapath_monitor_notif)) +CMD_VERSIONS(stats_oper_notif, + CMD_VER_ENTRY(3, iwl_system_statistics_notif_oper)) +CMD_VERSIONS(stats_oper_part1_notif, + CMD_VER_ENTRY(4, iwl_system_statistics_part1_notif_oper)) +CMD_VERSIONS(bt_coex_notif, + CMD_VER_ENTRY(1, iwl_bt_coex_profile_notif)) +CMD_VERSIONS(beacon_notification, + CMD_VER_ENTRY(6, iwl_extended_beacon_notif)) +CMD_VERSIONS(emlsr_mode_notif, + CMD_VER_ENTRY(1, iwl_esr_mode_notif)) +CMD_VERSIONS(emlsr_trans_fail_notif, + CMD_VER_ENTRY(1, iwl_esr_trans_fail_notif)) +CMD_VERSIONS(uapsd_misbehaving_ap_notif, + CMD_VER_ENTRY(1, iwl_uapsd_misbehaving_ap_notif)) +CMD_VERSIONS(time_msmt_notif, + CMD_VER_ENTRY(1, iwl_time_msmt_notify)) +CMD_VERSIONS(time_sync_confirm_notif, + CMD_VER_ENTRY(1, iwl_time_msmt_cfm_notify)) +CMD_VERSIONS(omi_status_notif, + CMD_VER_ENTRY(1, iwl_omi_send_status_notif)) +CMD_VERSIONS(ftm_resp_notif, CMD_VER_ENTRY(9, iwl_tof_range_rsp_ntfy)) + +DEFINE_SIMPLE_CANCELLATION(session_prot, iwl_session_prot_notif, mac_link_id) +DEFINE_SIMPLE_CANCELLATION(tlc, iwl_tlc_update_notif, sta_id) +DEFINE_SIMPLE_CANCELLATION(channel_switch_start, + iwl_channel_switch_start_notif, link_id) +DEFINE_SIMPLE_CANCELLATION(channel_switch_error, + iwl_channel_switch_error_notif, link_id) +DEFINE_SIMPLE_CANCELLATION(datapath_monitor, iwl_datapath_monitor_notif, + link_id) +DEFINE_SIMPLE_CANCELLATION(roc, iwl_roc_notif, activity) +DEFINE_SIMPLE_CANCELLATION(scan_complete, iwl_umac_scan_complete, uid) +DEFINE_SIMPLE_CANCELLATION(probe_resp_data, iwl_probe_resp_data_notif, + mac_id) +DEFINE_SIMPLE_CANCELLATION(uapsd_misbehaving_ap, iwl_uapsd_misbehaving_ap_notif, + mac_id) +#define iwl_mld_cancel_omi_status_notif iwl_mld_always_cancel +DEFINE_SIMPLE_CANCELLATION(ftm_resp, iwl_tof_range_rsp_ntfy, request_id) + +/** + * DOC: Handlers for fw notifications + * + * Here are listed the notifications IDs (including the group ID), the handler + * of the notification and how it should be called: + * + * - RX_HANDLER_SYNC: will be called as part of the Rx path + * - RX_HANDLER_ASYNC: will be handled in a working with the wiphy_lock held + * + * This means that if the firmware sends two notifications A and B in that + * order and notification A is RX_HANDLER_ASYNC and notification is + * RX_HANDLER_SYNC, the handler of B will likely be called before the handler + * of A. + * + * This list should be in order of frequency for performance purposes. + * The handler can be one from two contexts, see &iwl_rx_handler_context + * + * A handler can declare that it relies on a specific object in which case it + * can be cancelled in case the object is deleted. In order to use this + * mechanism, a cancellation function is needed. The cancellation function must + * receive an object id (the index of that object in the firmware) and a + * notification payload. It'll return true if that specific notification should + * be cancelled upon the obliteration of the specific instance of the object. + * + * DEFINE_SIMPLE_CANCELLATION allows to easily create a cancellation function + * that wills simply return true if a given object id matches the object id in + * the firmware notification. + */ + +VISIBLE_IF_IWLWIFI_KUNIT +const struct iwl_rx_handler iwl_mld_rx_handlers[] = { + RX_HANDLER_NO_OBJECT(LEGACY_GROUP, TX_CMD, tx_resp_notif, + RX_HANDLER_SYNC) + RX_HANDLER_NO_OBJECT(LEGACY_GROUP, BA_NOTIF, compressed_ba_notif, + RX_HANDLER_SYNC) + RX_HANDLER_OF_SCAN(LEGACY_GROUP, SCAN_COMPLETE_UMAC, + scan_complete_notif) + RX_HANDLER_NO_OBJECT(LEGACY_GROUP, SCAN_ITERATION_COMPLETE_UMAC, + scan_iter_complete_notif, + RX_HANDLER_SYNC) + RX_HANDLER_NO_VAL(LEGACY_GROUP, MATCH_FOUND_NOTIFICATION, + match_found_notif, RX_HANDLER_SYNC) + + RX_HANDLER_NO_OBJECT(STATISTICS_GROUP, STATISTICS_OPER_NOTIF, + stats_oper_notif, RX_HANDLER_ASYNC) + RX_HANDLER_NO_OBJECT(STATISTICS_GROUP, STATISTICS_OPER_PART1_NOTIF, + stats_oper_part1_notif, RX_HANDLER_ASYNC) + + RX_HANDLER_NO_OBJECT(LEGACY_GROUP, MFUART_LOAD_NOTIFICATION, + mfuart_notif, RX_HANDLER_SYNC) + + RX_HANDLER_NO_OBJECT(PHY_OPS_GROUP, DTS_MEASUREMENT_NOTIF_WIDE, + temp_notif, RX_HANDLER_ASYNC) + RX_HANDLER_OF_LINK(MAC_CONF_GROUP, SESSION_PROTECTION_NOTIF, + session_prot_notif) + RX_HANDLER_OF_LINK(MAC_CONF_GROUP, MISSED_BEACONS_NOTIF, + missed_beacon_notif) + RX_HANDLER_OF_STA(DATA_PATH_GROUP, TLC_MNG_UPDATE_NOTIF, tlc_notif) + RX_HANDLER_OF_LINK(MAC_CONF_GROUP, CHANNEL_SWITCH_START_NOTIF, + channel_switch_start_notif) + RX_HANDLER_OF_LINK(MAC_CONF_GROUP, CHANNEL_SWITCH_ERROR_NOTIF, + channel_switch_error_notif) + RX_HANDLER_OF_ROC(MAC_CONF_GROUP, ROC_NOTIF, roc_notif) + RX_HANDLER_NO_OBJECT(DATA_PATH_GROUP, MU_GROUP_MGMT_NOTIF, + mu_mimo_grp_notif, RX_HANDLER_SYNC) + RX_HANDLER_NO_OBJECT(PROT_OFFLOAD_GROUP, STORED_BEACON_NTF, + stored_beacon_notif, RX_HANDLER_SYNC) + RX_HANDLER_OF_VIF(MAC_CONF_GROUP, PROBE_RESPONSE_DATA_NOTIF, + probe_resp_data_notif) + RX_HANDLER_NO_OBJECT(PHY_OPS_GROUP, CT_KILL_NOTIFICATION, + ct_kill_notif, RX_HANDLER_ASYNC) + RX_HANDLER_OF_LINK(DATA_PATH_GROUP, MONITOR_NOTIF, + datapath_monitor_notif) + RX_HANDLER_NO_OBJECT(LEGACY_GROUP, MCC_CHUB_UPDATE_CMD, update_mcc, + RX_HANDLER_ASYNC) + RX_HANDLER_NO_OBJECT(BT_COEX_GROUP, PROFILE_NOTIF, + bt_coex_notif, RX_HANDLER_ASYNC) + RX_HANDLER_NO_OBJECT(LEGACY_GROUP, BEACON_NOTIFICATION, + beacon_notification, RX_HANDLER_ASYNC) + RX_HANDLER_NO_OBJECT(DATA_PATH_GROUP, ESR_MODE_NOTIF, + emlsr_mode_notif, RX_HANDLER_ASYNC) + RX_HANDLER_NO_OBJECT(MAC_CONF_GROUP, EMLSR_TRANS_FAIL_NOTIF, + emlsr_trans_fail_notif, RX_HANDLER_ASYNC) + RX_HANDLER_OF_VIF(LEGACY_GROUP, PSM_UAPSD_AP_MISBEHAVING_NOTIFICATION, + uapsd_misbehaving_ap_notif) + RX_HANDLER_NO_OBJECT(LEGACY_GROUP, + WNM_80211V_TIMING_MEASUREMENT_NOTIFICATION, + time_msmt_notif, RX_HANDLER_SYNC) + RX_HANDLER_NO_OBJECT(LEGACY_GROUP, + WNM_80211V_TIMING_MEASUREMENT_CONFIRM_NOTIFICATION, + time_sync_confirm_notif, RX_HANDLER_ASYNC) + RX_HANDLER_OF_LINK(DATA_PATH_GROUP, OMI_SEND_STATUS_NOTIF, + omi_status_notif) + RX_HANDLER_OF_FTM_REQ(LOCATION_GROUP, TOF_RANGE_RESPONSE_NOTIF, + ftm_resp_notif) +}; +EXPORT_SYMBOL_IF_IWLWIFI_KUNIT(iwl_mld_rx_handlers); + +#if IS_ENABLED(CONFIG_IWLWIFI_KUNIT_TESTS) +const unsigned int iwl_mld_rx_handlers_num = ARRAY_SIZE(iwl_mld_rx_handlers); +EXPORT_SYMBOL_IF_IWLWIFI_KUNIT(iwl_mld_rx_handlers_num); +#endif + +static bool +iwl_mld_notif_is_valid(struct iwl_mld *mld, struct iwl_rx_packet *pkt, + const struct iwl_rx_handler *handler) +{ + unsigned int size = iwl_rx_packet_payload_len(pkt); + size_t notif_ver; + + /* If n_sizes == 0, it indicates that a validation function may be used + * or that no validation is required. + */ + if (!handler->n_sizes) { + if (handler->val_fn) + return handler->val_fn(mld, pkt); + return true; + } + + notif_ver = iwl_fw_lookup_notif_ver(mld->fw, + iwl_cmd_groupid(handler->cmd_id), + iwl_cmd_opcode(handler->cmd_id), + IWL_FW_CMD_VER_UNKNOWN); + + for (int i = 0; i < handler->n_sizes; i++) { + if (handler->sizes[i].ver != notif_ver) + continue; + + if (IWL_FW_CHECK(mld, size < handler->sizes[i].size, + "unexpected notification 0x%04x size %d, need %d\n", + handler->cmd_id, size, handler->sizes[i].size)) + return false; + return true; + } + + IWL_FW_CHECK_FAILED(mld, + "notif 0x%04x ver %zu missing expected size, use version %u size\n", + handler->cmd_id, notif_ver, + handler->sizes[handler->n_sizes - 1].ver); + + return size < handler->sizes[handler->n_sizes - 1].size; +} + +struct iwl_async_handler_entry { + struct list_head list; + struct iwl_rx_cmd_buffer rxb; + const struct iwl_rx_handler *rx_h; +}; + +static void +iwl_mld_log_async_handler_op(struct iwl_mld *mld, const char *op, + struct iwl_rx_cmd_buffer *rxb) +{ + struct iwl_rx_packet *pkt = rxb_addr(rxb); + + IWL_DEBUG_HC(mld, + "%s async handler for notif %s (%.2x.%2x, seq 0x%x)\n", + op, iwl_get_cmd_string(mld->trans, + WIDE_ID(pkt->hdr.group_id, pkt->hdr.cmd)), + pkt->hdr.group_id, pkt->hdr.cmd, + le16_to_cpu(pkt->hdr.sequence)); +} + +static void iwl_mld_rx_notif(struct iwl_mld *mld, + struct iwl_rx_cmd_buffer *rxb, + struct iwl_rx_packet *pkt) +{ + for (int i = 0; i < ARRAY_SIZE(iwl_mld_rx_handlers); i++) { + const struct iwl_rx_handler *rx_h = &iwl_mld_rx_handlers[i]; + struct iwl_async_handler_entry *entry; + + if (rx_h->cmd_id != WIDE_ID(pkt->hdr.group_id, pkt->hdr.cmd)) + continue; + + if (!iwl_mld_notif_is_valid(mld, pkt, rx_h)) + return; + + if (rx_h->context == RX_HANDLER_SYNC) { + rx_h->fn(mld, pkt); + break; + } + + entry = kzalloc(sizeof(*entry), GFP_ATOMIC); + /* we can't do much... */ + if (!entry) + return; + + /* Set the async handler entry */ + entry->rxb._page = rxb_steal_page(rxb); + entry->rxb._offset = rxb->_offset; + entry->rxb._rx_page_order = rxb->_rx_page_order; + + entry->rx_h = rx_h; + + /* Add it to the list and queue the work */ + spin_lock(&mld->async_handlers_lock); + list_add_tail(&entry->list, &mld->async_handlers_list); + spin_unlock(&mld->async_handlers_lock); + + wiphy_work_queue(mld->hw->wiphy, + &mld->async_handlers_wk); + + iwl_mld_log_async_handler_op(mld, "Queued", rxb); + break; + } + + iwl_notification_wait_notify(&mld->notif_wait, pkt); +} + +void iwl_mld_rx(struct iwl_op_mode *op_mode, struct napi_struct *napi, + struct iwl_rx_cmd_buffer *rxb) +{ + struct iwl_rx_packet *pkt = rxb_addr(rxb); + struct iwl_mld *mld = IWL_OP_MODE_GET_MLD(op_mode); + u16 cmd_id = WIDE_ID(pkt->hdr.group_id, pkt->hdr.cmd); + + if (likely(cmd_id == WIDE_ID(LEGACY_GROUP, REPLY_RX_MPDU_CMD))) + iwl_mld_rx_mpdu(mld, napi, rxb, 0); + else if (cmd_id == WIDE_ID(LEGACY_GROUP, FRAME_RELEASE)) + iwl_mld_handle_frame_release_notif(mld, napi, pkt, 0); + else if (cmd_id == WIDE_ID(LEGACY_GROUP, BAR_FRAME_RELEASE)) + iwl_mld_handle_bar_frame_release_notif(mld, napi, pkt, 0); + else if (unlikely(cmd_id == WIDE_ID(DATA_PATH_GROUP, + RX_QUEUES_NOTIFICATION))) + iwl_mld_handle_rx_queues_sync_notif(mld, napi, pkt, 0); + else if (cmd_id == WIDE_ID(DATA_PATH_GROUP, RX_NO_DATA_NOTIF)) + iwl_mld_rx_monitor_no_data(mld, napi, pkt, 0); + else + iwl_mld_rx_notif(mld, rxb, pkt); +} + +void iwl_mld_rx_rss(struct iwl_op_mode *op_mode, struct napi_struct *napi, + struct iwl_rx_cmd_buffer *rxb, unsigned int queue) +{ + struct iwl_rx_packet *pkt = rxb_addr(rxb); + struct iwl_mld *mld = IWL_OP_MODE_GET_MLD(op_mode); + u16 cmd_id = WIDE_ID(pkt->hdr.group_id, pkt->hdr.cmd); + + if (unlikely(queue >= mld->trans->num_rx_queues)) + return; + + if (likely(cmd_id == WIDE_ID(LEGACY_GROUP, REPLY_RX_MPDU_CMD))) + iwl_mld_rx_mpdu(mld, napi, rxb, queue); + else if (unlikely(cmd_id == WIDE_ID(DATA_PATH_GROUP, + RX_QUEUES_NOTIFICATION))) + iwl_mld_handle_rx_queues_sync_notif(mld, napi, pkt, queue); + else if (unlikely(cmd_id == WIDE_ID(LEGACY_GROUP, FRAME_RELEASE))) + iwl_mld_handle_frame_release_notif(mld, napi, pkt, queue); +} + +void iwl_mld_delete_handlers(struct iwl_mld *mld, const u16 *cmds, int n_cmds) +{ + struct iwl_async_handler_entry *entry, *tmp; + + spin_lock_bh(&mld->async_handlers_lock); + list_for_each_entry_safe(entry, tmp, &mld->async_handlers_list, list) { + bool match = false; + + for (int i = 0; i < n_cmds; i++) { + if (entry->rx_h->cmd_id == cmds[i]) { + match = true; + break; + } + } + + if (!match) + continue; + + iwl_mld_log_async_handler_op(mld, "Delete", &entry->rxb); + iwl_free_rxb(&entry->rxb); + list_del(&entry->list); + kfree(entry); + } + spin_unlock_bh(&mld->async_handlers_lock); +} + +void iwl_mld_async_handlers_wk(struct wiphy *wiphy, struct wiphy_work *wk) +{ + struct iwl_mld *mld = + container_of(wk, struct iwl_mld, async_handlers_wk); + struct iwl_async_handler_entry *entry, *tmp; + LIST_HEAD(local_list); + + /* Sync with Rx path with a lock. Remove all the entries from this + * list, add them to a local one (lock free), and then handle them. + */ + spin_lock_bh(&mld->async_handlers_lock); + list_splice_init(&mld->async_handlers_list, &local_list); + spin_unlock_bh(&mld->async_handlers_lock); + + list_for_each_entry_safe(entry, tmp, &local_list, list) { + iwl_mld_log_async_handler_op(mld, "Handle", &entry->rxb); + entry->rx_h->fn(mld, rxb_addr(&entry->rxb)); + iwl_free_rxb(&entry->rxb); + list_del(&entry->list); + kfree(entry); + } +} + +void iwl_mld_purge_async_handlers_list(struct iwl_mld *mld) +{ + struct iwl_async_handler_entry *entry, *tmp; + + spin_lock_bh(&mld->async_handlers_lock); + list_for_each_entry_safe(entry, tmp, &mld->async_handlers_list, list) { + iwl_mld_log_async_handler_op(mld, "Purged", &entry->rxb); + iwl_free_rxb(&entry->rxb); + list_del(&entry->list); + kfree(entry); + } + spin_unlock_bh(&mld->async_handlers_lock); +} + +void iwl_mld_cancel_notifications_of_object(struct iwl_mld *mld, + enum iwl_mld_object_type obj_type, + u32 obj_id) +{ + struct iwl_async_handler_entry *entry, *tmp; + LIST_HEAD(cancel_list); + + lockdep_assert_wiphy(mld->wiphy); + + if (WARN_ON(obj_type == IWL_MLD_OBJECT_TYPE_NONE)) + return; + + /* Sync with RX path and remove matching entries from the async list */ + spin_lock_bh(&mld->async_handlers_lock); + list_for_each_entry_safe(entry, tmp, &mld->async_handlers_list, list) { + const struct iwl_rx_handler *rx_h = entry->rx_h; + + if (rx_h->obj_type != obj_type || WARN_ON(!rx_h->cancel)) + continue; + + if (rx_h->cancel(mld, rxb_addr(&entry->rxb), obj_id)) { + iwl_mld_log_async_handler_op(mld, "Cancel", &entry->rxb); + list_del(&entry->list); + list_add_tail(&entry->list, &cancel_list); + } + } + + spin_unlock_bh(&mld->async_handlers_lock); + + /* Free the matching entries outside of the spinlock */ + list_for_each_entry_safe(entry, tmp, &cancel_list, list) { + iwl_free_rxb(&entry->rxb); + list_del(&entry->list); + kfree(entry); + } +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/notif.h b/drivers/net/wireless/intel/iwlwifi/mld/notif.h new file mode 100644 index 0000000000000..2eaa1d4e138ee --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/notif.h @@ -0,0 +1,35 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024-2025 Intel Corporation + */ +#ifndef __iwl_mld_notif_h__ +#define __iwl_mld_notif_h__ + +struct iwl_mld; + +void iwl_mld_rx(struct iwl_op_mode *op_mode, struct napi_struct *napi, + struct iwl_rx_cmd_buffer *rxb); + +void iwl_mld_rx_rss(struct iwl_op_mode *op_mode, struct napi_struct *napi, + struct iwl_rx_cmd_buffer *rxb, unsigned int queue); + +void iwl_mld_async_handlers_wk(struct wiphy *wiphy, struct wiphy_work *wk); + +void iwl_mld_purge_async_handlers_list(struct iwl_mld *mld); + +enum iwl_mld_object_type { + IWL_MLD_OBJECT_TYPE_NONE, + IWL_MLD_OBJECT_TYPE_LINK, + IWL_MLD_OBJECT_TYPE_STA, + IWL_MLD_OBJECT_TYPE_VIF, + IWL_MLD_OBJECT_TYPE_ROC, + IWL_MLD_OBJECT_TYPE_SCAN, + IWL_MLD_OBJECT_TYPE_FTM_REQ, +}; + +void iwl_mld_cancel_notifications_of_object(struct iwl_mld *mld, + enum iwl_mld_object_type obj_type, + u32 obj_id); +void iwl_mld_delete_handlers(struct iwl_mld *mld, const u16 *cmds, int n_cmds); + +#endif /* __iwl_mld_notif_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/phy.c b/drivers/net/wireless/intel/iwlwifi/mld/phy.c new file mode 100644 index 0000000000000..c38f101628de3 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/phy.c @@ -0,0 +1,116 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024-2025 Intel Corporation + */ +#include + +#include "phy.h" +#include "hcmd.h" +#include "fw/api/phy-ctxt.h" + +int iwl_mld_allocate_fw_phy_id(struct iwl_mld *mld) +{ + int id; + unsigned long used = mld->used_phy_ids; + + for_each_clear_bit(id, &used, NUM_PHY_CTX) { + mld->used_phy_ids |= BIT(id); + return id; + } + + return -ENOSPC; +} +EXPORT_SYMBOL_IF_IWLWIFI_KUNIT(iwl_mld_allocate_fw_phy_id); + +struct cfg80211_chan_def * +iwl_mld_get_chandef_from_chanctx(struct ieee80211_chanctx_conf *ctx) +{ + bool use_def = cfg80211_channel_is_psc(ctx->def.chan) || + (ctx->def.chan->band == NL80211_BAND_6GHZ && + ctx->def.width >= NL80211_CHAN_WIDTH_80); + + return use_def ? &ctx->def : &ctx->min_def; +} +EXPORT_SYMBOL_IF_IWLWIFI_KUNIT(iwl_mld_get_chandef_from_chanctx); + +static u8 +iwl_mld_nl80211_width_to_fw(enum nl80211_chan_width width) +{ + switch (width) { + case NL80211_CHAN_WIDTH_20_NOHT: + case NL80211_CHAN_WIDTH_20: + return IWL_PHY_CHANNEL_MODE20; + case NL80211_CHAN_WIDTH_40: + return IWL_PHY_CHANNEL_MODE40; + case NL80211_CHAN_WIDTH_80: + return IWL_PHY_CHANNEL_MODE80; + case NL80211_CHAN_WIDTH_160: + return IWL_PHY_CHANNEL_MODE160; + case NL80211_CHAN_WIDTH_320: + return IWL_PHY_CHANNEL_MODE320; + default: + WARN(1, "Invalid channel width=%u", width); + return IWL_PHY_CHANNEL_MODE20; + } +} + +/* Maps the driver specific control channel position (relative to the center + * freq) definitions to the fw values + */ +u8 iwl_mld_get_fw_ctrl_pos(const struct cfg80211_chan_def *chandef) +{ + int offs = chandef->chan->center_freq - chandef->center_freq1; + int abs_offs = abs(offs); + u8 ret; + + if (offs == 0) { + /* The FW is expected to check the control channel position only + * when in HT/VHT and the channel width is not 20MHz. Return + * this value as the default one. + */ + return 0; + } + + /* this results in a value 0-7, i.e. fitting into 0b0111 */ + ret = (abs_offs - 10) / 20; + /* But we need the value to be in 0b1011 because 0b0100 is + * IWL_PHY_CTRL_POS_ABOVE, so shift bit 2 up to land in + * IWL_PHY_CTRL_POS_OFFS_EXT (0b1000) + */ + ret = (ret & IWL_PHY_CTRL_POS_OFFS_MSK) | + ((ret & BIT(2)) << 1); + /* and add the above bit */ + ret |= (offs > 0) * IWL_PHY_CTRL_POS_ABOVE; + + return ret; +} + +int iwl_mld_phy_fw_action(struct iwl_mld *mld, + struct ieee80211_chanctx_conf *ctx, u32 action) +{ + struct iwl_mld_phy *phy = iwl_mld_phy_from_mac80211(ctx); + struct cfg80211_chan_def *chandef = &phy->chandef; + struct iwl_phy_context_cmd cmd = { + .id_and_color = cpu_to_le32(phy->fw_id), + .action = cpu_to_le32(action), + .puncture_mask = cpu_to_le16(chandef->punctured), + /* Channel info */ + .ci.channel = cpu_to_le32(chandef->chan->hw_value), + .ci.band = iwl_mld_nl80211_band_to_fw(chandef->chan->band), + .ci.width = iwl_mld_nl80211_width_to_fw(chandef->width), + .ci.ctrl_pos = iwl_mld_get_fw_ctrl_pos(chandef), + }; + int ret; + + if (ctx->ap.chan) { + cmd.sbb_bandwidth = + iwl_mld_nl80211_width_to_fw(ctx->ap.width); + cmd.sbb_ctrl_channel_loc = iwl_mld_get_fw_ctrl_pos(&ctx->ap); + } + + ret = iwl_mld_send_cmd_pdu(mld, PHY_CONTEXT_CMD, &cmd); + if (ret) + IWL_ERR(mld, "Failed to send PHY_CONTEXT_CMD ret = %d\n", ret); + + return ret; +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/phy.h b/drivers/net/wireless/intel/iwlwifi/mld/phy.h new file mode 100644 index 0000000000000..3dfb8ca994e2e --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/phy.h @@ -0,0 +1,49 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024-2025 Intel Corporation + */ +#ifndef __iwl_mld_phy_h__ +#define __iwl_mld_phy_h__ + +#include "mld.h" + +/** + * struct iwl_mld_phy - PHY configuration parameters + * + * @fw_id: fw id of the phy. + * @chandef: the last chandef that mac80211 configured the driver + * with. Used to detect a no-op when the chanctx changes. + * @channel_load_by_us: channel load on this channel caused by + * the NIC itself, as indicated by firmware + */ +struct iwl_mld_phy { + /* Add here fields that need clean up on hw restart */ + struct_group(zeroed_on_hw_restart, + u8 fw_id; + struct cfg80211_chan_def chandef; + ); + /* And here fields that survive a hw restart */ + u32 channel_load_by_us; +}; + +static inline struct iwl_mld_phy * +iwl_mld_phy_from_mac80211(struct ieee80211_chanctx_conf *channel) +{ + return (void *)channel->drv_priv; +} + +/* Cleanup function for struct iwl_mld_phy, will be called in restart */ +static inline void +iwl_mld_cleanup_phy(struct iwl_mld *mld, struct iwl_mld_phy *phy) +{ + CLEANUP_STRUCT(phy); +} + +int iwl_mld_allocate_fw_phy_id(struct iwl_mld *mld); +int iwl_mld_phy_fw_action(struct iwl_mld *mld, + struct ieee80211_chanctx_conf *ctx, u32 action); +struct cfg80211_chan_def * +iwl_mld_get_chandef_from_chanctx(struct ieee80211_chanctx_conf *ctx); +u8 iwl_mld_get_fw_ctrl_pos(const struct cfg80211_chan_def *chandef); + +#endif /* __iwl_mld_phy_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/power.c b/drivers/net/wireless/intel/iwlwifi/mld/power.c new file mode 100644 index 0000000000000..ed7c0319f239b --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/power.c @@ -0,0 +1,395 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024-2025 Intel Corporation + */ +#include + +#include "mld.h" +#include "hcmd.h" +#include "power.h" +#include "iface.h" +#include "link.h" +#include "constants.h" + +static void iwl_mld_vif_ps_iterator(void *data, u8 *mac, + struct ieee80211_vif *vif) +{ + bool *ps_enable = (bool *)data; + + if (vif->type != NL80211_IFTYPE_STATION) + return; + + *ps_enable &= vif->cfg.ps; +} + +int iwl_mld_update_device_power(struct iwl_mld *mld, bool d3) +{ + struct iwl_device_power_cmd cmd = {}; + bool enable_ps = false; + + if (iwlmld_mod_params.power_scheme != IWL_POWER_SCHEME_CAM) { + enable_ps = true; + + /* Disable power save if any STA interface has + * power save turned off + */ + ieee80211_iterate_active_interfaces_mtx(mld->hw, + IEEE80211_IFACE_ITER_NORMAL, + iwl_mld_vif_ps_iterator, + &enable_ps); + } + + if (enable_ps) + cmd.flags |= + cpu_to_le16(DEVICE_POWER_FLAGS_POWER_SAVE_ENA_MSK); + + if (d3) + cmd.flags |= + cpu_to_le16(DEVICE_POWER_FLAGS_NO_SLEEP_TILL_D3_MSK); + + IWL_DEBUG_POWER(mld, + "Sending device power command with flags = 0x%X\n", + cmd.flags); + + return iwl_mld_send_cmd_pdu(mld, POWER_TABLE_CMD, &cmd); +} + +int iwl_mld_enable_beacon_filter(struct iwl_mld *mld, + const struct ieee80211_bss_conf *link_conf, + bool d3) +{ + struct iwl_beacon_filter_cmd cmd = { + IWL_BF_CMD_CONFIG_DEFAULTS, + .bf_enable_beacon_filter = cpu_to_le32(1), + .ba_enable_beacon_abort = cpu_to_le32(1), + }; + + if (ieee80211_vif_type_p2p(link_conf->vif) != NL80211_IFTYPE_STATION) + return 0; + +#ifdef CONFIG_IWLWIFI_DEBUGFS + if (iwl_mld_vif_from_mac80211(link_conf->vif)->disable_bf) + return 0; +#endif + + if (link_conf->cqm_rssi_thold) { + cmd.bf_energy_delta = + cpu_to_le32(link_conf->cqm_rssi_hyst); + /* fw uses an absolute value for this */ + cmd.bf_roaming_state = + cpu_to_le32(-link_conf->cqm_rssi_thold); + } + + if (d3) + cmd.ba_escape_timer = cpu_to_le32(IWL_BA_ESCAPE_TIMER_D3); + + return iwl_mld_send_cmd_pdu(mld, REPLY_BEACON_FILTERING_CMD, + &cmd); +} + +int iwl_mld_disable_beacon_filter(struct iwl_mld *mld, + struct ieee80211_vif *vif) +{ + struct iwl_beacon_filter_cmd cmd = {}; + + if (ieee80211_vif_type_p2p(vif) != NL80211_IFTYPE_STATION) + return 0; + + return iwl_mld_send_cmd_pdu(mld, REPLY_BEACON_FILTERING_CMD, + &cmd); +} + +static bool iwl_mld_power_is_radar(struct iwl_mld *mld, + const struct ieee80211_bss_conf *link_conf) +{ + const struct ieee80211_chanctx_conf *chanctx_conf; + + chanctx_conf = wiphy_dereference(mld->wiphy, link_conf->chanctx_conf); + + if (WARN_ON(!chanctx_conf)) + return false; + + return chanctx_conf->def.chan->flags & IEEE80211_CHAN_RADAR; +} + +static void iwl_mld_power_configure_uapsd(struct iwl_mld *mld, + struct iwl_mld_link *link, + struct iwl_mac_power_cmd *cmd, + bool ps_poll) +{ + bool tid_found = false; + + cmd->rx_data_timeout_uapsd = + cpu_to_le32(IWL_MLD_UAPSD_RX_DATA_TIMEOUT); + cmd->tx_data_timeout_uapsd = + cpu_to_le32(IWL_MLD_UAPSD_TX_DATA_TIMEOUT); + + /* set advanced pm flag with no uapsd ACs to enable ps-poll */ + if (ps_poll) { + cmd->flags |= cpu_to_le16(POWER_FLAGS_ADVANCE_PM_ENA_MSK); + return; + } + + for (enum ieee80211_ac_numbers ac = IEEE80211_AC_VO; + ac <= IEEE80211_AC_BK; + ac++) { + if (!link->queue_params[ac].uapsd) + continue; + + cmd->flags |= + cpu_to_le16(POWER_FLAGS_ADVANCE_PM_ENA_MSK | + POWER_FLAGS_UAPSD_MISBEHAVING_ENA_MSK); + + cmd->uapsd_ac_flags |= BIT(ac); + + /* QNDP TID - the highest TID with no admission control */ + if (!tid_found && !link->queue_params[ac].acm) { + tid_found = true; + switch (ac) { + case IEEE80211_AC_VO: + cmd->qndp_tid = 6; + break; + case IEEE80211_AC_VI: + cmd->qndp_tid = 5; + break; + case IEEE80211_AC_BE: + cmd->qndp_tid = 0; + break; + case IEEE80211_AC_BK: + cmd->qndp_tid = 1; + break; + } + } + } + + if (cmd->uapsd_ac_flags == (BIT(IEEE80211_AC_VO) | + BIT(IEEE80211_AC_VI) | + BIT(IEEE80211_AC_BE) | + BIT(IEEE80211_AC_BK))) { + cmd->flags |= cpu_to_le16(POWER_FLAGS_SNOOZE_ENA_MSK); + cmd->snooze_interval = cpu_to_le16(IWL_MLD_PS_SNOOZE_INTERVAL); + cmd->snooze_window = cpu_to_le16(IWL_MLD_PS_SNOOZE_WINDOW); + } + + cmd->uapsd_max_sp = mld->hw->uapsd_max_sp_len; +} + +static void +iwl_mld_power_config_skip_dtim(struct iwl_mld *mld, + const struct ieee80211_bss_conf *link_conf, + struct iwl_mac_power_cmd *cmd) +{ + unsigned int dtimper_tu; + unsigned int dtimper; + unsigned int skip; + + dtimper = link_conf->dtim_period ?: 1; + dtimper_tu = dtimper * link_conf->beacon_int; + + if (dtimper >= 10 || iwl_mld_power_is_radar(mld, link_conf)) + return; + + if (WARN_ON(!dtimper_tu)) + return; + + /* configure skip over dtim up to 900 TU DTIM interval */ + skip = max_t(int, 1, 900 / dtimper_tu); + + cmd->skip_dtim_periods = skip; + cmd->flags |= cpu_to_le16(POWER_FLAGS_SKIP_OVER_DTIM_MSK); +} + +#define POWER_KEEP_ALIVE_PERIOD_SEC 25 +static void iwl_mld_power_build_cmd(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct iwl_mac_power_cmd *cmd, + bool d3) +{ + int dtimper, bi; + int keep_alive; + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct ieee80211_bss_conf *link_conf = &vif->bss_conf; + struct iwl_mld_link *link = &mld_vif->deflink; + bool ps_poll = false; + + cmd->id_and_color = cpu_to_le32(mld_vif->fw_id); + + if (ieee80211_vif_is_mld(vif)) { + int link_id; + + if (WARN_ON(!vif->active_links)) + return; + + /* The firmware consumes one single configuration for the vif + * and can't differentiate between links, just pick the lowest + * link_id's configuration and use that. + */ + link_id = __ffs(vif->active_links); + link_conf = link_conf_dereference_check(vif, link_id); + link = iwl_mld_link_dereference_check(mld_vif, link_id); + + if (WARN_ON(!link_conf || !link)) + return; + } + dtimper = link_conf->dtim_period; + bi = link_conf->beacon_int; + + /* Regardless of power management state the driver must set + * keep alive period. FW will use it for sending keep alive NDPs + * immediately after association. Check that keep alive period + * is at least 3 * DTIM + */ + keep_alive = DIV_ROUND_UP(ieee80211_tu_to_usec(3 * dtimper * bi), + USEC_PER_SEC); + keep_alive = max(keep_alive, POWER_KEEP_ALIVE_PERIOD_SEC); + cmd->keep_alive_seconds = cpu_to_le16(keep_alive); + + if (iwlmld_mod_params.power_scheme != IWL_POWER_SCHEME_CAM) + cmd->flags |= cpu_to_le16(POWER_FLAGS_POWER_SAVE_ENA_MSK); + + if (!vif->cfg.ps || iwl_mld_tdls_sta_count(mld) > 0) + return; + + cmd->flags |= cpu_to_le16(POWER_FLAGS_POWER_MANAGEMENT_ENA_MSK); + + /* firmware supports LPRX for beacons at rate 1 Mbps or 6 Mbps only */ + if (link_conf->beacon_rate && + (link_conf->beacon_rate->bitrate == 10 || + link_conf->beacon_rate->bitrate == 60)) { + cmd->flags |= cpu_to_le16(POWER_FLAGS_LPRX_ENA_MSK); + cmd->lprx_rssi_threshold = POWER_LPRX_RSSI_THRESHOLD; + } + + if (d3) { + iwl_mld_power_config_skip_dtim(mld, link_conf, cmd); + cmd->rx_data_timeout = + cpu_to_le32(IWL_MLD_WOWLAN_PS_RX_DATA_TIMEOUT); + cmd->tx_data_timeout = + cpu_to_le32(IWL_MLD_WOWLAN_PS_TX_DATA_TIMEOUT); + } else if (iwl_mld_vif_low_latency(mld_vif) && vif->p2p) { + cmd->tx_data_timeout = + cpu_to_le32(IWL_MLD_SHORT_PS_TX_DATA_TIMEOUT); + cmd->rx_data_timeout = + cpu_to_le32(IWL_MLD_SHORT_PS_RX_DATA_TIMEOUT); + } else { + cmd->rx_data_timeout = + cpu_to_le32(IWL_MLD_DEFAULT_PS_RX_DATA_TIMEOUT); + cmd->tx_data_timeout = + cpu_to_le32(IWL_MLD_DEFAULT_PS_TX_DATA_TIMEOUT); + } + + /* uAPSD is only enabled for specific certifications. For those cases, + * mac80211 will allow uAPSD. Always call iwl_mld_power_configure_uapsd + * which will look at what mac80211 is saying. + */ +#ifdef CONFIG_IWLWIFI_DEBUGFS + ps_poll = mld_vif->use_ps_poll; +#endif + iwl_mld_power_configure_uapsd(mld, link, cmd, ps_poll); +} + +int iwl_mld_update_mac_power(struct iwl_mld *mld, struct ieee80211_vif *vif, + bool d3) +{ + struct iwl_mac_power_cmd cmd = {}; + + iwl_mld_power_build_cmd(mld, vif, &cmd, d3); + + return iwl_mld_send_cmd_pdu(mld, MAC_PM_POWER_TABLE, &cmd); +} + +static void +iwl_mld_tpe_sta_cmd_data(struct iwl_txpower_constraints_cmd *cmd, + const struct ieee80211_bss_conf *link) +{ + u8 i; + + /* NOTE: the 0 here is IEEE80211_TPE_CAT_6GHZ_DEFAULT, + * we fully ignore IEEE80211_TPE_CAT_6GHZ_SUBORDINATE + */ + + BUILD_BUG_ON(ARRAY_SIZE(cmd->psd_pwr) != + ARRAY_SIZE(link->tpe.psd_local[0].power)); + + /* if not valid, mac80211 puts default (max value) */ + for (i = 0; i < ARRAY_SIZE(cmd->psd_pwr); i++) + cmd->psd_pwr[i] = min(link->tpe.psd_local[0].power[i], + link->tpe.psd_reg_client[0].power[i]); + + BUILD_BUG_ON(ARRAY_SIZE(cmd->eirp_pwr) != + ARRAY_SIZE(link->tpe.max_local[0].power)); + + for (i = 0; i < ARRAY_SIZE(cmd->eirp_pwr); i++) + cmd->eirp_pwr[i] = min(link->tpe.max_local[0].power[i], + link->tpe.max_reg_client[0].power[i]); +} + +void +iwl_mld_send_ap_tx_power_constraint_cmd(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link) +{ + struct iwl_txpower_constraints_cmd cmd = {}; + struct iwl_mld_link *mld_link = iwl_mld_link_from_mac80211(link); + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + if (!mld_link->active) + return; + + if (link->chanreq.oper.chan->band != NL80211_BAND_6GHZ) + return; + + cmd.link_id = cpu_to_le16(mld_link->fw_id); + memset(cmd.psd_pwr, DEFAULT_TPE_TX_POWER, sizeof(cmd.psd_pwr)); + memset(cmd.eirp_pwr, DEFAULT_TPE_TX_POWER, sizeof(cmd.eirp_pwr)); + + if (vif->type == NL80211_IFTYPE_AP) { + cmd.ap_type = cpu_to_le16(IWL_6GHZ_AP_TYPE_VLP); + } else if (link->power_type == IEEE80211_REG_UNSET_AP) { + return; + } else { + cmd.ap_type = cpu_to_le16(link->power_type - 1); + iwl_mld_tpe_sta_cmd_data(&cmd, link); + } + + ret = iwl_mld_send_cmd_pdu(mld, + WIDE_ID(PHY_OPS_GROUP, + AP_TX_POWER_CONSTRAINTS_CMD), + &cmd); + if (ret) + IWL_ERR(mld, + "failed to send AP_TX_POWER_CONSTRAINTS_CMD (%d)\n", + ret); +} + +int iwl_mld_set_tx_power(struct iwl_mld *mld, + struct ieee80211_bss_conf *link_conf, + s16 tx_power) +{ + u32 cmd_id = REDUCE_TX_POWER_CMD; + struct iwl_mld_link *mld_link = iwl_mld_link_from_mac80211(link_conf); + u16 u_tx_power = tx_power == IWL_DEFAULT_MAX_TX_POWER ? + IWL_DEV_MAX_TX_POWER : 8 * tx_power; + struct iwl_dev_tx_power_cmd cmd = { + /* Those fields sit on the same place for v9 and v10 */ + .common.set_mode = cpu_to_le32(IWL_TX_POWER_MODE_SET_LINK), + .common.pwr_restriction = cpu_to_le16(u_tx_power), + }; + u8 cmd_ver = iwl_fw_lookup_cmd_ver(mld->fw, cmd_id, + IWL_FW_CMD_VER_UNKNOWN); + int len = sizeof(cmd.common); + + if (WARN_ON(!mld_link)) + return -ENODEV; + + cmd.common.link_id = cpu_to_le32(mld_link->fw_id); + + if (cmd_ver == 10) + len += sizeof(cmd.v10); + else if (cmd_ver == 9) + len += sizeof(cmd.v9); + + return iwl_mld_send_cmd_pdu(mld, cmd_id, &cmd, len); +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/power.h b/drivers/net/wireless/intel/iwlwifi/mld/power.h new file mode 100644 index 0000000000000..05ce27bef1067 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/power.h @@ -0,0 +1,33 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024 Intel Corporation + */ +#ifndef __iwl_mld_power_h__ +#define __iwl_mld_power_h__ + +#include + +#include "mld.h" + +int iwl_mld_update_device_power(struct iwl_mld *mld, bool d3); + +int iwl_mld_enable_beacon_filter(struct iwl_mld *mld, + const struct ieee80211_bss_conf *link_conf, + bool d3); + +int iwl_mld_disable_beacon_filter(struct iwl_mld *mld, + struct ieee80211_vif *vif); + +int iwl_mld_update_mac_power(struct iwl_mld *mld, struct ieee80211_vif *vif, + bool d3); + +void +iwl_mld_send_ap_tx_power_constraint_cmd(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link); + +int iwl_mld_set_tx_power(struct iwl_mld *mld, + struct ieee80211_bss_conf *link_conf, + s16 tx_power); + +#endif /* __iwl_mld_power_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/ptp.c b/drivers/net/wireless/intel/iwlwifi/mld/ptp.c new file mode 100644 index 0000000000000..d5c3f853d96c1 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/ptp.c @@ -0,0 +1,321 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2025 Intel Corporation + */ + +#include "mld.h" +#include "iwl-debug.h" +#include "hcmd.h" +#include "ptp.h" +#include + +/* The scaled_ppm parameter is ppm (parts per million) with a 16-bit fractional + * part, which means that a value of 1 in one of those fields actually means + * 2^-16 ppm, and 2^16=65536 is 1 ppm. + */ +#define PTP_SCALE_FACTOR 65536000000ULL + +#define IWL_PTP_GP2_WRAP 0x100000000ULL +#define IWL_PTP_WRAP_TIME (3600 * HZ) +#define IWL_PTP_WRAP_THRESHOLD_USEC (5000) + +static int iwl_mld_get_systime(struct iwl_mld *mld, u32 *gp2) +{ + *gp2 = iwl_read_prph(mld->trans, mld->trans->cfg->gp2_reg_addr); + + if (*gp2 == 0x5a5a5a5a) + return -EINVAL; + + return 0; +} + +static void iwl_mld_ptp_update_new_read(struct iwl_mld *mld, u32 gp2) +{ + IWL_DEBUG_PTP(mld, "PTP: last_gp2=%u, new gp2 read=%u\n", + mld->ptp_data.last_gp2, gp2); + + /* If the difference is above the threshold, assume it's a wraparound. + * Otherwise assume it's an old read and ignore it. + */ + if (gp2 < mld->ptp_data.last_gp2) { + if (mld->ptp_data.last_gp2 - gp2 < + IWL_PTP_WRAP_THRESHOLD_USEC) { + IWL_DEBUG_PTP(mld, + "PTP: ignore old read (gp2=%u, last_gp2=%u)\n", + gp2, mld->ptp_data.last_gp2); + return; + } + + mld->ptp_data.wrap_counter++; + IWL_DEBUG_PTP(mld, + "PTP: wraparound detected (new counter=%u)\n", + mld->ptp_data.wrap_counter); + } + + mld->ptp_data.last_gp2 = gp2; + schedule_delayed_work(&mld->ptp_data.dwork, IWL_PTP_WRAP_TIME); +} + +u64 iwl_mld_ptp_get_adj_time(struct iwl_mld *mld, u64 base_time_ns) +{ + struct ptp_data *data = &mld->ptp_data; + u64 scale_time_gp2_ns = mld->ptp_data.scale_update_gp2 * NSEC_PER_USEC; + u64 res; + u64 diff; + s64 scaled_diff; + + lockdep_assert_held(&data->lock); + + iwl_mld_ptp_update_new_read(mld, + div64_u64(base_time_ns, NSEC_PER_USEC)); + + base_time_ns = base_time_ns + + (data->wrap_counter * IWL_PTP_GP2_WRAP * NSEC_PER_USEC); + + /* It is possible that a GP2 timestamp was received from fw before the + * last scale update. + */ + if (base_time_ns < scale_time_gp2_ns) { + diff = scale_time_gp2_ns - base_time_ns; + scaled_diff = -mul_u64_u64_div_u64(diff, + data->scaled_freq, + PTP_SCALE_FACTOR); + } else { + diff = base_time_ns - scale_time_gp2_ns; + scaled_diff = mul_u64_u64_div_u64(diff, + data->scaled_freq, + PTP_SCALE_FACTOR); + } + + IWL_DEBUG_PTP(mld, "base_time=%llu diff ns=%llu scaled_diff_ns=%lld\n", + (unsigned long long)base_time_ns, + (unsigned long long)diff, (long long)scaled_diff); + + res = data->scale_update_adj_time_ns + data->delta + scaled_diff; + + IWL_DEBUG_PTP(mld, "scale_update_ns=%llu delta=%lld adj=%llu\n", + (unsigned long long)data->scale_update_adj_time_ns, + (long long)data->delta, (unsigned long long)res); + return res; +} + +static int iwl_mld_ptp_gettime(struct ptp_clock_info *ptp, + struct timespec64 *ts) +{ + struct iwl_mld *mld = container_of(ptp, struct iwl_mld, + ptp_data.ptp_clock_info); + struct ptp_data *data = &mld->ptp_data; + u32 gp2; + u64 ns; + + if (iwl_mld_get_systime(mld, &gp2)) { + IWL_DEBUG_PTP(mld, "PTP: gettime: failed to read systime\n"); + return -EIO; + } + + spin_lock_bh(&data->lock); + ns = iwl_mld_ptp_get_adj_time(mld, (u64)gp2 * NSEC_PER_USEC); + spin_unlock_bh(&data->lock); + + *ts = ns_to_timespec64(ns); + return 0; +} + +static int iwl_mld_ptp_adjtime(struct ptp_clock_info *ptp, s64 delta) +{ + struct iwl_mld *mld = container_of(ptp, struct iwl_mld, + ptp_data.ptp_clock_info); + struct ptp_data *data = &mld->ptp_data; + + spin_lock_bh(&data->lock); + data->delta += delta; + IWL_DEBUG_PTP(mld, "delta=%lld, new delta=%lld\n", (long long)delta, + (long long)data->delta); + spin_unlock_bh(&data->lock); + return 0; +} + +static int iwl_mld_ptp_adjfine(struct ptp_clock_info *ptp, long scaled_ppm) +{ + struct iwl_mld *mld = container_of(ptp, struct iwl_mld, + ptp_data.ptp_clock_info); + struct ptp_data *data = &mld->ptp_data; + u32 gp2; + + /* Must call iwl_mld_ptp_get_adj_time() before updating + * data->scale_update_gp2 or data->scaled_freq since + * scale_update_adj_time_ns should reflect the previous scaled_freq. + */ + if (iwl_mld_get_systime(mld, &gp2)) { + IWL_DEBUG_PTP(mld, "adjfine: failed to read systime\n"); + return -EBUSY; + } + + spin_lock_bh(&data->lock); + data->scale_update_adj_time_ns = + iwl_mld_ptp_get_adj_time(mld, gp2 * NSEC_PER_USEC); + data->scale_update_gp2 = gp2; + + /* scale_update_adj_time_ns now relects the configured delta, the + * wrap_counter and the previous scaled frequency. Thus delta and + * wrap_counter should be reset, and the scale frequency is updated + * to the new frequency. + */ + data->delta = 0; + data->wrap_counter = 0; + data->scaled_freq = PTP_SCALE_FACTOR + scaled_ppm; + IWL_DEBUG_PTP(mld, "adjfine: scaled_ppm=%ld new=%llu\n", + scaled_ppm, (unsigned long long)data->scaled_freq); + spin_unlock_bh(&data->lock); + return 0; +} + +static void iwl_mld_ptp_work(struct work_struct *wk) +{ + struct iwl_mld *mld = container_of(wk, struct iwl_mld, + ptp_data.dwork.work); + struct ptp_data *data = &mld->ptp_data; + u32 gp2; + + spin_lock_bh(&data->lock); + if (!iwl_mld_get_systime(mld, &gp2)) + iwl_mld_ptp_update_new_read(mld, gp2); + else + IWL_DEBUG_PTP(mld, "PTP work: failed to read GP2\n"); + spin_unlock_bh(&data->lock); +} + +static int +iwl_mld_get_crosstimestamp_fw(struct iwl_mld *mld, u32 *gp2, u64 *sys_time) +{ + struct iwl_synced_time_cmd synced_time_cmd = { + .operation = cpu_to_le32(IWL_SYNCED_TIME_OPERATION_READ_BOTH) + }; + struct iwl_host_cmd cmd = { + .id = WIDE_ID(DATA_PATH_GROUP, WNM_PLATFORM_PTM_REQUEST_CMD), + .flags = CMD_WANT_SKB, + .data[0] = &synced_time_cmd, + .len[0] = sizeof(synced_time_cmd), + }; + struct iwl_synced_time_rsp *resp; + struct iwl_rx_packet *pkt; + int ret; + u64 gp2_10ns; + + wiphy_lock(mld->wiphy); + ret = iwl_mld_send_cmd(mld, &cmd); + wiphy_unlock(mld->wiphy); + if (ret) + return ret; + + pkt = cmd.resp_pkt; + + if (iwl_rx_packet_payload_len(pkt) != sizeof(*resp)) { + IWL_DEBUG_PTP(mld, "PTP: Invalid PTM command response\n"); + iwl_free_resp(&cmd); + return -EIO; + } + + resp = (void *)pkt->data; + + gp2_10ns = (u64)le32_to_cpu(resp->gp2_timestamp_hi) << 32 | + le32_to_cpu(resp->gp2_timestamp_lo); + *gp2 = div_u64(gp2_10ns, 100); + + *sys_time = (u64)le32_to_cpu(resp->platform_timestamp_hi) << 32 | + le32_to_cpu(resp->platform_timestamp_lo); + + iwl_free_resp(&cmd); + return ret; +} + +static int +iwl_mld_phc_get_crosstimestamp(struct ptp_clock_info *ptp, + struct system_device_crosststamp *xtstamp) +{ + struct iwl_mld *mld = container_of(ptp, struct iwl_mld, + ptp_data.ptp_clock_info); + struct ptp_data *data = &mld->ptp_data; + int ret = 0; + /* Raw value read from GP2 register in usec */ + u32 gp2; + /* GP2 value in ns*/ + s64 gp2_ns; + /* System (wall) time */ + ktime_t sys_time; + + memset(xtstamp, 0, sizeof(struct system_device_crosststamp)); + + ret = iwl_mld_get_crosstimestamp_fw(mld, &gp2, &sys_time); + if (ret) { + IWL_DEBUG_PTP(mld, + "PTP: fw get_crosstimestamp failed (ret=%d)\n", + ret); + return ret; + } + + spin_lock_bh(&data->lock); + gp2_ns = iwl_mld_ptp_get_adj_time(mld, (u64)gp2 * NSEC_PER_USEC); + spin_unlock_bh(&data->lock); + + IWL_DEBUG_PTP(mld, + "Got Sync Time: GP2:%u, last_GP2: %u, GP2_ns: %lld, sys_time: %lld\n", + gp2, mld->ptp_data.last_gp2, gp2_ns, (s64)sys_time); + + /* System monotonic raw time is not used */ + xtstamp->device = ns_to_ktime(gp2_ns); + xtstamp->sys_realtime = sys_time; + + return ret; +} + +void iwl_mld_ptp_init(struct iwl_mld *mld) +{ + if (WARN_ON(mld->ptp_data.ptp_clock)) + return; + + spin_lock_init(&mld->ptp_data.lock); + INIT_DELAYED_WORK(&mld->ptp_data.dwork, iwl_mld_ptp_work); + + mld->ptp_data.ptp_clock_info.owner = THIS_MODULE; + mld->ptp_data.ptp_clock_info.gettime64 = iwl_mld_ptp_gettime; + mld->ptp_data.ptp_clock_info.max_adj = 0x7fffffff; + mld->ptp_data.ptp_clock_info.adjtime = iwl_mld_ptp_adjtime; + mld->ptp_data.ptp_clock_info.adjfine = iwl_mld_ptp_adjfine; + mld->ptp_data.scaled_freq = PTP_SCALE_FACTOR; + mld->ptp_data.ptp_clock_info.getcrosststamp = + iwl_mld_phc_get_crosstimestamp; + + /* Give a short 'friendly name' to identify the PHC clock */ + snprintf(mld->ptp_data.ptp_clock_info.name, + sizeof(mld->ptp_data.ptp_clock_info.name), + "%s", "iwlwifi-PTP"); + + mld->ptp_data.ptp_clock = + ptp_clock_register(&mld->ptp_data.ptp_clock_info, mld->dev); + + if (IS_ERR_OR_NULL(mld->ptp_data.ptp_clock)) { + IWL_ERR(mld, "Failed to register PHC clock (%ld)\n", + PTR_ERR(mld->ptp_data.ptp_clock)); + mld->ptp_data.ptp_clock = NULL; + } else { + IWL_INFO(mld, "Registered PHC clock: %s, with index: %d\n", + mld->ptp_data.ptp_clock_info.name, + ptp_clock_index(mld->ptp_data.ptp_clock)); + } +} + +void iwl_mld_ptp_remove(struct iwl_mld *mld) +{ + if (mld->ptp_data.ptp_clock) { + IWL_INFO(mld, "Unregistering PHC clock: %s, with index: %d\n", + mld->ptp_data.ptp_clock_info.name, + ptp_clock_index(mld->ptp_data.ptp_clock)); + + ptp_clock_unregister(mld->ptp_data.ptp_clock); + mld->ptp_data.ptp_clock = NULL; + mld->ptp_data.last_gp2 = 0; + mld->ptp_data.wrap_counter = 0; + cancel_delayed_work_sync(&mld->ptp_data.dwork); + } +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/ptp.h b/drivers/net/wireless/intel/iwlwifi/mld/ptp.h new file mode 100644 index 0000000000000..f3d18dd304e53 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/ptp.h @@ -0,0 +1,45 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2025 Intel Corporation + */ +#ifndef __iwl_mld_ptp_h__ +#define __iwl_mld_ptp_h__ + +#include + +/** + * struct ptp_data - PTP hardware clock data + * + * @ptp_clock: struct ptp_clock pointer returned by the ptp_clock_register() + * function. + * @ptp_clock_info: struct ptp_clock_info that describes a PTP hardware clock + * @lock: protects the time adjustments data + * @delta: delta between hardware clock and ptp clock in nanoseconds + * @scale_update_gp2: GP2 time when the scale was last updated + * @scale_update_adj_time_ns: adjusted time when the scale was last updated, + * in nanoseconds + * @scaled_freq: clock frequency offset, scaled to 65536000000 + * @last_gp2: the last GP2 reading from the hardware, used for tracking GP2 + * wraparounds + * @wrap_counter: number of wraparounds since scale_update_adj_time_ns + * @dwork: worker scheduled every 1 hour to detect workarounds + */ +struct ptp_data { + struct ptp_clock *ptp_clock; + struct ptp_clock_info ptp_clock_info; + + spinlock_t lock; + s64 delta; + u32 scale_update_gp2; + u64 scale_update_adj_time_ns; + u64 scaled_freq; + u32 last_gp2; + u32 wrap_counter; + struct delayed_work dwork; +}; + +void iwl_mld_ptp_init(struct iwl_mld *mld); +void iwl_mld_ptp_remove(struct iwl_mld *mld); +u64 iwl_mld_ptp_get_adj_time(struct iwl_mld *mld, u64 base_time_ns); + +#endif /* __iwl_mld_ptp_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/regulatory.c b/drivers/net/wireless/intel/iwlwifi/mld/regulatory.c new file mode 100644 index 0000000000000..a75af8c1e8ab0 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/regulatory.c @@ -0,0 +1,393 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024-2025 Intel Corporation + */ + +#include + +#include "fw/regulatory.h" +#include "fw/acpi.h" +#include "fw/uefi.h" + +#include "regulatory.h" +#include "mld.h" +#include "hcmd.h" + +void iwl_mld_get_bios_tables(struct iwl_mld *mld) +{ + int ret; + + iwl_acpi_get_guid_lock_status(&mld->fwrt); + + ret = iwl_bios_get_ppag_table(&mld->fwrt); + if (ret < 0) { + IWL_DEBUG_RADIO(mld, + "PPAG BIOS table invalid or unavailable. (%d)\n", + ret); + } + + ret = iwl_bios_get_wrds_table(&mld->fwrt); + if (ret < 0) { + IWL_DEBUG_RADIO(mld, + "WRDS SAR BIOS table invalid or unavailable. (%d)\n", + ret); + + /* If not available, don't fail and don't bother with EWRD and + * WGDS + */ + + if (!iwl_bios_get_wgds_table(&mld->fwrt)) { + /* If basic SAR is not available, we check for WGDS, + * which should *not* be available either. If it is + * available, issue an error, because we can't use SAR + * Geo without basic SAR. + */ + IWL_ERR(mld, "BIOS contains WGDS but no WRDS\n"); + } + + } else { + ret = iwl_bios_get_ewrd_table(&mld->fwrt); + /* If EWRD is not available, we can still use + * WRDS, so don't fail. + */ + if (ret < 0) + IWL_DEBUG_RADIO(mld, + "EWRD SAR BIOS table invalid or unavailable. (%d)\n", + ret); + + ret = iwl_bios_get_wgds_table(&mld->fwrt); + if (ret < 0) + IWL_DEBUG_RADIO(mld, + "Geo SAR BIOS table invalid or unavailable. (%d)\n", + ret); + /* we don't fail if the table is not available */ + } + + ret = iwl_uefi_get_uats_table(mld->trans, &mld->fwrt); + if (ret) + IWL_DEBUG_RADIO(mld, "failed to read UATS table (%d)\n", ret); +} + +static int iwl_mld_geo_sar_init(struct iwl_mld *mld) +{ + u32 cmd_id = WIDE_ID(PHY_OPS_GROUP, PER_CHAIN_LIMIT_OFFSET_CMD); + union iwl_geo_tx_power_profiles_cmd cmd; + u16 len; + u32 n_bands; + __le32 sk = cpu_to_le32(0); + int ret; + u8 cmd_ver = iwl_fw_lookup_cmd_ver(mld->fw, cmd_id, + IWL_FW_CMD_VER_UNKNOWN); + + BUILD_BUG_ON(offsetof(struct iwl_geo_tx_power_profiles_cmd_v4, ops) != + offsetof(struct iwl_geo_tx_power_profiles_cmd_v5, ops)); + + cmd.v4.ops = cpu_to_le32(IWL_PER_CHAIN_OFFSET_SET_TABLES); + + /* Only set to South Korea if the table revision is 1 */ + if (mld->fwrt.geo_rev == 1) + sk = cpu_to_le32(1); + + if (cmd_ver == 5) { + len = sizeof(cmd.v5); + n_bands = ARRAY_SIZE(cmd.v5.table[0]); + cmd.v5.table_revision = sk; + } else if (cmd_ver == 4) { + len = sizeof(cmd.v4); + n_bands = ARRAY_SIZE(cmd.v4.table[0]); + cmd.v4.table_revision = sk; + } else { + return -EOPNOTSUPP; + } + + BUILD_BUG_ON(offsetof(struct iwl_geo_tx_power_profiles_cmd_v4, table) != + offsetof(struct iwl_geo_tx_power_profiles_cmd_v5, table)); + /* the table is at the same position for all versions, so set use v4 */ + ret = iwl_sar_geo_fill_table(&mld->fwrt, &cmd.v4.table[0][0], + n_bands, BIOS_GEO_MAX_PROFILE_NUM); + + /* It is a valid scenario to not support SAR, or miss wgds table, + * but in that case there is no need to send the command. + */ + if (ret) + return 0; + + return iwl_mld_send_cmd_pdu(mld, cmd_id, &cmd, len); +} + +int iwl_mld_config_sar_profile(struct iwl_mld *mld, int prof_a, int prof_b) +{ + u32 cmd_id = REDUCE_TX_POWER_CMD; + struct iwl_dev_tx_power_cmd cmd = { + .common.set_mode = cpu_to_le32(IWL_TX_POWER_MODE_SET_CHAINS), + }; + __le16 *per_chain; + int ret; + u16 len = sizeof(cmd.common); + u32 n_subbands; + u8 cmd_ver = iwl_fw_lookup_cmd_ver(mld->fw, cmd_id, + IWL_FW_CMD_VER_UNKNOWN); + + if (cmd_ver == 10) { + len += sizeof(cmd.v10); + n_subbands = IWL_NUM_SUB_BANDS_V2; + per_chain = &cmd.v10.per_chain[0][0][0]; + cmd.v10.flags = + cpu_to_le32(mld->fwrt.reduced_power_flags); + } else if (cmd_ver == 9) { + len += sizeof(cmd.v9); + n_subbands = IWL_NUM_SUB_BANDS_V1; + per_chain = &cmd.v9.per_chain[0][0]; + } else { + return -EOPNOTSUPP; + } + + /* TODO: CDB - support IWL_NUM_CHAIN_TABLES_V2 */ + ret = iwl_sar_fill_profile(&mld->fwrt, per_chain, + IWL_NUM_CHAIN_TABLES, + n_subbands, prof_a, prof_b); + /* return on error or if the profile is disabled (positive number) */ + if (ret) + return ret; + + return iwl_mld_send_cmd_pdu(mld, cmd_id, &cmd, len); +} + +int iwl_mld_init_sar(struct iwl_mld *mld) +{ + int chain_a_prof = 1; + int chain_b_prof = 1; + int ret; + + /* If no profile was chosen by the user yet, choose profile 1 (WRDS) as + * default for both chains + */ + if (mld->fwrt.sar_chain_a_profile && mld->fwrt.sar_chain_b_profile) { + chain_a_prof = mld->fwrt.sar_chain_a_profile; + chain_b_prof = mld->fwrt.sar_chain_b_profile; + } + + ret = iwl_mld_config_sar_profile(mld, chain_a_prof, chain_b_prof); + if (ret < 0) + return ret; + + if (ret) + return 0; + + return iwl_mld_geo_sar_init(mld); +} + +int iwl_mld_init_sgom(struct iwl_mld *mld) +{ + int ret; + struct iwl_host_cmd cmd = { + .id = WIDE_ID(REGULATORY_AND_NVM_GROUP, + SAR_OFFSET_MAPPING_TABLE_CMD), + .data[0] = &mld->fwrt.sgom_table, + .len[0] = sizeof(mld->fwrt.sgom_table), + .dataflags[0] = IWL_HCMD_DFL_NOCOPY, + }; + + if (!mld->fwrt.sgom_enabled) { + IWL_DEBUG_RADIO(mld, "SGOM table is disabled\n"); + return 0; + } + + ret = iwl_mld_send_cmd(mld, &cmd); + if (ret) + IWL_ERR(mld, + "failed to send SAR_OFFSET_MAPPING_CMD (%d)\n", ret); + + return ret; +} + +static int iwl_mld_ppag_send_cmd(struct iwl_mld *mld) +{ + union iwl_ppag_table_cmd cmd = {}; + int ret, len; + + ret = iwl_fill_ppag_table(&mld->fwrt, &cmd, &len); + /* Not supporting PPAG table is a valid scenario */ + if (ret < 0) + return 0; + + IWL_DEBUG_RADIO(mld, "Sending PER_PLATFORM_ANT_GAIN_CMD\n"); + ret = iwl_mld_send_cmd_pdu(mld, WIDE_ID(PHY_OPS_GROUP, + PER_PLATFORM_ANT_GAIN_CMD), + &cmd, len); + if (ret < 0) + IWL_ERR(mld, "failed to send PER_PLATFORM_ANT_GAIN_CMD (%d)\n", + ret); + + return ret; +} + +int iwl_mld_init_ppag(struct iwl_mld *mld) +{ + /* no need to read the table, done in INIT stage */ + + if (!(iwl_is_ppag_approved(&mld->fwrt))) + return 0; + + return iwl_mld_ppag_send_cmd(mld); +} + +void iwl_mld_configure_lari(struct iwl_mld *mld) +{ + struct iwl_fw_runtime *fwrt = &mld->fwrt; + struct iwl_lari_config_change_cmd cmd = { + .config_bitmap = iwl_get_lari_config_bitmap(fwrt), + }; + int ret; + u32 value; + + ret = iwl_bios_get_dsm(fwrt, DSM_FUNC_11AX_ENABLEMENT, &value); + if (!ret) + cmd.oem_11ax_allow_bitmap = cpu_to_le32(value); + + ret = iwl_bios_get_dsm(fwrt, DSM_FUNC_ENABLE_UNII4_CHAN, &value); + if (!ret) + cmd.oem_unii4_allow_bitmap = + cpu_to_le32(value &= DSM_UNII4_ALLOW_BITMAP); + + ret = iwl_bios_get_dsm(fwrt, DSM_FUNC_ACTIVATE_CHANNEL, &value); + if (!ret) + cmd.chan_state_active_bitmap = cpu_to_le32(value); + + ret = iwl_bios_get_dsm(fwrt, DSM_FUNC_ENABLE_6E, &value); + if (!ret) + cmd.oem_uhb_allow_bitmap = cpu_to_le32(value); + + ret = iwl_bios_get_dsm(fwrt, DSM_FUNC_FORCE_DISABLE_CHANNELS, &value); + if (!ret) + cmd.force_disable_channels_bitmap = cpu_to_le32(value); + + ret = iwl_bios_get_dsm(fwrt, DSM_FUNC_ENERGY_DETECTION_THRESHOLD, + &value); + if (!ret) + cmd.edt_bitmap = cpu_to_le32(value); + + ret = iwl_bios_get_wbem(fwrt, &value); + if (!ret) + cmd.oem_320mhz_allow_bitmap = cpu_to_le32(value); + + ret = iwl_bios_get_dsm(fwrt, DSM_FUNC_ENABLE_11BE, &value); + if (!ret) + cmd.oem_11be_allow_bitmap = cpu_to_le32(value); + + if (!cmd.config_bitmap && + !cmd.oem_uhb_allow_bitmap && + !cmd.oem_11ax_allow_bitmap && + !cmd.oem_unii4_allow_bitmap && + !cmd.chan_state_active_bitmap && + !cmd.force_disable_channels_bitmap && + !cmd.edt_bitmap && + !cmd.oem_320mhz_allow_bitmap && + !cmd.oem_11be_allow_bitmap) + return; + + IWL_DEBUG_RADIO(mld, + "sending LARI_CONFIG_CHANGE, config_bitmap=0x%x, oem_11ax_allow_bitmap=0x%x\n", + le32_to_cpu(cmd.config_bitmap), + le32_to_cpu(cmd.oem_11ax_allow_bitmap)); + IWL_DEBUG_RADIO(mld, + "sending LARI_CONFIG_CHANGE, oem_unii4_allow_bitmap=0x%x, chan_state_active_bitmap=0x%x\n", + le32_to_cpu(cmd.oem_unii4_allow_bitmap), + le32_to_cpu(cmd.chan_state_active_bitmap)); + IWL_DEBUG_RADIO(mld, + "sending LARI_CONFIG_CHANGE, oem_uhb_allow_bitmap=0x%x, force_disable_channels_bitmap=0x%x\n", + le32_to_cpu(cmd.oem_uhb_allow_bitmap), + le32_to_cpu(cmd.force_disable_channels_bitmap)); + IWL_DEBUG_RADIO(mld, + "sending LARI_CONFIG_CHANGE, edt_bitmap=0x%x, oem_320mhz_allow_bitmap=0x%x\n", + le32_to_cpu(cmd.edt_bitmap), + le32_to_cpu(cmd.oem_320mhz_allow_bitmap)); + IWL_DEBUG_RADIO(mld, + "sending LARI_CONFIG_CHANGE, oem_11be_allow_bitmap=0x%x\n", + le32_to_cpu(cmd.oem_11be_allow_bitmap)); + + ret = iwl_mld_send_cmd_pdu(mld, WIDE_ID(REGULATORY_AND_NVM_GROUP, + LARI_CONFIG_CHANGE), &cmd); + if (ret) + IWL_DEBUG_RADIO(mld, + "Failed to send LARI_CONFIG_CHANGE (%d)\n", + ret); +} + +void iwl_mld_init_uats(struct iwl_mld *mld) +{ + int ret; + struct iwl_host_cmd cmd = { + .id = WIDE_ID(REGULATORY_AND_NVM_GROUP, + MCC_ALLOWED_AP_TYPE_CMD), + .data[0] = &mld->fwrt.uats_table, + .len[0] = sizeof(mld->fwrt.uats_table), + .dataflags[0] = IWL_HCMD_DFL_NOCOPY, + }; + + if (!mld->fwrt.uats_valid) + return; + + ret = iwl_mld_send_cmd(mld, &cmd); + if (ret) + IWL_ERR(mld, "failed to send MCC_ALLOWED_AP_TYPE_CMD (%d)\n", + ret); +} + +void iwl_mld_init_tas(struct iwl_mld *mld) +{ + int ret; + struct iwl_tas_data data = {}; + struct iwl_tas_config_cmd cmd = {}; + u32 cmd_id = WIDE_ID(REGULATORY_AND_NVM_GROUP, TAS_CONFIG); + + BUILD_BUG_ON(ARRAY_SIZE(data.block_list_array) != + IWL_WTAS_BLACK_LIST_MAX); + BUILD_BUG_ON(ARRAY_SIZE(cmd.block_list_array) != + IWL_WTAS_BLACK_LIST_MAX); + + if (!fw_has_capa(&mld->fw->ucode_capa, IWL_UCODE_TLV_CAPA_TAS_CFG)) { + IWL_DEBUG_RADIO(mld, "TAS not enabled in FW\n"); + return; + } + + ret = iwl_bios_get_tas_table(&mld->fwrt, &data); + if (ret < 0) { + IWL_DEBUG_RADIO(mld, + "TAS table invalid or unavailable. (%d)\n", + ret); + return; + } + + if (!iwl_is_tas_approved()) { + IWL_DEBUG_RADIO(mld, + "System vendor '%s' is not in the approved list, disabling TAS in US and Canada.\n", + dmi_get_system_info(DMI_SYS_VENDOR) ?: ""); + if ((!iwl_add_mcc_to_tas_block_list(data.block_list_array, + &data.block_list_size, + IWL_MCC_US)) || + (!iwl_add_mcc_to_tas_block_list(data.block_list_array, + &data.block_list_size, + IWL_MCC_CANADA))) { + IWL_DEBUG_RADIO(mld, + "Unable to add US/Canada to TAS block list, disabling TAS\n"); + return; + } + } else { + IWL_DEBUG_RADIO(mld, + "System vendor '%s' is in the approved list.\n", + dmi_get_system_info(DMI_SYS_VENDOR) ?: ""); + } + + cmd.block_list_size = cpu_to_le16(data.block_list_size); + for (u8 i = 0; i < data.block_list_size; i++) + cmd.block_list_array[i] = + cpu_to_le16(data.block_list_array[i]); + cmd.tas_config_info.table_source = data.table_source; + cmd.tas_config_info.table_revision = data.table_revision; + cmd.tas_config_info.value = cpu_to_le32(data.tas_selection); + + ret = iwl_mld_send_cmd_pdu(mld, cmd_id, &cmd); + if (ret) + IWL_DEBUG_RADIO(mld, "failed to send TAS_CONFIG (%d)\n", ret); +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/regulatory.h b/drivers/net/wireless/intel/iwlwifi/mld/regulatory.h new file mode 100644 index 0000000000000..3b01c645adda3 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/regulatory.h @@ -0,0 +1,23 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024-2025 Intel Corporation + */ +#ifndef __iwl_mld_regulatory_h__ +#define __iwl_mld_regulatory_h__ + +#include "mld.h" + +void iwl_mld_get_bios_tables(struct iwl_mld *mld); +void iwl_mld_configure_lari(struct iwl_mld *mld); +void iwl_mld_init_uats(struct iwl_mld *mld); +void iwl_mld_init_tas(struct iwl_mld *mld); + +int iwl_mld_init_ppag(struct iwl_mld *mld); + +int iwl_mld_init_sgom(struct iwl_mld *mld); + +int iwl_mld_init_sar(struct iwl_mld *mld); + +int iwl_mld_config_sar_profile(struct iwl_mld *mld, int prof_a, int prof_b); + +#endif /* __iwl_mld_regulatory_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/roc.c b/drivers/net/wireless/intel/iwlwifi/mld/roc.c new file mode 100644 index 0000000000000..b87faca23cebe --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/roc.c @@ -0,0 +1,224 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024 - 2025 Intel Corporation + */ +#include +#include + +#include "mld.h" +#include "roc.h" +#include "hcmd.h" +#include "iface.h" +#include "sta.h" +#include "mlo.h" + +#include "fw/api/context.h" +#include "fw/api/time-event.h" + +#define AUX_ROC_MAX_DELAY MSEC_TO_TU(200) + +static void +iwl_mld_vif_iter_emlsr_block_roc(void *data, u8 *mac, struct ieee80211_vif *vif) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + int *result = data; + int ret; + + ret = iwl_mld_block_emlsr_sync(mld_vif->mld, vif, + IWL_MLD_EMLSR_BLOCKED_ROC, + iwl_mld_get_primary_link(vif)); + if (ret) + *result = ret; +} + +int iwl_mld_start_roc(struct ieee80211_hw *hw, struct ieee80211_vif *vif, + struct ieee80211_channel *channel, int duration, + enum ieee80211_roc_type type) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mld_int_sta *aux_sta; + struct iwl_roc_req cmd = { + .action = cpu_to_le32(FW_CTXT_ACTION_ADD), + }; + u8 ver = iwl_fw_lookup_cmd_ver(mld->fw, + WIDE_ID(MAC_CONF_GROUP, ROC_CMD), 0); + u16 cmd_len = ver < 6 ? sizeof(struct iwl_roc_req_v5) : sizeof(cmd); + enum iwl_roc_activity activity; + int ret = 0; + + lockdep_assert_wiphy(mld->wiphy); + + ieee80211_iterate_active_interfaces_mtx(mld->hw, + IEEE80211_IFACE_ITER_NORMAL, + iwl_mld_vif_iter_emlsr_block_roc, + &ret); + if (ret) + return ret; + + /* TODO: task=Hotspot 2.0 */ + if (vif->type != NL80211_IFTYPE_P2P_DEVICE) { + IWL_ERR(mld, "NOT SUPPORTED: ROC on vif->type %d\n", + vif->type); + + return -EOPNOTSUPP; + } + + switch (type) { + case IEEE80211_ROC_TYPE_NORMAL: + activity = ROC_ACTIVITY_P2P_DISC; + break; + case IEEE80211_ROC_TYPE_MGMT_TX: + activity = ROC_ACTIVITY_P2P_NEG; + break; + default: + WARN_ONCE(1, "Got an invalid P2P ROC type\n"); + return -EINVAL; + } + + if (WARN_ON(mld_vif->roc_activity != ROC_NUM_ACTIVITIES)) + return -EBUSY; + + /* No MLO on P2P device */ + aux_sta = &mld_vif->deflink.aux_sta; + + ret = iwl_mld_add_aux_sta(mld, aux_sta); + if (ret) + return ret; + + cmd.activity = cpu_to_le32(activity); + cmd.sta_id = cpu_to_le32(aux_sta->sta_id); + cmd.channel_info.channel = cpu_to_le32(channel->hw_value); + cmd.channel_info.band = iwl_mld_nl80211_band_to_fw(channel->band); + cmd.channel_info.width = IWL_PHY_CHANNEL_MODE20; + /* TODO: task=Hotspot 2.0, revisit those parameters when we add an ROC + * on the BSS vif + */ + cmd.max_delay = cpu_to_le32(AUX_ROC_MAX_DELAY); + cmd.duration = cpu_to_le32(MSEC_TO_TU(duration)); + + memcpy(cmd.node_addr, vif->addr, ETH_ALEN); + + ret = iwl_mld_send_cmd_pdu(mld, WIDE_ID(MAC_CONF_GROUP, ROC_CMD), + &cmd, cmd_len); + if (ret) { + IWL_ERR(mld, "Couldn't send the ROC_CMD\n"); + return ret; + } + mld_vif->roc_activity = activity; + + return 0; +} + +static void +iwl_mld_vif_iter_emlsr_unblock_roc(void *data, u8 *mac, + struct ieee80211_vif *vif) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + + iwl_mld_unblock_emlsr(mld_vif->mld, vif, IWL_MLD_EMLSR_BLOCKED_ROC); +} + +static void iwl_mld_destroy_roc(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct iwl_mld_vif *mld_vif) +{ + mld_vif->roc_activity = ROC_NUM_ACTIVITIES; + + ieee80211_iterate_active_interfaces_mtx(mld->hw, + IEEE80211_IFACE_ITER_NORMAL, + iwl_mld_vif_iter_emlsr_unblock_roc, + NULL); + + /* wait until every tx has seen that roc_activity has been reset */ + synchronize_net(); + /* from here, no new tx will be added + * we can flush the Tx on the queues + */ + + iwl_mld_flush_link_sta_txqs(mld, mld_vif->deflink.aux_sta.sta_id); + + iwl_mld_remove_aux_sta(mld, vif, &vif->bss_conf); +} + +int iwl_mld_cancel_roc(struct ieee80211_hw *hw, + struct ieee80211_vif *vif) +{ + struct iwl_mld *mld = IWL_MAC80211_GET_MLD(hw); + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_roc_req cmd = { + .action = cpu_to_le32(FW_CTXT_ACTION_REMOVE), + }; + u8 ver = iwl_fw_lookup_cmd_ver(mld->fw, + WIDE_ID(MAC_CONF_GROUP, ROC_CMD), 0); + u16 cmd_len = ver < 6 ? sizeof(struct iwl_roc_req_v5) : sizeof(cmd); + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + /* TODO: task=Hotspot 2.0 */ + if (WARN_ON(vif->type != NL80211_IFTYPE_P2P_DEVICE)) + return -EOPNOTSUPP; + + /* No roc activity running it's probably already done */ + if (mld_vif->roc_activity == ROC_NUM_ACTIVITIES) + return 0; + + cmd.activity = cpu_to_le32(mld_vif->roc_activity); + + ret = iwl_mld_send_cmd_pdu(mld, WIDE_ID(MAC_CONF_GROUP, ROC_CMD), + &cmd, cmd_len); + if (ret) + IWL_ERR(mld, "Couldn't send the command to cancel the ROC\n"); + + /* We may have raced with the firmware expiring the ROC instance at + * this very moment. In that case, we can have a notification in the + * async processing queue. However, none can arrive _after_ this as + * ROC_CMD was sent synchronously, i.e. we waited for a response and + * the firmware cannot refer to this ROC after the response. Thus, + * if we just cancel the notification (if there's one) we'll be at a + * clean state for any possible next ROC. + */ + iwl_mld_cancel_notifications_of_object(mld, IWL_MLD_OBJECT_TYPE_ROC, + mld_vif->roc_activity); + + iwl_mld_destroy_roc(mld, vif, mld_vif); + + return 0; +} + +void iwl_mld_handle_roc_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + const struct iwl_roc_notif *notif = (void *)pkt->data; + u32 activity = le32_to_cpu(notif->activity); + /* TODO: task=Hotspot 2.0 - roc can run on BSS */ + struct ieee80211_vif *vif = mld->p2p_device_vif; + struct iwl_mld_vif *mld_vif; + + if (WARN_ON(!vif)) + return; + + mld_vif = iwl_mld_vif_from_mac80211(vif); + /* It is possible that the ROC was canceled + * but the notification was already fired. + */ + if (mld_vif->roc_activity != activity) + return; + + if (le32_to_cpu(notif->success) && + le32_to_cpu(notif->started)) { + /* We had a successful start */ + ieee80211_ready_on_channel(mld->hw); + } else { + /* ROC was not successful, tell the firmware to remove it */ + if (le32_to_cpu(notif->started)) + iwl_mld_cancel_roc(mld->hw, vif); + else + iwl_mld_destroy_roc(mld, vif, mld_vif); + /* we need to let know mac80211 about end OR + * an unsuccessful start + */ + ieee80211_remain_on_channel_expired(mld->hw); + } +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/roc.h b/drivers/net/wireless/intel/iwlwifi/mld/roc.h new file mode 100644 index 0000000000000..985d7f20a3d7f --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/roc.h @@ -0,0 +1,20 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024 Intel Corporation + */ +#ifndef __iwl_mld_roc_h__ +#define __iwl_mld_roc_h__ + +#include + +int iwl_mld_start_roc(struct ieee80211_hw *hw, struct ieee80211_vif *vif, + struct ieee80211_channel *channel, int duration, + enum ieee80211_roc_type type); + +int iwl_mld_cancel_roc(struct ieee80211_hw *hw, + struct ieee80211_vif *vif); + +void iwl_mld_handle_roc_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt); + +#endif /* __iwl_mld_roc_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/rx.c b/drivers/net/wireless/intel/iwlwifi/mld/rx.c new file mode 100644 index 0000000000000..c4f189bcece21 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/rx.c @@ -0,0 +1,2060 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024-2025 Intel Corporation + */ + +#include +#include + +#include "mld.h" +#include "sta.h" +#include "agg.h" +#include "rx.h" +#include "hcmd.h" +#include "iface.h" +#include "time_sync.h" +#include "fw/dbg.h" +#include "fw/api/rx.h" + +/* stores relevant PHY data fields extracted from iwl_rx_mpdu_desc */ +struct iwl_mld_rx_phy_data { + enum iwl_rx_phy_info_type info_type; + __le32 data0; + __le32 data1; + __le32 data2; + __le32 data3; + __le32 eht_data4; + __le32 data5; + __le16 data4; + bool first_subframe; + bool with_data; + __le32 rx_vec[4]; + u32 rate_n_flags; + u32 gp2_on_air_rise; + u16 phy_info; + u8 energy_a, energy_b; + u8 channel; +}; + +static void +iwl_mld_fill_phy_data(struct iwl_rx_mpdu_desc *desc, + struct iwl_mld_rx_phy_data *phy_data) +{ + phy_data->phy_info = le16_to_cpu(desc->phy_info); + phy_data->rate_n_flags = le32_to_cpu(desc->v3.rate_n_flags); + phy_data->gp2_on_air_rise = le32_to_cpu(desc->v3.gp2_on_air_rise); + phy_data->channel = desc->v3.channel; + phy_data->energy_a = desc->v3.energy_a; + phy_data->energy_b = desc->v3.energy_b; + phy_data->data0 = desc->v3.phy_data0; + phy_data->data1 = desc->v3.phy_data1; + phy_data->data2 = desc->v3.phy_data2; + phy_data->data3 = desc->v3.phy_data3; + phy_data->data4 = desc->phy_data4; + phy_data->eht_data4 = desc->phy_eht_data4; + phy_data->data5 = desc->v3.phy_data5; + phy_data->with_data = true; +} + +static inline int iwl_mld_check_pn(struct iwl_mld *mld, struct sk_buff *skb, + int queue, struct ieee80211_sta *sta) +{ + struct ieee80211_hdr *hdr = (void *)skb_mac_header(skb); + struct ieee80211_rx_status *stats = IEEE80211_SKB_RXCB(skb); + struct iwl_mld_sta *mld_sta; + struct iwl_mld_ptk_pn *ptk_pn; + int res; + u8 tid, keyidx; + u8 pn[IEEE80211_CCMP_PN_LEN]; + u8 *extiv; + + /* multicast and non-data only arrives on default queue; avoid checking + * for default queue - we don't want to replicate all the logic that's + * necessary for checking the PN on fragmented frames, leave that + * to mac80211 + */ + if (queue == 0 || !ieee80211_is_data(hdr->frame_control) || + is_multicast_ether_addr(hdr->addr1)) + return 0; + + if (!(stats->flag & RX_FLAG_DECRYPTED)) + return 0; + + /* if we are here - this for sure is either CCMP or GCMP */ + if (!sta) { + IWL_DEBUG_DROP(mld, + "expected hw-decrypted unicast frame for station\n"); + return -1; + } + + mld_sta = iwl_mld_sta_from_mac80211(sta); + + extiv = (u8 *)hdr + ieee80211_hdrlen(hdr->frame_control); + keyidx = extiv[3] >> 6; + + ptk_pn = rcu_dereference(mld_sta->ptk_pn[keyidx]); + if (!ptk_pn) + return -1; + + if (ieee80211_is_data_qos(hdr->frame_control)) + tid = ieee80211_get_tid(hdr); + else + tid = 0; + + /* we don't use HCCA/802.11 QoS TSPECs, so drop such frames */ + if (tid >= IWL_MAX_TID_COUNT) + return -1; + + /* load pn */ + pn[0] = extiv[7]; + pn[1] = extiv[6]; + pn[2] = extiv[5]; + pn[3] = extiv[4]; + pn[4] = extiv[1]; + pn[5] = extiv[0]; + + res = memcmp(pn, ptk_pn->q[queue].pn[tid], IEEE80211_CCMP_PN_LEN); + if (res < 0) + return -1; + if (!res && !(stats->flag & RX_FLAG_ALLOW_SAME_PN)) + return -1; + + memcpy(ptk_pn->q[queue].pn[tid], pn, IEEE80211_CCMP_PN_LEN); + stats->flag |= RX_FLAG_PN_VALIDATED; + + return 0; +} + +/* iwl_mld_pass_packet_to_mac80211 - passes the packet for mac80211 */ +void iwl_mld_pass_packet_to_mac80211(struct iwl_mld *mld, + struct napi_struct *napi, + struct sk_buff *skb, int queue, + struct ieee80211_sta *sta) +{ + KUNIT_STATIC_STUB_REDIRECT(iwl_mld_pass_packet_to_mac80211, + mld, napi, skb, queue, sta); + + if (unlikely(iwl_mld_check_pn(mld, skb, queue, sta))) { + kfree_skb(skb); + return; + } + + ieee80211_rx_napi(mld->hw, sta, skb, napi); +} +EXPORT_SYMBOL_IF_IWLWIFI_KUNIT(iwl_mld_pass_packet_to_mac80211); + +static void iwl_mld_fill_signal(struct iwl_mld *mld, + struct ieee80211_rx_status *rx_status, + struct iwl_mld_rx_phy_data *phy_data) +{ + u32 rate_n_flags = phy_data->rate_n_flags; + int energy_a = phy_data->energy_a; + int energy_b = phy_data->energy_b; + int max_energy; + + energy_a = energy_a ? -energy_a : S8_MIN; + energy_b = energy_b ? -energy_b : S8_MIN; + max_energy = max(energy_a, energy_b); + + IWL_DEBUG_STATS(mld, "energy in A %d B %d, and max %d\n", + energy_a, energy_b, max_energy); + + rx_status->signal = max_energy; + rx_status->chains = + (rate_n_flags & RATE_MCS_ANT_AB_MSK) >> RATE_MCS_ANT_POS; + rx_status->chain_signal[0] = energy_a; + rx_status->chain_signal[1] = energy_b; +} + +static void +iwl_mld_decode_he_phy_ru_alloc(struct iwl_mld_rx_phy_data *phy_data, + struct ieee80211_radiotap_he *he, + struct ieee80211_radiotap_he_mu *he_mu, + struct ieee80211_rx_status *rx_status) +{ + /* Unfortunately, we have to leave the mac80211 data + * incorrect for the case that we receive an HE-MU + * transmission and *don't* have the HE phy data (due + * to the bits being used for TSF). This shouldn't + * happen though as management frames where we need + * the TSF/timers are not be transmitted in HE-MU. + */ + u8 ru = le32_get_bits(phy_data->data1, IWL_RX_PHY_DATA1_HE_RU_ALLOC_MASK); + u32 rate_n_flags = phy_data->rate_n_flags; + u32 he_type = rate_n_flags & RATE_MCS_HE_TYPE_MSK; + u8 offs = 0; + + rx_status->bw = RATE_INFO_BW_HE_RU; + + he->data1 |= cpu_to_le16(IEEE80211_RADIOTAP_HE_DATA1_BW_RU_ALLOC_KNOWN); + + switch (ru) { + case 0 ... 36: + rx_status->he_ru = NL80211_RATE_INFO_HE_RU_ALLOC_26; + offs = ru; + break; + case 37 ... 52: + rx_status->he_ru = NL80211_RATE_INFO_HE_RU_ALLOC_52; + offs = ru - 37; + break; + case 53 ... 60: + rx_status->he_ru = NL80211_RATE_INFO_HE_RU_ALLOC_106; + offs = ru - 53; + break; + case 61 ... 64: + rx_status->he_ru = NL80211_RATE_INFO_HE_RU_ALLOC_242; + offs = ru - 61; + break; + case 65 ... 66: + rx_status->he_ru = NL80211_RATE_INFO_HE_RU_ALLOC_484; + offs = ru - 65; + break; + case 67: + rx_status->he_ru = NL80211_RATE_INFO_HE_RU_ALLOC_996; + break; + case 68: + rx_status->he_ru = NL80211_RATE_INFO_HE_RU_ALLOC_2x996; + break; + } + he->data2 |= le16_encode_bits(offs, + IEEE80211_RADIOTAP_HE_DATA2_RU_OFFSET); + he->data2 |= cpu_to_le16(IEEE80211_RADIOTAP_HE_DATA2_PRISEC_80_KNOWN | + IEEE80211_RADIOTAP_HE_DATA2_RU_OFFSET_KNOWN); + if (phy_data->data1 & cpu_to_le32(IWL_RX_PHY_DATA1_HE_RU_ALLOC_SEC80)) + he->data2 |= + cpu_to_le16(IEEE80211_RADIOTAP_HE_DATA2_PRISEC_80_SEC); + +#define CHECK_BW(bw) \ + BUILD_BUG_ON(IEEE80211_RADIOTAP_HE_MU_FLAGS2_BW_FROM_SIG_A_BW_ ## bw ## MHZ != \ + RATE_MCS_CHAN_WIDTH_##bw >> RATE_MCS_CHAN_WIDTH_POS); \ + BUILD_BUG_ON(IEEE80211_RADIOTAP_HE_DATA6_TB_PPDU_BW_ ## bw ## MHZ != \ + RATE_MCS_CHAN_WIDTH_##bw >> RATE_MCS_CHAN_WIDTH_POS) + CHECK_BW(20); + CHECK_BW(40); + CHECK_BW(80); + CHECK_BW(160); + + if (he_mu) + he_mu->flags2 |= + le16_encode_bits(u32_get_bits(rate_n_flags, + RATE_MCS_CHAN_WIDTH_MSK), + IEEE80211_RADIOTAP_HE_MU_FLAGS2_BW_FROM_SIG_A_BW); + else if (he_type == RATE_MCS_HE_TYPE_TRIG) + he->data6 |= + cpu_to_le16(IEEE80211_RADIOTAP_HE_DATA6_TB_PPDU_BW_KNOWN) | + le16_encode_bits(u32_get_bits(rate_n_flags, + RATE_MCS_CHAN_WIDTH_MSK), + IEEE80211_RADIOTAP_HE_DATA6_TB_PPDU_BW); +} + +static void +iwl_mld_decode_he_mu_ext(struct iwl_mld_rx_phy_data *phy_data, + struct ieee80211_radiotap_he_mu *he_mu) +{ + u32 phy_data2 = le32_to_cpu(phy_data->data2); + u32 phy_data3 = le32_to_cpu(phy_data->data3); + u16 phy_data4 = le16_to_cpu(phy_data->data4); + u32 rate_n_flags = phy_data->rate_n_flags; + + if (u32_get_bits(phy_data4, IWL_RX_PHY_DATA4_HE_MU_EXT_CH1_CRC_OK)) { + he_mu->flags1 |= + cpu_to_le16(IEEE80211_RADIOTAP_HE_MU_FLAGS1_CH1_RU_KNOWN | + IEEE80211_RADIOTAP_HE_MU_FLAGS1_CH1_CTR_26T_RU_KNOWN); + + he_mu->flags1 |= + le16_encode_bits(u32_get_bits(phy_data4, + IWL_RX_PHY_DATA4_HE_MU_EXT_CH1_CTR_RU), + IEEE80211_RADIOTAP_HE_MU_FLAGS1_CH1_CTR_26T_RU); + + he_mu->ru_ch1[0] = u32_get_bits(phy_data2, + IWL_RX_PHY_DATA2_HE_MU_EXT_CH1_RU0); + he_mu->ru_ch1[1] = u32_get_bits(phy_data3, + IWL_RX_PHY_DATA3_HE_MU_EXT_CH1_RU1); + he_mu->ru_ch1[2] = u32_get_bits(phy_data2, + IWL_RX_PHY_DATA2_HE_MU_EXT_CH1_RU2); + he_mu->ru_ch1[3] = u32_get_bits(phy_data3, + IWL_RX_PHY_DATA3_HE_MU_EXT_CH1_RU3); + } + + if (u32_get_bits(phy_data4, IWL_RX_PHY_DATA4_HE_MU_EXT_CH2_CRC_OK) && + (rate_n_flags & RATE_MCS_CHAN_WIDTH_MSK) != RATE_MCS_CHAN_WIDTH_20) { + he_mu->flags1 |= + cpu_to_le16(IEEE80211_RADIOTAP_HE_MU_FLAGS1_CH2_RU_KNOWN | + IEEE80211_RADIOTAP_HE_MU_FLAGS1_CH2_CTR_26T_RU_KNOWN); + + he_mu->flags2 |= + le16_encode_bits(u32_get_bits(phy_data4, + IWL_RX_PHY_DATA4_HE_MU_EXT_CH2_CTR_RU), + IEEE80211_RADIOTAP_HE_MU_FLAGS2_CH2_CTR_26T_RU); + + he_mu->ru_ch2[0] = u32_get_bits(phy_data2, + IWL_RX_PHY_DATA2_HE_MU_EXT_CH2_RU0); + he_mu->ru_ch2[1] = u32_get_bits(phy_data3, + IWL_RX_PHY_DATA3_HE_MU_EXT_CH2_RU1); + he_mu->ru_ch2[2] = u32_get_bits(phy_data2, + IWL_RX_PHY_DATA2_HE_MU_EXT_CH2_RU2); + he_mu->ru_ch2[3] = u32_get_bits(phy_data3, + IWL_RX_PHY_DATA3_HE_MU_EXT_CH2_RU3); + } +} + +static void +iwl_mld_decode_he_phy_data(struct iwl_mld_rx_phy_data *phy_data, + struct ieee80211_radiotap_he *he, + struct ieee80211_radiotap_he_mu *he_mu, + struct ieee80211_rx_status *rx_status, + int queue) +{ + switch (phy_data->info_type) { + case IWL_RX_PHY_INFO_TYPE_NONE: + case IWL_RX_PHY_INFO_TYPE_CCK: + case IWL_RX_PHY_INFO_TYPE_OFDM_LGCY: + case IWL_RX_PHY_INFO_TYPE_HT: + case IWL_RX_PHY_INFO_TYPE_VHT_SU: + case IWL_RX_PHY_INFO_TYPE_VHT_MU: + case IWL_RX_PHY_INFO_TYPE_EHT_MU: + case IWL_RX_PHY_INFO_TYPE_EHT_TB: + case IWL_RX_PHY_INFO_TYPE_EHT_MU_EXT: + case IWL_RX_PHY_INFO_TYPE_EHT_TB_EXT: + return; + case IWL_RX_PHY_INFO_TYPE_HE_TB_EXT: + he->data1 |= cpu_to_le16(IEEE80211_RADIOTAP_HE_DATA1_SPTL_REUSE_KNOWN | + IEEE80211_RADIOTAP_HE_DATA1_SPTL_REUSE2_KNOWN | + IEEE80211_RADIOTAP_HE_DATA1_SPTL_REUSE3_KNOWN | + IEEE80211_RADIOTAP_HE_DATA1_SPTL_REUSE4_KNOWN); + he->data4 |= le16_encode_bits(le32_get_bits(phy_data->data2, + IWL_RX_PHY_DATA2_HE_TB_EXT_SPTL_REUSE1), + IEEE80211_RADIOTAP_HE_DATA4_TB_SPTL_REUSE1); + he->data4 |= le16_encode_bits(le32_get_bits(phy_data->data2, + IWL_RX_PHY_DATA2_HE_TB_EXT_SPTL_REUSE2), + IEEE80211_RADIOTAP_HE_DATA4_TB_SPTL_REUSE2); + he->data4 |= le16_encode_bits(le32_get_bits(phy_data->data2, + IWL_RX_PHY_DATA2_HE_TB_EXT_SPTL_REUSE3), + IEEE80211_RADIOTAP_HE_DATA4_TB_SPTL_REUSE3); + he->data4 |= le16_encode_bits(le32_get_bits(phy_data->data2, + IWL_RX_PHY_DATA2_HE_TB_EXT_SPTL_REUSE4), + IEEE80211_RADIOTAP_HE_DATA4_TB_SPTL_REUSE4); + fallthrough; + case IWL_RX_PHY_INFO_TYPE_HE_SU: + case IWL_RX_PHY_INFO_TYPE_HE_MU: + case IWL_RX_PHY_INFO_TYPE_HE_MU_EXT: + case IWL_RX_PHY_INFO_TYPE_HE_TB: + /* HE common */ + he->data1 |= cpu_to_le16(IEEE80211_RADIOTAP_HE_DATA1_LDPC_XSYMSEG_KNOWN | + IEEE80211_RADIOTAP_HE_DATA1_DOPPLER_KNOWN | + IEEE80211_RADIOTAP_HE_DATA1_BSS_COLOR_KNOWN); + he->data2 |= cpu_to_le16(IEEE80211_RADIOTAP_HE_DATA2_PRE_FEC_PAD_KNOWN | + IEEE80211_RADIOTAP_HE_DATA2_PE_DISAMBIG_KNOWN | + IEEE80211_RADIOTAP_HE_DATA2_TXOP_KNOWN | + IEEE80211_RADIOTAP_HE_DATA2_NUM_LTF_SYMS_KNOWN); + he->data3 |= le16_encode_bits(le32_get_bits(phy_data->data0, + IWL_RX_PHY_DATA0_HE_BSS_COLOR_MASK), + IEEE80211_RADIOTAP_HE_DATA3_BSS_COLOR); + if (phy_data->info_type != IWL_RX_PHY_INFO_TYPE_HE_TB && + phy_data->info_type != IWL_RX_PHY_INFO_TYPE_HE_TB_EXT) { + he->data1 |= cpu_to_le16(IEEE80211_RADIOTAP_HE_DATA1_UL_DL_KNOWN); + he->data3 |= le16_encode_bits(le32_get_bits(phy_data->data0, + IWL_RX_PHY_DATA0_HE_UPLINK), + IEEE80211_RADIOTAP_HE_DATA3_UL_DL); + } + he->data3 |= le16_encode_bits(le32_get_bits(phy_data->data0, + IWL_RX_PHY_DATA0_HE_LDPC_EXT_SYM), + IEEE80211_RADIOTAP_HE_DATA3_LDPC_XSYMSEG); + he->data5 |= le16_encode_bits(le32_get_bits(phy_data->data0, + IWL_RX_PHY_DATA0_HE_PRE_FEC_PAD_MASK), + IEEE80211_RADIOTAP_HE_DATA5_PRE_FEC_PAD); + he->data5 |= le16_encode_bits(le32_get_bits(phy_data->data0, + IWL_RX_PHY_DATA0_HE_PE_DISAMBIG), + IEEE80211_RADIOTAP_HE_DATA5_PE_DISAMBIG); + he->data5 |= le16_encode_bits(le32_get_bits(phy_data->data1, + IWL_RX_PHY_DATA1_HE_LTF_NUM_MASK), + IEEE80211_RADIOTAP_HE_DATA5_NUM_LTF_SYMS); + he->data6 |= le16_encode_bits(le32_get_bits(phy_data->data0, + IWL_RX_PHY_DATA0_HE_TXOP_DUR_MASK), + IEEE80211_RADIOTAP_HE_DATA6_TXOP); + he->data6 |= le16_encode_bits(le32_get_bits(phy_data->data0, + IWL_RX_PHY_DATA0_HE_DOPPLER), + IEEE80211_RADIOTAP_HE_DATA6_DOPPLER); + break; + } + + switch (phy_data->info_type) { + case IWL_RX_PHY_INFO_TYPE_HE_MU_EXT: + case IWL_RX_PHY_INFO_TYPE_HE_MU: + case IWL_RX_PHY_INFO_TYPE_HE_SU: + he->data1 |= cpu_to_le16(IEEE80211_RADIOTAP_HE_DATA1_SPTL_REUSE_KNOWN); + he->data4 |= le16_encode_bits(le32_get_bits(phy_data->data0, + IWL_RX_PHY_DATA0_HE_SPATIAL_REUSE_MASK), + IEEE80211_RADIOTAP_HE_DATA4_SU_MU_SPTL_REUSE); + break; + default: + /* nothing here */ + break; + } + + switch (phy_data->info_type) { + case IWL_RX_PHY_INFO_TYPE_HE_MU_EXT: + he_mu->flags1 |= + le16_encode_bits(le16_get_bits(phy_data->data4, + IWL_RX_PHY_DATA4_HE_MU_EXT_SIGB_DCM), + IEEE80211_RADIOTAP_HE_MU_FLAGS1_SIG_B_DCM); + he_mu->flags1 |= + le16_encode_bits(le16_get_bits(phy_data->data4, + IWL_RX_PHY_DATA4_HE_MU_EXT_SIGB_MCS_MASK), + IEEE80211_RADIOTAP_HE_MU_FLAGS1_SIG_B_MCS); + he_mu->flags2 |= + le16_encode_bits(le16_get_bits(phy_data->data4, + IWL_RX_PHY_DATA4_HE_MU_EXT_PREAMBLE_PUNC_TYPE_MASK), + IEEE80211_RADIOTAP_HE_MU_FLAGS2_PUNC_FROM_SIG_A_BW); + iwl_mld_decode_he_mu_ext(phy_data, he_mu); + fallthrough; + case IWL_RX_PHY_INFO_TYPE_HE_MU: + he_mu->flags2 |= + le16_encode_bits(le32_get_bits(phy_data->data1, + IWL_RX_PHY_DATA1_HE_MU_SIBG_SYM_OR_USER_NUM_MASK), + IEEE80211_RADIOTAP_HE_MU_FLAGS2_SIG_B_SYMS_USERS); + he_mu->flags2 |= + le16_encode_bits(le32_get_bits(phy_data->data1, + IWL_RX_PHY_DATA1_HE_MU_SIGB_COMPRESSION), + IEEE80211_RADIOTAP_HE_MU_FLAGS2_SIG_B_COMP); + fallthrough; + case IWL_RX_PHY_INFO_TYPE_HE_TB: + case IWL_RX_PHY_INFO_TYPE_HE_TB_EXT: + iwl_mld_decode_he_phy_ru_alloc(phy_data, he, he_mu, rx_status); + break; + case IWL_RX_PHY_INFO_TYPE_HE_SU: + he->data1 |= cpu_to_le16(IEEE80211_RADIOTAP_HE_DATA1_BEAM_CHANGE_KNOWN); + he->data3 |= le16_encode_bits(le32_get_bits(phy_data->data0, + IWL_RX_PHY_DATA0_HE_BEAM_CHNG), + IEEE80211_RADIOTAP_HE_DATA3_BEAM_CHANGE); + break; + default: + /* nothing */ + break; + } +} + +static void iwl_mld_rx_he(struct iwl_mld *mld, struct sk_buff *skb, + struct iwl_mld_rx_phy_data *phy_data, + int queue) +{ + struct ieee80211_rx_status *rx_status = IEEE80211_SKB_RXCB(skb); + struct ieee80211_radiotap_he *he = NULL; + struct ieee80211_radiotap_he_mu *he_mu = NULL; + u32 rate_n_flags = phy_data->rate_n_flags; + u32 he_type = rate_n_flags & RATE_MCS_HE_TYPE_MSK; + u8 ltf; + static const struct ieee80211_radiotap_he known = { + .data1 = cpu_to_le16(IEEE80211_RADIOTAP_HE_DATA1_DATA_MCS_KNOWN | + IEEE80211_RADIOTAP_HE_DATA1_DATA_DCM_KNOWN | + IEEE80211_RADIOTAP_HE_DATA1_STBC_KNOWN | + IEEE80211_RADIOTAP_HE_DATA1_CODING_KNOWN), + .data2 = cpu_to_le16(IEEE80211_RADIOTAP_HE_DATA2_GI_KNOWN | + IEEE80211_RADIOTAP_HE_DATA2_TXBF_KNOWN), + }; + static const struct ieee80211_radiotap_he_mu mu_known = { + .flags1 = cpu_to_le16(IEEE80211_RADIOTAP_HE_MU_FLAGS1_SIG_B_MCS_KNOWN | + IEEE80211_RADIOTAP_HE_MU_FLAGS1_SIG_B_DCM_KNOWN | + IEEE80211_RADIOTAP_HE_MU_FLAGS1_SIG_B_SYMS_USERS_KNOWN | + IEEE80211_RADIOTAP_HE_MU_FLAGS1_SIG_B_COMP_KNOWN), + .flags2 = cpu_to_le16(IEEE80211_RADIOTAP_HE_MU_FLAGS2_PUNC_FROM_SIG_A_BW_KNOWN | + IEEE80211_RADIOTAP_HE_MU_FLAGS2_BW_FROM_SIG_A_BW_KNOWN), + }; + u16 phy_info = phy_data->phy_info; + + he = skb_put_data(skb, &known, sizeof(known)); + rx_status->flag |= RX_FLAG_RADIOTAP_HE; + + if (phy_data->info_type == IWL_RX_PHY_INFO_TYPE_HE_MU || + phy_data->info_type == IWL_RX_PHY_INFO_TYPE_HE_MU_EXT) { + he_mu = skb_put_data(skb, &mu_known, sizeof(mu_known)); + rx_status->flag |= RX_FLAG_RADIOTAP_HE_MU; + } + + /* report the AMPDU-EOF bit on single frames */ + if (!queue && !(phy_info & IWL_RX_MPDU_PHY_AMPDU)) { + rx_status->flag |= RX_FLAG_AMPDU_DETAILS; + rx_status->flag |= RX_FLAG_AMPDU_EOF_BIT_KNOWN; + if (phy_data->data0 & cpu_to_le32(IWL_RX_PHY_DATA0_HE_DELIM_EOF)) + rx_status->flag |= RX_FLAG_AMPDU_EOF_BIT; + } + + if (phy_info & IWL_RX_MPDU_PHY_TSF_OVERLOAD) + iwl_mld_decode_he_phy_data(phy_data, he, he_mu, rx_status, + queue); + + /* update aggregation data for monitor sake on default queue */ + if (!queue && (phy_info & IWL_RX_MPDU_PHY_TSF_OVERLOAD) && + (phy_info & IWL_RX_MPDU_PHY_AMPDU) && phy_data->first_subframe) { + rx_status->flag |= RX_FLAG_AMPDU_EOF_BIT_KNOWN; + if (phy_data->data0 & cpu_to_le32(IWL_RX_PHY_DATA0_EHT_DELIM_EOF)) + rx_status->flag |= RX_FLAG_AMPDU_EOF_BIT; + } + + if (he_type == RATE_MCS_HE_TYPE_EXT_SU && + rate_n_flags & RATE_MCS_HE_106T_MSK) { + rx_status->bw = RATE_INFO_BW_HE_RU; + rx_status->he_ru = NL80211_RATE_INFO_HE_RU_ALLOC_106; + } + + /* actually data is filled in mac80211 */ + if (he_type == RATE_MCS_HE_TYPE_SU || + he_type == RATE_MCS_HE_TYPE_EXT_SU) + he->data1 |= + cpu_to_le16(IEEE80211_RADIOTAP_HE_DATA1_BW_RU_ALLOC_KNOWN); + +#define CHECK_TYPE(F) \ + BUILD_BUG_ON(IEEE80211_RADIOTAP_HE_DATA1_FORMAT_ ## F != \ + (RATE_MCS_HE_TYPE_ ## F >> RATE_MCS_HE_TYPE_POS)) + + CHECK_TYPE(SU); + CHECK_TYPE(EXT_SU); + CHECK_TYPE(MU); + CHECK_TYPE(TRIG); + + he->data1 |= cpu_to_le16(he_type >> RATE_MCS_HE_TYPE_POS); + + if (rate_n_flags & RATE_MCS_BF_MSK) + he->data5 |= cpu_to_le16(IEEE80211_RADIOTAP_HE_DATA5_TXBF); + + switch ((rate_n_flags & RATE_MCS_HE_GI_LTF_MSK) >> + RATE_MCS_HE_GI_LTF_POS) { + case 0: + if (he_type == RATE_MCS_HE_TYPE_TRIG) + rx_status->he_gi = NL80211_RATE_INFO_HE_GI_1_6; + else + rx_status->he_gi = NL80211_RATE_INFO_HE_GI_0_8; + if (he_type == RATE_MCS_HE_TYPE_MU) + ltf = IEEE80211_RADIOTAP_HE_DATA5_LTF_SIZE_4X; + else + ltf = IEEE80211_RADIOTAP_HE_DATA5_LTF_SIZE_1X; + break; + case 1: + if (he_type == RATE_MCS_HE_TYPE_TRIG) + rx_status->he_gi = NL80211_RATE_INFO_HE_GI_1_6; + else + rx_status->he_gi = NL80211_RATE_INFO_HE_GI_0_8; + ltf = IEEE80211_RADIOTAP_HE_DATA5_LTF_SIZE_2X; + break; + case 2: + if (he_type == RATE_MCS_HE_TYPE_TRIG) { + rx_status->he_gi = NL80211_RATE_INFO_HE_GI_3_2; + ltf = IEEE80211_RADIOTAP_HE_DATA5_LTF_SIZE_4X; + } else { + rx_status->he_gi = NL80211_RATE_INFO_HE_GI_1_6; + ltf = IEEE80211_RADIOTAP_HE_DATA5_LTF_SIZE_2X; + } + break; + case 3: + rx_status->he_gi = NL80211_RATE_INFO_HE_GI_3_2; + ltf = IEEE80211_RADIOTAP_HE_DATA5_LTF_SIZE_4X; + break; + case 4: + rx_status->he_gi = NL80211_RATE_INFO_HE_GI_0_8; + ltf = IEEE80211_RADIOTAP_HE_DATA5_LTF_SIZE_4X; + break; + default: + ltf = IEEE80211_RADIOTAP_HE_DATA5_LTF_SIZE_UNKNOWN; + } + + he->data5 |= le16_encode_bits(ltf, + IEEE80211_RADIOTAP_HE_DATA5_LTF_SIZE); +} + +static void iwl_mld_decode_lsig(struct sk_buff *skb, + struct iwl_mld_rx_phy_data *phy_data) +{ + struct ieee80211_rx_status *rx_status = IEEE80211_SKB_RXCB(skb); + struct ieee80211_radiotap_lsig *lsig; + + switch (phy_data->info_type) { + case IWL_RX_PHY_INFO_TYPE_HT: + case IWL_RX_PHY_INFO_TYPE_VHT_SU: + case IWL_RX_PHY_INFO_TYPE_VHT_MU: + case IWL_RX_PHY_INFO_TYPE_HE_TB_EXT: + case IWL_RX_PHY_INFO_TYPE_HE_SU: + case IWL_RX_PHY_INFO_TYPE_HE_MU: + case IWL_RX_PHY_INFO_TYPE_HE_MU_EXT: + case IWL_RX_PHY_INFO_TYPE_HE_TB: + case IWL_RX_PHY_INFO_TYPE_EHT_MU: + case IWL_RX_PHY_INFO_TYPE_EHT_TB: + case IWL_RX_PHY_INFO_TYPE_EHT_MU_EXT: + case IWL_RX_PHY_INFO_TYPE_EHT_TB_EXT: + lsig = skb_put(skb, sizeof(*lsig)); + lsig->data1 = cpu_to_le16(IEEE80211_RADIOTAP_LSIG_DATA1_LENGTH_KNOWN); + lsig->data2 = le16_encode_bits(le32_get_bits(phy_data->data1, + IWL_RX_PHY_DATA1_LSIG_LEN_MASK), + IEEE80211_RADIOTAP_LSIG_DATA2_LENGTH); + rx_status->flag |= RX_FLAG_RADIOTAP_LSIG; + break; + default: + break; + } +} + +/* Put a TLV on the skb and return data pointer + * + * Also pad the len to 4 and zero out all data part + */ +static void * +iwl_mld_radiotap_put_tlv(struct sk_buff *skb, u16 type, u16 len) +{ + struct ieee80211_radiotap_tlv *tlv; + + tlv = skb_put(skb, sizeof(*tlv)); + tlv->type = cpu_to_le16(type); + tlv->len = cpu_to_le16(len); + return skb_put_zero(skb, ALIGN(len, 4)); +} + +#define LE32_DEC_ENC(value, dec_bits, enc_bits) \ + le32_encode_bits(le32_get_bits(value, dec_bits), enc_bits) + +#define IWL_MLD_ENC_USIG_VALUE_MASK(usig, in_value, dec_bits, enc_bits) do { \ + typeof(enc_bits) _enc_bits = enc_bits; \ + typeof(usig) _usig = usig; \ + (_usig)->mask |= cpu_to_le32(_enc_bits); \ + (_usig)->value |= LE32_DEC_ENC(in_value, dec_bits, _enc_bits); \ +} while (0) + +#define __IWL_MLD_ENC_EHT_RU(rt_data, rt_ru, fw_data, fw_ru) \ + eht->data[(rt_data)] |= \ + (cpu_to_le32 \ + (IEEE80211_RADIOTAP_EHT_DATA ## rt_data ## _RU_ALLOC_CC_ ## rt_ru ## _KNOWN) | \ + LE32_DEC_ENC(data ## fw_data, \ + IWL_RX_PHY_DATA ## fw_data ## _EHT_MU_EXT_RU_ALLOC_ ## fw_ru, \ + IEEE80211_RADIOTAP_EHT_DATA ## rt_data ## _RU_ALLOC_CC_ ## rt_ru)) + +#define _IWL_MLD_ENC_EHT_RU(rt_data, rt_ru, fw_data, fw_ru) \ + __IWL_MLD_ENC_EHT_RU(rt_data, rt_ru, fw_data, fw_ru) + +#define IEEE80211_RADIOTAP_RU_DATA_1_1_1 1 +#define IEEE80211_RADIOTAP_RU_DATA_2_1_1 2 +#define IEEE80211_RADIOTAP_RU_DATA_1_1_2 2 +#define IEEE80211_RADIOTAP_RU_DATA_2_1_2 2 +#define IEEE80211_RADIOTAP_RU_DATA_1_2_1 3 +#define IEEE80211_RADIOTAP_RU_DATA_2_2_1 3 +#define IEEE80211_RADIOTAP_RU_DATA_1_2_2 3 +#define IEEE80211_RADIOTAP_RU_DATA_2_2_2 4 + +#define IWL_RX_RU_DATA_A1 2 +#define IWL_RX_RU_DATA_A2 2 +#define IWL_RX_RU_DATA_B1 2 +#define IWL_RX_RU_DATA_B2 4 +#define IWL_RX_RU_DATA_C1 3 +#define IWL_RX_RU_DATA_C2 3 +#define IWL_RX_RU_DATA_D1 4 +#define IWL_RX_RU_DATA_D2 4 + +#define IWL_MLD_ENC_EHT_RU(rt_ru, fw_ru) \ + _IWL_MLD_ENC_EHT_RU(IEEE80211_RADIOTAP_RU_DATA_ ## rt_ru, \ + rt_ru, \ + IWL_RX_RU_DATA_ ## fw_ru, \ + fw_ru) + +static void iwl_mld_decode_eht_ext_mu(struct iwl_mld *mld, + struct iwl_mld_rx_phy_data *phy_data, + struct ieee80211_rx_status *rx_status, + struct ieee80211_radiotap_eht *eht, + struct ieee80211_radiotap_eht_usig *usig) +{ + if (phy_data->with_data) { + __le32 data1 = phy_data->data1; + __le32 data2 = phy_data->data2; + __le32 data3 = phy_data->data3; + __le32 data4 = phy_data->eht_data4; + __le32 data5 = phy_data->data5; + u32 phy_bw = phy_data->rate_n_flags & RATE_MCS_CHAN_WIDTH_MSK; + + IWL_MLD_ENC_USIG_VALUE_MASK(usig, data5, + IWL_RX_PHY_DATA5_EHT_TYPE_AND_COMP, + IEEE80211_RADIOTAP_EHT_USIG2_MU_B0_B1_PPDU_TYPE); + IWL_MLD_ENC_USIG_VALUE_MASK(usig, data5, + IWL_RX_PHY_DATA5_EHT_MU_PUNC_CH_CODE, + IEEE80211_RADIOTAP_EHT_USIG2_MU_B3_B7_PUNCTURED_INFO); + IWL_MLD_ENC_USIG_VALUE_MASK(usig, data4, + IWL_RX_PHY_DATA4_EHT_MU_EXT_SIGB_MCS, + IEEE80211_RADIOTAP_EHT_USIG2_MU_B9_B10_SIG_MCS); + IWL_MLD_ENC_USIG_VALUE_MASK + (usig, data1, IWL_RX_PHY_DATA1_EHT_MU_NUM_SIG_SYM_USIGA2, + IEEE80211_RADIOTAP_EHT_USIG2_MU_B11_B15_EHT_SIG_SYMBOLS); + + eht->user_info[0] |= + cpu_to_le32(IEEE80211_RADIOTAP_EHT_USER_INFO_STA_ID_KNOWN) | + LE32_DEC_ENC(data5, IWL_RX_PHY_DATA5_EHT_MU_STA_ID_USR, + IEEE80211_RADIOTAP_EHT_USER_INFO_STA_ID); + + eht->known |= cpu_to_le32(IEEE80211_RADIOTAP_EHT_KNOWN_NR_NON_OFDMA_USERS_M); + eht->data[7] |= LE32_DEC_ENC + (data5, IWL_RX_PHY_DATA5_EHT_MU_NUM_USR_NON_OFDMA, + IEEE80211_RADIOTAP_EHT_DATA7_NUM_OF_NON_OFDMA_USERS); + + /* + * Hardware labels the content channels/RU allocation values + * as follows: + * Content Channel 1 Content Channel 2 + * 20 MHz: A1 + * 40 MHz: A1 B1 + * 80 MHz: A1 C1 B1 D1 + * 160 MHz: A1 C1 A2 C2 B1 D1 B2 D2 + * 320 MHz: A1 C1 A2 C2 A3 C3 A4 C4 B1 D1 B2 D2 B3 D3 B4 D4 + * + * However firmware can only give us A1-D2, so the higher + * frequencies are missing. + */ + + switch (phy_bw) { + case RATE_MCS_CHAN_WIDTH_320: + /* additional values are missing in RX metadata */ + fallthrough; + case RATE_MCS_CHAN_WIDTH_160: + /* content channel 1 */ + IWL_MLD_ENC_EHT_RU(1_2_1, A2); + IWL_MLD_ENC_EHT_RU(1_2_2, C2); + /* content channel 2 */ + IWL_MLD_ENC_EHT_RU(2_2_1, B2); + IWL_MLD_ENC_EHT_RU(2_2_2, D2); + fallthrough; + case RATE_MCS_CHAN_WIDTH_80: + /* content channel 1 */ + IWL_MLD_ENC_EHT_RU(1_1_2, C1); + /* content channel 2 */ + IWL_MLD_ENC_EHT_RU(2_1_2, D1); + fallthrough; + case RATE_MCS_CHAN_WIDTH_40: + /* content channel 2 */ + IWL_MLD_ENC_EHT_RU(2_1_1, B1); + fallthrough; + case RATE_MCS_CHAN_WIDTH_20: + IWL_MLD_ENC_EHT_RU(1_1_1, A1); + break; + } + } else { + __le32 usig_a1 = phy_data->rx_vec[0]; + __le32 usig_a2 = phy_data->rx_vec[1]; + + IWL_MLD_ENC_USIG_VALUE_MASK(usig, usig_a1, + IWL_RX_USIG_A1_DISREGARD, + IEEE80211_RADIOTAP_EHT_USIG1_MU_B20_B24_DISREGARD); + IWL_MLD_ENC_USIG_VALUE_MASK(usig, usig_a1, + IWL_RX_USIG_A1_VALIDATE, + IEEE80211_RADIOTAP_EHT_USIG1_MU_B25_VALIDATE); + IWL_MLD_ENC_USIG_VALUE_MASK(usig, usig_a2, + IWL_RX_USIG_A2_EHT_PPDU_TYPE, + IEEE80211_RADIOTAP_EHT_USIG2_MU_B0_B1_PPDU_TYPE); + IWL_MLD_ENC_USIG_VALUE_MASK(usig, usig_a2, + IWL_RX_USIG_A2_EHT_USIG2_VALIDATE_B2, + IEEE80211_RADIOTAP_EHT_USIG2_MU_B2_VALIDATE); + IWL_MLD_ENC_USIG_VALUE_MASK(usig, usig_a2, + IWL_RX_USIG_A2_EHT_PUNC_CHANNEL, + IEEE80211_RADIOTAP_EHT_USIG2_MU_B3_B7_PUNCTURED_INFO); + IWL_MLD_ENC_USIG_VALUE_MASK(usig, usig_a2, + IWL_RX_USIG_A2_EHT_USIG2_VALIDATE_B8, + IEEE80211_RADIOTAP_EHT_USIG2_MU_B8_VALIDATE); + IWL_MLD_ENC_USIG_VALUE_MASK(usig, usig_a2, + IWL_RX_USIG_A2_EHT_SIG_MCS, + IEEE80211_RADIOTAP_EHT_USIG2_MU_B9_B10_SIG_MCS); + IWL_MLD_ENC_USIG_VALUE_MASK + (usig, usig_a2, IWL_RX_USIG_A2_EHT_SIG_SYM_NUM, + IEEE80211_RADIOTAP_EHT_USIG2_MU_B11_B15_EHT_SIG_SYMBOLS); + IWL_MLD_ENC_USIG_VALUE_MASK(usig, usig_a2, + IWL_RX_USIG_A2_EHT_CRC_OK, + IEEE80211_RADIOTAP_EHT_USIG2_MU_B16_B19_CRC); + } +} + +static void iwl_mld_decode_eht_ext_tb(struct iwl_mld *mld, + struct iwl_mld_rx_phy_data *phy_data, + struct ieee80211_rx_status *rx_status, + struct ieee80211_radiotap_eht *eht, + struct ieee80211_radiotap_eht_usig *usig) +{ + if (phy_data->with_data) { + __le32 data5 = phy_data->data5; + + IWL_MLD_ENC_USIG_VALUE_MASK(usig, data5, + IWL_RX_PHY_DATA5_EHT_TYPE_AND_COMP, + IEEE80211_RADIOTAP_EHT_USIG2_TB_B0_B1_PPDU_TYPE); + IWL_MLD_ENC_USIG_VALUE_MASK(usig, data5, + IWL_RX_PHY_DATA5_EHT_TB_SPATIAL_REUSE1, + IEEE80211_RADIOTAP_EHT_USIG2_TB_B3_B6_SPATIAL_REUSE_1); + + IWL_MLD_ENC_USIG_VALUE_MASK(usig, data5, + IWL_RX_PHY_DATA5_EHT_TB_SPATIAL_REUSE2, + IEEE80211_RADIOTAP_EHT_USIG2_TB_B7_B10_SPATIAL_REUSE_2); + } else { + __le32 usig_a1 = phy_data->rx_vec[0]; + __le32 usig_a2 = phy_data->rx_vec[1]; + + IWL_MLD_ENC_USIG_VALUE_MASK(usig, usig_a1, + IWL_RX_USIG_A1_DISREGARD, + IEEE80211_RADIOTAP_EHT_USIG1_TB_B20_B25_DISREGARD); + IWL_MLD_ENC_USIG_VALUE_MASK(usig, usig_a2, + IWL_RX_USIG_A2_EHT_PPDU_TYPE, + IEEE80211_RADIOTAP_EHT_USIG2_TB_B0_B1_PPDU_TYPE); + IWL_MLD_ENC_USIG_VALUE_MASK(usig, usig_a2, + IWL_RX_USIG_A2_EHT_USIG2_VALIDATE_B2, + IEEE80211_RADIOTAP_EHT_USIG2_TB_B2_VALIDATE); + IWL_MLD_ENC_USIG_VALUE_MASK(usig, usig_a2, + IWL_RX_USIG_A2_EHT_TRIG_SPATIAL_REUSE_1, + IEEE80211_RADIOTAP_EHT_USIG2_TB_B3_B6_SPATIAL_REUSE_1); + IWL_MLD_ENC_USIG_VALUE_MASK(usig, usig_a2, + IWL_RX_USIG_A2_EHT_TRIG_SPATIAL_REUSE_2, + IEEE80211_RADIOTAP_EHT_USIG2_TB_B7_B10_SPATIAL_REUSE_2); + IWL_MLD_ENC_USIG_VALUE_MASK(usig, usig_a2, + IWL_RX_USIG_A2_EHT_TRIG_USIG2_DISREGARD, + IEEE80211_RADIOTAP_EHT_USIG2_TB_B11_B15_DISREGARD); + IWL_MLD_ENC_USIG_VALUE_MASK(usig, usig_a2, + IWL_RX_USIG_A2_EHT_CRC_OK, + IEEE80211_RADIOTAP_EHT_USIG2_TB_B16_B19_CRC); + } +} + +static void iwl_mld_decode_eht_ru(struct iwl_mld *mld, + struct ieee80211_rx_status *rx_status, + struct ieee80211_radiotap_eht *eht) +{ + u32 ru = le32_get_bits(eht->data[8], + IEEE80211_RADIOTAP_EHT_DATA8_RU_ALLOC_TB_FMT_B7_B1); + enum nl80211_eht_ru_alloc nl_ru; + + /* Using D1.5 Table 9-53a - Encoding of PS160 and RU Allocation subfields + * in an EHT variant User Info field + */ + + switch (ru) { + case 0 ... 36: + nl_ru = NL80211_RATE_INFO_EHT_RU_ALLOC_26; + break; + case 37 ... 52: + nl_ru = NL80211_RATE_INFO_EHT_RU_ALLOC_52; + break; + case 53 ... 60: + nl_ru = NL80211_RATE_INFO_EHT_RU_ALLOC_106; + break; + case 61 ... 64: + nl_ru = NL80211_RATE_INFO_EHT_RU_ALLOC_242; + break; + case 65 ... 66: + nl_ru = NL80211_RATE_INFO_EHT_RU_ALLOC_484; + break; + case 67: + nl_ru = NL80211_RATE_INFO_EHT_RU_ALLOC_996; + break; + case 68: + nl_ru = NL80211_RATE_INFO_EHT_RU_ALLOC_2x996; + break; + case 69: + nl_ru = NL80211_RATE_INFO_EHT_RU_ALLOC_4x996; + break; + case 70 ... 81: + nl_ru = NL80211_RATE_INFO_EHT_RU_ALLOC_52P26; + break; + case 82 ... 89: + nl_ru = NL80211_RATE_INFO_EHT_RU_ALLOC_106P26; + break; + case 90 ... 93: + nl_ru = NL80211_RATE_INFO_EHT_RU_ALLOC_484P242; + break; + case 94 ... 95: + nl_ru = NL80211_RATE_INFO_EHT_RU_ALLOC_996P484; + break; + case 96 ... 99: + nl_ru = NL80211_RATE_INFO_EHT_RU_ALLOC_996P484P242; + break; + case 100 ... 103: + nl_ru = NL80211_RATE_INFO_EHT_RU_ALLOC_2x996P484; + break; + case 104: + nl_ru = NL80211_RATE_INFO_EHT_RU_ALLOC_3x996; + break; + case 105 ... 106: + nl_ru = NL80211_RATE_INFO_EHT_RU_ALLOC_3x996P484; + break; + default: + return; + } + + rx_status->bw = RATE_INFO_BW_EHT_RU; + rx_status->eht.ru = nl_ru; +} + +static void iwl_mld_decode_eht_phy_data(struct iwl_mld *mld, + struct iwl_mld_rx_phy_data *phy_data, + struct ieee80211_rx_status *rx_status, + struct ieee80211_radiotap_eht *eht, + struct ieee80211_radiotap_eht_usig *usig) + +{ + __le32 data0 = phy_data->data0; + __le32 data1 = phy_data->data1; + __le32 usig_a1 = phy_data->rx_vec[0]; + u8 info_type = phy_data->info_type; + + /* Not in EHT range */ + if (info_type < IWL_RX_PHY_INFO_TYPE_EHT_MU || + info_type > IWL_RX_PHY_INFO_TYPE_EHT_TB_EXT) + return; + + usig->common |= cpu_to_le32 + (IEEE80211_RADIOTAP_EHT_USIG_COMMON_UL_DL_KNOWN | + IEEE80211_RADIOTAP_EHT_USIG_COMMON_BSS_COLOR_KNOWN); + if (phy_data->with_data) { + usig->common |= LE32_DEC_ENC(data0, + IWL_RX_PHY_DATA0_EHT_UPLINK, + IEEE80211_RADIOTAP_EHT_USIG_COMMON_UL_DL); + usig->common |= LE32_DEC_ENC(data0, + IWL_RX_PHY_DATA0_EHT_BSS_COLOR_MASK, + IEEE80211_RADIOTAP_EHT_USIG_COMMON_BSS_COLOR); + } else { + usig->common |= LE32_DEC_ENC(usig_a1, + IWL_RX_USIG_A1_UL_FLAG, + IEEE80211_RADIOTAP_EHT_USIG_COMMON_UL_DL); + usig->common |= LE32_DEC_ENC(usig_a1, + IWL_RX_USIG_A1_BSS_COLOR, + IEEE80211_RADIOTAP_EHT_USIG_COMMON_BSS_COLOR); + } + + usig->common |= + cpu_to_le32(IEEE80211_RADIOTAP_EHT_USIG_COMMON_VALIDATE_BITS_CHECKED); + usig->common |= + LE32_DEC_ENC(data0, IWL_RX_PHY_DATA0_EHT_VALIDATE, + IEEE80211_RADIOTAP_EHT_USIG_COMMON_VALIDATE_BITS_OK); + + eht->known |= cpu_to_le32(IEEE80211_RADIOTAP_EHT_KNOWN_SPATIAL_REUSE); + eht->data[0] |= LE32_DEC_ENC(data0, + IWL_RX_PHY_DATA0_ETH_SPATIAL_REUSE_MASK, + IEEE80211_RADIOTAP_EHT_DATA0_SPATIAL_REUSE); + + /* All RU allocating size/index is in TB format */ + eht->known |= cpu_to_le32(IEEE80211_RADIOTAP_EHT_KNOWN_RU_ALLOC_TB_FMT); + eht->data[8] |= LE32_DEC_ENC(data0, IWL_RX_PHY_DATA0_EHT_PS160, + IEEE80211_RADIOTAP_EHT_DATA8_RU_ALLOC_TB_FMT_PS_160); + eht->data[8] |= LE32_DEC_ENC(data1, IWL_RX_PHY_DATA1_EHT_RU_ALLOC_B0, + IEEE80211_RADIOTAP_EHT_DATA8_RU_ALLOC_TB_FMT_B0); + eht->data[8] |= LE32_DEC_ENC(data1, IWL_RX_PHY_DATA1_EHT_RU_ALLOC_B1_B7, + IEEE80211_RADIOTAP_EHT_DATA8_RU_ALLOC_TB_FMT_B7_B1); + + iwl_mld_decode_eht_ru(mld, rx_status, eht); + + /* We only get here in case of IWL_RX_MPDU_PHY_TSF_OVERLOAD is set + * which is on only in case of monitor mode so no need to check monitor + * mode + */ + eht->known |= cpu_to_le32(IEEE80211_RADIOTAP_EHT_KNOWN_PRIMARY_80); + eht->data[1] |= + le32_encode_bits(mld->monitor.p80, + IEEE80211_RADIOTAP_EHT_DATA1_PRIMARY_80); + + usig->common |= cpu_to_le32(IEEE80211_RADIOTAP_EHT_USIG_COMMON_TXOP_KNOWN); + if (phy_data->with_data) + usig->common |= LE32_DEC_ENC(data0, IWL_RX_PHY_DATA0_EHT_TXOP_DUR_MASK, + IEEE80211_RADIOTAP_EHT_USIG_COMMON_TXOP); + else + usig->common |= LE32_DEC_ENC(usig_a1, IWL_RX_USIG_A1_TXOP_DURATION, + IEEE80211_RADIOTAP_EHT_USIG_COMMON_TXOP); + + eht->known |= cpu_to_le32(IEEE80211_RADIOTAP_EHT_KNOWN_LDPC_EXTRA_SYM_OM); + eht->data[0] |= LE32_DEC_ENC(data0, IWL_RX_PHY_DATA0_EHT_LDPC_EXT_SYM, + IEEE80211_RADIOTAP_EHT_DATA0_LDPC_EXTRA_SYM_OM); + + eht->known |= cpu_to_le32(IEEE80211_RADIOTAP_EHT_KNOWN_PRE_PADD_FACOR_OM); + eht->data[0] |= LE32_DEC_ENC(data0, IWL_RX_PHY_DATA0_EHT_PRE_FEC_PAD_MASK, + IEEE80211_RADIOTAP_EHT_DATA0_PRE_PADD_FACOR_OM); + + eht->known |= cpu_to_le32(IEEE80211_RADIOTAP_EHT_KNOWN_PE_DISAMBIGUITY_OM); + eht->data[0] |= LE32_DEC_ENC(data0, IWL_RX_PHY_DATA0_EHT_PE_DISAMBIG, + IEEE80211_RADIOTAP_EHT_DATA0_PE_DISAMBIGUITY_OM); + + /* TODO: what about IWL_RX_PHY_DATA0_EHT_BW320_SLOT */ + + if (!le32_get_bits(data0, IWL_RX_PHY_DATA0_EHT_SIGA_CRC_OK)) + usig->common |= cpu_to_le32(IEEE80211_RADIOTAP_EHT_USIG_COMMON_BAD_USIG_CRC); + + usig->common |= cpu_to_le32(IEEE80211_RADIOTAP_EHT_USIG_COMMON_PHY_VER_KNOWN); + usig->common |= LE32_DEC_ENC(data0, IWL_RX_PHY_DATA0_EHT_PHY_VER, + IEEE80211_RADIOTAP_EHT_USIG_COMMON_PHY_VER); + + /* + * TODO: what about TB - IWL_RX_PHY_DATA1_EHT_TB_PILOT_TYPE, + * IWL_RX_PHY_DATA1_EHT_TB_LOW_SS + */ + + eht->known |= cpu_to_le32(IEEE80211_RADIOTAP_EHT_KNOWN_EHT_LTF); + eht->data[0] |= LE32_DEC_ENC(data1, IWL_RX_PHY_DATA1_EHT_SIG_LTF_NUM, + IEEE80211_RADIOTAP_EHT_DATA0_EHT_LTF); + + if (info_type == IWL_RX_PHY_INFO_TYPE_EHT_TB_EXT || + info_type == IWL_RX_PHY_INFO_TYPE_EHT_TB) + iwl_mld_decode_eht_ext_tb(mld, phy_data, rx_status, eht, usig); + + if (info_type == IWL_RX_PHY_INFO_TYPE_EHT_MU_EXT || + info_type == IWL_RX_PHY_INFO_TYPE_EHT_MU) + iwl_mld_decode_eht_ext_mu(mld, phy_data, rx_status, eht, usig); +} + +static void iwl_mld_rx_eht(struct iwl_mld *mld, struct sk_buff *skb, + struct iwl_mld_rx_phy_data *phy_data, + int queue) +{ + struct ieee80211_rx_status *rx_status = IEEE80211_SKB_RXCB(skb); + struct ieee80211_radiotap_eht *eht; + struct ieee80211_radiotap_eht_usig *usig; + size_t eht_len = sizeof(*eht); + + u32 rate_n_flags = phy_data->rate_n_flags; + u32 he_type = rate_n_flags & RATE_MCS_HE_TYPE_MSK; + /* EHT and HE have the same values for LTF */ + u8 ltf = IEEE80211_RADIOTAP_HE_DATA5_LTF_SIZE_UNKNOWN; + u16 phy_info = phy_data->phy_info; + u32 bw; + + /* u32 for 1 user_info */ + if (phy_data->with_data) + eht_len += sizeof(u32); + + eht = iwl_mld_radiotap_put_tlv(skb, IEEE80211_RADIOTAP_EHT, eht_len); + + usig = iwl_mld_radiotap_put_tlv(skb, IEEE80211_RADIOTAP_EHT_USIG, + sizeof(*usig)); + rx_status->flag |= RX_FLAG_RADIOTAP_TLV_AT_END; + usig->common |= + cpu_to_le32(IEEE80211_RADIOTAP_EHT_USIG_COMMON_BW_KNOWN); + + /* specific handling for 320MHz */ + bw = u32_get_bits(rate_n_flags, RATE_MCS_CHAN_WIDTH_MSK); + if (bw == RATE_MCS_CHAN_WIDTH_320_VAL) + bw += le32_get_bits(phy_data->data0, + IWL_RX_PHY_DATA0_EHT_BW320_SLOT); + + usig->common |= cpu_to_le32 + (FIELD_PREP(IEEE80211_RADIOTAP_EHT_USIG_COMMON_BW, bw)); + + /* report the AMPDU-EOF bit on single frames */ + if (!queue && !(phy_info & IWL_RX_MPDU_PHY_AMPDU)) { + rx_status->flag |= RX_FLAG_AMPDU_DETAILS; + rx_status->flag |= RX_FLAG_AMPDU_EOF_BIT_KNOWN; + if (phy_data->data0 & + cpu_to_le32(IWL_RX_PHY_DATA0_EHT_DELIM_EOF)) + rx_status->flag |= RX_FLAG_AMPDU_EOF_BIT; + } + + if (phy_info & IWL_RX_MPDU_PHY_TSF_OVERLOAD) + iwl_mld_decode_eht_phy_data(mld, phy_data, rx_status, eht, usig); + +#define CHECK_TYPE(F) \ + BUILD_BUG_ON(IEEE80211_RADIOTAP_HE_DATA1_FORMAT_ ## F != \ + (RATE_MCS_HE_TYPE_ ## F >> RATE_MCS_HE_TYPE_POS)) + + CHECK_TYPE(SU); + CHECK_TYPE(EXT_SU); + CHECK_TYPE(MU); + CHECK_TYPE(TRIG); + + switch (u32_get_bits(rate_n_flags, RATE_MCS_HE_GI_LTF_MSK)) { + case 0: + if (he_type == RATE_MCS_HE_TYPE_TRIG) { + rx_status->eht.gi = NL80211_RATE_INFO_EHT_GI_1_6; + ltf = IEEE80211_RADIOTAP_HE_DATA5_LTF_SIZE_1X; + } else { + rx_status->eht.gi = NL80211_RATE_INFO_EHT_GI_0_8; + ltf = IEEE80211_RADIOTAP_HE_DATA5_LTF_SIZE_2X; + } + break; + case 1: + rx_status->eht.gi = NL80211_RATE_INFO_EHT_GI_1_6; + ltf = IEEE80211_RADIOTAP_HE_DATA5_LTF_SIZE_2X; + break; + case 2: + ltf = IEEE80211_RADIOTAP_HE_DATA5_LTF_SIZE_4X; + if (he_type == RATE_MCS_HE_TYPE_TRIG) + rx_status->eht.gi = NL80211_RATE_INFO_EHT_GI_3_2; + else + rx_status->eht.gi = NL80211_RATE_INFO_EHT_GI_0_8; + break; + case 3: + if (he_type != RATE_MCS_HE_TYPE_TRIG) { + ltf = IEEE80211_RADIOTAP_HE_DATA5_LTF_SIZE_4X; + rx_status->eht.gi = NL80211_RATE_INFO_EHT_GI_3_2; + } + break; + default: + /* nothing here */ + break; + } + + if (ltf != IEEE80211_RADIOTAP_HE_DATA5_LTF_SIZE_UNKNOWN) { + eht->known |= cpu_to_le32(IEEE80211_RADIOTAP_EHT_KNOWN_GI); + eht->data[0] |= cpu_to_le32 + (FIELD_PREP(IEEE80211_RADIOTAP_EHT_DATA0_LTF, + ltf) | + FIELD_PREP(IEEE80211_RADIOTAP_EHT_DATA0_GI, + rx_status->eht.gi)); + } + + if (!phy_data->with_data) { + eht->known |= cpu_to_le32(IEEE80211_RADIOTAP_EHT_KNOWN_NSS_S | + IEEE80211_RADIOTAP_EHT_KNOWN_BEAMFORMED_S); + eht->data[7] |= + le32_encode_bits(le32_get_bits(phy_data->rx_vec[2], + RX_NO_DATA_RX_VEC2_EHT_NSTS_MSK), + IEEE80211_RADIOTAP_EHT_DATA7_NSS_S); + if (rate_n_flags & RATE_MCS_BF_MSK) + eht->data[7] |= + cpu_to_le32(IEEE80211_RADIOTAP_EHT_DATA7_BEAMFORMED_S); + } else { + eht->user_info[0] |= + cpu_to_le32(IEEE80211_RADIOTAP_EHT_USER_INFO_MCS_KNOWN | + IEEE80211_RADIOTAP_EHT_USER_INFO_CODING_KNOWN | + IEEE80211_RADIOTAP_EHT_USER_INFO_NSS_KNOWN_O | + IEEE80211_RADIOTAP_EHT_USER_INFO_BEAMFORMING_KNOWN_O | + IEEE80211_RADIOTAP_EHT_USER_INFO_DATA_FOR_USER); + + if (rate_n_flags & RATE_MCS_BF_MSK) + eht->user_info[0] |= + cpu_to_le32(IEEE80211_RADIOTAP_EHT_USER_INFO_BEAMFORMING_O); + + if (rate_n_flags & RATE_MCS_LDPC_MSK) + eht->user_info[0] |= + cpu_to_le32(IEEE80211_RADIOTAP_EHT_USER_INFO_CODING); + + eht->user_info[0] |= cpu_to_le32 + (FIELD_PREP(IEEE80211_RADIOTAP_EHT_USER_INFO_MCS, + u32_get_bits(rate_n_flags, + RATE_VHT_MCS_RATE_CODE_MSK)) | + FIELD_PREP(IEEE80211_RADIOTAP_EHT_USER_INFO_NSS_O, + u32_get_bits(rate_n_flags, + RATE_MCS_NSS_MSK))); + } +} + +#ifdef CONFIG_IWLWIFI_DEBUGFS +static void iwl_mld_add_rtap_sniffer_config(struct iwl_mld *mld, + struct sk_buff *skb) +{ + struct ieee80211_rx_status *rx_status = IEEE80211_SKB_RXCB(skb); + struct ieee80211_radiotap_vendor_content *radiotap; + const u16 vendor_data_len = sizeof(mld->monitor.cur_aid); + + if (!mld->monitor.cur_aid) + return; + + radiotap = + iwl_mld_radiotap_put_tlv(skb, + IEEE80211_RADIOTAP_VENDOR_NAMESPACE, + sizeof(*radiotap) + vendor_data_len); + + /* Intel OUI */ + radiotap->oui[0] = 0xf6; + radiotap->oui[1] = 0x54; + radiotap->oui[2] = 0x25; + /* radiotap sniffer config sub-namespace */ + radiotap->oui_subtype = 1; + radiotap->vendor_type = 0; + + /* fill the data now */ + memcpy(radiotap->data, &mld->monitor.cur_aid, + sizeof(mld->monitor.cur_aid)); + + rx_status->flag |= RX_FLAG_RADIOTAP_TLV_AT_END; +} +#endif + +static void iwl_mld_rx_fill_status(struct iwl_mld *mld, struct sk_buff *skb, + struct iwl_mld_rx_phy_data *phy_data, + struct iwl_rx_mpdu_desc *mpdu_desc, + struct ieee80211_hdr *hdr, + int queue) +{ + struct ieee80211_rx_status *rx_status = IEEE80211_SKB_RXCB(skb); + u32 format = phy_data->rate_n_flags & RATE_MCS_MOD_TYPE_MSK; + u32 rate_n_flags = phy_data->rate_n_flags; + u8 stbc = u32_get_bits(rate_n_flags, RATE_MCS_STBC_MSK); + bool is_sgi = rate_n_flags & RATE_MCS_SGI_MSK; + + if (WARN_ON_ONCE(phy_data->with_data && (!mpdu_desc || !hdr))) + return; + + /* Keep packets with CRC errors (and with overrun) for monitor mode + * (otherwise the firmware discards them) but mark them as bad. + */ + if (phy_data->with_data && + (!(mpdu_desc->status & cpu_to_le32(IWL_RX_MPDU_STATUS_CRC_OK)) || + !(mpdu_desc->status & cpu_to_le32(IWL_RX_MPDU_STATUS_OVERRUN_OK)))) { + IWL_DEBUG_RX(mld, "Bad CRC or FIFO: 0x%08X.\n", + le32_to_cpu(mpdu_desc->status)); + rx_status->flag |= RX_FLAG_FAILED_FCS_CRC; + } + + phy_data->info_type = IWL_RX_PHY_INFO_TYPE_NONE; + + if (phy_data->with_data && + likely(!(phy_data->phy_info & IWL_RX_MPDU_PHY_TSF_OVERLOAD))) { + rx_status->mactime = + le64_to_cpu(mpdu_desc->v3.tsf_on_air_rise); + + /* TSF as indicated by the firmware is at INA time */ + rx_status->flag |= RX_FLAG_MACTIME_PLCP_START; + } else { + phy_data->info_type = + le32_get_bits(phy_data->data1, + IWL_RX_PHY_DATA1_INFO_TYPE_MASK); + } + + /* management stuff on default queue */ + if (!queue && phy_data->with_data && + unlikely(ieee80211_is_beacon(hdr->frame_control) || + ieee80211_is_probe_resp(hdr->frame_control))) { + rx_status->boottime_ns = ktime_get_boottime_ns(); + + if (mld->scan.pass_all_sched_res == SCHED_SCAN_PASS_ALL_STATE_ENABLED) + mld->scan.pass_all_sched_res = SCHED_SCAN_PASS_ALL_STATE_FOUND; + } + + /* set the preamble flag if appropriate */ + if (format == RATE_MCS_CCK_MSK && + phy_data->phy_info & IWL_RX_MPDU_PHY_SHORT_PREAMBLE) + rx_status->enc_flags |= RX_ENC_FLAG_SHORTPRE; + + iwl_mld_fill_signal(mld, rx_status, phy_data); + + /* This may be overridden by iwl_mld_rx_he() to HE_RU */ + switch (rate_n_flags & RATE_MCS_CHAN_WIDTH_MSK) { + case RATE_MCS_CHAN_WIDTH_20: + break; + case RATE_MCS_CHAN_WIDTH_40: + rx_status->bw = RATE_INFO_BW_40; + break; + case RATE_MCS_CHAN_WIDTH_80: + rx_status->bw = RATE_INFO_BW_80; + break; + case RATE_MCS_CHAN_WIDTH_160: + rx_status->bw = RATE_INFO_BW_160; + break; + case RATE_MCS_CHAN_WIDTH_320: + rx_status->bw = RATE_INFO_BW_320; + break; + } + + /* must be before L-SIG data */ + if (format == RATE_MCS_HE_MSK) + iwl_mld_rx_he(mld, skb, phy_data, queue); + + iwl_mld_decode_lsig(skb, phy_data); + + rx_status->device_timestamp = phy_data->gp2_on_air_rise; + + /* using TLV format and must be after all fixed len fields */ + if (format == RATE_MCS_EHT_MSK) + iwl_mld_rx_eht(mld, skb, phy_data, queue); + +#ifdef CONFIG_IWLWIFI_DEBUGFS + if (unlikely(mld->monitor.on)) + iwl_mld_add_rtap_sniffer_config(mld, skb); +#endif + + if (format != RATE_MCS_CCK_MSK && is_sgi) + rx_status->enc_flags |= RX_ENC_FLAG_SHORT_GI; + + if (rate_n_flags & RATE_MCS_LDPC_MSK) + rx_status->enc_flags |= RX_ENC_FLAG_LDPC; + + switch (format) { + case RATE_MCS_HT_MSK: + rx_status->encoding = RX_ENC_HT; + rx_status->rate_idx = RATE_HT_MCS_INDEX(rate_n_flags); + rx_status->enc_flags |= stbc << RX_ENC_FLAG_STBC_SHIFT; + break; + case RATE_MCS_VHT_MSK: + case RATE_MCS_HE_MSK: + case RATE_MCS_EHT_MSK: + if (format == RATE_MCS_VHT_MSK) { + rx_status->encoding = RX_ENC_VHT; + } else if (format == RATE_MCS_HE_MSK) { + rx_status->encoding = RX_ENC_HE; + rx_status->he_dcm = + !!(rate_n_flags & RATE_HE_DUAL_CARRIER_MODE_MSK); + } else if (format == RATE_MCS_EHT_MSK) { + rx_status->encoding = RX_ENC_EHT; + } + + rx_status->nss = u32_get_bits(rate_n_flags, RATE_MCS_NSS_MSK) + 1; + rx_status->rate_idx = rate_n_flags & RATE_MCS_CODE_MSK; + rx_status->enc_flags |= stbc << RX_ENC_FLAG_STBC_SHIFT; + break; + default: { + int rate = + iwl_mld_legacy_hw_idx_to_mac80211_idx(rate_n_flags, + rx_status->band); + + /* valid rate */ + if (rate >= 0 && rate <= 0xFF) { + rx_status->rate_idx = rate; + break; + } + + /* invalid rate */ + rx_status->rate_idx = 0; + + if (net_ratelimit()) + IWL_ERR(mld, "invalid rate_n_flags=0x%x, band=%d\n", + rate_n_flags, rx_status->band); + break; + } + } +} + +/* iwl_mld_create_skb adds the rxb to a new skb */ +static int iwl_mld_build_rx_skb(struct iwl_mld *mld, struct sk_buff *skb, + struct ieee80211_hdr *hdr, u16 len, + u8 crypt_len, struct iwl_rx_cmd_buffer *rxb) +{ + struct iwl_rx_packet *pkt = rxb_addr(rxb); + struct iwl_rx_mpdu_desc *desc = (void *)pkt->data; + unsigned int headlen, fraglen, pad_len = 0; + unsigned int hdrlen = ieee80211_hdrlen(hdr->frame_control); + u8 mic_crc_len = u8_get_bits(desc->mac_flags1, + IWL_RX_MPDU_MFLG1_MIC_CRC_LEN_MASK) << 1; + + if (desc->mac_flags2 & IWL_RX_MPDU_MFLG2_PAD) { + len -= 2; + pad_len = 2; + } + + /* For non monitor interface strip the bytes the RADA might not have + * removed (it might be disabled, e.g. for mgmt frames). As a monitor + * interface cannot exist with other interfaces, this removal is safe + * and sufficient, in monitor mode there's no decryption being done. + */ + if (len > mic_crc_len && !ieee80211_hw_check(mld->hw, RX_INCLUDES_FCS)) + len -= mic_crc_len; + + /* If frame is small enough to fit in skb->head, pull it completely. + * If not, only pull ieee80211_hdr (including crypto if present, and + * an additional 8 bytes for SNAP/ethertype, see below) so that + * splice() or TCP coalesce are more efficient. + * + * Since, in addition, ieee80211_data_to_8023() always pull in at + * least 8 bytes (possibly more for mesh) we can do the same here + * to save the cost of doing it later. That still doesn't pull in + * the actual IP header since the typical case has a SNAP header. + * If the latter changes (there are efforts in the standards group + * to do so) we should revisit this and ieee80211_data_to_8023(). + */ + headlen = (len <= skb_tailroom(skb)) ? len : hdrlen + crypt_len + 8; + + /* The firmware may align the packet to DWORD. + * The padding is inserted after the IV. + * After copying the header + IV skip the padding if + * present before copying packet data. + */ + hdrlen += crypt_len; + + if (unlikely(headlen < hdrlen)) + return -EINVAL; + + /* Since data doesn't move data while putting data on skb and that is + * the only way we use, data + len is the next place that hdr would + * be put + */ + skb_set_mac_header(skb, skb->len); + skb_put_data(skb, hdr, hdrlen); + skb_put_data(skb, (u8 *)hdr + hdrlen + pad_len, headlen - hdrlen); + + if (skb->ip_summed == CHECKSUM_COMPLETE) { + struct { + u8 hdr[6]; + __be16 type; + } __packed *shdr = (void *)((u8 *)hdr + hdrlen + pad_len); + + if (unlikely(headlen - hdrlen < sizeof(*shdr) || + !ether_addr_equal(shdr->hdr, rfc1042_header) || + (shdr->type != htons(ETH_P_IP) && + shdr->type != htons(ETH_P_ARP) && + shdr->type != htons(ETH_P_IPV6) && + shdr->type != htons(ETH_P_8021Q) && + shdr->type != htons(ETH_P_PAE) && + shdr->type != htons(ETH_P_TDLS)))) + skb->ip_summed = CHECKSUM_NONE; + } + + fraglen = len - headlen; + + if (fraglen) { + int offset = (u8 *)hdr + headlen + pad_len - + (u8 *)rxb_addr(rxb) + rxb_offset(rxb); + + skb_add_rx_frag(skb, 0, rxb_steal_page(rxb), offset, + fraglen, rxb->truesize); + } + + return 0; +} + +/* returns true if a packet is a duplicate or invalid tid and + * should be dropped. Updates AMSDU PN tracking info + */ +VISIBLE_IF_IWLWIFI_KUNIT +bool +iwl_mld_is_dup(struct iwl_mld *mld, struct ieee80211_sta *sta, + struct ieee80211_hdr *hdr, + const struct iwl_rx_mpdu_desc *mpdu_desc, + struct ieee80211_rx_status *rx_status, int queue) +{ + struct iwl_mld_sta *mld_sta; + struct iwl_mld_rxq_dup_data *dup_data; + u8 tid, sub_frame_idx; + + if (WARN_ON(!sta)) + return false; + + mld_sta = iwl_mld_sta_from_mac80211(sta); + + if (WARN_ON_ONCE(!mld_sta->dup_data)) + return false; + + dup_data = &mld_sta->dup_data[queue]; + + /* Drop duplicate 802.11 retransmissions + * (IEEE 802.11-2020: 10.3.2.14 "Duplicate detection and recovery") + */ + if (ieee80211_is_ctl(hdr->frame_control) || + ieee80211_is_any_nullfunc(hdr->frame_control) || + is_multicast_ether_addr(hdr->addr1)) + return false; + + if (ieee80211_is_data_qos(hdr->frame_control)) { + /* frame has qos control */ + tid = ieee80211_get_tid(hdr); + if (tid >= IWL_MAX_TID_COUNT) + return true; + } else { + tid = IWL_MAX_TID_COUNT; + } + + /* If this wasn't a part of an A-MSDU the sub-frame index will be 0 */ + sub_frame_idx = mpdu_desc->amsdu_info & + IWL_RX_MPDU_AMSDU_SUBFRAME_IDX_MASK; + + if (IWL_FW_CHECK(mld, + sub_frame_idx > 0 && + !(mpdu_desc->mac_flags2 & IWL_RX_MPDU_MFLG2_AMSDU), + "got sub_frame_idx=%d but A-MSDU flag is not set\n", + sub_frame_idx)) + return true; + + if (unlikely(ieee80211_has_retry(hdr->frame_control) && + dup_data->last_seq[tid] == hdr->seq_ctrl && + dup_data->last_sub_frame_idx[tid] >= sub_frame_idx)) + return true; + + /* Allow same PN as the first subframe for following sub frames */ + if (dup_data->last_seq[tid] == hdr->seq_ctrl && + sub_frame_idx > dup_data->last_sub_frame_idx[tid]) + rx_status->flag |= RX_FLAG_ALLOW_SAME_PN; + + dup_data->last_seq[tid] = hdr->seq_ctrl; + dup_data->last_sub_frame_idx[tid] = sub_frame_idx; + + rx_status->flag |= RX_FLAG_DUP_VALIDATED; + + return false; +} +EXPORT_SYMBOL_IF_IWLWIFI_KUNIT(iwl_mld_is_dup); + +static void iwl_mld_update_last_rx_timestamp(struct iwl_mld *mld, u8 baid) +{ + unsigned long now = jiffies; + unsigned long timeout; + struct iwl_mld_baid_data *ba_data; + + ba_data = rcu_dereference(mld->fw_id_to_ba[baid]); + if (!ba_data) { + IWL_DEBUG_HT(mld, "BAID %d not found in map\n", baid); + return; + } + + if (!ba_data->timeout) + return; + + /* To minimize cache bouncing between RX queues, avoid frequent updates + * to last_rx_timestamp. update it only when the timeout period has + * passed. The worst-case scenario is the session expiring after + * approximately 2 * timeout, which is negligible (the update is + * atomic). + */ + timeout = TU_TO_JIFFIES(ba_data->timeout); + if (time_is_before_jiffies(ba_data->last_rx_timestamp + timeout)) + ba_data->last_rx_timestamp = now; +} + +/* Processes received packets for a station. + * Sets *drop to true if the packet should be dropped. + * Returns the station if found, or NULL otherwise. + */ +static struct ieee80211_sta * +iwl_mld_rx_with_sta(struct iwl_mld *mld, struct ieee80211_hdr *hdr, + struct sk_buff *skb, + const struct iwl_rx_mpdu_desc *mpdu_desc, + const struct iwl_rx_packet *pkt, int queue, bool *drop) +{ + struct ieee80211_sta *sta = NULL; + struct ieee80211_link_sta *link_sta = NULL; + struct ieee80211_rx_status *rx_status; + u8 baid; + + if (mpdu_desc->status & cpu_to_le32(IWL_RX_MPDU_STATUS_SRC_STA_FOUND)) { + u8 sta_id = le32_get_bits(mpdu_desc->status, + IWL_RX_MPDU_STATUS_STA_ID); + + if (IWL_FW_CHECK(mld, + sta_id >= mld->fw->ucode_capa.num_stations, + "rx_mpdu: invalid sta_id %d\n", sta_id)) + return NULL; + + link_sta = rcu_dereference(mld->fw_id_to_link_sta[sta_id]); + if (!IS_ERR_OR_NULL(link_sta)) + sta = link_sta->sta; + } else if (!is_multicast_ether_addr(hdr->addr2)) { + /* Passing NULL is fine since we prevent two stations with the + * same address from being added. + */ + sta = ieee80211_find_sta_by_ifaddr(mld->hw, hdr->addr2, NULL); + } + + /* we may not have any station yet */ + if (!sta) + return NULL; + + rx_status = IEEE80211_SKB_RXCB(skb); + + if (link_sta && sta->valid_links) { + rx_status->link_valid = true; + rx_status->link_id = link_sta->link_id; + } + + /* fill checksum */ + if (ieee80211_is_data(hdr->frame_control) && + pkt->len_n_flags & cpu_to_le32(FH_RSCSR_RPA_EN)) { + u16 hwsum = be16_to_cpu(mpdu_desc->v3.raw_xsum); + + skb->ip_summed = CHECKSUM_COMPLETE; + skb->csum = csum_unfold(~(__force __sum16)hwsum); + } + + if (iwl_mld_is_dup(mld, sta, hdr, mpdu_desc, rx_status, queue)) { + IWL_DEBUG_DROP(mld, "Dropping duplicate packet 0x%x\n", + le16_to_cpu(hdr->seq_ctrl)); + *drop = true; + return NULL; + } + + baid = le32_get_bits(mpdu_desc->reorder_data, + IWL_RX_MPDU_REORDER_BAID_MASK); + if (baid != IWL_RX_REORDER_DATA_INVALID_BAID) + iwl_mld_update_last_rx_timestamp(mld, baid); + + if (link_sta && ieee80211_is_data(hdr->frame_control)) { + u8 sub_frame_idx = mpdu_desc->amsdu_info & + IWL_RX_MPDU_AMSDU_SUBFRAME_IDX_MASK; + + /* 0 means not an A-MSDU, and 1 means a new A-MSDU */ + if (!sub_frame_idx || sub_frame_idx == 1) + iwl_mld_count_mpdu_rx(link_sta, queue, 1); + + if (!is_multicast_ether_addr(hdr->addr1)) + iwl_mld_low_latency_update_counters(mld, hdr, sta, + queue); + } + + return sta; +} + +#define KEY_IDX_LEN 2 + +static int iwl_mld_rx_mgmt_prot(struct ieee80211_sta *sta, + struct ieee80211_hdr *hdr, + struct ieee80211_rx_status *rx_status, + u32 mpdu_status, + u32 mpdu_len) +{ + struct wireless_dev *wdev; + struct iwl_mld_sta *mld_sta; + struct iwl_mld_vif *mld_vif; + u8 keyidx; + struct ieee80211_key_conf *key; + const u8 *frame = (void *)hdr; + + if ((mpdu_status & IWL_RX_MPDU_STATUS_SEC_MASK) == + IWL_RX_MPDU_STATUS_SEC_NONE) + return 0; + + /* For non-beacon, we don't really care. But beacons may + * be filtered out, and we thus need the firmware's replay + * detection, otherwise beacons the firmware previously + * filtered could be replayed, or something like that, and + * it can filter a lot - though usually only if nothing has + * changed. + */ + if (!ieee80211_is_beacon(hdr->frame_control)) + return 0; + + if (!sta) + return -1; + + mld_sta = iwl_mld_sta_from_mac80211(sta); + mld_vif = iwl_mld_vif_from_mac80211(mld_sta->vif); + + /* key mismatch - will also report !MIC_OK but we shouldn't count it */ + if (!(mpdu_status & IWL_RX_MPDU_STATUS_KEY_VALID)) + goto report; + + /* good cases */ + if (likely(mpdu_status & IWL_RX_MPDU_STATUS_MIC_OK && + !(mpdu_status & IWL_RX_MPDU_STATUS_REPLAY_ERROR))) { + rx_status->flag |= RX_FLAG_DECRYPTED; + return 0; + } + + /* both keys will have the same cipher and MIC length, use + * whichever one is available + */ + key = rcu_dereference(mld_vif->bigtks[0]); + if (!key) { + key = rcu_dereference(mld_vif->bigtks[1]); + if (!key) + goto report; + } + + if (mpdu_len < key->icv_len + IEEE80211_GMAC_PN_LEN + KEY_IDX_LEN) + goto report; + + /* get the real key ID */ + keyidx = frame[mpdu_len - key->icv_len - IEEE80211_GMAC_PN_LEN - KEY_IDX_LEN]; + /* and if that's the other key, look it up */ + if (keyidx != key->keyidx) { + /* shouldn't happen since firmware checked, but be safe + * in case the MIC length is wrong too, for example + */ + if (keyidx != 6 && keyidx != 7) + return -1; + + key = rcu_dereference(mld_vif->bigtks[keyidx - 6]); + if (!key) + goto report; + } + + /* Report status to mac80211 */ + if (!(mpdu_status & IWL_RX_MPDU_STATUS_MIC_OK)) + ieee80211_key_mic_failure(key); + else if (mpdu_status & IWL_RX_MPDU_STATUS_REPLAY_ERROR) + ieee80211_key_replay(key); +report: + wdev = ieee80211_vif_to_wdev(mld_sta->vif); + if (wdev->netdev) + cfg80211_rx_unprot_mlme_mgmt(wdev->netdev, (void *)hdr, + mpdu_len); + + return -1; +} + +static int iwl_mld_rx_crypto(struct iwl_mld *mld, + struct ieee80211_sta *sta, + struct ieee80211_hdr *hdr, + struct ieee80211_rx_status *rx_status, + struct iwl_rx_mpdu_desc *desc, int queue, + u32 pkt_flags, u8 *crypto_len) +{ + u32 status = le32_to_cpu(desc->status); + + if (unlikely(ieee80211_is_mgmt(hdr->frame_control) && + !ieee80211_has_protected(hdr->frame_control))) + return iwl_mld_rx_mgmt_prot(sta, hdr, rx_status, status, + le16_to_cpu(desc->mpdu_len)); + + if (!ieee80211_has_protected(hdr->frame_control) || + (status & IWL_RX_MPDU_STATUS_SEC_MASK) == + IWL_RX_MPDU_STATUS_SEC_NONE) + return 0; + + switch (status & IWL_RX_MPDU_STATUS_SEC_MASK) { + case IWL_RX_MPDU_STATUS_SEC_CCM: + case IWL_RX_MPDU_STATUS_SEC_GCM: + BUILD_BUG_ON(IEEE80211_CCMP_PN_LEN != IEEE80211_GCMP_PN_LEN); + if (!(status & IWL_RX_MPDU_STATUS_MIC_OK)) { + IWL_DEBUG_DROP(mld, + "Dropping packet, bad MIC (CCM/GCM)\n"); + return -1; + } + + rx_status->flag |= RX_FLAG_DECRYPTED | RX_FLAG_MIC_STRIPPED; + *crypto_len = IEEE80211_CCMP_HDR_LEN; + return 0; + case IWL_RX_MPDU_STATUS_SEC_TKIP: + if (!(status & IWL_RX_MPDU_STATUS_ICV_OK)) + return -1; + + if (!(status & RX_MPDU_RES_STATUS_MIC_OK)) + rx_status->flag |= RX_FLAG_MMIC_ERROR; + + if (pkt_flags & FH_RSCSR_RADA_EN) { + rx_status->flag |= RX_FLAG_ICV_STRIPPED; + rx_status->flag |= RX_FLAG_MMIC_STRIPPED; + } + + *crypto_len = IEEE80211_TKIP_IV_LEN; + rx_status->flag |= RX_FLAG_DECRYPTED; + return 0; + default: + break; + } + + return 0; +} + +static void iwl_mld_rx_update_ampdu_ref(struct iwl_mld *mld, + struct iwl_mld_rx_phy_data *phy_data, + struct ieee80211_rx_status *rx_status) +{ + bool toggle_bit = + phy_data->phy_info & IWL_RX_MPDU_PHY_AMPDU_TOGGLE; + + rx_status->flag |= RX_FLAG_AMPDU_DETAILS; + /* Toggle is switched whenever new aggregation starts. Make + * sure ampdu_reference is never 0 so we can later use it to + * see if the frame was really part of an A-MPDU or not. + */ + if (toggle_bit != mld->monitor.ampdu_toggle) { + mld->monitor.ampdu_ref++; + if (mld->monitor.ampdu_ref == 0) + mld->monitor.ampdu_ref++; + mld->monitor.ampdu_toggle = toggle_bit; + phy_data->first_subframe = true; + } + rx_status->ampdu_reference = mld->monitor.ampdu_ref; +} + +static void +iwl_mld_fill_rx_status_band_freq(struct iwl_mld_rx_phy_data *phy_data, + struct iwl_rx_mpdu_desc *mpdu_desc, + struct ieee80211_rx_status *rx_status) +{ + enum nl80211_band band; + + band = BAND_IN_RX_STATUS(mpdu_desc->mac_phy_idx); + rx_status->band = iwl_mld_phy_band_to_nl80211(band); + rx_status->freq = ieee80211_channel_to_frequency(phy_data->channel, + rx_status->band); +} + +void iwl_mld_rx_mpdu(struct iwl_mld *mld, struct napi_struct *napi, + struct iwl_rx_cmd_buffer *rxb, int queue) +{ + struct iwl_rx_packet *pkt = rxb_addr(rxb); + struct iwl_mld_rx_phy_data phy_data = {}; + struct iwl_rx_mpdu_desc *mpdu_desc = (void *)pkt->data; + struct ieee80211_sta *sta; + struct ieee80211_hdr *hdr; + struct sk_buff *skb; + size_t mpdu_desc_size = sizeof(*mpdu_desc); + bool drop = false; + u8 crypto_len = 0; + u32 pkt_len = iwl_rx_packet_payload_len(pkt); + u32 mpdu_len; + enum iwl_mld_reorder_result reorder_res; + struct ieee80211_rx_status *rx_status; + + if (unlikely(mld->fw_status.in_hw_restart)) + return; + + if (IWL_FW_CHECK(mld, pkt_len < mpdu_desc_size, + "Bad REPLY_RX_MPDU_CMD size (%d)\n", pkt_len)) + return; + + mpdu_len = le16_to_cpu(mpdu_desc->mpdu_len); + + if (IWL_FW_CHECK(mld, mpdu_len + mpdu_desc_size > pkt_len, + "FW lied about packet len (%d)\n", pkt_len)) + return; + + /* Don't use dev_alloc_skb(), we'll have enough headroom once + * ieee80211_hdr pulled. + */ + skb = alloc_skb(128, GFP_ATOMIC); + if (!skb) { + IWL_ERR(mld, "alloc_skb failed\n"); + return; + } + + hdr = (void *)(pkt->data + mpdu_desc_size); + + iwl_mld_fill_phy_data(mpdu_desc, &phy_data); + + if (mpdu_desc->mac_flags2 & IWL_RX_MPDU_MFLG2_PAD) { + /* If the device inserted padding it means that (it thought) + * the 802.11 header wasn't a multiple of 4 bytes long. In + * this case, reserve two bytes at the start of the SKB to + * align the payload properly in case we end up copying it. + */ + skb_reserve(skb, 2); + } + + rx_status = IEEE80211_SKB_RXCB(skb); + + /* this is needed early */ + iwl_mld_fill_rx_status_band_freq(&phy_data, mpdu_desc, rx_status); + + rcu_read_lock(); + + sta = iwl_mld_rx_with_sta(mld, hdr, skb, mpdu_desc, pkt, queue, &drop); + if (drop) + goto drop; + + /* update aggregation data for monitor sake on default queue */ + if (!queue && (phy_data.phy_info & IWL_RX_MPDU_PHY_AMPDU)) + iwl_mld_rx_update_ampdu_ref(mld, &phy_data, rx_status); + + iwl_mld_rx_fill_status(mld, skb, &phy_data, mpdu_desc, hdr, queue); + + if (iwl_mld_rx_crypto(mld, sta, hdr, rx_status, mpdu_desc, queue, + le32_to_cpu(pkt->len_n_flags), &crypto_len)) + goto drop; + + if (iwl_mld_build_rx_skb(mld, skb, hdr, mpdu_len, crypto_len, rxb)) + goto drop; + + /* time sync frame is saved and will be released later when the + * notification with the timestamps arrives. + */ + if (iwl_mld_time_sync_frame(mld, skb, hdr->addr2)) + goto out; + + reorder_res = iwl_mld_reorder(mld, napi, queue, sta, skb, mpdu_desc); + switch (reorder_res) { + case IWL_MLD_PASS_SKB: + break; + case IWL_MLD_DROP_SKB: + goto drop; + case IWL_MLD_BUFFERED_SKB: + goto out; + default: + WARN_ON(1); + goto drop; + } + + iwl_mld_pass_packet_to_mac80211(mld, napi, skb, queue, sta); + + goto out; + +drop: + kfree_skb(skb); +out: + rcu_read_unlock(); +} + +#define SYNC_RX_QUEUE_TIMEOUT (HZ) +void iwl_mld_sync_rx_queues(struct iwl_mld *mld, + enum iwl_mld_internal_rxq_notif_type type, + const void *notif_payload, u32 notif_payload_size) +{ + u8 num_rx_queues = mld->trans->num_rx_queues; + struct { + struct iwl_rxq_sync_cmd sync_cmd; + struct iwl_mld_internal_rxq_notif notif; + } __packed cmd = { + .sync_cmd.rxq_mask = cpu_to_le32(BIT(num_rx_queues) - 1), + .sync_cmd.count = + cpu_to_le32(sizeof(struct iwl_mld_internal_rxq_notif) + + notif_payload_size), + .notif.type = type, + .notif.cookie = mld->rxq_sync.cookie, + }; + struct iwl_host_cmd hcmd = { + .id = WIDE_ID(DATA_PATH_GROUP, TRIGGER_RX_QUEUES_NOTIF_CMD), + .data[0] = &cmd, + .len[0] = sizeof(cmd), + .data[1] = notif_payload, + .len[1] = notif_payload_size, + }; + int ret; + + /* size must be a multiple of DWORD */ + if (WARN_ON(cmd.sync_cmd.count & cpu_to_le32(3))) + return; + + mld->rxq_sync.state = (1 << num_rx_queues) - 1; + + ret = iwl_mld_send_cmd(mld, &hcmd); + if (ret) { + IWL_ERR(mld, "Failed to trigger RX queues sync (%d)\n", ret); + goto out; + } + + ret = wait_event_timeout(mld->rxq_sync.waitq, + READ_ONCE(mld->rxq_sync.state) == 0, + SYNC_RX_QUEUE_TIMEOUT); + WARN_ONCE(!ret, "RXQ sync failed: state=0x%lx, cookie=%d\n", + mld->rxq_sync.state, mld->rxq_sync.cookie); + +out: + mld->rxq_sync.state = 0; + mld->rxq_sync.cookie++; +} + +void iwl_mld_handle_rx_queues_sync_notif(struct iwl_mld *mld, + struct napi_struct *napi, + struct iwl_rx_packet *pkt, int queue) +{ + struct iwl_rxq_sync_notification *notif; + struct iwl_mld_internal_rxq_notif *internal_notif; + u32 len = iwl_rx_packet_payload_len(pkt); + size_t combined_notif_len = sizeof(*notif) + sizeof(*internal_notif); + + notif = (void *)pkt->data; + internal_notif = (void *)notif->payload; + + if (IWL_FW_CHECK(mld, len < combined_notif_len, + "invalid notification size %u (%zu)\n", + len, combined_notif_len)) + return; + + len -= combined_notif_len; + + if (IWL_FW_CHECK(mld, mld->rxq_sync.cookie != internal_notif->cookie, + "received expired RX queue sync message (cookie=%d expected=%d q[%d])\n", + internal_notif->cookie, mld->rxq_sync.cookie, queue)) + return; + + switch (internal_notif->type) { + case IWL_MLD_RXQ_EMPTY: + IWL_FW_CHECK(mld, len, + "invalid empty notification size %d\n", len); + break; + case IWL_MLD_RXQ_NOTIF_DEL_BA: + if (IWL_FW_CHECK(mld, len != sizeof(struct iwl_mld_delba_data), + "invalid delba notification size %u (%zu)\n", + len, sizeof(struct iwl_mld_delba_data))) + break; + iwl_mld_del_ba(mld, queue, (void *)internal_notif->payload); + break; + default: + WARN_ON_ONCE(1); + } + + IWL_FW_CHECK(mld, !test_and_clear_bit(queue, &mld->rxq_sync.state), + "RXQ sync: queue %d responded a second time!\n", queue); + + if (READ_ONCE(mld->rxq_sync.state) == 0) + wake_up(&mld->rxq_sync.waitq); +} + +void iwl_mld_rx_monitor_no_data(struct iwl_mld *mld, struct napi_struct *napi, + struct iwl_rx_packet *pkt, int queue) +{ + struct iwl_rx_no_data_ver_3 *desc; + struct iwl_mld_rx_phy_data phy_data; + struct ieee80211_rx_status *rx_status; + struct sk_buff *skb; + u32 format, rssi; + + if (unlikely(mld->fw_status.in_hw_restart)) + return; + + if (IWL_FW_CHECK(mld, iwl_rx_packet_payload_len(pkt) < sizeof(*desc), + "Bad RX_NO_DATA_NOTIF size (%d)\n", + iwl_rx_packet_payload_len(pkt))) + return; + + desc = (void *)pkt->data; + + rssi = le32_to_cpu(desc->rssi); + phy_data.energy_a = u32_get_bits(rssi, RX_NO_DATA_CHAIN_A_MSK); + phy_data.energy_b = u32_get_bits(rssi, RX_NO_DATA_CHAIN_B_MSK); + phy_data.channel = u32_get_bits(rssi, RX_NO_DATA_CHANNEL_MSK); + phy_data.data0 = desc->phy_info[0]; + phy_data.data1 = desc->phy_info[1]; + phy_data.phy_info = IWL_RX_MPDU_PHY_TSF_OVERLOAD; + phy_data.gp2_on_air_rise = le32_to_cpu(desc->on_air_rise_time); + phy_data.rate_n_flags = le32_to_cpu(desc->rate); + phy_data.with_data = false; + + BUILD_BUG_ON(sizeof(phy_data.rx_vec) != sizeof(desc->rx_vec)); + memcpy(phy_data.rx_vec, desc->rx_vec, sizeof(phy_data.rx_vec)); + + format = phy_data.rate_n_flags & RATE_MCS_MOD_TYPE_MSK; + + /* Don't use dev_alloc_skb(), we'll have enough headroom once + * ieee80211_hdr pulled. + */ + skb = alloc_skb(128, GFP_ATOMIC); + if (!skb) { + IWL_ERR(mld, "alloc_skb failed\n"); + return; + } + + rx_status = IEEE80211_SKB_RXCB(skb); + + /* 0-length PSDU */ + rx_status->flag |= RX_FLAG_NO_PSDU; + + /* mark as failed PLCP on any errors to skip checks in mac80211 */ + if (le32_get_bits(desc->info, RX_NO_DATA_INFO_ERR_MSK) != + RX_NO_DATA_INFO_ERR_NONE) + rx_status->flag |= RX_FLAG_FAILED_PLCP_CRC; + + switch (le32_get_bits(desc->info, RX_NO_DATA_INFO_TYPE_MSK)) { + case RX_NO_DATA_INFO_TYPE_NDP: + rx_status->zero_length_psdu_type = + IEEE80211_RADIOTAP_ZERO_LEN_PSDU_SOUNDING; + break; + case RX_NO_DATA_INFO_TYPE_MU_UNMATCHED: + case RX_NO_DATA_INFO_TYPE_TB_UNMATCHED: + rx_status->zero_length_psdu_type = + IEEE80211_RADIOTAP_ZERO_LEN_PSDU_NOT_CAPTURED; + break; + default: + rx_status->zero_length_psdu_type = + IEEE80211_RADIOTAP_ZERO_LEN_PSDU_VENDOR; + break; + } + + rx_status->band = phy_data.channel > 14 ? NL80211_BAND_5GHZ : + NL80211_BAND_2GHZ; + + rx_status->freq = ieee80211_channel_to_frequency(phy_data.channel, + rx_status->band); + + iwl_mld_rx_fill_status(mld, skb, &phy_data, NULL, NULL, queue); + + /* No more radiotap info should be added after this point. + * Mark it as mac header for upper layers to know where + * the radiotap header ends. + */ + skb_set_mac_header(skb, skb->len); + + /* Override the nss from the rx_vec since the rate_n_flags has + * only 1 bit for the nss which gives a max of 2 ss but there + * may be up to 8 spatial streams. + */ + switch (format) { + case RATE_MCS_VHT_MSK: + rx_status->nss = + le32_get_bits(desc->rx_vec[0], + RX_NO_DATA_RX_VEC0_VHT_NSTS_MSK) + 1; + break; + case RATE_MCS_HE_MSK: + rx_status->nss = + le32_get_bits(desc->rx_vec[0], + RX_NO_DATA_RX_VEC0_HE_NSTS_MSK) + 1; + break; + case RATE_MCS_EHT_MSK: + rx_status->nss = + le32_get_bits(desc->rx_vec[2], + RX_NO_DATA_RX_VEC2_EHT_NSTS_MSK) + 1; + } + + /* pass the packet to mac80211 */ + rcu_read_lock(); + ieee80211_rx_napi(mld->hw, NULL, skb, napi); + rcu_read_unlock(); +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/rx.h b/drivers/net/wireless/intel/iwlwifi/mld/rx.h new file mode 100644 index 0000000000000..2beabd7e70b15 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/rx.h @@ -0,0 +1,72 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024-2025 Intel Corporation + */ +#ifndef __iwl_mld_rx_h__ +#define __iwl_mld_rx_h__ + +#include "mld.h" + +/** + * enum iwl_mld_internal_rxq_notif_type - RX queue sync notif types + * + * @IWL_MLD_RXQ_EMPTY: empty sync notification + * @IWL_MLD_RXQ_NOTIF_DEL_BA: notify RSS queues of delBA + */ +enum iwl_mld_internal_rxq_notif_type { + IWL_MLD_RXQ_EMPTY, + IWL_MLD_RXQ_NOTIF_DEL_BA, +}; + +/** + * struct iwl_mld_internal_rxq_notif - @iwl_rxq_sync_cmd internal data. + * This data is echoed by the firmware to all RSS queues and should be DWORD + * aligned. FW is agnostic to the data, so there are no endianness requirements + * + * @type: one of &iwl_mld_internal_rxq_notif_type + * @cookie: unique internal cookie to identify old notifications + * @reserved: reserved for alignment + * @payload: data to send to RX queues based on the type (may be empty) + */ +struct iwl_mld_internal_rxq_notif { + u8 type; + u8 reserved[3]; + u32 cookie; + u8 payload[]; +} __packed; + +/** + * struct iwl_mld_rx_queues_sync - RX queues sync data + * + * @waitq: wait queue for RX queues sync completion + * @cookie: unique id to correlate sync requests with responses + * @state: bitmask representing the sync state of RX queues + * all RX queues bits are set before sending the command, and the + * corresponding queue bit cleared upon handling the notification + */ +struct iwl_mld_rx_queues_sync { + wait_queue_head_t waitq; + u32 cookie; + unsigned long state; +}; + +void iwl_mld_rx_mpdu(struct iwl_mld *mld, struct napi_struct *napi, + struct iwl_rx_cmd_buffer *rxb, int queue); + +void iwl_mld_sync_rx_queues(struct iwl_mld *mld, + enum iwl_mld_internal_rxq_notif_type type, + const void *notif_payload, u32 notif_payload_size); + +void iwl_mld_handle_rx_queues_sync_notif(struct iwl_mld *mld, + struct napi_struct *napi, + struct iwl_rx_packet *pkt, int queue); + +void iwl_mld_pass_packet_to_mac80211(struct iwl_mld *mld, + struct napi_struct *napi, + struct sk_buff *skb, int queue, + struct ieee80211_sta *sta); + +void iwl_mld_rx_monitor_no_data(struct iwl_mld *mld, struct napi_struct *napi, + struct iwl_rx_packet *pkt, int queue); + +#endif /* __iwl_mld_agg_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/scan.c b/drivers/net/wireless/intel/iwlwifi/mld/scan.c new file mode 100644 index 0000000000000..5f4ede92028b4 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/scan.c @@ -0,0 +1,2006 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024-2025 Intel Corporation + */ +#include + +#include "mld.h" +#include "scan.h" +#include "hcmd.h" +#include "iface.h" +#include "phy.h" +#include "mlo.h" + +#include "fw/api/scan.h" +#include "fw/dbg.h" + +#define IWL_SCAN_DWELL_ACTIVE 10 +#define IWL_SCAN_DWELL_PASSIVE 110 +#define IWL_SCAN_NUM_OF_FRAGS 3 + +/* adaptive dwell max budget time [TU] for full scan */ +#define IWL_SCAN_ADWELL_MAX_BUDGET_FULL_SCAN 300 + +/* adaptive dwell max budget time [TU] for directed scan */ +#define IWL_SCAN_ADWELL_MAX_BUDGET_DIRECTED_SCAN 100 + +/* adaptive dwell default high band APs number */ +#define IWL_SCAN_ADWELL_DEFAULT_HB_N_APS 8 + +/* adaptive dwell default low band APs number */ +#define IWL_SCAN_ADWELL_DEFAULT_LB_N_APS 2 + +/* adaptive dwell default APs number for P2P social channels (1, 6, 11) */ +#define IWL_SCAN_ADWELL_DEFAULT_N_APS_SOCIAL 10 + +/* adaptive dwell number of APs override for P2P friendly GO channels */ +#define IWL_SCAN_ADWELL_N_APS_GO_FRIENDLY 10 + +/* adaptive dwell number of APs override for P2P social channels */ +#define IWL_SCAN_ADWELL_N_APS_SOCIAL_CHS 2 + +/* adaptive dwell number of APs override mask for p2p friendly GO */ +#define IWL_SCAN_ADWELL_N_APS_GO_FRIENDLY_BIT BIT(20) + +/* adaptive dwell number of APs override mask for social channels */ +#define IWL_SCAN_ADWELL_N_APS_SOCIAL_CHS_BIT BIT(21) + +#define SCAN_TIMEOUT_MSEC (30000 * HZ) + +/* minimal number of 2GHz and 5GHz channels in the regular scan request */ +#define IWL_MLD_6GHZ_PASSIVE_SCAN_MIN_CHANS 4 + +enum iwl_mld_scan_type { + IWL_SCAN_TYPE_NOT_SET, + IWL_SCAN_TYPE_UNASSOC, + IWL_SCAN_TYPE_WILD, + IWL_SCAN_TYPE_MILD, + IWL_SCAN_TYPE_FRAGMENTED, + IWL_SCAN_TYPE_FAST_BALANCE, +}; + +struct iwl_mld_scan_timing_params { + u32 suspend_time; + u32 max_out_time; +}; + +static const struct iwl_mld_scan_timing_params scan_timing[] = { + [IWL_SCAN_TYPE_UNASSOC] = { + .suspend_time = 0, + .max_out_time = 0, + }, + [IWL_SCAN_TYPE_WILD] = { + .suspend_time = 30, + .max_out_time = 120, + }, + [IWL_SCAN_TYPE_MILD] = { + .suspend_time = 120, + .max_out_time = 120, + }, + [IWL_SCAN_TYPE_FRAGMENTED] = { + .suspend_time = 95, + .max_out_time = 44, + }, + [IWL_SCAN_TYPE_FAST_BALANCE] = { + .suspend_time = 30, + .max_out_time = 37, + }, +}; + +struct iwl_mld_scan_params { + enum iwl_mld_scan_type type; + u32 n_channels; + u16 delay; + int n_ssids; + struct cfg80211_ssid *ssids; + struct ieee80211_channel **channels; + u32 flags; + u8 *mac_addr; + u8 *mac_addr_mask; + bool no_cck; + bool pass_all; + int n_match_sets; + struct iwl_scan_probe_req preq; + struct cfg80211_match_set *match_sets; + int n_scan_plans; + struct cfg80211_sched_scan_plan *scan_plans; + bool iter_notif; + bool respect_p2p_go; + u8 fw_link_id; + struct cfg80211_scan_6ghz_params *scan_6ghz_params; + u32 n_6ghz_params; + bool scan_6ghz; + bool enable_6ghz_passive; + u8 bssid[ETH_ALEN] __aligned(2); +}; + +struct iwl_mld_scan_respect_p2p_go_iter_data { + struct ieee80211_vif *current_vif; + bool p2p_go; +}; + +static void iwl_mld_scan_respect_p2p_go_iter(void *_data, u8 *mac, + struct ieee80211_vif *vif) +{ + struct iwl_mld_scan_respect_p2p_go_iter_data *data = _data; + + /* exclude the given vif */ + if (vif == data->current_vif) + return; + + /* TODO: CDB check the band of the GO */ + if (ieee80211_vif_type_p2p(vif) == NL80211_IFTYPE_P2P_GO && + iwl_mld_vif_from_mac80211(vif)->ap_ibss_active) + data->p2p_go = true; +} + +static bool iwl_mld_get_respect_p2p_go(struct iwl_mld *mld, + struct ieee80211_vif *vif, + bool low_latency) +{ + struct iwl_mld_scan_respect_p2p_go_iter_data data = { + .current_vif = vif, + .p2p_go = false, + }; + + if (!low_latency) + return false; + + ieee80211_iterate_active_interfaces_mtx(mld->hw, + IEEE80211_IFACE_ITER_NORMAL, + iwl_mld_scan_respect_p2p_go_iter, + &data); + + return data.p2p_go; +} + +struct iwl_mld_scan_iter_data { + struct ieee80211_vif *current_vif; + bool active_vif; + bool is_dcm_with_p2p_go; + bool global_low_latency; +}; + +static void iwl_mld_scan_iterator(void *_data, u8 *mac, + struct ieee80211_vif *vif) +{ + struct iwl_mld_scan_iter_data *data = _data; + struct ieee80211_vif *curr_vif = data->current_vif; + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mld_vif *curr_mld_vif; + unsigned long curr_vif_active_links; + u16 link_id; + + data->global_low_latency |= iwl_mld_vif_low_latency(mld_vif); + + if ((ieee80211_vif_is_mld(vif) && vif->active_links) || + (vif->type != NL80211_IFTYPE_P2P_DEVICE && + mld_vif->deflink.active)) + data->active_vif = true; + + if (vif == curr_vif) + return; + + if (ieee80211_vif_type_p2p(vif) != NL80211_IFTYPE_P2P_GO) + return; + + /* Currently P2P GO can't be AP MLD so the logic below assumes that */ + WARN_ON_ONCE(ieee80211_vif_is_mld(vif)); + + curr_vif_active_links = + ieee80211_vif_is_mld(curr_vif) ? curr_vif->active_links : 1; + + curr_mld_vif = iwl_mld_vif_from_mac80211(curr_vif); + + for_each_set_bit(link_id, &curr_vif_active_links, + IEEE80211_MLD_MAX_NUM_LINKS) { + struct iwl_mld_link *curr_mld_link = + iwl_mld_link_dereference_check(curr_mld_vif, link_id); + + if (WARN_ON(!curr_mld_link)) + return; + + if (rcu_access_pointer(curr_mld_link->chan_ctx) && + rcu_access_pointer(mld_vif->deflink.chan_ctx) != + rcu_access_pointer(curr_mld_link->chan_ctx)) { + data->is_dcm_with_p2p_go = true; + return; + } + } +} + +static enum +iwl_mld_scan_type iwl_mld_get_scan_type(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct iwl_mld_scan_iter_data *data) +{ + enum iwl_mld_traffic_load load = mld->scan.traffic_load.status; + + /* A scanning AP interface probably wants to generate a survey to do + * ACS (automatic channel selection). + * Force a non-fragmented scan in that case. + */ + if (ieee80211_vif_type_p2p(vif) == NL80211_IFTYPE_AP) + return IWL_SCAN_TYPE_WILD; + + if (!data->active_vif) + return IWL_SCAN_TYPE_UNASSOC; + + if ((load == IWL_MLD_TRAFFIC_HIGH || data->global_low_latency) && + vif->type != NL80211_IFTYPE_P2P_DEVICE) + return IWL_SCAN_TYPE_FRAGMENTED; + + /* In case of DCM with P2P GO set all scan requests as + * fast-balance scan + */ + if (vif->type == NL80211_IFTYPE_STATION && + data->is_dcm_with_p2p_go) + return IWL_SCAN_TYPE_FAST_BALANCE; + + if (load >= IWL_MLD_TRAFFIC_MEDIUM || data->global_low_latency) + return IWL_SCAN_TYPE_MILD; + + return IWL_SCAN_TYPE_WILD; +} + +static u8 * +iwl_mld_scan_add_2ghz_elems(struct iwl_mld *mld, const u8 *ies, + size_t len, u8 *const pos) +{ + static const u8 before_ds_params[] = { + WLAN_EID_SSID, + WLAN_EID_SUPP_RATES, + WLAN_EID_REQUEST, + WLAN_EID_EXT_SUPP_RATES, + }; + size_t offs; + u8 *newpos = pos; + + offs = ieee80211_ie_split(ies, len, + before_ds_params, + ARRAY_SIZE(before_ds_params), + 0); + + memcpy(newpos, ies, offs); + newpos += offs; + + /* Add a placeholder for DS Parameter Set element */ + *newpos++ = WLAN_EID_DS_PARAMS; + *newpos++ = 1; + *newpos++ = 0; + + memcpy(newpos, ies + offs, len - offs); + newpos += len - offs; + + return newpos; +} + +static void +iwl_mld_scan_add_tpc_report_elem(u8 *pos) +{ + pos[0] = WLAN_EID_VENDOR_SPECIFIC; + pos[1] = WFA_TPC_IE_LEN - 2; + pos[2] = (WLAN_OUI_MICROSOFT >> 16) & 0xff; + pos[3] = (WLAN_OUI_MICROSOFT >> 8) & 0xff; + pos[4] = WLAN_OUI_MICROSOFT & 0xff; + pos[5] = WLAN_OUI_TYPE_MICROSOFT_TPC; + pos[6] = 0; + /* pos[7] - tx power will be inserted by the FW */ + pos[7] = 0; + pos[8] = 0; +} + +static u32 +iwl_mld_scan_ooc_priority(enum iwl_mld_scan_status scan_status) +{ + if (scan_status == IWL_MLD_SCAN_REGULAR) + return IWL_SCAN_PRIORITY_EXT_6; + if (scan_status == IWL_MLD_SCAN_INT_MLO) + return IWL_SCAN_PRIORITY_EXT_4; + + return IWL_SCAN_PRIORITY_EXT_2; +} + +static bool +iwl_mld_scan_is_regular(struct iwl_mld_scan_params *params) +{ + return params->n_scan_plans == 1 && + params->scan_plans[0].iterations == 1; +} + +static bool +iwl_mld_scan_is_fragmented(enum iwl_mld_scan_type type) +{ + return (type == IWL_SCAN_TYPE_FRAGMENTED || + type == IWL_SCAN_TYPE_FAST_BALANCE); +} + +static int +iwl_mld_scan_uid_by_status(struct iwl_mld *mld, int status) +{ + for (int i = 0; i < ARRAY_SIZE(mld->scan.uid_status); i++) + if (mld->scan.uid_status[i] == status) + return i; + + return -ENOENT; +} + +static const char * +iwl_mld_scan_ebs_status_str(enum iwl_scan_ebs_status status) +{ + switch (status) { + case IWL_SCAN_EBS_SUCCESS: + return "successful"; + case IWL_SCAN_EBS_INACTIVE: + return "inactive"; + case IWL_SCAN_EBS_FAILED: + case IWL_SCAN_EBS_CHAN_NOT_FOUND: + default: + return "failed"; + } +} + +static int +iwl_mld_scan_ssid_exist(u8 *ssid, u8 ssid_len, struct iwl_ssid_ie *ssid_list) +{ + for (int i = 0; i < PROBE_OPTION_MAX; i++) { + if (!ssid_list[i].len) + return -1; + if (ssid_list[i].len == ssid_len && + !memcmp(ssid_list[i].ssid, ssid, ssid_len)) + return i; + } + + return -1; +} + +static bool +iwl_mld_scan_fits(struct iwl_mld *mld, int n_ssids, + struct ieee80211_scan_ies *ies, int n_channels) +{ + return ((n_ssids <= PROBE_OPTION_MAX) && + (n_channels <= mld->fw->ucode_capa.n_scan_channels) & + (ies->common_ie_len + ies->len[NL80211_BAND_2GHZ] + + ies->len[NL80211_BAND_5GHZ] + ies->len[NL80211_BAND_6GHZ] <= + iwl_mld_scan_max_template_size())); +} + +static void +iwl_mld_scan_build_probe_req(struct iwl_mld *mld, struct ieee80211_vif *vif, + struct ieee80211_scan_ies *ies, + struct iwl_mld_scan_params *params) +{ + struct ieee80211_mgmt *frame = (void *)params->preq.buf; + u8 *pos, *newpos; + const u8 *mac_addr = params->flags & NL80211_SCAN_FLAG_RANDOM_ADDR ? + params->mac_addr : NULL; + + if (mac_addr) + get_random_mask_addr(frame->sa, mac_addr, + params->mac_addr_mask); + else + memcpy(frame->sa, vif->addr, ETH_ALEN); + + frame->frame_control = cpu_to_le16(IEEE80211_STYPE_PROBE_REQ); + eth_broadcast_addr(frame->da); + ether_addr_copy(frame->bssid, params->bssid); + frame->seq_ctrl = 0; + + pos = frame->u.probe_req.variable; + *pos++ = WLAN_EID_SSID; + *pos++ = 0; + + params->preq.mac_header.offset = 0; + params->preq.mac_header.len = cpu_to_le16(24 + 2); + + /* Insert DS parameter set element on 2.4 GHz band */ + newpos = iwl_mld_scan_add_2ghz_elems(mld, + ies->ies[NL80211_BAND_2GHZ], + ies->len[NL80211_BAND_2GHZ], + pos); + params->preq.band_data[0].offset = cpu_to_le16(pos - params->preq.buf); + params->preq.band_data[0].len = cpu_to_le16(newpos - pos); + pos = newpos; + + memcpy(pos, ies->ies[NL80211_BAND_5GHZ], + ies->len[NL80211_BAND_5GHZ]); + params->preq.band_data[1].offset = cpu_to_le16(pos - params->preq.buf); + params->preq.band_data[1].len = + cpu_to_le16(ies->len[NL80211_BAND_5GHZ]); + pos += ies->len[NL80211_BAND_5GHZ]; + + memcpy(pos, ies->ies[NL80211_BAND_6GHZ], + ies->len[NL80211_BAND_6GHZ]); + params->preq.band_data[2].offset = cpu_to_le16(pos - params->preq.buf); + params->preq.band_data[2].len = + cpu_to_le16(ies->len[NL80211_BAND_6GHZ]); + pos += ies->len[NL80211_BAND_6GHZ]; + + memcpy(pos, ies->common_ies, ies->common_ie_len); + params->preq.common_data.offset = cpu_to_le16(pos - params->preq.buf); + + iwl_mld_scan_add_tpc_report_elem(pos + ies->common_ie_len); + params->preq.common_data.len = cpu_to_le16(ies->common_ie_len + + WFA_TPC_IE_LEN); +} + +static u16 +iwl_mld_scan_get_cmd_gen_flags(struct iwl_mld *mld, + struct iwl_mld_scan_params *params, + struct ieee80211_vif *vif, + enum iwl_mld_scan_status scan_status) +{ + u16 flags = 0; + + /* If no direct SSIDs are provided perform a passive scan. Otherwise, + * if there is a single SSID which is not the broadcast SSID, assume + * that the scan is intended for roaming purposes and thus enable Rx on + * all chains to improve chances of hearing the beacons/probe responses. + */ + if (params->n_ssids == 0) + flags |= IWL_UMAC_SCAN_GEN_FLAGS_V2_FORCE_PASSIVE; + else if (params->n_ssids == 1 && params->ssids[0].ssid_len) + flags |= IWL_UMAC_SCAN_GEN_FLAGS_V2_USE_ALL_RX_CHAINS; + + if (params->pass_all) + flags |= IWL_UMAC_SCAN_GEN_FLAGS_V2_PASS_ALL; + else + flags |= IWL_UMAC_SCAN_GEN_FLAGS_V2_MATCH; + + if (iwl_mld_scan_is_fragmented(params->type)) + flags |= IWL_UMAC_SCAN_GEN_FLAGS_V2_FRAGMENTED_LMAC1; + + if (!iwl_mld_scan_is_regular(params)) + flags |= IWL_UMAC_SCAN_GEN_FLAGS_V2_PERIODIC; + + if (params->iter_notif || + mld->scan.pass_all_sched_res == SCHED_SCAN_PASS_ALL_STATE_ENABLED) + flags |= IWL_UMAC_SCAN_GEN_FLAGS_V2_NTFY_ITER_COMPLETE; + + if (scan_status == IWL_MLD_SCAN_SCHED || + scan_status == IWL_MLD_SCAN_NETDETECT) + flags |= IWL_UMAC_SCAN_GEN_FLAGS_V2_PREEMPTIVE; + + if (params->flags & (NL80211_SCAN_FLAG_ACCEPT_BCAST_PROBE_RESP | + NL80211_SCAN_FLAG_OCE_PROBE_REQ_HIGH_TX_RATE | + NL80211_SCAN_FLAG_FILS_MAX_CHANNEL_TIME)) + flags |= IWL_UMAC_SCAN_GEN_FLAGS_V2_OCE; + + if ((scan_status == IWL_MLD_SCAN_SCHED || + scan_status == IWL_MLD_SCAN_NETDETECT) && + params->flags & NL80211_SCAN_FLAG_COLOCATED_6GHZ) + flags |= IWL_UMAC_SCAN_GEN_FLAGS_V2_TRIGGER_UHB_SCAN; + + if (params->enable_6ghz_passive) + flags |= IWL_UMAC_SCAN_GEN_FLAGS_V2_6GHZ_PASSIVE_SCAN; + + flags |= IWL_UMAC_SCAN_GEN_FLAGS_V2_ADAPTIVE_DWELL; + + return flags; +} + +static u8 +iwl_mld_scan_get_cmd_gen_flags2(struct iwl_mld *mld, + struct iwl_mld_scan_params *params, + struct ieee80211_vif *vif, u16 gen_flags) +{ + u8 flags = 0; + + /* TODO: CDB */ + if (params->respect_p2p_go) + flags |= IWL_UMAC_SCAN_GEN_PARAMS_FLAGS2_RESPECT_P2P_GO_LB | + IWL_UMAC_SCAN_GEN_PARAMS_FLAGS2_RESPECT_P2P_GO_HB; + + if (params->scan_6ghz) + flags |= IWL_UMAC_SCAN_GEN_PARAMS_FLAGS2_DONT_TOGGLE_ANT; + + return flags; +} + +static void +iwl_mld_scan_cmd_set_dwell(struct iwl_mld *mld, + struct iwl_scan_general_params_v11 *gp, + struct iwl_mld_scan_params *params) +{ + const struct iwl_mld_scan_timing_params *timing = + &scan_timing[params->type]; + + gp->adwell_default_social_chn = + IWL_SCAN_ADWELL_DEFAULT_N_APS_SOCIAL; + gp->adwell_default_2g = IWL_SCAN_ADWELL_DEFAULT_LB_N_APS; + gp->adwell_default_5g = IWL_SCAN_ADWELL_DEFAULT_HB_N_APS; + + if (params->n_ssids && params->ssids[0].ssid_len) + gp->adwell_max_budget = + cpu_to_le16(IWL_SCAN_ADWELL_MAX_BUDGET_DIRECTED_SCAN); + else + gp->adwell_max_budget = + cpu_to_le16(IWL_SCAN_ADWELL_MAX_BUDGET_FULL_SCAN); + + gp->scan_priority = cpu_to_le32(IWL_SCAN_PRIORITY_EXT_6); + + gp->max_out_of_time[SCAN_LB_LMAC_IDX] = cpu_to_le32(timing->max_out_time); + gp->suspend_time[SCAN_LB_LMAC_IDX] = cpu_to_le32(timing->suspend_time); + + gp->active_dwell[SCAN_LB_LMAC_IDX] = IWL_SCAN_DWELL_ACTIVE; + gp->passive_dwell[SCAN_LB_LMAC_IDX] = IWL_SCAN_DWELL_PASSIVE; + gp->active_dwell[SCAN_HB_LMAC_IDX] = IWL_SCAN_DWELL_ACTIVE; + gp->passive_dwell[SCAN_HB_LMAC_IDX] = IWL_SCAN_DWELL_PASSIVE; + + IWL_DEBUG_SCAN(mld, + "Scan: adwell_max_budget=%d max_out_of_time=%d suspend_time=%d\n", + gp->adwell_max_budget, + gp->max_out_of_time[SCAN_LB_LMAC_IDX], + gp->suspend_time[SCAN_LB_LMAC_IDX]); +} + +static void +iwl_mld_scan_cmd_set_gen_params(struct iwl_mld *mld, + struct iwl_mld_scan_params *params, + struct ieee80211_vif *vif, + struct iwl_scan_general_params_v11 *gp, + enum iwl_mld_scan_status scan_status) +{ + u16 gen_flags = iwl_mld_scan_get_cmd_gen_flags(mld, params, vif, + scan_status); + u8 gen_flags2 = iwl_mld_scan_get_cmd_gen_flags2(mld, params, vif, + gen_flags); + + IWL_DEBUG_SCAN(mld, "General: flags=0x%x, flags2=0x%x\n", + gen_flags, gen_flags2); + + gp->flags = cpu_to_le16(gen_flags); + gp->flags2 = gen_flags2; + + iwl_mld_scan_cmd_set_dwell(mld, gp, params); + + if (gen_flags & IWL_UMAC_SCAN_GEN_FLAGS_V2_FRAGMENTED_LMAC1) + gp->num_of_fragments[SCAN_LB_LMAC_IDX] = IWL_SCAN_NUM_OF_FRAGS; + + if (params->fw_link_id != IWL_MLD_INVALID_FW_ID) + gp->scan_start_mac_or_link_id = params->fw_link_id; +} + +static int +iwl_mld_scan_cmd_set_sched_params(struct iwl_mld_scan_params *params, + struct iwl_scan_umac_schedule *schedule, + __le16 *delay) +{ + if (WARN_ON(!params->n_scan_plans || + params->n_scan_plans > IWL_MAX_SCHED_SCAN_PLANS)) + return -EINVAL; + + for (int i = 0; i < params->n_scan_plans; i++) { + struct cfg80211_sched_scan_plan *scan_plan = + ¶ms->scan_plans[i]; + + schedule[i].iter_count = scan_plan->iterations; + schedule[i].interval = + cpu_to_le16(scan_plan->interval); + } + + /* If the number of iterations of the last scan plan is set to zero, + * it should run infinitely. However, this is not always the case. + * For example, when regular scan is requested the driver sets one scan + * plan with one iteration. + */ + if (!schedule[params->n_scan_plans - 1].iter_count) + schedule[params->n_scan_plans - 1].iter_count = 0xff; + + *delay = cpu_to_le16(params->delay); + + return 0; +} + +/* We insert the SSIDs in an inverted order, because the FW will + * invert it back. + */ +static void +iwl_mld_scan_cmd_build_ssids(struct iwl_mld_scan_params *params, + struct iwl_ssid_ie *ssids, u32 *ssid_bitmap) +{ + int i, j; + int index; + u32 tmp_bitmap = 0; + + /* copy SSIDs from match list. iwl_config_sched_scan_profiles() + * uses the order of these ssids to config match list. + */ + for (i = 0, j = params->n_match_sets - 1; + j >= 0 && i < PROBE_OPTION_MAX; + i++, j--) { + /* skip empty SSID match_sets */ + if (!params->match_sets[j].ssid.ssid_len) + continue; + + ssids[i].id = WLAN_EID_SSID; + ssids[i].len = params->match_sets[j].ssid.ssid_len; + memcpy(ssids[i].ssid, params->match_sets[j].ssid.ssid, + ssids[i].len); + } + + /* add SSIDs from scan SSID list */ + for (j = params->n_ssids - 1; + j >= 0 && i < PROBE_OPTION_MAX; + i++, j--) { + index = iwl_mld_scan_ssid_exist(params->ssids[j].ssid, + params->ssids[j].ssid_len, + ssids); + if (index < 0) { + ssids[i].id = WLAN_EID_SSID; + ssids[i].len = params->ssids[j].ssid_len; + memcpy(ssids[i].ssid, params->ssids[j].ssid, + ssids[i].len); + tmp_bitmap |= BIT(i); + } else { + tmp_bitmap |= BIT(index); + } + } + + if (ssid_bitmap) + *ssid_bitmap = tmp_bitmap; +} + +static void +iwl_mld_scan_fill_6g_chan_list(struct iwl_mld_scan_params *params, + struct iwl_scan_probe_params_v4 *pp) +{ + int j, idex_s = 0, idex_b = 0; + struct cfg80211_scan_6ghz_params *scan_6ghz_params = + params->scan_6ghz_params; + + for (j = 0; + j < params->n_ssids && idex_s < SCAN_SHORT_SSID_MAX_SIZE; + j++) { + if (!params->ssids[j].ssid_len) + continue; + + pp->short_ssid[idex_s] = + cpu_to_le32(~crc32_le(~0, params->ssids[j].ssid, + params->ssids[j].ssid_len)); + + /* hidden 6ghz scan */ + pp->direct_scan[idex_s].id = WLAN_EID_SSID; + pp->direct_scan[idex_s].len = params->ssids[j].ssid_len; + memcpy(pp->direct_scan[idex_s].ssid, params->ssids[j].ssid, + params->ssids[j].ssid_len); + idex_s++; + } + + /* Populate the arrays of the short SSIDs and the BSSIDs using the 6GHz + * collocated parameters. This might not be optimal, as this processing + * does not (yet) correspond to the actual channels, so it is possible + * that some entries would be left out. + */ + for (j = 0; j < params->n_6ghz_params; j++) { + int k; + + /* First, try to place the short SSID */ + if (scan_6ghz_params[j].short_ssid_valid) { + for (k = 0; k < idex_s; k++) { + if (pp->short_ssid[k] == + cpu_to_le32(scan_6ghz_params[j].short_ssid)) + break; + } + + if (k == idex_s && idex_s < SCAN_SHORT_SSID_MAX_SIZE) { + pp->short_ssid[idex_s++] = + cpu_to_le32(scan_6ghz_params[j].short_ssid); + } + } + + /* try to place BSSID for the same entry */ + for (k = 0; k < idex_b; k++) { + if (!memcmp(&pp->bssid_array[k], + scan_6ghz_params[j].bssid, ETH_ALEN)) + break; + } + + if (k == idex_b && idex_b < SCAN_BSSID_MAX_SIZE && + !WARN_ONCE(!is_valid_ether_addr(scan_6ghz_params[j].bssid), + "scan: invalid BSSID at index %u, index_b=%u\n", + j, idex_b)) { + memcpy(&pp->bssid_array[idex_b++], + scan_6ghz_params[j].bssid, ETH_ALEN); + } + } + + pp->short_ssid_num = idex_s; + pp->bssid_num = idex_b; +} + +static void +iwl_mld_scan_cmd_set_probe_params(struct iwl_mld_scan_params *params, + struct iwl_scan_probe_params_v4 *pp, + u32 *bitmap_ssid) +{ + pp->preq = params->preq; + + if (params->scan_6ghz) { + iwl_mld_scan_fill_6g_chan_list(params, pp); + return; + } + + /* relevant only for 2.4 GHz /5 GHz scan */ + iwl_mld_scan_cmd_build_ssids(params, pp->direct_scan, bitmap_ssid); +} + +static bool +iwl_mld_scan_use_ebs(struct iwl_mld *mld, struct ieee80211_vif *vif, + bool low_latency) +{ + const struct iwl_ucode_capabilities *capa = &mld->fw->ucode_capa; + + /* We can only use EBS if: + * 1. the feature is supported. + * 2. the last EBS was successful. + * 3. it's not a p2p find operation. + * 4. we are not in low latency mode, + * or if fragmented ebs is supported by the FW + * 5. the VIF is not an AP interface (scan wants survey results) + */ + return ((capa->flags & IWL_UCODE_TLV_FLAGS_EBS_SUPPORT) && + !mld->scan.last_ebs_failed && + vif->type != NL80211_IFTYPE_P2P_DEVICE && + (!low_latency || fw_has_api(capa, IWL_UCODE_TLV_API_FRAG_EBS)) && + ieee80211_vif_type_p2p(vif) != NL80211_IFTYPE_AP); +} + +static u8 +iwl_mld_scan_cmd_set_chan_flags(struct iwl_mld *mld, + struct iwl_mld_scan_params *params, + struct ieee80211_vif *vif, + bool low_latency) +{ + u8 flags = 0; + + flags |= IWL_SCAN_CHANNEL_FLAG_ENABLE_CHAN_ORDER; + + if (iwl_mld_scan_use_ebs(mld, vif, low_latency)) + flags |= IWL_SCAN_CHANNEL_FLAG_EBS | + IWL_SCAN_CHANNEL_FLAG_EBS_ACCURATE | + IWL_SCAN_CHANNEL_FLAG_CACHE_ADD; + + /* set fragmented ebs for fragmented scan */ + if (iwl_mld_scan_is_fragmented(params->type)) + flags |= IWL_SCAN_CHANNEL_FLAG_EBS_FRAG; + + /* Force EBS in case the scan is a fragmented and there is a need + * to take P2P GO operation into consideration during scan operation. + */ + /* TODO: CDB */ + if (iwl_mld_scan_is_fragmented(params->type) && + params->respect_p2p_go) { + IWL_DEBUG_SCAN(mld, "Respect P2P GO. Force EBS\n"); + flags |= IWL_SCAN_CHANNEL_FLAG_FORCE_EBS; + } + + return flags; +} + +static const u8 p2p_go_friendly_chs[] = { + 36, 40, 44, 48, 149, 153, 157, 161, 165, +}; + +static const u8 social_chs[] = { + 1, 6, 11 +}; + +static u32 iwl_mld_scan_ch_n_aps_flag(enum nl80211_iftype vif_type, u8 ch_id) +{ + if (vif_type != NL80211_IFTYPE_P2P_DEVICE) + return 0; + + for (int i = 0; i < ARRAY_SIZE(p2p_go_friendly_chs); i++) { + if (ch_id == p2p_go_friendly_chs[i]) + return IWL_SCAN_ADWELL_N_APS_GO_FRIENDLY_BIT; + } + + for (int i = 0; i < ARRAY_SIZE(social_chs); i++) { + if (ch_id == social_chs[i]) + return IWL_SCAN_ADWELL_N_APS_SOCIAL_CHS_BIT; + } + + return 0; +} + +static void +iwl_mld_scan_cmd_set_channels(struct iwl_mld *mld, + struct ieee80211_channel **channels, + struct iwl_scan_channel_params_v7 *cp, + int n_channels, u32 flags, + enum nl80211_iftype vif_type) +{ + for (int i = 0; i < n_channels; i++) { + enum nl80211_band band = channels[i]->band; + struct iwl_scan_channel_cfg_umac *cfg = &cp->channel_config[i]; + u8 iwl_band = iwl_mld_nl80211_band_to_fw(band); + u32 n_aps_flag = + iwl_mld_scan_ch_n_aps_flag(vif_type, + channels[i]->hw_value); + + if (IWL_MLD_ADAPTIVE_DWELL_NUM_APS_OVERRIDE) + n_aps_flag = IWL_SCAN_ADWELL_N_APS_GO_FRIENDLY_BIT; + + cfg->flags = cpu_to_le32(flags | n_aps_flag); + cfg->channel_num = channels[i]->hw_value; + if (cfg80211_channel_is_psc(channels[i])) + cfg->flags = 0; + + if (band == NL80211_BAND_6GHZ) { + /* 6 GHz channels should only appear in a scan request + * that has scan_6ghz set. The only exception is MLO + * scan, which has to be passive. + */ + WARN_ON_ONCE(cfg->flags != 0); + cfg->flags = + cpu_to_le32(IWL_UHB_CHAN_CFG_FLAG_FORCE_PASSIVE); + } + + cfg->v2.iter_count = 1; + cfg->v2.iter_interval = 0; + cfg->flags |= cpu_to_le32(iwl_band << + IWL_CHAN_CFG_FLAGS_BAND_POS); + } +} + +static u8 +iwl_mld_scan_cfg_channels_6g(struct iwl_mld *mld, + struct iwl_mld_scan_params *params, + u32 n_channels, + struct iwl_scan_probe_params_v4 *pp, + struct iwl_scan_channel_params_v7 *cp, + enum nl80211_iftype vif_type) +{ + struct cfg80211_scan_6ghz_params *scan_6ghz_params = + params->scan_6ghz_params; + u32 i; + u8 ch_cnt; + + for (i = 0, ch_cnt = 0; i < params->n_channels; i++) { + struct iwl_scan_channel_cfg_umac *cfg = + &cp->channel_config[ch_cnt]; + + u32 s_ssid_bitmap = 0, bssid_bitmap = 0, flags = 0; + u8 k, n_s_ssids = 0, n_bssids = 0; + u8 max_s_ssids, max_bssids; + bool force_passive = false, found = false, allow_passive = true, + unsolicited_probe_on_chan = false, psc_no_listen = false; + s8 psd_20 = IEEE80211_RNR_TBTT_PARAMS_PSD_RESERVED; + + /* Avoid performing passive scan on non PSC channels unless the + * scan is specifically a passive scan, i.e., no SSIDs + * configured in the scan command. + */ + if (!cfg80211_channel_is_psc(params->channels[i]) && + !params->n_6ghz_params && params->n_ssids) + continue; + + cfg->channel_num = params->channels[i]->hw_value; + cfg->flags |= + cpu_to_le32(PHY_BAND_6 << IWL_CHAN_CFG_FLAGS_BAND_POS); + + cfg->v5.iter_count = 1; + cfg->v5.iter_interval = 0; + + for (u32 j = 0; j < params->n_6ghz_params; j++) { + s8 tmp_psd_20; + + if (!(scan_6ghz_params[j].channel_idx == i)) + continue; + + unsolicited_probe_on_chan |= + scan_6ghz_params[j].unsolicited_probe; + + /* Use the highest PSD value allowed as advertised by + * APs for this channel + */ + tmp_psd_20 = scan_6ghz_params[j].psd_20; + if (tmp_psd_20 != + IEEE80211_RNR_TBTT_PARAMS_PSD_RESERVED && + (psd_20 == + IEEE80211_RNR_TBTT_PARAMS_PSD_RESERVED || + psd_20 < tmp_psd_20)) + psd_20 = tmp_psd_20; + + psc_no_listen |= scan_6ghz_params[j].psc_no_listen; + } + + /* In the following cases apply passive scan: + * 1. Non fragmented scan: + * - PSC channel with NO_LISTEN_FLAG on should be treated + * like non PSC channel + * - Non PSC channel with more than 3 short SSIDs or more + * than 9 BSSIDs. + * - Non PSC Channel with unsolicited probe response and + * more than 2 short SSIDs or more than 6 BSSIDs. + * - PSC channel with more than 2 short SSIDs or more than + * 6 BSSIDs. + * 2. Fragmented scan: + * - PSC channel with more than 1 SSID or 3 BSSIDs. + * - Non PSC channel with more than 2 SSIDs or 6 BSSIDs. + * - Non PSC channel with unsolicited probe response and + * more than 1 SSID or more than 3 BSSIDs. + */ + if (!iwl_mld_scan_is_fragmented(params->type)) { + if (!cfg80211_channel_is_psc(params->channels[i]) || + psc_no_listen) { + if (unsolicited_probe_on_chan) { + max_s_ssids = 2; + max_bssids = 6; + } else { + max_s_ssids = 3; + max_bssids = 9; + } + } else { + max_s_ssids = 2; + max_bssids = 6; + } + } else if (cfg80211_channel_is_psc(params->channels[i])) { + max_s_ssids = 1; + max_bssids = 3; + } else { + if (unsolicited_probe_on_chan) { + max_s_ssids = 1; + max_bssids = 3; + } else { + max_s_ssids = 2; + max_bssids = 6; + } + } + + /* To optimize the scan time, i.e., reduce the scan dwell time + * on each channel, the below logic tries to set 3 direct BSSID + * probe requests for each broadcast probe request with a short + * SSID. + */ + for (u32 j = 0; j < params->n_6ghz_params; j++) { + if (!(scan_6ghz_params[j].channel_idx == i)) + continue; + + found = false; + + for (k = 0; + k < pp->short_ssid_num && n_s_ssids < max_s_ssids; + k++) { + if (!scan_6ghz_params[j].unsolicited_probe && + le32_to_cpu(pp->short_ssid[k]) == + scan_6ghz_params[j].short_ssid) { + /* Relevant short SSID bit set */ + if (s_ssid_bitmap & BIT(k)) { + found = true; + break; + } + + /* Prefer creating BSSID entries unless + * the short SSID probe can be done in + * the same channel dwell iteration. + * + * We also need to create a short SSID + * entry for any hidden AP. + */ + if (3 * n_s_ssids > n_bssids && + !pp->direct_scan[k].len) + break; + + /* Hidden AP, cannot do passive scan */ + if (pp->direct_scan[k].len) + allow_passive = false; + + s_ssid_bitmap |= BIT(k); + n_s_ssids++; + found = true; + break; + } + } + + if (found) + continue; + + for (k = 0; k < pp->bssid_num; k++) { + if (memcmp(&pp->bssid_array[k], + scan_6ghz_params[j].bssid, + ETH_ALEN)) + continue; + + if (bssid_bitmap & BIT(k)) + break; + + if (n_bssids < max_bssids) { + bssid_bitmap |= BIT(k); + n_bssids++; + } else { + force_passive = TRUE; + } + + break; + } + } + + if (cfg80211_channel_is_psc(params->channels[i]) && + psc_no_listen) + flags |= IWL_UHB_CHAN_CFG_FLAG_PSC_CHAN_NO_LISTEN; + + if (unsolicited_probe_on_chan) + flags |= IWL_UHB_CHAN_CFG_FLAG_UNSOLICITED_PROBE_RES; + + if ((allow_passive && force_passive) || + (!(bssid_bitmap | s_ssid_bitmap) && + !cfg80211_channel_is_psc(params->channels[i]))) + flags |= IWL_UHB_CHAN_CFG_FLAG_FORCE_PASSIVE; + else + flags |= bssid_bitmap | (s_ssid_bitmap << 16); + + cfg->flags |= cpu_to_le32(flags); + cfg->v5.psd_20 = psd_20; + + ch_cnt++; + } + + if (params->n_channels > ch_cnt) + IWL_DEBUG_SCAN(mld, + "6GHz: reducing number channels: (%u->%u)\n", + params->n_channels, ch_cnt); + + return ch_cnt; +} + +static int +iwl_mld_scan_cmd_set_6ghz_chan_params(struct iwl_mld *mld, + struct iwl_mld_scan_params *params, + struct ieee80211_vif *vif, + struct iwl_scan_req_params_v17 *scan_p, + enum iwl_mld_scan_status scan_status) +{ + struct iwl_scan_channel_params_v7 *chan_p = &scan_p->channel_params; + struct iwl_scan_probe_params_v4 *probe_p = &scan_p->probe_params; + + chan_p->flags = iwl_mld_scan_get_cmd_gen_flags(mld, params, vif, + scan_status); + chan_p->count = iwl_mld_scan_cfg_channels_6g(mld, params, + params->n_channels, + probe_p, chan_p, + vif->type); + if (!chan_p->count) + return -EINVAL; + + if (!params->n_ssids || + (params->n_ssids == 1 && !params->ssids[0].ssid_len)) + chan_p->flags |= IWL_SCAN_CHANNEL_FLAG_6G_PSC_NO_FILTER; + + return 0; +} + +static int +iwl_mld_scan_cmd_set_chan_params(struct iwl_mld *mld, + struct iwl_mld_scan_params *params, + struct ieee80211_vif *vif, + struct iwl_scan_req_params_v17 *scan_p, + bool low_latency, + enum iwl_mld_scan_status scan_status, + u32 channel_cfg_flags) +{ + struct iwl_scan_channel_params_v7 *cp = &scan_p->channel_params; + struct ieee80211_supported_band *sband = + &mld->nvm_data->bands[NL80211_BAND_6GHZ]; + + cp->n_aps_override[0] = IWL_SCAN_ADWELL_N_APS_GO_FRIENDLY; + cp->n_aps_override[1] = IWL_SCAN_ADWELL_N_APS_SOCIAL_CHS; + + if (IWL_MLD_ADAPTIVE_DWELL_NUM_APS_OVERRIDE) + cp->n_aps_override[0] = IWL_MLD_ADAPTIVE_DWELL_NUM_APS_OVERRIDE; + + if (params->scan_6ghz) + return iwl_mld_scan_cmd_set_6ghz_chan_params(mld, params, + vif, scan_p, + scan_status); + + /* relevant only for 2.4 GHz/5 GHz scan */ + cp->flags = iwl_mld_scan_cmd_set_chan_flags(mld, params, vif, + low_latency); + cp->count = params->n_channels; + + iwl_mld_scan_cmd_set_channels(mld, params->channels, cp, + params->n_channels, channel_cfg_flags, + vif->type); + + if (!params->enable_6ghz_passive) + return 0; + + /* fill 6 GHz passive scan cfg */ + for (int i = 0; i < sband->n_channels; i++) { + struct ieee80211_channel *channel = + &sband->channels[i]; + struct iwl_scan_channel_cfg_umac *cfg = + &cp->channel_config[cp->count]; + + if (!cfg80211_channel_is_psc(channel)) + continue; + + cfg->channel_num = channel->hw_value; + cfg->v5.iter_count = 1; + cfg->v5.iter_interval = 0; + cfg->v5.psd_20 = + IEEE80211_RNR_TBTT_PARAMS_PSD_RESERVED; + cfg->flags = cpu_to_le32(PHY_BAND_6 << + IWL_CHAN_CFG_FLAGS_BAND_POS); + cp->count++; + } + + return 0; +} + +static int +iwl_mld_scan_build_cmd(struct iwl_mld *mld, struct ieee80211_vif *vif, + struct iwl_mld_scan_params *params, + enum iwl_mld_scan_status scan_status, + bool low_latency) +{ + struct iwl_scan_req_umac_v17 *cmd = mld->scan.cmd; + struct iwl_scan_req_params_v17 *scan_p = &cmd->scan_params; + u32 bitmap_ssid = 0; + int uid, ret; + + memset(mld->scan.cmd, 0, mld->scan.cmd_size); + + /* find a free UID entry */ + uid = iwl_mld_scan_uid_by_status(mld, IWL_MLD_SCAN_NONE); + if (uid < 0) + return uid; + + cmd->uid = cpu_to_le32(uid); + cmd->ooc_priority = + cpu_to_le32(iwl_mld_scan_ooc_priority(scan_status)); + + iwl_mld_scan_cmd_set_gen_params(mld, params, vif, + &scan_p->general_params, scan_status); + + ret = iwl_mld_scan_cmd_set_sched_params(params, + scan_p->periodic_params.schedule, + &scan_p->periodic_params.delay); + if (ret) + return ret; + + iwl_mld_scan_cmd_set_probe_params(params, &scan_p->probe_params, + &bitmap_ssid); + + ret = iwl_mld_scan_cmd_set_chan_params(mld, params, vif, scan_p, + low_latency, scan_status, + bitmap_ssid); + if (ret) + return ret; + + return uid; +} + +static bool +iwl_mld_scan_pass_all(struct iwl_mld *mld, + struct cfg80211_sched_scan_request *req) +{ + if (req->n_match_sets && req->match_sets[0].ssid.ssid_len) { + IWL_DEBUG_SCAN(mld, + "Sending scheduled scan with filtering, n_match_sets %d\n", + req->n_match_sets); + mld->scan.pass_all_sched_res = SCHED_SCAN_PASS_ALL_STATE_DISABLED; + return false; + } + + IWL_DEBUG_SCAN(mld, "Sending Scheduled scan without filtering\n"); + mld->scan.pass_all_sched_res = SCHED_SCAN_PASS_ALL_STATE_ENABLED; + + return true; +} + +static int +iwl_mld_config_sched_scan_profiles(struct iwl_mld *mld, + struct cfg80211_sched_scan_request *req) +{ + struct iwl_host_cmd hcmd = { + .id = SCAN_OFFLOAD_UPDATE_PROFILES_CMD, + .dataflags[0] = IWL_HCMD_DFL_NOCOPY, + }; + struct iwl_scan_offload_profile *profile; + struct iwl_scan_offload_profile_cfg_data *cfg_data; + struct iwl_scan_offload_profile_cfg *profile_cfg; + struct iwl_scan_offload_blocklist *blocklist; + u32 blocklist_size = IWL_SCAN_MAX_BLACKLIST_LEN * sizeof(*blocklist); + u32 cmd_size = blocklist_size + sizeof(*profile_cfg); + u8 *cmd; + int ret; + + if (WARN_ON(req->n_match_sets > IWL_SCAN_MAX_PROFILES_V2)) + return -EIO; + + cmd = kzalloc(cmd_size, GFP_KERNEL); + if (!cmd) + return -ENOMEM; + + hcmd.data[0] = cmd; + hcmd.len[0] = cmd_size; + + blocklist = (struct iwl_scan_offload_blocklist *)cmd; + profile_cfg = (struct iwl_scan_offload_profile_cfg *)(cmd + blocklist_size); + + /* No blocklist configuration */ + cfg_data = &profile_cfg->data; + cfg_data->num_profiles = req->n_match_sets; + cfg_data->active_clients = SCAN_CLIENT_SCHED_SCAN; + cfg_data->pass_match = SCAN_CLIENT_SCHED_SCAN; + cfg_data->match_notify = SCAN_CLIENT_SCHED_SCAN; + + if (!req->n_match_sets || !req->match_sets[0].ssid.ssid_len) + cfg_data->any_beacon_notify = SCAN_CLIENT_SCHED_SCAN; + + for (int i = 0; i < req->n_match_sets; i++) { + profile = &profile_cfg->profiles[i]; + + /* Support any cipher and auth algorithm */ + profile->unicast_cipher = 0xff; + profile->auth_alg = IWL_AUTH_ALGO_UNSUPPORTED | + IWL_AUTH_ALGO_NONE | IWL_AUTH_ALGO_PSK | + IWL_AUTH_ALGO_8021X | IWL_AUTH_ALGO_SAE | + IWL_AUTH_ALGO_8021X_SHA384 | IWL_AUTH_ALGO_OWE; + profile->network_type = IWL_NETWORK_TYPE_ANY; + profile->band_selection = IWL_SCAN_OFFLOAD_SELECT_ANY; + profile->client_bitmap = SCAN_CLIENT_SCHED_SCAN; + profile->ssid_index = i; + } + + IWL_DEBUG_SCAN(mld, + "Sending scheduled scan profile config (n_match_sets=%u)\n", + req->n_match_sets); + + ret = iwl_mld_send_cmd(mld, &hcmd); + + kfree(cmd); + + return ret; +} + +static int +iwl_mld_sched_scan_handle_non_psc_channels(struct iwl_mld_scan_params *params, + bool *non_psc_included) +{ + int i, j; + + *non_psc_included = false; + /* for 6 GHZ band only PSC channels need to be added */ + for (i = 0; i < params->n_channels; i++) { + struct ieee80211_channel *channel = params->channels[i]; + + if (channel->band == NL80211_BAND_6GHZ && + !cfg80211_channel_is_psc(channel)) { + *non_psc_included = true; + break; + } + } + + if (!*non_psc_included) + return 0; + + params->channels = + kmemdup(params->channels, + sizeof(params->channels[0]) * params->n_channels, + GFP_KERNEL); + if (!params->channels) + return -ENOMEM; + + for (i = j = 0; i < params->n_channels; i++) { + if (params->channels[i]->band == NL80211_BAND_6GHZ && + !cfg80211_channel_is_psc(params->channels[i])) + continue; + params->channels[j++] = params->channels[i]; + } + + params->n_channels = j; + + return 0; +} + +static void +iwl_mld_scan_6ghz_passive_scan(struct iwl_mld *mld, + struct iwl_mld_scan_params *params, + struct ieee80211_vif *vif) +{ + struct ieee80211_supported_band *sband = + &mld->nvm_data->bands[NL80211_BAND_6GHZ]; + u32 n_disabled, i; + + params->enable_6ghz_passive = false; + + /* 6 GHz passive scan may be enabled in the first 2.4 GHz/5 GHz scan + * phase to discover geo location if no AP's are found. Skip it when + * we're in the 6 GHz scan phase. + */ + if (params->scan_6ghz) + return; + + /* 6 GHz passive scan allowed only on station interface */ + if (vif->type != NL80211_IFTYPE_STATION) { + IWL_DEBUG_SCAN(mld, + "6GHz passive scan: not station interface\n"); + return; + } + + /* 6 GHz passive scan is allowed in a defined time interval following + * HW reset or resume flow, or while not associated and a large + * interval has passed since the last 6 GHz passive scan. + */ + if ((vif->cfg.assoc || + time_after(mld->scan.last_6ghz_passive_jiffies + + (IWL_MLD_6GHZ_PASSIVE_SCAN_TIMEOUT * HZ), jiffies)) && + (time_before(mld->scan.last_start_time_jiffies + + (IWL_MLD_6GHZ_PASSIVE_SCAN_ASSOC_TIMEOUT * HZ), + jiffies))) { + IWL_DEBUG_SCAN(mld, "6GHz passive scan: %s\n", + vif->cfg.assoc ? "associated" : + "timeout did not expire"); + return; + } + + /* not enough channels in the regular scan request */ + if (params->n_channels < IWL_MLD_6GHZ_PASSIVE_SCAN_MIN_CHANS) { + IWL_DEBUG_SCAN(mld, + "6GHz passive scan: not enough channels %d\n", + params->n_channels); + return; + } + + for (i = 0; i < params->n_ssids; i++) { + if (!params->ssids[i].ssid_len) + break; + } + + /* not a wildcard scan, so cannot enable passive 6 GHz scan */ + if (i == params->n_ssids) { + IWL_DEBUG_SCAN(mld, + "6GHz passive scan: no wildcard SSID\n"); + return; + } + + if (!sband || !sband->n_channels) { + IWL_DEBUG_SCAN(mld, + "6GHz passive scan: no 6GHz channels\n"); + return; + } + + for (i = 0, n_disabled = 0; i < sband->n_channels; i++) { + if (sband->channels[i].flags & (IEEE80211_CHAN_DISABLED)) + n_disabled++; + } + + /* Not all the 6 GHz channels are disabled, so no need for 6 GHz + * passive scan + */ + if (n_disabled != sband->n_channels) { + IWL_DEBUG_SCAN(mld, + "6GHz passive scan: 6GHz channels enabled\n"); + return; + } + + /* all conditions to enable 6 GHz passive scan are satisfied */ + IWL_DEBUG_SCAN(mld, "6GHz passive scan: can be enabled\n"); + params->enable_6ghz_passive = true; +} + +static void +iwl_mld_scan_set_link_id(struct iwl_mld *mld, struct ieee80211_vif *vif, + struct iwl_mld_scan_params *params, + s8 tsf_report_link_id, + enum iwl_mld_scan_status scan_status) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mld_link *link; + + if (tsf_report_link_id < 0) { + if (vif->active_links) + tsf_report_link_id = __ffs(vif->active_links); + else + tsf_report_link_id = 0; + } + + link = iwl_mld_link_dereference_check(mld_vif, tsf_report_link_id); + if (!WARN_ON(!link)) { + params->fw_link_id = link->fw_id; + /* we to store fw_link_id only for regular scan, + * and use it in scan complete notif + */ + if (scan_status == IWL_MLD_SCAN_REGULAR) + mld->scan.fw_link_id = link->fw_id; + } else { + mld->scan.fw_link_id = IWL_MLD_INVALID_FW_ID; + params->fw_link_id = IWL_MLD_INVALID_FW_ID; + } +} + +static int +_iwl_mld_single_scan_start(struct iwl_mld *mld, struct ieee80211_vif *vif, + struct cfg80211_scan_request *req, + struct ieee80211_scan_ies *ies, + enum iwl_mld_scan_status scan_status) +{ + struct iwl_host_cmd hcmd = { + .id = WIDE_ID(LONG_GROUP, SCAN_REQ_UMAC), + .len = { mld->scan.cmd_size, }, + .data = { mld->scan.cmd, }, + .dataflags = { IWL_HCMD_DFL_NOCOPY, }, + }; + struct iwl_mld_scan_iter_data scan_iter_data = { + .current_vif = vif, + }; + struct cfg80211_sched_scan_plan scan_plan = {.iterations = 1}; + struct iwl_mld_scan_params params = {}; + int ret, uid; + + /* we should have failed registration if scan_cmd was NULL */ + if (WARN_ON(!mld->scan.cmd)) + return -ENOMEM; + + if (!iwl_mld_scan_fits(mld, req->n_ssids, ies, req->n_channels)) + return -ENOBUFS; + + ieee80211_iterate_active_interfaces_mtx(mld->hw, + IEEE80211_IFACE_ITER_NORMAL, + iwl_mld_scan_iterator, + &scan_iter_data); + + params.type = iwl_mld_get_scan_type(mld, vif, &scan_iter_data); + params.n_ssids = req->n_ssids; + params.flags = req->flags; + params.n_channels = req->n_channels; + params.delay = 0; + params.ssids = req->ssids; + params.channels = req->channels; + params.mac_addr = req->mac_addr; + params.mac_addr_mask = req->mac_addr_mask; + params.no_cck = req->no_cck; + params.pass_all = true; + params.n_match_sets = 0; + params.match_sets = NULL; + params.scan_plans = &scan_plan; + params.n_scan_plans = 1; + + params.n_6ghz_params = req->n_6ghz_params; + params.scan_6ghz_params = req->scan_6ghz_params; + params.scan_6ghz = req->scan_6ghz; + + ether_addr_copy(params.bssid, req->bssid); + /* TODO: CDB - per-band flag */ + params.respect_p2p_go = + iwl_mld_get_respect_p2p_go(mld, vif, + scan_iter_data.global_low_latency); + + if (req->duration) + params.iter_notif = true; + + iwl_mld_scan_set_link_id(mld, vif, ¶ms, req->tsf_report_link_id, + scan_status); + + iwl_mld_scan_build_probe_req(mld, vif, ies, ¶ms); + + iwl_mld_scan_6ghz_passive_scan(mld, ¶ms, vif); + + uid = iwl_mld_scan_build_cmd(mld, vif, ¶ms, scan_status, + scan_iter_data.global_low_latency); + if (uid < 0) + return uid; + + ret = iwl_mld_send_cmd(mld, &hcmd); + if (ret) { + IWL_ERR(mld, "Scan failed! ret %d\n", ret); + return ret; + } + + IWL_DEBUG_SCAN(mld, "Scan request send success: status=%u, uid=%u\n", + scan_status, uid); + + mld->scan.uid_status[uid] = scan_status; + mld->scan.status |= scan_status; + + if (params.enable_6ghz_passive) + mld->scan.last_6ghz_passive_jiffies = jiffies; + + return 0; +} + +static int +iwl_mld_scan_send_abort_cmd_status(struct iwl_mld *mld, int uid, u32 *status) +{ + struct iwl_umac_scan_abort abort_cmd = { + .uid = cpu_to_le32(uid), + }; + struct iwl_host_cmd cmd = { + .id = WIDE_ID(LONG_GROUP, SCAN_ABORT_UMAC), + .flags = CMD_WANT_SKB, + .data = { &abort_cmd }, + .len[0] = sizeof(abort_cmd), + }; + struct iwl_rx_packet *pkt; + struct iwl_cmd_response *resp; + u32 resp_len; + int ret; + + ret = iwl_mld_send_cmd(mld, &cmd); + if (ret) + return ret; + + pkt = cmd.resp_pkt; + + resp_len = iwl_rx_packet_payload_len(pkt); + if (IWL_FW_CHECK(mld, resp_len != sizeof(*resp), + "Scan Abort: unexpected response length %d\n", + resp_len)) { + ret = -EIO; + goto out; + } + + resp = (void *)pkt->data; + *status = le32_to_cpu(resp->status); + +out: + iwl_free_resp(&cmd); + return ret; +} + +static int +iwl_mld_scan_abort(struct iwl_mld *mld, int type, int uid, bool *wait) +{ + enum iwl_umac_scan_abort_status status; + int ret; + + *wait = true; + + IWL_DEBUG_SCAN(mld, "Sending scan abort, uid %u\n", uid); + + ret = iwl_mld_scan_send_abort_cmd_status(mld, uid, &status); + + IWL_DEBUG_SCAN(mld, "Scan abort: ret=%d status=%u\n", ret, status); + + /* We don't need to wait to scan complete in the following cases: + * 1. Driver failed to send the scan abort cmd. + * 2. The FW is no longer familiar with the scan that needs to be + * stopped. It is expected that the scan complete notification was + * already received but not yet processed. + * + * In both cases the flow should continue similar to the case that the + * scan was really aborted. + */ + if (ret || status == IWL_UMAC_SCAN_ABORT_STATUS_NOT_FOUND) + *wait = false; + + return ret; +} + +static int +iwl_mld_scan_stop_wait(struct iwl_mld *mld, int type, int uid) +{ + struct iwl_notification_wait wait_scan_done; + static const u16 scan_comp_notif[] = { SCAN_COMPLETE_UMAC }; + bool wait = true; + int ret; + + iwl_init_notification_wait(&mld->notif_wait, &wait_scan_done, + scan_comp_notif, + ARRAY_SIZE(scan_comp_notif), + NULL, NULL); + + IWL_DEBUG_SCAN(mld, "Preparing to stop scan, type=%x\n", type); + + ret = iwl_mld_scan_abort(mld, type, uid, &wait); + if (ret) { + IWL_DEBUG_SCAN(mld, "couldn't stop scan type=%d\n", type); + goto return_no_wait; + } + + if (!wait) { + IWL_DEBUG_SCAN(mld, "no need to wait for scan type=%d\n", type); + goto return_no_wait; + } + + return iwl_wait_notification(&mld->notif_wait, &wait_scan_done, HZ); + +return_no_wait: + iwl_remove_notification(&mld->notif_wait, &wait_scan_done); + return ret; +} + +int iwl_mld_sched_scan_start(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct cfg80211_sched_scan_request *req, + struct ieee80211_scan_ies *ies, + int type) +{ + struct iwl_host_cmd hcmd = { + .id = WIDE_ID(LONG_GROUP, SCAN_REQ_UMAC), + .len = { mld->scan.cmd_size, }, + .data = { mld->scan.cmd, }, + .dataflags = { IWL_HCMD_DFL_NOCOPY, }, + }; + struct iwl_mld_scan_params params = {}; + struct iwl_mld_scan_iter_data scan_iter_data = { + .current_vif = vif, + }; + bool non_psc_included = false; + int ret, uid; + + /* we should have failed registration if scan_cmd was NULL */ + if (WARN_ON(!mld->scan.cmd)) + return -ENOMEM; + + /* FW supports only a single periodic scan */ + if (mld->scan.status & (IWL_MLD_SCAN_SCHED | IWL_MLD_SCAN_NETDETECT)) + return -EBUSY; + + ieee80211_iterate_active_interfaces_mtx(mld->hw, + IEEE80211_IFACE_ITER_NORMAL, + iwl_mld_scan_iterator, + &scan_iter_data); + + params.type = iwl_mld_get_scan_type(mld, vif, &scan_iter_data); + params.flags = req->flags; + params.n_ssids = req->n_ssids; + params.ssids = req->ssids; + params.n_channels = req->n_channels; + params.channels = req->channels; + params.mac_addr = req->mac_addr; + params.mac_addr_mask = req->mac_addr_mask; + params.no_cck = false; + params.pass_all = iwl_mld_scan_pass_all(mld, req); + params.n_match_sets = req->n_match_sets; + params.match_sets = req->match_sets; + params.n_scan_plans = req->n_scan_plans; + params.scan_plans = req->scan_plans; + /* TODO: CDB - per-band flag */ + params.respect_p2p_go = + iwl_mld_get_respect_p2p_go(mld, vif, + scan_iter_data.global_low_latency); + + /* UMAC scan supports up to 16-bit delays, trim it down to 16-bits */ + params.delay = req->delay > U16_MAX ? U16_MAX : req->delay; + + eth_broadcast_addr(params.bssid); + + ret = iwl_mld_config_sched_scan_profiles(mld, req); + if (ret) + return ret; + + iwl_mld_scan_build_probe_req(mld, vif, ies, ¶ms); + + ret = iwl_mld_sched_scan_handle_non_psc_channels(¶ms, + &non_psc_included); + if (ret) + goto out; + + if (!iwl_mld_scan_fits(mld, req->n_ssids, ies, params.n_channels)) { + ret = -ENOBUFS; + goto out; + } + + uid = iwl_mld_scan_build_cmd(mld, vif, ¶ms, type, + scan_iter_data.global_low_latency); + if (uid < 0) { + ret = uid; + goto out; + } + + ret = iwl_mld_send_cmd(mld, &hcmd); + if (!ret) { + IWL_DEBUG_SCAN(mld, + "Sched scan request send success: type=%u, uid=%u\n", + type, uid); + mld->scan.uid_status[uid] = type; + mld->scan.status |= type; + } else { + IWL_ERR(mld, "Sched scan failed! ret %d\n", ret); + mld->scan.pass_all_sched_res = SCHED_SCAN_PASS_ALL_STATE_DISABLED; + } + +out: + if (non_psc_included) + kfree(params.channels); + return ret; +} + +int iwl_mld_scan_stop(struct iwl_mld *mld, int type, bool notify) +{ + int uid, ret; + + IWL_DEBUG_SCAN(mld, + "Request to stop scan: type=0x%x, status=0x%x\n", + type, mld->scan.status); + + if (!(mld->scan.status & type)) + return 0; + + uid = iwl_mld_scan_uid_by_status(mld, type); + /* must be valid, we just checked it's running */ + if (WARN_ON_ONCE(uid < 0)) + return uid; + + ret = iwl_mld_scan_stop_wait(mld, type, uid); + if (ret) + IWL_DEBUG_SCAN(mld, "Failed to stop scan\n"); + + /* Clear the scan status so the next scan requests will + * succeed and mark the scan as stopping, so that the Rx + * handler doesn't do anything, as the scan was stopped from + * above. Also remove the handler to not notify mac80211 + * erroneously after a new scan starts, for example. + */ + mld->scan.status &= ~type; + mld->scan.uid_status[uid] = IWL_MLD_SCAN_NONE; + iwl_mld_cancel_notifications_of_object(mld, IWL_MLD_OBJECT_TYPE_SCAN, + uid); + + if (type == IWL_MLD_SCAN_REGULAR) { + if (notify) { + struct cfg80211_scan_info info = { + .aborted = true, + }; + + ieee80211_scan_completed(mld->hw, &info); + } + } else if (notify) { + ieee80211_sched_scan_stopped(mld->hw); + mld->scan.pass_all_sched_res = SCHED_SCAN_PASS_ALL_STATE_DISABLED; + } + + return ret; +} + +int iwl_mld_regular_scan_start(struct iwl_mld *mld, struct ieee80211_vif *vif, + struct cfg80211_scan_request *req, + struct ieee80211_scan_ies *ies) +{ + return _iwl_mld_single_scan_start(mld, vif, req, ies, + IWL_MLD_SCAN_REGULAR); +} + +static void iwl_mld_int_mlo_scan_start(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_channel **channels, + size_t n_channels) +{ + struct cfg80211_scan_request *req __free(kfree) = NULL; + struct ieee80211_scan_ies ies = {}; + size_t size; + int ret; + + IWL_DEBUG_SCAN(mld, "Starting Internal MLO scan: n_channels=%zu\n", + n_channels); + + size = struct_size(req, channels, n_channels); + req = kzalloc(size, GFP_KERNEL); + if (!req) + return; + + /* set the requested channels */ + for (int i = 0; i < n_channels; i++) + req->channels[i] = channels[i]; + + req->n_channels = n_channels; + + /* set the rates */ + for (int i = 0; i < NUM_NL80211_BANDS; i++) + if (mld->wiphy->bands[i]) + req->rates[i] = + (1 << mld->wiphy->bands[i]->n_bitrates) - 1; + + req->wdev = ieee80211_vif_to_wdev(vif); + req->wiphy = mld->wiphy; + req->scan_start = jiffies; + req->tsf_report_link_id = -1; + + ret = _iwl_mld_single_scan_start(mld, vif, req, &ies, + IWL_MLD_SCAN_INT_MLO); + + IWL_DEBUG_SCAN(mld, "Internal MLO scan: ret=%d\n", ret); +} + +void iwl_mld_int_mlo_scan(struct iwl_mld *mld, struct ieee80211_vif *vif) +{ + struct ieee80211_channel *channels[IEEE80211_MLD_MAX_NUM_LINKS]; + unsigned long usable_links = ieee80211_vif_usable_links(vif); + size_t n_channels = 0; + u8 link_id; + + lockdep_assert_wiphy(mld->wiphy); + + if (!vif->cfg.assoc || !ieee80211_vif_is_mld(vif) || + hweight16(vif->valid_links) == 1) + return; + + if (mld->scan.status & IWL_MLD_SCAN_INT_MLO) { + IWL_DEBUG_SCAN(mld, "Internal MLO scan is already running\n"); + return; + } + + for_each_set_bit(link_id, &usable_links, IEEE80211_MLD_MAX_NUM_LINKS) { + struct ieee80211_bss_conf *link_conf = + link_conf_dereference_check(vif, link_id); + + if (WARN_ON_ONCE(!link_conf)) + continue; + + channels[n_channels++] = link_conf->chanreq.oper.chan; + } + + if (!n_channels) + return; + + iwl_mld_int_mlo_scan_start(mld, vif, channels, n_channels); +} + +void iwl_mld_handle_scan_iter_complete_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + struct iwl_umac_scan_iter_complete_notif *notif = (void *)pkt->data; + u32 uid = __le32_to_cpu(notif->uid); + + if (IWL_FW_CHECK(mld, uid >= ARRAY_SIZE(mld->scan.uid_status), + "FW reports out-of-range scan UID %d\n", uid)) + return; + + if (mld->scan.uid_status[uid] == IWL_MLD_SCAN_REGULAR) + mld->scan.start_tsf = le64_to_cpu(notif->start_tsf); + + IWL_DEBUG_SCAN(mld, + "UMAC Scan iteration complete: status=0x%x scanned_channels=%d\n", + notif->status, notif->scanned_channels); + + if (mld->scan.pass_all_sched_res == SCHED_SCAN_PASS_ALL_STATE_FOUND) { + IWL_DEBUG_SCAN(mld, "Pass all scheduled scan results found\n"); + ieee80211_sched_scan_results(mld->hw); + mld->scan.pass_all_sched_res = SCHED_SCAN_PASS_ALL_STATE_ENABLED; + } + + IWL_DEBUG_SCAN(mld, + "UMAC Scan iteration complete: scan started at %llu (TSF)\n", + le64_to_cpu(notif->start_tsf)); +} + +void iwl_mld_handle_match_found_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + IWL_DEBUG_SCAN(mld, "Scheduled scan results\n"); + ieee80211_sched_scan_results(mld->hw); +} + +void iwl_mld_handle_scan_complete_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + struct iwl_umac_scan_complete *notif = (void *)pkt->data; + bool aborted = (notif->status == IWL_SCAN_OFFLOAD_ABORTED); + u32 uid = __le32_to_cpu(notif->uid); + + if (IWL_FW_CHECK(mld, uid >= ARRAY_SIZE(mld->scan.uid_status), + "FW reports out-of-range scan UID %d\n", uid)) + return; + + IWL_DEBUG_SCAN(mld, + "Scan completed: uid=%u type=%u, status=%s, EBS=%s\n", + uid, mld->scan.uid_status[uid], + notif->status == IWL_SCAN_OFFLOAD_COMPLETED ? + "completed" : "aborted", + iwl_mld_scan_ebs_status_str(notif->ebs_status)); + IWL_DEBUG_SCAN(mld, "Scan completed: scan_status=0x%x\n", + mld->scan.status); + IWL_DEBUG_SCAN(mld, + "Scan completed: line=%u, iter=%u, elapsed time=%u\n", + notif->last_schedule, notif->last_iter, + __le32_to_cpu(notif->time_from_last_iter)); + + if (IWL_FW_CHECK(mld, !(mld->scan.uid_status[uid] & mld->scan.status), + "FW reports scan UID %d we didn't trigger\n", uid)) + return; + + /* if the scan is already stopping, we don't need to notify mac80211 */ + if (mld->scan.uid_status[uid] == IWL_MLD_SCAN_REGULAR) { + struct cfg80211_scan_info info = { + .aborted = aborted, + .scan_start_tsf = mld->scan.start_tsf, + }; + int fw_link_id = mld->scan.fw_link_id; + struct ieee80211_bss_conf *link_conf = NULL; + + if (fw_link_id != IWL_MLD_INVALID_FW_ID) + link_conf = + wiphy_dereference(mld->wiphy, + mld->fw_id_to_bss_conf[fw_link_id]); + + /* It is possible that by the time the scan is complete the + * link was already removed and is not valid. + */ + if (link_conf) + ether_addr_copy(info.tsf_bssid, link_conf->bssid); + else + IWL_DEBUG_SCAN(mld, "Scan link is no longer valid\n"); + + ieee80211_scan_completed(mld->hw, &info); + } else if (mld->scan.uid_status[uid] == IWL_MLD_SCAN_SCHED) { + ieee80211_sched_scan_stopped(mld->hw); + mld->scan.pass_all_sched_res = SCHED_SCAN_PASS_ALL_STATE_DISABLED; + } else if (mld->scan.uid_status[uid] == IWL_MLD_SCAN_INT_MLO) { + IWL_DEBUG_SCAN(mld, "Internal MLO scan completed\n"); + mld->scan.last_mlo_scan_jiffies = jiffies; + + /* + * We limit link selection to internal MLO scans as otherwise + * we do not know whether all channels were covered. + */ + iwl_mld_select_links(mld); + } + + mld->scan.status &= ~mld->scan.uid_status[uid]; + + IWL_DEBUG_SCAN(mld, "Scan completed: after update: scan_status=0x%x\n", + mld->scan.status); + + mld->scan.uid_status[uid] = IWL_MLD_SCAN_NONE; + + if (notif->ebs_status != IWL_SCAN_EBS_SUCCESS && + notif->ebs_status != IWL_SCAN_EBS_INACTIVE) + mld->scan.last_ebs_failed = true; +} + +/* This function is used in nic restart flow, to inform mac80211 about scans + * that were aborted by restart flow or by an assert. + */ +void iwl_mld_report_scan_aborted(struct iwl_mld *mld) +{ + int uid; + + uid = iwl_mld_scan_uid_by_status(mld, IWL_MLD_SCAN_REGULAR); + if (uid >= 0) { + struct cfg80211_scan_info info = { + .aborted = true, + }; + + ieee80211_scan_completed(mld->hw, &info); + mld->scan.uid_status[uid] = IWL_MLD_SCAN_NONE; + } + + uid = iwl_mld_scan_uid_by_status(mld, IWL_MLD_SCAN_SCHED); + if (uid >= 0) { + mld->scan.pass_all_sched_res = SCHED_SCAN_PASS_ALL_STATE_DISABLED; + mld->scan.uid_status[uid] = IWL_MLD_SCAN_NONE; + + /* sched scan will be restarted by mac80211 in reconfig. + * report to mac80211 that sched scan stopped only if we won't + * restart the firmware. + */ + if (!iwlwifi_mod_params.fw_restart) + ieee80211_sched_scan_stopped(mld->hw); + } + + uid = iwl_mld_scan_uid_by_status(mld, IWL_MLD_SCAN_INT_MLO); + if (uid >= 0) { + IWL_DEBUG_SCAN(mld, "Internal MLO scan aborted\n"); + mld->scan.uid_status[uid] = IWL_MLD_SCAN_NONE; + } + + BUILD_BUG_ON(IWL_MLD_SCAN_NONE != 0); + memset(mld->scan.uid_status, 0, sizeof(mld->scan.uid_status)); +} + +int iwl_mld_alloc_scan_cmd(struct iwl_mld *mld) +{ + u8 scan_cmd_ver = iwl_fw_lookup_cmd_ver(mld->fw, SCAN_REQ_UMAC, + IWL_FW_CMD_VER_UNKNOWN); + size_t scan_cmd_size; + + if (scan_cmd_ver == 17) { + scan_cmd_size = sizeof(struct iwl_scan_req_umac_v17); + } else { + IWL_ERR(mld, "Unexpected scan cmd version %d\n", scan_cmd_ver); + return -EINVAL; + } + + mld->scan.cmd = kmalloc(scan_cmd_size, GFP_KERNEL); + if (!mld->scan.cmd) + return -ENOMEM; + + mld->scan.cmd_size = scan_cmd_size; + + return 0; +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/scan.h b/drivers/net/wireless/intel/iwlwifi/mld/scan.h new file mode 100644 index 0000000000000..cb589e0023193 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/scan.h @@ -0,0 +1,135 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024-2025 Intel Corporation + */ +#ifndef __iwl_mld_scan_h__ +#define __iwl_mld_scan_h__ + +int iwl_mld_alloc_scan_cmd(struct iwl_mld *mld); + +int iwl_mld_regular_scan_start(struct iwl_mld *mld, struct ieee80211_vif *vif, + struct cfg80211_scan_request *req, + struct ieee80211_scan_ies *ies); + +void iwl_mld_int_mlo_scan(struct iwl_mld *mld, struct ieee80211_vif *vif); + +void iwl_mld_handle_scan_iter_complete_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt); + +int iwl_mld_scan_stop(struct iwl_mld *mld, int type, bool notify); + +int iwl_mld_sched_scan_start(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct cfg80211_sched_scan_request *req, + struct ieee80211_scan_ies *ies, + int type); + +void iwl_mld_handle_match_found_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt); + +void iwl_mld_handle_scan_complete_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt); + +#define WFA_TPC_IE_LEN 9 + +static inline int iwl_mld_scan_max_template_size(void) +{ +#define MAC_HDR_LEN 24 +#define DS_IE_LEN 3 +#define SSID_IE_LEN 2 + +/* driver create the 802.11 header, WFA TPC IE, DS parameter and SSID IE */ +#define DRIVER_TOTAL_IES_LEN \ + (MAC_HDR_LEN + WFA_TPC_IE_LEN + DS_IE_LEN + SSID_IE_LEN) + + BUILD_BUG_ON(SCAN_OFFLOAD_PROBE_REQ_SIZE < DRIVER_TOTAL_IES_LEN); + + return SCAN_OFFLOAD_PROBE_REQ_SIZE - DRIVER_TOTAL_IES_LEN; +} + +void iwl_mld_report_scan_aborted(struct iwl_mld *mld); + +enum iwl_mld_scan_status { + IWL_MLD_SCAN_NONE = 0, + IWL_MLD_SCAN_REGULAR = BIT(0), + IWL_MLD_SCAN_SCHED = BIT(1), + IWL_MLD_SCAN_NETDETECT = BIT(2), + IWL_MLD_SCAN_INT_MLO = BIT(3), +}; + +/* enum iwl_mld_pass_all_sched_results_states - Defines the states for + * handling/passing scheduled scan results to mac80211 + * @SCHED_SCAN_PASS_ALL_STATE_DISABLED: Don't pass all scan results, only when + * a match found. + * @SCHED_SCAN_PASS_ALL_STATE_ENABLED: Pass all scan results is enabled + * (no filtering). + * @SCHED_SCAN_PASS_ALL_STATE_FOUND: A scan result is found, pass it on the + * next scan iteration complete notification. + */ +enum iwl_mld_pass_all_sched_results_states { + SCHED_SCAN_PASS_ALL_STATE_DISABLED, + SCHED_SCAN_PASS_ALL_STATE_ENABLED, + SCHED_SCAN_PASS_ALL_STATE_FOUND, +}; + +/** + * enum iwl_mld_traffic_load - Levels of traffic load + * + * @IWL_MLD_TRAFFIC_LOW: low traffic load + * @IWL_MLD_TRAFFIC_MEDIUM: medium traffic load + * @IWL_MLD_TRAFFIC_HIGH: high traffic load + */ +enum iwl_mld_traffic_load { + IWL_MLD_TRAFFIC_LOW, + IWL_MLD_TRAFFIC_MEDIUM, + IWL_MLD_TRAFFIC_HIGH, +}; + +/** + * struct iwl_mld_scan - Scan data + * @status: scan status, a combination of %enum iwl_mld_scan_status, + * reflects the %scan.uid_status array. + * @uid_status: array to track the scan status per uid. + * @start_tsf: start time of last scan in TSF of the link that requested + * the scan. + * @last_ebs_failed: true if the last EBS (Energy Based Scan) failed. + * @pass_all_sched_res: see %enum iwl_mld_pass_all_sched_results_states. + * @fw_link_id: the current (regular) scan fw link id, used by scan + * complete notif. + * @traffic_load: traffic load related data + * @traffic_load.last_stats_ts_usec: The timestamp of the last statistics + * notification, used to calculate the elapsed time between two + * notifications and determine the traffic load + * @traffic_load.status: The current traffic load status, see + * &enum iwl_mld_traffic_load + * @cmd_size: size of %cmd. + * @cmd: pointer to scan cmd buffer (allocated once in op mode start). + * @last_6ghz_passive_jiffies: stores the last 6GHz passive scan time + * in jiffies. + * @last_start_time_jiffies: stores the last start time in jiffies + * (interface up/reset/resume). + * @last_mlo_scan_jiffies: end time of the last MLO scan in jiffies. + */ +struct iwl_mld_scan { + /* Add here fields that need clean up on restart */ + struct_group(zeroed_on_hw_restart, + unsigned int status; + u32 uid_status[IWL_MAX_UMAC_SCANS]; + u64 start_tsf; + bool last_ebs_failed; + enum iwl_mld_pass_all_sched_results_states pass_all_sched_res; + u8 fw_link_id; + struct { + u32 last_stats_ts_usec; + enum iwl_mld_traffic_load status; + } traffic_load; + ); + /* And here fields that survive a fw restart */ + size_t cmd_size; + void *cmd; + unsigned long last_6ghz_passive_jiffies; + unsigned long last_start_time_jiffies; + unsigned long last_mlo_scan_jiffies; +}; + +#endif /* __iwl_mld_scan_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/session-protect.c b/drivers/net/wireless/intel/iwlwifi/mld/session-protect.c new file mode 100644 index 0000000000000..dbb5615dc3f64 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/session-protect.c @@ -0,0 +1,222 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024 Intel Corporation + */ +#include "session-protect.h" +#include "fw/api/time-event.h" +#include "fw/api/context.h" +#include "iface.h" +#include + +void iwl_mld_handle_session_prot_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + struct iwl_session_prot_notif *notif = (void *)pkt->data; + int fw_link_id = le32_to_cpu(notif->mac_link_id); + struct ieee80211_bss_conf *link_conf = + iwl_mld_fw_id_to_link_conf(mld, fw_link_id); + struct ieee80211_vif *vif; + struct iwl_mld_vif *mld_vif; + struct iwl_mld_session_protect *session_protect; + + if (WARN_ON(!link_conf)) + return; + + vif = link_conf->vif; + mld_vif = iwl_mld_vif_from_mac80211(vif); + session_protect = &mld_vif->session_protect; + + if (!le32_to_cpu(notif->status)) { + memset(session_protect, 0, sizeof(*session_protect)); + } else if (le32_to_cpu(notif->start)) { + /* End_jiffies indicates an active session */ + session_protect->session_requested = false; + session_protect->end_jiffies = + TU_TO_EXP_TIME(session_protect->duration); + /* !session_protect->end_jiffies means inactive session */ + if (!session_protect->end_jiffies) + session_protect->end_jiffies = 1; + } else { + memset(session_protect, 0, sizeof(*session_protect)); + } +} + +static int _iwl_mld_schedule_session_protection(struct iwl_mld *mld, + struct ieee80211_vif *vif, + u32 duration, u32 min_duration, + int link_id) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mld_link *link = + iwl_mld_link_dereference_check(mld_vif, link_id); + struct iwl_mld_session_protect *session_protect = + &mld_vif->session_protect; + struct iwl_session_prot_cmd cmd = { + .id_and_color = cpu_to_le32(link->fw_id), + .action = cpu_to_le32(FW_CTXT_ACTION_ADD), + .conf_id = cpu_to_le32(SESSION_PROTECT_CONF_ASSOC), + .duration_tu = cpu_to_le32(MSEC_TO_TU(duration)), + }; + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + WARN(hweight16(vif->active_links) > 1, + "Session protection isn't allowed with more than one active link"); + + if (session_protect->end_jiffies && + time_after(session_protect->end_jiffies, + TU_TO_EXP_TIME(min_duration))) { + IWL_DEBUG_TE(mld, "We have ample in the current session: %u\n", + jiffies_to_msecs(session_protect->end_jiffies - + jiffies)); + return -EALREADY; + } + + IWL_DEBUG_TE(mld, "Add a new session protection, duration %d TU\n", + le32_to_cpu(cmd.duration_tu)); + + ret = iwl_mld_send_cmd_pdu(mld, WIDE_ID(MAC_CONF_GROUP, + SESSION_PROTECTION_CMD), &cmd); + + if (ret) + return ret; + + /* end_jiffies will be updated when handling session_prot_notif */ + session_protect->end_jiffies = 0; + session_protect->duration = duration; + session_protect->session_requested = true; + + return 0; +} + +void iwl_mld_schedule_session_protection(struct iwl_mld *mld, + struct ieee80211_vif *vif, + u32 duration, u32 min_duration, + int link_id) +{ + int ret; + + ret = _iwl_mld_schedule_session_protection(mld, vif, duration, + min_duration, link_id); + if (ret && ret != -EALREADY) + IWL_ERR(mld, + "Couldn't send the SESSION_PROTECTION_CMD (%d)\n", + ret); +} + +struct iwl_mld_session_start_data { + struct iwl_mld *mld; + struct ieee80211_bss_conf *link_conf; + bool success; +}; + +static bool iwl_mld_session_start_fn(struct iwl_notif_wait_data *notif_wait, + struct iwl_rx_packet *pkt, void *_data) +{ + struct iwl_session_prot_notif *notif = (void *)pkt->data; + unsigned int pkt_len = iwl_rx_packet_payload_len(pkt); + struct iwl_mld_session_start_data *data = _data; + struct ieee80211_bss_conf *link_conf; + struct iwl_mld *mld = data->mld; + int fw_link_id; + + if (IWL_FW_CHECK(mld, pkt_len < sizeof(*notif), + "short session prot notif (%d)\n", + pkt_len)) + return false; + + fw_link_id = le32_to_cpu(notif->mac_link_id); + link_conf = iwl_mld_fw_id_to_link_conf(mld, fw_link_id); + + if (link_conf != data->link_conf) + return false; + + if (!le32_to_cpu(notif->status)) + return true; + + if (notif->start) { + data->success = true; + return true; + } + + return false; +} + +int iwl_mld_start_session_protection(struct iwl_mld *mld, + struct ieee80211_vif *vif, + u32 duration, u32 min_duration, + int link_id, unsigned long timeout) +{ + static const u16 start_notif[] = { SESSION_PROTECTION_NOTIF }; + struct iwl_notification_wait start_wait; + struct iwl_mld_session_start_data data = { + .mld = mld, + .link_conf = wiphy_dereference(mld->wiphy, + vif->link_conf[link_id]), + }; + int ret; + + if (WARN_ON(!data.link_conf)) + return -EINVAL; + + iwl_init_notification_wait(&mld->notif_wait, &start_wait, + start_notif, ARRAY_SIZE(start_notif), + iwl_mld_session_start_fn, &data); + + ret = _iwl_mld_schedule_session_protection(mld, vif, duration, + min_duration, link_id); + + if (ret) { + iwl_remove_notification(&mld->notif_wait, &start_wait); + return ret == -EALREADY ? 0 : ret; + } + + ret = iwl_wait_notification(&mld->notif_wait, &start_wait, timeout); + if (ret) + return ret; + return data.success ? 0 : -EIO; +} + +int iwl_mld_cancel_session_protection(struct iwl_mld *mld, + struct ieee80211_vif *vif, + int link_id) +{ + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + struct iwl_mld_link *link = + iwl_mld_link_dereference_check(mld_vif, link_id); + struct iwl_mld_session_protect *session_protect = + &mld_vif->session_protect; + struct iwl_session_prot_cmd cmd = { + .action = cpu_to_le32(FW_CTXT_ACTION_REMOVE), + .conf_id = cpu_to_le32(SESSION_PROTECT_CONF_ASSOC), + }; + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + /* If there isn't an active session or a requested one for this + * link do nothing + */ + if (!session_protect->session_requested && + !session_protect->end_jiffies) + return 0; + + if (WARN_ON(!link)) + return -EINVAL; + + cmd.id_and_color = cpu_to_le32(link->fw_id); + + ret = iwl_mld_send_cmd_pdu(mld, + WIDE_ID(MAC_CONF_GROUP, + SESSION_PROTECTION_CMD), &cmd); + if (ret) { + IWL_ERR(mld, + "Couldn't send the SESSION_PROTECTION_CMD\n"); + return ret; + } + + memset(session_protect, 0, sizeof(*session_protect)); + + return 0; +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/session-protect.h b/drivers/net/wireless/intel/iwlwifi/mld/session-protect.h new file mode 100644 index 0000000000000..642bec8451a16 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/session-protect.h @@ -0,0 +1,102 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024-2025 Intel Corporation + */ + +#ifndef __session_protect_h__ +#define __session_protect_h__ + +#include "mld.h" +#include "hcmd.h" +#include +#include "fw/api/mac-cfg.h" + +/** + * DOC: session protection + * + * Session protection is an API from the firmware that allows the driver to + * request time on medium. This is needed before the association when we need + * to be on medium for the association frame exchange. Once we configure the + * firmware as 'associated', the firmware will allocate time on medium without + * needed a session protection. + * + * TDLS discover uses this API as well even after association to ensure that + * other activities internal to the firmware will not interrupt our presence + * on medium. + */ + +/** + * struct iwl_mld_session_protect - session protection parameters + * @end_jiffies: expected end_jiffies of current session protection. + * 0 if not active + * @duration: the duration in tu of current session + * @session_requested: A session protection command was sent and wasn't yet + * answered + */ +struct iwl_mld_session_protect { + unsigned long end_jiffies; + u32 duration; + bool session_requested; +}; + +#define IWL_MLD_SESSION_PROTECTION_ASSOC_TIME_MS 900 +#define IWL_MLD_SESSION_PROTECTION_MIN_TIME_MS 400 + +/** + * iwl_mld_handle_session_prot_notif - handles %SESSION_PROTECTION_NOTIF + * @mld: the mld component + * @pkt: the RX packet containing the notification + */ +void iwl_mld_handle_session_prot_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt); + +/** + * iwl_mld_schedule_session_protection - schedule a session protection + * @mld: the mld component + * @vif: the virtual interface for which the protection issued + * @duration: the requested duration of the protection + * @min_duration: the minimum duration of the protection + * @link_id: The link to schedule a session protection for + */ +void iwl_mld_schedule_session_protection(struct iwl_mld *mld, + struct ieee80211_vif *vif, + u32 duration, u32 min_duration, + int link_id); + +/** + * iwl_mld_start_session_protection - start a session protection + * @mld: the mld component + * @vif: the virtual interface for which the protection issued + * @duration: the requested duration of the protection + * @min_duration: the minimum duration of the protection + * @link_id: The link to schedule a session protection for + * @timeout: timeout for waiting + * + * This schedules the session protection, and waits for it to start + * (with timeout) + * + * Returns: 0 if successful, error code otherwise + */ +int iwl_mld_start_session_protection(struct iwl_mld *mld, + struct ieee80211_vif *vif, + u32 duration, u32 min_duration, + int link_id, unsigned long timeout); + +/** + * iwl_mld_cancel_session_protection - cancel the session protection. + * @mld: the mld component + * @vif: the virtual interface for which the session is issued + * @link_id: cancel the session protection for given link + * + * This functions cancels the session protection which is an act of good + * citizenship. If it is not needed any more it should be canceled because + * the other mac contexts wait for the medium during that time. + * + * Returns: 0 if successful, error code otherwise + * + */ +int iwl_mld_cancel_session_protection(struct iwl_mld *mld, + struct ieee80211_vif *vif, + int link_id); + +#endif /* __session_protect_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/sta.c b/drivers/net/wireless/intel/iwlwifi/mld/sta.c new file mode 100644 index 0000000000000..f266a81dd29b9 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/sta.c @@ -0,0 +1,1265 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024-2025 Intel Corporation + */ + +#include +#include + +#include "sta.h" +#include "hcmd.h" +#include "iface.h" +#include "mlo.h" +#include "key.h" +#include "agg.h" +#include "tlc.h" +#include "fw/api/sta.h" +#include "fw/api/mac.h" +#include "fw/api/rx.h" + +int iwl_mld_fw_sta_id_from_link_sta(struct iwl_mld *mld, + struct ieee80211_link_sta *link_sta) +{ + struct iwl_mld_link_sta *mld_link_sta; + + /* This function should only be used with the wiphy lock held, + * In other cases, it is not guaranteed that the link_sta will exist + * in the driver too, and it is checked here. + */ + lockdep_assert_wiphy(mld->wiphy); + + /* This is not meant to be called with a NULL pointer */ + if (WARN_ON(!link_sta)) + return -ENOENT; + + mld_link_sta = iwl_mld_link_sta_from_mac80211(link_sta); + if (!mld_link_sta) { + WARN_ON(!iwl_mld_error_before_recovery(mld)); + return -ENOENT; + } + + return mld_link_sta->fw_id; +} + +static void +iwl_mld_fill_ampdu_size_and_dens(struct ieee80211_link_sta *link_sta, + struct ieee80211_bss_conf *link, + __le32 *tx_ampdu_max_size, + __le32 *tx_ampdu_spacing) +{ + u32 agg_size = 0, mpdu_dens = 0; + + if (WARN_ON(!link_sta || !link)) + return; + + /* Note that we always use only legacy & highest supported PPDUs, so + * of Draft P802.11be D.30 Table 10-12a--Fields used for calculating + * the maximum A-MPDU size of various PPDU types in different bands, + * we only need to worry about the highest supported PPDU type here. + */ + + if (link_sta->ht_cap.ht_supported) { + agg_size = link_sta->ht_cap.ampdu_factor; + mpdu_dens = link_sta->ht_cap.ampdu_density; + } + + if (link->chanreq.oper.chan->band == NL80211_BAND_6GHZ) { + /* overwrite HT values on 6 GHz */ + mpdu_dens = + le16_get_bits(link_sta->he_6ghz_capa.capa, + IEEE80211_HE_6GHZ_CAP_MIN_MPDU_START); + agg_size = + le16_get_bits(link_sta->he_6ghz_capa.capa, + IEEE80211_HE_6GHZ_CAP_MAX_AMPDU_LEN_EXP); + } else if (link_sta->vht_cap.vht_supported) { + /* if VHT supported overwrite HT value */ + agg_size = + u32_get_bits(link_sta->vht_cap.cap, + IEEE80211_VHT_CAP_MAX_A_MPDU_LENGTH_EXPONENT_MASK); + } + + /* D6.0 10.12.2 A-MPDU length limit rules + * A STA indicates the maximum length of the A-MPDU preEOF padding + * that it can receive in an HE PPDU in the Maximum A-MPDU Length + * Exponent field in its HT Capabilities, VHT Capabilities, + * and HE 6 GHz Band Capabilities elements (if present) and the + * Maximum AMPDU Length Exponent Extension field in its HE + * Capabilities element + */ + if (link_sta->he_cap.has_he) + agg_size += + u8_get_bits(link_sta->he_cap.he_cap_elem.mac_cap_info[3], + IEEE80211_HE_MAC_CAP3_MAX_AMPDU_LEN_EXP_MASK); + + if (link_sta->eht_cap.has_eht) + agg_size += + u8_get_bits(link_sta->eht_cap.eht_cap_elem.mac_cap_info[1], + IEEE80211_EHT_MAC_CAP1_MAX_AMPDU_LEN_MASK); + + /* Limit to max A-MPDU supported by FW */ + agg_size = min_t(u32, agg_size, + STA_FLG_MAX_AGG_SIZE_4M >> STA_FLG_MAX_AGG_SIZE_SHIFT); + + *tx_ampdu_max_size = cpu_to_le32(agg_size); + *tx_ampdu_spacing = cpu_to_le32(mpdu_dens); +} + +static u8 iwl_mld_get_uapsd_acs(struct ieee80211_sta *sta) +{ + u8 uapsd_acs = 0; + + if (sta->uapsd_queues & IEEE80211_WMM_IE_STA_QOSINFO_AC_BK) + uapsd_acs |= BIT(AC_BK); + if (sta->uapsd_queues & IEEE80211_WMM_IE_STA_QOSINFO_AC_BE) + uapsd_acs |= BIT(AC_BE); + if (sta->uapsd_queues & IEEE80211_WMM_IE_STA_QOSINFO_AC_VI) + uapsd_acs |= BIT(AC_VI); + if (sta->uapsd_queues & IEEE80211_WMM_IE_STA_QOSINFO_AC_VO) + uapsd_acs |= BIT(AC_VO); + + return uapsd_acs | uapsd_acs << 4; +} + +static u8 iwl_mld_he_get_ppe_val(u8 *ppe, u8 ppe_pos_bit) +{ + u8 byte_num = ppe_pos_bit / 8; + u8 bit_num = ppe_pos_bit % 8; + u8 residue_bits; + u8 res; + + if (bit_num <= 5) + return (ppe[byte_num] >> bit_num) & + (BIT(IEEE80211_PPE_THRES_INFO_PPET_SIZE) - 1); + + /* If bit_num > 5, we have to combine bits with next byte. + * Calculate how many bits we need to take from current byte (called + * here "residue_bits"), and add them to bits from next byte. + */ + + residue_bits = 8 - bit_num; + + res = (ppe[byte_num + 1] & + (BIT(IEEE80211_PPE_THRES_INFO_PPET_SIZE - residue_bits) - 1)) << + residue_bits; + res += (ppe[byte_num] >> bit_num) & (BIT(residue_bits) - 1); + + return res; +} + +static void iwl_mld_parse_ppe(struct iwl_mld *mld, + struct iwl_he_pkt_ext_v2 *pkt_ext, u8 nss, + u8 ru_index_bitmap, u8 *ppe, u8 ppe_pos_bit, + bool inheritance) +{ + /* FW currently supports only nss == MAX_HE_SUPP_NSS + * + * If nss > MAX: we can ignore values we don't support + * If nss < MAX: we can set zeros in other streams + */ + if (nss > MAX_HE_SUPP_NSS) { + IWL_DEBUG_INFO(mld, "Got NSS = %d - trimming to %d\n", nss, + MAX_HE_SUPP_NSS); + nss = MAX_HE_SUPP_NSS; + } + + for (int i = 0; i < nss; i++) { + u8 ru_index_tmp = ru_index_bitmap << 1; + u8 low_th = IWL_HE_PKT_EXT_NONE, high_th = IWL_HE_PKT_EXT_NONE; + + for (u8 bw = 0; + bw < ARRAY_SIZE(pkt_ext->pkt_ext_qam_th[i]); + bw++) { + ru_index_tmp >>= 1; + + /* According to the 11be spec, if for a specific BW the PPE Thresholds + * isn't present - it should inherit the thresholds from the last + * BW for which we had PPE Thresholds. In 11ax though, we don't have + * this inheritance - continue in this case + */ + if (!(ru_index_tmp & 1)) { + if (inheritance) + goto set_thresholds; + else + continue; + } + + high_th = iwl_mld_he_get_ppe_val(ppe, ppe_pos_bit); + ppe_pos_bit += IEEE80211_PPE_THRES_INFO_PPET_SIZE; + low_th = iwl_mld_he_get_ppe_val(ppe, ppe_pos_bit); + ppe_pos_bit += IEEE80211_PPE_THRES_INFO_PPET_SIZE; + +set_thresholds: + pkt_ext->pkt_ext_qam_th[i][bw][0] = low_th; + pkt_ext->pkt_ext_qam_th[i][bw][1] = high_th; + } + } +} + +static void iwl_mld_set_pkt_ext_from_he_ppe(struct iwl_mld *mld, + struct ieee80211_link_sta *link_sta, + struct iwl_he_pkt_ext_v2 *pkt_ext, + bool inheritance) +{ + u8 nss = (link_sta->he_cap.ppe_thres[0] & + IEEE80211_PPE_THRES_NSS_MASK) + 1; + u8 *ppe = &link_sta->he_cap.ppe_thres[0]; + u8 ru_index_bitmap = + u8_get_bits(*ppe, + IEEE80211_PPE_THRES_RU_INDEX_BITMASK_MASK); + /* Starting after PPE header */ + u8 ppe_pos_bit = IEEE80211_HE_PPE_THRES_INFO_HEADER_SIZE; + + iwl_mld_parse_ppe(mld, pkt_ext, nss, ru_index_bitmap, ppe, ppe_pos_bit, + inheritance); +} + +static int +iwl_mld_set_pkt_ext_from_nominal_padding(struct iwl_he_pkt_ext_v2 *pkt_ext, + u8 nominal_padding) +{ + int low_th = -1; + int high_th = -1; + + /* all the macros are the same for EHT and HE */ + switch (nominal_padding) { + case IEEE80211_EHT_PHY_CAP5_COMMON_NOMINAL_PKT_PAD_0US: + low_th = IWL_HE_PKT_EXT_NONE; + high_th = IWL_HE_PKT_EXT_NONE; + break; + case IEEE80211_EHT_PHY_CAP5_COMMON_NOMINAL_PKT_PAD_8US: + low_th = IWL_HE_PKT_EXT_BPSK; + high_th = IWL_HE_PKT_EXT_NONE; + break; + case IEEE80211_EHT_PHY_CAP5_COMMON_NOMINAL_PKT_PAD_16US: + case IEEE80211_EHT_PHY_CAP5_COMMON_NOMINAL_PKT_PAD_20US: + low_th = IWL_HE_PKT_EXT_NONE; + high_th = IWL_HE_PKT_EXT_BPSK; + break; + } + + if (low_th < 0 || high_th < 0) + return -EINVAL; + + /* Set the PPE thresholds accordingly */ + for (int i = 0; i < MAX_HE_SUPP_NSS; i++) { + for (u8 bw = 0; + bw < ARRAY_SIZE(pkt_ext->pkt_ext_qam_th[i]); + bw++) { + pkt_ext->pkt_ext_qam_th[i][bw][0] = low_th; + pkt_ext->pkt_ext_qam_th[i][bw][1] = high_th; + } + } + + return 0; +} + +static void iwl_mld_get_optimal_ppe_info(struct iwl_he_pkt_ext_v2 *pkt_ext, + u8 nominal_padding) +{ + for (int i = 0; i < MAX_HE_SUPP_NSS; i++) { + for (u8 bw = 0; bw < ARRAY_SIZE(pkt_ext->pkt_ext_qam_th[i]); + bw++) { + u8 *qam_th = &pkt_ext->pkt_ext_qam_th[i][bw][0]; + + if (nominal_padding > + IEEE80211_EHT_PHY_CAP5_COMMON_NOMINAL_PKT_PAD_8US && + qam_th[1] == IWL_HE_PKT_EXT_NONE) + qam_th[1] = IWL_HE_PKT_EXT_4096QAM; + else if (nominal_padding == + IEEE80211_EHT_PHY_CAP5_COMMON_NOMINAL_PKT_PAD_8US && + qam_th[0] == IWL_HE_PKT_EXT_NONE && + qam_th[1] == IWL_HE_PKT_EXT_NONE) + qam_th[0] = IWL_HE_PKT_EXT_4096QAM; + } + } +} + +static void iwl_mld_fill_pkt_ext(struct iwl_mld *mld, + struct ieee80211_link_sta *link_sta, + struct iwl_he_pkt_ext_v2 *pkt_ext) +{ + if (WARN_ON(!link_sta)) + return; + + /* Initialize the PPE thresholds to "None" (7), as described in Table + * 9-262ac of 80211.ax/D3.0. + */ + memset(pkt_ext, IWL_HE_PKT_EXT_NONE, sizeof(*pkt_ext)); + + if (link_sta->eht_cap.has_eht) { + u8 nominal_padding = + u8_get_bits(link_sta->eht_cap.eht_cap_elem.phy_cap_info[5], + IEEE80211_EHT_PHY_CAP5_COMMON_NOMINAL_PKT_PAD_MASK); + + /* If PPE Thresholds exists, parse them into a FW-familiar + * format. + */ + if (link_sta->eht_cap.eht_cap_elem.phy_cap_info[5] & + IEEE80211_EHT_PHY_CAP5_PPE_THRESHOLD_PRESENT) { + u8 nss = (link_sta->eht_cap.eht_ppe_thres[0] & + IEEE80211_EHT_PPE_THRES_NSS_MASK) + 1; + u8 *ppe = &link_sta->eht_cap.eht_ppe_thres[0]; + u8 ru_index_bitmap = + u16_get_bits(*ppe, + IEEE80211_EHT_PPE_THRES_RU_INDEX_BITMASK_MASK); + /* Starting after PPE header */ + u8 ppe_pos_bit = IEEE80211_EHT_PPE_THRES_INFO_HEADER_SIZE; + + iwl_mld_parse_ppe(mld, pkt_ext, nss, ru_index_bitmap, + ppe, ppe_pos_bit, true); + /* EHT PPE Thresholds doesn't exist - set the API according to + * HE PPE Tresholds + */ + } else if (link_sta->he_cap.he_cap_elem.phy_cap_info[6] & + IEEE80211_HE_PHY_CAP6_PPE_THRESHOLD_PRESENT) { + /* Even though HE Capabilities IE doesn't contain PPE + * Thresholds for BW 320Mhz, thresholds for this BW will + * be filled in with the same values as 160Mhz, due to + * the inheritance, as required. + */ + iwl_mld_set_pkt_ext_from_he_ppe(mld, link_sta, pkt_ext, + true); + + /* According to the requirements, for MCSs 12-13 the + * maximum value between HE PPE Threshold and Common + * Nominal Packet Padding needs to be taken + */ + iwl_mld_get_optimal_ppe_info(pkt_ext, nominal_padding); + + /* if PPE Thresholds doesn't present in both EHT IE and HE IE - + * take the Thresholds from Common Nominal Packet Padding field + */ + } else { + iwl_mld_set_pkt_ext_from_nominal_padding(pkt_ext, + nominal_padding); + } + } else if (link_sta->he_cap.has_he) { + /* If PPE Thresholds exist, parse them into a FW-familiar format. */ + if (link_sta->he_cap.he_cap_elem.phy_cap_info[6] & + IEEE80211_HE_PHY_CAP6_PPE_THRESHOLD_PRESENT) { + iwl_mld_set_pkt_ext_from_he_ppe(mld, link_sta, pkt_ext, + false); + /* PPE Thresholds doesn't exist - set the API PPE values + * according to Common Nominal Packet Padding field. + */ + } else { + u8 nominal_padding = + u8_get_bits(link_sta->he_cap.he_cap_elem.phy_cap_info[9], + IEEE80211_HE_PHY_CAP9_NOMINAL_PKT_PADDING_MASK); + if (nominal_padding != IEEE80211_HE_PHY_CAP9_NOMINAL_PKT_PADDING_RESERVED) + iwl_mld_set_pkt_ext_from_nominal_padding(pkt_ext, + nominal_padding); + } + } + + for (int i = 0; i < MAX_HE_SUPP_NSS; i++) { + for (int bw = 0; + bw < ARRAY_SIZE(*pkt_ext->pkt_ext_qam_th[i]); + bw++) { + u8 *qam_th = + &pkt_ext->pkt_ext_qam_th[i][bw][0]; + + IWL_DEBUG_HT(mld, + "PPE table: nss[%d] bw[%d] PPET8 = %d, PPET16 = %d\n", + i, bw, qam_th[0], qam_th[1]); + } + } +} + +static u32 iwl_mld_get_htc_flags(struct ieee80211_link_sta *link_sta) +{ + u8 *mac_cap_info = + &link_sta->he_cap.he_cap_elem.mac_cap_info[0]; + u32 htc_flags = 0; + + if (mac_cap_info[0] & IEEE80211_HE_MAC_CAP0_HTC_HE) + htc_flags |= IWL_HE_HTC_SUPPORT; + if ((mac_cap_info[1] & IEEE80211_HE_MAC_CAP1_LINK_ADAPTATION) || + (mac_cap_info[2] & IEEE80211_HE_MAC_CAP2_LINK_ADAPTATION)) { + u8 link_adap = + ((mac_cap_info[2] & + IEEE80211_HE_MAC_CAP2_LINK_ADAPTATION) << 1) + + (mac_cap_info[1] & + IEEE80211_HE_MAC_CAP1_LINK_ADAPTATION); + + if (link_adap == 2) + htc_flags |= + IWL_HE_HTC_LINK_ADAP_UNSOLICITED; + else if (link_adap == 3) + htc_flags |= IWL_HE_HTC_LINK_ADAP_BOTH; + } + if (mac_cap_info[2] & IEEE80211_HE_MAC_CAP2_BSR) + htc_flags |= IWL_HE_HTC_BSR_SUPP; + if (mac_cap_info[3] & IEEE80211_HE_MAC_CAP3_OMI_CONTROL) + htc_flags |= IWL_HE_HTC_OMI_SUPP; + if (mac_cap_info[4] & IEEE80211_HE_MAC_CAP4_BQR) + htc_flags |= IWL_HE_HTC_BQR_SUPP; + + return htc_flags; +} + +static int iwl_mld_send_sta_cmd(struct iwl_mld *mld, + const struct iwl_sta_cfg_cmd *cmd) +{ + int ret = iwl_mld_send_cmd_pdu(mld, + WIDE_ID(MAC_CONF_GROUP, STA_CONFIG_CMD), + cmd); + if (ret) + IWL_ERR(mld, "STA_CONFIG_CMD send failed, ret=0x%x\n", ret); + return ret; +} + +static int +iwl_mld_add_modify_sta_cmd(struct iwl_mld *mld, + struct ieee80211_link_sta *link_sta) +{ + struct ieee80211_sta *sta = link_sta->sta; + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(sta); + struct ieee80211_bss_conf *link; + struct iwl_mld_link *mld_link; + struct iwl_sta_cfg_cmd cmd = {}; + int fw_id = iwl_mld_fw_sta_id_from_link_sta(mld, link_sta); + + lockdep_assert_wiphy(mld->wiphy); + + link = link_conf_dereference_protected(mld_sta->vif, + link_sta->link_id); + + mld_link = iwl_mld_link_from_mac80211(link); + + if (WARN_ON(!link || !mld_link) || fw_id < 0) + return -EINVAL; + + cmd.sta_id = cpu_to_le32(fw_id); + cmd.station_type = cpu_to_le32(mld_sta->sta_type); + cmd.link_id = cpu_to_le32(mld_link->fw_id); + + memcpy(&cmd.peer_mld_address, sta->addr, ETH_ALEN); + memcpy(&cmd.peer_link_address, link_sta->addr, ETH_ALEN); + + if (mld_sta->sta_state >= IEEE80211_STA_ASSOC) + cmd.assoc_id = cpu_to_le32(sta->aid); + + if (sta->mfp || mld_sta->sta_state < IEEE80211_STA_AUTHORIZED) + cmd.mfp = cpu_to_le32(1); + + switch (link_sta->rx_nss) { + case 1: + cmd.mimo = cpu_to_le32(0); + break; + case 2 ... 8: + cmd.mimo = cpu_to_le32(1); + break; + } + + switch (link_sta->smps_mode) { + case IEEE80211_SMPS_AUTOMATIC: + case IEEE80211_SMPS_NUM_MODES: + WARN_ON(1); + break; + case IEEE80211_SMPS_STATIC: + /* override NSS */ + cmd.mimo = cpu_to_le32(0); + break; + case IEEE80211_SMPS_DYNAMIC: + cmd.mimo_protection = cpu_to_le32(1); + break; + case IEEE80211_SMPS_OFF: + /* nothing */ + break; + } + + iwl_mld_fill_ampdu_size_and_dens(link_sta, link, + &cmd.tx_ampdu_max_size, + &cmd.tx_ampdu_spacing); + + if (sta->wme) { + cmd.sp_length = + cpu_to_le32(sta->max_sp ? sta->max_sp * 2 : 128); + cmd.uapsd_acs = cpu_to_le32(iwl_mld_get_uapsd_acs(sta)); + } + + if (link_sta->he_cap.has_he) { + cmd.trig_rnd_alloc = + cpu_to_le32(link->uora_exists ? 1 : 0); + + /* PPE Thresholds */ + iwl_mld_fill_pkt_ext(mld, link_sta, &cmd.pkt_ext); + + /* HTC flags */ + cmd.htc_flags = + cpu_to_le32(iwl_mld_get_htc_flags(link_sta)); + + if (link_sta->he_cap.he_cap_elem.mac_cap_info[2] & + IEEE80211_HE_MAC_CAP2_ACK_EN) + cmd.ack_enabled = cpu_to_le32(1); + } + + return iwl_mld_send_sta_cmd(mld, &cmd); +} + +IWL_MLD_ALLOC_FN(link_sta, link_sta) + +static int +iwl_mld_add_link_sta(struct iwl_mld *mld, struct ieee80211_link_sta *link_sta) +{ + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(link_sta->sta); + struct iwl_mld_link_sta *mld_link_sta; + int ret; + u8 fw_id; + + lockdep_assert_wiphy(mld->wiphy); + + /* We will fail to add it to the FW anyway */ + if (iwl_mld_error_before_recovery(mld)) + return -ENODEV; + + mld_link_sta = iwl_mld_link_sta_from_mac80211(link_sta); + + /* We need to preserve the fw sta ids during a restart, since the fw + * will recover SN/PN for them, this is why the mld_link_sta exists. + */ + if (mld_link_sta) { + /* But if we are not restarting, this is not OK */ + WARN_ON(!mld->fw_status.in_hw_restart); + + /* Avoid adding a STA that is already in FW to avoid an assert */ + if (WARN_ON(mld_link_sta->in_fw)) + return -EINVAL; + + fw_id = mld_link_sta->fw_id; + goto add_to_fw; + } + + /* Allocate a fw id and map it to the link_sta */ + ret = iwl_mld_allocate_link_sta_fw_id(mld, &fw_id, link_sta); + if (ret) + return ret; + + if (link_sta == &link_sta->sta->deflink) { + mld_link_sta = &mld_sta->deflink; + } else { + mld_link_sta = kzalloc(sizeof(*mld_link_sta), GFP_KERNEL); + if (!mld_link_sta) + return -ENOMEM; + } + + mld_link_sta->fw_id = fw_id; + rcu_assign_pointer(mld_sta->link[link_sta->link_id], mld_link_sta); + +add_to_fw: + ret = iwl_mld_add_modify_sta_cmd(mld, link_sta); + if (ret) { + RCU_INIT_POINTER(mld->fw_id_to_link_sta[fw_id], NULL); + RCU_INIT_POINTER(mld_sta->link[link_sta->link_id], NULL); + if (link_sta != &link_sta->sta->deflink) + kfree(mld_link_sta); + return ret; + } + mld_link_sta->in_fw = true; + + return 0; +} + +static int iwl_mld_rm_sta_from_fw(struct iwl_mld *mld, u8 fw_sta_id) +{ + struct iwl_remove_sta_cmd cmd = { + .sta_id = cpu_to_le32(fw_sta_id), + }; + int ret; + + ret = iwl_mld_send_cmd_pdu(mld, + WIDE_ID(MAC_CONF_GROUP, STA_REMOVE_CMD), + &cmd); + if (ret) + IWL_ERR(mld, "Failed to remove station. Id=%d\n", fw_sta_id); + + return ret; +} + +static void +iwl_mld_remove_link_sta(struct iwl_mld *mld, + struct ieee80211_link_sta *link_sta) +{ + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(link_sta->sta); + struct iwl_mld_link_sta *mld_link_sta = + iwl_mld_link_sta_from_mac80211(link_sta); + + if (WARN_ON(!mld_link_sta)) + return; + + iwl_mld_rm_sta_from_fw(mld, mld_link_sta->fw_id); + mld_link_sta->in_fw = false; + + /* Now that the STA doesn't exist in FW, we don't expect any new + * notifications for it. Cancel the ones that are already pending + */ + iwl_mld_cancel_notifications_of_object(mld, IWL_MLD_OBJECT_TYPE_STA, + mld_link_sta->fw_id); + + /* This will not be done upon reconfig, so do it also when + * failed to remove from fw + */ + RCU_INIT_POINTER(mld->fw_id_to_link_sta[mld_link_sta->fw_id], NULL); + RCU_INIT_POINTER(mld_sta->link[link_sta->link_id], NULL); + if (mld_link_sta != &mld_sta->deflink) + kfree_rcu(mld_link_sta, rcu_head); +} + +int iwl_mld_update_all_link_stations(struct iwl_mld *mld, + struct ieee80211_sta *sta) +{ + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(sta); + struct ieee80211_link_sta *link_sta; + int link_id; + + for_each_sta_active_link(mld_sta->vif, sta, link_sta, link_id) { + int ret = iwl_mld_add_modify_sta_cmd(mld, link_sta); + + if (ret) + return ret; + } + return 0; +} + +static void iwl_mld_destroy_sta(struct ieee80211_sta *sta) +{ + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(sta); + + kfree(mld_sta->dup_data); + kfree(mld_sta->mpdu_counters); +} + +static int +iwl_mld_alloc_dup_data(struct iwl_mld *mld, struct iwl_mld_sta *mld_sta) +{ + struct iwl_mld_rxq_dup_data *dup_data; + + if (mld->fw_status.in_hw_restart) + return 0; + + dup_data = kcalloc(mld->trans->num_rx_queues, sizeof(*dup_data), + GFP_KERNEL); + if (!dup_data) + return -ENOMEM; + + /* Initialize all the last_seq values to 0xffff which can never + * compare equal to the frame's seq_ctrl in the check in + * iwl_mld_is_dup() since the lower 4 bits are the fragment + * number and fragmented packets don't reach that function. + * + * This thus allows receiving a packet with seqno 0 and the + * retry bit set as the very first packet on a new TID. + */ + for (int q = 0; q < mld->trans->num_rx_queues; q++) + memset(dup_data[q].last_seq, 0xff, + sizeof(dup_data[q].last_seq)); + mld_sta->dup_data = dup_data; + + return 0; +} + +static void iwl_mld_alloc_mpdu_counters(struct iwl_mld *mld, + struct ieee80211_sta *sta) +{ + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(sta); + struct ieee80211_vif *vif = mld_sta->vif; + + if (mld->fw_status.in_hw_restart) + return; + + /* MPDUs are counted only when EMLSR is possible */ + if (ieee80211_vif_type_p2p(vif) != NL80211_IFTYPE_STATION || + sta->tdls || !ieee80211_vif_is_mld(vif)) + return; + + mld_sta->mpdu_counters = kcalloc(mld->trans->num_rx_queues, + sizeof(*mld_sta->mpdu_counters), + GFP_KERNEL); + if (!mld_sta->mpdu_counters) + return; + + for (int q = 0; q < mld->trans->num_rx_queues; q++) + spin_lock_init(&mld_sta->mpdu_counters[q].lock); +} + +static int +iwl_mld_init_sta(struct iwl_mld *mld, struct ieee80211_sta *sta, + struct ieee80211_vif *vif, enum iwl_fw_sta_type type) +{ + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(sta); + + mld_sta->vif = vif; + mld_sta->sta_type = type; + mld_sta->mld = mld; + + if (!mld->fw_status.in_hw_restart) + for (int i = 0; i < ARRAY_SIZE(sta->txq); i++) + iwl_mld_init_txq(iwl_mld_txq_from_mac80211(sta->txq[i])); + + iwl_mld_alloc_mpdu_counters(mld, sta); + + iwl_mld_toggle_tx_ant(mld, &mld_sta->data_tx_ant); + + return iwl_mld_alloc_dup_data(mld, mld_sta); +} + +int iwl_mld_add_sta(struct iwl_mld *mld, struct ieee80211_sta *sta, + struct ieee80211_vif *vif, enum iwl_fw_sta_type type) +{ + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(sta); + struct ieee80211_link_sta *link_sta; + int link_id; + int ret; + + ret = iwl_mld_init_sta(mld, sta, vif, type); + if (ret) + return ret; + + /* We could have add only the deflink link_sta, but it will not work + * in the restart case if the single link that is active during + * reconfig is not the deflink one. + */ + for_each_sta_active_link(mld_sta->vif, sta, link_sta, link_id) { + ret = iwl_mld_add_link_sta(mld, link_sta); + if (ret) + goto destroy_sta; + } + + return 0; + +destroy_sta: + iwl_mld_destroy_sta(sta); + + return ret; +} + +void iwl_mld_flush_sta_txqs(struct iwl_mld *mld, struct ieee80211_sta *sta) +{ + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(sta); + struct ieee80211_link_sta *link_sta; + int link_id; + + for_each_sta_active_link(mld_sta->vif, sta, link_sta, link_id) { + int fw_sta_id = iwl_mld_fw_sta_id_from_link_sta(mld, link_sta); + + if (fw_sta_id < 0) + continue; + + iwl_mld_flush_link_sta_txqs(mld, fw_sta_id); + } +} + +void iwl_mld_wait_sta_txqs_empty(struct iwl_mld *mld, struct ieee80211_sta *sta) +{ + /* Avoid a warning in iwl_trans_wait_txq_empty if are anyway on the way + * to a restart. + */ + if (iwl_mld_error_before_recovery(mld)) + return; + + for (int i = 0; i < ARRAY_SIZE(sta->txq); i++) { + struct iwl_mld_txq *mld_txq = + iwl_mld_txq_from_mac80211(sta->txq[i]); + + if (!mld_txq->status.allocated) + continue; + + iwl_trans_wait_txq_empty(mld->trans, mld_txq->fw_id); + } +} + +void iwl_mld_remove_sta(struct iwl_mld *mld, struct ieee80211_sta *sta) +{ + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(sta); + struct ieee80211_vif *vif = mld_sta->vif; + struct ieee80211_link_sta *link_sta; + u8 link_id; + + lockdep_assert_wiphy(mld->wiphy); + + /* Tell the HW to flush the queues */ + iwl_mld_flush_sta_txqs(mld, sta); + + /* Wait for trans to empty its queues */ + iwl_mld_wait_sta_txqs_empty(mld, sta); + + /* Now we can remove the queues */ + for (int i = 0; i < ARRAY_SIZE(sta->txq); i++) + iwl_mld_remove_txq(mld, sta->txq[i]); + + for_each_sta_active_link(vif, sta, link_sta, link_id) { + /* Mac8011 will remove the groupwise keys after the sta is + * removed, but FW expects all the keys to be removed before + * the STA is, so remove them all here. + */ + if (vif->type == NL80211_IFTYPE_STATION) + iwl_mld_remove_ap_keys(mld, vif, sta, link_id); + + /* Remove the link_sta */ + iwl_mld_remove_link_sta(mld, link_sta); + } + + iwl_mld_destroy_sta(sta); +} + +u32 iwl_mld_fw_sta_id_mask(struct iwl_mld *mld, struct ieee80211_sta *sta) +{ + struct ieee80211_vif *vif = iwl_mld_sta_from_mac80211(sta)->vif; + struct ieee80211_link_sta *link_sta; + unsigned int link_id; + u32 result = 0; + + KUNIT_STATIC_STUB_REDIRECT(iwl_mld_fw_sta_id_mask, mld, sta); + + /* This function should only be used with the wiphy lock held, + * In other cases, it is not guaranteed that the link_sta will exist + * in the driver too, and it is checked in + * iwl_mld_fw_sta_id_from_link_sta. + */ + lockdep_assert_wiphy(mld->wiphy); + + for_each_sta_active_link(vif, sta, link_sta, link_id) { + int fw_id = iwl_mld_fw_sta_id_from_link_sta(mld, link_sta); + + if (!(fw_id < 0)) + result |= BIT(fw_id); + } + + return result; +} +EXPORT_SYMBOL_IF_IWLWIFI_KUNIT(iwl_mld_fw_sta_id_mask); + +static void iwl_mld_count_mpdu(struct ieee80211_link_sta *link_sta, int queue, + u32 count, bool tx) +{ + struct iwl_mld_per_q_mpdu_counter *queue_counter; + struct iwl_mld_per_link_mpdu_counter *link_counter; + struct iwl_mld_vif *mld_vif; + struct iwl_mld_sta *mld_sta; + struct iwl_mld_link *mld_link; + struct iwl_mld *mld; + int total_mpdus = 0; + + if (WARN_ON(!link_sta)) + return; + + mld_sta = iwl_mld_sta_from_mac80211(link_sta->sta); + if (!mld_sta->mpdu_counters) + return; + + mld_vif = iwl_mld_vif_from_mac80211(mld_sta->vif); + mld_link = iwl_mld_link_dereference_check(mld_vif, link_sta->link_id); + + if (WARN_ON_ONCE(!mld_link)) + return; + + queue_counter = &mld_sta->mpdu_counters[queue]; + + mld = mld_vif->mld; + + /* If it the window is over, first clear the counters. + * When we are not blocked by TPT, the window is managed by check_tpt_wk + */ + if ((mld_vif->emlsr.blocked_reasons & IWL_MLD_EMLSR_BLOCKED_TPT) && + time_is_before_jiffies(queue_counter->window_start_time + + IWL_MLD_TPT_COUNT_WINDOW)) { + memset(queue_counter->per_link, 0, + sizeof(queue_counter->per_link)); + queue_counter->window_start_time = jiffies; + + IWL_DEBUG_INFO(mld, "MPDU counters are cleared\n"); + } + + link_counter = &queue_counter->per_link[mld_link->fw_id]; + + spin_lock_bh(&queue_counter->lock); + + /* Update the statistics for this TPT measurement window */ + if (tx) + link_counter->tx += count; + else + link_counter->rx += count; + + /* + * Next, evaluate whether we should queue an unblock, + * skip this if we are not blocked due to low throughput. + */ + if (!(mld_vif->emlsr.blocked_reasons & IWL_MLD_EMLSR_BLOCKED_TPT)) + goto unlock; + + for (int i = 0; i <= IWL_FW_MAX_LINK_ID; i++) + total_mpdus += tx ? queue_counter->per_link[i].tx : + queue_counter->per_link[i].rx; + + /* Unblock is already queued if the threshold was reached before */ + if (total_mpdus - count >= IWL_MLD_ENTER_EMLSR_TPT_THRESH) + goto unlock; + + if (total_mpdus >= IWL_MLD_ENTER_EMLSR_TPT_THRESH) + wiphy_work_queue(mld->wiphy, &mld_vif->emlsr.unblock_tpt_wk); + +unlock: + spin_unlock_bh(&queue_counter->lock); +} + +/* must be called under rcu_read_lock() */ +void iwl_mld_count_mpdu_rx(struct ieee80211_link_sta *link_sta, int queue, + u32 count) +{ + iwl_mld_count_mpdu(link_sta, queue, count, false); +} + +/* must be called under rcu_read_lock() */ +void iwl_mld_count_mpdu_tx(struct ieee80211_link_sta *link_sta, u32 count) +{ + /* use queue 0 for all TX */ + iwl_mld_count_mpdu(link_sta, 0, count, true); +} + +static int iwl_mld_allocate_internal_txq(struct iwl_mld *mld, + struct iwl_mld_int_sta *internal_sta, + u8 tid) +{ + u32 sta_mask = BIT(internal_sta->sta_id); + int queue, size; + + size = max_t(u32, IWL_MGMT_QUEUE_SIZE, + mld->trans->cfg->min_txq_size); + + queue = iwl_trans_txq_alloc(mld->trans, 0, sta_mask, tid, size, + IWL_WATCHDOG_DISABLED); + + if (queue >= 0) + IWL_DEBUG_TX_QUEUES(mld, + "Enabling TXQ #%d for sta mask 0x%x tid %d\n", + queue, sta_mask, tid); + return queue; +} + +static int iwl_mld_send_aux_sta_cmd(struct iwl_mld *mld, + const struct iwl_mld_int_sta *internal_sta) +{ + struct iwl_aux_sta_cmd cmd = { + .sta_id = cpu_to_le32(internal_sta->sta_id), + /* TODO: CDB - properly set the lmac_id */ + .lmac_id = cpu_to_le32(IWL_LMAC_24G_INDEX), + }; + + return iwl_mld_send_cmd_pdu(mld, WIDE_ID(MAC_CONF_GROUP, AUX_STA_CMD), + &cmd); +} + +static int +iwl_mld_add_internal_sta_to_fw(struct iwl_mld *mld, + const struct iwl_mld_int_sta *internal_sta, + u8 fw_link_id, + const u8 *addr) +{ + struct iwl_sta_cfg_cmd cmd = {}; + + if (internal_sta->sta_type == STATION_TYPE_AUX) + return iwl_mld_send_aux_sta_cmd(mld, internal_sta); + + cmd.sta_id = cpu_to_le32((u8)internal_sta->sta_id); + cmd.link_id = cpu_to_le32(fw_link_id); + cmd.station_type = cpu_to_le32(internal_sta->sta_type); + + /* FW doesn't allow to add a IGTK/BIGTK if the sta isn't marked as MFP. + * On the other hand, FW will never check this flag during RX since + * an AP/GO doesn't receive protected broadcast management frames. + * So, we can set it unconditionally. + */ + if (internal_sta->sta_type == STATION_TYPE_BCAST_MGMT) + cmd.mfp = cpu_to_le32(1); + + if (addr) { + memcpy(cmd.peer_mld_address, addr, ETH_ALEN); + memcpy(cmd.peer_link_address, addr, ETH_ALEN); + } + + return iwl_mld_send_sta_cmd(mld, &cmd); +} + +static int iwl_mld_add_internal_sta(struct iwl_mld *mld, + struct iwl_mld_int_sta *internal_sta, + enum iwl_fw_sta_type sta_type, + u8 fw_link_id, const u8 *addr, u8 tid) +{ + int ret, queue_id; + + ret = iwl_mld_allocate_link_sta_fw_id(mld, + &internal_sta->sta_id, + ERR_PTR(-EINVAL)); + if (ret) + return ret; + + internal_sta->sta_type = sta_type; + + ret = iwl_mld_add_internal_sta_to_fw(mld, internal_sta, fw_link_id, + addr); + if (ret) + goto err; + + queue_id = iwl_mld_allocate_internal_txq(mld, internal_sta, tid); + if (queue_id < 0) { + iwl_mld_rm_sta_from_fw(mld, internal_sta->sta_id); + ret = queue_id; + goto err; + } + + internal_sta->queue_id = queue_id; + + return 0; +err: + iwl_mld_free_internal_sta(mld, internal_sta); + return ret; +} + +int iwl_mld_add_bcast_sta(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link) +{ + struct iwl_mld_link *mld_link = iwl_mld_link_from_mac80211(link); + const u8 bcast_addr[] = {0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF}; + const u8 *addr; + + if (WARN_ON(!mld_link)) + return -EINVAL; + + if (WARN_ON(vif->type != NL80211_IFTYPE_AP && + vif->type != NL80211_IFTYPE_ADHOC)) + return -EINVAL; + + addr = vif->type == NL80211_IFTYPE_ADHOC ? link->bssid : bcast_addr; + + return iwl_mld_add_internal_sta(mld, &mld_link->bcast_sta, + STATION_TYPE_BCAST_MGMT, + mld_link->fw_id, addr, + IWL_MGMT_TID); +} + +int iwl_mld_add_mcast_sta(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link) +{ + struct iwl_mld_link *mld_link = iwl_mld_link_from_mac80211(link); + const u8 mcast_addr[] = {0x03, 0x00, 0x00, 0x00, 0x00, 0x00}; + + if (WARN_ON(!mld_link)) + return -EINVAL; + + if (WARN_ON(vif->type != NL80211_IFTYPE_AP && + vif->type != NL80211_IFTYPE_ADHOC)) + return -EINVAL; + + return iwl_mld_add_internal_sta(mld, &mld_link->mcast_sta, + STATION_TYPE_MCAST, + mld_link->fw_id, mcast_addr, 0); +} + +int iwl_mld_add_aux_sta(struct iwl_mld *mld, + struct iwl_mld_int_sta *internal_sta) +{ + return iwl_mld_add_internal_sta(mld, internal_sta, STATION_TYPE_AUX, + 0, NULL, IWL_MAX_TID_COUNT); +} + +static void iwl_mld_remove_internal_sta(struct iwl_mld *mld, + struct iwl_mld_int_sta *internal_sta, + bool flush, u8 tid) +{ + if (WARN_ON_ONCE(internal_sta->sta_id == IWL_INVALID_STA || + internal_sta->queue_id == IWL_MLD_INVALID_QUEUE)) + return; + + if (flush) + iwl_mld_flush_link_sta_txqs(mld, internal_sta->sta_id); + + iwl_mld_free_txq(mld, BIT(internal_sta->sta_id), + tid, internal_sta->queue_id); + + iwl_mld_rm_sta_from_fw(mld, internal_sta->sta_id); + + iwl_mld_free_internal_sta(mld, internal_sta); +} + +void iwl_mld_remove_bcast_sta(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link) +{ + struct iwl_mld_link *mld_link = iwl_mld_link_from_mac80211(link); + + if (WARN_ON(!mld_link)) + return; + + if (WARN_ON(vif->type != NL80211_IFTYPE_AP && + vif->type != NL80211_IFTYPE_ADHOC)) + return; + + iwl_mld_remove_internal_sta(mld, &mld_link->bcast_sta, true, + IWL_MGMT_TID); +} + +void iwl_mld_remove_mcast_sta(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link) +{ + struct iwl_mld_link *mld_link = iwl_mld_link_from_mac80211(link); + + if (WARN_ON(!mld_link)) + return; + + if (WARN_ON(vif->type != NL80211_IFTYPE_AP && + vif->type != NL80211_IFTYPE_ADHOC)) + return; + + iwl_mld_remove_internal_sta(mld, &mld_link->mcast_sta, true, 0); +} + +void iwl_mld_remove_aux_sta(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link) +{ + struct iwl_mld_link *mld_link = iwl_mld_link_from_mac80211(link); + + if (WARN_ON(!mld_link)) + return; + + /* TODO: Hotspot 2.0 */ + if (WARN_ON(vif->type != NL80211_IFTYPE_P2P_DEVICE)) + return; + + iwl_mld_remove_internal_sta(mld, &mld_link->aux_sta, false, + IWL_MAX_TID_COUNT); +} + +static int iwl_mld_update_sta_resources(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + u32 old_sta_mask, + u32 new_sta_mask) +{ + int ret; + + ret = iwl_mld_update_sta_txqs(mld, sta, old_sta_mask, new_sta_mask); + if (ret) + return ret; + + ret = iwl_mld_update_sta_keys(mld, vif, sta, old_sta_mask, new_sta_mask); + if (ret) + return ret; + + return iwl_mld_update_sta_baids(mld, old_sta_mask, new_sta_mask); +} + +int iwl_mld_update_link_stas(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + u16 old_links, u16 new_links) +{ + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(sta); + struct iwl_mld_link_sta *mld_link_sta; + unsigned long links_to_add = ~old_links & new_links; + unsigned long links_to_rem = old_links & ~new_links; + unsigned long old_links_long = old_links; + unsigned long sta_mask_added = 0; + u32 current_sta_mask = 0, sta_mask_to_rem = 0; + unsigned int link_id, sta_id; + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + for_each_set_bit(link_id, &old_links_long, + IEEE80211_MLD_MAX_NUM_LINKS) { + mld_link_sta = + iwl_mld_link_sta_dereference_check(mld_sta, link_id); + + if (WARN_ON(!mld_link_sta)) + return -EINVAL; + + current_sta_mask |= BIT(mld_link_sta->fw_id); + if (links_to_rem & BIT(link_id)) + sta_mask_to_rem |= BIT(mld_link_sta->fw_id); + } + + if (sta_mask_to_rem) { + ret = iwl_mld_update_sta_resources(mld, vif, sta, + current_sta_mask, + current_sta_mask & + ~sta_mask_to_rem); + if (ret) + return ret; + + current_sta_mask &= ~sta_mask_to_rem; + } + + for_each_set_bit(link_id, &links_to_rem, IEEE80211_MLD_MAX_NUM_LINKS) { + struct ieee80211_link_sta *link_sta = + link_sta_dereference_protected(sta, link_id); + + if (WARN_ON(!link_sta)) + return -EINVAL; + + iwl_mld_remove_link_sta(mld, link_sta); + } + + for_each_set_bit(link_id, &links_to_add, IEEE80211_MLD_MAX_NUM_LINKS) { + struct ieee80211_link_sta *link_sta = + link_sta_dereference_protected(sta, link_id); + struct ieee80211_bss_conf *link; + + if (WARN_ON(!link_sta)) + return -EINVAL; + + ret = iwl_mld_add_link_sta(mld, link_sta); + if (ret) + goto remove_added_link_stas; + + mld_link_sta = + iwl_mld_link_sta_dereference_check(mld_sta, + link_id); + + link = link_conf_dereference_protected(mld_sta->vif, + link_sta->link_id); + iwl_mld_config_tlc_link(mld, vif, link, link_sta); + + sta_mask_added |= BIT(mld_link_sta->fw_id); + } + + if (sta_mask_added) { + ret = iwl_mld_update_sta_resources(mld, vif, sta, + current_sta_mask, + current_sta_mask | + sta_mask_added); + if (ret) + goto remove_added_link_stas; + } + + /* We couldn't activate the links before it has a STA. Now we can */ + for_each_set_bit(link_id, &links_to_add, IEEE80211_MLD_MAX_NUM_LINKS) { + struct ieee80211_bss_conf *link = + link_conf_dereference_protected(mld_sta->vif, link_id); + + if (WARN_ON(!link)) + continue; + + iwl_mld_activate_link(mld, link); + } + + return 0; + +remove_added_link_stas: + for_each_set_bit(sta_id, &sta_mask_added, mld->fw->ucode_capa.num_stations) { + struct ieee80211_link_sta *link_sta = + wiphy_dereference(mld->wiphy, + mld->fw_id_to_link_sta[sta_id]); + + if (WARN_ON(!link_sta)) + continue; + + iwl_mld_remove_link_sta(mld, link_sta); + } + + return ret; +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/sta.h b/drivers/net/wireless/intel/iwlwifi/mld/sta.h new file mode 100644 index 0000000000000..ddcffd7b9fde1 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/sta.h @@ -0,0 +1,266 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024-2025 Intel Corporation + */ + +#ifndef __iwl_mld_sta_h__ +#define __iwl_mld_sta_h__ + +#include + +#include "mld.h" +#include "tx.h" + +/** + * struct iwl_mld_rxq_dup_data - Duplication detection data, per STA & Rx queue + * @last_seq: last sequence per tid. + * @last_sub_frame_idx: the index of the last subframe in an A-MSDU. This value + * will be zero if the packet is not part of an A-MSDU. + */ +struct iwl_mld_rxq_dup_data { + __le16 last_seq[IWL_MAX_TID_COUNT + 1]; + u8 last_sub_frame_idx[IWL_MAX_TID_COUNT + 1]; +} ____cacheline_aligned_in_smp; + +/** + * struct iwl_mld_link_sta - link-level station + * + * This represents the link-level sta - the driver level equivalent to the + * ieee80211_link_sta + * + * @last_rate_n_flags: rate_n_flags from the last &iwl_tlc_update_notif + * @signal_avg: the signal average coming from the firmware + * @in_fw: whether the link STA is uploaded to the FW (false during restart) + * @rcu_head: RCU head for freeing this object + * @fw_id: the FW id of this link sta. + */ +struct iwl_mld_link_sta { + /* Add here fields that need clean up on restart */ + struct_group(zeroed_on_hw_restart, + u32 last_rate_n_flags; + bool in_fw; + s8 signal_avg; + ); + /* And here fields that survive a fw restart */ + struct rcu_head rcu_head; + u32 fw_id; +}; + +#define iwl_mld_link_sta_dereference_check(mld_sta, link_id) \ + rcu_dereference_check((mld_sta)->link[link_id], \ + lockdep_is_held(&mld_sta->mld->wiphy->mtx)) + +#define for_each_mld_link_sta(mld_sta, link_sta, link_id) \ + for (link_id = 0; link_id < ARRAY_SIZE((mld_sta)->link); \ + link_id++) \ + if ((link_sta = \ + iwl_mld_link_sta_dereference_check(mld_sta, link_id))) + +#define IWL_NUM_DEFAULT_KEYS 4 + +/* struct iwl_mld_ptk_pn - Holds Packet Number (PN) per TID. + * @rcu_head: RCU head for freeing this data. + * @pn: Array storing PN for each TID. + */ +struct iwl_mld_ptk_pn { + struct rcu_head rcu_head; + struct { + u8 pn[IWL_MAX_TID_COUNT][IEEE80211_CCMP_PN_LEN]; + } ____cacheline_aligned_in_smp q[]; +}; + +/** + * struct iwl_mld_per_link_mpdu_counter - per-link TX/RX MPDU counters + * + * @tx: Number of TX MPDUs. + * @rx: Number of RX MPDUs. + */ +struct iwl_mld_per_link_mpdu_counter { + u32 tx; + u32 rx; +}; + +/** + * struct iwl_mld_per_q_mpdu_counter - per-queue MPDU counter + * + * @lock: Needed to protect the counters when modified from statistics. + * @per_link: per-link counters. + * @window_start_time: timestamp of the counting-window start + */ +struct iwl_mld_per_q_mpdu_counter { + spinlock_t lock; + struct iwl_mld_per_link_mpdu_counter per_link[IWL_FW_MAX_LINK_ID + 1]; + unsigned long window_start_time; +} ____cacheline_aligned_in_smp; + +/** + * struct iwl_mld_sta - representation of a station in the driver. + * + * This represent the MLD-level sta, and will not be added to the FW. + * Embedded in ieee80211_sta. + * + * @vif: pointer the vif object. + * @sta_state: station state according to enum %ieee80211_sta_state + * @sta_type: type of this station. See &enum iwl_fw_sta_type + * @mld: a pointer to the iwl_mld object + * @dup_data: per queue duplicate packet detection data + * @data_tx_ant: stores the last TX antenna index; used for setting + * TX rate_n_flags for injected data frames (toggles on every TX failure). + * @tid_to_baid: a simple map of TID to Block-Ack fw id + * @deflink: This holds the default link STA information, for non MLO STA all + * link specific STA information is accessed through @deflink or through + * link[0] which points to address of @deflink. For MLO Link STA + * the first added link STA will point to deflink. + * @link: reference to Link Sta entries. For Non MLO STA, except 1st link, + * i.e link[0] all links would be assigned to NULL by default and + * would access link information via @deflink or link[0]. For MLO + * STA, first link STA being added will point its link pointer to + * @deflink address and remaining would be allocated and the address + * would be assigned to link[link_id] where link_id is the id assigned + * by the AP. + * @ptk_pn: Array of pointers to PTK PN data, used to track the Packet Number + * per key index and per queue (TID). + * @mpdu_counters: RX/TX MPDUs counters for each queue. + */ +struct iwl_mld_sta { + /* Add here fields that need clean up on restart */ + struct_group(zeroed_on_hw_restart, + enum ieee80211_sta_state sta_state; + enum iwl_fw_sta_type sta_type; + ); + /* And here fields that survive a fw restart */ + struct iwl_mld *mld; + struct ieee80211_vif *vif; + struct iwl_mld_rxq_dup_data *dup_data; + u8 tid_to_baid[IWL_MAX_TID_COUNT]; + u8 data_tx_ant; + + struct iwl_mld_link_sta deflink; + struct iwl_mld_link_sta __rcu *link[IEEE80211_MLD_MAX_NUM_LINKS]; + struct iwl_mld_ptk_pn __rcu *ptk_pn[IWL_NUM_DEFAULT_KEYS]; + struct iwl_mld_per_q_mpdu_counter *mpdu_counters; +}; + +static inline struct iwl_mld_sta * +iwl_mld_sta_from_mac80211(struct ieee80211_sta *sta) +{ + return (void *)sta->drv_priv; +} + +static inline void +iwl_mld_cleanup_sta(void *data, struct ieee80211_sta *sta) +{ + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(sta); + struct iwl_mld_link_sta *mld_link_sta; + u8 link_id; + + for (int i = 0; i < ARRAY_SIZE(sta->txq); i++) + CLEANUP_STRUCT(iwl_mld_txq_from_mac80211(sta->txq[i])); + + for_each_mld_link_sta(mld_sta, mld_link_sta, link_id) { + CLEANUP_STRUCT(mld_link_sta); + + if (!ieee80211_vif_is_mld(mld_sta->vif)) { + /* not an MLD STA; only has the deflink with ID zero */ + WARN_ON(link_id); + continue; + } + + if (mld_sta->vif->active_links & BIT(link_id)) + continue; + + /* Should not happen as link removal should always succeed */ + WARN_ON(1); + RCU_INIT_POINTER(mld_sta->link[link_id], NULL); + RCU_INIT_POINTER(mld_sta->mld->fw_id_to_link_sta[mld_link_sta->fw_id], + NULL); + if (mld_link_sta != &mld_sta->deflink) + kfree_rcu(mld_link_sta, rcu_head); + } + + CLEANUP_STRUCT(mld_sta); +} + +static inline struct iwl_mld_link_sta * +iwl_mld_link_sta_from_mac80211(struct ieee80211_link_sta *link_sta) +{ + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(link_sta->sta); + + return iwl_mld_link_sta_dereference_check(mld_sta, link_sta->link_id); +} + +int iwl_mld_add_sta(struct iwl_mld *mld, struct ieee80211_sta *sta, + struct ieee80211_vif *vif, enum iwl_fw_sta_type type); +void iwl_mld_remove_sta(struct iwl_mld *mld, struct ieee80211_sta *sta); +int iwl_mld_fw_sta_id_from_link_sta(struct iwl_mld *mld, + struct ieee80211_link_sta *link_sta); +u32 iwl_mld_fw_sta_id_mask(struct iwl_mld *mld, struct ieee80211_sta *sta); +int iwl_mld_update_all_link_stations(struct iwl_mld *mld, + struct ieee80211_sta *sta); +void iwl_mld_flush_sta_txqs(struct iwl_mld *mld, struct ieee80211_sta *sta); +void iwl_mld_wait_sta_txqs_empty(struct iwl_mld *mld, + struct ieee80211_sta *sta); +void iwl_mld_count_mpdu_rx(struct ieee80211_link_sta *link_sta, int queue, + u32 count); +void iwl_mld_count_mpdu_tx(struct ieee80211_link_sta *link_sta, u32 count); + +/** + * struct iwl_mld_int_sta - representation of an internal station + * (a station that exist in FW and in driver, but not in mac80211) + * + * @sta_id: the index of the station in the fw + * @queue_id: the if of the queue used by the station + * @sta_type: station type. One of &iwl_fw_sta_type + */ +struct iwl_mld_int_sta { + u8 sta_id; + u32 queue_id; + enum iwl_fw_sta_type sta_type; +}; + +static inline void +iwl_mld_init_internal_sta(struct iwl_mld_int_sta *internal_sta) +{ + internal_sta->sta_id = IWL_INVALID_STA; + internal_sta->queue_id = IWL_MLD_INVALID_QUEUE; +} + +static inline void +iwl_mld_free_internal_sta(struct iwl_mld *mld, + struct iwl_mld_int_sta *internal_sta) +{ + if (WARN_ON(internal_sta->sta_id == IWL_INVALID_STA)) + return; + + RCU_INIT_POINTER(mld->fw_id_to_link_sta[internal_sta->sta_id], NULL); + iwl_mld_init_internal_sta(internal_sta); +} + +int iwl_mld_add_bcast_sta(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link); + +int iwl_mld_add_mcast_sta(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link); + +int iwl_mld_add_aux_sta(struct iwl_mld *mld, + struct iwl_mld_int_sta *internal_sta); + +void iwl_mld_remove_bcast_sta(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link); + +void iwl_mld_remove_mcast_sta(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link); + +void iwl_mld_remove_aux_sta(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link); + +int iwl_mld_update_link_stas(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + u16 old_links, u16 new_links); +#endif /* __iwl_mld_sta_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/stats.c b/drivers/net/wireless/intel/iwlwifi/mld/stats.c new file mode 100644 index 0000000000000..842b9b9fdd8c7 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/stats.c @@ -0,0 +1,509 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024-2025 Intel Corporation + */ + +#include "mld.h" +#include "stats.h" +#include "sta.h" +#include "mlo.h" +#include "hcmd.h" +#include "iface.h" +#include "scan.h" +#include "phy.h" +#include "fw/api/stats.h" + +static int iwl_mld_send_fw_stats_cmd(struct iwl_mld *mld, u32 cfg_mask, + u32 cfg_time, u32 type_mask) +{ + u32 cmd_id = WIDE_ID(SYSTEM_GROUP, SYSTEM_STATISTICS_CMD); + struct iwl_system_statistics_cmd stats_cmd = { + .cfg_mask = cpu_to_le32(cfg_mask), + .config_time_sec = cpu_to_le32(cfg_time), + .type_id_mask = cpu_to_le32(type_mask), + }; + + return iwl_mld_send_cmd_pdu(mld, cmd_id, &stats_cmd); +} + +int iwl_mld_clear_stats_in_fw(struct iwl_mld *mld) +{ + u32 cfg_mask = IWL_STATS_CFG_FLG_ON_DEMAND_NTFY_MSK; + u32 type_mask = IWL_STATS_NTFY_TYPE_ID_OPER | + IWL_STATS_NTFY_TYPE_ID_OPER_PART1; + + return iwl_mld_send_fw_stats_cmd(mld, cfg_mask, 0, type_mask); +} + +static void +iwl_mld_fill_stats_from_oper_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt, + u8 fw_sta_id, struct station_info *sinfo) +{ + const struct iwl_system_statistics_notif_oper *notif = + (void *)&pkt->data; + const struct iwl_stats_ntfy_per_sta *per_sta = + ¬if->per_sta[fw_sta_id]; + struct ieee80211_link_sta *link_sta; + struct iwl_mld_link_sta *mld_link_sta; + + /* 0 isn't a valid value, but FW might send 0. + * In that case, set the latest non-zero value we stored + */ + rcu_read_lock(); + + link_sta = rcu_dereference(mld->fw_id_to_link_sta[fw_sta_id]); + if (IS_ERR_OR_NULL(link_sta)) + goto unlock; + + mld_link_sta = iwl_mld_link_sta_from_mac80211(link_sta); + if (WARN_ON(!mld_link_sta)) + goto unlock; + + if (per_sta->average_energy) + mld_link_sta->signal_avg = + -(s8)le32_to_cpu(per_sta->average_energy); + + sinfo->signal_avg = mld_link_sta->signal_avg; + sinfo->filled |= BIT_ULL(NL80211_STA_INFO_SIGNAL_AVG); + +unlock: + rcu_read_unlock(); +} + +struct iwl_mld_stats_data { + u8 fw_sta_id; + struct station_info *sinfo; + struct iwl_mld *mld; +}; + +static bool iwl_mld_wait_stats_handler(struct iwl_notif_wait_data *notif_data, + struct iwl_rx_packet *pkt, void *data) +{ + struct iwl_mld_stats_data *stats_data = data; + u16 cmd = WIDE_ID(pkt->hdr.group_id, pkt->hdr.cmd); + + switch (cmd) { + case WIDE_ID(STATISTICS_GROUP, STATISTICS_OPER_NOTIF): + iwl_mld_fill_stats_from_oper_notif(stats_data->mld, pkt, + stats_data->fw_sta_id, + stats_data->sinfo); + break; + case WIDE_ID(STATISTICS_GROUP, STATISTICS_OPER_PART1_NOTIF): + break; + case WIDE_ID(SYSTEM_GROUP, SYSTEM_STATISTICS_END_NOTIF): + return true; + } + + return false; +} + +static int +iwl_mld_fw_stats_to_mac80211(struct iwl_mld *mld, struct iwl_mld_sta *mld_sta, + struct station_info *sinfo) +{ + u32 cfg_mask = IWL_STATS_CFG_FLG_ON_DEMAND_NTFY_MSK | + IWL_STATS_CFG_FLG_RESET_MSK; + u32 type_mask = IWL_STATS_NTFY_TYPE_ID_OPER | + IWL_STATS_NTFY_TYPE_ID_OPER_PART1; + static const u16 notifications[] = { + WIDE_ID(STATISTICS_GROUP, STATISTICS_OPER_NOTIF), + WIDE_ID(STATISTICS_GROUP, STATISTICS_OPER_PART1_NOTIF), + WIDE_ID(SYSTEM_GROUP, SYSTEM_STATISTICS_END_NOTIF), + }; + struct iwl_mld_stats_data wait_stats_data = { + /* We don't support drv_sta_statistics in EMLSR */ + .fw_sta_id = mld_sta->deflink.fw_id, + .sinfo = sinfo, + .mld = mld, + }; + struct iwl_notification_wait stats_wait; + int ret; + + iwl_init_notification_wait(&mld->notif_wait, &stats_wait, + notifications, ARRAY_SIZE(notifications), + iwl_mld_wait_stats_handler, + &wait_stats_data); + + ret = iwl_mld_send_fw_stats_cmd(mld, cfg_mask, 0, type_mask); + if (ret) { + iwl_remove_notification(&mld->notif_wait, &stats_wait); + return ret; + } + + /* Wait 500ms for OPERATIONAL, PART1, and END notifications, + * which should be sufficient for the firmware to gather data + * from all LMACs and send notifications to the host. + */ + ret = iwl_wait_notification(&mld->notif_wait, &stats_wait, HZ / 2); + if (ret) + return ret; + + /* When periodic statistics are sent, FW will clear its statistics DB. + * If the statistics request here happens shortly afterwards, + * the response will contain data collected over a short time + * interval. The response we got here shouldn't be processed by + * the general statistics processing because it's incomplete. + * So, we delete it from the list so it won't be processed. + */ + iwl_mld_delete_handlers(mld, notifications, ARRAY_SIZE(notifications)); + + return 0; +} + +#define PERIODIC_STATS_SECONDS 5 + +int iwl_mld_request_periodic_fw_stats(struct iwl_mld *mld, bool enable) +{ + u32 cfg_mask = enable ? 0 : IWL_STATS_CFG_FLG_DISABLE_NTFY_MSK; + u32 type_mask = enable ? (IWL_STATS_NTFY_TYPE_ID_OPER | + IWL_STATS_NTFY_TYPE_ID_OPER_PART1) : 0; + u32 cfg_time = enable ? PERIODIC_STATS_SECONDS : 0; + + return iwl_mld_send_fw_stats_cmd(mld, cfg_mask, cfg_time, type_mask); +} + +static void iwl_mld_sta_stats_fill_txrate(struct iwl_mld_sta *mld_sta, + struct station_info *sinfo) +{ + struct rate_info *rinfo = &sinfo->txrate; + u32 rate_n_flags = mld_sta->deflink.last_rate_n_flags; + u32 format = rate_n_flags & RATE_MCS_MOD_TYPE_MSK; + u32 gi_ltf; + + sinfo->filled |= BIT_ULL(NL80211_STA_INFO_TX_BITRATE); + + switch (rate_n_flags & RATE_MCS_CHAN_WIDTH_MSK) { + case RATE_MCS_CHAN_WIDTH_20: + rinfo->bw = RATE_INFO_BW_20; + break; + case RATE_MCS_CHAN_WIDTH_40: + rinfo->bw = RATE_INFO_BW_40; + break; + case RATE_MCS_CHAN_WIDTH_80: + rinfo->bw = RATE_INFO_BW_80; + break; + case RATE_MCS_CHAN_WIDTH_160: + rinfo->bw = RATE_INFO_BW_160; + break; + case RATE_MCS_CHAN_WIDTH_320: + rinfo->bw = RATE_INFO_BW_320; + break; + } + + if (format == RATE_MCS_CCK_MSK || format == RATE_MCS_LEGACY_OFDM_MSK) { + int rate = u32_get_bits(rate_n_flags, RATE_LEGACY_RATE_MSK); + + /* add the offset needed to get to the legacy ofdm indices */ + if (format == RATE_MCS_LEGACY_OFDM_MSK) + rate += IWL_FIRST_OFDM_RATE; + + switch (rate) { + case IWL_RATE_1M_INDEX: + rinfo->legacy = 10; + break; + case IWL_RATE_2M_INDEX: + rinfo->legacy = 20; + break; + case IWL_RATE_5M_INDEX: + rinfo->legacy = 55; + break; + case IWL_RATE_11M_INDEX: + rinfo->legacy = 110; + break; + case IWL_RATE_6M_INDEX: + rinfo->legacy = 60; + break; + case IWL_RATE_9M_INDEX: + rinfo->legacy = 90; + break; + case IWL_RATE_12M_INDEX: + rinfo->legacy = 120; + break; + case IWL_RATE_18M_INDEX: + rinfo->legacy = 180; + break; + case IWL_RATE_24M_INDEX: + rinfo->legacy = 240; + break; + case IWL_RATE_36M_INDEX: + rinfo->legacy = 360; + break; + case IWL_RATE_48M_INDEX: + rinfo->legacy = 480; + break; + case IWL_RATE_54M_INDEX: + rinfo->legacy = 540; + } + return; + } + + rinfo->nss = u32_get_bits(rate_n_flags, RATE_MCS_NSS_MSK) + 1; + + if (format == RATE_MCS_HT_MSK) + rinfo->mcs = RATE_HT_MCS_INDEX(rate_n_flags); + else + rinfo->mcs = u32_get_bits(rate_n_flags, RATE_MCS_CODE_MSK); + + if (rate_n_flags & RATE_MCS_SGI_MSK) + rinfo->flags |= RATE_INFO_FLAGS_SHORT_GI; + + switch (format) { + case RATE_MCS_EHT_MSK: + rinfo->flags |= RATE_INFO_FLAGS_EHT_MCS; + break; + case RATE_MCS_HE_MSK: + gi_ltf = u32_get_bits(rate_n_flags, RATE_MCS_HE_GI_LTF_MSK); + + rinfo->flags |= RATE_INFO_FLAGS_HE_MCS; + + if (rate_n_flags & RATE_MCS_HE_106T_MSK) { + rinfo->bw = RATE_INFO_BW_HE_RU; + rinfo->he_ru_alloc = NL80211_RATE_INFO_HE_RU_ALLOC_106; + } + + switch (rate_n_flags & RATE_MCS_HE_TYPE_MSK) { + case RATE_MCS_HE_TYPE_SU: + case RATE_MCS_HE_TYPE_EXT_SU: + if (gi_ltf == 0 || gi_ltf == 1) + rinfo->he_gi = NL80211_RATE_INFO_HE_GI_0_8; + else if (gi_ltf == 2) + rinfo->he_gi = NL80211_RATE_INFO_HE_GI_1_6; + else if (gi_ltf == 3) + rinfo->he_gi = NL80211_RATE_INFO_HE_GI_3_2; + else + rinfo->he_gi = NL80211_RATE_INFO_HE_GI_0_8; + break; + case RATE_MCS_HE_TYPE_MU: + if (gi_ltf == 0 || gi_ltf == 1) + rinfo->he_gi = NL80211_RATE_INFO_HE_GI_0_8; + else if (gi_ltf == 2) + rinfo->he_gi = NL80211_RATE_INFO_HE_GI_1_6; + else + rinfo->he_gi = NL80211_RATE_INFO_HE_GI_3_2; + break; + case RATE_MCS_HE_TYPE_TRIG: + if (gi_ltf == 0 || gi_ltf == 1) + rinfo->he_gi = NL80211_RATE_INFO_HE_GI_1_6; + else + rinfo->he_gi = NL80211_RATE_INFO_HE_GI_3_2; + break; + } + + if (rate_n_flags & RATE_HE_DUAL_CARRIER_MODE_MSK) + rinfo->he_dcm = 1; + break; + case RATE_MCS_HT_MSK: + rinfo->flags |= RATE_INFO_FLAGS_MCS; + break; + case RATE_MCS_VHT_MSK: + rinfo->flags |= RATE_INFO_FLAGS_VHT_MCS; + break; + } +} + +void iwl_mld_mac80211_sta_statistics(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + struct station_info *sinfo) +{ + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(sta); + + /* This API is not EMLSR ready, so we cannot provide complete + * information if EMLSR is active + */ + if (hweight16(vif->active_links) > 1) + return; + + if (iwl_mld_fw_stats_to_mac80211(mld_sta->mld, mld_sta, sinfo)) + return; + + iwl_mld_sta_stats_fill_txrate(mld_sta, sinfo); + + /* TODO: NL80211_STA_INFO_BEACON_RX */ + + /* TODO: NL80211_STA_INFO_BEACON_SIGNAL_AVG */ +} + +#define IWL_MLD_TRAFFIC_LOAD_MEDIUM_THRESH 10 /* percentage */ +#define IWL_MLD_TRAFFIC_LOAD_HIGH_THRESH 50 /* percentage */ +#define IWL_MLD_TRAFFIC_LOAD_MIN_WINDOW_USEC (500 * 1000) + +static u8 iwl_mld_stats_load_percentage(u32 last_ts_usec, u32 curr_ts_usec, + u32 total_airtime_usec) +{ + u32 elapsed_usec = curr_ts_usec - last_ts_usec; + + if (elapsed_usec < IWL_MLD_TRAFFIC_LOAD_MIN_WINDOW_USEC) + return 0; + + return (100 * total_airtime_usec / elapsed_usec); +} + +static void iwl_mld_stats_recalc_traffic_load(struct iwl_mld *mld, + u32 total_airtime_usec, + u32 curr_ts_usec) +{ + u32 last_ts_usec = mld->scan.traffic_load.last_stats_ts_usec; + u8 load_prec; + + /* Skip the calculation as this is the first notification received */ + if (!last_ts_usec) + goto out; + + load_prec = iwl_mld_stats_load_percentage(last_ts_usec, curr_ts_usec, + total_airtime_usec); + + if (load_prec > IWL_MLD_TRAFFIC_LOAD_HIGH_THRESH) + mld->scan.traffic_load.status = IWL_MLD_TRAFFIC_HIGH; + else if (load_prec > IWL_MLD_TRAFFIC_LOAD_MEDIUM_THRESH) + mld->scan.traffic_load.status = IWL_MLD_TRAFFIC_MEDIUM; + else + mld->scan.traffic_load.status = IWL_MLD_TRAFFIC_LOW; + +out: + mld->scan.traffic_load.last_stats_ts_usec = curr_ts_usec; +} + +static void iwl_mld_update_link_sig(struct ieee80211_vif *vif, int sig, + struct ieee80211_bss_conf *bss_conf) +{ + struct iwl_mld *mld = iwl_mld_vif_from_mac80211(vif)->mld; + int exit_emlsr_thresh; + + if (sig == 0) { + IWL_DEBUG_RX(mld, "RSSI is 0 - skip signal based decision\n"); + return; + } + + /* TODO: task=statistics handle CQM notifications */ + + if (!iwl_mld_vif_has_emlsr_cap(vif)) + return; + + /* Handle inactive EMLSR, check whether to switch links */ + if (!iwl_mld_emlsr_active(vif)) { + if (sig < IWL_MLD_LOW_RSSI_MLO_SCAN_THRESH) + iwl_mld_trigger_link_selection(mld, vif); + return; + } + + /* We are in EMLSR, check if we need to exit */ + exit_emlsr_thresh = + iwl_mld_get_emlsr_rssi_thresh(mld, &bss_conf->chanreq.oper, + true); + + if (sig < exit_emlsr_thresh) + iwl_mld_exit_emlsr(mld, vif, IWL_MLD_EMLSR_EXIT_LOW_RSSI, + iwl_mld_get_other_link(vif, + bss_conf->link_id)); +} + +static void +iwl_mld_process_per_link_stats(struct iwl_mld *mld, + const struct iwl_stats_ntfy_per_link *per_link, + u32 curr_ts_usec) +{ + u32 total_airtime_usec = 0; + + for (u32 fw_id = 0; + fw_id < ARRAY_SIZE(mld->fw_id_to_bss_conf); + fw_id++) { + const struct iwl_stats_ntfy_per_link *link_stats; + struct ieee80211_bss_conf *bss_conf; + int sig; + + bss_conf = wiphy_dereference(mld->wiphy, + mld->fw_id_to_bss_conf[fw_id]); + if (!bss_conf || bss_conf->vif->type != NL80211_IFTYPE_STATION) + continue; + + link_stats = &per_link[fw_id]; + + total_airtime_usec += le32_to_cpu(link_stats->air_time); + + sig = -le32_to_cpu(link_stats->beacon_filter_average_energy); + iwl_mld_update_link_sig(bss_conf->vif, sig, bss_conf); + + /* TODO: parse more fields here (task=statistics)*/ + } + + iwl_mld_stats_recalc_traffic_load(mld, total_airtime_usec, + curr_ts_usec); +} + +static void +iwl_mld_process_per_sta_stats(struct iwl_mld *mld, + const struct iwl_stats_ntfy_per_sta *per_sta) +{ + for (int i = 0; i < mld->fw->ucode_capa.num_stations; i++) { + struct ieee80211_link_sta *link_sta = + wiphy_dereference(mld->wiphy, + mld->fw_id_to_link_sta[i]); + struct iwl_mld_link_sta *mld_link_sta; + s8 avg_energy = + -(s8)le32_to_cpu(per_sta[i].average_energy); + + if (IS_ERR_OR_NULL(link_sta) || !avg_energy) + continue; + + mld_link_sta = iwl_mld_link_sta_from_mac80211(link_sta); + if (WARN_ON(!mld_link_sta)) + continue; + + mld_link_sta->signal_avg = avg_energy; + } +} + +static void iwl_mld_fill_chanctx_stats(struct ieee80211_hw *hw, + struct ieee80211_chanctx_conf *ctx, + void *data) +{ + struct iwl_mld_phy *phy = iwl_mld_phy_from_mac80211(ctx); + const struct iwl_stats_ntfy_per_phy *per_phy = data; + + if (WARN_ON(phy->fw_id >= IWL_STATS_MAX_PHY_OPERATIONAL)) + return; + + phy->channel_load_by_us = + le32_to_cpu(per_phy[phy->fw_id].channel_load_by_us); + + /* TODO: channel load not by us (task=statistics) */ +} + +static void +iwl_mld_process_per_phy_stats(struct iwl_mld *mld, + const struct iwl_stats_ntfy_per_phy *per_phy) +{ + ieee80211_iter_chan_contexts_mtx(mld->hw, + iwl_mld_fill_chanctx_stats, + (void *)(uintptr_t)per_phy); + + /* channel_load_by_us may have been updated, so recheck */ + iwl_mld_emlsr_check_chan_load(mld); +} + +void iwl_mld_handle_stats_oper_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + const struct iwl_system_statistics_notif_oper *stats = + (void *)&pkt->data; + u32 curr_ts_usec = le32_to_cpu(stats->time_stamp); + + BUILD_BUG_ON(ARRAY_SIZE(stats->per_sta) != IWL_STATION_COUNT_MAX); + BUILD_BUG_ON(ARRAY_SIZE(stats->per_link) < + ARRAY_SIZE(mld->fw_id_to_bss_conf)); + + iwl_mld_process_per_link_stats(mld, stats->per_link, curr_ts_usec); + iwl_mld_process_per_sta_stats(mld, stats->per_sta); + iwl_mld_process_per_phy_stats(mld, stats->per_phy); + + iwl_mld_check_omi_bw_reduction(mld); +} + +void iwl_mld_handle_stats_oper_part1_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + /* TODO */ +} + diff --git a/drivers/net/wireless/intel/iwlwifi/mld/stats.h b/drivers/net/wireless/intel/iwlwifi/mld/stats.h new file mode 100644 index 0000000000000..de434e66d5555 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/stats.h @@ -0,0 +1,22 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024 Intel Corporation + */ +#ifndef __iwl_mld_stats_h__ +#define __iwl_mld_stats_h__ + +int iwl_mld_request_periodic_fw_stats(struct iwl_mld *mld, bool enable); + +void iwl_mld_mac80211_sta_statistics(struct ieee80211_hw *hw, + struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + struct station_info *sinfo); + +void iwl_mld_handle_stats_oper_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt); +void iwl_mld_handle_stats_oper_part1_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt); + +int iwl_mld_clear_stats_in_fw(struct iwl_mld *mld); + +#endif /* __iwl_mld_stats_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/tests/Makefile b/drivers/net/wireless/intel/iwlwifi/mld/tests/Makefile new file mode 100644 index 0000000000000..36317feb923ba --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/tests/Makefile @@ -0,0 +1,5 @@ +# SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +iwlmld-tests-y += module.o hcmd.o utils.o link.o rx.o agg.o link-selection.o + +ccflags-y += -I$(src)/../ +obj-$(CONFIG_IWLWIFI_KUNIT_TESTS) += iwlmld-tests.o diff --git a/drivers/net/wireless/intel/iwlwifi/mld/tests/agg.c b/drivers/net/wireless/intel/iwlwifi/mld/tests/agg.c new file mode 100644 index 0000000000000..1fd664be1a7c2 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/tests/agg.c @@ -0,0 +1,663 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * KUnit tests for channel helper functions + * + * Copyright (C) 2024 Intel Corporation + */ +#include +#include +#include + +#include "utils.h" +#include "mld.h" +#include "sta.h" +#include "agg.h" +#include "rx.h" + +#define FC_QOS_DATA (IEEE80211_FTYPE_DATA | IEEE80211_STYPE_QOS_DATA) +#define BA_WINDOW_SIZE 64 +#define QUEUE 0 + +static const struct reorder_buffer_case { + const char *desc; + struct { + /* ieee80211_hdr fields */ + u16 fc; + u8 tid; + bool multicast; + /* iwl_rx_mpdu_desc fields */ + u16 nssn; + /* used also for setting hdr::seq_ctrl */ + u16 sn; + u8 baid; + bool amsdu; + bool last_subframe; + bool old_sn; + bool dup; + } rx_pkt; + struct { + bool valid; + u16 head_sn; + u8 baid; + u16 num_entries; + /* The test prepares the reorder buffer with fake skbs based + * on the sequence numbers provided in @entries array. + */ + struct { + u16 sn; + /* Set add_subframes > 0 to simulate an A-MSDU by + * queueing additional @add_subframes skbs in the + * appropriate reorder buffer entry (based on the @sn) + */ + u8 add_subframes; + } entries[BA_WINDOW_SIZE]; + } reorder_buf_state; + struct { + enum iwl_mld_reorder_result reorder_res; + u16 head_sn; + u16 num_stored; + u16 skb_release_order[BA_WINDOW_SIZE]; + u16 skb_release_order_count; + } expected; +} reorder_buffer_cases[] = { + { + .desc = "RX packet with invalid BAID", + .rx_pkt = { + .fc = FC_QOS_DATA, + .baid = IWL_RX_REORDER_DATA_INVALID_BAID, + }, + .reorder_buf_state = { + .valid = true, + }, + .expected = { + /* Invalid BAID should not be buffered. The frame is + * passed to the network stack immediately. + */ + .reorder_res = IWL_MLD_PASS_SKB, + .num_stored = 0, + }, + }, + { + .desc = "RX Multicast packet", + .rx_pkt = { + .fc = FC_QOS_DATA, + .multicast = true, + }, + .reorder_buf_state = { + .valid = true, + }, + .expected = { + /* Multicast packets are not buffered. The packet is + * passed to the network stack immediately. + */ + .reorder_res = IWL_MLD_PASS_SKB, + .num_stored = 0, + }, + }, + { + .desc = "RX non-QoS data", + .rx_pkt = { + .fc = IEEE80211_FTYPE_DATA, + }, + .reorder_buf_state = { + .valid = true, + }, + .expected = { + /* non-QoS data frames do not require reordering. + * The packet is passed to the network stack + * immediately. + */ + .reorder_res = IWL_MLD_PASS_SKB, + }, + }, + { + .desc = "RX old SN packet, reorder buffer is not yet valid", + .rx_pkt = { + .fc = FC_QOS_DATA, + .old_sn = true, + }, + .reorder_buf_state = { + .valid = false, + }, + .expected = { + /* The buffer is invalid and the RX packet has an old + * SN. The packet is passed to the network stack + * immediately. + */ + .reorder_res = IWL_MLD_PASS_SKB, + }, + }, + { + .desc = "RX old SN packet, reorder buffer valid", + .rx_pkt = { + .fc = FC_QOS_DATA, + .old_sn = true, + }, + .reorder_buf_state = { + .valid = true, + .head_sn = 100, + }, + .expected = { + /* The buffer is valid and the RX packet has an old SN. + * The packet should be dropped. + */ + .reorder_res = IWL_MLD_DROP_SKB, + .num_stored = 0, + .head_sn = 100, + }, + }, + { + .desc = "RX duplicate packet", + .rx_pkt = { + .fc = FC_QOS_DATA, + .dup = true, + }, + .reorder_buf_state = { + .valid = true, + .head_sn = 100, + }, + .expected = { + /* Duplicate packets should be dropped */ + .reorder_res = IWL_MLD_DROP_SKB, + .num_stored = 0, + .head_sn = 100, + }, + }, + { + .desc = "RX In-order packet, sn < nssn", + .rx_pkt = { + .fc = FC_QOS_DATA, + .sn = 100, + .nssn = 101, + }, + .reorder_buf_state = { + .valid = true, + .head_sn = 100, + }, + .expected = { + /* 1. Reorder buffer is empty. + * 2. RX packet SN is in order and less than NSSN. + * Packet is released to the network stack immediately + * and buffer->head_sn is updated to NSSN. + */ + .reorder_res = IWL_MLD_PASS_SKB, + .num_stored = 0, + .head_sn = 101, + }, + }, + { + .desc = "RX In-order packet, sn == head_sn", + .rx_pkt = { + .fc = FC_QOS_DATA, + .sn = 101, + .nssn = 100, + }, + .reorder_buf_state = { + .valid = true, + .head_sn = 101, + }, + .expected = { + /* 1. Reorder buffer is empty. + * 2. RX packet SN is equal to buffer->head_sn. + * Packet is released to the network stack immediately + * and buffer->head_sn is incremented. + */ + .reorder_res = IWL_MLD_PASS_SKB, + .num_stored = 0, + .head_sn = 102, + }, + }, + { + .desc = "RX In-order packet, IEEE80211_MAX_SN wrap around", + .rx_pkt = { + .fc = FC_QOS_DATA, + .sn = IEEE80211_MAX_SN, + .nssn = IEEE80211_MAX_SN - 1, + }, + .reorder_buf_state = { + .valid = true, + .head_sn = IEEE80211_MAX_SN, + }, + .expected = { + /* 1. Reorder buffer is empty. + * 2. RX SN == buffer->head_sn == IEEE80211_MAX_SN + * Packet is released to the network stack immediately + * and buffer->head_sn is incremented correctly (wraps + * around to 0). + */ + .reorder_res = IWL_MLD_PASS_SKB, + .num_stored = 0, + .head_sn = 0, + }, + }, + { + .desc = "RX Out-of-order packet, pending packet in buffer", + .rx_pkt = { + .fc = FC_QOS_DATA, + .sn = 100, + .nssn = 102, + }, + .reorder_buf_state = { + .valid = true, + .head_sn = 100, + .num_entries = 1, + .entries[0].sn = 101, + }, + .expected = { + /* 1. Reorder buffer contains one packet with SN=101. + * 2. RX packet SN = buffer->head_sn. + * Both packets are released (in order) to the network + * stack as there are no gaps. + */ + .reorder_res = IWL_MLD_BUFFERED_SKB, + .num_stored = 0, + .head_sn = 102, + .skb_release_order = {100, 101}, + .skb_release_order_count = 2, + }, + }, + { + .desc = "RX Out-of-order packet, pending packet in buffer (wrap around)", + .rx_pkt = { + .fc = FC_QOS_DATA, + .sn = 0, + .nssn = 1, + }, + .reorder_buf_state = { + .valid = true, + .head_sn = IEEE80211_MAX_SN - 1, + .num_entries = 1, + .entries[0].sn = IEEE80211_MAX_SN, + }, + .expected = { + /* 1. Reorder buffer contains one packet with + * SN=IEEE80211_MAX_SN. + * 2. RX Packet SN = 0 (after wrap around) + * Both packets are released (in order) to the network + * stack as there are no gaps. + */ + .reorder_res = IWL_MLD_BUFFERED_SKB, + .num_stored = 0, + .head_sn = 1, + .skb_release_order = { 4095, 0 }, + .skb_release_order_count = 2, + }, + }, + { + .desc = "RX Out-of-order packet, filling 1/2 holes in buffer, release RX packet", + .rx_pkt = { + .fc = FC_QOS_DATA, + .sn = 100, + .nssn = 101, + }, + .reorder_buf_state = { + .valid = true, + .head_sn = 100, + .num_entries = 1, + .entries[0].sn = 102, + }, + .expected = { + /* 1. Reorder buffer contains one packet with SN=102. + * 2. There are 2 holes at SN={100, 101}. + * Only the RX packet (SN=100) is released, there is + * still a hole at 101. + */ + .reorder_res = IWL_MLD_BUFFERED_SKB, + .num_stored = 1, + .head_sn = 101, + .skb_release_order = {100}, + .skb_release_order_count = 1, + }, + }, + { + .desc = "RX Out-of-order packet, filling 1/2 holes, release 2 packets", + .rx_pkt = { + .fc = FC_QOS_DATA, + .sn = 102, + .nssn = 103, + }, + .reorder_buf_state = { + .valid = true, + .head_sn = 100, + .num_entries = 3, + .entries[0].sn = 101, + .entries[1].sn = 104, + .entries[2].sn = 105, + }, + .expected = { + /* 1. Reorder buffer contains three packets. + * 2. RX packet fills one of two holes (at SN=102). + * Two packets are released (until the next hole at + * SN=103). + */ + .reorder_res = IWL_MLD_BUFFERED_SKB, + .num_stored = 2, + .head_sn = 103, + .skb_release_order = {101, 102}, + .skb_release_order_count = 2, + }, + }, + { + .desc = "RX Out-of-order packet, filling 1/1 holes, no packets released", + .rx_pkt = { + .fc = FC_QOS_DATA, + .sn = 102, + .nssn = 100, + }, + .reorder_buf_state = { + .valid = true, + .head_sn = 100, + .num_entries = 3, + .entries[0].sn = 101, + .entries[1].sn = 103, + .entries[2].sn = 104, + }, + .expected = { + /* 1. Reorder buffer contains three packets: + * SN={101, 103, 104}. + * 2. RX packet fills a hole (SN=102), but NSSN is + * smaller than buffered frames. + * No packets can be released yet and buffer->head_sn + * is not updated. + */ + .reorder_res = IWL_MLD_BUFFERED_SKB, + .num_stored = 4, + .head_sn = 100, + }, + }, + { + .desc = "RX In-order A-MSDU, last subframe", + .rx_pkt = { + .fc = FC_QOS_DATA, + .sn = 100, + .nssn = 101, + .amsdu = true, + .last_subframe = true, + }, + .reorder_buf_state = { + .valid = true, + .head_sn = 100, + .num_entries = 1, + .entries[0] = { + .sn = 100, + .add_subframes = 1, + }, + }, + .expected = { + /* 1. Reorder buffer contains a 2-sub frames A-MSDU + * at SN=100. + * 2. RX packet is the last SN=100 A-MSDU subframe + * All packets are released in order (3 x SN=100). + */ + .reorder_res = IWL_MLD_BUFFERED_SKB, + .num_stored = 0, + .head_sn = 101, + .skb_release_order = {100, 100, 100}, + .skb_release_order_count = 3, + }, + }, + { + .desc = "RX In-order A-MSDU, not the last subframe", + .rx_pkt = { + .fc = FC_QOS_DATA, + .sn = 100, + .nssn = 101, + .amsdu = true, + .last_subframe = false, + }, + .reorder_buf_state = { + .valid = true, + .head_sn = 100, + .num_entries = 1, + .entries[0] = { + .sn = 100, + .add_subframes = 1, + }, + }, + .expected = { + /* 1. Reorder buffer contains a 2-sub frames A-MSDU + * at SN=100. + * 2. RX packet additional SN=100 A-MSDU subframe, + * but not the last one + * No packets are released and head_sn is not updated. + */ + .reorder_res = IWL_MLD_BUFFERED_SKB, + .num_stored = 3, + .head_sn = 100, + }, + }, +}; + +KUNIT_ARRAY_PARAM_DESC(test_reorder_buffer, reorder_buffer_cases, desc); + +static struct sk_buff_head g_released_skbs; +static u16 g_num_released_skbs; + +/* Add released skbs from reorder buffer to a global list; This allows us + * to verify the correct release order of packets after they pass through the + * simulated reorder logic. + */ +static void +fake_iwl_mld_pass_packet_to_mac80211(struct iwl_mld *mld, + struct napi_struct *napi, + struct sk_buff *skb, int queue, + struct ieee80211_sta *sta) +{ + __skb_queue_tail(&g_released_skbs, skb); + g_num_released_skbs++; +} + +static u32 +fake_iwl_mld_fw_sta_id_mask(struct iwl_mld *mld, struct ieee80211_sta *sta) +{ + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(sta); + struct iwl_mld_link_sta *mld_link_sta; + u8 link_id; + u32 sta_mask = 0; + + /* This is the expectation in the real function */ + lockdep_assert_wiphy(mld->wiphy); + + /* We can't use for_each_sta_active_link */ + for_each_mld_link_sta(mld_sta, mld_link_sta, link_id) + sta_mask |= BIT(mld_link_sta->fw_id); + return sta_mask; +} + +static struct iwl_rx_mpdu_desc *setup_mpdu_desc(void) +{ + struct kunit *test = kunit_get_current_test(); + const struct reorder_buffer_case *param = + (const void *)(test->param_value); + struct iwl_rx_mpdu_desc *mpdu_desc; + + KUNIT_ALLOC_AND_ASSERT(test, mpdu_desc); + + mpdu_desc->reorder_data |= + cpu_to_le32(FIELD_PREP(IWL_RX_MPDU_REORDER_BAID_MASK, + param->rx_pkt.baid)); + mpdu_desc->reorder_data |= + cpu_to_le32(FIELD_PREP(IWL_RX_MPDU_REORDER_SN_MASK, + param->rx_pkt.sn)); + mpdu_desc->reorder_data |= + cpu_to_le32(FIELD_PREP(IWL_RX_MPDU_REORDER_NSSN_MASK, + param->rx_pkt.nssn)); + if (param->rx_pkt.old_sn) + mpdu_desc->reorder_data |= + cpu_to_le32(IWL_RX_MPDU_REORDER_BA_OLD_SN); + + if (param->rx_pkt.dup) + mpdu_desc->status |= cpu_to_le32(IWL_RX_MPDU_STATUS_DUPLICATE); + + if (param->rx_pkt.amsdu) { + mpdu_desc->mac_flags2 |= IWL_RX_MPDU_MFLG2_AMSDU; + if (param->rx_pkt.last_subframe) + mpdu_desc->amsdu_info |= + IWL_RX_MPDU_AMSDU_LAST_SUBFRAME; + } + + return mpdu_desc; +} + +static struct sk_buff *alloc_and_setup_skb(u16 fc, u16 seq_ctrl, u8 tid, + bool mcast) +{ + struct kunit *test = kunit_get_current_test(); + struct ieee80211_hdr hdr = { + .frame_control = cpu_to_le16(fc), + .seq_ctrl = cpu_to_le16(seq_ctrl), + }; + struct sk_buff *skb; + + skb = kunit_zalloc_skb(test, 128, GFP_KERNEL); + KUNIT_ASSERT_NOT_NULL(test, skb); + + if (ieee80211_is_data_qos(hdr.frame_control)) { + u8 *qc = ieee80211_get_qos_ctl(&hdr); + + qc[0] = tid & IEEE80211_QOS_CTL_TID_MASK; + } + + if (mcast) + hdr.addr1[0] = 0x1; + + skb_set_mac_header(skb, skb->len); + skb_put_data(skb, &hdr, ieee80211_hdrlen(hdr.frame_control)); + + return skb; +} + +static struct iwl_mld_reorder_buffer * +setup_reorder_buffer(struct iwl_mld_baid_data *baid_data) +{ + struct kunit *test = kunit_get_current_test(); + const struct reorder_buffer_case *param = + (const void *)(test->param_value); + struct iwl_mld_reorder_buffer *buffer = baid_data->reorder_buf; + struct iwl_mld_reorder_buf_entry *entries = baid_data->entries; + struct sk_buff *fake_skb; + + buffer->valid = param->reorder_buf_state.valid; + buffer->head_sn = param->reorder_buf_state.head_sn; + buffer->queue = QUEUE; + + for (int i = 0; i < baid_data->buf_size; i++) + __skb_queue_head_init(&entries[i].frames); + + for (int i = 0; i < param->reorder_buf_state.num_entries; i++) { + u16 sn = param->reorder_buf_state.entries[i].sn; + int index = sn % baid_data->buf_size; + u8 add_subframes = + param->reorder_buf_state.entries[i].add_subframes; + /* create 1 skb per entry + additional skbs per num of + * requested subframes + */ + u8 num_skbs = 1 + add_subframes; + + for (int j = 0; j < num_skbs; j++) { + fake_skb = alloc_and_setup_skb(FC_QOS_DATA, sn, 0, + false); + __skb_queue_tail(&entries[index].frames, fake_skb); + buffer->num_stored++; + } + } + + return buffer; +} + +static struct iwl_mld_reorder_buffer *setup_ba_data(struct ieee80211_sta *sta) +{ + struct kunit *test = kunit_get_current_test(); + struct iwl_mld *mld = test->priv; + const struct reorder_buffer_case *param = + (const void *)(test->param_value); + struct iwl_mld_baid_data *baid_data = NULL; + struct iwl_mld_reorder_buffer *buffer; + u32 reorder_buf_size = BA_WINDOW_SIZE * sizeof(baid_data->entries[0]); + u8 baid = param->reorder_buf_state.baid; + + /* Assuming only 1 RXQ */ + KUNIT_ALLOC_AND_ASSERT_SIZE(test, baid_data, + sizeof(*baid_data) + reorder_buf_size); + + baid_data->baid = baid; + baid_data->tid = param->rx_pkt.tid; + baid_data->buf_size = BA_WINDOW_SIZE; + + wiphy_lock(mld->wiphy); + baid_data->sta_mask = iwl_mld_fw_sta_id_mask(mld, sta); + wiphy_unlock(mld->wiphy); + + baid_data->entries_per_queue = BA_WINDOW_SIZE; + + buffer = setup_reorder_buffer(baid_data); + + KUNIT_EXPECT_NULL(test, rcu_access_pointer(mld->fw_id_to_ba[baid])); + rcu_assign_pointer(mld->fw_id_to_ba[baid], baid_data); + + return buffer; +} + +static void test_reorder_buffer(struct kunit *test) +{ + struct iwl_mld *mld = test->priv; + const struct reorder_buffer_case *param = + (const void *)(test->param_value); + struct iwl_rx_mpdu_desc *mpdu_desc; + struct ieee80211_vif *vif; + struct ieee80211_sta *sta; + struct sk_buff *skb; + struct iwl_mld_reorder_buffer *buffer; + enum iwl_mld_reorder_result reorder_res; + u16 skb_release_order_count = param->expected.skb_release_order_count; + u16 skb_idx = 0; + + /* Init globals and activate stubs */ + __skb_queue_head_init(&g_released_skbs); + g_num_released_skbs = 0; + kunit_activate_static_stub(test, iwl_mld_fw_sta_id_mask, + fake_iwl_mld_fw_sta_id_mask); + kunit_activate_static_stub(test, iwl_mld_pass_packet_to_mac80211, + fake_iwl_mld_pass_packet_to_mac80211); + + vif = iwlmld_kunit_add_vif(false, NL80211_IFTYPE_STATION); + sta = iwlmld_kunit_setup_sta(vif, IEEE80211_STA_AUTHORIZED, -1); + + /* Prepare skb, mpdu_desc, BA data and the reorder buffer */ + skb = alloc_and_setup_skb(param->rx_pkt.fc, param->rx_pkt.sn, + param->rx_pkt.tid, param->rx_pkt.multicast); + buffer = setup_ba_data(sta); + mpdu_desc = setup_mpdu_desc(); + + rcu_read_lock(); + reorder_res = iwl_mld_reorder(mld, NULL, QUEUE, sta, skb, mpdu_desc); + rcu_read_unlock(); + + KUNIT_ASSERT_EQ(test, reorder_res, param->expected.reorder_res); + KUNIT_ASSERT_EQ(test, buffer->num_stored, param->expected.num_stored); + KUNIT_ASSERT_EQ(test, buffer->head_sn, param->expected.head_sn); + + /* Verify skbs release order */ + KUNIT_ASSERT_EQ(test, skb_release_order_count, g_num_released_skbs); + while ((skb = __skb_dequeue(&g_released_skbs))) { + struct ieee80211_hdr *hdr = (void *)skb_mac_header(skb); + + KUNIT_ASSERT_EQ(test, le16_to_cpu(hdr->seq_ctrl), + param->expected.skb_release_order[skb_idx]); + skb_idx++; + } + KUNIT_ASSERT_EQ(test, skb_idx, skb_release_order_count); +} + +static struct kunit_case reorder_buffer_test_cases[] = { + KUNIT_CASE_PARAM(test_reorder_buffer, test_reorder_buffer_gen_params), + {}, +}; + +static struct kunit_suite reorder_buffer = { + .name = "iwlmld-reorder-buffer", + .test_cases = reorder_buffer_test_cases, + .init = iwlmld_kunit_test_init, +}; + +kunit_test_suite(reorder_buffer); diff --git a/drivers/net/wireless/intel/iwlwifi/mld/tests/hcmd.c b/drivers/net/wireless/intel/iwlwifi/mld/tests/hcmd.c new file mode 100644 index 0000000000000..4e189bf8b3fb4 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/tests/hcmd.c @@ -0,0 +1,62 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * KUnit tests for channel helper functions + * + * Copyright (C) 2024 Intel Corporation + */ +#include + +#include +#include "mld.h" + +MODULE_IMPORT_NS("EXPORTED_FOR_KUNIT_TESTING"); + +static void test_hcmd_names_sorted(struct kunit *test) +{ + int i; + + for (i = 0; i < global_iwl_mld_goups_size; i++) { + const struct iwl_hcmd_arr *arr = &iwl_mld_groups[i]; + int j; + + if (!arr->arr) + continue; + for (j = 0; j < arr->size - 1; j++) + KUNIT_EXPECT_LE(test, arr->arr[j].cmd_id, + arr->arr[j + 1].cmd_id); + } +} + +static void test_hcmd_names_for_rx(struct kunit *test) +{ + static struct iwl_trans t = { + .command_groups = iwl_mld_groups, + }; + + t.command_groups_size = global_iwl_mld_goups_size; + + for (unsigned int i = 0; i < iwl_mld_rx_handlers_num; i++) { + const struct iwl_rx_handler *rxh; + const char *name; + + rxh = &iwl_mld_rx_handlers[i]; + + name = iwl_get_cmd_string(&t, rxh->cmd_id); + KUNIT_EXPECT_NOT_NULL(test, name); + KUNIT_EXPECT_NE_MSG(test, strcmp(name, "UNKNOWN"), 0, + "ID 0x%04x is UNKNOWN", rxh->cmd_id); + } +} + +static struct kunit_case hcmd_names_cases[] = { + KUNIT_CASE(test_hcmd_names_sorted), + KUNIT_CASE(test_hcmd_names_for_rx), + {}, +}; + +static struct kunit_suite hcmd_names = { + .name = "iwlmld-hcmd-names", + .test_cases = hcmd_names_cases, +}; + +kunit_test_suite(hcmd_names); diff --git a/drivers/net/wireless/intel/iwlwifi/mld/tests/link-selection.c b/drivers/net/wireless/intel/iwlwifi/mld/tests/link-selection.c new file mode 100644 index 0000000000000..d835550c1a6bd --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/tests/link-selection.c @@ -0,0 +1,171 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * KUnit tests for link selection functions + * + * Copyright (C) 2025 Intel Corporation + */ +#include + +#include "utils.h" +#include "mld.h" +#include "link.h" +#include "iface.h" +#include "phy.h" + +static const struct link_grading_test_case { + const char *desc; + struct { + struct { + u8 link_id; + const struct cfg80211_chan_def *chandef; + bool active; + s32 signal; + bool has_chan_util_elem; + u8 chan_util; /* 0-255 , used only if has_chan_util_elem is true */ + u8 chan_load_by_us; /* 0-100, used only if active is true */; + } link; + } input; + unsigned int expected_grade; +} link_grading_cases[] = { + { + .desc = "channel util of 128 (50%)", + .input.link = { + .link_id = 0, + .chandef = &chandef_2ghz, + .active = false, + .has_chan_util_elem = true, + .chan_util = 128, + }, + .expected_grade = 86, + }, + { + .desc = "channel util of 180 (70%)", + .input.link = { + .link_id = 0, + .chandef = &chandef_2ghz, + .active = false, + .has_chan_util_elem = true, + .chan_util = 180, + }, + .expected_grade = 51, + }, + { + .desc = "channel util of 180 (70%), channel load by us of 10%", + .input.link = { + .link_id = 0, + .chandef = &chandef_2ghz, + .has_chan_util_elem = true, + .chan_util = 180, + .active = true, + .chan_load_by_us = 10, + }, + .expected_grade = 67, + }, + { + .desc = "no channel util element", + .input.link = { + .link_id = 0, + .chandef = &chandef_2ghz, + .active = true, + }, + .expected_grade = 120, + }, +}; + +KUNIT_ARRAY_PARAM_DESC(link_grading, link_grading_cases, desc); + +static void setup_link(struct ieee80211_bss_conf *link) +{ + struct kunit *test = kunit_get_current_test(); + struct iwl_mld *mld = test->priv; + const struct link_grading_test_case *test_param = + (const void *)(test->param_value); + + KUNIT_ALLOC_AND_ASSERT(test, link->bss); + + link->bss->signal = DBM_TO_MBM(test_param->input.link.signal); + + link->chanreq.oper = *test_param->input.link.chandef; + + if (test_param->input.link.has_chan_util_elem) { + struct cfg80211_bss_ies *ies; + struct ieee80211_bss_load_elem bss_load = { + .channel_util = test_param->input.link.chan_util, + }; + struct element *elem = + iwlmld_kunit_gen_element(WLAN_EID_QBSS_LOAD, + &bss_load, + sizeof(bss_load)); + unsigned int elem_len = sizeof(*elem) + sizeof(bss_load); + + KUNIT_ALLOC_AND_ASSERT_SIZE(test, ies, sizeof(*ies) + elem_len); + memcpy(ies->data, elem, elem_len); + ies->len = elem_len; + rcu_assign_pointer(link->bss->beacon_ies, ies); + rcu_assign_pointer(link->bss->ies, ies); + } + + if (test_param->input.link.active) { + struct ieee80211_chanctx_conf *chan_ctx = + wiphy_dereference(mld->wiphy, link->chanctx_conf); + struct iwl_mld_phy *phy; + + KUNIT_ASSERT_NOT_NULL(test, chan_ctx); + + phy = iwl_mld_phy_from_mac80211(chan_ctx); + + phy->channel_load_by_us = test_param->input.link.chan_load_by_us; + } +} + +static void test_link_grading(struct kunit *test) +{ + struct iwl_mld *mld = test->priv; + const struct link_grading_test_case *test_param = + (const void *)(test->param_value); + struct ieee80211_vif *vif; + struct ieee80211_bss_conf *link; + unsigned int actual_grade; + u8 assoc_link_id; + /* Extract test case parameters */ + u8 link_id = test_param->input.link.link_id; + enum nl80211_band band = test_param->input.link.chandef->chan->band; + bool active = test_param->input.link.active; + u16 valid_links; + + /* If the link is not active, use a different link as the assoc link */ + if (active) { + assoc_link_id = link_id; + valid_links = BIT(link_id); + } else { + assoc_link_id = BIT(ffz(BIT(link_id))); + valid_links = BIT(assoc_link_id) | BIT(link_id); + } + + vif = iwlmld_kunit_setup_mlo_assoc(valid_links, assoc_link_id, band); + + wiphy_lock(mld->wiphy); + link = wiphy_dereference(mld->wiphy, vif->link_conf[link_id]); + KUNIT_ASSERT_NOT_NULL(test, link); + + setup_link(link); + + actual_grade = iwl_mld_get_link_grade(mld, link); + wiphy_unlock(mld->wiphy); + + /* Assert that the returned grade matches the expected grade */ + KUNIT_EXPECT_EQ(test, actual_grade, test_param->expected_grade); +} + +static struct kunit_case link_selection_cases[] = { + KUNIT_CASE_PARAM(test_link_grading, link_grading_gen_params), + {}, +}; + +static struct kunit_suite link_selection = { + .name = "iwlmld-link-selection-tests", + .test_cases = link_selection_cases, + .init = iwlmld_kunit_test_init, +}; + +kunit_test_suite(link_selection); diff --git a/drivers/net/wireless/intel/iwlwifi/mld/tests/link.c b/drivers/net/wireless/intel/iwlwifi/mld/tests/link.c new file mode 100644 index 0000000000000..6e251dbc1dfe1 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/tests/link.c @@ -0,0 +1,103 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * KUnit tests for channel helper functions + * + * Copyright (C) 2024 Intel Corporation + */ +#include + +#include "utils.h" +#include "mld.h" +#include "link.h" +#include "iface.h" +#include "fw/api/mac-cfg.h" + +static const struct missed_beacon_test_case { + const char *desc; + struct { + struct iwl_missed_beacons_notif notif; + bool emlsr; + } input; + struct { + bool disconnected; + bool emlsr; + } output; +} missed_beacon_cases[] = { + { + .desc = "no EMLSR, no disconnect", + .input.notif = { + .consec_missed_beacons = cpu_to_le32(4), + }, + }, + { + .desc = "no EMLSR, no beacon loss since Rx, no disconnect", + .input.notif = { + .consec_missed_beacons = cpu_to_le32(20), + }, + }, + { + .desc = "no EMLSR, beacon loss since Rx, disconnect", + .input.notif = { + .consec_missed_beacons = cpu_to_le32(20), + .consec_missed_beacons_since_last_rx = + cpu_to_le32(10), + }, + .output.disconnected = true, + }, +}; + +KUNIT_ARRAY_PARAM_DESC(test_missed_beacon, missed_beacon_cases, desc); + +static void fake_ieee80211_connection_loss(struct ieee80211_vif *vif) +{ + vif->cfg.assoc = false; +} + +static void test_missed_beacon(struct kunit *test) +{ + struct iwl_mld *mld = test->priv; + struct iwl_missed_beacons_notif *notif; + const struct missed_beacon_test_case *test_param = + (const void *)(test->param_value); + struct ieee80211_vif *vif; + struct iwl_rx_packet *pkt; + + kunit_activate_static_stub(test, ieee80211_connection_loss, + fake_ieee80211_connection_loss); + pkt = iwl_mld_kunit_create_pkt(test_param->input.notif); + notif = (void *)pkt->data; + + if (test_param->input.emlsr) { + vif = iwlmld_kunit_assoc_emlsr(0x3, NL80211_BAND_5GHZ, + NL80211_BAND_6GHZ); + } else { + struct iwl_mld_vif *mld_vif; + + vif = iwlmld_kunit_setup_non_mlo_assoc(NL80211_BAND_6GHZ); + mld_vif = iwl_mld_vif_from_mac80211(vif); + notif->link_id = cpu_to_le32(mld_vif->deflink.fw_id); + } + + wiphy_lock(mld->wiphy); + + iwl_mld_handle_missed_beacon_notif(mld, pkt); + + wiphy_unlock(mld->wiphy); + + KUNIT_ASSERT_NE(test, vif->cfg.assoc, test_param->output.disconnected); + + /* TODO: add test cases for esr and check */ +} + +static struct kunit_case link_cases[] = { + KUNIT_CASE_PARAM(test_missed_beacon, test_missed_beacon_gen_params), + {}, +}; + +static struct kunit_suite link = { + .name = "iwlmld-link", + .test_cases = link_cases, + .init = iwlmld_kunit_test_init, +}; + +kunit_test_suite(link); diff --git a/drivers/net/wireless/intel/iwlwifi/mld/tests/module.c b/drivers/net/wireless/intel/iwlwifi/mld/tests/module.c new file mode 100644 index 0000000000000..5d9818587b238 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/tests/module.c @@ -0,0 +1,11 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * This is just module boilerplate for the iwlmld kunit module. + * + * Copyright (C) 2024 Intel Corporation + */ +#include + +MODULE_IMPORT_NS("IWLWIFI"); +MODULE_LICENSE("GPL"); +MODULE_DESCRIPTION("kunit tests for iwlmld"); diff --git a/drivers/net/wireless/intel/iwlwifi/mld/tests/rx.c b/drivers/net/wireless/intel/iwlwifi/mld/tests/rx.c new file mode 100644 index 0000000000000..20cb4e03ab413 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/tests/rx.c @@ -0,0 +1,353 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * KUnit tests for channel helper functions + * + * Copyright (C) 2024 Intel Corporation + */ +#include +#include "utils.h" +#include "iwl-trans.h" +#include "mld.h" +#include "sta.h" + +static const struct is_dup_case { + const char *desc; + struct { + /* ieee80211_hdr fields */ + __le16 fc; + __le16 seq; + u8 tid; + bool multicast; + /* iwl_rx_mpdu_desc fields */ + bool is_amsdu; + u8 sub_frame_idx; + } rx_pkt; + struct { + __le16 last_seq; + u8 last_sub_frame_idx; + u8 tid; + } dup_data_state; + struct { + bool is_dup; + u32 rx_status_flag; + } result; +} is_dup_cases[] = { + { + .desc = "Control frame", + .rx_pkt = { + .fc = __cpu_to_le16(IEEE80211_FTYPE_CTL), + }, + .result = { + .is_dup = false, + .rx_status_flag = 0, + } + }, + { + .desc = "Null func frame", + .rx_pkt = { + .fc = __cpu_to_le16(IEEE80211_FTYPE_DATA | + IEEE80211_STYPE_NULLFUNC), + }, + .result = { + .is_dup = false, + .rx_status_flag = 0, + } + }, + { + .desc = "Multicast data", + .rx_pkt = { + .fc = __cpu_to_le16(IEEE80211_FTYPE_DATA), + .multicast = true, + }, + .result = { + .is_dup = false, + .rx_status_flag = 0, + } + }, + { + .desc = "QoS null func frame", + .rx_pkt = { + .fc = __cpu_to_le16(IEEE80211_FTYPE_DATA | + IEEE80211_STYPE_QOS_NULLFUNC), + }, + .result = { + .is_dup = false, + .rx_status_flag = 0, + } + }, + { + .desc = "QoS data new sequence", + .rx_pkt = { + .fc = __cpu_to_le16(IEEE80211_FTYPE_DATA | + IEEE80211_STYPE_QOS_DATA), + .seq = __cpu_to_le16(0x101), + }, + .dup_data_state = { + .last_seq = __cpu_to_le16(0x100), + .last_sub_frame_idx = 0, + }, + .result = { + .is_dup = false, + .rx_status_flag = RX_FLAG_DUP_VALIDATED, + }, + }, + { + .desc = "QoS data same sequence, no retry", + .rx_pkt = { + .fc = __cpu_to_le16(IEEE80211_FTYPE_DATA | + IEEE80211_STYPE_QOS_DATA), + .seq = __cpu_to_le16(0x100), + }, + .dup_data_state = { + .last_seq = __cpu_to_le16(0x100), + .last_sub_frame_idx = 0, + }, + .result = { + .is_dup = false, + .rx_status_flag = RX_FLAG_DUP_VALIDATED, + }, + }, + { + .desc = "QoS data same sequence, has retry", + .rx_pkt = { + .fc = __cpu_to_le16(IEEE80211_FTYPE_DATA | + IEEE80211_STYPE_QOS_DATA | + IEEE80211_FCTL_RETRY), + .seq = __cpu_to_le16(0x100), + }, + .dup_data_state = { + .last_seq = __cpu_to_le16(0x100), + .last_sub_frame_idx = 0, + }, + .result = { + .is_dup = true, + .rx_status_flag = 0, + }, + }, + { + .desc = "QoS data invalid tid", + .rx_pkt = { + .fc = __cpu_to_le16(IEEE80211_FTYPE_DATA | + IEEE80211_STYPE_QOS_DATA), + .seq = __cpu_to_le16(0x100), + .tid = IWL_MAX_TID_COUNT + 1, + }, + .result = { + .is_dup = true, + .rx_status_flag = 0, + }, + }, + { + .desc = "non-QoS data, same sequence, same tid, no retry", + .rx_pkt = { + /* Driver will use tid = IWL_MAX_TID_COUNT */ + .fc = __cpu_to_le16(IEEE80211_FTYPE_DATA), + .seq = __cpu_to_le16(0x100), + }, + .dup_data_state = { + .tid = IWL_MAX_TID_COUNT, + .last_seq = __cpu_to_le16(0x100), + .last_sub_frame_idx = 0, + }, + .result = { + .is_dup = false, + .rx_status_flag = RX_FLAG_DUP_VALIDATED, + }, + }, + { + .desc = "non-QoS data, same sequence, same tid, has retry", + .rx_pkt = { + /* Driver will use tid = IWL_MAX_TID_COUNT */ + .fc = __cpu_to_le16(IEEE80211_FTYPE_DATA | + IEEE80211_FCTL_RETRY), + .seq = __cpu_to_le16(0x100), + }, + .dup_data_state = { + .tid = IWL_MAX_TID_COUNT, + .last_seq = __cpu_to_le16(0x100), + .last_sub_frame_idx = 0, + }, + .result = { + .is_dup = true, + .rx_status_flag = 0, + }, + }, + { + .desc = "non-QoS data, same sequence on different tid's", + .rx_pkt = { + /* Driver will use tid = IWL_MAX_TID_COUNT */ + .fc = __cpu_to_le16(IEEE80211_FTYPE_DATA), + .seq = __cpu_to_le16(0x100), + }, + .dup_data_state = { + .tid = 7, + .last_seq = __cpu_to_le16(0x100), + .last_sub_frame_idx = 0, + }, + .result = { + .is_dup = false, + .rx_status_flag = RX_FLAG_DUP_VALIDATED, + }, + }, + { + .desc = "A-MSDU new subframe, allow same PN", + .rx_pkt = { + .fc = __cpu_to_le16(IEEE80211_FTYPE_DATA | + IEEE80211_STYPE_QOS_DATA), + .seq = __cpu_to_le16(0x100), + .is_amsdu = true, + .sub_frame_idx = 1, + }, + .dup_data_state = { + .last_seq = __cpu_to_le16(0x100), + .last_sub_frame_idx = 0, + }, + .result = { + .is_dup = false, + .rx_status_flag = RX_FLAG_ALLOW_SAME_PN | + RX_FLAG_DUP_VALIDATED, + }, + }, + { + .desc = "A-MSDU subframe with smaller idx, disallow same PN", + .rx_pkt = { + .fc = __cpu_to_le16(IEEE80211_FTYPE_DATA | + IEEE80211_STYPE_QOS_DATA), + .seq = __cpu_to_le16(0x100), + .is_amsdu = true, + .sub_frame_idx = 1, + }, + .dup_data_state = { + .last_seq = __cpu_to_le16(0x100), + .last_sub_frame_idx = 2, + }, + .result = { + .is_dup = false, + .rx_status_flag = RX_FLAG_DUP_VALIDATED, + }, + }, + { + .desc = "A-MSDU same subframe, no retry, disallow same PN", + .rx_pkt = { + .fc = __cpu_to_le16(IEEE80211_FTYPE_DATA | + IEEE80211_STYPE_QOS_DATA), + .seq = __cpu_to_le16(0x100), + .is_amsdu = true, + .sub_frame_idx = 0, + }, + .dup_data_state = { + .last_seq = __cpu_to_le16(0x100), + .last_sub_frame_idx = 0, + }, + .result = { + .is_dup = false, + .rx_status_flag = RX_FLAG_DUP_VALIDATED, + }, + }, + { + .desc = "A-MSDU same subframe, has retry", + .rx_pkt = { + .fc = __cpu_to_le16(IEEE80211_FTYPE_DATA | + IEEE80211_STYPE_QOS_DATA | + IEEE80211_FCTL_RETRY), + .seq = __cpu_to_le16(0x100), + .is_amsdu = true, + .sub_frame_idx = 0, + }, + .dup_data_state = { + .last_seq = __cpu_to_le16(0x100), + .last_sub_frame_idx = 0, + }, + .result = { + .is_dup = true, + .rx_status_flag = 0, + }, + }, +}; + +KUNIT_ARRAY_PARAM_DESC(test_is_dup, is_dup_cases, desc); + +static void +setup_dup_data_state(struct ieee80211_sta *sta) +{ + struct kunit *test = kunit_get_current_test(); + const struct is_dup_case *param = (const void *)(test->param_value); + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(sta); + u8 tid = param->dup_data_state.tid; + struct iwl_mld_rxq_dup_data *dup_data; + + /* Allocate dup_data only for 1 queue */ + KUNIT_ALLOC_AND_ASSERT(test, dup_data); + + /* Initialize dup data, see iwl_mld_alloc_dup_data */ + memset(dup_data->last_seq, 0xff, sizeof(dup_data->last_seq)); + + dup_data->last_seq[tid] = param->dup_data_state.last_seq; + dup_data->last_sub_frame_idx[tid] = + param->dup_data_state.last_sub_frame_idx; + + mld_sta->dup_data = dup_data; +} + +static void setup_rx_pkt(const struct is_dup_case *param, + struct ieee80211_hdr *hdr, + struct iwl_rx_mpdu_desc *mpdu_desc) +{ + u8 tid = param->rx_pkt.tid; + + /* Set "new rx packet" header */ + hdr->frame_control = param->rx_pkt.fc; + hdr->seq_ctrl = param->rx_pkt.seq; + + if (ieee80211_is_data_qos(hdr->frame_control)) { + u8 *qc = ieee80211_get_qos_ctl(hdr); + + qc[0] = tid & IEEE80211_QOS_CTL_TID_MASK; + } + + if (param->rx_pkt.multicast) + hdr->addr1[0] = 0x1; + + /* Set mpdu_desc */ + mpdu_desc->amsdu_info = param->rx_pkt.sub_frame_idx & + IWL_RX_MPDU_AMSDU_SUBFRAME_IDX_MASK; + if (param->rx_pkt.is_amsdu) + mpdu_desc->mac_flags2 |= IWL_RX_MPDU_MFLG2_AMSDU; +} + +static void test_is_dup(struct kunit *test) +{ + const struct is_dup_case *param = (const void *)(test->param_value); + struct iwl_mld *mld = test->priv; + struct iwl_rx_mpdu_desc mpdu_desc = { }; + struct ieee80211_rx_status rx_status = { }; + struct ieee80211_vif *vif; + struct ieee80211_sta *sta; + struct ieee80211_hdr hdr; + + vif = iwlmld_kunit_add_vif(false, NL80211_IFTYPE_STATION); + sta = iwlmld_kunit_setup_sta(vif, IEEE80211_STA_AUTHORIZED, -1); + + /* Prepare test case state */ + setup_dup_data_state(sta); + setup_rx_pkt(param, &hdr, &mpdu_desc); + + KUNIT_EXPECT_EQ(test, + iwl_mld_is_dup(mld, sta, &hdr, &mpdu_desc, &rx_status, + 0), /* assuming only 1 queue */ + param->result.is_dup); + KUNIT_EXPECT_EQ(test, rx_status.flag, param->result.rx_status_flag); +} + +static struct kunit_case is_dup_test_cases[] = { + KUNIT_CASE_PARAM(test_is_dup, test_is_dup_gen_params), + {}, +}; + +static struct kunit_suite is_dup = { + .name = "iwlmld-rx-is-dup", + .test_cases = is_dup_test_cases, + .init = iwlmld_kunit_test_init, +}; + +kunit_test_suite(is_dup); diff --git a/drivers/net/wireless/intel/iwlwifi/mld/tests/utils.c b/drivers/net/wireless/intel/iwlwifi/mld/tests/utils.c new file mode 100644 index 0000000000000..a8c1e6c721388 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/tests/utils.c @@ -0,0 +1,475 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * KUnit tests for channel helper functions + * + * Copyright (C) 2024-2025 Intel Corporation + */ +#include +#include + +#include "utils.h" + +#include + +#include "fw/api/scan.h" +#include "fw/api/mac-cfg.h" +#include "iwl-trans.h" +#include "mld.h" +#include "iface.h" +#include "link.h" +#include "phy.h" +#include "sta.h" + +int iwlmld_kunit_test_init(struct kunit *test) +{ + struct iwl_mld *mld; + struct iwl_trans *trans; + const struct iwl_cfg *cfg; + struct iwl_fw *fw; + struct ieee80211_hw *hw; + + KUNIT_ALLOC_AND_ASSERT(test, trans); + KUNIT_ALLOC_AND_ASSERT(test, trans->dev); + KUNIT_ALLOC_AND_ASSERT(test, cfg); + KUNIT_ALLOC_AND_ASSERT(test, fw); + KUNIT_ALLOC_AND_ASSERT(test, hw); + KUNIT_ALLOC_AND_ASSERT(test, hw->wiphy); + + mutex_init(&hw->wiphy->mtx); + + /* Allocate and initialize the mld structure */ + KUNIT_ALLOC_AND_ASSERT(test, mld); + iwl_construct_mld(mld, trans, cfg, fw, hw, NULL); + + fw->ucode_capa.num_stations = IWL_STATION_COUNT_MAX; + fw->ucode_capa.num_links = IWL_FW_MAX_LINK_ID + 1; + + mld->fwrt.trans = trans; + mld->fwrt.fw = fw; + mld->fwrt.dev = trans->dev; + + /* TODO: add priv_size to hw allocation and setup hw->priv to enable + * testing mac80211 callbacks + */ + + KUNIT_ALLOC_AND_ASSERT(test, mld->nvm_data); + KUNIT_ALLOC_AND_ASSERT_SIZE(test, mld->scan.cmd, + sizeof(struct iwl_scan_req_umac_v17)); + mld->scan.cmd_size = sizeof(struct iwl_scan_req_umac_v17); + + /* This is not the state at the end of the regular opmode_start, + * but it is more common to need it. Explicitly undo this if needed. + */ + mld->trans->state = IWL_TRANS_FW_ALIVE; + mld->fw_status.running = true; + + /* Avoid passing mld struct around */ + test->priv = mld; + return 0; +} + +IWL_MLD_ALLOC_FN(link, bss_conf) + +static void iwlmld_kunit_init_link(struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link, + struct iwl_mld_link *mld_link, int link_id) +{ + struct kunit *test = kunit_get_current_test(); + struct iwl_mld *mld = test->priv; + struct iwl_mld_vif *mld_vif = iwl_mld_vif_from_mac80211(vif); + int ret; + + /* setup mac80211 link */ + rcu_assign_pointer(vif->link_conf[link_id], link); + link->link_id = link_id; + link->vif = vif; + link->beacon_int = 100; + link->dtim_period = 3; + link->qos = true; + + /* and mld_link */ + ret = iwl_mld_allocate_link_fw_id(mld, &mld_link->fw_id, link); + KUNIT_ASSERT_EQ(test, ret, 0); + rcu_assign_pointer(mld_vif->link[link_id], mld_link); + rcu_assign_pointer(vif->link_conf[link_id], link); +} + +IWL_MLD_ALLOC_FN(vif, vif) + +/* Helper function to add and initialize a VIF for KUnit tests */ +struct ieee80211_vif *iwlmld_kunit_add_vif(bool mlo, enum nl80211_iftype type) +{ + struct kunit *test = kunit_get_current_test(); + struct iwl_mld *mld = test->priv; + struct ieee80211_vif *vif; + struct iwl_mld_vif *mld_vif; + int ret; + + /* TODO: support more types */ + KUNIT_ASSERT_EQ(test, type, NL80211_IFTYPE_STATION); + + KUNIT_ALLOC_AND_ASSERT_SIZE(test, vif, + sizeof(*vif) + sizeof(*mld_vif)); + + vif->type = type; + mld_vif = iwl_mld_vif_from_mac80211(vif); + mld_vif->mld = mld; + + ret = iwl_mld_allocate_vif_fw_id(mld, &mld_vif->fw_id, vif); + KUNIT_ASSERT_EQ(test, ret, 0); + + /* TODO: revisit (task=EHT) */ + if (mlo) + return vif; + + /* Initialize the default link */ + iwlmld_kunit_init_link(vif, &vif->bss_conf, &mld_vif->deflink, 0); + + return vif; +} + +/* Use only for MLO vif */ +struct ieee80211_bss_conf * +iwlmld_kunit_add_link(struct ieee80211_vif *vif, int link_id) +{ + struct kunit *test = kunit_get_current_test(); + struct ieee80211_bss_conf *link; + struct iwl_mld_link *mld_link; + + KUNIT_ALLOC_AND_ASSERT(test, link); + KUNIT_ALLOC_AND_ASSERT(test, mld_link); + + iwlmld_kunit_init_link(vif, link, mld_link, link_id); + vif->valid_links |= BIT(link_id); + + return link; +} + +struct ieee80211_chanctx_conf * +iwlmld_kunit_add_chanctx_from_def(struct cfg80211_chan_def *def) +{ + struct kunit *test = kunit_get_current_test(); + struct iwl_mld *mld = test->priv; + struct ieee80211_chanctx_conf *ctx; + struct iwl_mld_phy *phy; + int fw_id; + + KUNIT_ALLOC_AND_ASSERT_SIZE(test, ctx, sizeof(*ctx) + sizeof(*phy)); + + /* Setup the chanctx conf */ + ctx->def = *def; + ctx->min_def = *def; + ctx->ap = *def; + + /* and the iwl_mld_phy */ + phy = iwl_mld_phy_from_mac80211(ctx); + + fw_id = iwl_mld_allocate_fw_phy_id(mld); + KUNIT_ASSERT_GE(test, fw_id, 0); + + phy->fw_id = fw_id; + phy->chandef = *iwl_mld_get_chandef_from_chanctx(ctx); + + return ctx; +} + +void iwlmld_kunit_assign_chanctx_to_link(struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link, + struct ieee80211_chanctx_conf *ctx) +{ + struct kunit *test = kunit_get_current_test(); + struct iwl_mld *mld = test->priv; + struct iwl_mld_link *mld_link; + + KUNIT_EXPECT_NULL(test, rcu_access_pointer(link->chanctx_conf)); + rcu_assign_pointer(link->chanctx_conf, ctx); + + lockdep_assert_wiphy(mld->wiphy); + + mld_link = iwl_mld_link_from_mac80211(link); + + KUNIT_EXPECT_NULL(test, rcu_access_pointer(mld_link->chan_ctx)); + KUNIT_EXPECT_FALSE(test, mld_link->active); + + rcu_assign_pointer(mld_link->chan_ctx, ctx); + mld_link->active = true; + + if (ieee80211_vif_is_mld(vif)) + vif->active_links |= BIT(link->link_id); +} + +IWL_MLD_ALLOC_FN(link_sta, link_sta) + +static void iwlmld_kunit_add_link_sta(struct ieee80211_sta *sta, + struct ieee80211_link_sta *link_sta, + struct iwl_mld_link_sta *mld_link_sta, + u8 link_id) +{ + struct kunit *test = kunit_get_current_test(); + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(sta); + struct iwl_mld *mld = test->priv; + u8 fw_id; + int ret; + + /* initialize mac80211's link_sta */ + link_sta->link_id = link_id; + rcu_assign_pointer(sta->link[link_id], link_sta); + + link_sta->sta = sta; + + /* and the iwl_mld_link_sta */ + ret = iwl_mld_allocate_link_sta_fw_id(mld, &fw_id, link_sta); + KUNIT_ASSERT_EQ(test, ret, 0); + mld_link_sta->fw_id = fw_id; + + rcu_assign_pointer(mld_sta->link[link_id], mld_link_sta); +} + +static struct ieee80211_link_sta * +iwlmld_kunit_alloc_link_sta(struct ieee80211_sta *sta, int link_id) +{ + struct kunit *test = kunit_get_current_test(); + struct ieee80211_link_sta *link_sta; + struct iwl_mld_link_sta *mld_link_sta; + + /* Only valid for MLO */ + KUNIT_ASSERT_TRUE(test, sta->valid_links); + + KUNIT_ALLOC_AND_ASSERT(test, link_sta); + KUNIT_ALLOC_AND_ASSERT(test, mld_link_sta); + + iwlmld_kunit_add_link_sta(sta, link_sta, mld_link_sta, link_id); + + sta->valid_links |= BIT(link_id); + + return link_sta; +} + +/* Allocate and initialize a STA with the first link_sta */ +static struct ieee80211_sta * +iwlmld_kunit_add_sta(struct ieee80211_vif *vif, int link_id) +{ + struct kunit *test = kunit_get_current_test(); + struct ieee80211_sta *sta; + struct iwl_mld_sta *mld_sta; + + /* Allocate memory for ieee80211_sta with embedded iwl_mld_sta */ + KUNIT_ALLOC_AND_ASSERT_SIZE(test, sta, sizeof(*sta) + sizeof(*mld_sta)); + + /* TODO: allocate and initialize the TXQs ? */ + + mld_sta = iwl_mld_sta_from_mac80211(sta); + mld_sta->vif = vif; + mld_sta->mld = test->priv; + + /* TODO: adjust for internal stations */ + mld_sta->sta_type = STATION_TYPE_PEER; + + if (link_id >= 0) { + iwlmld_kunit_add_link_sta(sta, &sta->deflink, + &mld_sta->deflink, link_id); + sta->valid_links = BIT(link_id); + } else { + iwlmld_kunit_add_link_sta(sta, &sta->deflink, + &mld_sta->deflink, 0); + } + return sta; +} + +/* Move s STA to a state */ +static void iwlmld_kunit_move_sta_state(struct ieee80211_vif *vif, + struct ieee80211_sta *sta, + enum ieee80211_sta_state state) +{ + struct kunit *test = kunit_get_current_test(); + struct iwl_mld_sta *mld_sta; + struct iwl_mld_vif *mld_vif; + + /* The sta will be removed automatically at the end of the test */ + KUNIT_ASSERT_NE(test, state, IEEE80211_STA_NOTEXIST); + + mld_sta = iwl_mld_sta_from_mac80211(sta); + mld_sta->sta_state = state; + + mld_vif = iwl_mld_vif_from_mac80211(mld_sta->vif); + mld_vif->authorized = state == IEEE80211_STA_AUTHORIZED; + + if (vif->type == NL80211_IFTYPE_STATION && !sta->tdls) + mld_vif->ap_sta = sta; +} + +struct ieee80211_sta *iwlmld_kunit_setup_sta(struct ieee80211_vif *vif, + enum ieee80211_sta_state state, + int link_id) +{ + struct kunit *test = kunit_get_current_test(); + struct ieee80211_sta *sta; + + /* The sta will be removed automatically at the end of the test */ + KUNIT_ASSERT_NE(test, state, IEEE80211_STA_NOTEXIST); + + /* First - allocate and init the STA */ + sta = iwlmld_kunit_add_sta(vif, link_id); + + /* Now move it all the way to the wanted state */ + for (enum ieee80211_sta_state _state = IEEE80211_STA_NONE; + _state <= state; _state++) + iwlmld_kunit_move_sta_state(vif, sta, state); + + return sta; +} + +static void iwlmld_kunit_set_vif_associated(struct ieee80211_vif *vif) +{ + /* TODO: setup chanreq */ + /* TODO setup capabilities */ + + vif->cfg.assoc = 1; +} + +static struct ieee80211_vif * +iwlmld_kunit_setup_assoc(bool mlo, u8 link_id, enum nl80211_band band) +{ + struct kunit *test = kunit_get_current_test(); + struct iwl_mld *mld = test->priv; + struct ieee80211_vif *vif; + struct ieee80211_bss_conf *link; + struct ieee80211_chanctx_conf *chan_ctx; + + KUNIT_ASSERT_TRUE(test, mlo || link_id == 0); + + vif = iwlmld_kunit_add_vif(mlo, NL80211_IFTYPE_STATION); + + if (mlo) + link = iwlmld_kunit_add_link(vif, link_id); + else + link = &vif->bss_conf; + + chan_ctx = iwlmld_kunit_add_chanctx(band); + + wiphy_lock(mld->wiphy); + iwlmld_kunit_assign_chanctx_to_link(vif, link, chan_ctx); + wiphy_unlock(mld->wiphy); + + /* The AP sta will now be pointer to by mld_vif->ap_sta */ + iwlmld_kunit_setup_sta(vif, IEEE80211_STA_AUTHORIZED, link_id); + + iwlmld_kunit_set_vif_associated(vif); + + return vif; +} + +struct ieee80211_vif *iwlmld_kunit_setup_mlo_assoc(u16 valid_links, + u8 assoc_link_id, + enum nl80211_band band) +{ + struct kunit *test = kunit_get_current_test(); + struct ieee80211_vif *vif; + + KUNIT_ASSERT_TRUE(test, + hweight16(valid_links) == 1 || + hweight16(valid_links) == 2); + KUNIT_ASSERT_TRUE(test, valid_links & BIT(assoc_link_id)); + + vif = iwlmld_kunit_setup_assoc(true, assoc_link_id, band); + + /* Add the other link, if applicable */ + if (hweight16(valid_links) > 1) { + u8 other_link_id = ffs(valid_links & ~BIT(assoc_link_id)) - 1; + + iwlmld_kunit_add_link(vif, other_link_id); + } + + return vif; +} + +struct ieee80211_vif *iwlmld_kunit_setup_non_mlo_assoc(enum nl80211_band band) +{ + return iwlmld_kunit_setup_assoc(false, 0, band); +} + +struct iwl_rx_packet * +_iwl_mld_kunit_create_pkt(const void *notif, size_t notif_sz) +{ + struct kunit *test = kunit_get_current_test(); + struct iwl_rx_packet *pkt; + + KUNIT_ALLOC_AND_ASSERT_SIZE(test, pkt, sizeof(pkt) + notif_sz); + + memcpy(pkt->data, notif, notif_sz); + pkt->len_n_flags = cpu_to_le32(sizeof(pkt->hdr) + notif_sz); + + return pkt; +} + +struct ieee80211_vif *iwlmld_kunit_assoc_emlsr(u16 valid_links, + enum nl80211_band band1, + enum nl80211_band band2) +{ + struct kunit *test = kunit_get_current_test(); + struct iwl_mld *mld = test->priv; + struct ieee80211_vif *vif; + struct ieee80211_bss_conf *link; + struct ieee80211_chanctx_conf *chan_ctx; + struct ieee80211_sta *sta; + struct iwl_mld_vif *mld_vif; + u8 assoc_link_id, other_link_id; + + KUNIT_ASSERT_TRUE(test, hweight16(valid_links) == 2); + + assoc_link_id = ffs(valid_links) - 1; + other_link_id = ffs(valid_links & ~BIT(assoc_link_id)) - 1; + + vif = iwlmld_kunit_setup_mlo_assoc(valid_links, assoc_link_id, band1); + mld_vif = iwl_mld_vif_from_mac80211(vif); + + /* Activate second link */ + wiphy_lock(mld->wiphy); + + link = wiphy_dereference(mld->wiphy, vif->link_conf[other_link_id]); + KUNIT_EXPECT_NOT_NULL(test, link); + + chan_ctx = iwlmld_kunit_add_chanctx(band2); + iwlmld_kunit_assign_chanctx_to_link(vif, link, chan_ctx); + + wiphy_unlock(mld->wiphy); + + /* And other link sta */ + sta = mld_vif->ap_sta; + KUNIT_EXPECT_NOT_NULL(test, sta); + + iwlmld_kunit_alloc_link_sta(sta, other_link_id); + + return vif; +} + +struct element *iwlmld_kunit_gen_element(u8 id, const void *data, size_t len) +{ + struct kunit *test = kunit_get_current_test(); + struct element *elem; + + KUNIT_ALLOC_AND_ASSERT_SIZE(test, elem, sizeof(*elem) + len); + + elem->id = id; + elem->datalen = len; + memcpy(elem->data, data, len); + + return elem; +} + +struct iwl_mld_phy *iwlmld_kunit_get_phy_of_link(struct ieee80211_vif *vif, + u8 link_id) +{ + struct kunit *test = kunit_get_current_test(); + struct iwl_mld *mld = test->priv; + struct ieee80211_chanctx_conf *chanctx; + struct ieee80211_bss_conf *link = + wiphy_dereference(mld->wiphy, vif->link_conf[link_id]); + + KUNIT_EXPECT_NOT_NULL(test, link); + + chanctx = wiphy_dereference(mld->wiphy, link->chanctx_conf); + KUNIT_EXPECT_NOT_NULL(test, chanctx); + + return iwl_mld_phy_from_mac80211(chanctx); +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/tests/utils.h b/drivers/net/wireless/intel/iwlwifi/mld/tests/utils.h new file mode 100644 index 0000000000000..9b1f2ada997f3 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/tests/utils.h @@ -0,0 +1,124 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024-2025 Intel Corporation + */ + +#ifndef __iwl_mld_kunit_utils_h__ +#define __iwl_mld_kunit_utils_h__ + +#include +#include + +struct iwl_mld; + +int iwlmld_kunit_test_init(struct kunit *test); + +enum nl80211_iftype; + +struct ieee80211_vif *iwlmld_kunit_add_vif(bool mlo, enum nl80211_iftype type); + +struct ieee80211_bss_conf * +iwlmld_kunit_add_link(struct ieee80211_vif *vif, int link_id); + +#define KUNIT_ALLOC_AND_ASSERT_SIZE(test, ptr, size) \ +do { \ + (ptr) = kunit_kzalloc((test), (size), GFP_KERNEL); \ + KUNIT_ASSERT_NOT_NULL((test), (ptr)); \ +} while (0) + +#define KUNIT_ALLOC_AND_ASSERT(test, ptr) \ + KUNIT_ALLOC_AND_ASSERT_SIZE(test, ptr, sizeof(*(ptr))) + +#define CHANNEL(_name, _band, _freq) \ +static struct ieee80211_channel _name = { \ + .band = (_band), \ + .center_freq = (_freq), \ + .hw_value = (_freq), \ +} + +#define CHANDEF(_name, _channel, _freq1, _width) \ +__maybe_unused static struct cfg80211_chan_def _name = { \ + .chan = &(_channel), \ + .center_freq1 = (_freq1), \ + .width = (_width), \ +} + +CHANNEL(chan_2ghz, NL80211_BAND_2GHZ, 2412); +CHANNEL(chan_5ghz, NL80211_BAND_5GHZ, 5200); +CHANNEL(chan_6ghz, NL80211_BAND_6GHZ, 6115); +/* Feel free to add more */ + +CHANDEF(chandef_2ghz, chan_2ghz, 2412, NL80211_CHAN_WIDTH_20); +CHANDEF(chandef_5ghz, chan_5ghz, 5200, NL80211_CHAN_WIDTH_40); +CHANDEF(chandef_6ghz, chan_6ghz, 6115, NL80211_CHAN_WIDTH_160); +/* Feel free to add more */ + +//struct cfg80211_chan_def; + +struct ieee80211_chanctx_conf * +iwlmld_kunit_add_chanctx_from_def(struct cfg80211_chan_def *def); + +static inline struct ieee80211_chanctx_conf * +iwlmld_kunit_add_chanctx(enum nl80211_band band) +{ + struct cfg80211_chan_def *chandef; + + switch (band) { + case NL80211_BAND_2GHZ: + chandef = &chandef_2ghz; + break; + case NL80211_BAND_5GHZ: + chandef = &chandef_5ghz; + break; + default: + case NL80211_BAND_6GHZ: + chandef = &chandef_6ghz; + break; + } + + return iwlmld_kunit_add_chanctx_from_def(chandef); +} + +void iwlmld_kunit_assign_chanctx_to_link(struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link, + struct ieee80211_chanctx_conf *ctx); + +/* Allocate a sta, initialize it and move it to the wanted state */ +struct ieee80211_sta *iwlmld_kunit_setup_sta(struct ieee80211_vif *vif, + enum ieee80211_sta_state state, + int link_id); + +struct ieee80211_vif *iwlmld_kunit_setup_mlo_assoc(u16 valid_links, + u8 assoc_link_id, + enum nl80211_band band); +struct ieee80211_vif *iwlmld_kunit_setup_non_mlo_assoc(enum nl80211_band band); + +struct iwl_rx_packet * +_iwl_mld_kunit_create_pkt(const void *notif, size_t notif_sz); + +#define iwl_mld_kunit_create_pkt(_notif) \ + _iwl_mld_kunit_create_pkt(&(_notif), sizeof(_notif)) + +struct ieee80211_vif *iwlmld_kunit_assoc_emlsr(u16 valid_links, + enum nl80211_band band1, + enum nl80211_band band2); + +struct element *iwlmld_kunit_gen_element(u8 id, const void *data, size_t len); + +/** + * iwlmld_kunit_get_phy_of_link - Get the phy of a link + * + * @vif: The vif to get the phy from. + * @link_id: The id of the link to get the phy for. + * + * given a vif and link id, return the phy pointer of that link. + * This assumes that the link exists, and that it had a chanctx + * assigned. + * If this is not the case, the test will fail. + * + * Return: phy pointer. + */ +struct iwl_mld_phy *iwlmld_kunit_get_phy_of_link(struct ieee80211_vif *vif, + u8 link_id); + +#endif /* __iwl_mld_kunit_utils_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/thermal.c b/drivers/net/wireless/intel/iwlwifi/mld/thermal.c new file mode 100644 index 0000000000000..1909953a9be98 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/thermal.c @@ -0,0 +1,438 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024-2025 Intel Corporation + */ +#ifdef CONFIG_THERMAL +#include +#include +#endif + +#include "fw/api/phy.h" + +#include "thermal.h" +#include "mld.h" +#include "hcmd.h" + +#define IWL_MLD_CT_KILL_DURATION (5 * HZ) + +void iwl_mld_handle_ct_kill_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + const struct ct_kill_notif *notif = (const void *)pkt->data; + + IWL_ERR(mld, + "CT Kill notification: temp = %d, DTS = 0x%x, Scheme 0x%x - Enter CT Kill\n", + le16_to_cpu(notif->temperature), notif->dts, + notif->scheme); + + iwl_mld_set_ctkill(mld, true); + + wiphy_delayed_work_queue(mld->wiphy, &mld->ct_kill_exit_wk, + round_jiffies_relative(IWL_MLD_CT_KILL_DURATION)); +} + +static void iwl_mld_exit_ctkill(struct wiphy *wiphy, struct wiphy_work *wk) +{ + struct iwl_mld *mld; + + mld = container_of(wk, struct iwl_mld, ct_kill_exit_wk.work); + + IWL_ERR(mld, "Exit CT Kill\n"); + iwl_mld_set_ctkill(mld, false); +} + +void iwl_mld_handle_temp_notif(struct iwl_mld *mld, struct iwl_rx_packet *pkt) +{ + const struct iwl_dts_measurement_notif *notif = + (const void *)pkt->data; + int temp; + u32 ths_crossed; + + temp = le32_to_cpu(notif->temp); + + /* shouldn't be negative, but since it's s32, make sure it isn't */ + if (IWL_FW_CHECK(mld, temp < 0, "negative temperature %d\n", temp)) + return; + + ths_crossed = le32_to_cpu(notif->threshold_idx); + + /* 0xFF in ths_crossed means the notification is not related + * to a trip, so we can ignore it here. + */ + if (ths_crossed == 0xFF) + return; + + IWL_DEBUG_TEMP(mld, "Temp = %d Threshold crossed = %d\n", + temp, ths_crossed); + + if (IWL_FW_CHECK(mld, ths_crossed >= IWL_MAX_DTS_TRIPS, + "bad threshold: %d\n", ths_crossed)) + return; + +#ifdef CONFIG_THERMAL + if (mld->tzone) + thermal_zone_device_update(mld->tzone, THERMAL_TRIP_VIOLATED); +#endif /* CONFIG_THERMAL */ +} + +#ifdef CONFIG_THERMAL +static int iwl_mld_get_temp(struct iwl_mld *mld, s32 *temp) +{ + struct iwl_host_cmd cmd = { + .id = WIDE_ID(PHY_OPS_GROUP, CMD_DTS_MEASUREMENT_TRIGGER_WIDE), + .flags = CMD_WANT_SKB, + }; + const struct iwl_dts_measurement_resp *resp; + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + ret = iwl_mld_send_cmd(mld, &cmd); + if (ret) { + IWL_ERR(mld, + "Failed to send the temperature measurement command (err=%d)\n", + ret); + return ret; + } + + if (iwl_rx_packet_payload_len(cmd.resp_pkt) < sizeof(*resp)) { + IWL_ERR(mld, + "Failed to get a valid response to DTS measurement\n"); + ret = -EIO; + goto free_resp; + } + + resp = (const void *)cmd.resp_pkt->data; + *temp = le32_to_cpu(resp->temp); + + IWL_DEBUG_TEMP(mld, + "Got temperature measurement response: temp=%d\n", + *temp); + +free_resp: + iwl_free_resp(&cmd); + return ret; +} + +static int compare_temps(const void *a, const void *b) +{ + return ((s16)le16_to_cpu(*(__le16 *)a) - + (s16)le16_to_cpu(*(__le16 *)b)); +} + +struct iwl_trip_walk_data { + __le16 *thresholds; + int count; +}; + +static int iwl_trip_temp_iter(struct thermal_trip *trip, void *arg) +{ + struct iwl_trip_walk_data *twd = arg; + + if (trip->temperature == THERMAL_TEMP_INVALID) + return 0; + + twd->thresholds[twd->count++] = + cpu_to_le16((s16)(trip->temperature / 1000)); + return 0; +} +#endif + +int iwl_mld_config_temp_report_ths(struct iwl_mld *mld) +{ + struct temp_report_ths_cmd cmd = {0}; + int ret; +#ifdef CONFIG_THERMAL + struct iwl_trip_walk_data twd = { + .thresholds = cmd.thresholds, + .count = 0 + }; + + if (!mld->tzone) + goto send; + + /* The thermal core holds an array of temperature trips that are + * unsorted and uncompressed, the FW should get it compressed and + * sorted. + */ + + /* compress trips to cmd array, remove uninitialized values*/ + for_each_thermal_trip(mld->tzone, iwl_trip_temp_iter, &twd); + + cmd.num_temps = cpu_to_le32(twd.count); + if (twd.count) + sort(cmd.thresholds, twd.count, sizeof(s16), + compare_temps, NULL); + +send: +#endif + lockdep_assert_wiphy(mld->wiphy); + + ret = iwl_mld_send_cmd_pdu(mld, WIDE_ID(PHY_OPS_GROUP, + TEMP_REPORTING_THRESHOLDS_CMD), + &cmd); + if (ret) + IWL_ERR(mld, "TEMP_REPORT_THS_CMD command failed (err=%d)\n", + ret); + + return ret; +} + +#ifdef CONFIG_THERMAL +static int iwl_mld_tzone_get_temp(struct thermal_zone_device *device, + int *temperature) +{ + struct iwl_mld *mld = thermal_zone_device_priv(device); + int temp; + int ret = 0; + + wiphy_lock(mld->wiphy); + + if (!mld->fw_status.running) { + /* Tell the core that there is no valid temperature value to + * return, but it need not worry about this. + */ + *temperature = THERMAL_TEMP_INVALID; + goto unlock; + } + + ret = iwl_mld_get_temp(mld, &temp); + if (ret) + goto unlock; + + *temperature = temp * 1000; +unlock: + wiphy_unlock(mld->wiphy); + return ret; +} + +static int iwl_mld_tzone_set_trip_temp(struct thermal_zone_device *device, + const struct thermal_trip *trip, + int temp) +{ + struct iwl_mld *mld = thermal_zone_device_priv(device); + int ret; + + wiphy_lock(mld->wiphy); + + if (!mld->fw_status.running) { + ret = -EIO; + goto unlock; + } + + if ((temp / 1000) > S16_MAX) { + ret = -EINVAL; + goto unlock; + } + + ret = iwl_mld_config_temp_report_ths(mld); +unlock: + wiphy_unlock(mld->wiphy); + return ret; +} + +static struct thermal_zone_device_ops tzone_ops = { + .get_temp = iwl_mld_tzone_get_temp, + .set_trip_temp = iwl_mld_tzone_set_trip_temp, +}; + +static void iwl_mld_thermal_zone_register(struct iwl_mld *mld) +{ + int ret; + char name[16]; + static atomic_t counter = ATOMIC_INIT(0); + struct thermal_trip trips[IWL_MAX_DTS_TRIPS] = { + [0 ... IWL_MAX_DTS_TRIPS - 1] = { + .temperature = THERMAL_TEMP_INVALID, + .type = THERMAL_TRIP_PASSIVE, + .flags = THERMAL_TRIP_FLAG_RW_TEMP, + }, + }; + + BUILD_BUG_ON(ARRAY_SIZE(name) >= THERMAL_NAME_LENGTH); + + sprintf(name, "iwlwifi_%u", atomic_inc_return(&counter) & 0xFF); + mld->tzone = + thermal_zone_device_register_with_trips(name, trips, + IWL_MAX_DTS_TRIPS, + mld, &tzone_ops, + NULL, 0, 0); + if (IS_ERR(mld->tzone)) { + IWL_DEBUG_TEMP(mld, + "Failed to register to thermal zone (err = %ld)\n", + PTR_ERR(mld->tzone)); + mld->tzone = NULL; + return; + } + + ret = thermal_zone_device_enable(mld->tzone); + if (ret) { + IWL_DEBUG_TEMP(mld, "Failed to enable thermal zone\n"); + thermal_zone_device_unregister(mld->tzone); + } +} + +/* budget in mWatt */ +static const u32 iwl_mld_cdev_budgets[] = { + 2400, /* cooling state 0 */ + 2000, /* cooling state 1 */ + 1800, /* cooling state 2 */ + 1600, /* cooling state 3 */ + 1400, /* cooling state 4 */ + 1200, /* cooling state 5 */ + 1000, /* cooling state 6 */ + 900, /* cooling state 7 */ + 800, /* cooling state 8 */ + 700, /* cooling state 9 */ + 650, /* cooling state 10 */ + 600, /* cooling state 11 */ + 550, /* cooling state 12 */ + 500, /* cooling state 13 */ + 450, /* cooling state 14 */ + 400, /* cooling state 15 */ + 350, /* cooling state 16 */ + 300, /* cooling state 17 */ + 250, /* cooling state 18 */ + 200, /* cooling state 19 */ + 150, /* cooling state 20 */ +}; + +int iwl_mld_config_ctdp(struct iwl_mld *mld, u32 state, + enum iwl_ctdp_cmd_operation op) +{ + struct iwl_ctdp_cmd cmd = { + .operation = cpu_to_le32(op), + .budget = cpu_to_le32(iwl_mld_cdev_budgets[state]), + .window_size = 0, + }; + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + ret = iwl_mld_send_cmd_pdu(mld, WIDE_ID(PHY_OPS_GROUP, CTDP_CONFIG_CMD), + &cmd); + + if (ret) { + IWL_ERR(mld, "cTDP command failed (err=%d)\n", ret); + return ret; + } + + if (op == CTDP_CMD_OPERATION_START) + mld->cooling_dev.cur_state = state; + + return 0; +} + +static int iwl_mld_tcool_get_max_state(struct thermal_cooling_device *cdev, + unsigned long *state) +{ + *state = ARRAY_SIZE(iwl_mld_cdev_budgets) - 1; + + return 0; +} + +static int iwl_mld_tcool_get_cur_state(struct thermal_cooling_device *cdev, + unsigned long *state) +{ + struct iwl_mld *mld = (struct iwl_mld *)(cdev->devdata); + + *state = mld->cooling_dev.cur_state; + + return 0; +} + +static int iwl_mld_tcool_set_cur_state(struct thermal_cooling_device *cdev, + unsigned long new_state) +{ + struct iwl_mld *mld = (struct iwl_mld *)(cdev->devdata); + int ret; + + wiphy_lock(mld->wiphy); + + if (!mld->fw_status.running) { + ret = -EIO; + goto unlock; + } + + if (new_state >= ARRAY_SIZE(iwl_mld_cdev_budgets)) { + ret = -EINVAL; + goto unlock; + } + + ret = iwl_mld_config_ctdp(mld, new_state, CTDP_CMD_OPERATION_START); + +unlock: + wiphy_unlock(mld->wiphy); + return ret; +} + +static const struct thermal_cooling_device_ops tcooling_ops = { + .get_max_state = iwl_mld_tcool_get_max_state, + .get_cur_state = iwl_mld_tcool_get_cur_state, + .set_cur_state = iwl_mld_tcool_set_cur_state, +}; + +static void iwl_mld_cooling_device_register(struct iwl_mld *mld) +{ + char name[] = "iwlwifi"; + + BUILD_BUG_ON(ARRAY_SIZE(name) >= THERMAL_NAME_LENGTH); + + mld->cooling_dev.cdev = + thermal_cooling_device_register(name, + mld, + &tcooling_ops); + + if (IS_ERR(mld->cooling_dev.cdev)) { + IWL_DEBUG_TEMP(mld, + "Failed to register to cooling device (err = %ld)\n", + PTR_ERR(mld->cooling_dev.cdev)); + mld->cooling_dev.cdev = NULL; + return; + } +} + +static void iwl_mld_thermal_zone_unregister(struct iwl_mld *mld) +{ + if (!mld->tzone) + return; + + IWL_DEBUG_TEMP(mld, "Thermal zone device unregister\n"); + if (mld->tzone) { + thermal_zone_device_unregister(mld->tzone); + mld->tzone = NULL; + } +} + +static void iwl_mld_cooling_device_unregister(struct iwl_mld *mld) +{ + if (!mld->cooling_dev.cdev) + return; + + IWL_DEBUG_TEMP(mld, "Cooling device unregister\n"); + if (mld->cooling_dev.cdev) { + thermal_cooling_device_unregister(mld->cooling_dev.cdev); + mld->cooling_dev.cdev = NULL; + } +} +#endif /* CONFIG_THERMAL */ + +void iwl_mld_thermal_initialize(struct iwl_mld *mld) +{ + wiphy_delayed_work_init(&mld->ct_kill_exit_wk, iwl_mld_exit_ctkill); + +#ifdef CONFIG_THERMAL + iwl_mld_cooling_device_register(mld); + iwl_mld_thermal_zone_register(mld); +#endif +} + +void iwl_mld_thermal_exit(struct iwl_mld *mld) +{ + wiphy_delayed_work_cancel(mld->wiphy, &mld->ct_kill_exit_wk); + +#ifdef CONFIG_THERMAL + iwl_mld_cooling_device_unregister(mld); + iwl_mld_thermal_zone_unregister(mld); +#endif +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/thermal.h b/drivers/net/wireless/intel/iwlwifi/mld/thermal.h new file mode 100644 index 0000000000000..8c8893331b05c --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/thermal.h @@ -0,0 +1,36 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024 Intel Corporation + */ +#ifndef __iwl_mld_thermal_h__ +#define __iwl_mld_thermal_h__ + +#include "iwl-trans.h" + +struct iwl_mld; + +#ifdef CONFIG_THERMAL +#include + +/* + * struct iwl_mld_cooling_device + * @cur_state: current state + * @cdev: struct thermal cooling device + */ +struct iwl_mld_cooling_device { + u32 cur_state; + struct thermal_cooling_device *cdev; +}; + +int iwl_mld_config_ctdp(struct iwl_mld *mld, u32 state, + enum iwl_ctdp_cmd_operation op); +#endif + +void iwl_mld_handle_temp_notif(struct iwl_mld *mld, struct iwl_rx_packet *pkt); +void iwl_mld_handle_ct_kill_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt); +int iwl_mld_config_temp_report_ths(struct iwl_mld *mld); +void iwl_mld_thermal_initialize(struct iwl_mld *mld); +void iwl_mld_thermal_exit(struct iwl_mld *mld); + +#endif /* __iwl_mld_thermal_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/time_sync.c b/drivers/net/wireless/intel/iwlwifi/mld/time_sync.c new file mode 100644 index 0000000000000..50799f9bfccba --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/time_sync.c @@ -0,0 +1,240 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2025 Intel Corporation + */ + +#include "mld.h" +#include "hcmd.h" +#include "ptp.h" +#include "time_sync.h" +#include + +static int iwl_mld_init_time_sync(struct iwl_mld *mld, u32 protocols, + const u8 *addr) +{ + struct iwl_mld_time_sync_data *time_sync = kzalloc(sizeof(*time_sync), + GFP_KERNEL); + + if (!time_sync) + return -ENOMEM; + + time_sync->active_protocols = protocols; + ether_addr_copy(time_sync->peer_addr, addr); + skb_queue_head_init(&time_sync->frame_list); + rcu_assign_pointer(mld->time_sync, time_sync); + + return 0; +} + +int iwl_mld_time_sync_fw_config(struct iwl_mld *mld) +{ + struct iwl_time_sync_cfg_cmd cmd = {}; + struct iwl_mld_time_sync_data *time_sync; + int err; + + time_sync = wiphy_dereference(mld->wiphy, mld->time_sync); + if (!time_sync) + return -EINVAL; + + cmd.protocols = cpu_to_le32(time_sync->active_protocols); + ether_addr_copy(cmd.peer_addr, time_sync->peer_addr); + + err = iwl_mld_send_cmd_pdu(mld, + WIDE_ID(DATA_PATH_GROUP, + WNM_80211V_TIMING_MEASUREMENT_CONFIG_CMD), + &cmd); + if (err) + IWL_ERR(mld, "Failed to send time sync cfg cmd: %d\n", err); + + return err; +} + +int iwl_mld_time_sync_config(struct iwl_mld *mld, const u8 *addr, u32 protocols) +{ + struct iwl_mld_time_sync_data *time_sync; + int err; + + time_sync = wiphy_dereference(mld->wiphy, mld->time_sync); + + /* The fw only supports one peer. We do allow reconfiguration of the + * same peer for cases of fw reset etc. + */ + if (time_sync && time_sync->active_protocols && + !ether_addr_equal(addr, time_sync->peer_addr)) { + IWL_DEBUG_INFO(mld, "Time sync: reject config for peer: %pM\n", + addr); + return -ENOBUFS; + } + + if (protocols & ~(IWL_TIME_SYNC_PROTOCOL_TM | + IWL_TIME_SYNC_PROTOCOL_FTM)) + return -EINVAL; + + IWL_DEBUG_INFO(mld, "Time sync: set peer addr=%pM\n", addr); + + iwl_mld_deinit_time_sync(mld); + err = iwl_mld_init_time_sync(mld, protocols, addr); + if (err) + return err; + + err = iwl_mld_time_sync_fw_config(mld); + return err; +} + +void iwl_mld_deinit_time_sync(struct iwl_mld *mld) +{ + struct iwl_mld_time_sync_data *time_sync = + wiphy_dereference(mld->wiphy, mld->time_sync); + + if (!time_sync) + return; + + RCU_INIT_POINTER(mld->time_sync, NULL); + skb_queue_purge(&time_sync->frame_list); + kfree_rcu(time_sync, rcu_head); +} + +bool iwl_mld_time_sync_frame(struct iwl_mld *mld, struct sk_buff *skb, u8 *addr) +{ + struct iwl_mld_time_sync_data *time_sync; + + rcu_read_lock(); + time_sync = rcu_dereference(mld->time_sync); + if (time_sync && ether_addr_equal(time_sync->peer_addr, addr) && + (ieee80211_is_timing_measurement(skb) || ieee80211_is_ftm(skb))) { + skb_queue_tail(&time_sync->frame_list, skb); + rcu_read_unlock(); + return true; + } + rcu_read_unlock(); + + return false; +} + +static bool iwl_mld_is_skb_match(struct sk_buff *skb, u8 *addr, u8 dialog_token) +{ + struct ieee80211_mgmt *mgmt = (void *)skb->data; + u8 skb_dialog_token; + + if (ieee80211_is_timing_measurement(skb)) + skb_dialog_token = mgmt->u.action.u.wnm_timing_msr.dialog_token; + else + skb_dialog_token = mgmt->u.action.u.ftm.dialog_token; + + if ((ether_addr_equal(mgmt->sa, addr) || + ether_addr_equal(mgmt->da, addr)) && + skb_dialog_token == dialog_token) + return true; + + return false; +} + +static struct sk_buff *iwl_mld_time_sync_find_skb(struct iwl_mld *mld, u8 *addr, + u8 dialog_token) +{ + struct iwl_mld_time_sync_data *time_sync; + struct sk_buff *skb; + + rcu_read_lock(); + + time_sync = rcu_dereference(mld->time_sync); + if (IWL_FW_CHECK(mld, !time_sync, + "Time sync notification but time sync is not initialized\n")) { + rcu_read_unlock(); + return NULL; + } + + /* The notifications are expected to arrive in the same order of the + * frames. If the incoming notification doesn't match the first SKB + * in the queue, it means there was no time sync notification for this + * SKB and it can be dropped. + */ + while ((skb = skb_dequeue(&time_sync->frame_list))) { + if (iwl_mld_is_skb_match(skb, addr, dialog_token)) + break; + + kfree_skb(skb); + skb = NULL; + IWL_DEBUG_DROP(mld, + "Time sync: drop SKB without matching notification\n"); + } + rcu_read_unlock(); + + return skb; +} + +static u64 iwl_mld_get_64_bit(__le32 high, __le32 low) +{ + return ((u64)le32_to_cpu(high) << 32) | le32_to_cpu(low); +} + +void iwl_mld_handle_time_msmt_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + struct ptp_data *data = &mld->ptp_data; + struct iwl_time_msmt_notify *notif = (void *)pkt->data; + struct ieee80211_rx_status *rx_status; + struct skb_shared_hwtstamps *shwt; + u64 ts_10ns; + struct sk_buff *skb = + iwl_mld_time_sync_find_skb(mld, notif->peer_addr, + le32_to_cpu(notif->dialog_token)); + u64 adj_time; + + if (IWL_FW_CHECK(mld, !skb, "Time sync event but no pending skb\n")) + return; + + spin_lock_bh(&data->lock); + ts_10ns = iwl_mld_get_64_bit(notif->t2_hi, notif->t2_lo); + adj_time = iwl_mld_ptp_get_adj_time(mld, ts_10ns * 10); + shwt = skb_hwtstamps(skb); + shwt->hwtstamp = ktime_set(0, adj_time); + + ts_10ns = iwl_mld_get_64_bit(notif->t3_hi, notif->t3_lo); + adj_time = iwl_mld_ptp_get_adj_time(mld, ts_10ns * 10); + rx_status = IEEE80211_SKB_RXCB(skb); + rx_status->ack_tx_hwtstamp = ktime_set(0, adj_time); + spin_unlock_bh(&data->lock); + + IWL_DEBUG_INFO(mld, + "Time sync: RX event - report frame t2=%llu t3=%llu\n", + ktime_to_ns(shwt->hwtstamp), + ktime_to_ns(rx_status->ack_tx_hwtstamp)); + ieee80211_rx_napi(mld->hw, NULL, skb, NULL); +} + +void iwl_mld_handle_time_sync_confirm_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + struct ptp_data *data = &mld->ptp_data; + struct iwl_time_msmt_cfm_notify *notif = (void *)pkt->data; + struct ieee80211_tx_status status = {}; + struct skb_shared_hwtstamps *shwt; + u64 ts_10ns, adj_time; + + status.skb = + iwl_mld_time_sync_find_skb(mld, notif->peer_addr, + le32_to_cpu(notif->dialog_token)); + + if (IWL_FW_CHECK(mld, !status.skb, + "Time sync confirm but no pending skb\n")) + return; + + spin_lock_bh(&data->lock); + ts_10ns = iwl_mld_get_64_bit(notif->t1_hi, notif->t1_lo); + adj_time = iwl_mld_ptp_get_adj_time(mld, ts_10ns * 10); + shwt = skb_hwtstamps(status.skb); + shwt->hwtstamp = ktime_set(0, adj_time); + + ts_10ns = iwl_mld_get_64_bit(notif->t4_hi, notif->t4_lo); + adj_time = iwl_mld_ptp_get_adj_time(mld, ts_10ns * 10); + status.info = IEEE80211_SKB_CB(status.skb); + status.ack_hwtstamp = ktime_set(0, adj_time); + spin_unlock_bh(&data->lock); + + IWL_DEBUG_INFO(mld, + "Time sync: TX event - report frame t1=%llu t4=%llu\n", + ktime_to_ns(shwt->hwtstamp), + ktime_to_ns(status.ack_hwtstamp)); + ieee80211_tx_status_ext(mld->hw, &status); +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/time_sync.h b/drivers/net/wireless/intel/iwlwifi/mld/time_sync.h new file mode 100644 index 0000000000000..2d4c5512e9615 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/time_sync.h @@ -0,0 +1,26 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2025 Intel Corporation + */ +#ifndef __iwl_mld_time_sync_h__ +#define __iwl_mld_time_sync_h__ + +struct iwl_mld_time_sync_data { + struct rcu_head rcu_head; + u8 peer_addr[ETH_ALEN]; + u32 active_protocols; + struct sk_buff_head frame_list; +}; + +int iwl_mld_time_sync_config(struct iwl_mld *mld, const u8 *addr, + u32 protocols); +int iwl_mld_time_sync_fw_config(struct iwl_mld *mld); +void iwl_mld_deinit_time_sync(struct iwl_mld *mld); +void iwl_mld_handle_time_msmt_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt); +bool iwl_mld_time_sync_frame(struct iwl_mld *mld, struct sk_buff *skb, + u8 *addr); +void iwl_mld_handle_time_sync_confirm_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt); + +#endif /* __iwl_mld_time_sync_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/tlc.c b/drivers/net/wireless/intel/iwlwifi/mld/tlc.c new file mode 100644 index 0000000000000..85ec358f1417d --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/tlc.c @@ -0,0 +1,717 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024 Intel Corporation + */ + +#include + +#include "tlc.h" +#include "hcmd.h" +#include "sta.h" + +#include "fw/api/rs.h" +#include "fw/api/context.h" +#include "fw/api/dhc.h" + +static u8 iwl_mld_fw_bw_from_sta_bw(const struct ieee80211_link_sta *link_sta) +{ + switch (link_sta->bandwidth) { + case IEEE80211_STA_RX_BW_320: + return IWL_TLC_MNG_CH_WIDTH_320MHZ; + case IEEE80211_STA_RX_BW_160: + return IWL_TLC_MNG_CH_WIDTH_160MHZ; + case IEEE80211_STA_RX_BW_80: + return IWL_TLC_MNG_CH_WIDTH_80MHZ; + case IEEE80211_STA_RX_BW_40: + return IWL_TLC_MNG_CH_WIDTH_40MHZ; + case IEEE80211_STA_RX_BW_20: + default: + return IWL_TLC_MNG_CH_WIDTH_20MHZ; + } +} + +static __le16 +iwl_mld_get_tlc_cmd_flags(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_link_sta *link_sta, + const struct ieee80211_sta_he_cap *own_he_cap, + const struct ieee80211_sta_eht_cap *own_eht_cap) +{ + struct ieee80211_sta_ht_cap *ht_cap = &link_sta->ht_cap; + struct ieee80211_sta_vht_cap *vht_cap = &link_sta->vht_cap; + struct ieee80211_sta_he_cap *he_cap = &link_sta->he_cap; + bool has_vht = vht_cap->vht_supported; + u16 flags = 0; + + /* STBC flags */ + if (mld->cfg->ht_params->stbc && + (hweight8(iwl_mld_get_valid_tx_ant(mld)) > 1)) { + if (he_cap->has_he && he_cap->he_cap_elem.phy_cap_info[2] & + IEEE80211_HE_PHY_CAP2_STBC_RX_UNDER_80MHZ) + flags |= IWL_TLC_MNG_CFG_FLAGS_STBC_MSK; + else if (vht_cap->cap & IEEE80211_VHT_CAP_RXSTBC_MASK) + flags |= IWL_TLC_MNG_CFG_FLAGS_STBC_MSK; + else if (ht_cap->cap & IEEE80211_HT_CAP_RX_STBC) + flags |= IWL_TLC_MNG_CFG_FLAGS_STBC_MSK; + } + + /* LDPC */ + if (mld->cfg->ht_params->ldpc && + ((ht_cap->cap & IEEE80211_HT_CAP_LDPC_CODING) || + (has_vht && (vht_cap->cap & IEEE80211_VHT_CAP_RXLDPC)))) + flags |= IWL_TLC_MNG_CFG_FLAGS_LDPC_MSK; + + if (he_cap->has_he && (he_cap->he_cap_elem.phy_cap_info[1] & + IEEE80211_HE_PHY_CAP1_LDPC_CODING_IN_PAYLOAD)) + flags |= IWL_TLC_MNG_CFG_FLAGS_LDPC_MSK; + + if (own_he_cap && + !(own_he_cap->he_cap_elem.phy_cap_info[1] & + IEEE80211_HE_PHY_CAP1_LDPC_CODING_IN_PAYLOAD)) + flags &= ~IWL_TLC_MNG_CFG_FLAGS_LDPC_MSK; + + /* DCM */ + if (he_cap->has_he && + (he_cap->he_cap_elem.phy_cap_info[3] & + IEEE80211_HE_PHY_CAP3_DCM_MAX_CONST_RX_MASK && + own_he_cap && + own_he_cap->he_cap_elem.phy_cap_info[3] & + IEEE80211_HE_PHY_CAP3_DCM_MAX_CONST_TX_MASK)) + flags |= IWL_TLC_MNG_CFG_FLAGS_HE_DCM_NSS_1_MSK; + + /* Extra EHT LTF */ + if (own_eht_cap && + own_eht_cap->eht_cap_elem.phy_cap_info[5] & + IEEE80211_EHT_PHY_CAP5_SUPP_EXTRA_EHT_LTF && + link_sta->eht_cap.has_eht && + link_sta->eht_cap.eht_cap_elem.phy_cap_info[5] & + IEEE80211_EHT_PHY_CAP5_SUPP_EXTRA_EHT_LTF) { + flags |= IWL_TLC_MNG_CFG_FLAGS_EHT_EXTRA_LTF_MSK; + } + + return cpu_to_le16(flags); +} + +static u8 iwl_mld_get_fw_chains(struct iwl_mld *mld) +{ + u8 chains = iwl_mld_get_valid_tx_ant(mld); + u8 fw_chains = 0; + + if (chains & ANT_A) + fw_chains |= IWL_TLC_MNG_CHAIN_A_MSK; + if (chains & ANT_B) + fw_chains |= IWL_TLC_MNG_CHAIN_B_MSK; + + return fw_chains; +} + +static u8 iwl_mld_get_fw_sgi(struct ieee80211_link_sta *link_sta) +{ + struct ieee80211_sta_ht_cap *ht_cap = &link_sta->ht_cap; + struct ieee80211_sta_vht_cap *vht_cap = &link_sta->vht_cap; + struct ieee80211_sta_he_cap *he_cap = &link_sta->he_cap; + u8 sgi_chwidths = 0; + + /* If the association supports HE, HT/VHT rates will never be used for + * Tx and therefor there's no need to set the + * sgi-per-channel-width-support bits + */ + if (he_cap->has_he) + return 0; + + if (ht_cap->cap & IEEE80211_HT_CAP_SGI_20) + sgi_chwidths |= BIT(IWL_TLC_MNG_CH_WIDTH_20MHZ); + if (ht_cap->cap & IEEE80211_HT_CAP_SGI_40) + sgi_chwidths |= BIT(IWL_TLC_MNG_CH_WIDTH_40MHZ); + if (vht_cap->cap & IEEE80211_VHT_CAP_SHORT_GI_80) + sgi_chwidths |= BIT(IWL_TLC_MNG_CH_WIDTH_80MHZ); + if (vht_cap->cap & IEEE80211_VHT_CAP_SHORT_GI_160) + sgi_chwidths |= BIT(IWL_TLC_MNG_CH_WIDTH_160MHZ); + + return sgi_chwidths; +} + +static int +iwl_mld_get_highest_fw_mcs(const struct ieee80211_sta_vht_cap *vht_cap, + int nss) +{ + u16 rx_mcs = le16_to_cpu(vht_cap->vht_mcs.rx_mcs_map) & + (0x3 << (2 * (nss - 1))); + rx_mcs >>= (2 * (nss - 1)); + + switch (rx_mcs) { + case IEEE80211_VHT_MCS_SUPPORT_0_7: + return IWL_TLC_MNG_HT_RATE_MCS7; + case IEEE80211_VHT_MCS_SUPPORT_0_8: + return IWL_TLC_MNG_HT_RATE_MCS8; + case IEEE80211_VHT_MCS_SUPPORT_0_9: + return IWL_TLC_MNG_HT_RATE_MCS9; + default: + WARN_ON_ONCE(1); + break; + } + + return 0; +} + +static void +iwl_mld_fill_vht_rates(const struct ieee80211_link_sta *link_sta, + const struct ieee80211_sta_vht_cap *vht_cap, + struct iwl_tlc_config_cmd_v4 *cmd) +{ + u16 supp; + int i, highest_mcs; + u8 max_nss = link_sta->rx_nss; + struct ieee80211_vht_cap ieee_vht_cap = { + .vht_cap_info = cpu_to_le32(vht_cap->cap), + .supp_mcs = vht_cap->vht_mcs, + }; + + /* the station support only a single receive chain */ + if (link_sta->smps_mode == IEEE80211_SMPS_STATIC) + max_nss = 1; + + for (i = 0; i < max_nss && i < IWL_TLC_NSS_MAX; i++) { + int nss = i + 1; + + highest_mcs = iwl_mld_get_highest_fw_mcs(vht_cap, nss); + if (!highest_mcs) + continue; + + supp = BIT(highest_mcs + 1) - 1; + if (link_sta->bandwidth == IEEE80211_STA_RX_BW_20) + supp &= ~BIT(IWL_TLC_MNG_HT_RATE_MCS9); + + cmd->ht_rates[i][IWL_TLC_MCS_PER_BW_80] = cpu_to_le16(supp); + /* Check if VHT extended NSS indicates that the bandwidth/NSS + * configuration is supported - only for MCS 0 since we already + * decoded the MCS bits anyway ourselves. + */ + if (link_sta->bandwidth == IEEE80211_STA_RX_BW_160 && + ieee80211_get_vht_max_nss(&ieee_vht_cap, + IEEE80211_VHT_CHANWIDTH_160MHZ, + 0, true, nss) >= nss) + cmd->ht_rates[i][IWL_TLC_MCS_PER_BW_160] = + cmd->ht_rates[i][IWL_TLC_MCS_PER_BW_80]; + } +} + +static u16 iwl_mld_he_mac80211_mcs_to_fw_mcs(u16 mcs) +{ + switch (mcs) { + case IEEE80211_HE_MCS_SUPPORT_0_7: + return BIT(IWL_TLC_MNG_HT_RATE_MCS7 + 1) - 1; + case IEEE80211_HE_MCS_SUPPORT_0_9: + return BIT(IWL_TLC_MNG_HT_RATE_MCS9 + 1) - 1; + case IEEE80211_HE_MCS_SUPPORT_0_11: + return BIT(IWL_TLC_MNG_HT_RATE_MCS11 + 1) - 1; + case IEEE80211_HE_MCS_NOT_SUPPORTED: + return 0; + } + + WARN(1, "invalid HE MCS %d\n", mcs); + return 0; +} + +static void +iwl_mld_fill_he_rates(const struct ieee80211_link_sta *link_sta, + const struct ieee80211_sta_he_cap *own_he_cap, + struct iwl_tlc_config_cmd_v4 *cmd) +{ + const struct ieee80211_sta_he_cap *he_cap = &link_sta->he_cap; + u16 mcs_160 = le16_to_cpu(he_cap->he_mcs_nss_supp.rx_mcs_160); + u16 mcs_80 = le16_to_cpu(he_cap->he_mcs_nss_supp.rx_mcs_80); + u16 tx_mcs_80 = le16_to_cpu(own_he_cap->he_mcs_nss_supp.tx_mcs_80); + u16 tx_mcs_160 = le16_to_cpu(own_he_cap->he_mcs_nss_supp.tx_mcs_160); + int i; + u8 nss = link_sta->rx_nss; + + /* the station support only a single receive chain */ + if (link_sta->smps_mode == IEEE80211_SMPS_STATIC) + nss = 1; + + for (i = 0; i < nss && i < IWL_TLC_NSS_MAX; i++) { + u16 _mcs_160 = (mcs_160 >> (2 * i)) & 0x3; + u16 _mcs_80 = (mcs_80 >> (2 * i)) & 0x3; + u16 _tx_mcs_160 = (tx_mcs_160 >> (2 * i)) & 0x3; + u16 _tx_mcs_80 = (tx_mcs_80 >> (2 * i)) & 0x3; + + /* If one side doesn't support - mark both as not supporting */ + if (_mcs_80 == IEEE80211_HE_MCS_NOT_SUPPORTED || + _tx_mcs_80 == IEEE80211_HE_MCS_NOT_SUPPORTED) { + _mcs_80 = IEEE80211_HE_MCS_NOT_SUPPORTED; + _tx_mcs_80 = IEEE80211_HE_MCS_NOT_SUPPORTED; + } + if (_mcs_80 > _tx_mcs_80) + _mcs_80 = _tx_mcs_80; + cmd->ht_rates[i][IWL_TLC_MCS_PER_BW_80] = + cpu_to_le16(iwl_mld_he_mac80211_mcs_to_fw_mcs(_mcs_80)); + + /* If one side doesn't support - mark both as not supporting */ + if (_mcs_160 == IEEE80211_HE_MCS_NOT_SUPPORTED || + _tx_mcs_160 == IEEE80211_HE_MCS_NOT_SUPPORTED) { + _mcs_160 = IEEE80211_HE_MCS_NOT_SUPPORTED; + _tx_mcs_160 = IEEE80211_HE_MCS_NOT_SUPPORTED; + } + if (_mcs_160 > _tx_mcs_160) + _mcs_160 = _tx_mcs_160; + cmd->ht_rates[i][IWL_TLC_MCS_PER_BW_160] = + cpu_to_le16(iwl_mld_he_mac80211_mcs_to_fw_mcs(_mcs_160)); + } +} + +static void iwl_mld_set_eht_mcs(__le16 ht_rates[][3], + enum IWL_TLC_MCS_PER_BW bw, + u8 max_nss, u16 mcs_msk) +{ + if (max_nss >= 2) + ht_rates[IWL_TLC_NSS_2][bw] |= cpu_to_le16(mcs_msk); + + if (max_nss >= 1) + ht_rates[IWL_TLC_NSS_1][bw] |= cpu_to_le16(mcs_msk); +} + +static const +struct ieee80211_eht_mcs_nss_supp_bw * +iwl_mld_get_eht_mcs_of_bw(enum IWL_TLC_MCS_PER_BW bw, + const struct ieee80211_eht_mcs_nss_supp *eht_mcs) +{ + switch (bw) { + case IWL_TLC_MCS_PER_BW_80: + return &eht_mcs->bw._80; + case IWL_TLC_MCS_PER_BW_160: + return &eht_mcs->bw._160; + case IWL_TLC_MCS_PER_BW_320: + return &eht_mcs->bw._320; + default: + return NULL; + } +} + +static u8 iwl_mld_get_eht_max_nss(u8 rx_nss, u8 tx_nss) +{ + u8 tx = u8_get_bits(tx_nss, IEEE80211_EHT_MCS_NSS_TX); + u8 rx = u8_get_bits(rx_nss, IEEE80211_EHT_MCS_NSS_RX); + /* the max nss that can be used, + * is the min with our tx capa and the peer rx capa. + */ + return min(tx, rx); +} + +#define MAX_NSS_MCS(mcs_num, rx, tx) \ + iwl_mld_get_eht_max_nss((rx)->rx_tx_mcs ##mcs_num## _max_nss, \ + (tx)->rx_tx_mcs ##mcs_num## _max_nss) + +static void +iwl_mld_fill_eht_rates(struct ieee80211_vif *vif, + const struct ieee80211_link_sta *link_sta, + const struct ieee80211_sta_he_cap *own_he_cap, + const struct ieee80211_sta_eht_cap *own_eht_cap, + struct iwl_tlc_config_cmd_v4 *cmd) +{ + /* peer RX mcs capa */ + const struct ieee80211_eht_mcs_nss_supp *eht_rx_mcs = + &link_sta->eht_cap.eht_mcs_nss_supp; + /* our TX mcs capa */ + const struct ieee80211_eht_mcs_nss_supp *eht_tx_mcs = + &own_eht_cap->eht_mcs_nss_supp; + + enum IWL_TLC_MCS_PER_BW bw; + struct ieee80211_eht_mcs_nss_supp_20mhz_only mcs_rx_20; + struct ieee80211_eht_mcs_nss_supp_20mhz_only mcs_tx_20; + + /* peer is 20 MHz only */ + if (vif->type == NL80211_IFTYPE_AP && + !(link_sta->he_cap.he_cap_elem.phy_cap_info[0] & + IEEE80211_HE_PHY_CAP0_CHANNEL_WIDTH_SET_MASK_ALL)) { + mcs_rx_20 = eht_rx_mcs->only_20mhz; + } else { + mcs_rx_20.rx_tx_mcs7_max_nss = + eht_rx_mcs->bw._80.rx_tx_mcs9_max_nss; + mcs_rx_20.rx_tx_mcs9_max_nss = + eht_rx_mcs->bw._80.rx_tx_mcs9_max_nss; + mcs_rx_20.rx_tx_mcs11_max_nss = + eht_rx_mcs->bw._80.rx_tx_mcs11_max_nss; + mcs_rx_20.rx_tx_mcs13_max_nss = + eht_rx_mcs->bw._80.rx_tx_mcs13_max_nss; + } + + /* NIC is capable of 20 MHz only */ + if (!(own_he_cap->he_cap_elem.phy_cap_info[0] & + IEEE80211_HE_PHY_CAP0_CHANNEL_WIDTH_SET_MASK_ALL)) { + mcs_tx_20 = eht_tx_mcs->only_20mhz; + } else { + mcs_tx_20.rx_tx_mcs7_max_nss = + eht_tx_mcs->bw._80.rx_tx_mcs9_max_nss; + mcs_tx_20.rx_tx_mcs9_max_nss = + eht_tx_mcs->bw._80.rx_tx_mcs9_max_nss; + mcs_tx_20.rx_tx_mcs11_max_nss = + eht_tx_mcs->bw._80.rx_tx_mcs11_max_nss; + mcs_tx_20.rx_tx_mcs13_max_nss = + eht_tx_mcs->bw._80.rx_tx_mcs13_max_nss; + } + + /* rates for 20/40/80 MHz */ + bw = IWL_TLC_MCS_PER_BW_80; + iwl_mld_set_eht_mcs(cmd->ht_rates, bw, + MAX_NSS_MCS(7, &mcs_rx_20, &mcs_tx_20), + GENMASK(7, 0)); + iwl_mld_set_eht_mcs(cmd->ht_rates, bw, + MAX_NSS_MCS(9, &mcs_rx_20, &mcs_tx_20), + GENMASK(9, 8)); + iwl_mld_set_eht_mcs(cmd->ht_rates, bw, + MAX_NSS_MCS(11, &mcs_rx_20, &mcs_tx_20), + GENMASK(11, 10)); + iwl_mld_set_eht_mcs(cmd->ht_rates, bw, + MAX_NSS_MCS(13, &mcs_rx_20, &mcs_tx_20), + GENMASK(13, 12)); + + /* rates for 160/320 MHz */ + for (bw = IWL_TLC_MCS_PER_BW_160; bw <= IWL_TLC_MCS_PER_BW_320; bw++) { + const struct ieee80211_eht_mcs_nss_supp_bw *mcs_rx = + iwl_mld_get_eht_mcs_of_bw(bw, eht_rx_mcs); + const struct ieee80211_eht_mcs_nss_supp_bw *mcs_tx = + iwl_mld_get_eht_mcs_of_bw(bw, eht_tx_mcs); + + /* got unsupported index for bw */ + if (!mcs_rx || !mcs_tx) + continue; + + /* break out if we don't support the bandwidth */ + if (cmd->max_ch_width < (bw + IWL_TLC_MNG_CH_WIDTH_80MHZ)) + break; + + iwl_mld_set_eht_mcs(cmd->ht_rates, bw, + MAX_NSS_MCS(9, mcs_rx, mcs_tx), + GENMASK(9, 0)); + iwl_mld_set_eht_mcs(cmd->ht_rates, bw, + MAX_NSS_MCS(11, mcs_rx, mcs_tx), + GENMASK(11, 10)); + iwl_mld_set_eht_mcs(cmd->ht_rates, bw, + MAX_NSS_MCS(13, mcs_rx, mcs_tx), + GENMASK(13, 12)); + } + + /* the station support only a single receive chain */ + if (link_sta->smps_mode == IEEE80211_SMPS_STATIC || + link_sta->rx_nss < 2) + memset(cmd->ht_rates[IWL_TLC_NSS_2], 0, + sizeof(cmd->ht_rates[IWL_TLC_NSS_2])); +} + +static void +iwl_mld_fill_supp_rates(struct iwl_mld *mld, struct ieee80211_vif *vif, + struct ieee80211_link_sta *link_sta, + struct ieee80211_supported_band *sband, + const struct ieee80211_sta_he_cap *own_he_cap, + const struct ieee80211_sta_eht_cap *own_eht_cap, + struct iwl_tlc_config_cmd_v4 *cmd) +{ + int i; + u16 non_ht_rates = 0; + unsigned long rates_bitmap; + const struct ieee80211_sta_ht_cap *ht_cap = &link_sta->ht_cap; + const struct ieee80211_sta_vht_cap *vht_cap = &link_sta->vht_cap; + const struct ieee80211_sta_he_cap *he_cap = &link_sta->he_cap; + + /* non HT rates */ + rates_bitmap = link_sta->supp_rates[sband->band]; + for_each_set_bit(i, &rates_bitmap, BITS_PER_LONG) + non_ht_rates |= BIT(sband->bitrates[i].hw_value); + + cmd->non_ht_rates = cpu_to_le16(non_ht_rates); + cmd->mode = IWL_TLC_MNG_MODE_NON_HT; + + if (link_sta->eht_cap.has_eht && own_he_cap && own_eht_cap) { + cmd->mode = IWL_TLC_MNG_MODE_EHT; + iwl_mld_fill_eht_rates(vif, link_sta, own_he_cap, + own_eht_cap, cmd); + } else if (he_cap->has_he && own_he_cap) { + cmd->mode = IWL_TLC_MNG_MODE_HE; + iwl_mld_fill_he_rates(link_sta, own_he_cap, cmd); + } else if (vht_cap->vht_supported) { + cmd->mode = IWL_TLC_MNG_MODE_VHT; + iwl_mld_fill_vht_rates(link_sta, vht_cap, cmd); + } else if (ht_cap->ht_supported) { + cmd->mode = IWL_TLC_MNG_MODE_HT; + cmd->ht_rates[IWL_TLC_NSS_1][IWL_TLC_MCS_PER_BW_80] = + cpu_to_le16(ht_cap->mcs.rx_mask[0]); + + /* the station support only a single receive chain */ + if (link_sta->smps_mode == IEEE80211_SMPS_STATIC) + cmd->ht_rates[IWL_TLC_NSS_2][IWL_TLC_MCS_PER_BW_80] = + 0; + else + cmd->ht_rates[IWL_TLC_NSS_2][IWL_TLC_MCS_PER_BW_80] = + cpu_to_le16(ht_cap->mcs.rx_mask[1]); + } +} + +static void iwl_mld_send_tlc_cmd(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_link_sta *link_sta, + enum nl80211_band band) +{ + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(link_sta->sta); + struct ieee80211_supported_band *sband = mld->hw->wiphy->bands[band]; + const struct ieee80211_sta_he_cap *own_he_cap = + ieee80211_get_he_iftype_cap_vif(sband, vif); + const struct ieee80211_sta_eht_cap *own_eht_cap = + ieee80211_get_eht_iftype_cap_vif(sband, vif); + struct iwl_tlc_config_cmd_v4 cmd = { + /* For AP mode, use 20 MHz until the STA is authorized */ + .max_ch_width = mld_sta->sta_state > IEEE80211_STA_ASSOC ? + iwl_mld_fw_bw_from_sta_bw(link_sta) : + IWL_TLC_MNG_CH_WIDTH_20MHZ, + .flags = iwl_mld_get_tlc_cmd_flags(mld, vif, link_sta, + own_he_cap, own_eht_cap), + .chains = iwl_mld_get_fw_chains(mld), + .sgi_ch_width_supp = iwl_mld_get_fw_sgi(link_sta), + .max_mpdu_len = cpu_to_le16(link_sta->agg.max_rc_amsdu_len), + }; + int fw_sta_id = iwl_mld_fw_sta_id_from_link_sta(mld, link_sta); + int ret; + + if (fw_sta_id < 0) + return; + + cmd.sta_id = fw_sta_id; + + iwl_mld_fill_supp_rates(mld, vif, link_sta, sband, + own_he_cap, own_eht_cap, + &cmd); + + IWL_DEBUG_RATE(mld, + "TLC CONFIG CMD, sta_id=%d, max_ch_width=%d, mode=%d\n", + cmd.sta_id, cmd.max_ch_width, cmd.mode); + + /* Send async since this can be called within a RCU-read section */ + ret = iwl_mld_send_cmd_with_flags_pdu(mld, WIDE_ID(DATA_PATH_GROUP, + TLC_MNG_CONFIG_CMD), + CMD_ASYNC, &cmd); + if (ret) + IWL_ERR(mld, "Failed to send TLC cmd (%d)\n", ret); +} + +static void iwl_mld_recalc_amsdu_len(struct iwl_mld *mld, + struct ieee80211_link_sta *link_sta) +{ + const struct ieee80211_sta_ht_cap *ht_cap = &link_sta->ht_cap; + + /* For EHT, HE and VHT - we can use the value as it was calculated by + * mac80211. + */ + if (link_sta->eht_cap.has_eht || link_sta->he_cap.has_he || + link_sta->vht_cap.vht_supported) + goto recalc; + + /* But for HT, mac80211 doesn't enforce to 4095, so force it here */ + if (ht_cap->ht_supported && ht_cap->cap & IEEE80211_HT_CAP_MAX_AMSDU) + /* Agg is offloaded, so we need to assume that agg are enabled + * and max mpdu in ampdu is 4095 (spec 802.11-2016 9.3.2.1) + */ + link_sta->agg.max_amsdu_len = IEEE80211_MAX_MPDU_LEN_HT_BA; + +recalc: + link_sta->agg.max_rc_amsdu_len = link_sta->agg.max_amsdu_len; + ieee80211_sta_recalc_aggregates(link_sta->sta); +} + +int iwl_mld_send_tlc_dhc(struct iwl_mld *mld, u8 sta_id, u32 type, u32 data) +{ + struct { + struct iwl_dhc_cmd dhc; + struct iwl_dhc_tlc_cmd tlc; + } __packed cmd = { + .tlc.sta_id = sta_id, + .tlc.type = cpu_to_le32(type), + .tlc.data[0] = cpu_to_le32(data), + .dhc.length = cpu_to_le32(sizeof(cmd.tlc) >> 2), + .dhc.index_and_mask = + cpu_to_le32(DHC_TABLE_INTEGRATION | DHC_TARGET_UMAC | + DHC_INTEGRATION_TLC_DEBUG_CONFIG), + }; + int ret; + + ret = iwl_mld_send_cmd_with_flags_pdu(mld, + WIDE_ID(IWL_ALWAYS_LONG_GROUP, + DEBUG_HOST_COMMAND), + CMD_ASYNC, &cmd); + IWL_DEBUG_RATE(mld, "sta_id %d, type: 0x%X, value: 0x%X, ret%d\n", + sta_id, type, data, ret); + return ret; +} + +void iwl_mld_config_tlc_link(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link_conf, + struct ieee80211_link_sta *link_sta) +{ + enum nl80211_band band; + + if (WARN_ON_ONCE(!link_conf->chanreq.oper.chan)) + return; + + band = link_conf->chanreq.oper.chan->band; + + iwl_mld_recalc_amsdu_len(mld, link_sta); + + iwl_mld_send_tlc_cmd(mld, vif, link_sta, band); +} + +void iwl_mld_config_tlc(struct iwl_mld *mld, struct ieee80211_vif *vif, + struct ieee80211_sta *sta) +{ + struct ieee80211_bss_conf *link; + int link_id; + + lockdep_assert_wiphy(mld->wiphy); + + for_each_vif_active_link(vif, link, link_id) { + struct ieee80211_link_sta *link_sta = + link_sta_dereference_check(sta, link_id); + + if (!link || !link_sta) + continue; + + iwl_mld_config_tlc_link(mld, vif, link, link_sta); + } +} + +static u16 +iwl_mld_get_amsdu_size_of_tid(struct iwl_mld *mld, + struct ieee80211_link_sta *link_sta, + unsigned int tid) +{ + struct ieee80211_sta *sta = link_sta->sta; + struct ieee80211_vif *vif = iwl_mld_sta_from_mac80211(sta)->vif; + const u8 tid_to_mac80211_ac[] = { + IEEE80211_AC_BE, + IEEE80211_AC_BK, + IEEE80211_AC_BK, + IEEE80211_AC_BE, + IEEE80211_AC_VI, + IEEE80211_AC_VI, + IEEE80211_AC_VO, + IEEE80211_AC_VO, + }; + unsigned int result = link_sta->agg.max_rc_amsdu_len; + u8 ac, txf, lmac; + + lockdep_assert_wiphy(mld->wiphy); + + /* Don't send an AMSDU that will be longer than the TXF. + * Add a security margin of 256 for the TX command + headers. + * We also want to have the start of the next packet inside the + * fifo to be able to send bursts. + */ + + if (WARN_ON(tid >= ARRAY_SIZE(tid_to_mac80211_ac))) + return 0; + + ac = tid_to_mac80211_ac[tid]; + + /* For HE redirect to trigger based fifos */ + if (link_sta->he_cap.has_he) + ac += 4; + + txf = iwl_mld_mac80211_ac_to_fw_tx_fifo(ac); + + /* Only one link: take the lmac according to the band */ + if (hweight16(sta->valid_links) <= 1) { + enum nl80211_band band; + struct ieee80211_bss_conf *link = + wiphy_dereference(mld->wiphy, + vif->link_conf[link_sta->link_id]); + + if (WARN_ON(!link || !link->chanreq.oper.chan)) + band = NL80211_BAND_2GHZ; + else + band = link->chanreq.oper.chan->band; + lmac = iwl_mld_get_lmac_id(mld, band); + + /* More than one link but with 2 lmacs: take the minimum */ + } else if (fw_has_capa(&mld->fw->ucode_capa, + IWL_UCODE_TLV_CAPA_CDB_SUPPORT)) { + lmac = IWL_LMAC_5G_INDEX; + result = min_t(unsigned int, result, + mld->fwrt.smem_cfg.lmac[lmac].txfifo_size[txf] - 256); + lmac = IWL_LMAC_24G_INDEX; + /* More than one link but only one lmac */ + } else { + lmac = IWL_LMAC_24G_INDEX; + } + + return min_t(unsigned int, result, + mld->fwrt.smem_cfg.lmac[lmac].txfifo_size[txf] - 256); +} + +void iwl_mld_handle_tlc_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + struct iwl_tlc_update_notif *notif = (void *)pkt->data; + struct ieee80211_link_sta *link_sta; + u32 flags = le32_to_cpu(notif->flags); + u32 enabled; + u16 size; + + if (IWL_FW_CHECK(mld, notif->sta_id >= mld->fw->ucode_capa.num_stations, + "Invalid sta id (%d) in TLC notification\n", + notif->sta_id)) + return; + + link_sta = wiphy_dereference(mld->wiphy, + mld->fw_id_to_link_sta[notif->sta_id]); + + if (WARN(IS_ERR_OR_NULL(link_sta), + "link_sta of sta id (%d) doesn't exist\n", notif->sta_id)) + return; + + if (flags & IWL_TLC_NOTIF_FLAG_RATE) { + struct iwl_mld_link_sta *mld_link_sta = + iwl_mld_link_sta_from_mac80211(link_sta); + char pretty_rate[100]; + + if (WARN_ON(!mld_link_sta)) + return; + + mld_link_sta->last_rate_n_flags = le32_to_cpu(notif->rate); + + rs_pretty_print_rate(pretty_rate, sizeof(pretty_rate), + mld_link_sta->last_rate_n_flags); + IWL_DEBUG_RATE(mld, "TLC notif: new rate = %s\n", pretty_rate); + } + + /* We are done processing the notif */ + if (!(flags & IWL_TLC_NOTIF_FLAG_AMSDU)) + return; + + enabled = le32_to_cpu(notif->amsdu_enabled); + size = le32_to_cpu(notif->amsdu_size); + + if (size < 2000) { + size = 0; + enabled = 0; + } + + if (IWL_FW_CHECK(mld, size > link_sta->agg.max_amsdu_len, + "Invalid AMSDU len in TLC notif: %d (Max AMSDU len: %d)\n", + size, link_sta->agg.max_amsdu_len)) + return; + + link_sta->agg.max_rc_amsdu_len = size; + + for (int i = 0; i < IWL_MAX_TID_COUNT; i++) { + if (enabled & BIT(i)) + link_sta->agg.max_tid_amsdu_len[i] = + iwl_mld_get_amsdu_size_of_tid(mld, link_sta, i); + else + link_sta->agg.max_tid_amsdu_len[i] = 0; + } + + ieee80211_sta_recalc_aggregates(link_sta->sta); + + IWL_DEBUG_RATE(mld, + "AMSDU update. AMSDU size: %d, AMSDU selected size: %d, AMSDU TID bitmap 0x%X\n", + le32_to_cpu(notif->amsdu_size), size, enabled); +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/tlc.h b/drivers/net/wireless/intel/iwlwifi/mld/tlc.h new file mode 100644 index 0000000000000..c32f42e8840bb --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/tlc.h @@ -0,0 +1,23 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024 Intel Corporation + */ +#ifndef __iwl_mld_tlc_h__ +#define __iwl_mld_tlc_h__ + +#include "mld.h" + +void iwl_mld_config_tlc_link(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_bss_conf *link_conf, + struct ieee80211_link_sta *link_sta); + +void iwl_mld_config_tlc(struct iwl_mld *mld, struct ieee80211_vif *vif, + struct ieee80211_sta *sta); + +void iwl_mld_handle_tlc_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt); + +int iwl_mld_send_tlc_dhc(struct iwl_mld *mld, u8 sta_id, u32 type, u32 data); + +#endif /* __iwl_mld_tlc_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/mld/tx.c b/drivers/net/wireless/intel/iwlwifi/mld/tx.c new file mode 100644 index 0000000000000..543abe72e4652 --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/tx.c @@ -0,0 +1,1374 @@ +// SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause +/* + * Copyright (C) 2024 - 2025 Intel Corporation + */ +#include + +#include "tx.h" +#include "sta.h" +#include "hcmd.h" +#include "iwl-utils.h" +#include "iface.h" + +#include "fw/dbg.h" + +#include "fw/api/tx.h" +#include "fw/api/rs.h" +#include "fw/api/txq.h" +#include "fw/api/datapath.h" +#include "fw/api/time-event.h" + +#define MAX_ANT_NUM 2 + +/* Toggles between TX antennas. Receives the bitmask of valid TX antennas and + * the *index* used for the last TX, and returns the next valid *index* to use. + * In order to set it in the tx_cmd, must do BIT(idx). + */ +static u8 iwl_mld_next_ant(u8 valid, u8 last_idx) +{ + u8 index = last_idx; + + for (int i = 0; i < MAX_ANT_NUM; i++) { + index = (index + 1) % MAX_ANT_NUM; + if (valid & BIT(index)) + return index; + } + + WARN_ONCE(1, "Failed to toggle between antennas 0x%x", valid); + + return last_idx; +} + +void iwl_mld_toggle_tx_ant(struct iwl_mld *mld, u8 *ant) +{ + *ant = iwl_mld_next_ant(iwl_mld_get_valid_tx_ant(mld), *ant); +} + +static int +iwl_mld_get_queue_size(struct iwl_mld *mld, struct ieee80211_txq *txq) +{ + struct ieee80211_sta *sta = txq->sta; + struct ieee80211_link_sta *link_sta; + unsigned int link_id; + int max_size = IWL_DEFAULT_QUEUE_SIZE; + + lockdep_assert_wiphy(mld->wiphy); + + for_each_sta_active_link(txq->vif, sta, link_sta, link_id) { + if (link_sta->eht_cap.has_eht) { + max_size = IWL_DEFAULT_QUEUE_SIZE_EHT; + break; + } + + if (link_sta->he_cap.has_he) + max_size = IWL_DEFAULT_QUEUE_SIZE_HE; + } + + return max_size; +} + +static int iwl_mld_allocate_txq(struct iwl_mld *mld, struct ieee80211_txq *txq) +{ + u8 tid = txq->tid == IEEE80211_NUM_TIDS ? IWL_MGMT_TID : txq->tid; + u32 fw_sta_mask = iwl_mld_fw_sta_id_mask(mld, txq->sta); + /* We can't know when the station is asleep or awake, so we + * must disable the queue hang detection. + */ + unsigned int watchdog_timeout = txq->vif->type == NL80211_IFTYPE_AP ? + IWL_WATCHDOG_DISABLED : + mld->trans->trans_cfg->base_params->wd_timeout; + int queue, size; + + lockdep_assert_wiphy(mld->wiphy); + + if (tid == IWL_MGMT_TID) + size = max_t(u32, IWL_MGMT_QUEUE_SIZE, + mld->trans->cfg->min_txq_size); + else + size = iwl_mld_get_queue_size(mld, txq); + + queue = iwl_trans_txq_alloc(mld->trans, 0, fw_sta_mask, tid, size, + watchdog_timeout); + + if (queue >= 0) + IWL_DEBUG_TX_QUEUES(mld, + "Enabling TXQ #%d for sta mask 0x%x tid %d\n", + queue, fw_sta_mask, tid); + return queue; +} + +static int iwl_mld_add_txq(struct iwl_mld *mld, struct ieee80211_txq *txq) +{ + struct iwl_mld_txq *mld_txq = iwl_mld_txq_from_mac80211(txq); + int id; + + lockdep_assert_wiphy(mld->wiphy); + + /* This will alse send the SCD_QUEUE_CONFIG_CMD */ + id = iwl_mld_allocate_txq(mld, txq); + if (id < 0) + return id; + + mld_txq->fw_id = id; + mld_txq->status.allocated = true; + + rcu_assign_pointer(mld->fw_id_to_txq[id], txq); + + return 0; +} + +void iwl_mld_add_txq_list(struct iwl_mld *mld) +{ + lockdep_assert_wiphy(mld->wiphy); + + while (!list_empty(&mld->txqs_to_add)) { + struct ieee80211_txq *txq; + struct iwl_mld_txq *mld_txq = + list_first_entry(&mld->txqs_to_add, struct iwl_mld_txq, + list); + int failed; + + txq = container_of((void *)mld_txq, struct ieee80211_txq, + drv_priv); + + failed = iwl_mld_add_txq(mld, txq); + + local_bh_disable(); + spin_lock(&mld->add_txqs_lock); + list_del_init(&mld_txq->list); + spin_unlock(&mld->add_txqs_lock); + /* If the queue allocation failed, we can't transmit. Leave the + * frames on the txq, maybe the attempt to allocate the queue + * will succeed. + */ + if (!failed) + iwl_mld_tx_from_txq(mld, txq); + local_bh_enable(); + } +} + +void iwl_mld_add_txqs_wk(struct wiphy *wiphy, struct wiphy_work *wk) +{ + struct iwl_mld *mld = container_of(wk, struct iwl_mld, + add_txqs_wk); + + /* will reschedule to run after restart */ + if (mld->fw_status.in_hw_restart) + return; + + iwl_mld_add_txq_list(mld); +} + +void +iwl_mld_free_txq(struct iwl_mld *mld, u32 fw_sta_mask, u32 tid, u32 queue_id) +{ + struct iwl_scd_queue_cfg_cmd remove_cmd = { + .operation = cpu_to_le32(IWL_SCD_QUEUE_REMOVE), + .u.remove.tid = cpu_to_le32(tid), + .u.remove.sta_mask = cpu_to_le32(fw_sta_mask), + }; + + iwl_mld_send_cmd_pdu(mld, + WIDE_ID(DATA_PATH_GROUP, SCD_QUEUE_CONFIG_CMD), + &remove_cmd); + + iwl_trans_txq_free(mld->trans, queue_id); +} + +void iwl_mld_remove_txq(struct iwl_mld *mld, struct ieee80211_txq *txq) +{ + struct iwl_mld_txq *mld_txq = iwl_mld_txq_from_mac80211(txq); + u32 sta_msk, tid; + + lockdep_assert_wiphy(mld->wiphy); + + spin_lock_bh(&mld->add_txqs_lock); + if (!list_empty(&mld_txq->list)) + list_del_init(&mld_txq->list); + spin_unlock_bh(&mld->add_txqs_lock); + + if (!mld_txq->status.allocated || + WARN_ON(mld_txq->fw_id >= ARRAY_SIZE(mld->fw_id_to_txq))) + return; + + sta_msk = iwl_mld_fw_sta_id_mask(mld, txq->sta); + + tid = txq->tid == IEEE80211_NUM_TIDS ? IWL_MGMT_TID : + txq->tid; + + iwl_mld_free_txq(mld, sta_msk, tid, mld_txq->fw_id); + + RCU_INIT_POINTER(mld->fw_id_to_txq[mld_txq->fw_id], NULL); + mld_txq->status.allocated = false; +} + +#define OPT_HDR(type, skb, off) \ + (type *)(skb_network_header(skb) + (off)) + +static __le32 +iwl_mld_get_offload_assist(struct sk_buff *skb, bool amsdu) +{ + struct ieee80211_hdr *hdr = (void *)skb->data; + u16 mh_len = ieee80211_hdrlen(hdr->frame_control); + u16 offload_assist = 0; +#if IS_ENABLED(CONFIG_INET) + u8 protocol = 0; + + /* Do not compute checksum if already computed */ + if (skb->ip_summed != CHECKSUM_PARTIAL) + goto out; + + /* We do not expect to be requested to csum stuff we do not support */ + + /* TBD: do we also need to check + * !(mvm->hw->netdev_features & IWL_TX_CSUM_NETIF_FLAGS) now that all + * the devices we support has this flags? + */ + if (WARN_ONCE(skb->protocol != htons(ETH_P_IP) && + skb->protocol != htons(ETH_P_IPV6), + "No support for requested checksum\n")) { + skb_checksum_help(skb); + goto out; + } + + if (skb->protocol == htons(ETH_P_IP)) { + protocol = ip_hdr(skb)->protocol; + } else { +#if IS_ENABLED(CONFIG_IPV6) + struct ipv6hdr *ipv6h = + (struct ipv6hdr *)skb_network_header(skb); + unsigned int off = sizeof(*ipv6h); + + protocol = ipv6h->nexthdr; + while (protocol != NEXTHDR_NONE && ipv6_ext_hdr(protocol)) { + struct ipv6_opt_hdr *hp; + + /* only supported extension headers */ + if (protocol != NEXTHDR_ROUTING && + protocol != NEXTHDR_HOP && + protocol != NEXTHDR_DEST) { + skb_checksum_help(skb); + goto out; + } + + hp = OPT_HDR(struct ipv6_opt_hdr, skb, off); + protocol = hp->nexthdr; + off += ipv6_optlen(hp); + } + /* if we get here - protocol now should be TCP/UDP */ +#endif + } + + if (protocol != IPPROTO_TCP && protocol != IPPROTO_UDP) { + WARN_ON_ONCE(1); + skb_checksum_help(skb); + goto out; + } + + /* enable L4 csum */ + offload_assist |= BIT(TX_CMD_OFFLD_L4_EN); + + /* Set offset to IP header (snap). + * We don't support tunneling so no need to take care of inner header. + * Size is in words. + */ + offload_assist |= (4 << TX_CMD_OFFLD_IP_HDR); + + /* Do IPv4 csum for AMSDU only (no IP csum for Ipv6) */ + if (skb->protocol == htons(ETH_P_IP) && amsdu) { + ip_hdr(skb)->check = 0; + offload_assist |= BIT(TX_CMD_OFFLD_L3_EN); + } + + /* reset UDP/TCP header csum */ + if (protocol == IPPROTO_TCP) + tcp_hdr(skb)->check = 0; + else + udp_hdr(skb)->check = 0; + +out: +#endif + mh_len /= 2; + offload_assist |= mh_len << TX_CMD_OFFLD_MH_SIZE; + + if (amsdu) + offload_assist |= BIT(TX_CMD_OFFLD_AMSDU); + else if (ieee80211_hdrlen(hdr->frame_control) % 4) + /* padding is inserted later in transport */ + offload_assist |= BIT(TX_CMD_OFFLD_PAD); + + return cpu_to_le32(offload_assist); +} + +static void iwl_mld_get_basic_rates_and_band(struct iwl_mld *mld, + struct ieee80211_vif *vif, + struct ieee80211_tx_info *info, + unsigned long *basic_rates, + u8 *band) +{ + u32 link_id = u32_get_bits(info->control.flags, + IEEE80211_TX_CTRL_MLO_LINK); + + *basic_rates = vif->bss_conf.basic_rates; + *band = info->band; + + if (link_id == IEEE80211_LINK_UNSPECIFIED && + ieee80211_vif_is_mld(vif)) { + /* shouldn't do this when >1 link is active */ + WARN_ON(hweight16(vif->active_links) != 1); + link_id = __ffs(vif->active_links); + } + + if (link_id < IEEE80211_LINK_UNSPECIFIED) { + struct ieee80211_bss_conf *link_conf; + + rcu_read_lock(); + link_conf = rcu_dereference(vif->link_conf[link_id]); + if (link_conf) { + *basic_rates = link_conf->basic_rates; + if (link_conf->chanreq.oper.chan) + *band = link_conf->chanreq.oper.chan->band; + } + rcu_read_unlock(); + } +} + +u8 iwl_mld_get_lowest_rate(struct iwl_mld *mld, + struct ieee80211_tx_info *info, + struct ieee80211_vif *vif) +{ + struct ieee80211_supported_band *sband; + u16 lowest_cck = IWL_RATE_COUNT, lowest_ofdm = IWL_RATE_COUNT; + unsigned long basic_rates; + u8 band, rate; + u32 i; + + iwl_mld_get_basic_rates_and_band(mld, vif, info, &basic_rates, &band); + + sband = mld->hw->wiphy->bands[band]; + for_each_set_bit(i, &basic_rates, BITS_PER_LONG) { + u16 hw = sband->bitrates[i].hw_value; + + if (hw >= IWL_FIRST_OFDM_RATE) { + if (lowest_ofdm > hw) + lowest_ofdm = hw; + } else if (lowest_cck > hw) { + lowest_cck = hw; + } + } + + if (band == NL80211_BAND_2GHZ && !vif->p2p && + vif->type != NL80211_IFTYPE_P2P_DEVICE && + !(info->flags & IEEE80211_TX_CTL_NO_CCK_RATE)) { + if (lowest_cck != IWL_RATE_COUNT) + rate = lowest_cck; + else if (lowest_ofdm != IWL_RATE_COUNT) + rate = lowest_ofdm; + else + rate = IWL_FIRST_CCK_RATE; + } else if (lowest_ofdm != IWL_RATE_COUNT) { + rate = lowest_ofdm; + } else { + rate = IWL_FIRST_OFDM_RATE; + } + + return rate; +} + +static u32 iwl_mld_mac80211_rate_idx_to_fw(struct iwl_mld *mld, + struct ieee80211_tx_info *info, + int rate_idx) +{ + u32 rate_flags = 0; + u8 rate_plcp; + + /* if the rate isn't a well known legacy rate, take the lowest one */ + if (rate_idx < 0 || rate_idx >= IWL_RATE_COUNT_LEGACY) + rate_idx = iwl_mld_get_lowest_rate(mld, info, + info->control.vif); + + WARN_ON_ONCE(rate_idx < 0); + + /* Set CCK or OFDM flag */ + if (rate_idx <= IWL_LAST_CCK_RATE) + rate_flags |= RATE_MCS_CCK_MSK; + else + rate_flags |= RATE_MCS_LEGACY_OFDM_MSK; + + /* Legacy rates are indexed: + * 0 - 3 for CCK and 0 - 7 for OFDM + */ + rate_plcp = (rate_idx >= IWL_FIRST_OFDM_RATE ? + rate_idx - IWL_FIRST_OFDM_RATE : rate_idx); + + return (u32)rate_plcp | rate_flags; +} + +static u32 iwl_mld_get_tx_ant(struct iwl_mld *mld, + struct ieee80211_tx_info *info, + struct ieee80211_sta *sta, __le16 fc) +{ + if (sta && ieee80211_is_data(fc)) { + struct iwl_mld_sta *mld_sta = iwl_mld_sta_from_mac80211(sta); + + return BIT(mld_sta->data_tx_ant) << RATE_MCS_ANT_POS; + } + + return BIT(mld->mgmt_tx_ant) << RATE_MCS_ANT_POS; +} + +static u32 iwl_mld_get_inject_tx_rate(struct iwl_mld *mld, + struct ieee80211_tx_info *info, + struct ieee80211_sta *sta, + __le16 fc) +{ + struct ieee80211_tx_rate *rate = &info->control.rates[0]; + u32 result; + + /* we only care about legacy/HT/VHT so far, so we can + * build in v1 and use iwl_new_rate_from_v1() + * FIXME: in newer devices we only support the new rates, build + * the rate_n_flags in the new format here instead of using v1 and + * converting it. + */ + + if (rate->flags & IEEE80211_TX_RC_VHT_MCS) { + u8 mcs = ieee80211_rate_get_vht_mcs(rate); + u8 nss = ieee80211_rate_get_vht_nss(rate); + + result = RATE_MCS_VHT_MSK_V1; + result |= u32_encode_bits(mcs, RATE_VHT_MCS_RATE_CODE_MSK); + result |= u32_encode_bits(nss, RATE_MCS_NSS_MSK); + + if (rate->flags & IEEE80211_TX_RC_SHORT_GI) + result |= RATE_MCS_SGI_MSK_V1; + + if (rate->flags & IEEE80211_TX_RC_40_MHZ_WIDTH) + result |= u32_encode_bits(1, RATE_MCS_CHAN_WIDTH_MSK_V1); + else if (rate->flags & IEEE80211_TX_RC_80_MHZ_WIDTH) + result |= u32_encode_bits(2, RATE_MCS_CHAN_WIDTH_MSK_V1); + else if (rate->flags & IEEE80211_TX_RC_160_MHZ_WIDTH) + result |= u32_encode_bits(3, RATE_MCS_CHAN_WIDTH_MSK_V1); + + result = iwl_new_rate_from_v1(result); + } else if (rate->flags & IEEE80211_TX_RC_MCS) { + result = RATE_MCS_HT_MSK_V1; + result |= u32_encode_bits(rate->idx, + RATE_HT_MCS_RATE_CODE_MSK_V1 | + RATE_HT_MCS_NSS_MSK_V1); + if (rate->flags & IEEE80211_TX_RC_SHORT_GI) + result |= RATE_MCS_SGI_MSK_V1; + if (rate->flags & IEEE80211_TX_RC_40_MHZ_WIDTH) + result |= u32_encode_bits(1, RATE_MCS_CHAN_WIDTH_MSK_V1); + if (info->flags & IEEE80211_TX_CTL_LDPC) + result |= RATE_MCS_LDPC_MSK_V1; + if (u32_get_bits(info->flags, IEEE80211_TX_CTL_STBC)) + result |= RATE_MCS_STBC_MSK; + + result = iwl_new_rate_from_v1(result); + } else { + result = iwl_mld_mac80211_rate_idx_to_fw(mld, info, rate->idx); + } + + if (info->control.antennas) + result |= u32_encode_bits(info->control.antennas, + RATE_MCS_ANT_AB_MSK); + else + result |= iwl_mld_get_tx_ant(mld, info, sta, fc); + + return result; +} + +static u32 iwl_mld_get_tx_rate_n_flags(struct iwl_mld *mld, + struct ieee80211_tx_info *info, + struct ieee80211_sta *sta, __le16 fc) +{ + if (unlikely(info->control.flags & IEEE80211_TX_CTRL_RATE_INJECT)) + return iwl_mld_get_inject_tx_rate(mld, info, sta, fc); + + return iwl_mld_mac80211_rate_idx_to_fw(mld, info, -1) | + iwl_mld_get_tx_ant(mld, info, sta, fc); +} + +static void +iwl_mld_fill_tx_cmd_hdr(struct iwl_tx_cmd_gen3 *tx_cmd, + struct sk_buff *skb, bool amsdu) +{ + struct ieee80211_tx_info *info = IEEE80211_SKB_CB(skb); + struct ieee80211_hdr *hdr = (void *)skb->data; + struct ieee80211_vif *vif; + + /* Copy MAC header from skb into command buffer */ + memcpy(tx_cmd->hdr, hdr, ieee80211_hdrlen(hdr->frame_control)); + + if (!amsdu || !skb_is_gso(skb)) + return; + + /* As described in IEEE sta 802.11-2020, table 9-30 (Address + * field contents), A-MSDU address 3 should contain the BSSID + * address. + * + * In TSO, the skb header address 3 contains the original address 3 to + * correctly create all the A-MSDU subframes headers from it. + * Override now the address 3 in the command header with the BSSID. + * + * Note: we fill in the MLD address, but the firmware will do the + * necessary translation to link address after encryption. + */ + vif = info->control.vif; + switch (vif->type) { + case NL80211_IFTYPE_STATION: + ether_addr_copy(tx_cmd->hdr->addr3, vif->cfg.ap_addr); + break; + case NL80211_IFTYPE_AP: + ether_addr_copy(tx_cmd->hdr->addr3, vif->addr); + break; + default: + break; + } +} + +static void +iwl_mld_fill_tx_cmd(struct iwl_mld *mld, struct sk_buff *skb, + struct iwl_device_tx_cmd *dev_tx_cmd, + struct ieee80211_sta *sta) +{ + struct ieee80211_tx_info *info = IEEE80211_SKB_CB(skb); + struct ieee80211_hdr *hdr = (void *)skb->data; + struct iwl_mld_sta *mld_sta = sta ? iwl_mld_sta_from_mac80211(sta) : + NULL; + struct iwl_tx_cmd_gen3 *tx_cmd; + bool amsdu = ieee80211_is_data_qos(hdr->frame_control) && + (*ieee80211_get_qos_ctl(hdr) & + IEEE80211_QOS_CTL_A_MSDU_PRESENT); + u32 rate_n_flags = 0; + u16 flags = 0; + + dev_tx_cmd->hdr.cmd = TX_CMD; + + if (!info->control.hw_key) + flags |= IWL_TX_FLAGS_ENCRYPT_DIS; + + /* For data and mgmt packets rate info comes from the fw. + * Only set rate/antenna for injected frames with fixed rate, or + * when no sta is given. + */ + if (unlikely(!sta || + info->control.flags & IEEE80211_TX_CTRL_RATE_INJECT)) { + flags |= IWL_TX_FLAGS_CMD_RATE; + rate_n_flags = iwl_mld_get_tx_rate_n_flags(mld, info, sta, + hdr->frame_control); + } else if (!ieee80211_is_data(hdr->frame_control) || + (mld_sta && + mld_sta->sta_state < IEEE80211_STA_AUTHORIZED)) { + /* These are important frames */ + flags |= IWL_TX_FLAGS_HIGH_PRI; + } + + tx_cmd = (void *)dev_tx_cmd->payload; + + iwl_mld_fill_tx_cmd_hdr(tx_cmd, skb, amsdu); + + tx_cmd->offload_assist = iwl_mld_get_offload_assist(skb, amsdu); + + /* Total # bytes to be transmitted */ + tx_cmd->len = cpu_to_le16((u16)skb->len); + + tx_cmd->flags = cpu_to_le16(flags); + + tx_cmd->rate_n_flags = cpu_to_le32(rate_n_flags); +} + +/* Caller of this need to check that info->control.vif is not NULL */ +static struct iwl_mld_link * +iwl_mld_get_link_from_tx_info(struct ieee80211_tx_info *info) +{ + struct iwl_mld_vif *mld_vif = + iwl_mld_vif_from_mac80211(info->control.vif); + u32 link_id = u32_get_bits(info->control.flags, + IEEE80211_TX_CTRL_MLO_LINK); + + if (link_id == IEEE80211_LINK_UNSPECIFIED) { + if (info->control.vif->active_links) + link_id = ffs(info->control.vif->active_links) - 1; + else + link_id = 0; + } + + return rcu_dereference(mld_vif->link[link_id]); +} + +static int +iwl_mld_get_tx_queue_id(struct iwl_mld *mld, struct ieee80211_txq *txq, + struct sk_buff *skb) +{ + struct ieee80211_tx_info *info = IEEE80211_SKB_CB(skb); + struct ieee80211_hdr *hdr = (void *)skb->data; + __le16 fc = hdr->frame_control; + struct iwl_mld_vif *mld_vif; + struct iwl_mld_link *link; + + if (txq && txq->sta) + return iwl_mld_txq_from_mac80211(txq)->fw_id; + + if (!info->control.vif) + return IWL_MLD_INVALID_QUEUE; + + switch (info->control.vif->type) { + case NL80211_IFTYPE_AP: + case NL80211_IFTYPE_ADHOC: + link = iwl_mld_get_link_from_tx_info(info); + + if (WARN_ON(!link)) + break; + + /* ucast disassociate/deauth frames without a station might + * happen, especially with reason 7 ("Class 3 frame received + * from nonassociated STA"). + */ + if (ieee80211_is_mgmt(fc) && + (!ieee80211_is_bufferable_mmpdu(skb) || + ieee80211_is_deauth(fc) || ieee80211_is_disassoc(fc))) + return link->bcast_sta.queue_id; + + if (is_multicast_ether_addr(hdr->addr1) && + !ieee80211_has_order(fc)) + return link->mcast_sta.queue_id; + + WARN_ONCE(info->control.vif->type != NL80211_IFTYPE_ADHOC, + "Couldn't find a TXQ. fc=0x%02x", le16_to_cpu(fc)); + return link->bcast_sta.queue_id; + case NL80211_IFTYPE_P2P_DEVICE: + mld_vif = iwl_mld_vif_from_mac80211(info->control.vif); + + if (mld_vif->roc_activity == ROC_NUM_ACTIVITIES) { + IWL_DEBUG_DROP(mld, "Drop tx outside ROC\n"); + return IWL_MLD_INVALID_DROP_TX; + } + + WARN_ON(!ieee80211_is_mgmt(fc)); + + return mld_vif->deflink.aux_sta.queue_id; + default: + /* TODO: consider monitor (task=monitor) */ + WARN_ONCE(1, "Unsupported vif type\n"); + break; + } + + return IWL_MLD_INVALID_QUEUE; +} + +static void iwl_mld_probe_resp_set_noa(struct iwl_mld *mld, + struct sk_buff *skb) +{ + struct ieee80211_tx_info *info = IEEE80211_SKB_CB(skb); + struct iwl_mld_link *mld_link = + &iwl_mld_vif_from_mac80211(info->control.vif)->deflink; + struct iwl_probe_resp_data *resp_data; + u8 *pos; + + if (!info->control.vif->p2p) + return; + + rcu_read_lock(); + + resp_data = rcu_dereference(mld_link->probe_resp_data); + if (!resp_data) + goto out; + + if (!resp_data->notif.noa_active) + goto out; + + if (skb_tailroom(skb) < resp_data->noa_len) { + if (pskb_expand_head(skb, 0, resp_data->noa_len, GFP_ATOMIC)) { + IWL_ERR(mld, + "Failed to reallocate probe resp\n"); + goto out; + } + } + + pos = skb_put(skb, resp_data->noa_len); + + *pos++ = WLAN_EID_VENDOR_SPECIFIC; + /* Set length of IE body (not including ID and length itself) */ + *pos++ = resp_data->noa_len - 2; + *pos++ = (WLAN_OUI_WFA >> 16) & 0xff; + *pos++ = (WLAN_OUI_WFA >> 8) & 0xff; + *pos++ = WLAN_OUI_WFA & 0xff; + *pos++ = WLAN_OUI_TYPE_WFA_P2P; + + memcpy(pos, &resp_data->notif.noa_attr, + resp_data->noa_len - sizeof(struct ieee80211_vendor_ie)); + +out: + rcu_read_unlock(); +} + +/* This function must be called with BHs disabled */ +static int iwl_mld_tx_mpdu(struct iwl_mld *mld, struct sk_buff *skb, + struct ieee80211_txq *txq) +{ + struct ieee80211_hdr *hdr = (void *)skb->data; + struct ieee80211_tx_info *info = IEEE80211_SKB_CB(skb); + struct ieee80211_sta *sta = txq ? txq->sta : NULL; + struct iwl_device_tx_cmd *dev_tx_cmd; + int queue = iwl_mld_get_tx_queue_id(mld, txq, skb); + u8 tid = IWL_MAX_TID_COUNT; + + if (WARN_ONCE(queue == IWL_MLD_INVALID_QUEUE, "Invalid TX Queue id") || + queue == IWL_MLD_INVALID_DROP_TX) + return -1; + + if (unlikely(ieee80211_is_any_nullfunc(hdr->frame_control))) + return -1; + + dev_tx_cmd = iwl_trans_alloc_tx_cmd(mld->trans); + if (unlikely(!dev_tx_cmd)) + return -1; + + if (unlikely(ieee80211_is_probe_resp(hdr->frame_control))) { + if (IWL_MLD_NON_TRANSMITTING_AP) + return -1; + + iwl_mld_probe_resp_set_noa(mld, skb); + } + + iwl_mld_fill_tx_cmd(mld, skb, dev_tx_cmd, sta); + + if (ieee80211_is_data(hdr->frame_control)) { + if (ieee80211_is_data_qos(hdr->frame_control)) + tid = ieee80211_get_tid(hdr); + else + tid = IWL_TID_NON_QOS; + } + + IWL_DEBUG_TX(mld, "TX TID:%d from Q:%d len %d\n", + tid, queue, skb->len); + + /* From now on, we cannot access info->control */ + memset(&info->status, 0, sizeof(info->status)); + memset(info->driver_data, 0, sizeof(info->driver_data)); + + info->driver_data[1] = dev_tx_cmd; + + if (iwl_trans_tx(mld->trans, skb, dev_tx_cmd, queue)) + goto err; + + /* Update low-latency counter when a packet is queued instead + * of after TX, it makes sense for early low-latency detection + */ + if (sta) + iwl_mld_low_latency_update_counters(mld, hdr, sta, 0); + + return 0; + +err: + iwl_trans_free_tx_cmd(mld->trans, dev_tx_cmd); + IWL_DEBUG_TX(mld, "TX from Q:%d dropped\n", queue); + return -1; +} + +#ifdef CONFIG_INET + +/* This function handles the segmentation of a large TSO packet into multiple + * MPDUs, ensuring that the resulting segments conform to AMSDU limits and + * constraints. + */ +static int iwl_mld_tx_tso_segment(struct iwl_mld *mld, struct sk_buff *skb, + struct ieee80211_sta *sta, + struct sk_buff_head *mpdus_skbs) +{ + struct ieee80211_hdr *hdr = (void *)skb->data; + netdev_features_t netdev_flags = NETIF_F_CSUM_MASK | NETIF_F_SG; + unsigned int mss = skb_shinfo(skb)->gso_size; + unsigned int num_subframes, tcp_payload_len, subf_len; + u16 snap_ip_tcp, pad, max_tid_amsdu_len; + u8 tid; + + snap_ip_tcp = 8 + skb_network_header_len(skb) + tcp_hdrlen(skb); + + if (!ieee80211_is_data_qos(hdr->frame_control) || + !sta->cur->max_rc_amsdu_len) + return iwl_tx_tso_segment(skb, 1, netdev_flags, mpdus_skbs); + + /* Do not build AMSDU for IPv6 with extension headers. + * Ask stack to segment and checksum the generated MPDUs for us. + */ + if (skb->protocol == htons(ETH_P_IPV6) && + ((struct ipv6hdr *)skb_network_header(skb))->nexthdr != + IPPROTO_TCP) { + netdev_flags &= ~NETIF_F_CSUM_MASK; + return iwl_tx_tso_segment(skb, 1, netdev_flags, mpdus_skbs); + } + + tid = ieee80211_get_tid(hdr); + if (WARN_ON_ONCE(tid >= IWL_MAX_TID_COUNT)) + return -EINVAL; + + max_tid_amsdu_len = sta->cur->max_tid_amsdu_len[tid]; + if (!max_tid_amsdu_len) + return iwl_tx_tso_segment(skb, 1, netdev_flags, mpdus_skbs); + + /* Sub frame header + SNAP + IP header + TCP header + MSS */ + subf_len = sizeof(struct ethhdr) + snap_ip_tcp + mss; + pad = (4 - subf_len) & 0x3; + + /* If we have N subframes in the A-MSDU, then the A-MSDU's size is + * N * subf_len + (N - 1) * pad. + */ + num_subframes = (max_tid_amsdu_len + pad) / (subf_len + pad); + + if (sta->max_amsdu_subframes && + num_subframes > sta->max_amsdu_subframes) + num_subframes = sta->max_amsdu_subframes; + + tcp_payload_len = skb_tail_pointer(skb) - skb_transport_header(skb) - + tcp_hdrlen(skb) + skb->data_len; + + /* Make sure we have enough TBs for the A-MSDU: + * 2 for each subframe + * 1 more for each fragment + * 1 more for the potential data in the header + */ + if ((num_subframes * 2 + skb_shinfo(skb)->nr_frags + 1) > + mld->trans->max_skb_frags) + num_subframes = 1; + + if (num_subframes > 1) + *ieee80211_get_qos_ctl(hdr) |= IEEE80211_QOS_CTL_A_MSDU_PRESENT; + + /* This skb fits in one single A-MSDU */ + if (tcp_payload_len <= num_subframes * mss) { + __skb_queue_tail(mpdus_skbs, skb); + return 0; + } + + /* Trick the segmentation function to make it create SKBs that can fit + * into one A-MSDU. + */ + return iwl_tx_tso_segment(skb, num_subframes, netdev_flags, mpdus_skbs); +} + +/* Manages TSO (TCP Segmentation Offload) packet transmission by segmenting + * large packets when necessary and transmitting each segment as MPDU. + */ +static int iwl_mld_tx_tso(struct iwl_mld *mld, struct sk_buff *skb, + struct ieee80211_txq *txq) +{ + struct ieee80211_tx_info *info = IEEE80211_SKB_CB(skb); + struct sk_buff *orig_skb = skb; + struct sk_buff_head mpdus_skbs; + unsigned int payload_len; + int ret; + + if (WARN_ON(!txq || !txq->sta)) + return -1; + + payload_len = skb_tail_pointer(skb) - skb_transport_header(skb) - + tcp_hdrlen(skb) + skb->data_len; + + if (payload_len <= skb_shinfo(skb)->gso_size) + return iwl_mld_tx_mpdu(mld, skb, txq); + + if (!info->control.vif) + return -1; + + __skb_queue_head_init(&mpdus_skbs); + + ret = iwl_mld_tx_tso_segment(mld, skb, txq->sta, &mpdus_skbs); + if (ret) + return ret; + + WARN_ON(skb_queue_empty(&mpdus_skbs)); + + while (!skb_queue_empty(&mpdus_skbs)) { + skb = __skb_dequeue(&mpdus_skbs); + + ret = iwl_mld_tx_mpdu(mld, skb, txq); + if (!ret) + continue; + + /* Free skbs created as part of TSO logic that have not yet + * been dequeued + */ + __skb_queue_purge(&mpdus_skbs); + + /* skb here is not necessarily same as skb that entered + * this method, so free it explicitly. + */ + if (skb == orig_skb) + ieee80211_free_txskb(mld->hw, skb); + else + kfree_skb(skb); + + /* there was error, but we consumed skb one way or + * another, so return 0 + */ + return 0; + } + + return 0; +} +#else +static int iwl_mld_tx_tso(struct iwl_mld *mld, struct sk_buff *skb, + struct ieee80211_txq *txq) +{ + /* Impossible to get TSO without CONFIG_INET */ + WARN_ON(1); + + return -1; +} +#endif /* CONFIG_INET */ + +void iwl_mld_tx_skb(struct iwl_mld *mld, struct sk_buff *skb, + struct ieee80211_txq *txq) +{ + if (skb_is_gso(skb)) { + if (!iwl_mld_tx_tso(mld, skb, txq)) + return; + goto err; + } + + if (likely(!iwl_mld_tx_mpdu(mld, skb, txq))) + return; + +err: + ieee80211_free_txskb(mld->hw, skb); +} + +void iwl_mld_tx_from_txq(struct iwl_mld *mld, struct ieee80211_txq *txq) +{ + struct iwl_mld_txq *mld_txq = iwl_mld_txq_from_mac80211(txq); + struct sk_buff *skb = NULL; + u8 zero_addr[ETH_ALEN] = {}; + + /* + * No need for threads to be pending here, they can leave the first + * taker all the work. + * + * mld_txq->tx_request logic: + * + * If 0, no one is currently TXing, set to 1 to indicate current thread + * will now start TX and other threads should quit. + * + * If 1, another thread is currently TXing, set to 2 to indicate to + * that thread that there was another request. Since that request may + * have raced with the check whether the queue is empty, the TXing + * thread should check the queue's status one more time before leaving. + * This check is done in order to not leave any TX hanging in the queue + * until the next TX invocation (which may not even happen). + * + * If 2, another thread is currently TXing, and it will already double + * check the queue, so do nothing. + */ + if (atomic_fetch_add_unless(&mld_txq->tx_request, 1, 2)) + return; + + rcu_read_lock(); + do { + while (likely(!mld_txq->status.stop_full) && + (skb = ieee80211_tx_dequeue(mld->hw, txq))) + iwl_mld_tx_skb(mld, skb, txq); + } while (atomic_dec_return(&mld_txq->tx_request)); + + IWL_DEBUG_TX(mld, "TXQ of sta %pM tid %d is now empty\n", + txq->sta ? txq->sta->addr : zero_addr, txq->tid); + + rcu_read_unlock(); +} + +static void iwl_mld_hwrate_to_tx_rate(u32 rate_n_flags, + struct ieee80211_tx_info *info) +{ + enum nl80211_band band = info->band; + struct ieee80211_tx_rate *tx_rate = &info->status.rates[0]; + u32 sgi = rate_n_flags & RATE_MCS_SGI_MSK; + u32 chan_width = rate_n_flags & RATE_MCS_CHAN_WIDTH_MSK; + u32 format = rate_n_flags & RATE_MCS_MOD_TYPE_MSK; + + if (sgi) + tx_rate->flags |= IEEE80211_TX_RC_SHORT_GI; + + switch (chan_width) { + case RATE_MCS_CHAN_WIDTH_20: + break; + case RATE_MCS_CHAN_WIDTH_40: + tx_rate->flags |= IEEE80211_TX_RC_40_MHZ_WIDTH; + break; + case RATE_MCS_CHAN_WIDTH_80: + tx_rate->flags |= IEEE80211_TX_RC_80_MHZ_WIDTH; + break; + case RATE_MCS_CHAN_WIDTH_160: + tx_rate->flags |= IEEE80211_TX_RC_160_MHZ_WIDTH; + break; + default: + break; + } + + switch (format) { + case RATE_MCS_HT_MSK: + tx_rate->flags |= IEEE80211_TX_RC_MCS; + tx_rate->idx = RATE_HT_MCS_INDEX(rate_n_flags); + break; + case RATE_MCS_VHT_MSK: + ieee80211_rate_set_vht(tx_rate, + rate_n_flags & RATE_MCS_CODE_MSK, + FIELD_GET(RATE_MCS_NSS_MSK, + rate_n_flags) + 1); + tx_rate->flags |= IEEE80211_TX_RC_VHT_MCS; + break; + case RATE_MCS_HE_MSK: + /* mac80211 cannot do this without ieee80211_tx_status_ext() + * but it only matters for radiotap + */ + tx_rate->idx = 0; + break; + default: + tx_rate->idx = + iwl_mld_legacy_hw_idx_to_mac80211_idx(rate_n_flags, + band); + break; + } +} + +void iwl_mld_handle_tx_resp_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + struct iwl_tx_resp *tx_resp = (void *)pkt->data; + int txq_id = le16_to_cpu(tx_resp->tx_queue); + struct agg_tx_status *agg_status = &tx_resp->status; + u32 status = le16_to_cpu(agg_status->status); + u32 pkt_len = iwl_rx_packet_payload_len(pkt); + size_t notif_size = sizeof(*tx_resp) + sizeof(u32); + int sta_id = IWL_TX_RES_GET_RA(tx_resp->ra_tid); + int tid = IWL_TX_RES_GET_TID(tx_resp->ra_tid); + struct ieee80211_link_sta *link_sta; + struct iwl_mld_sta *mld_sta; + u16 ssn; + struct sk_buff_head skbs; + u8 skb_freed = 0; + bool mgmt = false; + bool tx_failure = (status & TX_STATUS_MSK) != TX_STATUS_SUCCESS; + + if (IWL_FW_CHECK(mld, tx_resp->frame_count != 1, + "Invalid tx_resp notif frame_count (%d)\n", + tx_resp->frame_count)) + return; + + /* validate the size of the variable part of the notif */ + if (IWL_FW_CHECK(mld, notif_size != pkt_len, + "Invalid tx_resp notif size (expected=%zu got=%u)\n", + notif_size, pkt_len)) + return; + + ssn = le32_to_cpup((__le32 *)agg_status + + tx_resp->frame_count) & 0xFFFF; + + __skb_queue_head_init(&skbs); + + /* we can free until ssn % q.n_bd not inclusive */ + iwl_trans_reclaim(mld->trans, txq_id, ssn, &skbs, false); + + while (!skb_queue_empty(&skbs)) { + struct sk_buff *skb = __skb_dequeue(&skbs); + struct ieee80211_tx_info *info = IEEE80211_SKB_CB(skb); + struct ieee80211_hdr *hdr = (void *)skb->data; + + skb_freed++; + + iwl_trans_free_tx_cmd(mld->trans, info->driver_data[1]); + + memset(&info->status, 0, sizeof(info->status)); + + info->flags &= ~(IEEE80211_TX_STAT_ACK | IEEE80211_TX_STAT_TX_FILTERED); + + /* inform mac80211 about what happened with the frame */ + switch (status & TX_STATUS_MSK) { + case TX_STATUS_SUCCESS: + case TX_STATUS_DIRECT_DONE: + info->flags |= IEEE80211_TX_STAT_ACK; + break; + default: + break; + } + + /* If we are freeing multiple frames, mark all the frames + * but the first one as acked, since they were acknowledged + * before + */ + if (skb_freed > 1) + info->flags |= IEEE80211_TX_STAT_ACK; + + if (tx_failure) { + enum iwl_fw_ini_time_point tp = + IWL_FW_INI_TIME_POINT_TX_FAILED; + + if (ieee80211_is_action(hdr->frame_control)) + tp = IWL_FW_INI_TIME_POINT_TX_WFD_ACTION_FRAME_FAILED; + else if (ieee80211_is_mgmt(hdr->frame_control)) + mgmt = true; + + iwl_dbg_tlv_time_point(&mld->fwrt, tp, NULL); + } + + iwl_mld_hwrate_to_tx_rate(le32_to_cpu(tx_resp->initial_rate), + info); + + if (likely(!iwl_mld_time_sync_frame(mld, skb, hdr->addr1))) + ieee80211_tx_status_skb(mld->hw, skb); + } + + IWL_DEBUG_TX_REPLY(mld, + "TXQ %d status 0x%08x ssn=%d initial_rate 0x%x retries %d\n", + txq_id, status, ssn, le32_to_cpu(tx_resp->initial_rate), + tx_resp->failure_frame); + + if (tx_failure && mgmt) + iwl_mld_toggle_tx_ant(mld, &mld->mgmt_tx_ant); + + if (IWL_FW_CHECK(mld, sta_id >= mld->fw->ucode_capa.num_stations, + "Got invalid sta_id (%d)\n", sta_id)) + return; + + rcu_read_lock(); + + link_sta = rcu_dereference(mld->fw_id_to_link_sta[sta_id]); + if (!link_sta) { + /* This can happen if the TX cmd was sent before pre_rcu_remove + * but the TX response was received after + */ + IWL_DEBUG_TX_REPLY(mld, + "Got valid sta_id (%d) but sta is NULL\n", + sta_id); + goto out; + } + + if (IS_ERR(link_sta)) + goto out; + + mld_sta = iwl_mld_sta_from_mac80211(link_sta->sta); + + if (tx_failure && mld_sta->sta_state < IEEE80211_STA_AUTHORIZED) + iwl_mld_toggle_tx_ant(mld, &mld_sta->data_tx_ant); + + if (tid < IWL_MAX_TID_COUNT) + iwl_mld_count_mpdu_tx(link_sta, 1); + +out: + rcu_read_unlock(); +} + +static void iwl_mld_tx_reclaim_txq(struct iwl_mld *mld, int txq, int index, + bool in_flush) +{ + struct sk_buff_head reclaimed_skbs; + + __skb_queue_head_init(&reclaimed_skbs); + + iwl_trans_reclaim(mld->trans, txq, index, &reclaimed_skbs, in_flush); + + while (!skb_queue_empty(&reclaimed_skbs)) { + struct sk_buff *skb = __skb_dequeue(&reclaimed_skbs); + struct ieee80211_tx_info *info = IEEE80211_SKB_CB(skb); + + iwl_trans_free_tx_cmd(mld->trans, info->driver_data[1]); + + memset(&info->status, 0, sizeof(info->status)); + + /* Packet was transmitted successfully, failures come as single + * frames because before failing a frame the firmware transmits + * it without aggregation at least once. + */ + if (!in_flush) + info->flags |= IEEE80211_TX_STAT_ACK; + else + info->flags &= ~IEEE80211_TX_STAT_ACK; + + ieee80211_tx_status_skb(mld->hw, skb); + } +} + +int iwl_mld_flush_link_sta_txqs(struct iwl_mld *mld, u32 fw_sta_id) +{ + struct iwl_tx_path_flush_cmd_rsp *rsp; + struct iwl_tx_path_flush_cmd flush_cmd = { + .sta_id = cpu_to_le32(fw_sta_id), + .tid_mask = cpu_to_le16(0xffff), + }; + struct iwl_host_cmd cmd = { + .id = TXPATH_FLUSH, + .len = { sizeof(flush_cmd), }, + .data = { &flush_cmd, }, + .flags = CMD_WANT_SKB, + }; + int ret, num_flushed_queues; + u32 resp_len; + + IWL_DEBUG_TX_QUEUES(mld, "flush for sta id %d tid mask 0x%x\n", + fw_sta_id, 0xffff); + + ret = iwl_mld_send_cmd(mld, &cmd); + if (ret) { + IWL_ERR(mld, "Failed to send flush command (%d)\n", ret); + return ret; + } + + resp_len = iwl_rx_packet_payload_len(cmd.resp_pkt); + if (IWL_FW_CHECK(mld, resp_len != sizeof(*rsp), + "Invalid TXPATH_FLUSH response len: %d\n", + resp_len)) { + ret = -EIO; + goto free_rsp; + } + + rsp = (void *)cmd.resp_pkt->data; + + if (IWL_FW_CHECK(mld, le16_to_cpu(rsp->sta_id) != fw_sta_id, + "sta_id %d != rsp_sta_id %d\n", fw_sta_id, + le16_to_cpu(rsp->sta_id))) { + ret = -EIO; + goto free_rsp; + } + + num_flushed_queues = le16_to_cpu(rsp->num_flushed_queues); + if (IWL_FW_CHECK(mld, num_flushed_queues > IWL_TX_FLUSH_QUEUE_RSP, + "num_flushed_queues %d\n", num_flushed_queues)) { + ret = -EIO; + goto free_rsp; + } + + for (int i = 0; i < num_flushed_queues; i++) { + struct iwl_flush_queue_info *queue_info = &rsp->queues[i]; + int read_after = le16_to_cpu(queue_info->read_after_flush); + int txq_id = le16_to_cpu(queue_info->queue_num); + + if (IWL_FW_CHECK(mld, + txq_id >= ARRAY_SIZE(mld->fw_id_to_txq), + "Invalid txq id %d\n", txq_id)) + continue; + + IWL_DEBUG_TX_QUEUES(mld, + "tid %d txq_id %d read-before %d read-after %d\n", + le16_to_cpu(queue_info->tid), txq_id, + le16_to_cpu(queue_info->read_before_flush), + read_after); + + iwl_mld_tx_reclaim_txq(mld, txq_id, read_after, true); + } + +free_rsp: + iwl_free_resp(&cmd); + return ret; +} + +int iwl_mld_ensure_queue(struct iwl_mld *mld, struct ieee80211_txq *txq) +{ + struct iwl_mld_txq *mld_txq = iwl_mld_txq_from_mac80211(txq); + int ret; + + lockdep_assert_wiphy(mld->wiphy); + + if (likely(mld_txq->status.allocated)) + return 0; + + ret = iwl_mld_add_txq(mld, txq); + + spin_lock_bh(&mld->add_txqs_lock); + if (!list_empty(&mld_txq->list)) + list_del_init(&mld_txq->list); + spin_unlock_bh(&mld->add_txqs_lock); + + return ret; +} + +int iwl_mld_update_sta_txqs(struct iwl_mld *mld, + struct ieee80211_sta *sta, + u32 old_sta_mask, u32 new_sta_mask) +{ + struct iwl_scd_queue_cfg_cmd cmd = { + .operation = cpu_to_le32(IWL_SCD_QUEUE_MODIFY), + .u.modify.old_sta_mask = cpu_to_le32(old_sta_mask), + .u.modify.new_sta_mask = cpu_to_le32(new_sta_mask), + }; + + lockdep_assert_wiphy(mld->wiphy); + + for (int tid = 0; tid <= IWL_MAX_TID_COUNT; tid++) { + struct ieee80211_txq *txq = + sta->txq[tid != IWL_MAX_TID_COUNT ? + tid : IEEE80211_NUM_TIDS]; + struct iwl_mld_txq *mld_txq = + iwl_mld_txq_from_mac80211(txq); + int ret; + + if (!mld_txq->status.allocated) + continue; + + if (tid == IWL_MAX_TID_COUNT) + cmd.u.modify.tid = cpu_to_le32(IWL_MGMT_TID); + else + cmd.u.modify.tid = cpu_to_le32(tid); + + ret = iwl_mld_send_cmd_pdu(mld, + WIDE_ID(DATA_PATH_GROUP, + SCD_QUEUE_CONFIG_CMD), + &cmd); + if (ret) + return ret; + } + + return 0; +} + +void iwl_mld_handle_compressed_ba_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt) +{ + struct iwl_compressed_ba_notif *ba_res = (void *)pkt->data; + u32 pkt_len = iwl_rx_packet_payload_len(pkt); + u16 tfd_cnt = le16_to_cpu(ba_res->tfd_cnt); + u8 sta_id = ba_res->sta_id; + struct ieee80211_link_sta *link_sta; + + if (!tfd_cnt) + return; + + if (IWL_FW_CHECK(mld, struct_size(ba_res, tfd, tfd_cnt) > pkt_len, + "Short BA notif (tfd_cnt=%d, size:0x%x)\n", + tfd_cnt, pkt_len)) + return; + + IWL_DEBUG_TX_REPLY(mld, + "BA notif received from sta_id=%d, flags=0x%x, sent:%d, acked:%d\n", + sta_id, le32_to_cpu(ba_res->flags), + le16_to_cpu(ba_res->txed), + le16_to_cpu(ba_res->done)); + + for (int i = 0; i < tfd_cnt; i++) { + struct iwl_compressed_ba_tfd *ba_tfd = &ba_res->tfd[i]; + int txq_id = le16_to_cpu(ba_tfd->q_num); + int index = le16_to_cpu(ba_tfd->tfd_index); + + if (IWL_FW_CHECK(mld, + txq_id >= ARRAY_SIZE(mld->fw_id_to_txq), + "Invalid txq id %d\n", txq_id)) + continue; + + iwl_mld_tx_reclaim_txq(mld, txq_id, index, false); + } + + if (IWL_FW_CHECK(mld, sta_id >= mld->fw->ucode_capa.num_stations, + "Got invalid sta_id (%d)\n", sta_id)) + return; + + rcu_read_lock(); + + link_sta = rcu_dereference(mld->fw_id_to_link_sta[sta_id]); + if (IWL_FW_CHECK(mld, IS_ERR_OR_NULL(link_sta), + "Got valid sta_id (%d) but link_sta is NULL\n", + sta_id)) + goto out; + + iwl_mld_count_mpdu_tx(link_sta, le16_to_cpu(ba_res->txed)); +out: + rcu_read_unlock(); +} diff --git a/drivers/net/wireless/intel/iwlwifi/mld/tx.h b/drivers/net/wireless/intel/iwlwifi/mld/tx.h new file mode 100644 index 0000000000000..520f15f9d33ca --- /dev/null +++ b/drivers/net/wireless/intel/iwlwifi/mld/tx.h @@ -0,0 +1,77 @@ +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ +/* + * Copyright (C) 2024 Intel Corporation + */ +#ifndef __iwl_mld_tx_h__ +#define __iwl_mld_tx_h__ + +#include "mld.h" + +#define IWL_MLD_INVALID_QUEUE 0xFFFF +#define IWL_MLD_INVALID_DROP_TX 0xFFFE + +/** + * struct iwl_mld_txq - TX Queue data + * + * @fw_id: the fw id of this txq. Only valid when &status.allocated is true. + * @status: bitmap of the txq status + * @status.allocated: Indicates that the queue was allocated. + * @status.stop_full: Indicates that the queue is full and should stop TXing. + * @list: list pointer, for &mld::txqs_to_add + * @tx_request: makes sure that if there are multiple threads that want to tx + * from this txq, only one of them will do all the TXing. + * This is needed to avoid spinning the trans txq lock, which is expensive + */ +struct iwl_mld_txq { + /* Add here fields that need clean up on restart */ + struct_group(zeroed_on_hw_restart, + u16 fw_id; + struct { + u8 allocated:1; + u8 stop_full:1; + } status; + ); + struct list_head list; + atomic_t tx_request; + /* And here fields that survive a fw restart */ +}; + +static inline void iwl_mld_init_txq(struct iwl_mld_txq *mld_txq) +{ + INIT_LIST_HEAD(&mld_txq->list); + atomic_set(&mld_txq->tx_request, 0); +} + +static inline struct iwl_mld_txq * +iwl_mld_txq_from_mac80211(struct ieee80211_txq *txq) +{ + return (void *)txq->drv_priv; +} + +void iwl_mld_add_txqs_wk(struct wiphy *wiphy, struct wiphy_work *wk); +void iwl_mld_remove_txq(struct iwl_mld *mld, struct ieee80211_txq *txq); +void iwl_mld_add_txq_list(struct iwl_mld *mld); +void +iwl_mld_free_txq(struct iwl_mld *mld, u32 fw_sta_mask, u32 tid, u32 queue_id); +void iwl_mld_tx_from_txq(struct iwl_mld *mld, struct ieee80211_txq *txq); +void iwl_mld_handle_tx_resp_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt); +int iwl_mld_flush_link_sta_txqs(struct iwl_mld *mld, u32 fw_sta_id); +int iwl_mld_ensure_queue(struct iwl_mld *mld, struct ieee80211_txq *txq); + +int iwl_mld_update_sta_txqs(struct iwl_mld *mld, + struct ieee80211_sta *sta, + u32 old_sta_mask, u32 new_sta_mask); + +void iwl_mld_handle_compressed_ba_notif(struct iwl_mld *mld, + struct iwl_rx_packet *pkt); +void iwl_mld_toggle_tx_ant(struct iwl_mld *mld, u8 *ant); + +u8 iwl_mld_get_lowest_rate(struct iwl_mld *mld, + struct ieee80211_tx_info *info, + struct ieee80211_vif *vif); + +void iwl_mld_tx_skb(struct iwl_mld *mld, struct sk_buff *skb, + struct ieee80211_txq *txq); + +#endif /* __iwl_mld_tx_h__ */ diff --git a/drivers/net/wireless/intel/iwlwifi/pcie/drv.c b/drivers/net/wireless/intel/iwlwifi/pcie/drv.c index cd347cd8176a7..03f7eb46bbc78 100644 --- a/drivers/net/wireless/intel/iwlwifi/pcie/drv.c +++ b/drivers/net/wireless/intel/iwlwifi/pcie/drv.c @@ -498,7 +498,8 @@ VISIBLE_IF_IWLWIFI_KUNIT const struct pci_device_id iwl_hw_card_ids[] = { /* Ma devices */ {IWL_PCI_DEVICE(0x2729, PCI_ANY_ID, iwl_ma_trans_cfg)}, {IWL_PCI_DEVICE(0x7E40, PCI_ANY_ID, iwl_ma_trans_cfg)}, - +#endif /* CONFIG_IWLMVM */ +#if IS_ENABLED(CONFIG_IWLMLD) /* Bz devices */ {IWL_PCI_DEVICE(0x272b, PCI_ANY_ID, iwl_gl_trans_cfg)}, {IWL_PCI_DEVICE(0xA840, 0x0000, iwl_bz_trans_cfg)}, @@ -543,7 +544,7 @@ VISIBLE_IF_IWLWIFI_KUNIT const struct pci_device_id iwl_hw_card_ids[] = { /* Dr devices */ {IWL_PCI_DEVICE(0x272F, PCI_ANY_ID, iwl_dr_trans_cfg)}, -#endif /* CONFIG_IWLMVM */ +#endif /* CONFIG_IWLMLD */ {0} }; @@ -1109,6 +1110,8 @@ VISIBLE_IF_IWLWIFI_KUNIT const struct iwl_dev_info iwl_dev_info_table[] = { 80, IWL_CFG_CORES_BT, IWL_CFG_NO_CDB, iwlax210_2ax_cfg_so_jf_b0, iwl9462_name), +#endif /* CONFIG_IWLMVM */ +#if IS_ENABLED(CONFIG_IWLMLD) /* Bz */ _IWL_DEV_INFO(IWL_CFG_ANY, IWL_CFG_ANY, IWL_CFG_MAC_TYPE_BZ, IWL_CFG_ANY, @@ -1230,7 +1233,7 @@ VISIBLE_IF_IWLWIFI_KUNIT const struct iwl_dev_info iwl_dev_info_table[] = { IWL_CFG_ANY, IWL_CFG_ANY, IWL_CFG_ANY, IWL_CFG_BW_ANY, IWL_CFG_ANY, IWL_CFG_ANY, iwl_cfg_br, iwl_br_name), -#endif /* CONFIG_IWLMVM */ +#endif /* CONFIG_IWLMLD */ }; EXPORT_SYMBOL_IF_IWLWIFI_KUNIT(iwl_dev_info_table);