[PATCH 1/5] drm/ttm: use an operation ctx for ttm_mem_global_alloc_page

[PATCH 1/5] drm/ttm: use an operation ctx for ttm_mem_global_alloc_page

[Roger He]

forward the operation context to ttm_mem_global_alloc_page as well,
and the ultimate goal is swapout enablement for per vm BOs.

Per vm BOs refer to all BOs which share same reservation object.

Change-Id: I4104a12e09a374b6477a0dd5a8fce26dce27a746
Signed-off-by: Roger He <Hongbo.He@amd.com>
---
 drivers/gpu/drm/ttm/ttm_memory.c         | 10 +++-------
 drivers/gpu/drm/ttm/ttm_page_alloc.c     |  6 +++++-
 drivers/gpu/drm/ttm/ttm_page_alloc_dma.c |  8 ++++++--
 include/drm/ttm/ttm_memory.h             |  3 ++-
 4 files changed, 16 insertions(+), 11 deletions(-)

diff --git a/drivers/gpu/drm/ttm/ttm_memory.c b/drivers/gpu/drm/ttm/ttm_memory.c
index 525d3b6..102b326 100644
--- a/drivers/gpu/drm/ttm/ttm_memory.c
+++ b/drivers/gpu/drm/ttm/ttm_memory.c
@@ -539,14 +539,10 @@ int ttm_mem_global_alloc(struct ttm_mem_global *glob, uint64_t memory,
 EXPORT_SYMBOL(ttm_mem_global_alloc);
 
 int ttm_mem_global_alloc_page(struct ttm_mem_global *glob,
-			      struct page *page, uint64_t size)
+			      struct page *page, uint64_t size,
+			      struct ttm_operation_ctx *ctx)
 {
-
 	struct ttm_mem_zone *zone = NULL;
-	struct ttm_operation_ctx ctx = {
-		.interruptible = false,
-		.no_wait_gpu = false
-	};
 
 	/**
 	 * Page allocations may be registed in a single zone
@@ -560,7 +556,7 @@ int ttm_mem_global_alloc_page(struct ttm_mem_global *glob,
 	if (glob->zone_dma32 && page_to_pfn(page) > 0x00100000UL)
 		zone = glob->zone_kernel;
 #endif
-	return ttm_mem_global_alloc_zone(glob, zone, size, &ctx);
+	return ttm_mem_global_alloc_zone(glob, zone, size, ctx);
 }
 
 void ttm_mem_global_free_page(struct ttm_mem_global *glob, struct page *page,
diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc.c b/drivers/gpu/drm/ttm/ttm_page_alloc.c
index b5ba644..8f93ff3 100644
--- a/drivers/gpu/drm/ttm/ttm_page_alloc.c
+++ b/drivers/gpu/drm/ttm/ttm_page_alloc.c
@@ -1061,6 +1061,10 @@ void ttm_page_alloc_fini(void)
 int ttm_pool_populate(struct ttm_tt *ttm)
 {
 	struct ttm_mem_global *mem_glob = ttm->glob->mem_glob;
+	struct ttm_operation_ctx ctx = {
+		.interruptible = false,
+		.no_wait_gpu = false
+	};
 	unsigned i;
 	int ret;
 
@@ -1076,7 +1080,7 @@ int ttm_pool_populate(struct ttm_tt *ttm)
 
 	for (i = 0; i < ttm->num_pages; ++i) {
 		ret = ttm_mem_global_alloc_page(mem_glob, ttm->pages[i],
-						PAGE_SIZE);
+						PAGE_SIZE, &ctx);
 		if (unlikely(ret != 0)) {
 			ttm_pool_unpopulate(ttm);
 			return -ENOMEM;
diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c b/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c
index bda00b2..8aac86a 100644
--- a/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c
+++ b/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c
@@ -927,6 +927,10 @@ int ttm_dma_populate(struct ttm_dma_tt *ttm_dma, struct device *dev)
 {
 	struct ttm_tt *ttm = &ttm_dma->ttm;
 	struct ttm_mem_global *mem_glob = ttm->glob->mem_glob;
+	struct ttm_operation_ctx ctx = {
+		.interruptible = false,
+		.no_wait_gpu = false
+	};
 	unsigned long num_pages = ttm->num_pages;
 	struct dma_pool *pool;
 	enum pool_type type;
@@ -962,7 +966,7 @@ int ttm_dma_populate(struct ttm_dma_tt *ttm_dma, struct device *dev)
 			break;
 
 		ret = ttm_mem_global_alloc_page(mem_glob, ttm->pages[i],
-						pool->size);
+						pool->size, &ctx);
 		if (unlikely(ret != 0)) {
 			ttm_dma_unpopulate(ttm_dma, dev);
 			return -ENOMEM;
@@ -998,7 +1002,7 @@ int ttm_dma_populate(struct ttm_dma_tt *ttm_dma, struct device *dev)
 		}
 
 		ret = ttm_mem_global_alloc_page(mem_glob, ttm->pages[i],
-						pool->size);
+						pool->size, &ctx);
 		if (unlikely(ret != 0)) {
 			ttm_dma_unpopulate(ttm_dma, dev);
 			return -ENOMEM;
diff --git a/include/drm/ttm/ttm_memory.h b/include/drm/ttm/ttm_memory.h
index 755c107..8936285 100644
--- a/include/drm/ttm/ttm_memory.h
+++ b/include/drm/ttm/ttm_memory.h
@@ -84,7 +84,8 @@ extern int ttm_mem_global_alloc(struct ttm_mem_global *glob, uint64_t memory,
 extern void ttm_mem_global_free(struct ttm_mem_global *glob,
 				uint64_t amount);
 extern int ttm_mem_global_alloc_page(struct ttm_mem_global *glob,
-				     struct page *page, uint64_t size);
+				     struct page *page, uint64_t size,
+				     struct ttm_operation_ctx *ctx);
 extern void ttm_mem_global_free_page(struct ttm_mem_global *glob,
 				     struct page *page, uint64_t size);
 extern size_t ttm_round_pot(size_t size);
-- 
2.7.4

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

[PATCH 2/5] drm/ttm: use an operation ctx for ttm_tt_populate in ttm_bo_driver

[Roger He]

forward the operation context to ttm_tt_populate as well,
and the ultimate goal is swapout enablement for per vm BOs.

Change-Id: If8dfa0f500429d1420e0da67eb6901f0bfbca57b
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Roger He <Hongbo.He@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c     |  7 ++++---
 drivers/gpu/drm/ast/ast_ttm.c               |  5 +++--
 drivers/gpu/drm/cirrus/cirrus_ttm.c         |  5 +++--
 drivers/gpu/drm/hisilicon/hibmc/hibmc_ttm.c |  5 +++--
 drivers/gpu/drm/mgag200/mgag200_ttm.c       |  5 +++--
 drivers/gpu/drm/nouveau/nouveau_bo.c        |  8 ++++----
 drivers/gpu/drm/qxl/qxl_ttm.c               |  5 +++--
 drivers/gpu/drm/radeon/radeon_ttm.c         |  9 +++++----
 drivers/gpu/drm/ttm/ttm_agp_backend.c       |  4 ++--
 drivers/gpu/drm/ttm/ttm_bo_util.c           | 11 ++++++++---
 drivers/gpu/drm/ttm/ttm_bo_vm.c             |  7 ++++++-
 drivers/gpu/drm/ttm/ttm_page_alloc.c        | 13 +++++--------
 drivers/gpu/drm/ttm/ttm_page_alloc_dma.c    | 11 ++++-------
 drivers/gpu/drm/ttm/ttm_tt.c                |  6 +++++-
 drivers/gpu/drm/virtio/virtgpu_object.c     |  6 +++++-
 drivers/gpu/drm/virtio/virtgpu_ttm.c        |  5 +++--
 drivers/gpu/drm/vmwgfx/vmwgfx_buffer.c      | 13 +++++--------
 drivers/gpu/drm/vmwgfx/vmwgfx_mob.c         | 13 +++++++++++--
 include/drm/ttm/ttm_bo_driver.h             |  5 +++--
 include/drm/ttm/ttm_page_alloc.h            | 11 +++++++----
 20 files changed, 92 insertions(+), 62 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index f1b7d98..044f5b5 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -990,7 +990,8 @@ static struct ttm_tt *amdgpu_ttm_tt_create(struct ttm_bo_device *bdev,
 	return &gtt->ttm.ttm;
 }
 
-static int amdgpu_ttm_tt_populate(struct ttm_tt *ttm)
+static int amdgpu_ttm_tt_populate(struct ttm_tt *ttm,
+			struct ttm_operation_ctx *ctx)
 {
 	struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev);
 	struct amdgpu_ttm_tt *gtt = (void *)ttm;
@@ -1018,11 +1019,11 @@ static int amdgpu_ttm_tt_populate(struct ttm_tt *ttm)
 
 #ifdef CONFIG_SWIOTLB
 	if (swiotlb_nr_tbl()) {
-		return ttm_dma_populate(&gtt->ttm, adev->dev);
+		return ttm_dma_populate(&gtt->ttm, adev->dev, ctx);
 	}
 #endif
 
-	return ttm_populate_and_map_pages(adev->dev, &gtt->ttm);
+	return ttm_populate_and_map_pages(adev->dev, &gtt->ttm, ctx);
 }
 
 static void amdgpu_ttm_tt_unpopulate(struct ttm_tt *ttm)
diff --git a/drivers/gpu/drm/ast/ast_ttm.c b/drivers/gpu/drm/ast/ast_ttm.c
index 28da7c2..1413e94 100644
--- a/drivers/gpu/drm/ast/ast_ttm.c
+++ b/drivers/gpu/drm/ast/ast_ttm.c
@@ -216,9 +216,10 @@ static struct ttm_tt *ast_ttm_tt_create(struct ttm_bo_device *bdev,
 	return tt;
 }
 
-static int ast_ttm_tt_populate(struct ttm_tt *ttm)
+static int ast_ttm_tt_populate(struct ttm_tt *ttm,
+			struct ttm_operation_ctx *ctx)
 {
-	return ttm_pool_populate(ttm);
+	return ttm_pool_populate(ttm, ctx);
 }
 
 static void ast_ttm_tt_unpopulate(struct ttm_tt *ttm)
diff --git a/drivers/gpu/drm/cirrus/cirrus_ttm.c b/drivers/gpu/drm/cirrus/cirrus_ttm.c
index 2a5b54d..65ae960 100644
--- a/drivers/gpu/drm/cirrus/cirrus_ttm.c
+++ b/drivers/gpu/drm/cirrus/cirrus_ttm.c
@@ -216,9 +216,10 @@ static struct ttm_tt *cirrus_ttm_tt_create(struct ttm_bo_device *bdev,
 	return tt;
 }
 
-static int cirrus_ttm_tt_populate(struct ttm_tt *ttm)
+static int cirrus_ttm_tt_populate(struct ttm_tt *ttm,
+		struct ttm_operation_ctx *ctx)
 {
-	return ttm_pool_populate(ttm);
+	return ttm_pool_populate(ttm, ctx);
 }
 
 static void cirrus_ttm_tt_unpopulate(struct ttm_tt *ttm)
diff --git a/drivers/gpu/drm/hisilicon/hibmc/hibmc_ttm.c b/drivers/gpu/drm/hisilicon/hibmc/hibmc_ttm.c
index ab4ee59..8516e00 100644
--- a/drivers/gpu/drm/hisilicon/hibmc/hibmc_ttm.c
+++ b/drivers/gpu/drm/hisilicon/hibmc/hibmc_ttm.c
@@ -223,9 +223,10 @@ static struct ttm_tt *hibmc_ttm_tt_create(struct ttm_bo_device *bdev,
 	return tt;
 }
 
-static int hibmc_ttm_tt_populate(struct ttm_tt *ttm)
+static int hibmc_ttm_tt_populate(struct ttm_tt *ttm,
+		struct ttm_operation_ctx *ctx)
 {
-	return ttm_pool_populate(ttm);
+	return ttm_pool_populate(ttm, ctx);
 }
 
 static void hibmc_ttm_tt_unpopulate(struct ttm_tt *ttm)
diff --git a/drivers/gpu/drm/mgag200/mgag200_ttm.c b/drivers/gpu/drm/mgag200/mgag200_ttm.c
index f03da63..6fa8076 100644
--- a/drivers/gpu/drm/mgag200/mgag200_ttm.c
+++ b/drivers/gpu/drm/mgag200/mgag200_ttm.c
@@ -216,9 +216,10 @@ static struct ttm_tt *mgag200_ttm_tt_create(struct ttm_bo_device *bdev,
 	return tt;
 }
 
-static int mgag200_ttm_tt_populate(struct ttm_tt *ttm)
+static int mgag200_ttm_tt_populate(struct ttm_tt *ttm,
+			struct ttm_operation_ctx *ctx)
 {
-	return ttm_pool_populate(ttm);
+	return ttm_pool_populate(ttm, ctx);
 }
 
 static void mgag200_ttm_tt_unpopulate(struct ttm_tt *ttm)
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
index 6b6fb20..b141c27 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -1547,7 +1547,7 @@ nouveau_ttm_fault_reserve_notify(struct ttm_buffer_object *bo)
 }
 
 static int
-nouveau_ttm_tt_populate(struct ttm_tt *ttm)
+nouveau_ttm_tt_populate(struct ttm_tt *ttm, struct ttm_operation_ctx *ctx)
 {
 	struct ttm_dma_tt *ttm_dma = (void *)ttm;
 	struct nouveau_drm *drm;
@@ -1572,17 +1572,17 @@ nouveau_ttm_tt_populate(struct ttm_tt *ttm)
 
 #if IS_ENABLED(CONFIG_AGP)
 	if (drm->agp.bridge) {
-		return ttm_agp_tt_populate(ttm);
+		return ttm_agp_tt_populate(ttm, ctx);
 	}
 #endif
 
 #if IS_ENABLED(CONFIG_SWIOTLB) && IS_ENABLED(CONFIG_X86)
 	if (swiotlb_nr_tbl()) {
-		return ttm_dma_populate((void *)ttm, dev);
+		return ttm_dma_populate((void *)ttm, dev, ctx);
 	}
 #endif
 
-	r = ttm_pool_populate(ttm);
+	r = ttm_pool_populate(ttm, ctx);
 	if (r) {
 		return r;
 	}
diff --git a/drivers/gpu/drm/qxl/qxl_ttm.c b/drivers/gpu/drm/qxl/qxl_ttm.c
index 78ce118..f72ce3b 100644
--- a/drivers/gpu/drm/qxl/qxl_ttm.c
+++ b/drivers/gpu/drm/qxl/qxl_ttm.c
@@ -291,14 +291,15 @@ static struct ttm_backend_func qxl_backend_func = {
 	.destroy = &qxl_ttm_backend_destroy,
 };
 
-static int qxl_ttm_tt_populate(struct ttm_tt *ttm)
+static int qxl_ttm_tt_populate(struct ttm_tt *ttm,
+			struct ttm_operation_ctx *ctx)
 {
 	int r;
 
 	if (ttm->state != tt_unpopulated)
 		return 0;
 
-	r = ttm_pool_populate(ttm);
+	r = ttm_pool_populate(ttm, ctx);
 	if (r)
 		return r;
 
diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c
index 557fd79..9424b81 100644
--- a/drivers/gpu/drm/radeon/radeon_ttm.c
+++ b/drivers/gpu/drm/radeon/radeon_ttm.c
@@ -721,7 +721,8 @@ static struct radeon_ttm_tt *radeon_ttm_tt_to_gtt(struct ttm_tt *ttm)
 	return (struct radeon_ttm_tt *)ttm;
 }
 
-static int radeon_ttm_tt_populate(struct ttm_tt *ttm)
+static int radeon_ttm_tt_populate(struct ttm_tt *ttm,
+			struct ttm_operation_ctx *ctx)
 {
 	struct radeon_ttm_tt *gtt = radeon_ttm_tt_to_gtt(ttm);
 	struct radeon_device *rdev;
@@ -750,17 +751,17 @@ static int radeon_ttm_tt_populate(struct ttm_tt *ttm)
 	rdev = radeon_get_rdev(ttm->bdev);
 #if IS_ENABLED(CONFIG_AGP)
 	if (rdev->flags & RADEON_IS_AGP) {
-		return ttm_agp_tt_populate(ttm);
+		return ttm_agp_tt_populate(ttm, ctx);
 	}
 #endif
 
 #ifdef CONFIG_SWIOTLB
 	if (swiotlb_nr_tbl()) {
-		return ttm_dma_populate(&gtt->ttm, rdev->dev);
+		return ttm_dma_populate(&gtt->ttm, rdev->dev, ctx);
 	}
 #endif
 
-	return ttm_populate_and_map_pages(rdev->dev, &gtt->ttm);
+	return ttm_populate_and_map_pages(rdev->dev, &gtt->ttm, ctx);
 }
 
 static void radeon_ttm_tt_unpopulate(struct ttm_tt *ttm)
diff --git a/drivers/gpu/drm/ttm/ttm_agp_backend.c b/drivers/gpu/drm/ttm/ttm_agp_backend.c
index 028ab60..3e795a0 100644
--- a/drivers/gpu/drm/ttm/ttm_agp_backend.c
+++ b/drivers/gpu/drm/ttm/ttm_agp_backend.c
@@ -133,12 +133,12 @@ struct ttm_tt *ttm_agp_tt_create(struct ttm_bo_device *bdev,
 }
 EXPORT_SYMBOL(ttm_agp_tt_create);
 
-int ttm_agp_tt_populate(struct ttm_tt *ttm)
+int ttm_agp_tt_populate(struct ttm_tt *ttm, struct ttm_operation_ctx *ctx)
 {
 	if (ttm->state != tt_unpopulated)
 		return 0;
 
-	return ttm_pool_populate(ttm);
+	return ttm_pool_populate(ttm, ctx);
 }
 EXPORT_SYMBOL(ttm_agp_tt_populate);
 
diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c
index 6e353df..b7eb507 100644
--- a/drivers/gpu/drm/ttm/ttm_bo_util.c
+++ b/drivers/gpu/drm/ttm/ttm_bo_util.c
@@ -376,7 +376,7 @@ int ttm_bo_move_memcpy(struct ttm_buffer_object *bo,
 	 * TTM might be null for moves within the same region.
 	 */
 	if (ttm && ttm->state == tt_unpopulated) {
-		ret = ttm->bdev->driver->ttm_tt_populate(ttm);
+		ret = ttm->bdev->driver->ttm_tt_populate(ttm, ctx);
 		if (ret)
 			goto out1;
 	}
@@ -545,14 +545,19 @@ static int ttm_bo_kmap_ttm(struct ttm_buffer_object *bo,
 			   unsigned long num_pages,
 			   struct ttm_bo_kmap_obj *map)
 {
-	struct ttm_mem_reg *mem = &bo->mem; pgprot_t prot;
+	struct ttm_mem_reg *mem = &bo->mem;
+	struct ttm_operation_ctx ctx = {
+		.interruptible = false,
+		.no_wait_gpu = false
+	};
 	struct ttm_tt *ttm = bo->ttm;
+	pgprot_t prot;
 	int ret;
 
 	BUG_ON(!ttm);
 
 	if (ttm->state == tt_unpopulated) {
-		ret = ttm->bdev->driver->ttm_tt_populate(ttm);
+		ret = ttm->bdev->driver->ttm_tt_populate(ttm, &ctx);
 		if (ret)
 			return ret;
 	}
diff --git a/drivers/gpu/drm/ttm/ttm_bo_vm.c b/drivers/gpu/drm/ttm/ttm_bo_vm.c
index c8ebb75..65dfcdd 100644
--- a/drivers/gpu/drm/ttm/ttm_bo_vm.c
+++ b/drivers/gpu/drm/ttm/ttm_bo_vm.c
@@ -215,12 +215,17 @@ static int ttm_bo_vm_fault(struct vm_fault *vmf)
 		cvma.vm_page_prot = ttm_io_prot(bo->mem.placement,
 						cvma.vm_page_prot);
 	} else {
+		struct ttm_operation_ctx ctx = {
+			.interruptible = false,
+			.no_wait_gpu = false
+		};
+
 		ttm = bo->ttm;
 		cvma.vm_page_prot = ttm_io_prot(bo->mem.placement,
 						cvma.vm_page_prot);
 
 		/* Allocate all page at once, most common usage */
-		if (ttm->bdev->driver->ttm_tt_populate(ttm)) {
+		if (ttm->bdev->driver->ttm_tt_populate(ttm, &ctx)) {
 			retval = VM_FAULT_OOM;
 			goto out_io_unlock;
 		}
diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc.c b/drivers/gpu/drm/ttm/ttm_page_alloc.c
index 8f93ff3..f1a3d55 100644
--- a/drivers/gpu/drm/ttm/ttm_page_alloc.c
+++ b/drivers/gpu/drm/ttm/ttm_page_alloc.c
@@ -1058,13 +1058,9 @@ void ttm_page_alloc_fini(void)
 	_manager = NULL;
 }
 
-int ttm_pool_populate(struct ttm_tt *ttm)
+int ttm_pool_populate(struct ttm_tt *ttm, struct ttm_operation_ctx *ctx)
 {
 	struct ttm_mem_global *mem_glob = ttm->glob->mem_glob;
-	struct ttm_operation_ctx ctx = {
-		.interruptible = false,
-		.no_wait_gpu = false
-	};
 	unsigned i;
 	int ret;
 
@@ -1080,7 +1076,7 @@ int ttm_pool_populate(struct ttm_tt *ttm)
 
 	for (i = 0; i < ttm->num_pages; ++i) {
 		ret = ttm_mem_global_alloc_page(mem_glob, ttm->pages[i],
-						PAGE_SIZE, &ctx);
+						PAGE_SIZE, ctx);
 		if (unlikely(ret != 0)) {
 			ttm_pool_unpopulate(ttm);
 			return -ENOMEM;
@@ -1117,12 +1113,13 @@ void ttm_pool_unpopulate(struct ttm_tt *ttm)
 }
 EXPORT_SYMBOL(ttm_pool_unpopulate);
 
-int ttm_populate_and_map_pages(struct device *dev, struct ttm_dma_tt *tt)
+int ttm_populate_and_map_pages(struct device *dev, struct ttm_dma_tt *tt,
+					struct ttm_operation_ctx *ctx)
 {
 	unsigned i, j;
 	int r;
 
-	r = ttm_pool_populate(&tt->ttm);
+	r = ttm_pool_populate(&tt->ttm, ctx);
 	if (r)
 		return r;
 
diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c b/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c
index 8aac86a..3ac5391 100644
--- a/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c
+++ b/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c
@@ -923,14 +923,11 @@ static gfp_t ttm_dma_pool_gfp_flags(struct ttm_dma_tt *ttm_dma, bool huge)
  * On success pages list will hold count number of correctly
  * cached pages. On failure will hold the negative return value (-ENOMEM, etc).
  */
-int ttm_dma_populate(struct ttm_dma_tt *ttm_dma, struct device *dev)
+int ttm_dma_populate(struct ttm_dma_tt *ttm_dma, struct device *dev,
+			struct ttm_operation_ctx *ctx)
 {
 	struct ttm_tt *ttm = &ttm_dma->ttm;
 	struct ttm_mem_global *mem_glob = ttm->glob->mem_glob;
-	struct ttm_operation_ctx ctx = {
-		.interruptible = false,
-		.no_wait_gpu = false
-	};
 	unsigned long num_pages = ttm->num_pages;
 	struct dma_pool *pool;
 	enum pool_type type;
@@ -966,7 +963,7 @@ int ttm_dma_populate(struct ttm_dma_tt *ttm_dma, struct device *dev)
 			break;
 
 		ret = ttm_mem_global_alloc_page(mem_glob, ttm->pages[i],
-						pool->size, &ctx);
+						pool->size, ctx);
 		if (unlikely(ret != 0)) {
 			ttm_dma_unpopulate(ttm_dma, dev);
 			return -ENOMEM;
@@ -1002,7 +999,7 @@ int ttm_dma_populate(struct ttm_dma_tt *ttm_dma, struct device *dev)
 		}
 
 		ret = ttm_mem_global_alloc_page(mem_glob, ttm->pages[i],
-						pool->size, &ctx);
+						pool->size, ctx);
 		if (unlikely(ret != 0)) {
 			ttm_dma_unpopulate(ttm_dma, dev);
 			return -ENOMEM;
diff --git a/drivers/gpu/drm/ttm/ttm_tt.c b/drivers/gpu/drm/ttm/ttm_tt.c
index 8ebc8d3..b48d7a0 100644
--- a/drivers/gpu/drm/ttm/ttm_tt.c
+++ b/drivers/gpu/drm/ttm/ttm_tt.c
@@ -263,6 +263,10 @@ void ttm_tt_unbind(struct ttm_tt *ttm)
 
 int ttm_tt_bind(struct ttm_tt *ttm, struct ttm_mem_reg *bo_mem)
 {
+	struct ttm_operation_ctx ctx = {
+		.interruptible = false,
+		.no_wait_gpu = false
+	};
 	int ret = 0;
 
 	if (!ttm)
@@ -271,7 +275,7 @@ int ttm_tt_bind(struct ttm_tt *ttm, struct ttm_mem_reg *bo_mem)
 	if (ttm->state == tt_bound)
 		return 0;
 
-	ret = ttm->bdev->driver->ttm_tt_populate(ttm);
+	ret = ttm->bdev->driver->ttm_tt_populate(ttm, &ctx);
 	if (ret)
 		return ret;
 
diff --git a/drivers/gpu/drm/virtio/virtgpu_object.c b/drivers/gpu/drm/virtio/virtgpu_object.c
index 6f66b73..0b90cdb 100644
--- a/drivers/gpu/drm/virtio/virtgpu_object.c
+++ b/drivers/gpu/drm/virtio/virtgpu_object.c
@@ -124,13 +124,17 @@ int virtio_gpu_object_get_sg_table(struct virtio_gpu_device *qdev,
 	int ret;
 	struct page **pages = bo->tbo.ttm->pages;
 	int nr_pages = bo->tbo.num_pages;
+	struct ttm_operation_ctx ctx = {
+		.interruptible = false,
+		.no_wait_gpu = false
+	};
 
 	/* wtf swapping */
 	if (bo->pages)
 		return 0;
 
 	if (bo->tbo.ttm->state == tt_unpopulated)
-		bo->tbo.ttm->bdev->driver->ttm_tt_populate(bo->tbo.ttm);
+		bo->tbo.ttm->bdev->driver->ttm_tt_populate(bo->tbo.ttm, &ctx);
 	bo->pages = kmalloc(sizeof(struct sg_table), GFP_KERNEL);
 	if (!bo->pages)
 		goto out;
diff --git a/drivers/gpu/drm/virtio/virtgpu_ttm.c b/drivers/gpu/drm/virtio/virtgpu_ttm.c
index 488c6bd..f6118f5 100644
--- a/drivers/gpu/drm/virtio/virtgpu_ttm.c
+++ b/drivers/gpu/drm/virtio/virtgpu_ttm.c
@@ -324,12 +324,13 @@ static struct ttm_backend_func virtio_gpu_backend_func = {
 	.destroy = &virtio_gpu_ttm_backend_destroy,
 };
 
-static int virtio_gpu_ttm_tt_populate(struct ttm_tt *ttm)
+static int virtio_gpu_ttm_tt_populate(struct ttm_tt *ttm,
+		struct ttm_operation_ctx *ctx)
 {
 	if (ttm->state != tt_unpopulated)
 		return 0;
 
-	return ttm_pool_populate(ttm);
+	return ttm_pool_populate(ttm, ctx);
 }
 
 static void virtio_gpu_ttm_tt_unpopulate(struct ttm_tt *ttm)
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_buffer.c
index ef97542..90b0d6b 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_buffer.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_buffer.c
@@ -635,16 +635,12 @@ static void vmw_ttm_destroy(struct ttm_tt *ttm)
 }
 
 
-static int vmw_ttm_populate(struct ttm_tt *ttm)
+static int vmw_ttm_populate(struct ttm_tt *ttm, struct ttm_operation_ctx *ctx)
 {
 	struct vmw_ttm_tt *vmw_tt =
 		container_of(ttm, struct vmw_ttm_tt, dma_ttm.ttm);
 	struct vmw_private *dev_priv = vmw_tt->dev_priv;
 	struct ttm_mem_global *glob = vmw_mem_glob(dev_priv);
-	struct ttm_operation_ctx ctx = {
-		.interruptible = true,
-		.no_wait_gpu = false
-	};
 	int ret;
 
 	if (ttm->state != tt_unpopulated)
@@ -653,15 +649,16 @@ static int vmw_ttm_populate(struct ttm_tt *ttm)
 	if (dev_priv->map_mode == vmw_dma_alloc_coherent) {
 		size_t size =
 			ttm_round_pot(ttm->num_pages * sizeof(dma_addr_t));
-		ret = ttm_mem_global_alloc(glob, size, &ctx);
+		ret = ttm_mem_global_alloc(glob, size, ctx);
 		if (unlikely(ret != 0))
 			return ret;
 
-		ret = ttm_dma_populate(&vmw_tt->dma_ttm, dev_priv->dev->dev);
+		ret = ttm_dma_populate(&vmw_tt->dma_ttm, dev_priv->dev->dev,
+					ctx);
 		if (unlikely(ret != 0))
 			ttm_mem_global_free(glob, size);
 	} else
-		ret = ttm_pool_populate(ttm);
+		ret = ttm_pool_populate(ttm, ctx);
 
 	return ret;
 }
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_mob.c b/drivers/gpu/drm/vmwgfx/vmwgfx_mob.c
index b17f08f..736ca47 100644
--- a/drivers/gpu/drm/vmwgfx/vmwgfx_mob.c
+++ b/drivers/gpu/drm/vmwgfx/vmwgfx_mob.c
@@ -240,6 +240,10 @@ static int vmw_otable_batch_setup(struct vmw_private *dev_priv,
 	unsigned long offset;
 	unsigned long bo_size;
 	struct vmw_otable *otables = batch->otables;
+	struct ttm_operation_ctx ctx = {
+		.interruptible = false,
+		.no_wait_gpu = false
+	};
 	SVGAOTableType i;
 	int ret;
 
@@ -264,7 +268,7 @@ static int vmw_otable_batch_setup(struct vmw_private *dev_priv,
 
 	ret = ttm_bo_reserve(batch->otable_bo, false, true, NULL);
 	BUG_ON(ret != 0);
-	ret = vmw_bo_driver.ttm_tt_populate(batch->otable_bo->ttm);
+	ret = vmw_bo_driver.ttm_tt_populate(batch->otable_bo->ttm, &ctx);
 	if (unlikely(ret != 0))
 		goto out_unreserve;
 	ret = vmw_bo_map_dma(batch->otable_bo);
@@ -430,6 +434,11 @@ static int vmw_mob_pt_populate(struct vmw_private *dev_priv,
 			       struct vmw_mob *mob)
 {
 	int ret;
+	struct ttm_operation_ctx ctx = {
+		.interruptible = false,
+		.no_wait_gpu = false
+	};
+
 	BUG_ON(mob->pt_bo != NULL);
 
 	ret = ttm_bo_create(&dev_priv->bdev, mob->num_pages * PAGE_SIZE,
@@ -442,7 +451,7 @@ static int vmw_mob_pt_populate(struct vmw_private *dev_priv,
 	ret = ttm_bo_reserve(mob->pt_bo, false, true, NULL);
 
 	BUG_ON(ret != 0);
-	ret = vmw_bo_driver.ttm_tt_populate(mob->pt_bo->ttm);
+	ret = vmw_bo_driver.ttm_tt_populate(mob->pt_bo->ttm, &ctx);
 	if (unlikely(ret != 0))
 		goto out_unreserve;
 	ret = vmw_bo_map_dma(mob->pt_bo);
diff --git a/include/drm/ttm/ttm_bo_driver.h b/include/drm/ttm/ttm_bo_driver.h
index 934fecf..84860ec 100644
--- a/include/drm/ttm/ttm_bo_driver.h
+++ b/include/drm/ttm/ttm_bo_driver.h
@@ -352,7 +352,8 @@ struct ttm_bo_driver {
 	 * Returns:
 	 * -ENOMEM: Out of memory.
 	 */
-	int (*ttm_tt_populate)(struct ttm_tt *ttm);
+	int (*ttm_tt_populate)(struct ttm_tt *ttm,
+			struct ttm_operation_ctx *ctx);
 
 	/**
 	 * ttm_tt_unpopulate
@@ -1077,7 +1078,7 @@ struct ttm_tt *ttm_agp_tt_create(struct ttm_bo_device *bdev,
 				 struct agp_bridge_data *bridge,
 				 unsigned long size, uint32_t page_flags,
 				 struct page *dummy_read_page);
-int ttm_agp_tt_populate(struct ttm_tt *ttm);
+int ttm_agp_tt_populate(struct ttm_tt *ttm, struct ttm_operation_ctx *ctx);
 void ttm_agp_tt_unpopulate(struct ttm_tt *ttm);
 #endif
 
diff --git a/include/drm/ttm/ttm_page_alloc.h b/include/drm/ttm/ttm_page_alloc.h
index 5938113..4d9b019 100644
--- a/include/drm/ttm/ttm_page_alloc.h
+++ b/include/drm/ttm/ttm_page_alloc.h
@@ -47,7 +47,7 @@ void ttm_page_alloc_fini(void);
  *
  * Add backing pages to all of @ttm
  */
-int ttm_pool_populate(struct ttm_tt *ttm);
+int ttm_pool_populate(struct ttm_tt *ttm, struct ttm_operation_ctx *ctx);
 
 /**
  * ttm_pool_unpopulate:
@@ -61,7 +61,8 @@ void ttm_pool_unpopulate(struct ttm_tt *ttm);
 /**
  * Populates and DMA maps pages to fullfil a ttm_dma_populate() request
  */
-int ttm_populate_and_map_pages(struct device *dev, struct ttm_dma_tt *tt);
+int ttm_populate_and_map_pages(struct device *dev, struct ttm_dma_tt *tt,
+				struct ttm_operation_ctx *ctx);
 
 /**
  * Unpopulates and DMA unmaps pages as part of a
@@ -89,7 +90,8 @@ void ttm_dma_page_alloc_fini(void);
  */
 int ttm_dma_page_alloc_debugfs(struct seq_file *m, void *data);
 
-int ttm_dma_populate(struct ttm_dma_tt *ttm_dma, struct device *dev);
+int ttm_dma_populate(struct ttm_dma_tt *ttm_dma, struct device *dev,
+			struct ttm_operation_ctx *ctx);
 void ttm_dma_unpopulate(struct ttm_dma_tt *ttm_dma, struct device *dev);
 
 #else
@@ -106,7 +108,8 @@ static inline int ttm_dma_page_alloc_debugfs(struct seq_file *m, void *data)
 	return 0;
 }
 static inline int ttm_dma_populate(struct ttm_dma_tt *ttm_dma,
-				   struct device *dev)
+				struct device *dev,
+				struct ttm_operation_ctx *ctx)
 {
 	return -ENOMEM;
 }
-- 
2.7.4

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

[PATCH 3/5] drm/ttm: use an operation ctx for ttm_tt_bind

[Roger He]

forward the operation context to ttm_tt_bind as well,
and the ultimate goal is swapout enablement for per vm BOs.

Change-Id: I42a7df8c50e1ce3b527ee9cb78809f8e58136f07
Signed-off-by: Roger He <Hongbo.He@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 2 +-
 drivers/gpu/drm/nouveau/nouveau_bo.c    | 2 +-
 drivers/gpu/drm/radeon/radeon_ttm.c     | 2 +-
 drivers/gpu/drm/ttm/ttm_bo.c            | 2 +-
 drivers/gpu/drm/ttm/ttm_bo_util.c       | 2 +-
 drivers/gpu/drm/ttm/ttm_tt.c            | 9 +++------
 include/drm/ttm/ttm_bo_driver.h         | 3 ++-
 7 files changed, 10 insertions(+), 12 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index 044f5b5..e4bb435 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -497,7 +497,7 @@ static int amdgpu_move_vram_ram(struct ttm_buffer_object *bo, bool evict,
 		goto out_cleanup;
 	}
 
-	r = ttm_tt_bind(bo->ttm, &tmp_mem);
+	r = ttm_tt_bind(bo->ttm, &tmp_mem, ctx);
 	if (unlikely(r)) {
 		goto out_cleanup;
 	}
diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c
index b141c27..a83ba6a 100644
--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -1218,7 +1218,7 @@ nouveau_bo_move_flipd(struct ttm_buffer_object *bo, bool evict, bool intr,
 	if (ret)
 		return ret;
 
-	ret = ttm_tt_bind(bo->ttm, &tmp_reg);
+	ret = ttm_tt_bind(bo->ttm, &tmp_reg, &ctx);
 	if (ret)
 		goto out;
 
diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c
index 9424b81..2b74ba1b 100644
--- a/drivers/gpu/drm/radeon/radeon_ttm.c
+++ b/drivers/gpu/drm/radeon/radeon_ttm.c
@@ -339,7 +339,7 @@ static int radeon_move_vram_ram(struct ttm_buffer_object *bo,
 		goto out_cleanup;
 	}
 
-	r = ttm_tt_bind(bo->ttm, &tmp_mem);
+	r = ttm_tt_bind(bo->ttm, &tmp_mem, &ctx);
 	if (unlikely(r)) {
 		goto out_cleanup;
 	}
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index c59f572..e7595b4 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -300,7 +300,7 @@ static int ttm_bo_handle_move_mem(struct ttm_buffer_object *bo,
 			goto out_err;
 
 		if (mem->mem_type != TTM_PL_SYSTEM) {
-			ret = ttm_tt_bind(bo->ttm, mem);
+			ret = ttm_tt_bind(bo->ttm, mem, ctx);
 			if (ret)
 				goto out_err;
 		}
diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c
index b7eb507..153de1b 100644
--- a/drivers/gpu/drm/ttm/ttm_bo_util.c
+++ b/drivers/gpu/drm/ttm/ttm_bo_util.c
@@ -73,7 +73,7 @@ int ttm_bo_move_ttm(struct ttm_buffer_object *bo,
 		return ret;
 
 	if (new_mem->mem_type != TTM_PL_SYSTEM) {
-		ret = ttm_tt_bind(ttm, new_mem);
+		ret = ttm_tt_bind(ttm, new_mem, ctx);
 		if (unlikely(ret != 0))
 			return ret;
 	}
diff --git a/drivers/gpu/drm/ttm/ttm_tt.c b/drivers/gpu/drm/ttm/ttm_tt.c
index b48d7a0..5a046a3 100644
--- a/drivers/gpu/drm/ttm/ttm_tt.c
+++ b/drivers/gpu/drm/ttm/ttm_tt.c
@@ -261,12 +261,9 @@ void ttm_tt_unbind(struct ttm_tt *ttm)
 	}
 }
 
-int ttm_tt_bind(struct ttm_tt *ttm, struct ttm_mem_reg *bo_mem)
+int ttm_tt_bind(struct ttm_tt *ttm, struct ttm_mem_reg *bo_mem,
+		struct ttm_operation_ctx *ctx)
 {
-	struct ttm_operation_ctx ctx = {
-		.interruptible = false,
-		.no_wait_gpu = false
-	};
 	int ret = 0;
 
 	if (!ttm)
@@ -275,7 +272,7 @@ int ttm_tt_bind(struct ttm_tt *ttm, struct ttm_mem_reg *bo_mem)
 	if (ttm->state == tt_bound)
 		return 0;
 
-	ret = ttm->bdev->driver->ttm_tt_populate(ttm, &ctx);
+	ret = ttm->bdev->driver->ttm_tt_populate(ttm, ctx);
 	if (ret)
 		return ret;
 
diff --git a/include/drm/ttm/ttm_bo_driver.h b/include/drm/ttm/ttm_bo_driver.h
index 84860ec..94064b1 100644
--- a/include/drm/ttm/ttm_bo_driver.h
+++ b/include/drm/ttm/ttm_bo_driver.h
@@ -650,7 +650,8 @@ void ttm_dma_tt_fini(struct ttm_dma_tt *ttm_dma);
  *
  * Bind the pages of @ttm to an aperture location identified by @bo_mem
  */
-int ttm_tt_bind(struct ttm_tt *ttm, struct ttm_mem_reg *bo_mem);
+int ttm_tt_bind(struct ttm_tt *ttm, struct ttm_mem_reg *bo_mem,
+		struct ttm_operation_ctx *ctx);
 
 /**
  * ttm_ttm_destroy:
-- 
2.7.4

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

[PATCH 4/5] drm/ttm: add new function to check if bo is allowable to evict or swapout

[Roger He]

extract a function as ttm_bo_evict_swapout_allowable since eviction and
swapout can share same logic

Change-Id: I80a475a93fceed8d66d74a1832c815a0756341ac
Signed-off-by: Roger He <Hongbo.He@amd.com>
---
 drivers/gpu/drm/ttm/ttm_bo.c | 29 +++++++++++++++++++----------
 1 file changed, 19 insertions(+), 10 deletions(-)

diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index e7595b4..313925c 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -706,6 +706,23 @@ bool ttm_bo_eviction_valuable(struct ttm_buffer_object *bo,
 }
 EXPORT_SYMBOL(ttm_bo_eviction_valuable);
 
+static bool ttm_bo_evict_swapout_allowable(struct ttm_buffer_object *bo,
+			struct ttm_operation_ctx *ctx, bool *locked)
+{
+	bool ret = false;
+
+	*locked = false;
+	if (bo->resv == ctx->resv) {
+		if (ctx->allow_reserved_eviction || !list_empty(&bo->ddestroy))
+			ret = true;
+	} else {
+		*locked = reservation_object_trylock(bo->resv);
+		ret = *locked;
+	}
+
+	return ret;
+}
+
 static int ttm_mem_evict_first(struct ttm_bo_device *bdev,
 			       uint32_t mem_type,
 			       const struct ttm_place *place,
@@ -721,21 +738,13 @@ static int ttm_mem_evict_first(struct ttm_bo_device *bdev,
 	spin_lock(&glob->lru_lock);
 	for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i) {
 		list_for_each_entry(bo, &man->lru[i], lru) {
-			if (bo->resv == ctx->resv) {
-				if (!ctx->allow_reserved_eviction &&
-				    list_empty(&bo->ddestroy))
-					continue;
-			} else {
-				locked = reservation_object_trylock(bo->resv);
-				if (!locked)
-					continue;
-			}
+			if (!ttm_bo_evict_swapout_allowable(bo, ctx, &locked))
+				continue;
 
 			if (place && !bdev->driver->eviction_valuable(bo,
 								      place)) {
 				if (locked)
 					reservation_object_unlock(bo->resv);
-				locked = false;
 				continue;
 			}
 			break;
-- 
2.7.4

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

[PATCH 5/5] drm/ttm: enable swapout of per VM BOs during allocation

[Roger He]

if the bo shares same reservation object then not lock it again
at swapout time to make it possible to swap out.

Change-Id: I1e87954564f38ad298bf6e4ff88c9f26f291a62d
Signed-off-by: Roger He <Hongbo.He@amd.com>
---
 drivers/gpu/drm/ttm/ttm_bo.c     | 15 +++++++++++----
 drivers/gpu/drm/ttm/ttm_memory.c | 12 ++++++++----
 include/drm/ttm/ttm_bo_api.h     |  3 ++-
 3 files changed, 21 insertions(+), 9 deletions(-)

diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index 313925c..ecb8916 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -1686,18 +1686,20 @@ EXPORT_SYMBOL(ttm_bo_synccpu_write_release);
  * A buffer object shrink method that tries to swap out the first
  * buffer object on the bo_global::swap_lru list.
  */
-int ttm_bo_swapout(struct ttm_bo_global *glob)
+int ttm_bo_swapout(struct ttm_bo_global *glob, struct ttm_operation_ctx *ctx)
 {
 	struct ttm_buffer_object *bo;
 	int ret = -EBUSY;
+	bool locked;
 	unsigned i;
 
 	spin_lock(&glob->lru_lock);
 	for (i = 0; i < TTM_MAX_BO_PRIORITY; ++i) {
 		list_for_each_entry(bo, &glob->swap_lru[i], swap) {
-			ret = reservation_object_trylock(bo->resv) ? 0 : -EBUSY;
-			if (!ret)
+			if (ttm_bo_evict_swapout_allowable(bo, ctx, &locked)) {
+				ret = 0;
 				break;
+			}
 		}
 		if (!ret)
 			break;
@@ -1773,7 +1775,12 @@ EXPORT_SYMBOL(ttm_bo_swapout);
 
 void ttm_bo_swapout_all(struct ttm_bo_device *bdev)
 {
-	while (ttm_bo_swapout(bdev->glob) == 0)
+	struct ttm_operation_ctx ctx = {
+		.interruptible = false,
+		.no_wait_gpu = false
+	};
+
+	while (ttm_bo_swapout(bdev->glob, &ctx) == 0)
 		;
 }
 EXPORT_SYMBOL(ttm_bo_swapout_all);
diff --git a/drivers/gpu/drm/ttm/ttm_memory.c b/drivers/gpu/drm/ttm/ttm_memory.c
index 102b326..aa0c381 100644
--- a/drivers/gpu/drm/ttm/ttm_memory.c
+++ b/drivers/gpu/drm/ttm/ttm_memory.c
@@ -211,7 +211,7 @@ static bool ttm_zones_above_swap_target(struct ttm_mem_global *glob,
  */
 
 static void ttm_shrink(struct ttm_mem_global *glob, bool from_wq,
-		       uint64_t extra)
+			uint64_t extra, struct ttm_operation_ctx *ctx)
 {
 	int ret;
 
@@ -219,7 +219,7 @@ static void ttm_shrink(struct ttm_mem_global *glob, bool from_wq,
 
 	while (ttm_zones_above_swap_target(glob, from_wq, extra)) {
 		spin_unlock(&glob->lock);
-		ret = ttm_bo_swapout(glob->bo_glob);
+		ret = ttm_bo_swapout(glob->bo_glob, ctx);
 		spin_lock(&glob->lock);
 		if (unlikely(ret != 0))
 			break;
@@ -230,10 +230,14 @@ static void ttm_shrink(struct ttm_mem_global *glob, bool from_wq,
 
 static void ttm_shrink_work(struct work_struct *work)
 {
+	struct ttm_operation_ctx ctx = {
+		.interruptible = false,
+		.no_wait_gpu = false
+	};
 	struct ttm_mem_global *glob =
 	    container_of(work, struct ttm_mem_global, work);
 
-	ttm_shrink(glob, true, 0ULL);
+	ttm_shrink(glob, true, 0ULL, &ctx);
 }
 
 static int ttm_mem_init_kernel_zone(struct ttm_mem_global *glob,
@@ -520,7 +524,7 @@ static int ttm_mem_global_alloc_zone(struct ttm_mem_global *glob,
 			return -ENOMEM;
 		if (unlikely(count-- == 0))
 			return -ENOMEM;
-		ttm_shrink(glob, false, memory + (memory >> 2) + 16);
+		ttm_shrink(glob, false, memory + (memory >> 2) + 16, ctx);
 	}
 
 	return 0;
diff --git a/include/drm/ttm/ttm_bo_api.h b/include/drm/ttm/ttm_bo_api.h
index 24a8db7..f1c74c2 100644
--- a/include/drm/ttm/ttm_bo_api.h
+++ b/include/drm/ttm/ttm_bo_api.h
@@ -752,7 +752,8 @@ ssize_t ttm_bo_io(struct ttm_bo_device *bdev, struct file *filp,
 		  const char __user *wbuf, char __user *rbuf,
 		  size_t count, loff_t *f_pos, bool write);
 
-int ttm_bo_swapout(struct ttm_bo_global *glob);
+int ttm_bo_swapout(struct ttm_bo_global *glob,
+			struct ttm_operation_ctx *ctx);
 void ttm_bo_swapout_all(struct ttm_bo_device *bdev);
 int ttm_bo_wait_unreserved(struct ttm_buffer_object *bo);
 #endif
-- 
2.7.4

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

Re: [PATCH 1/5] drm/ttm: use an operation ctx for ttm_mem_global_alloc_page

[Christian König]

Am 21.12.2017 um 10:42 schrieb Roger He:
> forward the operation context to ttm_mem_global_alloc_page as well,
> and the ultimate goal is swapout enablement for per vm BOs.

As Thomas commented as well, better write "reserved BOs" in the commit 
message.

The term "per vm BOs" is something completely amdgpu specific.

With that fixed the patch is Reviewed-by: Christian König 
<christian.koenig@amd.com>.

Regards,
Christian.

>
> Per vm BOs refer to all BOs which share same reservation object.
>
> Change-Id: I4104a12e09a374b6477a0dd5a8fce26dce27a746
> Signed-off-by: Roger He <Hongbo.He@amd.com>
> ---
>   drivers/gpu/drm/ttm/ttm_memory.c         | 10 +++-------
>   drivers/gpu/drm/ttm/ttm_page_alloc.c     |  6 +++++-
>   drivers/gpu/drm/ttm/ttm_page_alloc_dma.c |  8 ++++++--
>   include/drm/ttm/ttm_memory.h             |  3 ++-
>   4 files changed, 16 insertions(+), 11 deletions(-)
>
> diff --git a/drivers/gpu/drm/ttm/ttm_memory.c b/drivers/gpu/drm/ttm/ttm_memory.c
> index 525d3b6..102b326 100644
> --- a/drivers/gpu/drm/ttm/ttm_memory.c
> +++ b/drivers/gpu/drm/ttm/ttm_memory.c
> @@ -539,14 +539,10 @@ int ttm_mem_global_alloc(struct ttm_mem_global *glob, uint64_t memory,
>   EXPORT_SYMBOL(ttm_mem_global_alloc);
>   
>   int ttm_mem_global_alloc_page(struct ttm_mem_global *glob,
> -			      struct page *page, uint64_t size)
> +			      struct page *page, uint64_t size,
> +			      struct ttm_operation_ctx *ctx)
>   {
> -
>   	struct ttm_mem_zone *zone = NULL;
> -	struct ttm_operation_ctx ctx = {
> -		.interruptible = false,
> -		.no_wait_gpu = false
> -	};
>   
>   	/**
>   	 * Page allocations may be registed in a single zone
> @@ -560,7 +556,7 @@ int ttm_mem_global_alloc_page(struct ttm_mem_global *glob,
>   	if (glob->zone_dma32 && page_to_pfn(page) > 0x00100000UL)
>   		zone = glob->zone_kernel;
>   #endif
> -	return ttm_mem_global_alloc_zone(glob, zone, size, &ctx);
> +	return ttm_mem_global_alloc_zone(glob, zone, size, ctx);
>   }
>   
>   void ttm_mem_global_free_page(struct ttm_mem_global *glob, struct page *page,
> diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc.c b/drivers/gpu/drm/ttm/ttm_page_alloc.c
> index b5ba644..8f93ff3 100644
> --- a/drivers/gpu/drm/ttm/ttm_page_alloc.c
> +++ b/drivers/gpu/drm/ttm/ttm_page_alloc.c
> @@ -1061,6 +1061,10 @@ void ttm_page_alloc_fini(void)
>   int ttm_pool_populate(struct ttm_tt *ttm)
>   {
>   	struct ttm_mem_global *mem_glob = ttm->glob->mem_glob;
> +	struct ttm_operation_ctx ctx = {
> +		.interruptible = false,
> +		.no_wait_gpu = false
> +	};
>   	unsigned i;
>   	int ret;
>   
> @@ -1076,7 +1080,7 @@ int ttm_pool_populate(struct ttm_tt *ttm)
>   
>   	for (i = 0; i < ttm->num_pages; ++i) {
>   		ret = ttm_mem_global_alloc_page(mem_glob, ttm->pages[i],
> -						PAGE_SIZE);
> +						PAGE_SIZE, &ctx);
>   		if (unlikely(ret != 0)) {
>   			ttm_pool_unpopulate(ttm);
>   			return -ENOMEM;
> diff --git a/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c b/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c
> index bda00b2..8aac86a 100644
> --- a/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c
> +++ b/drivers/gpu/drm/ttm/ttm_page_alloc_dma.c
> @@ -927,6 +927,10 @@ int ttm_dma_populate(struct ttm_dma_tt *ttm_dma, struct device *dev)
>   {
>   	struct ttm_tt *ttm = &ttm_dma->ttm;
>   	struct ttm_mem_global *mem_glob = ttm->glob->mem_glob;
> +	struct ttm_operation_ctx ctx = {
> +		.interruptible = false,
> +		.no_wait_gpu = false
> +	};
>   	unsigned long num_pages = ttm->num_pages;
>   	struct dma_pool *pool;
>   	enum pool_type type;
> @@ -962,7 +966,7 @@ int ttm_dma_populate(struct ttm_dma_tt *ttm_dma, struct device *dev)
>   			break;
>   
>   		ret = ttm_mem_global_alloc_page(mem_glob, ttm->pages[i],
> -						pool->size);
> +						pool->size, &ctx);
>   		if (unlikely(ret != 0)) {
>   			ttm_dma_unpopulate(ttm_dma, dev);
>   			return -ENOMEM;
> @@ -998,7 +1002,7 @@ int ttm_dma_populate(struct ttm_dma_tt *ttm_dma, struct device *dev)
>   		}
>   
>   		ret = ttm_mem_global_alloc_page(mem_glob, ttm->pages[i],
> -						pool->size);
> +						pool->size, &ctx);
>   		if (unlikely(ret != 0)) {
>   			ttm_dma_unpopulate(ttm_dma, dev);
>   			return -ENOMEM;
> diff --git a/include/drm/ttm/ttm_memory.h b/include/drm/ttm/ttm_memory.h
> index 755c107..8936285 100644
> --- a/include/drm/ttm/ttm_memory.h
> +++ b/include/drm/ttm/ttm_memory.h
> @@ -84,7 +84,8 @@ extern int ttm_mem_global_alloc(struct ttm_mem_global *glob, uint64_t memory,
>   extern void ttm_mem_global_free(struct ttm_mem_global *glob,
>   				uint64_t amount);
>   extern int ttm_mem_global_alloc_page(struct ttm_mem_global *glob,
> -				     struct page *page, uint64_t size);
> +				     struct page *page, uint64_t size,
> +				     struct ttm_operation_ctx *ctx);
>   extern void ttm_mem_global_free_page(struct ttm_mem_global *glob,
>   				     struct page *page, uint64_t size);
>   extern size_t ttm_round_pot(size_t size);

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel