drm/amdgpu: add VRAM manager v2

Split VRAM allocations into 4MB blocks.

v2: fix typo in comment, some suggested cleanups
v3: document how to disable the feature, fix rebase issue

Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Edward O'Callaghan <funfunctor@folklore1984.net>
Tested-by: Mike Lothian <mike@fireburn.co.uk>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/amd/amdgpu/Makefile

@@ -24,7 +24,7 @@ amdgpu-y += amdgpu_device.o amdgpu_kms.o \
 	atombios_encoders.o amdgpu_sa.o atombios_i2c.o \
 	amdgpu_prime.o amdgpu_vm.o amdgpu_ib.o amdgpu_pll.o \
 	amdgpu_ucode.o amdgpu_bo_list.o amdgpu_ctx.o amdgpu_sync.o \
-	amdgpu_gtt_mgr.o
+	amdgpu_gtt_mgr.o amdgpu_vram_mgr.o
 
 # add asic specific block
 amdgpu-$(CONFIG_DRM_AMDGPU_CIK)+= cik.o cik_ih.o kv_smc.o kv_dpm.o \

drivers/gpu/drm/amd/amdgpu/amdgpu.h

@@ -97,6 +97,7 @@ extern char *amdgpu_disable_cu;
 extern int amdgpu_sclk_deep_sleep_en;
 extern char *amdgpu_virtual_display;
 extern unsigned amdgpu_pp_feature_mask;
+extern int amdgpu_vram_page_split;
 
 #define AMDGPU_WAIT_IDLE_TIMEOUT_IN_MS	        3000
 #define AMDGPU_MAX_USEC_TIMEOUT			100000	/* 100 ms */

drivers/gpu/drm/amd/amdgpu/amdgpu_device.c

@@ -1039,6 +1039,13 @@ static void amdgpu_check_arguments(struct amdgpu_device *adev)
 			 amdgpu_vm_block_size);
 		amdgpu_vm_block_size = 9;
 	}
+
+	if ((amdgpu_vram_page_split != -1 && amdgpu_vram_page_split < 16) ||
+	    !amdgpu_check_pot_argument(amdgpu_vram_page_split)) {
+		dev_warn(adev->dev, "invalid VRAM page split (%d)\n",
+			 amdgpu_vram_page_split);
+		amdgpu_vram_page_split = 1024;
+	}
 }
 
 /**

drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c

@@ -85,6 +85,7 @@ int amdgpu_vm_size = 64;
 int amdgpu_vm_block_size = -1;
 int amdgpu_vm_fault_stop = 0;
 int amdgpu_vm_debug = 0;
+int amdgpu_vram_page_split = 1024;
 int amdgpu_exp_hw_support = 0;
 int amdgpu_sched_jobs = 32;
 int amdgpu_sched_hw_submission = 2;
@@ -165,6 +166,9 @@ module_param_named(vm_fault_stop, amdgpu_vm_fault_stop, int, 0444);
 MODULE_PARM_DESC(vm_debug, "Debug VM handling (0 = disabled (default), 1 = enabled)");
 module_param_named(vm_debug, amdgpu_vm_debug, int, 0644);
 
+MODULE_PARM_DESC(vram_page_split, "Number of pages after we split VRAM allocations (default 1024, -1 = disable)");
+module_param_named(vram_page_split, amdgpu_vram_page_split, int, 0444);
+
 MODULE_PARM_DESC(exp_hw_support, "experimental hw support (1 = enable, 0 = disable (default))");
 module_param_named(exp_hw_support, amdgpu_exp_hw_support, int, 0444);
 

drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c

@@ -168,7 +168,7 @@ static int amdgpu_init_mem_type(struct ttm_bo_device *bdev, uint32_t type,
 		break;
 	case TTM_PL_VRAM:
 		/* "On-card" video ram */
-		man->func = &ttm_bo_manager_func;
+		man->func = &amdgpu_vram_mgr_func;
 		man->gpu_offset = adev->mc.vram_start;
 		man->flags = TTM_MEMTYPE_FLAG_FIXED |
 			     TTM_MEMTYPE_FLAG_MAPPABLE;

drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h

@@ -66,6 +66,7 @@ struct amdgpu_mman {
 };
 
 extern const struct ttm_mem_type_manager_func amdgpu_gtt_mgr_func;
+extern const struct ttm_mem_type_manager_func amdgpu_vram_mgr_func;
 
 int amdgpu_gtt_mgr_alloc(struct ttm_mem_type_manager *man,
 			 struct ttm_buffer_object *tbo,

drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c

@@ -0,0 +1,222 @@
+/*
+ * Copyright 2016 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Authors: Christian König
+ */
+
+#include <drm/drmP.h>
+#include "amdgpu.h"
+
+struct amdgpu_vram_mgr {
+	struct drm_mm mm;
+	spinlock_t lock;
+};
+
+/**
+ * amdgpu_vram_mgr_init - init VRAM manager and DRM MM
+ *
+ * @man: TTM memory type manager
+ * @p_size: maximum size of VRAM
+ *
+ * Allocate and initialize the VRAM manager.
+ */
+static int amdgpu_vram_mgr_init(struct ttm_mem_type_manager *man,
+				unsigned long p_size)
+{
+	struct amdgpu_vram_mgr *mgr;
+
+	mgr = kzalloc(sizeof(*mgr), GFP_KERNEL);
+	if (!mgr)
+		return -ENOMEM;
+
+	drm_mm_init(&mgr->mm, 0, p_size);
+	spin_lock_init(&mgr->lock);
+	man->priv = mgr;
+	return 0;
+}
+
+/**
+ * amdgpu_vram_mgr_fini - free and destroy VRAM manager
+ *
+ * @man: TTM memory type manager
+ *
+ * Destroy and free the VRAM manager, returns -EBUSY if ranges are still
+ * allocated inside it.
+ */
+static int amdgpu_vram_mgr_fini(struct ttm_mem_type_manager *man)
+{
+	struct amdgpu_vram_mgr *mgr = man->priv;
+
+	spin_lock(&mgr->lock);
+	if (!drm_mm_clean(&mgr->mm)) {
+		spin_unlock(&mgr->lock);
+		return -EBUSY;
+	}
+
+	drm_mm_takedown(&mgr->mm);
+	spin_unlock(&mgr->lock);
+	kfree(mgr);
+	man->priv = NULL;
+	return 0;
+}
+
+/**
+ * amdgpu_vram_mgr_new - allocate new ranges
+ *
+ * @man: TTM memory type manager
+ * @tbo: TTM BO we need this range for
+ * @place: placement flags and restrictions
+ * @mem: the resulting mem object
+ *
+ * Allocate VRAM for the given BO.
+ */
+static int amdgpu_vram_mgr_new(struct ttm_mem_type_manager *man,
+			       struct ttm_buffer_object *tbo,
+			       const struct ttm_place *place,
+			       struct ttm_mem_reg *mem)
+{
+	struct amdgpu_bo *bo = container_of(tbo, struct amdgpu_bo, tbo);
+	struct amdgpu_vram_mgr *mgr = man->priv;
+	struct drm_mm *mm = &mgr->mm;
+	struct drm_mm_node *nodes;
+	enum drm_mm_search_flags sflags = DRM_MM_SEARCH_DEFAULT;
+	enum drm_mm_allocator_flags aflags = DRM_MM_CREATE_DEFAULT;
+	unsigned long lpfn, num_nodes, pages_per_node, pages_left;
+	unsigned i;
+	int r;
+
+	lpfn = place->lpfn;
+	if (!lpfn)
+		lpfn = man->size;
+
+	if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS ||
+	    amdgpu_vram_page_split == -1) {
+		pages_per_node = ~0ul;
+		num_nodes = 1;
+	} else {
+		pages_per_node = max((uint32_t)amdgpu_vram_page_split,
+				     mem->page_alignment);
+		num_nodes = DIV_ROUND_UP(mem->num_pages, pages_per_node);
+	}
+
+	nodes = kcalloc(num_nodes, sizeof(*nodes), GFP_KERNEL);
+	if (!nodes)
+		return -ENOMEM;
+
+	if (place->flags & TTM_PL_FLAG_TOPDOWN) {
+		sflags = DRM_MM_SEARCH_BELOW;
+		aflags = DRM_MM_CREATE_TOP;
+	}
+
+	pages_left = mem->num_pages;
+
+	spin_lock(&mgr->lock);
+	for (i = 0; i < num_nodes; ++i) {
+		unsigned long pages = min(pages_left, pages_per_node);
+		uint32_t alignment = mem->page_alignment;
+
+		if (pages == pages_per_node)
+			alignment = pages_per_node;
+		else
+			sflags |= DRM_MM_SEARCH_BEST;
+
+		r = drm_mm_insert_node_in_range_generic(mm, &nodes[i], pages,
+							alignment, 0,
+							place->fpfn, lpfn,
+							sflags, aflags);
+		if (unlikely(r))
+			goto error;
+
+		pages_left -= pages;
+	}
+	spin_unlock(&mgr->lock);
+
+	mem->start = num_nodes == 1 ? nodes[0].start : AMDGPU_BO_INVALID_OFFSET;
+	mem->mm_node = nodes;
+
+	return 0;
+
+error:
+	while (i--)
+		drm_mm_remove_node(&nodes[i]);
+	spin_unlock(&mgr->lock);
+
+	kfree(nodes);
+	return r == -ENOSPC ? 0 : r;
+}
+
+/**
+ * amdgpu_vram_mgr_del - free ranges
+ *
+ * @man: TTM memory type manager
+ * @tbo: TTM BO we need this range for
+ * @place: placement flags and restrictions
+ * @mem: TTM memory object
+ *
+ * Free the allocated VRAM again.
+ */
+static void amdgpu_vram_mgr_del(struct ttm_mem_type_manager *man,
+				struct ttm_mem_reg *mem)
+{
+	struct amdgpu_vram_mgr *mgr = man->priv;
+	struct drm_mm_node *nodes = mem->mm_node;
+	unsigned pages = mem->num_pages;
+
+	if (!mem->mm_node)
+		return;
+
+	spin_lock(&mgr->lock);
+	while (pages) {
+		pages -= nodes->size;
+		drm_mm_remove_node(nodes);
+		++nodes;
+	}
+	spin_unlock(&mgr->lock);
+
+	kfree(mem->mm_node);
+	mem->mm_node = NULL;
+}
+
+/**
+ * amdgpu_vram_mgr_debug - dump VRAM table
+ *
+ * @man: TTM memory type manager
+ * @prefix: text prefix
+ *
+ * Dump the table content using printk.
+ */
+static void amdgpu_vram_mgr_debug(struct ttm_mem_type_manager *man,
+				  const char *prefix)
+{
+	struct amdgpu_vram_mgr *mgr = man->priv;
+
+	spin_lock(&mgr->lock);
+	drm_mm_debug_table(&mgr->mm, prefix);
+	spin_unlock(&mgr->lock);
+}
+
+const struct ttm_mem_type_manager_func amdgpu_vram_mgr_func = {
+	amdgpu_vram_mgr_init,
+	amdgpu_vram_mgr_fini,
+	amdgpu_vram_mgr_new,
+	amdgpu_vram_mgr_del,
+	amdgpu_vram_mgr_debug
+};