[PATCH i-g-t] tests/intel/xe_vm: fix xe_bb_size() conversion

[Zbigniew Kempczyński]

On Thu, Feb 01, 2024 at 10:06:37AM +0000, Matthew Auld wrote:
> The large* tests need to be able to partition the bo_size without
> breaking any alignment restrictions, and adding on the pre-fetch size
> breaks that.  Rather keep the passed in bo_size as-is and add the
> prefetch size as hidden padding at the end.
> 
> Signed-off-by: Matthew Auld <matthew.auld at intel.com>
> Cc: Zbigniew Kempczyński <zbigniew.kempczynski at intel.com>
> ---
>  tests/intel/xe_vm.c | 22 ++++++++++++----------
>  1 file changed, 12 insertions(+), 10 deletions(-)
> 
> diff --git a/tests/intel/xe_vm.c b/tests/intel/xe_vm.c
> index 67276b220..fe667e64d 100644
> --- a/tests/intel/xe_vm.c
> +++ b/tests/intel/xe_vm.c
> @@ -954,6 +954,7 @@ test_large_binds(int fd, struct drm_xe_engine_class_instance *eci,
>  		.num_syncs = 2,
>  		.syncs = to_user_pointer(sync),
>  	};
> +	size_t bo_size_prefetch, padding;
>  	uint64_t addr = 0x1ull << 30, base_addr = 0x1ull << 30;
>  	uint32_t vm;
>  	uint32_t exec_queues[MAX_N_EXEC_QUEUES];
> @@ -975,20 +976,21 @@ test_large_binds(int fd, struct drm_xe_engine_class_instance *eci,
>  	igt_assert(n_exec_queues <= MAX_N_EXEC_QUEUES);
>  	vm = xe_vm_create(fd, 0, 0);
>  
> -	bo_size = xe_bb_size(fd, bo_size);
> -
>  	if (flags & LARGE_BIND_FLAG_USERPTR) {
> -		map = aligned_alloc(xe_get_default_alignment(fd), bo_size);
> +		bo_size_prefetch = xe_bb_size(fd, bo_size);
> +		map = aligned_alloc(xe_get_default_alignment(fd), bo_size_prefetch);
>  		igt_assert(map);
>  	} else {
>  		igt_skip_on(xe_visible_vram_size(fd, 0) && bo_size >
>  			    xe_visible_vram_size(fd, 0));
>  
> -		bo = xe_bo_create(fd, vm, bo_size,
> +		bo_size_prefetch = xe_bb_size(fd, bo_size);
> +		bo = xe_bo_create(fd, vm, bo_size_prefetch,
>  				  vram_if_possible(fd, eci->gt_id),
>  				  DRM_XE_GEM_CREATE_FLAG_NEEDS_VISIBLE_VRAM);
>  		map = xe_bo_map(fd, bo, bo_size);
>  	}
> +	padding = bo_size_prefetch - bo_size;
>  
>  	for (i = 0; i < n_exec_queues; i++) {
>  		exec_queues[i] = xe_exec_queue_create(fd, vm, eci, 0);
> @@ -1001,19 +1003,19 @@ test_large_binds(int fd, struct drm_xe_engine_class_instance *eci,
>  			xe_vm_bind_userptr_async(fd, vm, 0, to_user_pointer(map),
>  						 addr, bo_size / 2, NULL, 0);
>  			xe_vm_bind_userptr_async(fd, vm, 0, to_user_pointer(map) + bo_size / 2,
> -						 addr + bo_size / 2, bo_size / 2,
> +						 addr + bo_size / 2, bo_size / 2 + padding,
>  						 sync, 1);
>  		} else {
>  			xe_vm_bind_userptr_async(fd, vm, 0, to_user_pointer(map),
> -						 addr, bo_size, sync, 1);
> +						 addr, bo_size + padding, sync, 1);
>  		}
>  	} else {
>  		if (flags & LARGE_BIND_FLAG_SPLIT) {
>  			xe_vm_bind_async(fd, vm, 0, bo, 0, addr, bo_size / 2, NULL, 0);
>  			xe_vm_bind_async(fd, vm, 0, bo, bo_size / 2, addr + bo_size / 2,
> -					 bo_size / 2, sync, 1);
> +					 bo_size / 2 + padding, sync, 1);
>  		} else {
> -			xe_vm_bind_async(fd, vm, 0, bo, 0, addr, bo_size, sync, 1);
> +			xe_vm_bind_async(fd, vm, 0, bo, 0, addr, bo_size + padding, sync, 1);
>  		}
>  	}
>  
> @@ -1061,9 +1063,9 @@ test_large_binds(int fd, struct drm_xe_engine_class_instance *eci,
>  		xe_vm_unbind_async(fd, vm, 0, 0, base_addr,
>  				   bo_size / 2, NULL, 0);
>  		xe_vm_unbind_async(fd, vm, 0, 0, base_addr + bo_size / 2,
> -				   bo_size / 2, sync, 1);
> +				   bo_size / 2 + padding, sync, 1);
>  	} else {
> -		xe_vm_unbind_async(fd, vm, 0, 0, base_addr, bo_size,
> +		xe_vm_unbind_async(fd, vm, 0, 0, base_addr, bo_size + padding,
>  				   sync, 1);
>  	}
>  	igt_assert(syncobj_wait(fd, &sync[0].handle, 1, INT64_MAX, 0, NULL));
> -- 
> 2.43.0
> 

LGTM:

Reviewed-by: Zbigniew Kempczyński <zbigniew.kempczynski at intel.com>

--
Zbigniew