Allows us to gut a BO of it's backing store when the driver says that it isn't needed any more.
Signed-off-by: Christian König christian.koenig@amd.com --- drivers/gpu/drm/ttm/ttm_bo.c | 15 ++++++++++++--- drivers/gpu/drm/ttm/ttm_bo_util.c | 24 ++++++++++++++++++++++++ include/drm/ttm/ttm_bo_driver.h | 9 +++++++++ 3 files changed, 45 insertions(+), 3 deletions(-)
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c index ad142a92eb80..98e06f8bf23b 100644 --- a/drivers/gpu/drm/ttm/ttm_bo.c +++ b/drivers/gpu/drm/ttm/ttm_bo.c @@ -622,14 +622,23 @@ static int ttm_bo_evict(struct ttm_buffer_object *bo,
reservation_object_assert_held(bo->resv);
+ placement.num_placement = 0; + placement.num_busy_placement = 0; + bdev->driver->evict_flags(bo, &placement); + + if (!placement.num_placement && !placement.num_busy_placement) { + ret = ttm_bo_pipeline_gutting(bo); + if (ret) + return ret; + + return ttm_tt_create(bo, false); + } + evict_mem = bo->mem; evict_mem.mm_node = NULL; evict_mem.bus.io_reserved_vm = false; evict_mem.bus.io_reserved_count = 0;
- placement.num_placement = 0; - placement.num_busy_placement = 0; - bdev->driver->evict_flags(bo, &placement); ret = ttm_bo_mem_space(bo, &placement, &evict_mem, ctx); if (ret) { if (ret != -ERESTARTSYS) { diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c index 6d6a3f46143b..1f730b3f18e5 100644 --- a/drivers/gpu/drm/ttm/ttm_bo_util.c +++ b/drivers/gpu/drm/ttm/ttm_bo_util.c @@ -801,3 +801,27 @@ int ttm_bo_pipeline_move(struct ttm_buffer_object *bo, return 0; } EXPORT_SYMBOL(ttm_bo_pipeline_move); + +int ttm_bo_pipeline_gutting(struct ttm_buffer_object *bo) +{ + struct ttm_buffer_object *ghost; + int ret; + + ret = ttm_buffer_object_transfer(bo, &ghost); + if (ret) + return ret; + + ret = reservation_object_copy_fences(ghost->resv, bo->resv); + /* Last resort, wait for the BO to be idle when we are OOM */ + if (ret) + ttm_bo_wait(bo, false, false); + + memset(&bo->mem, 0, sizeof(bo->mem)); + bo->mem.mem_type = TTM_PL_SYSTEM; + bo->ttm = NULL; + + ttm_bo_unreserve(ghost); + ttm_bo_unref(&ghost); + + return 0; +} diff --git a/include/drm/ttm/ttm_bo_driver.h b/include/drm/ttm/ttm_bo_driver.h index f8e2515b401f..39cd6b086d3a 100644 --- a/include/drm/ttm/ttm_bo_driver.h +++ b/include/drm/ttm/ttm_bo_driver.h @@ -849,6 +849,15 @@ int ttm_bo_pipeline_move(struct ttm_buffer_object *bo, struct dma_fence *fence, bool evict, struct ttm_mem_reg *new_mem);
+/** + * ttm_bo_pipeline_gutting. + * + * @bo: A pointer to a struct ttm_buffer_object. + * + * Pipelined gutting a BO of it's backing store. + */ +int ttm_bo_pipeline_gutting(struct ttm_buffer_object *bo); + /** * ttm_io_prot *
Instead of calculating the size in bytes just to recalculate the number of pages from it pass the BO directly to the function.
Signed-off-by: Christian König christian.koenig@amd.com --- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 8 ++++---- drivers/gpu/drm/ast/ast_ttm.c | 6 +++--- drivers/gpu/drm/bochs/bochs_mm.c | 5 ++--- drivers/gpu/drm/cirrus/cirrus_ttm.c | 6 +++--- drivers/gpu/drm/hisilicon/hibmc/hibmc_ttm.c | 5 ++--- drivers/gpu/drm/mgag200/mgag200_ttm.c | 6 +++--- drivers/gpu/drm/nouveau/nouveau_bo.c | 10 ++++------ drivers/gpu/drm/nouveau/nouveau_sgdma.c | 7 +++---- drivers/gpu/drm/nouveau/nouveau_ttm.h | 4 ++-- drivers/gpu/drm/qxl/qxl_ttm.c | 8 ++++---- drivers/gpu/drm/radeon/radeon_ttm.c | 12 ++++++------ drivers/gpu/drm/ttm/ttm_agp_backend.c | 6 +++--- drivers/gpu/drm/ttm/ttm_tt.c | 29 ++++++++++++++--------------- drivers/gpu/drm/virtio/virtgpu_ttm.c | 7 +++---- drivers/gpu/drm/vmwgfx/vmwgfx_buffer.c | 10 +++++----- drivers/staging/vboxvideo/vbox_ttm.c | 5 ++--- include/drm/ttm/ttm_bo_driver.h | 6 ++---- include/drm/ttm/ttm_tt.h | 22 ++++++++++------------ 18 files changed, 75 insertions(+), 87 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index 628cbc3ab46f..44ae2c60c9d4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -974,20 +974,20 @@ static struct ttm_backend_func amdgpu_backend_func = { .destroy = &amdgpu_ttm_backend_destroy, };
-static struct ttm_tt *amdgpu_ttm_tt_create(struct ttm_bo_device *bdev, - unsigned long size, uint32_t page_flags) +static struct ttm_tt *amdgpu_ttm_tt_create(struct ttm_buffer_object *bo, + uint32_t page_flags) { struct amdgpu_device *adev; struct amdgpu_ttm_tt *gtt;
- adev = amdgpu_ttm_adev(bdev); + adev = amdgpu_ttm_adev(bo->bdev);
gtt = kzalloc(sizeof(struct amdgpu_ttm_tt), GFP_KERNEL); if (gtt == NULL) { return NULL; } gtt->ttm.ttm.func = &amdgpu_backend_func; - if (ttm_sg_tt_init(>t->ttm, bdev, size, page_flags)) { + if (ttm_sg_tt_init(>t->ttm, bo, page_flags)) { kfree(gtt); return NULL; } diff --git a/drivers/gpu/drm/ast/ast_ttm.c b/drivers/gpu/drm/ast/ast_ttm.c index 211224f6bdd3..fe354ebf374d 100644 --- a/drivers/gpu/drm/ast/ast_ttm.c +++ b/drivers/gpu/drm/ast/ast_ttm.c @@ -199,8 +199,8 @@ static struct ttm_backend_func ast_tt_backend_func = { };
-static struct ttm_tt *ast_ttm_tt_create(struct ttm_bo_device *bdev, - unsigned long size, uint32_t page_flags) +static struct ttm_tt *ast_ttm_tt_create(struct ttm_buffer_object *bo, + uint32_t page_flags) { struct ttm_tt *tt;
@@ -208,7 +208,7 @@ static struct ttm_tt *ast_ttm_tt_create(struct ttm_bo_device *bdev, if (tt == NULL) return NULL; tt->func = &ast_tt_backend_func; - if (ttm_tt_init(tt, bdev, size, page_flags)) { + if (ttm_tt_init(tt, bo, page_flags)) { kfree(tt); return NULL; } diff --git a/drivers/gpu/drm/bochs/bochs_mm.c b/drivers/gpu/drm/bochs/bochs_mm.c index 73722484e12b..50c52b45448f 100644 --- a/drivers/gpu/drm/bochs/bochs_mm.c +++ b/drivers/gpu/drm/bochs/bochs_mm.c @@ -176,8 +176,7 @@ static struct ttm_backend_func bochs_tt_backend_func = { .destroy = &bochs_ttm_backend_destroy, };
-static struct ttm_tt *bochs_ttm_tt_create(struct ttm_bo_device *bdev, - unsigned long size, +static struct ttm_tt *bochs_ttm_tt_create(struct ttm_buffer_object *bo, uint32_t page_flags) { struct ttm_tt *tt; @@ -186,7 +185,7 @@ static struct ttm_tt *bochs_ttm_tt_create(struct ttm_bo_device *bdev, if (tt == NULL) return NULL; tt->func = &bochs_tt_backend_func; - if (ttm_tt_init(tt, bdev, size, page_flags)) { + if (ttm_tt_init(tt, bo, page_flags)) { kfree(tt); return NULL; } diff --git a/drivers/gpu/drm/cirrus/cirrus_ttm.c b/drivers/gpu/drm/cirrus/cirrus_ttm.c index 6cd0233b3bf8..f21953243790 100644 --- a/drivers/gpu/drm/cirrus/cirrus_ttm.c +++ b/drivers/gpu/drm/cirrus/cirrus_ttm.c @@ -199,8 +199,8 @@ static struct ttm_backend_func cirrus_tt_backend_func = { };
-static struct ttm_tt *cirrus_ttm_tt_create(struct ttm_bo_device *bdev, - unsigned long size, uint32_t page_flags) +static struct ttm_tt *cirrus_ttm_tt_create(struct ttm_buffer_object *bo, + uint32_t page_flags) { struct ttm_tt *tt;
@@ -208,7 +208,7 @@ static struct ttm_tt *cirrus_ttm_tt_create(struct ttm_bo_device *bdev, if (tt == NULL) return NULL; tt->func = &cirrus_tt_backend_func; - if (ttm_tt_init(tt, bdev, size, page_flags)) { + if (ttm_tt_init(tt, bo, page_flags)) { kfree(tt); return NULL; } diff --git a/drivers/gpu/drm/hisilicon/hibmc/hibmc_ttm.c b/drivers/gpu/drm/hisilicon/hibmc/hibmc_ttm.c index 8dfffdbb6b07..4871025f7573 100644 --- a/drivers/gpu/drm/hisilicon/hibmc/hibmc_ttm.c +++ b/drivers/gpu/drm/hisilicon/hibmc/hibmc_ttm.c @@ -200,8 +200,7 @@ static struct ttm_backend_func hibmc_tt_backend_func = { .destroy = &hibmc_ttm_backend_destroy, };
-static struct ttm_tt *hibmc_ttm_tt_create(struct ttm_bo_device *bdev, - unsigned long size, +static struct ttm_tt *hibmc_ttm_tt_create(struct ttm_buffer_object *bo, u32 page_flags) { struct ttm_tt *tt; @@ -213,7 +212,7 @@ static struct ttm_tt *hibmc_ttm_tt_create(struct ttm_bo_device *bdev, return NULL; } tt->func = &hibmc_tt_backend_func; - ret = ttm_tt_init(tt, bdev, size, page_flags); + ret = ttm_tt_init(tt, bo, page_flags); if (ret) { DRM_ERROR("failed to initialize ttm_tt: %d\n", ret); kfree(tt); diff --git a/drivers/gpu/drm/mgag200/mgag200_ttm.c b/drivers/gpu/drm/mgag200/mgag200_ttm.c index 69beb2046008..05570f0de4d7 100644 --- a/drivers/gpu/drm/mgag200/mgag200_ttm.c +++ b/drivers/gpu/drm/mgag200/mgag200_ttm.c @@ -199,8 +199,8 @@ static struct ttm_backend_func mgag200_tt_backend_func = { };
-static struct ttm_tt *mgag200_ttm_tt_create(struct ttm_bo_device *bdev, - unsigned long size, uint32_t page_flags) +static struct ttm_tt *mgag200_ttm_tt_create(struct ttm_buffer_object *bo, + uint32_t page_flags) { struct ttm_tt *tt;
@@ -208,7 +208,7 @@ static struct ttm_tt *mgag200_ttm_tt_create(struct ttm_bo_device *bdev, if (tt == NULL) return NULL; tt->func = &mgag200_tt_backend_func; - if (ttm_tt_init(tt, bdev, size, page_flags)) { + if (ttm_tt_init(tt, bo, page_flags)) { kfree(tt); return NULL; } diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c index 49cc8dfcb141..6f402c4f2bdd 100644 --- a/drivers/gpu/drm/nouveau/nouveau_bo.c +++ b/drivers/gpu/drm/nouveau/nouveau_bo.c @@ -604,19 +604,17 @@ nouveau_bo_wr32(struct nouveau_bo *nvbo, unsigned index, u32 val) }
static struct ttm_tt * -nouveau_ttm_tt_create(struct ttm_bo_device *bdev, unsigned long size, - uint32_t page_flags) +nouveau_ttm_tt_create(struct ttm_buffer_object *bo, uint32_t page_flags) { #if IS_ENABLED(CONFIG_AGP) - struct nouveau_drm *drm = nouveau_bdev(bdev); + struct nouveau_drm *drm = nouveau_bdev(bo->bdev);
if (drm->agp.bridge) { - return ttm_agp_tt_create(bdev, drm->agp.bridge, size, - page_flags); + return ttm_agp_tt_create(bo, drm->agp.bridge, page_flags); } #endif
- return nouveau_sgdma_create_ttm(bdev, size, page_flags); + return nouveau_sgdma_create_ttm(bo, page_flags); }
static int diff --git a/drivers/gpu/drm/nouveau/nouveau_sgdma.c b/drivers/gpu/drm/nouveau/nouveau_sgdma.c index 87b030437f4d..8ebdc74cc0ad 100644 --- a/drivers/gpu/drm/nouveau/nouveau_sgdma.c +++ b/drivers/gpu/drm/nouveau/nouveau_sgdma.c @@ -82,10 +82,9 @@ static struct ttm_backend_func nv50_sgdma_backend = { };
struct ttm_tt * -nouveau_sgdma_create_ttm(struct ttm_bo_device *bdev, - unsigned long size, uint32_t page_flags) +nouveau_sgdma_create_ttm(struct ttm_buffer_object *bo, uint32_t page_flags) { - struct nouveau_drm *drm = nouveau_bdev(bdev); + struct nouveau_drm *drm = nouveau_bdev(bo->bdev); struct nouveau_sgdma_be *nvbe;
nvbe = kzalloc(sizeof(*nvbe), GFP_KERNEL); @@ -97,7 +96,7 @@ nouveau_sgdma_create_ttm(struct ttm_bo_device *bdev, else nvbe->ttm.ttm.func = &nv50_sgdma_backend;
- if (ttm_dma_tt_init(&nvbe->ttm, bdev, size, page_flags)) + if (ttm_dma_tt_init(&nvbe->ttm, bo, page_flags)) /* * A failing ttm_dma_tt_init() will call ttm_tt_destroy() * and thus our nouveau_sgdma_destroy() hook, so we don't need diff --git a/drivers/gpu/drm/nouveau/nouveau_ttm.h b/drivers/gpu/drm/nouveau/nouveau_ttm.h index 64e484ee5ef1..89929ad8c7cd 100644 --- a/drivers/gpu/drm/nouveau/nouveau_ttm.h +++ b/drivers/gpu/drm/nouveau/nouveau_ttm.h @@ -12,8 +12,8 @@ extern const struct ttm_mem_type_manager_func nouveau_vram_manager; extern const struct ttm_mem_type_manager_func nouveau_gart_manager; extern const struct ttm_mem_type_manager_func nv04_gart_manager;
-struct ttm_tt *nouveau_sgdma_create_ttm(struct ttm_bo_device *, - unsigned long size, u32 page_flags); +struct ttm_tt *nouveau_sgdma_create_ttm(struct ttm_buffer_object *bo, + u32 page_flags);
int nouveau_ttm_init(struct nouveau_drm *drm); void nouveau_ttm_fini(struct nouveau_drm *drm); diff --git a/drivers/gpu/drm/qxl/qxl_ttm.c b/drivers/gpu/drm/qxl/qxl_ttm.c index 2ad70eb96207..ee2340e31f06 100644 --- a/drivers/gpu/drm/qxl/qxl_ttm.c +++ b/drivers/gpu/drm/qxl/qxl_ttm.c @@ -291,19 +291,19 @@ static struct ttm_backend_func qxl_backend_func = { .destroy = &qxl_ttm_backend_destroy, };
-static struct ttm_tt *qxl_ttm_tt_create(struct ttm_bo_device *bdev, - unsigned long size, uint32_t page_flags) +static struct ttm_tt *qxl_ttm_tt_create(struct ttm_buffer_object *bo, + uint32_t page_flags) { struct qxl_device *qdev; struct qxl_ttm_tt *gtt;
- qdev = qxl_get_qdev(bdev); + qdev = qxl_get_qdev(bo->bdev); gtt = kzalloc(sizeof(struct qxl_ttm_tt), GFP_KERNEL); if (gtt == NULL) return NULL; gtt->ttm.ttm.func = &qxl_backend_func; gtt->qdev = qdev; - if (ttm_dma_tt_init(>t->ttm, bdev, size, page_flags)) { + if (ttm_dma_tt_init(>t->ttm, bo, page_flags)) { kfree(gtt); return NULL; } diff --git a/drivers/gpu/drm/radeon/radeon_ttm.c b/drivers/gpu/drm/radeon/radeon_ttm.c index 009f55a2bbf9..8689fcca051c 100644 --- a/drivers/gpu/drm/radeon/radeon_ttm.c +++ b/drivers/gpu/drm/radeon/radeon_ttm.c @@ -686,17 +686,17 @@ static struct ttm_backend_func radeon_backend_func = { .destroy = &radeon_ttm_backend_destroy, };
-static struct ttm_tt *radeon_ttm_tt_create(struct ttm_bo_device *bdev, - unsigned long size, uint32_t page_flags) +static struct ttm_tt *radeon_ttm_tt_create(struct ttm_buffer_object *bo, + uint32_t page_flags) { struct radeon_device *rdev; struct radeon_ttm_tt *gtt;
- rdev = radeon_get_rdev(bdev); + rdev = radeon_get_rdev(bo->bdev); #if IS_ENABLED(CONFIG_AGP) if (rdev->flags & RADEON_IS_AGP) { - return ttm_agp_tt_create(bdev, rdev->ddev->agp->bridge, - size, page_flags); + return ttm_agp_tt_create(bo, rdev->ddev->agp->bridge, + page_flags); } #endif
@@ -706,7 +706,7 @@ static struct ttm_tt *radeon_ttm_tt_create(struct ttm_bo_device *bdev, } gtt->ttm.ttm.func = &radeon_backend_func; gtt->rdev = rdev; - if (ttm_dma_tt_init(>t->ttm, bdev, size, page_flags)) { + if (ttm_dma_tt_init(>t->ttm, bo, page_flags)) { kfree(gtt); return NULL; } diff --git a/drivers/gpu/drm/ttm/ttm_agp_backend.c b/drivers/gpu/drm/ttm/ttm_agp_backend.c index f7c2aefbec7c..7c2485fe88d8 100644 --- a/drivers/gpu/drm/ttm/ttm_agp_backend.c +++ b/drivers/gpu/drm/ttm/ttm_agp_backend.c @@ -110,9 +110,9 @@ static struct ttm_backend_func ttm_agp_func = { .destroy = ttm_agp_destroy, };
-struct ttm_tt *ttm_agp_tt_create(struct ttm_bo_device *bdev, +struct ttm_tt *ttm_agp_tt_create(struct ttm_buffer_object *bo, struct agp_bridge_data *bridge, - unsigned long size, uint32_t page_flags) + uint32_t page_flags) { struct ttm_agp_backend *agp_be;
@@ -124,7 +124,7 @@ struct ttm_tt *ttm_agp_tt_create(struct ttm_bo_device *bdev, agp_be->bridge = bridge; agp_be->ttm.func = &ttm_agp_func;
- if (ttm_tt_init(&agp_be->ttm, bdev, size, page_flags)) { + if (ttm_tt_init(&agp_be->ttm, bo, page_flags)) { kfree(agp_be); return NULL; } diff --git a/drivers/gpu/drm/ttm/ttm_tt.c b/drivers/gpu/drm/ttm/ttm_tt.c index 971133106ec2..65976238d24b 100644 --- a/drivers/gpu/drm/ttm/ttm_tt.c +++ b/drivers/gpu/drm/ttm/ttm_tt.c @@ -73,8 +73,7 @@ int ttm_tt_create(struct ttm_buffer_object *bo, bool zero_alloc) return -EINVAL; }
- bo->ttm = bdev->driver->ttm_tt_create(bdev, bo->num_pages << PAGE_SHIFT, - page_flags); + bo->ttm = bdev->driver->ttm_tt_create(bo, page_flags); if (unlikely(bo->ttm == NULL)) return -ENOMEM;
@@ -237,21 +236,21 @@ void ttm_tt_destroy(struct ttm_tt *ttm) ttm->func->destroy(ttm); }
-void ttm_tt_init_fields(struct ttm_tt *ttm, struct ttm_bo_device *bdev, - unsigned long size, uint32_t page_flags) +void ttm_tt_init_fields(struct ttm_tt *ttm, struct ttm_buffer_object *bo, + uint32_t page_flags) { - ttm->bdev = bdev; - ttm->num_pages = (size + PAGE_SIZE - 1) >> PAGE_SHIFT; + ttm->bdev = bo->bdev; + ttm->num_pages = bo->num_pages; ttm->caching_state = tt_cached; ttm->page_flags = page_flags; ttm->state = tt_unpopulated; ttm->swap_storage = NULL; }
-int ttm_tt_init(struct ttm_tt *ttm, struct ttm_bo_device *bdev, - unsigned long size, uint32_t page_flags) +int ttm_tt_init(struct ttm_tt *ttm, struct ttm_buffer_object *bo, + uint32_t page_flags) { - ttm_tt_init_fields(ttm, bdev, size, page_flags); + ttm_tt_init_fields(ttm, bo, page_flags);
if (ttm_tt_alloc_page_directory(ttm)) { ttm_tt_destroy(ttm); @@ -269,12 +268,12 @@ void ttm_tt_fini(struct ttm_tt *ttm) } EXPORT_SYMBOL(ttm_tt_fini);
-int ttm_dma_tt_init(struct ttm_dma_tt *ttm_dma, struct ttm_bo_device *bdev, - unsigned long size, uint32_t page_flags) +int ttm_dma_tt_init(struct ttm_dma_tt *ttm_dma, struct ttm_buffer_object *bo, + uint32_t page_flags) { struct ttm_tt *ttm = &ttm_dma->ttm;
- ttm_tt_init_fields(ttm, bdev, size, page_flags); + ttm_tt_init_fields(ttm, bo, page_flags);
INIT_LIST_HEAD(&ttm_dma->pages_list); if (ttm_dma_tt_alloc_page_directory(ttm_dma)) { @@ -286,13 +285,13 @@ int ttm_dma_tt_init(struct ttm_dma_tt *ttm_dma, struct ttm_bo_device *bdev, } EXPORT_SYMBOL(ttm_dma_tt_init);
-int ttm_sg_tt_init(struct ttm_dma_tt *ttm_dma, struct ttm_bo_device *bdev, - unsigned long size, uint32_t page_flags) +int ttm_sg_tt_init(struct ttm_dma_tt *ttm_dma, struct ttm_buffer_object *bo, + uint32_t page_flags) { struct ttm_tt *ttm = &ttm_dma->ttm; int ret;
- ttm_tt_init_fields(ttm, bdev, size, page_flags); + ttm_tt_init_fields(ttm, bo, page_flags);
INIT_LIST_HEAD(&ttm_dma->pages_list); if (page_flags & TTM_PAGE_FLAG_SG) diff --git a/drivers/gpu/drm/virtio/virtgpu_ttm.c b/drivers/gpu/drm/virtio/virtgpu_ttm.c index ee9839fbae66..01618afc6bce 100644 --- a/drivers/gpu/drm/virtio/virtgpu_ttm.c +++ b/drivers/gpu/drm/virtio/virtgpu_ttm.c @@ -324,20 +324,19 @@ static struct ttm_backend_func virtio_gpu_backend_func = { .destroy = &virtio_gpu_ttm_backend_destroy, };
-static struct ttm_tt *virtio_gpu_ttm_tt_create(struct ttm_bo_device *bdev, - unsigned long size, +static struct ttm_tt *virtio_gpu_ttm_tt_create(struct ttm_buffer_object *bo, uint32_t page_flags) { struct virtio_gpu_device *vgdev; struct virtio_gpu_ttm_tt *gtt;
- vgdev = virtio_gpu_get_vgdev(bdev); + vgdev = virtio_gpu_get_vgdev(bo->bdev); gtt = kzalloc(sizeof(struct virtio_gpu_ttm_tt), GFP_KERNEL); if (gtt == NULL) return NULL; gtt->ttm.ttm.func = &virtio_gpu_backend_func; gtt->vgdev = vgdev; - if (ttm_dma_tt_init(>t->ttm, bdev, size, page_flags)) { + if (ttm_dma_tt_init(>t->ttm, bo, page_flags)) { kfree(gtt); return NULL; } diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_buffer.c b/drivers/gpu/drm/vmwgfx/vmwgfx_buffer.c index fead3f2dbb46..7177eecb8c9f 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_buffer.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_buffer.c @@ -693,8 +693,8 @@ static struct ttm_backend_func vmw_ttm_func = { .destroy = vmw_ttm_destroy, };
-static struct ttm_tt *vmw_ttm_tt_create(struct ttm_bo_device *bdev, - unsigned long size, uint32_t page_flags) +static struct ttm_tt *vmw_ttm_tt_create(struct ttm_buffer_object *bo, + uint32_t page_flags) { struct vmw_ttm_tt *vmw_be; int ret; @@ -704,13 +704,13 @@ static struct ttm_tt *vmw_ttm_tt_create(struct ttm_bo_device *bdev, return NULL;
vmw_be->dma_ttm.ttm.func = &vmw_ttm_func; - vmw_be->dev_priv = container_of(bdev, struct vmw_private, bdev); + vmw_be->dev_priv = container_of(bo->bdev, struct vmw_private, bdev); vmw_be->mob = NULL;
if (vmw_be->dev_priv->map_mode == vmw_dma_alloc_coherent) - ret = ttm_dma_tt_init(&vmw_be->dma_ttm, bdev, size, page_flags); + ret = ttm_dma_tt_init(&vmw_be->dma_ttm, bo, page_flags); else - ret = ttm_tt_init(&vmw_be->dma_ttm.ttm, bdev, size, page_flags); + ret = ttm_tt_init(&vmw_be->dma_ttm.ttm, bo, page_flags); if (unlikely(ret != 0)) goto out_no_init;
diff --git a/drivers/staging/vboxvideo/vbox_ttm.c b/drivers/staging/vboxvideo/vbox_ttm.c index 2c7daa3d0f24..548edb7c494b 100644 --- a/drivers/staging/vboxvideo/vbox_ttm.c +++ b/drivers/staging/vboxvideo/vbox_ttm.c @@ -193,8 +193,7 @@ static struct ttm_backend_func vbox_tt_backend_func = { .destroy = &vbox_ttm_backend_destroy, };
-static struct ttm_tt *vbox_ttm_tt_create(struct ttm_bo_device *bdev, - unsigned long size, +static struct ttm_tt *vbox_ttm_tt_create(struct ttm_buffer_object *bo, u32 page_flags) { struct ttm_tt *tt; @@ -204,7 +203,7 @@ static struct ttm_tt *vbox_ttm_tt_create(struct ttm_bo_device *bdev, return NULL;
tt->func = &vbox_tt_backend_func; - if (ttm_tt_init(tt, bdev, size, page_flags)) { + if (ttm_tt_init(tt, bo, page_flags)) { kfree(tt); return NULL; } diff --git a/include/drm/ttm/ttm_bo_driver.h b/include/drm/ttm/ttm_bo_driver.h index 39cd6b086d3a..3234cc322e70 100644 --- a/include/drm/ttm/ttm_bo_driver.h +++ b/include/drm/ttm/ttm_bo_driver.h @@ -225,8 +225,7 @@ struct ttm_bo_driver { /** * ttm_tt_create * - * @bdev: pointer to a struct ttm_bo_device: - * @size: Size of the data needed backing. + * @bo: The buffer object to create the ttm for. * @page_flags: Page flags as identified by TTM_PAGE_FLAG_XX flags. * * Create a struct ttm_tt to back data with system memory pages. @@ -234,8 +233,7 @@ struct ttm_bo_driver { * Returns: * NULL: Out of memory. */ - struct ttm_tt *(*ttm_tt_create)(struct ttm_bo_device *bdev, - unsigned long size, + struct ttm_tt *(*ttm_tt_create)(struct ttm_buffer_object *bo, uint32_t page_flags);
/** diff --git a/include/drm/ttm/ttm_tt.h b/include/drm/ttm/ttm_tt.h index 1cf316a4257c..c0e928abf592 100644 --- a/include/drm/ttm/ttm_tt.h +++ b/include/drm/ttm/ttm_tt.h @@ -150,8 +150,7 @@ int ttm_tt_create(struct ttm_buffer_object *bo, bool zero_alloc); * ttm_tt_init * * @ttm: The struct ttm_tt. - * @bdev: pointer to a struct ttm_bo_device: - * @size: Size of the data needed backing. + * @bo: The buffer object we create the ttm for. * @page_flags: Page flags as identified by TTM_PAGE_FLAG_XX flags. * * Create a struct ttm_tt to back data with system memory pages. @@ -159,12 +158,12 @@ int ttm_tt_create(struct ttm_buffer_object *bo, bool zero_alloc); * Returns: * NULL: Out of memory. */ -int ttm_tt_init(struct ttm_tt *ttm, struct ttm_bo_device *bdev, - unsigned long size, uint32_t page_flags); -int ttm_dma_tt_init(struct ttm_dma_tt *ttm_dma, struct ttm_bo_device *bdev, - unsigned long size, uint32_t page_flags); -int ttm_sg_tt_init(struct ttm_dma_tt *ttm_dma, struct ttm_bo_device *bdev, - unsigned long size, uint32_t page_flags); +int ttm_tt_init(struct ttm_tt *ttm, struct ttm_buffer_object *bo, + uint32_t page_flags); +int ttm_dma_tt_init(struct ttm_dma_tt *ttm_dma, struct ttm_buffer_object *bo, + uint32_t page_flags); +int ttm_sg_tt_init(struct ttm_dma_tt *ttm_dma, struct ttm_buffer_object *bo, + uint32_t page_flags);
/** * ttm_tt_fini @@ -254,9 +253,8 @@ void ttm_tt_unpopulate(struct ttm_tt *ttm); /** * ttm_agp_tt_create * - * @bdev: Pointer to a struct ttm_bo_device. + * @bo: Buffer object we allocate the ttm for. * @bridge: The agp bridge this device is sitting on. - * @size: Size of the data needed backing. * @page_flags: Page flags as identified by TTM_PAGE_FLAG_XX flags. * * @@ -264,9 +262,9 @@ void ttm_tt_unpopulate(struct ttm_tt *ttm); * for TT memory. This function uses the linux agpgart interface to * bind and unbind memory backing a ttm_tt. */ -struct ttm_tt *ttm_agp_tt_create(struct ttm_bo_device *bdev, +struct ttm_tt *ttm_agp_tt_create(struct ttm_buffer_object *bo, struct agp_bridge_data *bridge, - unsigned long size, uint32_t page_flags); + uint32_t page_flags); int ttm_agp_tt_populate(struct ttm_tt *ttm, struct ttm_operation_ctx *ctx); void ttm_agp_tt_unpopulate(struct ttm_tt *ttm); #endif
Better to set this with all other fields as well.
Signed-off-by: Christian König christian.koenig@amd.com --- drivers/gpu/drm/ttm/ttm_tt.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-)
diff --git a/drivers/gpu/drm/ttm/ttm_tt.c b/drivers/gpu/drm/ttm/ttm_tt.c index 65976238d24b..7e672be987b5 100644 --- a/drivers/gpu/drm/ttm/ttm_tt.c +++ b/drivers/gpu/drm/ttm/ttm_tt.c @@ -77,9 +77,6 @@ int ttm_tt_create(struct ttm_buffer_object *bo, bool zero_alloc) if (unlikely(bo->ttm == NULL)) return -ENOMEM;
- if (bo->type == ttm_bo_type_sg) - bo->ttm->sg = bo->sg; - return 0; }
@@ -245,6 +242,7 @@ void ttm_tt_init_fields(struct ttm_tt *ttm, struct ttm_buffer_object *bo, ttm->page_flags = page_flags; ttm->state = tt_unpopulated; ttm->swap_storage = NULL; + ttm->sg = bo->sg; }
int ttm_tt_init(struct ttm_tt *ttm, struct ttm_buffer_object *bo,
Instead of moving this to the SYSTEM domain just drop the backing store and let the resulting allocation be freed.
Signed-off-by: Christian König christian.koenig@amd.com --- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 6 ++++++ 1 file changed, 6 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index 44ae2c60c9d4..eebc9f77b611 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -203,6 +203,12 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo, .flags = TTM_PL_MASK_CACHING | TTM_PL_FLAG_SYSTEM };
+ if (bo->type == ttm_bo_type_sg) { + placement->num_placement = 0; + placement->num_busy_placement = 0; + return; + } + if (!amdgpu_ttm_bo_is_amdgpu_bo(bo)) { placement->placement = &placements; placement->busy_placement = &placements;
Just set the GTT domain as mandatory, so that the BO is validated into it on first use. This allows us to setup the sg table later on.
Signed-off-by: Christian König christian.koenig@amd.com --- drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c index fb66b45548d3..44230165e181 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c @@ -106,10 +106,12 @@ amdgpu_gem_prime_import_sg_table(struct drm_device *dev,
ww_mutex_lock(&resv->lock, NULL); ret = amdgpu_bo_create(adev, attach->dmabuf->size, PAGE_SIZE, false, - AMDGPU_GEM_DOMAIN_GTT, 0, sg, resv, &bo); + AMDGPU_GEM_DOMAIN_CPU, 0, sg, resv, &bo); if (ret) goto error;
+ bo->allowed_domains = AMDGPU_GEM_DOMAIN_GTT; + bo->preferred_domains = AMDGPU_GEM_DOMAIN_GTT; if (attach->dmabuf->ops != &amdgpu_dmabuf_ops) bo->prime_shared_count = 1;
Drop the "kernel" and sg parameter and give the BO type to create explicit to amdgpu_bo_create instead of figuring it out from the parameters.
Signed-off-by: Christian König christian.koenig@amd.com --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 5 +-- drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c | 8 ++--- drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c | 7 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 6 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 46 +++++++++++---------------- drivers/gpu/drm/amd/amdgpu/amdgpu_object.h | 11 +++---- drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c | 7 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_test.c | 11 +++---- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 11 ++++--- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 8 ++--- 11 files changed, 58 insertions(+), 64 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 292c7e72820c..b1116b773516 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -441,7 +441,7 @@ struct amdgpu_sa_bo { void amdgpu_gem_force_release(struct amdgpu_device *adev); int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size, int alignment, u32 initial_domain, - u64 flags, bool kernel, + u64 flags, enum ttm_bo_type type, struct reservation_object *resv, struct drm_gem_object **obj);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c index 450426dbed92..7f096ed6e83d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c @@ -215,8 +215,9 @@ int alloc_gtt_mem(struct kgd_dev *kgd, size_t size, if ((*mem) == NULL) return -ENOMEM;
- r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_GTT, - AMDGPU_GEM_CREATE_CPU_GTT_USWC, NULL, NULL, &(*mem)->bo); + r = amdgpu_bo_create(adev, size, PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT, + AMDGPU_GEM_CREATE_CPU_GTT_USWC, ttm_bo_type_kernel, + NULL, &(*mem)->bo); if (r) { dev_err(adev->dev, "failed to allocate BO for amdkfd (%d)\n", r); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c index 2fb299afc12b..02b849be083b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c @@ -80,8 +80,8 @@ static void amdgpu_benchmark_move(struct amdgpu_device *adev, unsigned size, int time;
n = AMDGPU_BENCHMARK_ITERATIONS; - r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, sdomain, 0, NULL, - NULL, &sobj); + r = amdgpu_bo_create(adev, size, PAGE_SIZE,sdomain, 0, + ttm_bo_type_kernel, NULL, &sobj); if (r) { goto out_cleanup; } @@ -93,8 +93,8 @@ static void amdgpu_benchmark_move(struct amdgpu_device *adev, unsigned size, if (r) { goto out_cleanup; } - r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, ddomain, 0, NULL, - NULL, &dobj); + r = amdgpu_bo_create(adev, size, PAGE_SIZE, ddomain, 0, + ttm_bo_type_kernel, NULL, &dobj); if (r) { goto out_cleanup; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c index dc8d9f3216fa..cf0f186c6092 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c @@ -113,11 +113,12 @@ int amdgpu_gart_table_vram_alloc(struct amdgpu_device *adev) int r;
if (adev->gart.robj == NULL) { - r = amdgpu_bo_create(adev, adev->gart.table_size, - PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_VRAM, + r = amdgpu_bo_create(adev, adev->gart.table_size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, - NULL, NULL, &adev->gart.robj); + ttm_bo_type_kernel, NULL, + &adev->gart.robj); if (r) { return r; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index 55a840ae6d68..bb9b21266b67 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -45,7 +45,7 @@ void amdgpu_gem_object_free(struct drm_gem_object *gobj)
int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size, int alignment, u32 initial_domain, - u64 flags, bool kernel, + u64 flags, enum ttm_bo_type type, struct reservation_object *resv, struct drm_gem_object **obj) { @@ -59,8 +59,8 @@ int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size, }
retry: - r = amdgpu_bo_create(adev, size, alignment, kernel, initial_domain, - flags, NULL, resv, &bo); + r = amdgpu_bo_create(adev, size, alignment, initial_domain, + flags, type, resv, &bo); if (r) { if (r != -ERESTARTSYS) { if (flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 216799ccb545..8f5027e8241a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -187,10 +187,10 @@ int amdgpu_bo_create_reserved(struct amdgpu_device *adev, int r;
if (!*bo_ptr) { - r = amdgpu_bo_create(adev, size, align, true, domain, + r = amdgpu_bo_create(adev, size, align, domain, AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, - NULL, NULL, bo_ptr); + ttm_bo_type_kernel, NULL, bo_ptr); if (r) { dev_err(adev->dev, "(%d) failed to allocate kernel bo\n", r); @@ -331,21 +331,19 @@ static bool amdgpu_bo_validate_size(struct amdgpu_device *adev, return false; }
-static int amdgpu_bo_do_create(struct amdgpu_device *adev, - unsigned long size, int byte_align, - bool kernel, u32 domain, u64 flags, - struct sg_table *sg, +static int amdgpu_bo_do_create(struct amdgpu_device *adev, unsigned long size, + int byte_align, u32 domain, + u64 flags, enum ttm_bo_type type, struct reservation_object *resv, struct amdgpu_bo **bo_ptr) { struct ttm_operation_ctx ctx = { - .interruptible = !kernel, + .interruptible = (type != ttm_bo_type_kernel), .no_wait_gpu = false, .resv = resv, .flags = TTM_OPT_FLAG_ALLOW_RES_EVICT }; struct amdgpu_bo *bo; - enum ttm_bo_type type; unsigned long page_align; size_t acc_size; int r; @@ -356,13 +354,6 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, if (!amdgpu_bo_validate_size(adev, size, domain)) return -ENOMEM;
- if (kernel) { - type = ttm_bo_type_kernel; - } else if (sg) { - type = ttm_bo_type_sg; - } else { - type = ttm_bo_type_device; - } *bo_ptr = NULL;
acc_size = ttm_bo_dma_acc_size(&adev->mman.bdev, size, @@ -381,7 +372,8 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, AMDGPU_GEM_DOMAIN_GWS | AMDGPU_GEM_DOMAIN_OA); bo->allowed_domains = bo->preferred_domains; - if (!kernel && bo->allowed_domains == AMDGPU_GEM_DOMAIN_VRAM) + if (type != ttm_bo_type_kernel && + bo->allowed_domains == AMDGPU_GEM_DOMAIN_VRAM) bo->allowed_domains |= AMDGPU_GEM_DOMAIN_GTT;
bo->flags = flags; @@ -419,7 +411,7 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev,
r = ttm_bo_init_reserved(&adev->mman.bdev, &bo->tbo, size, type, &bo->placement, page_align, &ctx, acc_size, - sg, resv, &amdgpu_ttm_bo_destroy); + NULL, resv, &amdgpu_ttm_bo_destroy); if (unlikely(r != 0)) return r;
@@ -431,7 +423,7 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, else amdgpu_cs_report_moved_bytes(adev, ctx.bytes_moved, 0);
- if (kernel) + if (type == ttm_bo_type_kernel) bo->tbo.priority = 1;
if (flags & AMDGPU_GEM_CREATE_VRAM_CLEARED && @@ -475,12 +467,11 @@ static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, if (bo->shadow) return 0;
- r = amdgpu_bo_do_create(adev, size, byte_align, true, - AMDGPU_GEM_DOMAIN_GTT, + r = amdgpu_bo_do_create(adev, size, byte_align, AMDGPU_GEM_DOMAIN_GTT, AMDGPU_GEM_CREATE_CPU_GTT_USWC | AMDGPU_GEM_CREATE_SHADOW, - NULL, bo->tbo.resv, - &bo->shadow); + ttm_bo_type_kernel, + bo->tbo.resv, &bo->shadow); if (!r) { bo->shadow->parent = amdgpu_bo_ref(bo); mutex_lock(&adev->shadow_list_lock); @@ -491,18 +482,17 @@ static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, return r; }
-int amdgpu_bo_create(struct amdgpu_device *adev, - unsigned long size, int byte_align, - bool kernel, u32 domain, u64 flags, - struct sg_table *sg, +int amdgpu_bo_create(struct amdgpu_device *adev, unsigned long size, + int byte_align, u32 domain, + u64 flags, enum ttm_bo_type type, struct reservation_object *resv, struct amdgpu_bo **bo_ptr) { uint64_t parent_flags = flags & ~AMDGPU_GEM_CREATE_SHADOW; int r;
- r = amdgpu_bo_do_create(adev, size, byte_align, kernel, domain, - parent_flags, sg, resv, bo_ptr); + r = amdgpu_bo_do_create(adev, size, byte_align, domain, + parent_flags, type, resv, bo_ptr); if (r) return r;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h index 1cef944ef98d..deb10ecd89a1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h @@ -201,12 +201,11 @@ static inline bool amdgpu_bo_explicit_sync(struct amdgpu_bo *bo) return bo->flags & AMDGPU_GEM_CREATE_EXPLICIT_SYNC; }
-int amdgpu_bo_create(struct amdgpu_device *adev, - unsigned long size, int byte_align, - bool kernel, u32 domain, u64 flags, - struct sg_table *sg, - struct reservation_object *resv, - struct amdgpu_bo **bo_ptr); +int amdgpu_bo_create(struct amdgpu_device *adev, unsigned long size, + int byte_align, u32 domain, + u64 flags, enum ttm_bo_type type, + struct reservation_object *resv, + struct amdgpu_bo **bo_ptr); int amdgpu_bo_create_reserved(struct amdgpu_device *adev, unsigned long size, int align, u32 domain, struct amdgpu_bo **bo_ptr, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c index 44230165e181..1c9991738477 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c @@ -105,11 +105,14 @@ amdgpu_gem_prime_import_sg_table(struct drm_device *dev, int ret;
ww_mutex_lock(&resv->lock, NULL); - ret = amdgpu_bo_create(adev, attach->dmabuf->size, PAGE_SIZE, false, - AMDGPU_GEM_DOMAIN_CPU, 0, sg, resv, &bo); + ret = amdgpu_bo_create(adev, attach->dmabuf->size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_CPU, 0, ttm_bo_type_sg, + resv, &bo); if (ret) goto error;
+ bo->tbo.sg = sg; + bo->tbo.ttm->sg = sg; bo->allowed_domains = AMDGPU_GEM_DOMAIN_GTT; bo->preferred_domains = AMDGPU_GEM_DOMAIN_GTT; if (attach->dmabuf->ops != &amdgpu_dmabuf_ops) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c index f3d81b6fb499..2dbe87591f81 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c @@ -59,9 +59,8 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) goto out_cleanup; }
- r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_VRAM, 0, - NULL, NULL, &vram_obj); + r = amdgpu_bo_create(adev, size, PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, 0, + ttm_bo_type_kernel, NULL, &vram_obj); if (r) { DRM_ERROR("Failed to create VRAM object\n"); goto out_cleanup; @@ -80,9 +79,9 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) void **vram_start, **vram_end; struct dma_fence *fence = NULL;
- r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_GTT, 0, NULL, - NULL, gtt_obj + i); + r = amdgpu_bo_create(adev, size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_GTT, 0, + ttm_bo_type_kernel, NULL, gtt_obj + i); if (r) { DRM_ERROR("Failed to create GTT object %d\n", i); goto out_lclean; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index eebc9f77b611..1ec54629f277 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -1317,11 +1317,12 @@ static int amdgpu_ttm_fw_reserve_vram_init(struct amdgpu_device *adev) if (adev->fw_vram_usage.size > 0 && adev->fw_vram_usage.size <= vram_size) {
- r = amdgpu_bo_create(adev, adev->fw_vram_usage.size, - PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | - AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, NULL, NULL, - &adev->fw_vram_usage.reserved_bo); + r = amdgpu_bo_create(adev, adev->fw_vram_usage.size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | + AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, + ttm_bo_type_kernel, NULL, + &adev->fw_vram_usage.reserved_bo); if (r) goto error_create;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 0b237e027cab..24474294c92a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -413,9 +413,9 @@ static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev, if (!entry->base.bo) { r = amdgpu_bo_create(adev, amdgpu_vm_bo_size(adev, level), - AMDGPU_GPU_PAGE_SIZE, true, + AMDGPU_GPU_PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, flags, - NULL, resv, &pt); + ttm_bo_type_kernel, resv, &pt); if (r) return r;
@@ -2409,8 +2409,8 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, AMDGPU_GEM_CREATE_SHADOW);
size = amdgpu_vm_bo_size(adev, adev->vm_manager.root_level); - r = amdgpu_bo_create(adev, size, align, true, AMDGPU_GEM_DOMAIN_VRAM, - flags, NULL, NULL, &vm->root.base.bo); + r = amdgpu_bo_create(adev, size, align, AMDGPU_GEM_DOMAIN_VRAM, flags, + ttm_bo_type_kernel, NULL, &vm->root.base.bo); if (r) goto error_free_sched_entity;
NAK.
For KFD we need the ability to create a BO from an SG list that doesn't come from another BO. We use this for mapping pages from the doorbell aperture into GPUVM for GPU self-dispatch.
If you remove this now, I'll need to add it back in some form in a month or two when I get to that part of upstreaming KFD.
There may be other ways to implement this. Currently we need to create a BO for anything we want to map into a GPUVM page table, because the amdgpu_vm code is based around BOs. If there was a way to map physical addresses into GPUVM without creating a buffer object, that would work too.
Regards, Felix
On 2018-03-06 09:43 AM, Christian König wrote:
Drop the "kernel" and sg parameter and give the BO type to create explicit to amdgpu_bo_create instead of figuring it out from the parameters.
Signed-off-by: Christian König christian.koenig@amd.com
drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 5 +-- drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c | 8 ++--- drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c | 7 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 6 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 46 +++++++++++---------------- drivers/gpu/drm/amd/amdgpu/amdgpu_object.h | 11 +++---- drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c | 7 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_test.c | 11 +++---- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 11 ++++--- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 8 ++--- 11 files changed, 58 insertions(+), 64 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 292c7e72820c..b1116b773516 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -441,7 +441,7 @@ struct amdgpu_sa_bo { void amdgpu_gem_force_release(struct amdgpu_device *adev); int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size, int alignment, u32 initial_domain,
u64 flags, bool kernel,
u64 flags, enum ttm_bo_type type, struct reservation_object *resv, struct drm_gem_object **obj);diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c index 450426dbed92..7f096ed6e83d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c @@ -215,8 +215,9 @@ int alloc_gtt_mem(struct kgd_dev *kgd, size_t size, if ((*mem) == NULL) return -ENOMEM;
- r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_GTT,
AMDGPU_GEM_CREATE_CPU_GTT_USWC, NULL, NULL, &(*mem)->bo);
- r = amdgpu_bo_create(adev, size, PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT,
AMDGPU_GEM_CREATE_CPU_GTT_USWC, ttm_bo_type_kernel,NULL, &(*mem)->bo);diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c index 2fb299afc12b..02b849be083b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c @@ -80,8 +80,8 @@ static void amdgpu_benchmark_move(struct amdgpu_device *adev, unsigned size, int time;
n = AMDGPU_BENCHMARK_ITERATIONS;
- r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, sdomain, 0, NULL,
NULL, &sobj);
- r = amdgpu_bo_create(adev, size, PAGE_SIZE,sdomain, 0,
ttm_bo_type_kernel, NULL, &sobj);@@ -93,8 +93,8 @@ static void amdgpu_benchmark_move(struct amdgpu_device *adev, unsigned size, if (r) { goto out_cleanup; }
- r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, ddomain, 0, NULL,
NULL, &dobj);
- r = amdgpu_bo_create(adev, size, PAGE_SIZE, ddomain, 0,
ttm_bo_type_kernel, NULL, &dobj);diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c index dc8d9f3216fa..cf0f186c6092 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c @@ -113,11 +113,12 @@ int amdgpu_gart_table_vram_alloc(struct amdgpu_device *adev) int r;
if (adev->gart.robj == NULL) {
r = amdgpu_bo_create(adev, adev->gart.table_size,PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_VRAM,
r = amdgpu_bo_create(adev, adev->gart.table_size, PAGE_SIZE,AMDGPU_GEM_DOMAIN_VRAM, AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS,
NULL, NULL, &adev->gart.robj);
ttm_bo_type_kernel, NULL,&adev->gart.robj);diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index 55a840ae6d68..bb9b21266b67 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -45,7 +45,7 @@ void amdgpu_gem_object_free(struct drm_gem_object *gobj)
int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size, int alignment, u32 initial_domain,
u64 flags, bool kernel,
u64 flags, enum ttm_bo_type type, struct reservation_object *resv, struct drm_gem_object **obj){ @@ -59,8 +59,8 @@ int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size, }
retry:
- r = amdgpu_bo_create(adev, size, alignment, kernel, initial_domain,
flags, NULL, resv, &bo);
- r = amdgpu_bo_create(adev, size, alignment, initial_domain,
flags, type, resv, &bo);diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 216799ccb545..8f5027e8241a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -187,10 +187,10 @@ int amdgpu_bo_create_reserved(struct amdgpu_device *adev, int r;
if (!*bo_ptr) {
r = amdgpu_bo_create(adev, size, align, true, domain,
r = amdgpu_bo_create(adev, size, align, domain, AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS,
NULL, NULL, bo_ptr);
ttm_bo_type_kernel, NULL, bo_ptr);@@ -331,21 +331,19 @@ static bool amdgpu_bo_validate_size(struct amdgpu_device *adev, return false; }
-static int amdgpu_bo_do_create(struct amdgpu_device *adev,
unsigned long size, int byte_align,bool kernel, u32 domain, u64 flags,struct sg_table *sg,+static int amdgpu_bo_do_create(struct amdgpu_device *adev, unsigned long size,
int byte_align, u32 domain,u64 flags, enum ttm_bo_type type, struct reservation_object *resv, struct amdgpu_bo **bo_ptr){ struct ttm_operation_ctx ctx = {
.interruptible = !kernel,
.interruptible = (type != ttm_bo_type_kernel),
- enum ttm_bo_type type; unsigned long page_align; size_t acc_size; int r;
@@ -356,13 +354,6 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, if (!amdgpu_bo_validate_size(adev, size, domain)) return -ENOMEM;
if (kernel) {
type = ttm_bo_type_kernel;} else if (sg) {
type = ttm_bo_type_sg;} else {
type = ttm_bo_type_device;} *bo_ptr = NULL;
acc_size = ttm_bo_dma_acc_size(&adev->mman.bdev, size,
@@ -381,7 +372,8 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, AMDGPU_GEM_DOMAIN_GWS | AMDGPU_GEM_DOMAIN_OA); bo->allowed_domains = bo->preferred_domains;
- if (!kernel && bo->allowed_domains == AMDGPU_GEM_DOMAIN_VRAM)
if (type != ttm_bo_type_kernel &&
bo->allowed_domains == AMDGPU_GEM_DOMAIN_VRAM)bo->allowed_domains |= AMDGPU_GEM_DOMAIN_GTT;
bo->flags = flags;
@@ -419,7 +411,7 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev,
r = ttm_bo_init_reserved(&adev->mman.bdev, &bo->tbo, size, type, &bo->placement, page_align, &ctx, acc_size,
sg, resv, &amdgpu_ttm_bo_destroy);
NULL, resv, &amdgpu_ttm_bo_destroy);@@ -431,7 +423,7 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, else amdgpu_cs_report_moved_bytes(adev, ctx.bytes_moved, 0);
- if (kernel)
if (type == ttm_bo_type_kernel) bo->tbo.priority = 1;
if (flags & AMDGPU_GEM_CREATE_VRAM_CLEARED &&
@@ -475,12 +467,11 @@ static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, if (bo->shadow) return 0;
- r = amdgpu_bo_do_create(adev, size, byte_align, true,
AMDGPU_GEM_DOMAIN_GTT,
- r = amdgpu_bo_do_create(adev, size, byte_align, AMDGPU_GEM_DOMAIN_GTT, AMDGPU_GEM_CREATE_CPU_GTT_USWC | AMDGPU_GEM_CREATE_SHADOW,
NULL, bo->tbo.resv,&bo->shadow);
ttm_bo_type_kernel,bo->tbo.resv, &bo->shadow);@@ -491,18 +482,17 @@ static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, return r; }
-int amdgpu_bo_create(struct amdgpu_device *adev,
unsigned long size, int byte_align,bool kernel, u32 domain, u64 flags,struct sg_table *sg,+int amdgpu_bo_create(struct amdgpu_device *adev, unsigned long size,
int byte_align, u32 domain,u64 flags, enum ttm_bo_type type, struct reservation_object *resv, struct amdgpu_bo **bo_ptr){ uint64_t parent_flags = flags & ~AMDGPU_GEM_CREATE_SHADOW; int r;
- r = amdgpu_bo_do_create(adev, size, byte_align, kernel, domain,
parent_flags, sg, resv, bo_ptr);
- r = amdgpu_bo_do_create(adev, size, byte_align, domain,
parent_flags, type, resv, bo_ptr);diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h index 1cef944ef98d..deb10ecd89a1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h @@ -201,12 +201,11 @@ static inline bool amdgpu_bo_explicit_sync(struct amdgpu_bo *bo) return bo->flags & AMDGPU_GEM_CREATE_EXPLICIT_SYNC; }
-int amdgpu_bo_create(struct amdgpu_device *adev,
unsigned long size, int byte_align,bool kernel, u32 domain, u64 flags,struct sg_table *sg,struct reservation_object *resv,struct amdgpu_bo **bo_ptr);+int amdgpu_bo_create(struct amdgpu_device *adev, unsigned long size,
int byte_align, u32 domain,u64 flags, enum ttm_bo_type type,struct reservation_object *resv,struct amdgpu_bo **bo_ptr);int amdgpu_bo_create_reserved(struct amdgpu_device *adev, unsigned long size, int align, u32 domain, struct amdgpu_bo **bo_ptr, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c index 44230165e181..1c9991738477 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c @@ -105,11 +105,14 @@ amdgpu_gem_prime_import_sg_table(struct drm_device *dev, int ret;
ww_mutex_lock(&resv->lock, NULL);
- ret = amdgpu_bo_create(adev, attach->dmabuf->size, PAGE_SIZE, false,
AMDGPU_GEM_DOMAIN_CPU, 0, sg, resv, &bo);
ret = amdgpu_bo_create(adev, attach->dmabuf->size, PAGE_SIZE,
AMDGPU_GEM_DOMAIN_CPU, 0, ttm_bo_type_sg,resv, &bo);if (ret) goto error;
bo->tbo.sg = sg;
bo->tbo.ttm->sg = sg; bo->allowed_domains = AMDGPU_GEM_DOMAIN_GTT; bo->preferred_domains = AMDGPU_GEM_DOMAIN_GTT; if (attach->dmabuf->ops != &amdgpu_dmabuf_ops)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c index f3d81b6fb499..2dbe87591f81 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c @@ -59,9 +59,8 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) goto out_cleanup; }
- r = amdgpu_bo_create(adev, size, PAGE_SIZE, true,
AMDGPU_GEM_DOMAIN_VRAM, 0,NULL, NULL, &vram_obj);
- r = amdgpu_bo_create(adev, size, PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, 0,
ttm_bo_type_kernel, NULL, &vram_obj);@@ -80,9 +79,9 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) void **vram_start, **vram_end; struct dma_fence *fence = NULL;
r = amdgpu_bo_create(adev, size, PAGE_SIZE, true,AMDGPU_GEM_DOMAIN_GTT, 0, NULL,NULL, gtt_obj + i);
r = amdgpu_bo_create(adev, size, PAGE_SIZE,AMDGPU_GEM_DOMAIN_GTT, 0,ttm_bo_type_kernel, NULL, gtt_obj + i);diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index eebc9f77b611..1ec54629f277 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -1317,11 +1317,12 @@ static int amdgpu_ttm_fw_reserve_vram_init(struct amdgpu_device *adev) if (adev->fw_vram_usage.size > 0 && adev->fw_vram_usage.size <= vram_size) {
r = amdgpu_bo_create(adev, adev->fw_vram_usage.size,PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_VRAM,AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, NULL, NULL,&adev->fw_vram_usage.reserved_bo);
r = amdgpu_bo_create(adev, adev->fw_vram_usage.size, PAGE_SIZE,AMDGPU_GEM_DOMAIN_VRAM,AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS,ttm_bo_type_kernel, NULL,&adev->fw_vram_usage.reserved_bo);diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 0b237e027cab..24474294c92a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -413,9 +413,9 @@ static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev, if (!entry->base.bo) { r = amdgpu_bo_create(adev, amdgpu_vm_bo_size(adev, level),
AMDGPU_GPU_PAGE_SIZE, true,
AMDGPU_GPU_PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, flags,
NULL, resv, &pt);
ttm_bo_type_kernel, resv, &pt); if (r) return r;@@ -2409,8 +2409,8 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, AMDGPU_GEM_CREATE_SHADOW);
size = amdgpu_vm_bo_size(adev, adev->vm_manager.root_level);
- r = amdgpu_bo_create(adev, size, align, true, AMDGPU_GEM_DOMAIN_VRAM,
flags, NULL, NULL, &vm->root.base.bo);
- r = amdgpu_bo_create(adev, size, align, AMDGPU_GEM_DOMAIN_VRAM, flags,
ttm_bo_type_kernel, NULL, &vm->root.base.bo);
Am 06.03.2018 um 22:29 schrieb Felix Kuehling:
NAK.
For KFD we need the ability to create a BO from an SG list that doesn't come from another BO. We use this for mapping pages from the doorbell aperture into GPUVM for GPU self-dispatch.
You can still do this, see amdgpu_gem_prime_import_sg_table.
Just set the BO type to SG and then setting the SG table after creating it:
ret = amdgpu_bo_create(adev, attach->dmabuf->size, PAGE_SIZE, AMDGPU_GEM_DOMAIN_CPU, 0, ttm_bo_type_sg, resv, &bo); if (ret) goto error;
bo->tbo.sg = sg; bo->tbo.ttm->sg = sg;
Then validate the result into the GTT domain to actually use the SG table.
BTW: How do you create an SG table for the doorbell?
Regards, Christian.
If you remove this now, I'll need to add it back in some form in a month or two when I get to that part of upstreaming KFD.
There may be other ways to implement this. Currently we need to create a BO for anything we want to map into a GPUVM page table, because the amdgpu_vm code is based around BOs. If there was a way to map physical addresses into GPUVM without creating a buffer object, that would work too.
Regards, Felix
On 2018-03-06 09:43 AM, Christian König wrote:
Drop the "kernel" and sg parameter and give the BO type to create explicit to amdgpu_bo_create instead of figuring it out from the parameters.
Signed-off-by: Christian König christian.koenig@amd.com
drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 5 +-- drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c | 8 ++--- drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c | 7 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 6 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 46 +++++++++++---------------- drivers/gpu/drm/amd/amdgpu/amdgpu_object.h | 11 +++---- drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c | 7 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_test.c | 11 +++---- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 11 ++++--- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 8 ++--- 11 files changed, 58 insertions(+), 64 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 292c7e72820c..b1116b773516 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -441,7 +441,7 @@ struct amdgpu_sa_bo { void amdgpu_gem_force_release(struct amdgpu_device *adev); int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size, int alignment, u32 initial_domain,
u64 flags, bool kernel,
u64 flags, enum ttm_bo_type type, struct reservation_object *resv, struct drm_gem_object **obj);diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c index 450426dbed92..7f096ed6e83d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c @@ -215,8 +215,9 @@ int alloc_gtt_mem(struct kgd_dev *kgd, size_t size, if ((*mem) == NULL) return -ENOMEM;
- r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_GTT,
AMDGPU_GEM_CREATE_CPU_GTT_USWC, NULL, NULL, &(*mem)->bo);
- r = amdgpu_bo_create(adev, size, PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT,
AMDGPU_GEM_CREATE_CPU_GTT_USWC, ttm_bo_type_kernel,NULL, &(*mem)->bo);diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c index 2fb299afc12b..02b849be083b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c @@ -80,8 +80,8 @@ static void amdgpu_benchmark_move(struct amdgpu_device *adev, unsigned size, int time;
n = AMDGPU_BENCHMARK_ITERATIONS;
- r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, sdomain, 0, NULL,
NULL, &sobj);
- r = amdgpu_bo_create(adev, size, PAGE_SIZE,sdomain, 0,
ttm_bo_type_kernel, NULL, &sobj);@@ -93,8 +93,8 @@ static void amdgpu_benchmark_move(struct amdgpu_device *adev, unsigned size, if (r) { goto out_cleanup; }
- r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, ddomain, 0, NULL,
NULL, &dobj);
- r = amdgpu_bo_create(adev, size, PAGE_SIZE, ddomain, 0,
ttm_bo_type_kernel, NULL, &dobj);diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c index dc8d9f3216fa..cf0f186c6092 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c @@ -113,11 +113,12 @@ int amdgpu_gart_table_vram_alloc(struct amdgpu_device *adev) int r;
if (adev->gart.robj == NULL) {
r = amdgpu_bo_create(adev, adev->gart.table_size,PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_VRAM,
r = amdgpu_bo_create(adev, adev->gart.table_size, PAGE_SIZE,AMDGPU_GEM_DOMAIN_VRAM, AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS,
NULL, NULL, &adev->gart.robj);
ttm_bo_type_kernel, NULL,&adev->gart.robj);diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index 55a840ae6d68..bb9b21266b67 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -45,7 +45,7 @@ void amdgpu_gem_object_free(struct drm_gem_object *gobj)
int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size, int alignment, u32 initial_domain,
u64 flags, bool kernel,
u64 flags, enum ttm_bo_type type, struct reservation_object *resv, struct drm_gem_object **obj)@@ -59,8 +59,8 @@ int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size, }
retry:
- r = amdgpu_bo_create(adev, size, alignment, kernel, initial_domain,
flags, NULL, resv, &bo);
- r = amdgpu_bo_create(adev, size, alignment, initial_domain,
flags, type, resv, &bo);diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 216799ccb545..8f5027e8241a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -187,10 +187,10 @@ int amdgpu_bo_create_reserved(struct amdgpu_device *adev, int r;
if (!*bo_ptr) {
r = amdgpu_bo_create(adev, size, align, true, domain,
r = amdgpu_bo_create(adev, size, align, domain, AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS,
NULL, NULL, bo_ptr);
ttm_bo_type_kernel, NULL, bo_ptr);@@ -331,21 +331,19 @@ static bool amdgpu_bo_validate_size(struct amdgpu_device *adev, return false; }
-static int amdgpu_bo_do_create(struct amdgpu_device *adev,
unsigned long size, int byte_align,bool kernel, u32 domain, u64 flags,struct sg_table *sg,+static int amdgpu_bo_do_create(struct amdgpu_device *adev, unsigned long size,
int byte_align, u32 domain,u64 flags, enum ttm_bo_type type, struct reservation_object *resv, struct amdgpu_bo **bo_ptr)
.interruptible = !kernel,
.interruptible = (type != ttm_bo_type_kernel),
- enum ttm_bo_type type; unsigned long page_align; size_t acc_size; int r;
@@ -356,13 +354,6 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, if (!amdgpu_bo_validate_size(adev, size, domain)) return -ENOMEM;
if (kernel) {
type = ttm_bo_type_kernel;} else if (sg) {
type = ttm_bo_type_sg;} else {
type = ttm_bo_type_device;} *bo_ptr = NULL;
acc_size = ttm_bo_dma_acc_size(&adev->mman.bdev, size,
@@ -381,7 +372,8 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, AMDGPU_GEM_DOMAIN_GWS | AMDGPU_GEM_DOMAIN_OA); bo->allowed_domains = bo->preferred_domains;
- if (!kernel && bo->allowed_domains == AMDGPU_GEM_DOMAIN_VRAM)
if (type != ttm_bo_type_kernel &&
bo->allowed_domains == AMDGPU_GEM_DOMAIN_VRAM)bo->allowed_domains |= AMDGPU_GEM_DOMAIN_GTT;
bo->flags = flags;
@@ -419,7 +411,7 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev,
r = ttm_bo_init_reserved(&adev->mman.bdev, &bo->tbo, size, type, &bo->placement, page_align, &ctx, acc_size,
sg, resv, &amdgpu_ttm_bo_destroy);
NULL, resv, &amdgpu_ttm_bo_destroy);@@ -431,7 +423,7 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, else amdgpu_cs_report_moved_bytes(adev, ctx.bytes_moved, 0);
- if (kernel)
if (type == ttm_bo_type_kernel) bo->tbo.priority = 1;
if (flags & AMDGPU_GEM_CREATE_VRAM_CLEARED &&
@@ -475,12 +467,11 @@ static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, if (bo->shadow) return 0;
- r = amdgpu_bo_do_create(adev, size, byte_align, true,
AMDGPU_GEM_DOMAIN_GTT,
- r = amdgpu_bo_do_create(adev, size, byte_align, AMDGPU_GEM_DOMAIN_GTT, AMDGPU_GEM_CREATE_CPU_GTT_USWC | AMDGPU_GEM_CREATE_SHADOW,
NULL, bo->tbo.resv,&bo->shadow);
ttm_bo_type_kernel,bo->tbo.resv, &bo->shadow);@@ -491,18 +482,17 @@ static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, return r; }
-int amdgpu_bo_create(struct amdgpu_device *adev,
unsigned long size, int byte_align,bool kernel, u32 domain, u64 flags,struct sg_table *sg,+int amdgpu_bo_create(struct amdgpu_device *adev, unsigned long size,
int byte_align, u32 domain,u64 flags, enum ttm_bo_type type, struct reservation_object *resv, struct amdgpu_bo **bo_ptr)
- r = amdgpu_bo_do_create(adev, size, byte_align, kernel, domain,
parent_flags, sg, resv, bo_ptr);
- r = amdgpu_bo_do_create(adev, size, byte_align, domain,
parent_flags, type, resv, bo_ptr);diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h index 1cef944ef98d..deb10ecd89a1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h @@ -201,12 +201,11 @@ static inline bool amdgpu_bo_explicit_sync(struct amdgpu_bo *bo) return bo->flags & AMDGPU_GEM_CREATE_EXPLICIT_SYNC; }
-int amdgpu_bo_create(struct amdgpu_device *adev,
unsigned long size, int byte_align,bool kernel, u32 domain, u64 flags,struct sg_table *sg,struct reservation_object *resv,struct amdgpu_bo **bo_ptr);+int amdgpu_bo_create(struct amdgpu_device *adev, unsigned long size,
int byte_align, u32 domain,u64 flags, enum ttm_bo_type type,struct reservation_object *resv,struct amdgpu_bo **bo_ptr);diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c index 44230165e181..1c9991738477 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c @@ -105,11 +105,14 @@ amdgpu_gem_prime_import_sg_table(struct drm_device *dev, int ret;
ww_mutex_lock(&resv->lock, NULL);
- ret = amdgpu_bo_create(adev, attach->dmabuf->size, PAGE_SIZE, false,
AMDGPU_GEM_DOMAIN_CPU, 0, sg, resv, &bo);
ret = amdgpu_bo_create(adev, attach->dmabuf->size, PAGE_SIZE,
AMDGPU_GEM_DOMAIN_CPU, 0, ttm_bo_type_sg,resv, &bo);if (ret) goto error;
bo->tbo.sg = sg;
bo->tbo.ttm->sg = sg; bo->allowed_domains = AMDGPU_GEM_DOMAIN_GTT; bo->preferred_domains = AMDGPU_GEM_DOMAIN_GTT; if (attach->dmabuf->ops != &amdgpu_dmabuf_ops)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c index f3d81b6fb499..2dbe87591f81 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c @@ -59,9 +59,8 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) goto out_cleanup; }
- r = amdgpu_bo_create(adev, size, PAGE_SIZE, true,
AMDGPU_GEM_DOMAIN_VRAM, 0,NULL, NULL, &vram_obj);
- r = amdgpu_bo_create(adev, size, PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, 0,
ttm_bo_type_kernel, NULL, &vram_obj);@@ -80,9 +79,9 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) void **vram_start, **vram_end; struct dma_fence *fence = NULL;
r = amdgpu_bo_create(adev, size, PAGE_SIZE, true,AMDGPU_GEM_DOMAIN_GTT, 0, NULL,NULL, gtt_obj + i);
r = amdgpu_bo_create(adev, size, PAGE_SIZE,AMDGPU_GEM_DOMAIN_GTT, 0,ttm_bo_type_kernel, NULL, gtt_obj + i);diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index eebc9f77b611..1ec54629f277 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -1317,11 +1317,12 @@ static int amdgpu_ttm_fw_reserve_vram_init(struct amdgpu_device *adev) if (adev->fw_vram_usage.size > 0 && adev->fw_vram_usage.size <= vram_size) {
r = amdgpu_bo_create(adev, adev->fw_vram_usage.size,PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_VRAM,AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, NULL, NULL,&adev->fw_vram_usage.reserved_bo);
r = amdgpu_bo_create(adev, adev->fw_vram_usage.size, PAGE_SIZE,AMDGPU_GEM_DOMAIN_VRAM,AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED |AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS,ttm_bo_type_kernel, NULL,&adev->fw_vram_usage.reserved_bo);diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 0b237e027cab..24474294c92a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -413,9 +413,9 @@ static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev, if (!entry->base.bo) { r = amdgpu_bo_create(adev, amdgpu_vm_bo_size(adev, level),
AMDGPU_GPU_PAGE_SIZE, true,
AMDGPU_GPU_PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, flags,
NULL, resv, &pt);
ttm_bo_type_kernel, resv, &pt); if (r) return r;@@ -2409,8 +2409,8 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, AMDGPU_GEM_CREATE_SHADOW);
size = amdgpu_vm_bo_size(adev, adev->vm_manager.root_level);
- r = amdgpu_bo_create(adev, size, align, true, AMDGPU_GEM_DOMAIN_VRAM,
flags, NULL, NULL, &vm->root.base.bo);
- r = amdgpu_bo_create(adev, size, align, AMDGPU_GEM_DOMAIN_VRAM, flags,
ttm_bo_type_kernel, NULL, &vm->root.base.bo);
On 2018-03-07 04:16 AM, Christian König wrote:
Am 06.03.2018 um 22:29 schrieb Felix Kuehling:
NAK.
For KFD we need the ability to create a BO from an SG list that doesn't come from another BO. We use this for mapping pages from the doorbell aperture into GPUVM for GPU self-dispatch.
You can still do this, see amdgpu_gem_prime_import_sg_table.
Just set the BO type to SG and then setting the SG table after creating it:
ret = amdgpu_bo_create(adev, attach->dmabuf->size, PAGE_SIZE, AMDGPU_GEM_DOMAIN_CPU, 0, ttm_bo_type_sg, resv, &bo); if (ret) goto error; bo->tbo.sg = sg; bo->tbo.ttm->sg = sg;
Then validate the result into the GTT domain to actually use the SG table.
OK, I'll try implementing that before we merge your change.
BTW: How do you create an SG table for the doorbell?
KFD manages the doorbell aperture and assigns one or two pages of doorbells to each process. KFD knows the bus address of the doorbells and passes it to amdgpu, which creates the SG like this:
static struct sg_table *create_doorbell_sg(uint64_t addr, uint32_t size) { struct sg_table *sg = kmalloc(sizeof(*sg), GFP_KERNEL);
if (!sg) return NULL; if (sg_alloc_table(sg, 1, GFP_KERNEL)) { kfree(sg); return NULL; } sg->sgl->dma_address = addr; sg->sgl->length = size; #ifdef CONFIG_NEED_SG_DMA_LENGTH sg->sgl->dma_length = size; #endif return sg; }
Regards, Felix
Regards, Christian.
If you remove this now, I'll need to add it back in some form in a month or two when I get to that part of upstreaming KFD.
There may be other ways to implement this. Currently we need to create a BO for anything we want to map into a GPUVM page table, because the amdgpu_vm code is based around BOs. If there was a way to map physical addresses into GPUVM without creating a buffer object, that would work too.
Regards, Felix
On 2018-03-06 09:43 AM, Christian König wrote:
Drop the "kernel" and sg parameter and give the BO type to create explicit to amdgpu_bo_create instead of figuring it out from the parameters.
Signed-off-by: Christian König christian.koenig@amd.com
drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 5 +-- drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c | 8 ++--- drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c | 7 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 6 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 46 +++++++++++---------------- drivers/gpu/drm/amd/amdgpu/amdgpu_object.h | 11 +++---- drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c | 7 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_test.c | 11 +++---- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 11 ++++--- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 8 ++--- 11 files changed, 58 insertions(+), 64 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 292c7e72820c..b1116b773516 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -441,7 +441,7 @@ struct amdgpu_sa_bo { void amdgpu_gem_force_release(struct amdgpu_device *adev); int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size, int alignment, u32 initial_domain, - u64 flags, bool kernel, + u64 flags, enum ttm_bo_type type, struct reservation_object *resv, struct drm_gem_object **obj); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c index 450426dbed92..7f096ed6e83d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c @@ -215,8 +215,9 @@ int alloc_gtt_mem(struct kgd_dev *kgd, size_t size, if ((*mem) == NULL) return -ENOMEM; - r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_GTT, - AMDGPU_GEM_CREATE_CPU_GTT_USWC, NULL, NULL, &(*mem)->bo); + r = amdgpu_bo_create(adev, size, PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT, + AMDGPU_GEM_CREATE_CPU_GTT_USWC, ttm_bo_type_kernel, + NULL, &(*mem)->bo); if (r) { dev_err(adev->dev, "failed to allocate BO for amdkfd (%d)\n", r); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c index 2fb299afc12b..02b849be083b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c @@ -80,8 +80,8 @@ static void amdgpu_benchmark_move(struct amdgpu_device *adev, unsigned size, int time; n = AMDGPU_BENCHMARK_ITERATIONS; - r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, sdomain, 0, NULL, - NULL, &sobj); + r = amdgpu_bo_create(adev, size, PAGE_SIZE,sdomain, 0, + ttm_bo_type_kernel, NULL, &sobj); if (r) { goto out_cleanup; } @@ -93,8 +93,8 @@ static void amdgpu_benchmark_move(struct amdgpu_device *adev, unsigned size, if (r) { goto out_cleanup; } - r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, ddomain, 0, NULL, - NULL, &dobj); + r = amdgpu_bo_create(adev, size, PAGE_SIZE, ddomain, 0, + ttm_bo_type_kernel, NULL, &dobj); if (r) { goto out_cleanup; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c index dc8d9f3216fa..cf0f186c6092 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c @@ -113,11 +113,12 @@ int amdgpu_gart_table_vram_alloc(struct amdgpu_device *adev) int r; if (adev->gart.robj == NULL) { - r = amdgpu_bo_create(adev, adev->gart.table_size, - PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_VRAM, + r = amdgpu_bo_create(adev, adev->gart.table_size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, - NULL, NULL, &adev->gart.robj); + ttm_bo_type_kernel, NULL, + &adev->gart.robj); if (r) { return r; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index 55a840ae6d68..bb9b21266b67 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -45,7 +45,7 @@ void amdgpu_gem_object_free(struct drm_gem_object *gobj) int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size, int alignment, u32 initial_domain, - u64 flags, bool kernel, + u64 flags, enum ttm_bo_type type, struct reservation_object *resv, struct drm_gem_object **obj) { @@ -59,8 +59,8 @@ int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size, } retry: - r = amdgpu_bo_create(adev, size, alignment, kernel, initial_domain, - flags, NULL, resv, &bo); + r = amdgpu_bo_create(adev, size, alignment, initial_domain, + flags, type, resv, &bo); if (r) { if (r != -ERESTARTSYS) { if (flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 216799ccb545..8f5027e8241a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -187,10 +187,10 @@ int amdgpu_bo_create_reserved(struct amdgpu_device *adev, int r; if (!*bo_ptr) { - r = amdgpu_bo_create(adev, size, align, true, domain, + r = amdgpu_bo_create(adev, size, align, domain, AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, - NULL, NULL, bo_ptr); + ttm_bo_type_kernel, NULL, bo_ptr); if (r) { dev_err(adev->dev, "(%d) failed to allocate kernel bo\n", r); @@ -331,21 +331,19 @@ static bool amdgpu_bo_validate_size(struct amdgpu_device *adev, return false; } -static int amdgpu_bo_do_create(struct amdgpu_device *adev, - unsigned long size, int byte_align, - bool kernel, u32 domain, u64 flags, - struct sg_table *sg, +static int amdgpu_bo_do_create(struct amdgpu_device *adev, unsigned long size, + int byte_align, u32 domain, + u64 flags, enum ttm_bo_type type, struct reservation_object *resv, struct amdgpu_bo **bo_ptr) { struct ttm_operation_ctx ctx = { - .interruptible = !kernel, + .interruptible = (type != ttm_bo_type_kernel), .no_wait_gpu = false, .resv = resv, .flags = TTM_OPT_FLAG_ALLOW_RES_EVICT }; struct amdgpu_bo *bo; - enum ttm_bo_type type; unsigned long page_align; size_t acc_size; int r; @@ -356,13 +354,6 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, if (!amdgpu_bo_validate_size(adev, size, domain)) return -ENOMEM; - if (kernel) { - type = ttm_bo_type_kernel; - } else if (sg) { - type = ttm_bo_type_sg; - } else { - type = ttm_bo_type_device; - } *bo_ptr = NULL; acc_size = ttm_bo_dma_acc_size(&adev->mman.bdev, size, @@ -381,7 +372,8 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, AMDGPU_GEM_DOMAIN_GWS | AMDGPU_GEM_DOMAIN_OA); bo->allowed_domains = bo->preferred_domains; - if (!kernel && bo->allowed_domains == AMDGPU_GEM_DOMAIN_VRAM) + if (type != ttm_bo_type_kernel && + bo->allowed_domains == AMDGPU_GEM_DOMAIN_VRAM) bo->allowed_domains |= AMDGPU_GEM_DOMAIN_GTT; bo->flags = flags; @@ -419,7 +411,7 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, r = ttm_bo_init_reserved(&adev->mman.bdev, &bo->tbo, size, type, &bo->placement, page_align, &ctx, acc_size, - sg, resv, &amdgpu_ttm_bo_destroy); + NULL, resv, &amdgpu_ttm_bo_destroy); if (unlikely(r != 0)) return r; @@ -431,7 +423,7 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, else amdgpu_cs_report_moved_bytes(adev, ctx.bytes_moved, 0); - if (kernel) + if (type == ttm_bo_type_kernel) bo->tbo.priority = 1; if (flags & AMDGPU_GEM_CREATE_VRAM_CLEARED && @@ -475,12 +467,11 @@ static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, if (bo->shadow) return 0; - r = amdgpu_bo_do_create(adev, size, byte_align, true, - AMDGPU_GEM_DOMAIN_GTT, + r = amdgpu_bo_do_create(adev, size, byte_align, AMDGPU_GEM_DOMAIN_GTT, AMDGPU_GEM_CREATE_CPU_GTT_USWC | AMDGPU_GEM_CREATE_SHADOW, - NULL, bo->tbo.resv, - &bo->shadow); + ttm_bo_type_kernel, + bo->tbo.resv, &bo->shadow); if (!r) { bo->shadow->parent = amdgpu_bo_ref(bo); mutex_lock(&adev->shadow_list_lock); @@ -491,18 +482,17 @@ static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, return r; } -int amdgpu_bo_create(struct amdgpu_device *adev, - unsigned long size, int byte_align, - bool kernel, u32 domain, u64 flags, - struct sg_table *sg, +int amdgpu_bo_create(struct amdgpu_device *adev, unsigned long size, + int byte_align, u32 domain, + u64 flags, enum ttm_bo_type type, struct reservation_object *resv, struct amdgpu_bo **bo_ptr) { uint64_t parent_flags = flags & ~AMDGPU_GEM_CREATE_SHADOW; int r; - r = amdgpu_bo_do_create(adev, size, byte_align, kernel, domain, - parent_flags, sg, resv, bo_ptr); + r = amdgpu_bo_do_create(adev, size, byte_align, domain, + parent_flags, type, resv, bo_ptr); if (r) return r; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h index 1cef944ef98d..deb10ecd89a1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h @@ -201,12 +201,11 @@ static inline bool amdgpu_bo_explicit_sync(struct amdgpu_bo *bo) return bo->flags & AMDGPU_GEM_CREATE_EXPLICIT_SYNC; } -int amdgpu_bo_create(struct amdgpu_device *adev, - unsigned long size, int byte_align, - bool kernel, u32 domain, u64 flags, - struct sg_table *sg, - struct reservation_object *resv, - struct amdgpu_bo **bo_ptr); +int amdgpu_bo_create(struct amdgpu_device *adev, unsigned long size, + int byte_align, u32 domain, + u64 flags, enum ttm_bo_type type, + struct reservation_object *resv, + struct amdgpu_bo **bo_ptr); int amdgpu_bo_create_reserved(struct amdgpu_device *adev, unsigned long size, int align, u32 domain, struct amdgpu_bo **bo_ptr, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c index 44230165e181..1c9991738477 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c @@ -105,11 +105,14 @@ amdgpu_gem_prime_import_sg_table(struct drm_device *dev, int ret; ww_mutex_lock(&resv->lock, NULL); - ret = amdgpu_bo_create(adev, attach->dmabuf->size, PAGE_SIZE, false, - AMDGPU_GEM_DOMAIN_CPU, 0, sg, resv, &bo); + ret = amdgpu_bo_create(adev, attach->dmabuf->size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_CPU, 0, ttm_bo_type_sg, + resv, &bo); if (ret) goto error; + bo->tbo.sg = sg; + bo->tbo.ttm->sg = sg; bo->allowed_domains = AMDGPU_GEM_DOMAIN_GTT; bo->preferred_domains = AMDGPU_GEM_DOMAIN_GTT; if (attach->dmabuf->ops != &amdgpu_dmabuf_ops) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c index f3d81b6fb499..2dbe87591f81 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c @@ -59,9 +59,8 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) goto out_cleanup; } - r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_VRAM, 0, - NULL, NULL, &vram_obj); + r = amdgpu_bo_create(adev, size, PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, 0, + ttm_bo_type_kernel, NULL, &vram_obj); if (r) { DRM_ERROR("Failed to create VRAM object\n"); goto out_cleanup; @@ -80,9 +79,9 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) void **vram_start, **vram_end; struct dma_fence *fence = NULL; - r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, - AMDGPU_GEM_DOMAIN_GTT, 0, NULL, - NULL, gtt_obj + i); + r = amdgpu_bo_create(adev, size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_GTT, 0, + ttm_bo_type_kernel, NULL, gtt_obj + i); if (r) { DRM_ERROR("Failed to create GTT object %d\n", i); goto out_lclean; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index eebc9f77b611..1ec54629f277 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -1317,11 +1317,12 @@ static int amdgpu_ttm_fw_reserve_vram_init(struct amdgpu_device *adev) if (adev->fw_vram_usage.size > 0 && adev->fw_vram_usage.size <= vram_size) { - r = amdgpu_bo_create(adev, adev->fw_vram_usage.size, - PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | - AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, NULL, NULL, - &adev->fw_vram_usage.reserved_bo); + r = amdgpu_bo_create(adev, adev->fw_vram_usage.size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | + AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, + ttm_bo_type_kernel, NULL, + &adev->fw_vram_usage.reserved_bo); if (r) goto error_create; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 0b237e027cab..24474294c92a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -413,9 +413,9 @@ static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev, if (!entry->base.bo) { r = amdgpu_bo_create(adev, amdgpu_vm_bo_size(adev, level), - AMDGPU_GPU_PAGE_SIZE, true, + AMDGPU_GPU_PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, flags, - NULL, resv, &pt); + ttm_bo_type_kernel, resv, &pt); if (r) return r; @@ -2409,8 +2409,8 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, AMDGPU_GEM_CREATE_SHADOW); size = amdgpu_vm_bo_size(adev, adev->vm_manager.root_level); - r = amdgpu_bo_create(adev, size, align, true, AMDGPU_GEM_DOMAIN_VRAM, - flags, NULL, NULL, &vm->root.base.bo); + r = amdgpu_bo_create(adev, size, align, AMDGPU_GEM_DOMAIN_VRAM, flags, + ttm_bo_type_kernel, NULL, &vm->root.base.bo); if (r) goto error_free_sched_entity;
Patch 1,4,5: Acked-by: Roger He Hongbo.He@amd.com Patch 2,3,6: Reviewed-by: Roger He Hongbo.He@amd.com
-----Original Message----- From: dri-devel [mailto:dri-devel-bounces@lists.freedesktop.org] On Behalf Of Christian K?nig Sent: Tuesday, March 06, 2018 10:44 PM To: amd-gfx@lists.freedesktop.org; dri-devel@lists.freedesktop.org Subject: [PATCH 1/6] drm/ttm: add ttm_bo_pipeline_gutting
Allows us to gut a BO of it's backing store when the driver says that it isn't needed any more.
Signed-off-by: Christian König christian.koenig@amd.com --- drivers/gpu/drm/ttm/ttm_bo.c | 15 ++++++++++++--- drivers/gpu/drm/ttm/ttm_bo_util.c | 24 ++++++++++++++++++++++++ include/drm/ttm/ttm_bo_driver.h | 9 +++++++++ 3 files changed, 45 insertions(+), 3 deletions(-)
diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c index ad142a92eb80..98e06f8bf23b 100644 --- a/drivers/gpu/drm/ttm/ttm_bo.c +++ b/drivers/gpu/drm/ttm/ttm_bo.c @@ -622,14 +622,23 @@ static int ttm_bo_evict(struct ttm_buffer_object *bo,
reservation_object_assert_held(bo->resv);
+ placement.num_placement = 0; + placement.num_busy_placement = 0; + bdev->driver->evict_flags(bo, &placement); + + if (!placement.num_placement && !placement.num_busy_placement) { + ret = ttm_bo_pipeline_gutting(bo); + if (ret) + return ret; + + return ttm_tt_create(bo, false); + } + evict_mem = bo->mem; evict_mem.mm_node = NULL; evict_mem.bus.io_reserved_vm = false; evict_mem.bus.io_reserved_count = 0;
- placement.num_placement = 0; - placement.num_busy_placement = 0; - bdev->driver->evict_flags(bo, &placement); ret = ttm_bo_mem_space(bo, &placement, &evict_mem, ctx); if (ret) { if (ret != -ERESTARTSYS) { diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c index 6d6a3f46143b..1f730b3f18e5 100644 --- a/drivers/gpu/drm/ttm/ttm_bo_util.c +++ b/drivers/gpu/drm/ttm/ttm_bo_util.c @@ -801,3 +801,27 @@ int ttm_bo_pipeline_move(struct ttm_buffer_object *bo, return 0; } EXPORT_SYMBOL(ttm_bo_pipeline_move); + +int ttm_bo_pipeline_gutting(struct ttm_buffer_object *bo) { + struct ttm_buffer_object *ghost; + int ret; + + ret = ttm_buffer_object_transfer(bo, &ghost); + if (ret) + return ret; + + ret = reservation_object_copy_fences(ghost->resv, bo->resv); + /* Last resort, wait for the BO to be idle when we are OOM */ + if (ret) + ttm_bo_wait(bo, false, false); + + memset(&bo->mem, 0, sizeof(bo->mem)); + bo->mem.mem_type = TTM_PL_SYSTEM; + bo->ttm = NULL; + + ttm_bo_unreserve(ghost); + ttm_bo_unref(&ghost); + + return 0; +} diff --git a/include/drm/ttm/ttm_bo_driver.h b/include/drm/ttm/ttm_bo_driver.h index f8e2515b401f..39cd6b086d3a 100644 --- a/include/drm/ttm/ttm_bo_driver.h +++ b/include/drm/ttm/ttm_bo_driver.h @@ -849,6 +849,15 @@ int ttm_bo_pipeline_move(struct ttm_buffer_object *bo, struct dma_fence *fence, bool evict, struct ttm_mem_reg *new_mem);
+/** + * ttm_bo_pipeline_gutting. + * + * @bo: A pointer to a struct ttm_buffer_object. + * + * Pipelined gutting a BO of it's backing store. + */ +int ttm_bo_pipeline_gutting(struct ttm_buffer_object *bo); + /** * ttm_io_prot * -- 2.14.1
_______________________________________________ dri-devel mailing list dri-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/dri-devel
dri-devel@lists.freedesktop.org
-
Christian König -
Felix Kuehling -
He, Roger