diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_object.c')
-rw-r--r-- | drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 691 |
1 files changed, 388 insertions, 303 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index e3f16b49e970..aeb92e5677ac 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -31,6 +31,7 @@ */ #include <linux/list.h> #include <linux/slab.h> +#include <linux/dma-buf.h> #include <drm/amdgpu_drm.h> #include <drm/drm_cache.h> @@ -51,50 +52,44 @@ * */ -/** - * amdgpu_bo_subtract_pin_size - Remove BO from pin_size accounting - * - * @bo: &amdgpu_bo buffer object - * - * This function is called when a BO stops being pinned, and updates the - * &amdgpu_device pin_size values accordingly. - */ -static void amdgpu_bo_subtract_pin_size(struct amdgpu_bo *bo) -{ - struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); - - if (bo->tbo.mem.mem_type == TTM_PL_VRAM) { - atomic64_sub(amdgpu_bo_size(bo), &adev->vram_pin_size); - atomic64_sub(amdgpu_vram_mgr_bo_visible_size(bo), - &adev->visible_pin_size); - } else if (bo->tbo.mem.mem_type == TTM_PL_TT) { - atomic64_sub(amdgpu_bo_size(bo), &adev->gart_pin_size); - } -} - static void amdgpu_bo_destroy(struct ttm_buffer_object *tbo) { - struct amdgpu_device *adev = amdgpu_ttm_adev(tbo->bdev); struct amdgpu_bo *bo = ttm_to_amdgpu_bo(tbo); - if (bo->pin_count > 0) - amdgpu_bo_subtract_pin_size(bo); - amdgpu_bo_kunmap(bo); if (bo->tbo.base.import_attach) drm_prime_gem_destroy(&bo->tbo.base, bo->tbo.sg); drm_gem_object_release(&bo->tbo.base); + amdgpu_bo_unref(&bo->parent); + kvfree(bo); +} + +static void amdgpu_bo_user_destroy(struct ttm_buffer_object *tbo) +{ + struct amdgpu_bo *bo = ttm_to_amdgpu_bo(tbo); + struct amdgpu_bo_user *ubo; + + ubo = to_amdgpu_bo_user(bo); + kfree(ubo->metadata); + amdgpu_bo_destroy(tbo); +} + +static void amdgpu_bo_vm_destroy(struct ttm_buffer_object *tbo) +{ + struct amdgpu_device *adev = amdgpu_ttm_adev(tbo->bdev); + struct amdgpu_bo *bo = ttm_to_amdgpu_bo(tbo); + struct amdgpu_bo_vm *vmbo; + + vmbo = to_amdgpu_bo_vm(bo); /* in case amdgpu_device_recover_vram got NULL of bo->parent */ - if (!list_empty(&bo->shadow_list)) { + if (!list_empty(&vmbo->shadow_list)) { mutex_lock(&adev->shadow_list_lock); - list_del_init(&bo->shadow_list); + list_del_init(&vmbo->shadow_list); mutex_unlock(&adev->shadow_list_lock); } - amdgpu_bo_unref(&bo->parent); - kfree(bo->metadata); - kfree(bo); + amdgpu_bo_destroy(tbo); } /** @@ -109,8 +104,11 @@ static void amdgpu_bo_destroy(struct ttm_buffer_object *tbo) */ bool amdgpu_bo_is_amdgpu_bo(struct ttm_buffer_object *bo) { - if (bo->destroy == &amdgpu_bo_destroy) + if (bo->destroy == &amdgpu_bo_destroy || + bo->destroy == &amdgpu_bo_user_destroy || + bo->destroy == &amdgpu_bo_vm_destroy) return true; + return false; } @@ -135,8 +133,8 @@ void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain) places[c].fpfn = 0; places[c].lpfn = 0; - places[c].flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED | - TTM_PL_FLAG_VRAM; + places[c].mem_type = TTM_PL_VRAM; + places[c].flags = 0; if (flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) places[c].lpfn = visible_pfn; @@ -151,56 +149,54 @@ void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain) if (domain & AMDGPU_GEM_DOMAIN_GTT) { places[c].fpfn = 0; places[c].lpfn = 0; - places[c].flags = TTM_PL_FLAG_TT; - if (flags & AMDGPU_GEM_CREATE_CPU_GTT_USWC) - places[c].flags |= TTM_PL_FLAG_WC | - TTM_PL_FLAG_UNCACHED; - else - places[c].flags |= TTM_PL_FLAG_CACHED; + places[c].mem_type = + abo->flags & AMDGPU_GEM_CREATE_PREEMPTIBLE ? + AMDGPU_PL_PREEMPT : TTM_PL_TT; + places[c].flags = 0; c++; } if (domain & AMDGPU_GEM_DOMAIN_CPU) { places[c].fpfn = 0; places[c].lpfn = 0; - places[c].flags = TTM_PL_FLAG_SYSTEM; - if (flags & AMDGPU_GEM_CREATE_CPU_GTT_USWC) - places[c].flags |= TTM_PL_FLAG_WC | - TTM_PL_FLAG_UNCACHED; - else - places[c].flags |= TTM_PL_FLAG_CACHED; + places[c].mem_type = TTM_PL_SYSTEM; + places[c].flags = 0; c++; } if (domain & AMDGPU_GEM_DOMAIN_GDS) { places[c].fpfn = 0; places[c].lpfn = 0; - places[c].flags = TTM_PL_FLAG_UNCACHED | AMDGPU_PL_FLAG_GDS; + places[c].mem_type = AMDGPU_PL_GDS; + places[c].flags = 0; c++; } if (domain & AMDGPU_GEM_DOMAIN_GWS) { places[c].fpfn = 0; places[c].lpfn = 0; - places[c].flags = TTM_PL_FLAG_UNCACHED | AMDGPU_PL_FLAG_GWS; + places[c].mem_type = AMDGPU_PL_GWS; + places[c].flags = 0; c++; } if (domain & AMDGPU_GEM_DOMAIN_OA) { places[c].fpfn = 0; places[c].lpfn = 0; - places[c].flags = TTM_PL_FLAG_UNCACHED | AMDGPU_PL_FLAG_OA; + places[c].mem_type = AMDGPU_PL_OA; + places[c].flags = 0; c++; } if (!c) { places[c].fpfn = 0; places[c].lpfn = 0; - places[c].flags = TTM_PL_MASK_CACHING | TTM_PL_FLAG_SYSTEM; + places[c].mem_type = TTM_PL_SYSTEM; + places[c].flags = 0; c++; } - BUG_ON(c >= AMDGPU_BO_MAX_PLACEMENTS); + BUG_ON(c > AMDGPU_BO_MAX_PLACEMENTS); placement->num_placement = c; placement->placement = places; @@ -251,6 +247,7 @@ int amdgpu_bo_create_reserved(struct amdgpu_device *adev, bp.flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; bp.type = ttm_bo_type_kernel; bp.resv = NULL; + bp.bo_ptr_size = sizeof(struct amdgpu_bo); if (!*bo_ptr) { r = amdgpu_bo_create(adev, &bp, bo_ptr); @@ -373,6 +370,9 @@ int amdgpu_bo_create_kernel_at(struct amdgpu_device *adev, if (r) return r; + if ((*bo_ptr) == NULL) + return 0; + /* * Remove the original mem node and create a new one at the request * position. @@ -380,14 +380,14 @@ int amdgpu_bo_create_kernel_at(struct amdgpu_device *adev, if (cpu_addr) amdgpu_bo_kunmap(*bo_ptr); - ttm_bo_mem_put(&(*bo_ptr)->tbo, &(*bo_ptr)->tbo.mem); + ttm_resource_free(&(*bo_ptr)->tbo, &(*bo_ptr)->tbo.resource); for (i = 0; i < (*bo_ptr)->placement.num_placement; ++i) { (*bo_ptr)->placements[i].fpfn = offset >> PAGE_SHIFT; (*bo_ptr)->placements[i].lpfn = (offset + size) >> PAGE_SHIFT; } r = ttm_bo_mem_space(&(*bo_ptr)->tbo, &(*bo_ptr)->placement, - &(*bo_ptr)->tbo.mem, &ctx); + &(*bo_ptr)->tbo.resource, &ctx); if (r) goto error; @@ -441,14 +441,14 @@ void amdgpu_bo_free_kernel(struct amdgpu_bo **bo, u64 *gpu_addr, static bool amdgpu_bo_validate_size(struct amdgpu_device *adev, unsigned long size, u32 domain) { - struct ttm_mem_type_manager *man = NULL; + struct ttm_resource_manager *man = NULL; /* * If GTT is part of requested domains the check must succeed to * allow fall back to GTT */ if (domain & AMDGPU_GEM_DOMAIN_GTT) { - man = &adev->mman.bdev.man[TTM_PL_TT]; + man = ttm_manager_type(&adev->mman.bdev, TTM_PL_TT); if (size < (man->size << PAGE_SHIFT)) return true; @@ -457,7 +457,7 @@ static bool amdgpu_bo_validate_size(struct amdgpu_device *adev, } if (domain & AMDGPU_GEM_DOMAIN_VRAM) { - man = &adev->mman.bdev.man[TTM_PL_VRAM]; + man = ttm_manager_type(&adev->mman.bdev, TTM_PL_VRAM); if (size < (man->size << PAGE_SHIFT)) return true; @@ -509,20 +509,31 @@ bool amdgpu_bo_support_uswc(u64 bo_flags) #endif } -static int amdgpu_bo_do_create(struct amdgpu_device *adev, +/** + * amdgpu_bo_create - create an &amdgpu_bo buffer object + * @adev: amdgpu device object + * @bp: parameters to be used for the buffer object + * @bo_ptr: pointer to the buffer object pointer + * + * Creates an &amdgpu_bo buffer object. + * + * Returns: + * 0 for success or a negative error code on failure. + */ +int amdgpu_bo_create(struct amdgpu_device *adev, struct amdgpu_bo_param *bp, struct amdgpu_bo **bo_ptr) { struct ttm_operation_ctx ctx = { .interruptible = (bp->type != ttm_bo_type_kernel), .no_wait_gpu = bp->no_wait_gpu, - .resv = bp->resv, - .flags = bp->type != ttm_bo_type_kernel ? - TTM_OPT_FLAG_ALLOW_RES_EVICT : 0 + /* We opt to avoid OOM on system pages allocations */ + .gfp_retry_mayfail = true, + .allow_res_evict = bp->type != ttm_bo_type_kernel, + .resv = bp->resv }; struct amdgpu_bo *bo; unsigned long page_align, size = bp->size; - size_t acc_size; int r; /* Note that GDS/GWS/OA allocates 1 page per byte/resource. */ @@ -543,16 +554,13 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, if (!amdgpu_bo_validate_size(adev, size, bp->domain)) return -ENOMEM; - *bo_ptr = NULL; + BUG_ON(bp->bo_ptr_size < sizeof(struct amdgpu_bo)); - acc_size = ttm_bo_dma_acc_size(&adev->mman.bdev, size, - sizeof(struct amdgpu_bo)); - - bo = kzalloc(sizeof(struct amdgpu_bo), GFP_KERNEL); + *bo_ptr = NULL; + bo = kvzalloc(bp->bo_ptr_size, GFP_KERNEL); if (bo == NULL) return -ENOMEM; - drm_gem_private_object_init(adev->ddev, &bo->tbo.base, size); - INIT_LIST_HEAD(&bo->shadow_list); + drm_gem_private_object_init(adev_to_drm(adev), &bo->tbo.base, size); bo->vm_bo = NULL; bo->preferred_domains = bp->preferred_domain ? bp->preferred_domain : bp->domain; @@ -575,22 +583,25 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, if (bp->type == ttm_bo_type_kernel) bo->tbo.priority = 1; + if (!bp->destroy) + bp->destroy = &amdgpu_bo_destroy; + r = ttm_bo_init_reserved(&adev->mman.bdev, &bo->tbo, size, bp->type, - &bo->placement, page_align, &ctx, acc_size, - NULL, bp->resv, &amdgpu_bo_destroy); + &bo->placement, page_align, &ctx, NULL, + bp->resv, bp->destroy); if (unlikely(r != 0)) return r; if (!amdgpu_gmc_vram_full_visible(&adev->gmc) && - bo->tbo.mem.mem_type == TTM_PL_VRAM && - bo->tbo.mem.start < adev->gmc.visible_vram_size >> PAGE_SHIFT) + bo->tbo.resource->mem_type == TTM_PL_VRAM && + bo->tbo.resource->start < adev->gmc.visible_vram_size >> PAGE_SHIFT) amdgpu_cs_report_moved_bytes(adev, ctx.bytes_moved, ctx.bytes_moved); else amdgpu_cs_report_moved_bytes(adev, ctx.bytes_moved, 0); if (bp->flags & AMDGPU_GEM_CREATE_VRAM_CLEARED && - bo->tbo.mem.placement & TTM_PL_FLAG_VRAM) { + bo->tbo.resource->mem_type == TTM_PL_VRAM) { struct dma_fence *fence; r = amdgpu_fill_buffer(bo, 0, bo->tbo.base.resv, &fence); @@ -621,110 +632,82 @@ fail_unreserve: return r; } -static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, - unsigned long size, - struct amdgpu_bo *bo) +/** + * amdgpu_bo_create_user - create an &amdgpu_bo_user buffer object + * @adev: amdgpu device object + * @bp: parameters to be used for the buffer object + * @ubo_ptr: pointer to the buffer object pointer + * + * Create a BO to be used by user application; + * + * Returns: + * 0 for success or a negative error code on failure. + */ + +int amdgpu_bo_create_user(struct amdgpu_device *adev, + struct amdgpu_bo_param *bp, + struct amdgpu_bo_user **ubo_ptr) { - struct amdgpu_bo_param bp; + struct amdgpu_bo *bo_ptr; int r; - if (bo->shadow) - return 0; - - memset(&bp, 0, sizeof(bp)); - bp.size = size; - bp.domain = AMDGPU_GEM_DOMAIN_GTT; - bp.flags = AMDGPU_GEM_CREATE_CPU_GTT_USWC | - AMDGPU_GEM_CREATE_SHADOW; - bp.type = ttm_bo_type_kernel; - bp.resv = bo->tbo.base.resv; - - r = amdgpu_bo_do_create(adev, &bp, &bo->shadow); - if (!r) { - bo->shadow->parent = amdgpu_bo_ref(bo); - mutex_lock(&adev->shadow_list_lock); - list_add_tail(&bo->shadow->shadow_list, &adev->shadow_list); - mutex_unlock(&adev->shadow_list_lock); - } + bp->bo_ptr_size = sizeof(struct amdgpu_bo_user); + bp->destroy = &amdgpu_bo_user_destroy; + r = amdgpu_bo_create(adev, bp, &bo_ptr); + if (r) + return r; + *ubo_ptr = to_amdgpu_bo_user(bo_ptr); return r; } /** - * amdgpu_bo_create - create an &amdgpu_bo buffer object + * amdgpu_bo_create_vm - create an &amdgpu_bo_vm buffer object * @adev: amdgpu device object * @bp: parameters to be used for the buffer object - * @bo_ptr: pointer to the buffer object pointer + * @vmbo_ptr: pointer to the buffer object pointer * - * Creates an &amdgpu_bo buffer object; and if requested, also creates a - * shadow object. - * Shadow object is used to backup the original buffer object, and is always - * in GTT. + * Create a BO to be for GPUVM. * * Returns: * 0 for success or a negative error code on failure. */ -int amdgpu_bo_create(struct amdgpu_device *adev, - struct amdgpu_bo_param *bp, - struct amdgpu_bo **bo_ptr) + +int amdgpu_bo_create_vm(struct amdgpu_device *adev, + struct amdgpu_bo_param *bp, + struct amdgpu_bo_vm **vmbo_ptr) { - u64 flags = bp->flags; + struct amdgpu_bo *bo_ptr; int r; - bp->flags = bp->flags & ~AMDGPU_GEM_CREATE_SHADOW; - r = amdgpu_bo_do_create(adev, bp, bo_ptr); + /* bo_ptr_size will be determined by the caller and it depends on + * num of amdgpu_vm_pt entries. + */ + BUG_ON(bp->bo_ptr_size < sizeof(struct amdgpu_bo_vm)); + bp->destroy = &amdgpu_bo_vm_destroy; + r = amdgpu_bo_create(adev, bp, &bo_ptr); if (r) return r; - if ((flags & AMDGPU_GEM_CREATE_SHADOW) && !(adev->flags & AMD_IS_APU)) { - if (!bp->resv) - WARN_ON(dma_resv_lock((*bo_ptr)->tbo.base.resv, - NULL)); - - r = amdgpu_bo_create_shadow(adev, bp->size, *bo_ptr); - - if (!bp->resv) - dma_resv_unlock((*bo_ptr)->tbo.base.resv); - - if (r) - amdgpu_bo_unref(bo_ptr); - } - + *vmbo_ptr = to_amdgpu_bo_vm(bo_ptr); + INIT_LIST_HEAD(&(*vmbo_ptr)->shadow_list); return r; } /** - * amdgpu_bo_validate - validate an &amdgpu_bo buffer object - * @bo: pointer to the buffer object + * amdgpu_bo_add_to_shadow_list - add a BO to the shadow list * - * Sets placement according to domain; and changes placement and caching - * policy of the buffer object according to the placement. - * This is used for validating shadow bos. It calls ttm_bo_validate() to - * make sure the buffer is resident where it needs to be. + * @vmbo: BO that will be inserted into the shadow list * - * Returns: - * 0 for success or a negative error code on failure. + * Insert a BO to the shadow list. */ -int amdgpu_bo_validate(struct amdgpu_bo *bo) +void amdgpu_bo_add_to_shadow_list(struct amdgpu_bo_vm *vmbo) { - struct ttm_operation_ctx ctx = { false, false }; - uint32_t domain; - int r; + struct amdgpu_device *adev = amdgpu_ttm_adev(vmbo->bo.tbo.bdev); - if (bo->pin_count) - return 0; - - domain = bo->preferred_domains; - -retry: - amdgpu_bo_placement_from_domain(bo, domain); - r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); - if (unlikely(r == -ENOMEM) && domain != bo->allowed_domains) { - domain = bo->allowed_domains; - goto retry; - } - - return r; + mutex_lock(&adev->shadow_list_lock); + list_add_tail(&vmbo->shadow_list, &adev->shadow_list); + mutex_unlock(&adev->shadow_list_lock); } /** @@ -752,7 +735,7 @@ int amdgpu_bo_restore_shadow(struct amdgpu_bo *shadow, struct dma_fence **fence) return amdgpu_copy_buffer(ring, shadow_addr, parent_addr, amdgpu_bo_size(shadow), NULL, fence, - true, false); + true, false, false); } /** @@ -781,12 +764,12 @@ int amdgpu_bo_kmap(struct amdgpu_bo *bo, void **ptr) return 0; } - r = dma_resv_wait_timeout_rcu(bo->tbo.base.resv, false, false, - MAX_SCHEDULE_TIMEOUT); + r = dma_resv_wait_timeout(bo->tbo.base.resv, false, false, + MAX_SCHEDULE_TIMEOUT); if (r < 0) return r; - r = ttm_bo_kmap(&bo->tbo, 0, bo->tbo.num_pages, &bo->kmap); + r = ttm_bo_kmap(&bo->tbo, 0, bo->tbo.resource->num_pages, &bo->kmap); if (r) return r; @@ -896,28 +879,30 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, return -EINVAL; /* A shared bo cannot be migrated to VRAM */ - if (bo->prime_shared_count) { + if (bo->tbo.base.import_attach) { if (domain & AMDGPU_GEM_DOMAIN_GTT) domain = AMDGPU_GEM_DOMAIN_GTT; else return -EINVAL; } - /* This assumes only APU display buffers are pinned with (VRAM|GTT). - * See function amdgpu_display_supported_domains() - */ - domain = amdgpu_bo_get_preferred_pin_domain(adev, domain); - - if (bo->pin_count) { - uint32_t mem_type = bo->tbo.mem.mem_type; + if (bo->tbo.pin_count) { + uint32_t mem_type = bo->tbo.resource->mem_type; + uint32_t mem_flags = bo->tbo.resource->placement; if (!(domain & amdgpu_mem_type_to_domain(mem_type))) return -EINVAL; - bo->pin_count++; + if ((mem_type == TTM_PL_VRAM) && + (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS) && + !(mem_flags & TTM_PL_FLAG_CONTIGUOUS)) + return -EINVAL; + + ttm_bo_pin(&bo->tbo); if (max_offset != 0) { - u64 domain_start = bo->tbo.bdev->man[mem_type].gpu_offset; + u64 domain_start = amdgpu_ttm_domain_start(adev, + mem_type); WARN_ON_ONCE(max_offset < (amdgpu_bo_gpu_offset(bo) - domain_start)); } @@ -925,7 +910,14 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, return 0; } - bo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; + /* This assumes only APU display buffers are pinned with (VRAM|GTT). + * See function amdgpu_display_supported_domains() + */ + domain = amdgpu_bo_get_preferred_domain(adev, domain); + + if (bo->tbo.base.import_attach) + dma_buf_pin(bo->tbo.base.import_attach); + /* force to pin into visible video ram */ if (!(bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS)) bo->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; @@ -941,7 +933,6 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, if (!bo->placements[i].lpfn || (lpfn && lpfn < bo->placements[i].lpfn)) bo->placements[i].lpfn = lpfn; - bo->placements[i].flags |= TTM_PL_FLAG_NO_EVICT; } r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); @@ -950,9 +941,9 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, goto error; } - bo->pin_count = 1; + ttm_bo_pin(&bo->tbo); - domain = amdgpu_mem_type_to_domain(bo->tbo.mem.mem_type); + domain = amdgpu_mem_type_to_domain(bo->tbo.resource->mem_type); if (domain == AMDGPU_GEM_DOMAIN_VRAM) { atomic64_add(amdgpu_bo_size(bo), &adev->vram_pin_size); atomic64_add(amdgpu_vram_mgr_bo_visible_size(bo), @@ -979,6 +970,7 @@ error: */ int amdgpu_bo_pin(struct amdgpu_bo *bo, u32 domain) { + bo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; return amdgpu_bo_pin_restricted(bo, domain, 0, 0); } @@ -992,53 +984,24 @@ int amdgpu_bo_pin(struct amdgpu_bo *bo, u32 domain) * Returns: * 0 for success or a negative error code on failure. */ -int amdgpu_bo_unpin(struct amdgpu_bo *bo) +void amdgpu_bo_unpin(struct amdgpu_bo *bo) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); - struct ttm_operation_ctx ctx = { false, false }; - int r, i; - - if (WARN_ON_ONCE(!bo->pin_count)) { - dev_warn(adev->dev, "%p unpin not necessary\n", bo); - return 0; - } - bo->pin_count--; - if (bo->pin_count) - return 0; - amdgpu_bo_subtract_pin_size(bo); - - for (i = 0; i < bo->placement.num_placement; i++) { - bo->placements[i].lpfn = 0; - bo->placements[i].flags &= ~TTM_PL_FLAG_NO_EVICT; - } - r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); - if (unlikely(r)) - dev_err(adev->dev, "%p validate failed for unpin\n", bo); + ttm_bo_unpin(&bo->tbo); + if (bo->tbo.pin_count) + return; - return r; -} + if (bo->tbo.base.import_attach) + dma_buf_unpin(bo->tbo.base.import_attach); -/** - * amdgpu_bo_evict_vram - evict VRAM buffers - * @adev: amdgpu device object - * - * Evicts all VRAM buffers on the lru list of the memory type. - * Mainly used for evicting vram at suspend time. - * - * Returns: - * 0 for success or a negative error code on failure. - */ -int amdgpu_bo_evict_vram(struct amdgpu_device *adev) -{ - /* late 2.6.33 fix IGP hibernate - we need pm ops to do this correct */ -#ifndef CONFIG_HIBERNATION - if (adev->flags & AMD_IS_APU) { - /* Useless to evict on IGP chips */ - return 0; + if (bo->tbo.resource->mem_type == TTM_PL_VRAM) { + atomic64_sub(amdgpu_bo_size(bo), &adev->vram_pin_size); + atomic64_sub(amdgpu_vram_mgr_bo_visible_size(bo), + &adev->visible_pin_size); + } else if (bo->tbo.resource->mem_type == TTM_PL_TT) { + atomic64_sub(amdgpu_bo_size(bo), &adev->gart_pin_size); } -#endif - return ttm_bo_evict_mm(&adev->mman.bdev, TTM_PL_VRAM); } static const char *amdgpu_vram_names[] = { @@ -1052,6 +1015,7 @@ static const char *amdgpu_vram_names[] = { "DDR3", "DDR4", "GDDR6", + "DDR5" }; /** @@ -1065,13 +1029,17 @@ static const char *amdgpu_vram_names[] = { */ int amdgpu_bo_init(struct amdgpu_device *adev) { - /* reserve PAT memory space to WC for VRAM */ - arch_io_reserve_memtype_wc(adev->gmc.aper_base, - adev->gmc.aper_size); + /* On A+A platform, VRAM can be mapped as WB */ + if (!adev->gmc.xgmi.connected_to_cpu) { + /* reserve PAT memory space to WC for VRAM */ + arch_io_reserve_memtype_wc(adev->gmc.aper_base, + adev->gmc.aper_size); + + /* Add an MTRR for the VRAM */ + adev->gmc.vram_mtrr = arch_phys_wc_add(adev->gmc.aper_base, + adev->gmc.aper_size); + } - /* Add an MTRR for the VRAM */ - adev->gmc.vram_mtrr = arch_phys_wc_add(adev->gmc.aper_base, - adev->gmc.aper_size); DRM_INFO("Detected VRAM RAM=%lluM, BAR=%lluM\n", adev->gmc.mc_vram_size >> 20, (unsigned long long)adev->gmc.aper_size >> 20); @@ -1081,23 +1049,6 @@ int amdgpu_bo_init(struct amdgpu_device *adev) } /** - * amdgpu_bo_late_init - late init - * @adev: amdgpu device object - * - * Calls amdgpu_ttm_late_init() to free resources used earlier during - * initialization. - * - * Returns: - * 0 for success or a negative error code on failure. - */ -int amdgpu_bo_late_init(struct amdgpu_device *adev) -{ - amdgpu_ttm_late_init(adev); - - return 0; -} - -/** * amdgpu_bo_fini - tear down memory manager * @adev: amdgpu device object * @@ -1106,27 +1057,6 @@ int amdgpu_bo_late_init(struct amdgpu_device *adev) void amdgpu_bo_fini(struct amdgpu_device *adev) { amdgpu_ttm_fini(adev); - arch_phys_wc_del(adev->gmc.vram_mtrr); - arch_io_free_memtype_wc(adev->gmc.aper_base, adev->gmc.aper_size); -} - -/** - * amdgpu_bo_fbdev_mmap - mmap fbdev memory - * @bo: &amdgpu_bo buffer object - * @vma: vma as input from the fbdev mmap method - * - * Calls ttm_fbdev_mmap() to mmap fbdev memory if it is backed by a bo. - * - * Returns: - * 0 for success or a negative error code on failure. - */ -int amdgpu_bo_fbdev_mmap(struct amdgpu_bo *bo, - struct vm_area_struct *vma) -{ - if (vma->vm_pgoff != 0) - return -EACCES; - - return ttm_bo_mmap_obj(vma, &bo->tbo); } /** @@ -1143,12 +1073,15 @@ int amdgpu_bo_fbdev_mmap(struct amdgpu_bo *bo, int amdgpu_bo_set_tiling_flags(struct amdgpu_bo *bo, u64 tiling_flags) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); + struct amdgpu_bo_user *ubo; + BUG_ON(bo->tbo.type == ttm_bo_type_kernel); if (adev->family <= AMDGPU_FAMILY_CZ && AMDGPU_TILING_GET(tiling_flags, TILE_SPLIT) > 6) return -EINVAL; - bo->tiling_flags = tiling_flags; + ubo = to_amdgpu_bo_user(bo); + ubo->tiling_flags = tiling_flags; return 0; } @@ -1162,10 +1095,14 @@ int amdgpu_bo_set_tiling_flags(struct amdgpu_bo *bo, u64 tiling_flags) */ void amdgpu_bo_get_tiling_flags(struct amdgpu_bo *bo, u64 *tiling_flags) { + struct amdgpu_bo_user *ubo; + + BUG_ON(bo->tbo.type == ttm_bo_type_kernel); dma_resv_assert_held(bo->tbo.base.resv); + ubo = to_amdgpu_bo_user(bo); if (tiling_flags) - *tiling_flags = bo->tiling_flags; + *tiling_flags = ubo->tiling_flags; } /** @@ -1184,13 +1121,16 @@ void amdgpu_bo_get_tiling_flags(struct amdgpu_bo *bo, u64 *tiling_flags) int amdgpu_bo_set_metadata (struct amdgpu_bo *bo, void *metadata, uint32_t metadata_size, uint64_t flags) { + struct amdgpu_bo_user *ubo; void *buffer; + BUG_ON(bo->tbo.type == ttm_bo_type_kernel); + ubo = to_amdgpu_bo_user(bo); if (!metadata_size) { - if (bo->metadata_size) { - kfree(bo->metadata); - bo->metadata = NULL; - bo->metadata_size = 0; + if (ubo->metadata_size) { + kfree(ubo->metadata); + ubo->metadata = NULL; + ubo->metadata_size = 0; } return 0; } @@ -1202,10 +1142,10 @@ int amdgpu_bo_set_metadata (struct amdgpu_bo *bo, void *metadata, if (buffer == NULL) return -ENOMEM; - kfree(bo->metadata); - bo->metadata_flags = flags; - bo->metadata = buffer; - bo->metadata_size = metadata_size; + kfree(ubo->metadata); + ubo->metadata_flags = flags; + ubo->metadata = buffer; + ubo->metadata_size = metadata_size; return 0; } @@ -1229,21 +1169,26 @@ int amdgpu_bo_get_metadata(struct amdgpu_bo *bo, void *buffer, size_t buffer_size, uint32_t *metadata_size, uint64_t *flags) { + struct amdgpu_bo_user *ubo; + if (!buffer && !metadata_size) return -EINVAL; + BUG_ON(bo->tbo.type == ttm_bo_type_kernel); + ubo = to_amdgpu_bo_user(bo); + if (metadata_size) + *metadata_size = ubo->metadata_size; + if (buffer) { - if (buffer_size < bo->metadata_size) + if (buffer_size < ubo->metadata_size) return -EINVAL; - if (bo->metadata_size) - memcpy(buffer, bo->metadata, bo->metadata_size); + if (ubo->metadata_size) + memcpy(buffer, ubo->metadata, ubo->metadata_size); } - if (metadata_size) - *metadata_size = bo->metadata_size; if (flags) - *flags = bo->metadata_flags; + *flags = ubo->metadata_flags; return 0; } @@ -1260,11 +1205,11 @@ int amdgpu_bo_get_metadata(struct amdgpu_bo *bo, void *buffer, */ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo, bool evict, - struct ttm_mem_reg *new_mem) + struct ttm_resource *new_mem) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev); struct amdgpu_bo *abo; - struct ttm_mem_reg *old_mem = &bo->mem; + struct ttm_resource *old_mem = bo->resource; if (!amdgpu_bo_is_amdgpu_bo(bo)) return; @@ -1274,6 +1219,10 @@ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo, amdgpu_bo_kunmap(abo); + if (abo->tbo.base.dma_buf && !abo->tbo.base.import_attach && + bo->resource->mem_type != TTM_PL_SYSTEM) + dma_buf_move_notify(abo->tbo.base.dma_buf); + /* remember the eviction */ if (evict) atomic64_inc(&adev->num_evictions); @@ -1286,8 +1235,28 @@ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo, trace_amdgpu_bo_move(abo, new_mem->mem_type, old_mem->mem_type); } +void amdgpu_bo_get_memory(struct amdgpu_bo *bo, uint64_t *vram_mem, + uint64_t *gtt_mem, uint64_t *cpu_mem) +{ + unsigned int domain; + + domain = amdgpu_mem_type_to_domain(bo->tbo.resource->mem_type); + switch (domain) { + case AMDGPU_GEM_DOMAIN_VRAM: + *vram_mem += amdgpu_bo_size(bo); + break; + case AMDGPU_GEM_DOMAIN_GTT: + *gtt_mem += amdgpu_bo_size(bo); + break; + case AMDGPU_GEM_DOMAIN_CPU: + default: + *cpu_mem += amdgpu_bo_size(bo); + break; + } +} + /** - * amdgpu_bo_move_notify - notification about a BO being released + * amdgpu_bo_release_notify - notification about a BO being released * @bo: pointer to a buffer object * * Wipes VRAM buffers whose contents should not be leaked before the @@ -1307,7 +1276,13 @@ void amdgpu_bo_release_notify(struct ttm_buffer_object *bo) if (abo->kfd_bo) amdgpu_amdkfd_unreserve_memory_limit(abo); - if (bo->mem.mem_type != TTM_PL_VRAM || !bo->mem.mm_node || + /* We only remove the fence if the resv has individualized. */ + WARN_ON_ONCE(bo->type == ttm_bo_type_kernel + && bo->base.resv != &bo->base._resv); + if (bo->base.resv == &bo->base._resv) + amdgpu_amdkfd_remove_fence_on_pt_pd_bos(abo); + + if (bo->resource->mem_type != TTM_PL_VRAM || !(abo->flags & AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE)) return; @@ -1333,33 +1308,27 @@ void amdgpu_bo_release_notify(struct ttm_buffer_object *bo) * Returns: * 0 for success or a negative error code on failure. */ -int amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo) +vm_fault_t amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev); struct ttm_operation_ctx ctx = { false, false }; - struct amdgpu_bo *abo; - unsigned long offset, size; + struct amdgpu_bo *abo = ttm_to_amdgpu_bo(bo); + unsigned long offset; int r; - if (!amdgpu_bo_is_amdgpu_bo(bo)) - return 0; - - abo = ttm_to_amdgpu_bo(bo); - /* Remember that this BO was accessed by the CPU */ abo->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; - if (bo->mem.mem_type != TTM_PL_VRAM) + if (bo->resource->mem_type != TTM_PL_VRAM) return 0; - size = bo->mem.num_pages << PAGE_SHIFT; - offset = bo->mem.start << PAGE_SHIFT; - if ((offset + size) <= adev->gmc.visible_vram_size) + offset = bo->resource->start << PAGE_SHIFT; + if ((offset + bo->base.size) <= adev->gmc.visible_vram_size) return 0; /* Can't move a pinned BO to visible VRAM */ - if (abo->pin_count > 0) - return -EINVAL; + if (abo->tbo.pin_count > 0) + return VM_FAULT_SIGBUS; /* hurrah the memory is not visible ! */ atomic64_inc(&adev->num_vram_cpu_page_faults); @@ -1371,15 +1340,18 @@ int amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo) abo->placement.busy_placement = &abo->placements[1]; r = ttm_bo_validate(bo, &abo->placement, &ctx); - if (unlikely(r != 0)) - return r; + if (unlikely(r == -EBUSY || r == -ERESTARTSYS)) + return VM_FAULT_NOPAGE; + else if (unlikely(r)) + return VM_FAULT_SIGBUS; - offset = bo->mem.start << PAGE_SHIFT; + offset = bo->resource->start << PAGE_SHIFT; /* this should never happen */ - if (bo->mem.mem_type == TTM_PL_VRAM && - (offset + size) > adev->gmc.visible_vram_size) - return -EINVAL; + if (bo->resource->mem_type == TTM_PL_VRAM && + (offset + bo->base.size) > adev->gmc.visible_vram_size) + return VM_FAULT_SIGBUS; + ttm_bo_move_to_lru_tail_unlocked(bo); return 0; } @@ -1403,30 +1375,52 @@ void amdgpu_bo_fence(struct amdgpu_bo *bo, struct dma_fence *fence, } /** - * amdgpu_sync_wait_resv - Wait for BO reservation fences + * amdgpu_bo_sync_wait_resv - Wait for BO reservation fences * - * @bo: buffer object + * @adev: amdgpu device pointer + * @resv: reservation object to sync to + * @sync_mode: synchronization mode * @owner: fence owner * @intr: Whether the wait is interruptible * + * Extract the fences from the reservation object and waits for them to finish. + * * Returns: * 0 on success, errno otherwise. */ -int amdgpu_bo_sync_wait(struct amdgpu_bo *bo, void *owner, bool intr) +int amdgpu_bo_sync_wait_resv(struct amdgpu_device *adev, struct dma_resv *resv, + enum amdgpu_sync_mode sync_mode, void *owner, + bool intr) { - struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); struct amdgpu_sync sync; int r; amdgpu_sync_create(&sync); - amdgpu_sync_resv(adev, &sync, bo->tbo.base.resv, owner, false); + amdgpu_sync_resv(adev, &sync, resv, sync_mode, owner); r = amdgpu_sync_wait(&sync, intr); amdgpu_sync_free(&sync); - return r; } /** + * amdgpu_bo_sync_wait - Wrapper for amdgpu_bo_sync_wait_resv + * @bo: buffer object to wait for + * @owner: fence owner + * @intr: Whether the wait is interruptible + * + * Wrapper to wait for fences in a BO. + * Returns: + * 0 on success, errno otherwise. + */ +int amdgpu_bo_sync_wait(struct amdgpu_bo *bo, void *owner, bool intr) +{ + struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); + + return amdgpu_bo_sync_wait_resv(adev, bo->tbo.base.resv, + AMDGPU_SYNC_NE_OWNER, owner, intr); +} + +/** * amdgpu_bo_gpu_offset - return GPU offset of bo * @bo: amdgpu object for which we query the offset * @@ -1438,25 +1432,43 @@ int amdgpu_bo_sync_wait(struct amdgpu_bo *bo, void *owner, bool intr) */ u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo) { - WARN_ON_ONCE(bo->tbo.mem.mem_type == TTM_PL_SYSTEM); + WARN_ON_ONCE(bo->tbo.resource->mem_type == TTM_PL_SYSTEM); WARN_ON_ONCE(!dma_resv_is_locked(bo->tbo.base.resv) && - !bo->pin_count && bo->tbo.type != ttm_bo_type_kernel); - WARN_ON_ONCE(bo->tbo.mem.start == AMDGPU_BO_INVALID_OFFSET); - WARN_ON_ONCE(bo->tbo.mem.mem_type == TTM_PL_VRAM && + !bo->tbo.pin_count && bo->tbo.type != ttm_bo_type_kernel); + WARN_ON_ONCE(bo->tbo.resource->start == AMDGPU_BO_INVALID_OFFSET); + WARN_ON_ONCE(bo->tbo.resource->mem_type == TTM_PL_VRAM && !(bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS)); - return amdgpu_gmc_sign_extend(bo->tbo.offset); + return amdgpu_bo_gpu_offset_no_check(bo); } /** - * amdgpu_bo_get_preferred_pin_domain - get preferred domain for scanout + * amdgpu_bo_gpu_offset_no_check - return GPU offset of bo + * @bo: amdgpu object for which we query the offset + * + * Returns: + * current GPU offset of the object without raising warnings. + */ +u64 amdgpu_bo_gpu_offset_no_check(struct amdgpu_bo *bo) +{ + struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); + uint64_t offset; + + offset = (bo->tbo.resource->start << PAGE_SHIFT) + + amdgpu_ttm_domain_start(adev, bo->tbo.resource->mem_type); + + return amdgpu_gmc_sign_extend(offset); +} + +/** + * amdgpu_bo_get_preferred_domain - get preferred domain * @adev: amdgpu device object * @domain: allowed :ref:`memory domains <amdgpu_memory_domains>` * * Returns: - * Which of the allowed domains is preferred for pinning the BO for scanout. + * Which of the allowed domains is preferred for allocating the BO. */ -uint32_t amdgpu_bo_get_preferred_pin_domain(struct amdgpu_device *adev, +uint32_t amdgpu_bo_get_preferred_domain(struct amdgpu_device *adev, uint32_t domain) { if (domain == (AMDGPU_GEM_DOMAIN_VRAM | AMDGPU_GEM_DOMAIN_GTT)) { @@ -1466,3 +1478,76 @@ uint32_t amdgpu_bo_get_preferred_pin_domain(struct amdgpu_device *adev, } return domain; } + +#if defined(CONFIG_DEBUG_FS) +#define amdgpu_bo_print_flag(m, bo, flag) \ + do { \ + if (bo->flags & (AMDGPU_GEM_CREATE_ ## flag)) { \ + seq_printf((m), " " #flag); \ + } \ + } while (0) + +/** + * amdgpu_bo_print_info - print BO info in debugfs file + * + * @id: Index or Id of the BO + * @bo: Requested BO for printing info + * @m: debugfs file + * + * Print BO information in debugfs file + * + * Returns: + * Size of the BO in bytes. + */ +u64 amdgpu_bo_print_info(int id, struct amdgpu_bo *bo, struct seq_file *m) +{ + struct dma_buf_attachment *attachment; + struct dma_buf *dma_buf; + unsigned int domain; + const char *placement; + unsigned int pin_count; + u64 size; + + domain = amdgpu_mem_type_to_domain(bo->tbo.resource->mem_type); + switch (domain) { + case AMDGPU_GEM_DOMAIN_VRAM: + placement = "VRAM"; + break; + case AMDGPU_GEM_DOMAIN_GTT: + placement = " GTT"; + break; + case AMDGPU_GEM_DOMAIN_CPU: + default: + placement = " CPU"; + break; + } + + size = amdgpu_bo_size(bo); + seq_printf(m, "\t\t0x%08x: %12lld byte %s", + id, size, placement); + + pin_count = READ_ONCE(bo->tbo.pin_count); + if (pin_count) + seq_printf(m, " pin count %d", pin_count); + + dma_buf = READ_ONCE(bo->tbo.base.dma_buf); + attachment = READ_ONCE(bo->tbo.base.import_attach); + + if (attachment) + seq_printf(m, " imported from %p", dma_buf); + else if (dma_buf) + seq_printf(m, " exported as %p", dma_buf); + + amdgpu_bo_print_flag(m, bo, CPU_ACCESS_REQUIRED); + amdgpu_bo_print_flag(m, bo, NO_CPU_ACCESS); + amdgpu_bo_print_flag(m, bo, CPU_GTT_USWC); + amdgpu_bo_print_flag(m, bo, VRAM_CLEARED); + amdgpu_bo_print_flag(m, bo, VRAM_CONTIGUOUS); + amdgpu_bo_print_flag(m, bo, VM_ALWAYS_VALID); + amdgpu_bo_print_flag(m, bo, EXPLICIT_SYNC); + + seq_puts(m, "\n"); + + return size; +} +#endif |