diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c')
-rw-r--r-- | drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 66 |
1 files changed, 50 insertions, 16 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c index 08997092e7f7..cbd593f7d553 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c @@ -40,10 +40,10 @@ #define AMDGPU_USERPTR_RESTORE_DELAY_MS 1 /* - * Align VRAM allocations to 2MB to avoid fragmentation caused by 4K allocations in the tail 2MB + * Align VRAM availability to 2MB to avoid fragmentation caused by 4K allocations in the tail 2MB * BO chunk */ -#define VRAM_ALLOCATION_ALIGN (1 << 21) +#define VRAM_AVAILABLITY_ALIGN (1 << 21) /* Impose limit on how much memory KFD can use */ static struct { @@ -129,7 +129,7 @@ void amdgpu_amdkfd_reserve_system_mem(uint64_t size) * * Return: returns -ENOMEM in case of error, ZERO otherwise */ -static int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev, +int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev, uint64_t size, u32 alloc_flag) { uint64_t reserved_for_pt = @@ -149,7 +149,7 @@ static int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev, * to avoid fragmentation caused by 4K allocations in the tail * 2M BO chunk. */ - vram_needed = ALIGN(size, VRAM_ALLOCATION_ALIGN); + vram_needed = size; } else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_USERPTR) { system_mem_needed = size; } else if (!(alloc_flag & @@ -169,7 +169,7 @@ static int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev, kfd_mem_limit.max_system_mem_limit && !no_system_mem_limit) || (kfd_mem_limit.ttm_mem_used + ttm_mem_needed > kfd_mem_limit.max_ttm_mem_limit) || - (adev->kfd.vram_used + vram_needed > + (adev && adev->kfd.vram_used + vram_needed > adev->gmc.real_vram_size - atomic64_read(&adev->vram_pin_size) - reserved_for_pt)) { @@ -180,7 +180,12 @@ static int amdgpu_amdkfd_reserve_mem_limit(struct amdgpu_device *adev, /* Update memory accounting by decreasing available system * memory, TTM memory and GPU memory as computed above */ - adev->kfd.vram_used += vram_needed; + WARN_ONCE(vram_needed && !adev, + "adev reference can't be null when vram is used"); + if (adev) { + adev->kfd.vram_used += vram_needed; + adev->kfd.vram_used_aligned += ALIGN(vram_needed, VRAM_AVAILABLITY_ALIGN); + } kfd_mem_limit.system_mem_used += system_mem_needed; kfd_mem_limit.ttm_mem_used += ttm_mem_needed; @@ -189,7 +194,7 @@ release: return ret; } -static void unreserve_mem_limit(struct amdgpu_device *adev, +void amdgpu_amdkfd_unreserve_mem_limit(struct amdgpu_device *adev, uint64_t size, u32 alloc_flag) { spin_lock(&kfd_mem_limit.mem_limit_lock); @@ -198,7 +203,12 @@ static void unreserve_mem_limit(struct amdgpu_device *adev, kfd_mem_limit.system_mem_used -= size; kfd_mem_limit.ttm_mem_used -= size; } else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_VRAM) { - adev->kfd.vram_used -= ALIGN(size, VRAM_ALLOCATION_ALIGN); + WARN_ONCE(!adev, + "adev reference can't be null when alloc mem flags vram is set"); + if (adev) { + adev->kfd.vram_used -= size; + adev->kfd.vram_used_aligned -= ALIGN(size, VRAM_AVAILABLITY_ALIGN); + } } else if (alloc_flag & KFD_IOC_ALLOC_MEM_FLAGS_USERPTR) { kfd_mem_limit.system_mem_used -= size; } else if (!(alloc_flag & @@ -207,8 +217,7 @@ static void unreserve_mem_limit(struct amdgpu_device *adev, pr_err("%s: Invalid BO type %#x\n", __func__, alloc_flag); goto release; } - - WARN_ONCE(adev->kfd.vram_used < 0, + WARN_ONCE(adev && adev->kfd.vram_used < 0, "KFD VRAM memory accounting unbalanced"); WARN_ONCE(kfd_mem_limit.ttm_mem_used < 0, "KFD TTM memory accounting unbalanced"); @@ -225,7 +234,7 @@ void amdgpu_amdkfd_release_notify(struct amdgpu_bo *bo) u32 alloc_flags = bo->kfd_bo->alloc_flags; u64 size = amdgpu_bo_size(bo); - unreserve_mem_limit(adev, size, alloc_flags); + amdgpu_amdkfd_unreserve_mem_limit(adev, size, alloc_flags); kfree(bo->kfd_bo); } @@ -773,7 +782,6 @@ kfd_mem_attach_dmabuf(struct amdgpu_device *adev, struct kgd_mem *mem, *bo = gem_to_amdgpu_bo(gobj); (*bo)->flags |= AMDGPU_GEM_CREATE_PREEMPTIBLE; - (*bo)->parent = amdgpu_bo_ref(mem->bo); return 0; } @@ -1604,15 +1612,14 @@ size_t amdgpu_amdkfd_get_available_memory(struct amdgpu_device *adev) uint64_t reserved_for_pt = ESTIMATE_PT_SIZE(amdgpu_amdkfd_total_mem_size); size_t available; - spin_lock(&kfd_mem_limit.mem_limit_lock); available = adev->gmc.real_vram_size - - adev->kfd.vram_used + - adev->kfd.vram_used_aligned - atomic64_read(&adev->vram_pin_size) - reserved_for_pt; spin_unlock(&kfd_mem_limit.mem_limit_lock); - return ALIGN_DOWN(available, VRAM_ALLOCATION_ALIGN); + return ALIGN_DOWN(available, VRAM_AVAILABLITY_ALIGN); } int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( @@ -1749,7 +1756,7 @@ err_node_allow: /* Don't unreserve system mem limit twice */ goto err_reserve_limit; err_bo_create: - unreserve_mem_limit(adev, size, flags); + amdgpu_amdkfd_unreserve_mem_limit(adev, size, flags); err_reserve_limit: mutex_destroy(&(*mem)->lock); if (gobj) @@ -1770,6 +1777,7 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu( { struct amdkfd_process_info *process_info = mem->process_info; unsigned long bo_size = mem->bo->tbo.base.size; + bool use_release_notifier = (mem->bo->kfd_bo == mem); struct kfd_mem_attachment *entry, *tmp; struct bo_vm_reservation_context ctx; struct ttm_validate_buffer *bo_list_entry; @@ -1861,6 +1869,13 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu( */ drm_gem_object_put(&mem->bo->tbo.base); + /* + * For kgd_mem allocated in amdgpu_amdkfd_gpuvm_import_dmabuf(), + * explicitly free it here. + */ + if (!use_release_notifier) + kfree(mem); + return ret; } @@ -2883,3 +2898,22 @@ bool amdgpu_amdkfd_bo_mapped_to_dev(struct amdgpu_device *adev, struct kgd_mem * } return false; } + +#if defined(CONFIG_DEBUG_FS) + +int kfd_debugfs_kfd_mem_limits(struct seq_file *m, void *data) +{ + + spin_lock(&kfd_mem_limit.mem_limit_lock); + seq_printf(m, "System mem used %lldM out of %lluM\n", + (kfd_mem_limit.system_mem_used >> 20), + (kfd_mem_limit.max_system_mem_limit >> 20)); + seq_printf(m, "TTM mem used %lldM out of %lluM\n", + (kfd_mem_limit.ttm_mem_used >> 20), + (kfd_mem_limit.max_ttm_mem_limit >> 20)); + spin_unlock(&kfd_mem_limit.mem_limit_lock); + + return 0; +} + +#endif |