[PATCH] drm/amdgpu: Fix error handling in amdgpu_amdkfd_gpuvm_free_memory_of_gpu

[Felix Kuehling]

Am 2022-06-09 um 13:17 schrieb Ramesh Errabolu:
> Following error conditions are fixed:
>      Ensure calls to delete a list element are safe
>      Unpin MMIO and DOORBELL BOs only after map count goes to zero
>      Print a warning message if unreserving VMs encounters an error
>
> Signed-off-by: Ramesh Errabolu <Ramesh.Errabolu at amd.com>
> ---
>   .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c  | 38 ++++++++++---------
>   1 file changed, 20 insertions(+), 18 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
> index a1de900ba677..78b3efecc2f9 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c
> @@ -1013,14 +1013,14 @@ static void add_kgd_mem_to_kfd_bo_list(struct kgd_mem *mem,
>   	mutex_unlock(&process_info->lock);
>   }
>   
> -static void remove_kgd_mem_from_kfd_bo_list(struct kgd_mem *mem,
> +static void remove_kgd_mem_from_validate_list(struct kgd_mem *mem,
>   		struct amdkfd_process_info *process_info)
>   {
>   	struct ttm_validate_buffer *bo_list_entry;
>   
>   	bo_list_entry = &mem->validate_list;
>   	mutex_lock(&process_info->lock);
> -	list_del(&bo_list_entry->head);
> +	list_del_init(&bo_list_entry->head);
>   	mutex_unlock(&process_info->lock);
>   }
>   
> @@ -1796,7 +1796,7 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu(
>   
>   allocate_init_user_pages_failed:
>   err_pin_bo:
> -	remove_kgd_mem_from_kfd_bo_list(*mem, avm->process_info);
> +	remove_kgd_mem_from_validate_list(*mem, avm->process_info);
>   	drm_vma_node_revoke(&gobj->vma_node, drm_priv);
>   err_node_allow:
>   	/* Don't unreserve system mem limit twice */
> @@ -1825,20 +1825,12 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu(
>   	unsigned long bo_size = mem->bo->tbo.base.size;
>   	struct kfd_mem_attachment *entry, *tmp;
>   	struct bo_vm_reservation_context ctx;
> -	struct ttm_validate_buffer *bo_list_entry;
>   	unsigned int mapped_to_gpu_memory;
>   	int ret;
>   	bool is_imported = false;
>   
>   	mutex_lock(&mem->lock);
>   
> -	/* Unpin MMIO/DOORBELL BO's that were pinned during allocation */
> -	if (mem->alloc_flags &
> -	    (KFD_IOC_ALLOC_MEM_FLAGS_DOORBELL |
> -	     KFD_IOC_ALLOC_MEM_FLAGS_MMIO_REMAP)) {
> -		amdgpu_amdkfd_gpuvm_unpin_bo(mem->bo);
> -	}
> -
>   	mapped_to_gpu_memory = mem->mapped_to_gpu_memory;
>   	is_imported = mem->is_imported;
>   	mutex_unlock(&mem->lock);
> @@ -1852,11 +1844,10 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu(
>   		return -EBUSY;
>   	}
>   
> -	/* Make sure restore workers don't access the BO any more */
> -	bo_list_entry = &mem->validate_list;
> -	mutex_lock(&process_info->lock);
> -	list_del(&bo_list_entry->head);
> -	mutex_unlock(&process_info->lock);
> +	/* Make sure restore workers don't access the BO any more
> +	 * Ensure removal of BO from validate list is reentrant safe

The comment about being reentrant safe belongs in 
remove_kgd_mem_from_validate_list. That said, "reentrant safe" is not 
the correct term here. See 
https://en.wikipedia.org/wiki/Reentrancy_(computing). It refers to 
functions that can run multiple times concurrently, or be interrupted in 
the middle. Neither of those are applicable here.

The correct term here is "idempotent". See 
https://en.wikipedia.org/wiki/Idempotence. It requires that the function 
can be called multiple times sequentially without changing the results 
beyond the first call.

With that fixed, the patch is

Reviewed-by: Felix Kuehling <Felix.Kuehling at amd.com>


> +	 */
> +	remove_kgd_mem_from_validate_list(mem, process_info);
>   
>   	/* No more MMU notifiers */
>   	amdgpu_mn_unregister(mem->bo);
> @@ -1878,7 +1869,18 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu(
>   	list_for_each_entry_safe(entry, tmp, &mem->attachments, list)
>   		kfd_mem_detach(entry);
>   
> +	/* Return success even in case of error */
>   	ret = unreserve_bo_and_vms(&ctx, false, false);
> +	if (unlikely(ret)) {
> +		WARN_ONCE(ret, "Error in unreserving BO and associated VMs");
> +		ret = 0;
> +	}
> +
> +	/* Unpin MMIO/DOORBELL BO's that were pinned during allocation */
> +	if (mem->alloc_flags &
> +	    (KFD_IOC_ALLOC_MEM_FLAGS_DOORBELL |
> +	     KFD_IOC_ALLOC_MEM_FLAGS_MMIO_REMAP))
> +		amdgpu_amdkfd_gpuvm_unpin_bo(mem->bo);
>   
>   	/* Free the sync object */
>   	amdgpu_sync_free(&mem->sync);
> @@ -2814,7 +2816,7 @@ int amdgpu_amdkfd_add_gws_to_process(void *info, void *gws, struct kgd_mem **mem
>   bo_reservation_failure:
>   	mutex_unlock(&(*mem)->process_info->lock);
>   	amdgpu_sync_free(&(*mem)->sync);
> -	remove_kgd_mem_from_kfd_bo_list(*mem, process_info);
> +	remove_kgd_mem_from_validate_list(*mem, process_info);
>   	amdgpu_bo_unref(&gws_bo);
>   	mutex_destroy(&(*mem)->lock);
>   	kfree(*mem);
> @@ -2832,7 +2834,7 @@ int amdgpu_amdkfd_remove_gws_from_process(void *info, void *mem)
>   	/* Remove BO from process's validate list so restore worker won't touch
>   	 * it anymore
>   	 */
> -	remove_kgd_mem_from_kfd_bo_list(kgd_mem, process_info);
> +	remove_kgd_mem_from_validate_list(kgd_mem, process_info);
>   
>   	ret = amdgpu_bo_reserve(gws_bo, false);
>   	if (unlikely(ret)) {