[RFC PATCH] drm/amdgpu: move ttm bo->offset to amdgpu_bo

Christian König christian.koenig at amd.com
Tue Feb 11 11:50:18 UTC 2020


Am 11.02.20 um 12:17 schrieb Nirmoy Das:
> GPU address should belong to driver not in memory management.
> This patch moves ttm bo.offset and gpu_offset calculation to amdgpu driver.

Actually for amdgpu we should be able to get completely rid of 
bo->offset. All we need to do is to change amdgpu_bo_gpu_offset() and do 
the correct offset calculation there.

Apart from that looks good to me.

Regards,
Christian.

>
> Signed-off-by: Nirmoy Das <nirmoy.das at amd.com>
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c    |  2 +-
>   drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c |  4 +-
>   drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c        |  4 +-
>   drivers/gpu/drm/amd/amdgpu/amdgpu_display.c   |  2 +-
>   drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c        |  2 +-
>   drivers/gpu/drm/amd/amdgpu/amdgpu_object.c    | 22 ++++++++--
>   drivers/gpu/drm/amd/amdgpu/amdgpu_object.h    |  3 ++
>   drivers/gpu/drm/amd/amdgpu/amdgpu_test.c      |  2 +-
>   drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c       | 40 +++++++++----------
>   drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h       |  2 +-
>   drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c   |  8 ++--
>   .../gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c |  2 +-
>   12 files changed, 55 insertions(+), 38 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
> index 8609287620ea..613047ea9065 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
> @@ -263,7 +263,7 @@ int amdgpu_amdkfd_alloc_gtt_mem(struct kgd_dev *kgd, size_t size,
>   		goto allocate_mem_pin_bo_failed;
>   	}
>   
> -	r = amdgpu_ttm_alloc_gart(&bo->tbo);
> +	r = amdgpu_ttm_alloc_gart(bo);
>   	if (r) {
>   		dev_err(adev->dev, "%p bind failed\n", bo);
>   		goto allocate_mem_kmap_bo_failed;
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c
> index d9b35df33806..f29986231b27 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c
> @@ -98,7 +98,7 @@ static void amdgpu_benchmark_move(struct amdgpu_device *adev, unsigned size,
>   		amdgpu_bo_unreserve(sobj);
>   		goto out_cleanup;
>   	}
> -	r = amdgpu_ttm_alloc_gart(&sobj->tbo);
> +	r = amdgpu_ttm_alloc_gart(sobj);
>   	amdgpu_bo_unreserve(sobj);
>   	if (r) {
>   		goto out_cleanup;
> @@ -117,7 +117,7 @@ static void amdgpu_benchmark_move(struct amdgpu_device *adev, unsigned size,
>   		amdgpu_bo_unreserve(sobj);
>   		goto out_cleanup;
>   	}
> -	r = amdgpu_ttm_alloc_gart(&dobj->tbo);
> +	r = amdgpu_ttm_alloc_gart(dobj);
>   	amdgpu_bo_unreserve(dobj);
>   	if (r) {
>   		goto out_cleanup;
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
> index 80ba6dfc54e2..921719d92ab1 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
> @@ -642,7 +642,7 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p,
>   	if (!r && p->uf_entry.tv.bo) {
>   		struct amdgpu_bo *uf = ttm_to_amdgpu_bo(p->uf_entry.tv.bo);
>   
> -		r = amdgpu_ttm_alloc_gart(&uf->tbo);
> +		r = amdgpu_ttm_alloc_gart(uf);
>   		p->job->uf_addr += amdgpu_bo_gpu_offset(uf);
>   	}
>   
> @@ -1676,5 +1676,5 @@ int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser,
>   			return r;
>   	}
>   
> -	return amdgpu_ttm_alloc_gart(&(*bo)->tbo);
> +	return amdgpu_ttm_alloc_gart(*bo);
>   }
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c
> index ca9056b0472a..3d0aa2b2926f 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c
> @@ -199,7 +199,7 @@ int amdgpu_display_crtc_page_flip_target(struct drm_crtc *crtc,
>   		}
>   	}
>   
> -	r = amdgpu_ttm_alloc_gart(&new_abo->tbo);
> +	r = amdgpu_ttm_alloc_gart(new_abo);
>   	if (unlikely(r != 0)) {
>   		DRM_ERROR("%p bind failed\n", new_abo);
>   		goto unpin;
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c
> index 143753d237e7..a02b1cbd006d 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c
> @@ -174,7 +174,7 @@ static int amdgpufb_create_pinned_object(struct amdgpu_fbdev *rfbdev,
>   		goto out_unref;
>   	}
>   
> -	ret = amdgpu_ttm_alloc_gart(&abo->tbo);
> +	ret = amdgpu_ttm_alloc_gart(abo);
>   	if (ret) {
>   		amdgpu_bo_unreserve(abo);
>   		dev_err(adev->dev, "%p bind failed\n", abo);
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> index 6f60a581e3ba..650758a5b3aa 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
> @@ -274,7 +274,7 @@ int amdgpu_bo_create_reserved(struct amdgpu_device *adev,
>   		goto error_unreserve;
>   	}
>   
> -	r = amdgpu_ttm_alloc_gart(&(*bo_ptr)->tbo);
> +	r = amdgpu_ttm_alloc_gart(*bo_ptr);
>   	if (r) {
>   		dev_err(adev->dev, "%p bind failed\n", *bo_ptr);
>   		goto error_unpin;
> @@ -917,7 +917,7 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain,
>   		bo->pin_count++;
>   
>   		if (max_offset != 0) {
> -			u64 domain_start = bo->tbo.bdev->man[mem_type].gpu_offset;
> +			u64 domain_start = amdgpu_bo_gpu_start(adev, mem_type);
>   			WARN_ON_ONCE(max_offset <
>   				     (amdgpu_bo_gpu_offset(bo) - domain_start));
>   		}
> @@ -1248,6 +1248,18 @@ int amdgpu_bo_get_metadata(struct amdgpu_bo *bo, void *buffer,
>   	return 0;
>   }
>   
> +uint64_t amdgpu_bo_gpu_start(struct amdgpu_device *adev, uint32_t type)
> +{
> +	switch(type) {
> +		case TTM_PL_TT:
> +			return adev->gmc.gart_start;
> +		case TTM_PL_VRAM:
> +			return adev->gmc.vram_start;
> +	}
> +
> +	return 0;
> +}
> +
>   /**
>    * amdgpu_bo_move_notify - notification about a memory move
>    * @bo: pointer to a buffer object
> @@ -1278,6 +1290,10 @@ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo,
>   	if (evict)
>   		atomic64_inc(&adev->num_evictions);
>   
> +	if (new_mem)
> +		abo->offset = (new_mem->start << PAGE_SHIFT) +
> +			 amdgpu_bo_gpu_start(amdgpu_ttm_adev(bo->bdev), new_mem->mem_type);
> +
>   	/* update statistics */
>   	if (!new_mem)
>   		return;
> @@ -1467,7 +1483,7 @@ u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo)
>   	WARN_ON_ONCE(bo->tbo.mem.mem_type == TTM_PL_VRAM &&
>   		     !(bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS));
>   
> -	return amdgpu_gmc_sign_extend(bo->tbo.offset);
> +	return amdgpu_gmc_sign_extend(bo->offset);
>   }
>   
>   /**
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
> index 96d805889e8d..c7e1028fd31f 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
> @@ -94,6 +94,8 @@ struct amdgpu_bo {
>   	unsigned			prime_shared_count;
>   	/* per VM structure for page tables and with virtual addresses */
>   	struct amdgpu_vm_bo_base	*vm_bo;
> +	 /* GPU address space is independent of CPU word size */
> +	uint64_t			offset;
>   	/* Constant after initialization */
>   	struct amdgpu_bo		*parent;
>   	struct amdgpu_bo		*shadow;
> @@ -294,6 +296,7 @@ int amdgpu_bo_restore_shadow(struct amdgpu_bo *shadow,
>   uint32_t amdgpu_bo_get_preferred_pin_domain(struct amdgpu_device *adev,
>   					    uint32_t domain);
>   
> +uint64_t amdgpu_bo_gpu_start(struct amdgpu_device *adev, uint32_t type);
>   /*
>    * sub allocation
>    */
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c
> index 476f1f89aaad..dfab64bbb5fa 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c
> @@ -103,7 +103,7 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev)
>   			DRM_ERROR("Failed to pin GTT object %d\n", i);
>   			goto out_lclean_unres;
>   		}
> -		r = amdgpu_ttm_alloc_gart(&gtt_obj[i]->tbo);
> +		r = amdgpu_ttm_alloc_gart(gtt_obj[i]);
>   		if (r) {
>   			DRM_ERROR("%p bind failed\n", gtt_obj[i]);
>   			goto out_lclean_unpin;
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> index 2c1d1eb1a7e1..15d9e8f807b9 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
> @@ -103,7 +103,6 @@ static int amdgpu_init_mem_type(struct ttm_bo_device *bdev, uint32_t type,
>   	case TTM_PL_TT:
>   		/* GTT memory  */
>   		man->func = &amdgpu_gtt_mgr_func;
> -		man->gpu_offset = adev->gmc.gart_start;
>   		man->available_caching = TTM_PL_MASK_CACHING;
>   		man->default_caching = TTM_PL_FLAG_CACHED;
>   		man->flags = TTM_MEMTYPE_FLAG_MAPPABLE | TTM_MEMTYPE_FLAG_CMA;
> @@ -111,7 +110,6 @@ static int amdgpu_init_mem_type(struct ttm_bo_device *bdev, uint32_t type,
>   	case TTM_PL_VRAM:
>   		/* "On-card" video ram */
>   		man->func = &amdgpu_vram_mgr_func;
> -		man->gpu_offset = adev->gmc.vram_start;
>   		man->flags = TTM_MEMTYPE_FLAG_FIXED |
>   			     TTM_MEMTYPE_FLAG_MAPPABLE;
>   		man->available_caching = TTM_PL_FLAG_UNCACHED | TTM_PL_FLAG_WC;
> @@ -122,7 +120,6 @@ static int amdgpu_init_mem_type(struct ttm_bo_device *bdev, uint32_t type,
>   	case AMDGPU_PL_OA:
>   		/* On-chip GDS memory*/
>   		man->func = &ttm_bo_manager_func;
> -		man->gpu_offset = 0;
>   		man->flags = TTM_MEMTYPE_FLAG_FIXED | TTM_MEMTYPE_FLAG_CMA;
>   		man->available_caching = TTM_PL_FLAG_UNCACHED;
>   		man->default_caching = TTM_PL_FLAG_UNCACHED;
> @@ -270,7 +267,7 @@ static uint64_t amdgpu_mm_node_addr(struct ttm_buffer_object *bo,
>   
>   	if (mm_node->start != AMDGPU_BO_INVALID_OFFSET) {
>   		addr = mm_node->start << PAGE_SHIFT;
> -		addr += bo->bdev->man[mem->mem_type].gpu_offset;
> +		addr += amdgpu_bo_gpu_start(amdgpu_ttm_adev(bo->bdev), mem->mem_type);
>   	}
>   	return addr;
>   }
> @@ -1086,27 +1083,28 @@ static int amdgpu_ttm_backend_bind(struct ttm_tt *ttm,
>   /**
>    * amdgpu_ttm_alloc_gart - Allocate GART memory for buffer object
>    */
> -int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo)
> +int amdgpu_ttm_alloc_gart(struct amdgpu_bo *bo)
>   {
> -	struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev);
> +	struct ttm_buffer_object *tbo = &bo->tbo;
> +	struct amdgpu_device *adev = amdgpu_ttm_adev(tbo->bdev);
>   	struct ttm_operation_ctx ctx = { false, false };
> -	struct amdgpu_ttm_tt *gtt = (void*)bo->ttm;
> +	struct amdgpu_ttm_tt *gtt = (void*)tbo->ttm;
>   	struct ttm_mem_reg tmp;
>   	struct ttm_placement placement;
>   	struct ttm_place placements;
>   	uint64_t addr, flags;
>   	int r;
>   
> -	if (bo->mem.start != AMDGPU_BO_INVALID_OFFSET)
> +	if (tbo->mem.start != AMDGPU_BO_INVALID_OFFSET)
>   		return 0;
>   
> -	addr = amdgpu_gmc_agp_addr(bo);
> +	addr = amdgpu_gmc_agp_addr(tbo);
>   	if (addr != AMDGPU_BO_INVALID_OFFSET) {
> -		bo->mem.start = addr >> PAGE_SHIFT;
> +		tbo->mem.start = addr >> PAGE_SHIFT;
>   	} else {
>   
>   		/* allocate GART space */
> -		tmp = bo->mem;
> +		tmp = tbo->mem;
>   		tmp.mm_node = NULL;
>   		placement.num_placement = 1;
>   		placement.placement = &placements;
> @@ -1114,30 +1112,30 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo)
>   		placement.busy_placement = &placements;
>   		placements.fpfn = 0;
>   		placements.lpfn = adev->gmc.gart_size >> PAGE_SHIFT;
> -		placements.flags = (bo->mem.placement & ~TTM_PL_MASK_MEM) |
> +		placements.flags = (tbo->mem.placement & ~TTM_PL_MASK_MEM) |
>   			TTM_PL_FLAG_TT;
>   
> -		r = ttm_bo_mem_space(bo, &placement, &tmp, &ctx);
> +		r = ttm_bo_mem_space(tbo, &placement, &tmp, &ctx);
>   		if (unlikely(r))
>   			return r;
>   
>   		/* compute PTE flags for this buffer object */
> -		flags = amdgpu_ttm_tt_pte_flags(adev, bo->ttm, &tmp);
> +		flags = amdgpu_ttm_tt_pte_flags(adev, tbo->ttm, &tmp);
>   
>   		/* Bind pages */
>   		gtt->offset = (u64)tmp.start << PAGE_SHIFT;
> -		r = amdgpu_ttm_gart_bind(adev, bo, flags);
> +		r = amdgpu_ttm_gart_bind(adev, tbo, flags);
>   		if (unlikely(r)) {
> -			ttm_bo_mem_put(bo, &tmp);
> +			ttm_bo_mem_put(tbo, &tmp);
>   			return r;
>   		}
>   
> -		ttm_bo_mem_put(bo, &bo->mem);
> -		bo->mem = tmp;
> +		ttm_bo_mem_put(tbo, &tbo->mem);
> +		tbo->mem = tmp;
>   	}
>   
> -	bo->offset = (bo->mem.start << PAGE_SHIFT) +
> -		bo->bdev->man[bo->mem.mem_type].gpu_offset;
> +	bo->offset = (tbo->mem.start << PAGE_SHIFT) +
> +		amdgpu_bo_gpu_start(adev, tbo->mem.mem_type);
>   
>   	return 0;
>   }
> @@ -2149,7 +2147,7 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo,
>   	}
>   
>   	if (bo->tbo.mem.mem_type == TTM_PL_TT) {
> -		r = amdgpu_ttm_alloc_gart(&bo->tbo);
> +		r = amdgpu_ttm_alloc_gart(bo);
>   		if (r)
>   			return r;
>   	}
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h
> index f1ebd424510c..4f6e7ef2b004 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h
> @@ -100,7 +100,7 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo,
>   			struct dma_fence **fence);
>   
>   int amdgpu_mmap(struct file *filp, struct vm_area_struct *vma);
> -int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo);
> +int amdgpu_ttm_alloc_gart(struct amdgpu_bo *bo);
>   int amdgpu_ttm_recover_gart(struct ttm_buffer_object *tbo);
>   
>   #if IS_ENABLED(CONFIG_DRM_AMDGPU_USERPTR)
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c
> index 4cc7881f438c..c810cda7a06e 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm_sdma.c
> @@ -37,12 +37,12 @@ static int amdgpu_vm_sdma_map_table(struct amdgpu_bo *table)
>   {
>   	int r;
>   
> -	r = amdgpu_ttm_alloc_gart(&table->tbo);
> +	r = amdgpu_ttm_alloc_gart(table);
>   	if (r)
>   		return r;
>   
>   	if (table->shadow)
> -		r = amdgpu_ttm_alloc_gart(&table->shadow->tbo);
> +		r = amdgpu_ttm_alloc_gart(table->shadow);
>   
>   	return r;
>   }
> @@ -140,7 +140,7 @@ static void amdgpu_vm_sdma_copy_ptes(struct amdgpu_vm_update_params *p,
>   
>   	src += p->num_dw_left * 4;
>   
> -	pe += amdgpu_gmc_sign_extend(bo->tbo.offset);
> +	pe += amdgpu_gmc_sign_extend(bo->offset);
>   	trace_amdgpu_vm_copy_ptes(pe, src, count, p->direct);
>   
>   	amdgpu_vm_copy_pte(p->adev, ib, pe, src, count);
> @@ -167,7 +167,7 @@ static void amdgpu_vm_sdma_set_ptes(struct amdgpu_vm_update_params *p,
>   {
>   	struct amdgpu_ib *ib = p->job->ibs;
>   
> -	pe += amdgpu_gmc_sign_extend(bo->tbo.offset);
> +	pe += amdgpu_gmc_sign_extend(bo->offset);
>   	trace_amdgpu_vm_set_ptes(pe, addr, count, incr, flags, p->direct);
>   	if (count < 3) {
>   		amdgpu_vm_write_pte(p->adev, ib, pe, addr | flags,
> diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
> index 6786d34f7d04..277dbc6ba8ec 100644
> --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
> +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
> @@ -5212,7 +5212,7 @@ static int dm_plane_helper_prepare_fb(struct drm_plane *plane,
>   		return r;
>   	}
>   
> -	r = amdgpu_ttm_alloc_gart(&rbo->tbo);
> +	r = amdgpu_ttm_alloc_gart(rbo);
>   	if (unlikely(r != 0)) {
>   		amdgpu_bo_unpin(rbo);
>   		ttm_eu_backoff_reservation(&ticket, &list);



More information about the amd-gfx mailing list