From ef61a0405742a9f7f6051bc6fd2f017d87d07911 Mon Sep 17 00:00:00 2001 From: Jiaxun Yang Date: Fri, 1 Dec 2023 11:50:28 +0000 Subject: [PATCH 1/8] PCI: loongson: Limit MRRS to 256 This is a partial revert of 8b3517f88ff2 ("PCI: loongson: Prevent LS7A MRRS increases") for MIPS-based Loongson. Some MIPS Loongson systems don't support arbitrary Max_Read_Request_Size (MRRS) settings. 8b3517f88ff2 ("PCI: loongson: Prevent LS7A MRRS increases") worked around that by (1) assuming that firmware configured MRRS to the maximum supported value and (2) preventing the PCI core from increasing MRRS. Unfortunately, some firmware doesn't set that maximum MRRS correctly, which results in devices not being initialized correctly. One symptom, from the Debian report below, is this: ata4.00: exception Emask 0x0 SAct 0x20000000 SErr 0x0 action 0x6 frozen ata4.00: failed command: WRITE FPDMA QUEUED ata4.00: cmd 61/20:e8:00:f0:e1/00:00:00:00:00/40 tag 29 ncq dma 16384 out res 40/00:00:00:00:00/00:00:00:00:00/00 Emask 0x4 (timeout) ata4.00: status: { DRDY } ata4: hard resetting link Limit MRRS to 256 because MIPS Loongson with higher MRRS support is considered rare. This must be done at device enablement stage because the MRRS setting may get lost if PCI_COMMAND_MASTER on the parent bridge is cleared, and we are only sure parent bridge is enabled at this point. Fixes: 8b3517f88ff2 ("PCI: loongson: Prevent LS7A MRRS increases") Closes: https://bugzilla.kernel.org/show_bug.cgi?id=217680 Link: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1035587 Link: https://lore.kernel.org/r/20231201115028.84351-1-jiaxun.yang@flygoat.com Signed-off-by: Jiaxun Yang Signed-off-by: Bjorn Helgaas Acked-by: Huacai Chen Cc: stable@vger.kernel.org --- drivers/pci/controller/pci-loongson.c | 46 ++++++++++++++++++++++++--- 1 file changed, 41 insertions(+), 5 deletions(-) diff --git a/drivers/pci/controller/pci-loongson.c b/drivers/pci/controller/pci-loongson.c index d45e7b8dc530d..8b34ccff073a9 100644 --- a/drivers/pci/controller/pci-loongson.c +++ b/drivers/pci/controller/pci-loongson.c @@ -80,13 +80,49 @@ DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_LOONGSON, DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_LOONGSON, DEV_LS7A_LPC, system_bus_quirk); +/* + * Some Loongson PCIe ports have hardware limitations on their Maximum Read + * Request Size. They can't handle anything larger than this. Sane + * firmware will set proper MRRS at boot, so we only need no_inc_mrrs for + * bridges. However, some MIPS Loongson firmware doesn't set MRRS properly, + * so we have to enforce maximum safe MRRS, which is 256 bytes. + */ +#ifdef CONFIG_MIPS +static void loongson_set_min_mrrs_quirk(struct pci_dev *pdev) +{ + struct pci_bus *bus = pdev->bus; + struct pci_dev *bridge; + static const struct pci_device_id bridge_devids[] = { + { PCI_VDEVICE(LOONGSON, DEV_LS2K_PCIE_PORT0) }, + { PCI_VDEVICE(LOONGSON, DEV_LS7A_PCIE_PORT0) }, + { PCI_VDEVICE(LOONGSON, DEV_LS7A_PCIE_PORT1) }, + { PCI_VDEVICE(LOONGSON, DEV_LS7A_PCIE_PORT2) }, + { PCI_VDEVICE(LOONGSON, DEV_LS7A_PCIE_PORT3) }, + { PCI_VDEVICE(LOONGSON, DEV_LS7A_PCIE_PORT4) }, + { PCI_VDEVICE(LOONGSON, DEV_LS7A_PCIE_PORT5) }, + { PCI_VDEVICE(LOONGSON, DEV_LS7A_PCIE_PORT6) }, + { 0, }, + }; + + /* look for the matching bridge */ + while (!pci_is_root_bus(bus)) { + bridge = bus->self; + bus = bus->parent; + + if (pci_match_id(bridge_devids, bridge)) { + if (pcie_get_readrq(pdev) > 256) { + pci_info(pdev, "limiting MRRS to 256\n"); + pcie_set_readrq(pdev, 256); + } + break; + } + } +} +DECLARE_PCI_FIXUP_ENABLE(PCI_ANY_ID, PCI_ANY_ID, loongson_set_min_mrrs_quirk); +#endif + static void loongson_mrrs_quirk(struct pci_dev *pdev) { - /* - * Some Loongson PCIe ports have h/w limitations of maximum read - * request size. They can't handle anything larger than this. So - * force this limit on any devices attached under these ports. - */ struct pci_host_bridge *bridge = pci_find_host_bridge(pdev->bus); bridge->no_inc_mrrs = 1; From 718ab8226636a1a3a7d281f5d6a7ad7c925efe5a Mon Sep 17 00:00:00 2001 From: Johan Hovold Date: Tue, 28 Nov 2023 09:15:07 +0100 Subject: [PATCH 2/8] PCI/ASPM: Add pci_enable_link_state_locked() Add pci_enable_link_state_locked() for enabling link states that can be used in contexts where a pci_bus_sem read lock is already held (e.g. from pci_walk_bus()). This helper will be used to fix a couple of potential deadlocks where the current helper is called with the lock already held, hence the CC stable tag. Fixes: f492edb40b54 ("PCI: vmd: Add quirk to configure PCIe ASPM and LTR") Link: https://lore.kernel.org/r/20231128081512.19387-2-johan+linaro@kernel.org Signed-off-by: Johan Hovold [bhelgaas: include helper name in subject, commit log] Signed-off-by: Bjorn Helgaas Reviewed-by: Manivannan Sadhasivam Cc: # 6.3 Cc: Michael Bottini Cc: David E. Box --- drivers/pci/pcie/aspm.c | 53 +++++++++++++++++++++++++++++++---------- include/linux/pci.h | 3 +++ 2 files changed, 43 insertions(+), 13 deletions(-) diff --git a/drivers/pci/pcie/aspm.c b/drivers/pci/pcie/aspm.c index 50b04ae5c394e..5eb462772354e 100644 --- a/drivers/pci/pcie/aspm.c +++ b/drivers/pci/pcie/aspm.c @@ -1109,17 +1109,7 @@ int pci_disable_link_state(struct pci_dev *pdev, int state) } EXPORT_SYMBOL(pci_disable_link_state); -/** - * pci_enable_link_state - Clear and set the default device link state so that - * the link may be allowed to enter the specified states. Note that if the - * BIOS didn't grant ASPM control to the OS, this does nothing because we can't - * touch the LNKCTL register. Also note that this does not enable states - * disabled by pci_disable_link_state(). Return 0 or a negative errno. - * - * @pdev: PCI device - * @state: Mask of ASPM link states to enable - */ -int pci_enable_link_state(struct pci_dev *pdev, int state) +static int __pci_enable_link_state(struct pci_dev *pdev, int state, bool locked) { struct pcie_link_state *link = pcie_aspm_get_link(pdev); @@ -1136,7 +1126,8 @@ int pci_enable_link_state(struct pci_dev *pdev, int state) return -EPERM; } - down_read(&pci_bus_sem); + if (!locked) + down_read(&pci_bus_sem); mutex_lock(&aspm_lock); link->aspm_default = 0; if (state & PCIE_LINK_STATE_L0S) @@ -1157,12 +1148,48 @@ int pci_enable_link_state(struct pci_dev *pdev, int state) link->clkpm_default = (state & PCIE_LINK_STATE_CLKPM) ? 1 : 0; pcie_set_clkpm(link, policy_to_clkpm_state(link)); mutex_unlock(&aspm_lock); - up_read(&pci_bus_sem); + if (!locked) + up_read(&pci_bus_sem); return 0; } + +/** + * pci_enable_link_state - Clear and set the default device link state so that + * the link may be allowed to enter the specified states. Note that if the + * BIOS didn't grant ASPM control to the OS, this does nothing because we can't + * touch the LNKCTL register. Also note that this does not enable states + * disabled by pci_disable_link_state(). Return 0 or a negative errno. + * + * @pdev: PCI device + * @state: Mask of ASPM link states to enable + */ +int pci_enable_link_state(struct pci_dev *pdev, int state) +{ + return __pci_enable_link_state(pdev, state, false); +} EXPORT_SYMBOL(pci_enable_link_state); +/** + * pci_enable_link_state_locked - Clear and set the default device link state + * so that the link may be allowed to enter the specified states. Note that if + * the BIOS didn't grant ASPM control to the OS, this does nothing because we + * can't touch the LNKCTL register. Also note that this does not enable states + * disabled by pci_disable_link_state(). Return 0 or a negative errno. + * + * @pdev: PCI device + * @state: Mask of ASPM link states to enable + * + * Context: Caller holds pci_bus_sem read lock. + */ +int pci_enable_link_state_locked(struct pci_dev *pdev, int state) +{ + lockdep_assert_held_read(&pci_bus_sem); + + return __pci_enable_link_state(pdev, state, true); +} +EXPORT_SYMBOL(pci_enable_link_state_locked); + static int pcie_aspm_set_policy(const char *val, const struct kernel_param *kp) { diff --git a/include/linux/pci.h b/include/linux/pci.h index 60ca768bc8679..dea043bc1e383 100644 --- a/include/linux/pci.h +++ b/include/linux/pci.h @@ -1829,6 +1829,7 @@ extern bool pcie_ports_native; int pci_disable_link_state(struct pci_dev *pdev, int state); int pci_disable_link_state_locked(struct pci_dev *pdev, int state); int pci_enable_link_state(struct pci_dev *pdev, int state); +int pci_enable_link_state_locked(struct pci_dev *pdev, int state); void pcie_no_aspm(void); bool pcie_aspm_support_enabled(void); bool pcie_aspm_enabled(struct pci_dev *pdev); @@ -1839,6 +1840,8 @@ static inline int pci_disable_link_state_locked(struct pci_dev *pdev, int state) { return 0; } static inline int pci_enable_link_state(struct pci_dev *pdev, int state) { return 0; } +static inline int pci_enable_link_state_locked(struct pci_dev *pdev, int state) +{ return 0; } static inline void pcie_no_aspm(void) { } static inline bool pcie_aspm_support_enabled(void) { return false; } static inline bool pcie_aspm_enabled(struct pci_dev *pdev) { return false; } From 49de0dc87965079a8e2803ee4b39f9d946259423 Mon Sep 17 00:00:00 2001 From: Johan Hovold Date: Tue, 28 Nov 2023 09:15:08 +0100 Subject: [PATCH 3/8] PCI: vmd: Fix potential deadlock when enabling ASPM The vmd_pm_enable_quirk() helper is called from pci_walk_bus() during probe to enable ASPM for controllers with VMD_FEAT_BIOS_PM_QUIRK set. Since pci_walk_bus() already holds a pci_bus_sem read lock, use pci_enable_link_state_locked() to enable link states in order to avoid a potential deadlock (e.g. in case someone takes a write lock before reacquiring the read lock). Fixes: f492edb40b54 ("PCI: vmd: Add quirk to configure PCIe ASPM and LTR") Link: https://lore.kernel.org/r/20231128081512.19387-3-johan+linaro@kernel.org Signed-off-by: Johan Hovold [bhelgaas: add "potential" in subject since the deadlock has only been reported by lockdep, include helper name in commit log] Signed-off-by: Bjorn Helgaas Reviewed-by: Manivannan Sadhasivam Cc: # 6.3 Cc: Michael Bottini Cc: David E. Box --- drivers/pci/controller/vmd.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/pci/controller/vmd.c b/drivers/pci/controller/vmd.c index 94ba61fe1c441..0452cbc362eef 100644 --- a/drivers/pci/controller/vmd.c +++ b/drivers/pci/controller/vmd.c @@ -751,7 +751,7 @@ static int vmd_pm_enable_quirk(struct pci_dev *pdev, void *userdata) if (!(features & VMD_FEAT_BIOS_PM_QUIRK)) return 0; - pci_enable_link_state(pdev, PCIE_LINK_STATE_ALL); + pci_enable_link_state_locked(pdev, PCIE_LINK_STATE_ALL); pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_LTR); if (!pos) From f352ce99926048e12aa4281c32471031351aec98 Mon Sep 17 00:00:00 2001 From: Johan Hovold Date: Tue, 28 Nov 2023 09:15:09 +0100 Subject: [PATCH 4/8] PCI: qcom: Fix potential deadlock when enabling ASPM The qcom_pcie_enable_aspm() helper is called from pci_walk_bus() during host init to enable ASPM. Since pci_walk_bus() already holds a pci_bus_sem read lock, use pci_enable_link_state_locked() to enable link states in order to avoid a potential deadlock (e.g. in case someone takes a write lock before reacquiring the read lock). This issue was reported by lockdep: ============================================ WARNING: possible recursive locking detected 6.7.0-rc1 #4 Not tainted -------------------------------------------- kworker/u16:6/147 is trying to acquire lock: ffffbf3ff9d2cfa0 (pci_bus_sem){++++}-{3:3}, at: pci_enable_link_state+0x74/0x1e8 but task is already holding lock: ffffbf3ff9d2cfa0 (pci_bus_sem){++++}-{3:3}, at: pci_walk_bus+0x34/0xbc other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(pci_bus_sem); lock(pci_bus_sem); *** DEADLOCK *** Fixes: 9f4f3dfad8cf ("PCI: qcom: Enable ASPM for platforms supporting 1.9.0 ops") Link: https://lore.kernel.org/r/20231128081512.19387-4-johan+linaro@kernel.org Signed-off-by: Johan Hovold [bhelgaas: add "potential" in subject since the deadlock has only been reported by lockdep, include helper name in commit log] Signed-off-by: Bjorn Helgaas Reviewed-by: Manivannan Sadhasivam --- drivers/pci/controller/dwc/pcie-qcom.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/pci/controller/dwc/pcie-qcom.c b/drivers/pci/controller/dwc/pcie-qcom.c index 6902e97719d13..13dbdfac98f2f 100644 --- a/drivers/pci/controller/dwc/pcie-qcom.c +++ b/drivers/pci/controller/dwc/pcie-qcom.c @@ -970,7 +970,7 @@ static int qcom_pcie_enable_aspm(struct pci_dev *pdev, void *userdata) { /* Downstream devices need to be in D0 state before enabling PCI PM substates */ pci_set_power_state(pdev, PCI_D0); - pci_enable_link_state(pdev, PCIE_LINK_STATE_ALL); + pci_enable_link_state_locked(pdev, PCIE_LINK_STATE_ALL); return 0; } From 780f52e3213e5f05bb41adebe1f2214f2f86f4a3 Mon Sep 17 00:00:00 2001 From: Johan Hovold Date: Tue, 28 Nov 2023 09:15:10 +0100 Subject: [PATCH 5/8] PCI: qcom: Clean up ASPM comment Break up the newly added ASPM comment so that it fits within the soft 80 character limit and becomes more readable. Link: https://lore.kernel.org/r/20231128081512.19387-5-johan+linaro@kernel.org Signed-off-by: Johan Hovold Signed-off-by: Bjorn Helgaas --- drivers/pci/controller/dwc/pcie-qcom.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/drivers/pci/controller/dwc/pcie-qcom.c b/drivers/pci/controller/dwc/pcie-qcom.c index 13dbdfac98f2f..11c80555d9754 100644 --- a/drivers/pci/controller/dwc/pcie-qcom.c +++ b/drivers/pci/controller/dwc/pcie-qcom.c @@ -968,7 +968,10 @@ static int qcom_pcie_post_init_2_7_0(struct qcom_pcie *pcie) static int qcom_pcie_enable_aspm(struct pci_dev *pdev, void *userdata) { - /* Downstream devices need to be in D0 state before enabling PCI PM substates */ + /* + * Downstream devices need to be in D0 state before enabling PCI PM + * substates. + */ pci_set_power_state(pdev, PCI_D0); pci_enable_link_state_locked(pdev, PCIE_LINK_STATE_ALL); From e673d383bdba94c9924388086b91988254d39f19 Mon Sep 17 00:00:00 2001 From: Johan Hovold Date: Tue, 28 Nov 2023 09:15:11 +0100 Subject: [PATCH 6/8] PCI/ASPM: Clean up __pci_disable_link_state() 'sem' parameter Replace the current 'sem' parameter to the __pci_disable_link_state() helper with a more descriptive 'locked' parameter, which indicates whether a pci_bus_sem read lock is already held. Link: https://lore.kernel.org/r/20231128081512.19387-6-johan+linaro@kernel.org Signed-off-by: Johan Hovold [bhelgaas: include function name in subject, commit log] Signed-off-by: Bjorn Helgaas Reviewed-by: Manivannan Sadhasivam --- drivers/pci/pcie/aspm.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/drivers/pci/pcie/aspm.c b/drivers/pci/pcie/aspm.c index 5eb462772354e..d7a3ca555cc13 100644 --- a/drivers/pci/pcie/aspm.c +++ b/drivers/pci/pcie/aspm.c @@ -1041,7 +1041,7 @@ static struct pcie_link_state *pcie_aspm_get_link(struct pci_dev *pdev) return bridge->link_state; } -static int __pci_disable_link_state(struct pci_dev *pdev, int state, bool sem) +static int __pci_disable_link_state(struct pci_dev *pdev, int state, bool locked) { struct pcie_link_state *link = pcie_aspm_get_link(pdev); @@ -1060,7 +1060,7 @@ static int __pci_disable_link_state(struct pci_dev *pdev, int state, bool sem) return -EPERM; } - if (sem) + if (!locked) down_read(&pci_bus_sem); mutex_lock(&aspm_lock); if (state & PCIE_LINK_STATE_L0S) @@ -1082,7 +1082,7 @@ static int __pci_disable_link_state(struct pci_dev *pdev, int state, bool sem) link->clkpm_disable = 1; pcie_set_clkpm(link, policy_to_clkpm_state(link)); mutex_unlock(&aspm_lock); - if (sem) + if (!locked) up_read(&pci_bus_sem); return 0; @@ -1090,7 +1090,7 @@ static int __pci_disable_link_state(struct pci_dev *pdev, int state, bool sem) int pci_disable_link_state_locked(struct pci_dev *pdev, int state) { - return __pci_disable_link_state(pdev, state, false); + return __pci_disable_link_state(pdev, state, true); } EXPORT_SYMBOL(pci_disable_link_state_locked); @@ -1105,7 +1105,7 @@ EXPORT_SYMBOL(pci_disable_link_state_locked); */ int pci_disable_link_state(struct pci_dev *pdev, int state) { - return __pci_disable_link_state(pdev, state, true); + return __pci_disable_link_state(pdev, state, false); } EXPORT_SYMBOL(pci_disable_link_state); From 7ff2b7a1821b61c324626ad57c3664398fb0083d Mon Sep 17 00:00:00 2001 From: Johan Hovold Date: Tue, 28 Nov 2023 09:15:12 +0100 Subject: [PATCH 7/8] PCI/ASPM: Add pci_disable_link_state_locked() lockdep assert Add a lockdep assert to pci_disable_link_state_locked() which should only be called with a pci_bus_sem read lock held. Link: https://lore.kernel.org/r/20231128081512.19387-7-johan+linaro@kernel.org Signed-off-by: Johan Hovold [bhelgaas: include function name in subject, commit log] Signed-off-by: Bjorn Helgaas Reviewed-by: Manivannan Sadhasivam --- drivers/pci/pcie/aspm.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/drivers/pci/pcie/aspm.c b/drivers/pci/pcie/aspm.c index d7a3ca555cc13..5dab531c8654d 100644 --- a/drivers/pci/pcie/aspm.c +++ b/drivers/pci/pcie/aspm.c @@ -1090,6 +1090,8 @@ static int __pci_disable_link_state(struct pci_dev *pdev, int state, bool locked int pci_disable_link_state_locked(struct pci_dev *pdev, int state) { + lockdep_assert_held_read(&pci_bus_sem); + return __pci_disable_link_state(pdev, state, true); } EXPORT_SYMBOL(pci_disable_link_state_locked); From 5df12742b7e3aae2594a30a9d14d5d6e9e7699f4 Mon Sep 17 00:00:00 2001 From: Bjorn Helgaas Date: Thu, 14 Dec 2023 09:08:56 -0600 Subject: [PATCH 8/8] Revert "PCI: acpiphp: Reassign resources on bridge if necessary" This reverts commit 40613da52b13fb21c5566f10b287e0ca8c12c4e9 and the subsequent fix to it: cc22522fd55e ("PCI: acpiphp: Use pci_assign_unassigned_bridge_resources() only for non-root bus") 40613da52b13 fixed a problem where hot-adding a device with large BARs failed if the bridge windows programmed by firmware were not large enough. cc22522fd55e ("PCI: acpiphp: Use pci_assign_unassigned_bridge_resources() only for non-root bus") fixed a problem with 40613da52b13: an ACPI hot-add of a device on a PCI root bus (common in the virt world) or firmware sending ACPI Bus Check to non-existent Root Ports (e.g., on Dell Inspiron 7352/0W6WV0) caused a NULL pointer dereference and suspend/resume hangs. Unfortunately the combination of 40613da52b13 and cc22522fd55e caused other problems: - Fiona reported that hot-add of SCSI disks in QEMU virtual machine fails sometimes. - Dongli reported a similar problem with hot-add of SCSI disks. - Jonathan reported a console freeze during boot on bare metal due to an error in radeon GPU initialization. Revert both patches to avoid adding these problems. This means we will again see the problems with hot-adding devices with large BARs and the NULL pointer dereferences and suspend/resume issues that 40613da52b13 and cc22522fd55e were intended to fix. Fixes: 40613da52b13 ("PCI: acpiphp: Reassign resources on bridge if necessary") Fixes: cc22522fd55e ("PCI: acpiphp: Use pci_assign_unassigned_bridge_resources() only for non-root bus") Reported-by: Fiona Ebner Closes: https://lore.kernel.org/r/9eb669c0-d8f2-431d-a700-6da13053ae54@proxmox.com Reported-by: Dongli Zhang Closes: https://lore.kernel.org/r/3c4a446a-b167-11b8-f36f-d3c1b49b42e9@oracle.com Reported-by: Jonathan Woithe Closes: https://lore.kernel.org/r/ZXpaNCLiDM+Kv38H@marvin.atrad.com.au Signed-off-by: Bjorn Helgaas Acked-by: Michael S. Tsirkin Acked-by: Igor Mammedov Cc: --- drivers/pci/hotplug/acpiphp_glue.c | 9 +++------ 1 file changed, 3 insertions(+), 6 deletions(-) diff --git a/drivers/pci/hotplug/acpiphp_glue.c b/drivers/pci/hotplug/acpiphp_glue.c index 601129772b2d5..5b1f271c6034b 100644 --- a/drivers/pci/hotplug/acpiphp_glue.c +++ b/drivers/pci/hotplug/acpiphp_glue.c @@ -512,15 +512,12 @@ static void enable_slot(struct acpiphp_slot *slot, bool bridge) if (pass && dev->subordinate) { check_hotplug_bridge(slot, dev); pcibios_resource_survey_bus(dev->subordinate); - if (pci_is_root_bus(bus)) - __pci_bus_size_bridges(dev->subordinate, &add_list); + __pci_bus_size_bridges(dev->subordinate, + &add_list); } } } - if (pci_is_root_bus(bus)) - __pci_bus_assign_resources(bus, &add_list, NULL); - else - pci_assign_unassigned_bridge_resources(bus->self); + __pci_bus_assign_resources(bus, &add_list, NULL); } acpiphp_sanitize_bus(bus);