/* * Copyright 2009 Jerome Glisse. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including * without limitation the rights to use, copy, modify, merge, publish, * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE * USE OR OTHER DEALINGS IN THE SOFTWARE. * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. * */ /* * Authors: * Jerome Glisse * Thomas Hellstrom * Dave Airlie */ #include #include #include #include #include #include "amdgpu.h" #include "amdgpu_trace.h" #include "amdgpu_amdkfd.h" /** * DOC: amdgpu_object * * This defines the interfaces to operate on an &amdgpu_bo buffer object which * represents memory used by driver (VRAM, system memory, etc.). The driver * provides DRM/GEM APIs to userspace. DRM/GEM APIs then use these interfaces * to create/destroy/set buffer object which are then managed by the kernel TTM * memory manager. * The interfaces are also used internally by kernel clients, including gfx, * uvd, etc. for kernel managed allocations used by the GPU. * */ /** * 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); /* in case amdgpu_device_recover_vram got NULL of bo->parent */ if (!list_empty(&bo->shadow_list)) { mutex_lock(&adev->shadow_list_lock); list_del_init(&bo->shadow_list); mutex_unlock(&adev->shadow_list_lock); } amdgpu_bo_unref(&bo->parent); kfree(bo->metadata); kfree(bo); } /** * amdgpu_bo_is_amdgpu_bo - check if the buffer object is an &amdgpu_bo * @bo: buffer object to be checked * * Uses destroy function associated with the object to determine if this is * an &amdgpu_bo. * * Returns: * true if the object belongs to &amdgpu_bo, false if not. */ bool amdgpu_bo_is_amdgpu_bo(struct ttm_buffer_object *bo) { if (bo->destroy == &amdgpu_bo_destroy) return true; return false; } /** * amdgpu_bo_placement_from_domain - set buffer's placement * @abo: &amdgpu_bo buffer object whose placement is to be set * @domain: requested domain * * Sets buffer's placement according to requested domain and the buffer's * flags. */ void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain) { struct amdgpu_device *adev = amdgpu_ttm_adev(abo->tbo.bdev); struct ttm_placement *placement = &abo->placement; struct ttm_place *places = abo->placements; u64 flags = abo->flags; u32 c = 0; if (domain & AMDGPU_GEM_DOMAIN_VRAM) { unsigned visible_pfn = adev->gmc.visible_vram_size >> PAGE_SHIFT; places[c].fpfn = 0; places[c].lpfn = 0; places[c].flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED | TTM_PL_FLAG_VRAM; if (flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) places[c].lpfn = visible_pfn; else places[c].flags |= TTM_PL_FLAG_TOPDOWN; if (flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS) places[c].flags |= TTM_PL_FLAG_CONTIGUOUS; c++; } 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; 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; 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; 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; 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; c++; } if (!c) { places[c].fpfn = 0; places[c].lpfn = 0; places[c].flags = TTM_PL_MASK_CACHING | TTM_PL_FLAG_SYSTEM; c++; } BUG_ON(c >= AMDGPU_BO_MAX_PLACEMENTS); placement->num_placement = c; placement->placement = places; placement->num_busy_placement = c; placement->busy_placement = places; } /** * amdgpu_bo_create_reserved - create reserved BO for kernel use * * @adev: amdgpu device object * @size: size for the new BO * @align: alignment for the new BO * @domain: where to place it * @bo_ptr: used to initialize BOs in structures * @gpu_addr: GPU addr of the pinned BO * @cpu_addr: optional CPU address mapping * * Allocates and pins a BO for kernel internal use, and returns it still * reserved. * * Note: For bo_ptr new BO is only created if bo_ptr points to NULL. * * Returns: * 0 on success, negative error code otherwise. */ int amdgpu_bo_create_reserved(struct amdgpu_device *adev, unsigned long size, int align, u32 domain, struct amdgpu_bo **bo_ptr, u64 *gpu_addr, void **cpu_addr) { struct amdgpu_bo_param bp; bool free = false; int r; if (!size) { amdgpu_bo_unref(bo_ptr); return 0; } memset(&bp, 0, sizeof(bp)); bp.size = size; bp.byte_align = align; bp.domain = domain; bp.flags = cpu_addr ? AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED : AMDGPU_GEM_CREATE_NO_CPU_ACCESS; bp.flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; bp.type = ttm_bo_type_kernel; bp.resv = NULL; if (!*bo_ptr) { r = amdgpu_bo_create(adev, &bp, bo_ptr); if (r) { dev_err(adev->dev, "(%d) failed to allocate kernel bo\n", r); return r; } free = true; } r = amdgpu_bo_reserve(*bo_ptr, false); if (r) { dev_err(adev->dev, "(%d) failed to reserve kernel bo\n", r); goto error_free; } r = amdgpu_bo_pin(*bo_ptr, domain); if (r) { dev_err(adev->dev, "(%d) kernel bo pin failed\n", r); goto error_unreserve; } r = amdgpu_ttm_alloc_gart(&(*bo_ptr)->tbo); if (r) { dev_err(adev->dev, "%p bind failed\n", *bo_ptr); goto error_unpin; } if (gpu_addr) *gpu_addr = amdgpu_bo_gpu_offset(*bo_ptr); if (cpu_addr) { r = amdgpu_bo_kmap(*bo_ptr, cpu_addr); if (r) { dev_err(adev->dev, "(%d) kernel bo map failed\n", r); goto error_unpin; } } return 0; error_unpin: amdgpu_bo_unpin(*bo_ptr); error_unreserve: amdgpu_bo_unreserve(*bo_ptr); error_free: if (free) amdgpu_bo_unref(bo_ptr); return r; } /** * amdgpu_bo_create_kernel - create BO for kernel use * * @adev: amdgpu device object * @size: size for the new BO * @align: alignment for the new BO * @domain: where to place it * @bo_ptr: used to initialize BOs in structures * @gpu_addr: GPU addr of the pinned BO * @cpu_addr: optional CPU address mapping * * Allocates and pins a BO for kernel internal use. * * Note: For bo_ptr new BO is only created if bo_ptr points to NULL. * * Returns: * 0 on success, negative error code otherwise. */ int amdgpu_bo_create_kernel(struct amdgpu_device *adev, unsigned long size, int align, u32 domain, struct amdgpu_bo **bo_ptr, u64 *gpu_addr, void **cpu_addr) { int r; r = amdgpu_bo_create_reserved(adev, size, align, domain, bo_ptr, gpu_addr, cpu_addr); if (r) return r; if (*bo_ptr) amdgpu_bo_unreserve(*bo_ptr); return 0; } /** * amdgpu_bo_create_kernel_at - create BO for kernel use at specific location * * @adev: amdgpu device object * @offset: offset of the BO * @size: size of the BO * @domain: where to place it * @bo_ptr: used to initialize BOs in structures * @cpu_addr: optional CPU address mapping * * Creates a kernel BO at a specific offset in the address space of the domain. * * Returns: * 0 on success, negative error code otherwise. */ int amdgpu_bo_create_kernel_at(struct amdgpu_device *adev, uint64_t offset, uint64_t size, uint32_t domain, struct amdgpu_bo **bo_ptr, void **cpu_addr) { struct ttm_operation_ctx ctx = { false, false }; unsigned int i; int r; offset &= PAGE_MASK; size = ALIGN(size, PAGE_SIZE); r = amdgpu_bo_create_reserved(adev, size, PAGE_SIZE, domain, bo_ptr, NULL, cpu_addr); if (r) return r; /* * Remove the original mem node and create a new one at the request * position. */ if (cpu_addr) amdgpu_bo_kunmap(*bo_ptr); ttm_bo_mem_put(&(*bo_ptr)->tbo, &(*bo_ptr)->tbo.mem); 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); if (r) goto error; if (cpu_addr) { r = amdgpu_bo_kmap(*bo_ptr, cpu_addr); if (r) goto error; } amdgpu_bo_unreserve(*bo_ptr); return 0; error: amdgpu_bo_unreserve(*bo_ptr); amdgpu_bo_unref(bo_ptr); return r; } /** * amdgpu_bo_free_kernel - free BO for kernel use * * @bo: amdgpu BO to free * @gpu_addr: pointer to where the BO's GPU memory space address was stored * @cpu_addr: pointer to where the BO's CPU memory space address was stored * * unmaps and unpin a BO for kernel internal use. */ void amdgpu_bo_free_kernel(struct amdgpu_bo **bo, u64 *gpu_addr, void **cpu_addr) { if (*bo == NULL) return; if (likely(amdgpu_bo_reserve(*bo, true) == 0)) { if (cpu_addr) amdgpu_bo_kunmap(*bo); amdgpu_bo_unpin(*bo); amdgpu_bo_unreserve(*bo); } amdgpu_bo_unref(bo); if (gpu_addr) *gpu_addr = 0; if (cpu_addr) *cpu_addr = NULL; } /* Validate bo size is bit bigger then the request domain */ static bool amdgpu_bo_validate_size(struct amdgpu_device *adev, unsigned long size, u32 domain) { struct ttm_mem_type_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]; if (size < (man->size << PAGE_SHIFT)) return true; else goto fail; } if (domain & AMDGPU_GEM_DOMAIN_VRAM) { man = &adev->mman.bdev.man[TTM_PL_VRAM]; if (size < (man->size << PAGE_SHIFT)) return true; else goto fail; } /* TODO add more domains checks, such as AMDGPU_GEM_DOMAIN_CPU */ return true; fail: DRM_DEBUG("BO size %lu > total memory in domain: %llu\n", size, man->size << PAGE_SHIFT); return false; } bool amdgpu_bo_support_uswc(u64 bo_flags) { #ifdef CONFIG_X86_32 /* XXX: Write-combined CPU mappings of GTT seem broken on 32-bit * See https://bugs.freedesktop.org/show_bug.cgi?id=84627 */ return false; #elif defined(CONFIG_X86) && !defined(CONFIG_X86_PAT) /* Don't try to enable write-combining when it can't work, or things * may be slow * See https://bugs.freedesktop.org/show_bug.cgi?id=88758 */ #ifndef CONFIG_COMPILE_TEST #warning Please enable CONFIG_MTRR and CONFIG_X86_PAT for better performance \ thanks to write-combining #endif if (bo_flags & AMDGPU_GEM_CREATE_CPU_GTT_USWC) DRM_INFO_ONCE("Please enable CONFIG_MTRR and CONFIG_X86_PAT for " "better performance thanks to write-combining\n"); return false; #else /* For architectures that don't support WC memory, * mask out the WC flag from the BO */ if (!drm_arch_can_wc_memory()) return false; return true; #endif } static int amdgpu_bo_do_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 }; 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. */ if (bp->domain & (AMDGPU_GEM_DOMAIN_GWS | AMDGPU_GEM_DOMAIN_OA)) { /* GWS and OA don't need any alignment. */ page_align = bp->byte_align; size <<= PAGE_SHIFT; } else if (bp->domain & AMDGPU_GEM_DOMAIN_GDS) { /* Both size and alignment must be a multiple of 4. */ page_align = ALIGN(bp->byte_align, 4); size = ALIGN(size, 4) << PAGE_SHIFT; } else { /* Memory should be aligned at least to a page size. */ page_align = ALIGN(bp->byte_align, PAGE_SIZE) >> PAGE_SHIFT; size = ALIGN(size, PAGE_SIZE); } if (!amdgpu_bo_validate_size(adev, size, bp->domain)) return -ENOMEM; *bo_ptr = NULL; acc_size = ttm_bo_dma_acc_size(&adev->mman.bdev, size, sizeof(struct amdgpu_bo)); bo = kzalloc(sizeof(struct amdgpu_bo), GFP_KERNEL); if (bo == NULL) return -ENOMEM; drm_gem_private_object_init(adev->ddev, &bo->tbo.base, size); INIT_LIST_HEAD(&bo->shadow_list); bo->vm_bo = NULL; bo->preferred_domains = bp->preferred_domain ? bp->preferred_domain : bp->domain; bo->allowed_domains = bo->preferred_domains; if (bp->type != ttm_bo_type_kernel && bo->allowed_domains == AMDGPU_GEM_DOMAIN_VRAM) bo->allowed_domains |= AMDGPU_GEM_DOMAIN_GTT; bo->flags = bp->flags; if (!amdgpu_bo_support_uswc(bo->flags)) bo->flags &= ~AMDGPU_GEM_CREATE_CPU_GTT_USWC; bo->tbo.bdev = &adev->mman.bdev; if (bp->domain & (AMDGPU_GEM_DOMAIN_GWS | AMDGPU_GEM_DOMAIN_OA | AMDGPU_GEM_DOMAIN_GDS)) amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_CPU); else amdgpu_bo_placement_from_domain(bo, bp->domain); if (bp->type == ttm_bo_type_kernel) bo->tbo.priority = 1; 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); 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) 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) { struct dma_fence *fence; r = amdgpu_fill_buffer(bo, 0, bo->tbo.base.resv, &fence); if (unlikely(r)) goto fail_unreserve; amdgpu_bo_fence(bo, fence, false); dma_fence_put(bo->tbo.moving); bo->tbo.moving = dma_fence_get(fence); dma_fence_put(fence); } if (!bp->resv) amdgpu_bo_unreserve(bo); *bo_ptr = bo; trace_amdgpu_bo_create(bo); /* Treat CPU_ACCESS_REQUIRED only as a hint if given by UMD */ if (bp->type == ttm_bo_type_device) bo->flags &= ~AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; return 0; fail_unreserve: if (!bp->resv) dma_resv_unlock(bo->tbo.base.resv); amdgpu_bo_unref(&bo); return r; } static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, unsigned long size, struct amdgpu_bo *bo) { struct amdgpu_bo_param bp; 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); } return r; } /** * 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; and if requested, also creates a * shadow object. * Shadow object is used to backup the original buffer object, and is always * in GTT. * * 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) { u64 flags = bp->flags; int r; bp->flags = bp->flags & ~AMDGPU_GEM_CREATE_SHADOW; r = amdgpu_bo_do_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); } return r; } /** * amdgpu_bo_validate - validate an &amdgpu_bo buffer object * @bo: pointer to the buffer object * * 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. * * Returns: * 0 for success or a negative error code on failure. */ int amdgpu_bo_validate(struct amdgpu_bo *bo) { struct ttm_operation_ctx ctx = { false, false }; uint32_t domain; int r; 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; } /** * amdgpu_bo_restore_shadow - restore an &amdgpu_bo shadow * * @shadow: &amdgpu_bo shadow to be restored * @fence: dma_fence associated with the operation * * Copies a buffer object's shadow content back to the object. * This is used for recovering a buffer from its shadow in case of a gpu * reset where vram context may be lost. * * Returns: * 0 for success or a negative error code on failure. */ int amdgpu_bo_restore_shadow(struct amdgpu_bo *shadow, struct dma_fence **fence) { struct amdgpu_device *adev = amdgpu_ttm_adev(shadow->tbo.bdev); struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring; uint64_t shadow_addr, parent_addr; shadow_addr = amdgpu_bo_gpu_offset(shadow); parent_addr = amdgpu_bo_gpu_offset(shadow->parent); return amdgpu_copy_buffer(ring, shadow_addr, parent_addr, amdgpu_bo_size(shadow), NULL, fence, true, false, false); } /** * amdgpu_bo_kmap - map an &amdgpu_bo buffer object * @bo: &amdgpu_bo buffer object to be mapped * @ptr: kernel virtual address to be returned * * Calls ttm_bo_kmap() to set up the kernel virtual mapping; calls * amdgpu_bo_kptr() to get the kernel virtual address. * * Returns: * 0 for success or a negative error code on failure. */ int amdgpu_bo_kmap(struct amdgpu_bo *bo, void **ptr) { void *kptr; long r; if (bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS) return -EPERM; kptr = amdgpu_bo_kptr(bo); if (kptr) { if (ptr) *ptr = kptr; return 0; } r = dma_resv_wait_timeout_rcu(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); if (r) return r; if (ptr) *ptr = amdgpu_bo_kptr(bo); return 0; } /** * amdgpu_bo_kptr - returns a kernel virtual address of the buffer object * @bo: &amdgpu_bo buffer object * * Calls ttm_kmap_obj_virtual() to get the kernel virtual address * * Returns: * the virtual address of a buffer object area. */ void *amdgpu_bo_kptr(struct amdgpu_bo *bo) { bool is_iomem; return ttm_kmap_obj_virtual(&bo->kmap, &is_iomem); } /** * amdgpu_bo_kunmap - unmap an &amdgpu_bo buffer object * @bo: &amdgpu_bo buffer object to be unmapped * * Unmaps a kernel map set up by amdgpu_bo_kmap(). */ void amdgpu_bo_kunmap(struct amdgpu_bo *bo) { if (bo->kmap.bo) ttm_bo_kunmap(&bo->kmap); } /** * amdgpu_bo_ref - reference an &amdgpu_bo buffer object * @bo: &amdgpu_bo buffer object * * References the contained &ttm_buffer_object. * * Returns: * a refcounted pointer to the &amdgpu_bo buffer object. */ struct amdgpu_bo *amdgpu_bo_ref(struct amdgpu_bo *bo) { if (bo == NULL) return NULL; ttm_bo_get(&bo->tbo); return bo; } /** * amdgpu_bo_unref - unreference an &amdgpu_bo buffer object * @bo: &amdgpu_bo buffer object * * Unreferences the contained &ttm_buffer_object and clear the pointer */ void amdgpu_bo_unref(struct amdgpu_bo **bo) { struct ttm_buffer_object *tbo; if ((*bo) == NULL) return; tbo = &((*bo)->tbo); ttm_bo_put(tbo); *bo = NULL; } /** * amdgpu_bo_pin_restricted - pin an &amdgpu_bo buffer object * @bo: &amdgpu_bo buffer object to be pinned * @domain: domain to be pinned to * @min_offset: the start of requested address range * @max_offset: the end of requested address range * * Pins the buffer object according to requested domain and address range. If * the memory is unbound gart memory, binds the pages into gart table. Adjusts * pin_count and pin_size accordingly. * * Pinning means to lock pages in memory along with keeping them at a fixed * offset. It is required when a buffer can not be moved, for example, when * a display buffer is being scanned out. * * Compared with amdgpu_bo_pin(), this function gives more flexibility on * where to pin a buffer if there are specific restrictions on where a buffer * must be located. * * Returns: * 0 for success or a negative error code on failure. */ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, u64 min_offset, u64 max_offset) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); struct ttm_operation_ctx ctx = { false, false }; int r, i; if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm)) return -EPERM; if (WARN_ON_ONCE(min_offset > max_offset)) return -EINVAL; /* A shared bo cannot be migrated to VRAM */ if (bo->prime_shared_count) { 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 (!(domain & amdgpu_mem_type_to_domain(mem_type))) return -EINVAL; bo->pin_count++; if (max_offset != 0) { u64 domain_start = amdgpu_ttm_domain_start(adev, mem_type); WARN_ON_ONCE(max_offset < (amdgpu_bo_gpu_offset(bo) - domain_start)); } return 0; } if (bo->tbo.base.import_attach) dma_buf_pin(bo->tbo.base.import_attach); bo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; /* force to pin into visible video ram */ if (!(bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS)) bo->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; amdgpu_bo_placement_from_domain(bo, domain); for (i = 0; i < bo->placement.num_placement; i++) { unsigned fpfn, lpfn; fpfn = min_offset >> PAGE_SHIFT; lpfn = max_offset >> PAGE_SHIFT; if (fpfn > bo->placements[i].fpfn) bo->placements[i].fpfn = fpfn; 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); if (unlikely(r)) { dev_err(adev->dev, "%p pin failed\n", bo); goto error; } bo->pin_count = 1; domain = amdgpu_mem_type_to_domain(bo->tbo.mem.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), &adev->visible_pin_size); } else if (domain == AMDGPU_GEM_DOMAIN_GTT) { atomic64_add(amdgpu_bo_size(bo), &adev->gart_pin_size); } error: return r; } /** * amdgpu_bo_pin - pin an &amdgpu_bo buffer object * @bo: &amdgpu_bo buffer object to be pinned * @domain: domain to be pinned to * * A simple wrapper to amdgpu_bo_pin_restricted(). * Provides a simpler API for buffers that do not have any strict restrictions * on where a buffer must be located. * * Returns: * 0 for success or a negative error code on failure. */ int amdgpu_bo_pin(struct amdgpu_bo *bo, u32 domain) { return amdgpu_bo_pin_restricted(bo, domain, 0, 0); } /** * amdgpu_bo_unpin - unpin an &amdgpu_bo buffer object * @bo: &amdgpu_bo buffer object to be unpinned * * Decreases the pin_count, and clears the flags if pin_count reaches 0. * Changes placement and pin size accordingly. * * Returns: * 0 for success or a negative error code on failure. */ int 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); if (bo->tbo.base.import_attach) dma_buf_unpin(bo->tbo.base.import_attach); 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); return r; } /** * 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; } #endif return ttm_bo_evict_mm(&adev->mman.bdev, TTM_PL_VRAM); } static const char *amdgpu_vram_names[] = { "UNKNOWN", "GDDR1", "DDR2", "GDDR3", "GDDR4", "GDDR5", "HBM", "DDR3", "DDR4", "GDDR6", }; /** * amdgpu_bo_init - initialize memory manager * @adev: amdgpu device object * * Calls amdgpu_ttm_init() to initialize amdgpu memory manager. * * Returns: * 0 for success or a negative error code on failure. */ 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); /* 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); DRM_INFO("RAM width %dbits %s\n", adev->gmc.vram_width, amdgpu_vram_names[adev->gmc.vram_type]); return amdgpu_ttm_init(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 * * Reverses amdgpu_bo_init() to tear down memory manager. */ 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); } /** * amdgpu_bo_set_tiling_flags - set tiling flags * @bo: &amdgpu_bo buffer object * @tiling_flags: new flags * * Sets buffer object's tiling flags with the new one. Used by GEM ioctl or * kernel driver to set the tiling flags on a buffer. * * Returns: * 0 for success or a negative error code on failure. */ int amdgpu_bo_set_tiling_flags(struct amdgpu_bo *bo, u64 tiling_flags) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); if (adev->family <= AMDGPU_FAMILY_CZ && AMDGPU_TILING_GET(tiling_flags, TILE_SPLIT) > 6) return -EINVAL; bo->tiling_flags = tiling_flags; return 0; } /** * amdgpu_bo_get_tiling_flags - get tiling flags * @bo: &amdgpu_bo buffer object * @tiling_flags: returned flags * * Gets buffer object's tiling flags. Used by GEM ioctl or kernel driver to * set the tiling flags on a buffer. */ void amdgpu_bo_get_tiling_flags(struct amdgpu_bo *bo, u64 *tiling_flags) { dma_resv_assert_held(bo->tbo.base.resv); if (tiling_flags) *tiling_flags = bo->tiling_flags; } /** * amdgpu_bo_set_metadata - set metadata * @bo: &amdgpu_bo buffer object * @metadata: new metadata * @metadata_size: size of the new metadata * @flags: flags of the new metadata * * Sets buffer object's metadata, its size and flags. * Used via GEM ioctl. * * Returns: * 0 for success or a negative error code on failure. */ int amdgpu_bo_set_metadata (struct amdgpu_bo *bo, void *metadata, uint32_t metadata_size, uint64_t flags) { void *buffer; if (!metadata_size) { if (bo->metadata_size) { kfree(bo->metadata); bo->metadata = NULL; bo->metadata_size = 0; } return 0; } if (metadata == NULL) return -EINVAL; buffer = kmemdup(metadata, metadata_size, GFP_KERNEL); if (buffer == NULL) return -ENOMEM; kfree(bo->metadata); bo->metadata_flags = flags; bo->metadata = buffer; bo->metadata_size = metadata_size; return 0; } /** * amdgpu_bo_get_metadata - get metadata * @bo: &amdgpu_bo buffer object * @buffer: returned metadata * @buffer_size: size of the buffer * @metadata_size: size of the returned metadata * @flags: flags of the returned metadata * * Gets buffer object's metadata, its size and flags. buffer_size shall not be * less than metadata_size. * Used via GEM ioctl. * * Returns: * 0 for success or a negative error code on failure. */ int amdgpu_bo_get_metadata(struct amdgpu_bo *bo, void *buffer, size_t buffer_size, uint32_t *metadata_size, uint64_t *flags) { if (!buffer && !metadata_size) return -EINVAL; if (buffer) { if (buffer_size < bo->metadata_size) return -EINVAL; if (bo->metadata_size) memcpy(buffer, bo->metadata, bo->metadata_size); } if (metadata_size) *metadata_size = bo->metadata_size; if (flags) *flags = bo->metadata_flags; return 0; } /** * amdgpu_bo_move_notify - notification about a memory move * @bo: pointer to a buffer object * @evict: if this move is evicting the buffer from the graphics address space * @new_mem: new information of the bufer object * * Marks the corresponding &amdgpu_bo buffer object as invalid, also performs * bookkeeping. * TTM driver callback which is called when ttm moves a buffer. */ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo, bool evict, struct ttm_mem_reg *new_mem) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev); struct amdgpu_bo *abo; struct ttm_mem_reg *old_mem = &bo->mem; if (!amdgpu_bo_is_amdgpu_bo(bo)) return; abo = ttm_to_amdgpu_bo(bo); amdgpu_vm_bo_invalidate(adev, abo, evict); amdgpu_bo_kunmap(abo); if (abo->tbo.base.dma_buf && !abo->tbo.base.import_attach && bo->mem.mem_type != TTM_PL_SYSTEM) dma_buf_move_notify(abo->tbo.base.dma_buf); /* remember the eviction */ if (evict) atomic64_inc(&adev->num_evictions); /* update statistics */ if (!new_mem) return; /* move_notify is called before move happens */ trace_amdgpu_bo_move(abo, new_mem->mem_type, old_mem->mem_type); } /** * amdgpu_bo_move_notify - notification about a BO being released * @bo: pointer to a buffer object * * Wipes VRAM buffers whose contents should not be leaked before the * memory is released. */ void amdgpu_bo_release_notify(struct ttm_buffer_object *bo) { struct dma_fence *fence = NULL; struct amdgpu_bo *abo; int r; if (!amdgpu_bo_is_amdgpu_bo(bo)) return; abo = ttm_to_amdgpu_bo(bo); if (abo->kfd_bo) amdgpu_amdkfd_unreserve_memory_limit(abo); /* 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->mem.mem_type != TTM_PL_VRAM || !bo->mem.mm_node || !(abo->flags & AMDGPU_GEM_CREATE_VRAM_WIPE_ON_RELEASE)) return; dma_resv_lock(bo->base.resv, NULL); r = amdgpu_fill_buffer(abo, AMDGPU_POISON, bo->base.resv, &fence); if (!WARN_ON(r)) { amdgpu_bo_fence(abo, fence, false); dma_fence_put(fence); } dma_resv_unlock(bo->base.resv); } /** * amdgpu_bo_fault_reserve_notify - notification about a memory fault * @bo: pointer to a buffer object * * Notifies the driver we are taking a fault on this BO and have reserved it, * also performs bookkeeping. * TTM driver callback for dealing with vm faults. * * Returns: * 0 for success or a negative error code on failure. */ int 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; 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) return 0; size = bo->mem.num_pages << PAGE_SHIFT; offset = bo->mem.start << PAGE_SHIFT; if ((offset + size) <= adev->gmc.visible_vram_size) return 0; /* Can't move a pinned BO to visible VRAM */ if (abo->pin_count > 0) return -EINVAL; /* hurrah the memory is not visible ! */ atomic64_inc(&adev->num_vram_cpu_page_faults); amdgpu_bo_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_VRAM | AMDGPU_GEM_DOMAIN_GTT); /* Avoid costly evictions; only set GTT as a busy placement */ abo->placement.num_busy_placement = 1; abo->placement.busy_placement = &abo->placements[1]; r = ttm_bo_validate(bo, &abo->placement, &ctx); if (unlikely(r != 0)) return r; offset = bo->mem.start << PAGE_SHIFT; /* this should never happen */ if (bo->mem.mem_type == TTM_PL_VRAM && (offset + size) > adev->gmc.visible_vram_size) return -EINVAL; return 0; } /** * amdgpu_bo_fence - add fence to buffer object * * @bo: buffer object in question * @fence: fence to add * @shared: true if fence should be added shared * */ void amdgpu_bo_fence(struct amdgpu_bo *bo, struct dma_fence *fence, bool shared) { struct dma_resv *resv = bo->tbo.base.resv; if (shared) dma_resv_add_shared_fence(resv, fence); else dma_resv_add_excl_fence(resv, fence); } /** * amdgpu_bo_sync_wait_resv - Wait for BO reservation fences * * @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_resv(struct amdgpu_device *adev, struct dma_resv *resv, enum amdgpu_sync_mode sync_mode, void *owner, bool intr) { struct amdgpu_sync sync; int r; amdgpu_sync_create(&sync); 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 * * Note: object should either be pinned or reserved when calling this * function, it might be useful to add check for this for debugging. * * Returns: * current GPU offset of the object. */ u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo) { WARN_ON_ONCE(bo->tbo.mem.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->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS)); return amdgpu_bo_gpu_offset_no_check(bo); } /** * 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.mem.start << PAGE_SHIFT) + amdgpu_ttm_domain_start(adev, bo->tbo.mem.mem_type); return amdgpu_gmc_sign_extend(offset); } /** * amdgpu_bo_get_preferred_pin_domain - get preferred domain for scanout * @adev: amdgpu device object * @domain: allowed :ref:`memory domains ` * * Returns: * Which of the allowed domains is preferred for pinning the BO for scanout. */ uint32_t amdgpu_bo_get_preferred_pin_domain(struct amdgpu_device *adev, uint32_t domain) { if (domain == (AMDGPU_GEM_DOMAIN_VRAM | AMDGPU_GEM_DOMAIN_GTT)) { domain = AMDGPU_GEM_DOMAIN_VRAM; if (adev->gmc.real_vram_size <= AMDGPU_SG_THRESHOLD) domain = AMDGPU_GEM_DOMAIN_GTT; } return domain; }