drm/amdkfd: Reject zero-sized AQL queue allocations after size halving

KFD_IOC_ALLOC_MEMORY_OF_GPU with flag
KFD_IOC_ALLOC_MEM_FLAGS_AQL_QUEUE_MEM and size=1 triggers the AQL
wraparound workaround (size >>= 1), reducing size to 0. The resulting
zero passes through PAGE_ALIGN(0) = 0 without validation, bypassing the
per-process VRAM quota check in reserve_mem_limit()
(vram_used + 0 > vram_available is always false).

The fix adds post-halving zero-size validation in the primary
allocation path (amdgpu_amdkfd_gpuvm.c). The check happens after size
halving but before reserve_mem_limit(), and uses err_alignment_size
error path to properly clean up the allocated kgd_mem structure and
mutex.

Cc: stable@vger.kernel.org
Signed-off-by: Sunday Clement <Sunday.Clement@amd.com>
Reviewed-by: Alex Deucher <Alexander.Deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
This commit is contained in:
Sunday Clement
2026-08-06 10:59:34 -04:00
committed by Alex Deucher
parent 2ee9836545
commit 40ba09e111
2 changed files with 9 additions and 1 deletions

View File

@@ -1795,6 +1795,12 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
size >>= 1;
aligned_size = PAGE_ALIGN(size);
/* reject AQL queue with size < 2 */
if (!aligned_size) {
ret = -EINVAL;
goto err_alignment_size;
}
(*mem)->alloc_flags = flags;
amdgpu_sync_create(&(*mem)->sync);
@@ -1886,6 +1892,7 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
amdgpu_amdkfd_unreserve_mem_limit(adev, aligned_size, flags, xcp_id);
err_reserve_limit:
amdgpu_sync_free(&(*mem)->sync);
err_alignment_size:
mutex_destroy(&(*mem)->lock);
if (gobj)
drm_gem_object_put(gobj);

View File

@@ -1200,7 +1200,8 @@ static int kfd_ioctl_alloc_memory_of_gpu(struct file *filep,
if (flags & KFD_IOC_ALLOC_MEM_FLAGS_AQL_QUEUE_MEM)
size >>= 1;
atomic64_add(PAGE_ALIGN(size), &pdd->vram_usage);
size = PAGE_ALIGN(size);
atomic64_add(size, &pdd->vram_usage);
}
mutex_unlock(&p->mutex);