[PATCH 2/2] drm/xe/vm: move xe_svm_init() earlier
Matthew Brost
matthew.brost at intel.com
Thu May 15 00:08:28 UTC 2025
On Wed, May 14, 2025 at 04:24:26PM +0100, Matthew Auld wrote:
> In xe_vm_close_and_put() we need to be able to call xe_svm_fini(),
> however during vm creation we can call this on the error path, before
> having actually initialised the svm state, leading to various splats
> followed by a fatal NPD.
>
> Fixes: 6fd979c2f331 ("drm/xe: Add SVM init / close / fini to faulting VMs")
> Link: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/4967
> Signed-off-by: Matthew Auld <matthew.auld at intel.com>
> Cc: Matthew Brost <matthew.brost at intel.com>
Reviewed-by: Matthew Brost <matthew.brost at intel.com>
> ---
> drivers/gpu/drm/xe/xe_vm.c | 19 ++++++++++++-------
> 1 file changed, 12 insertions(+), 7 deletions(-)
>
> diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c
> index 168756fb140b..7140d8856bad 100644
> --- a/drivers/gpu/drm/xe/xe_vm.c
> +++ b/drivers/gpu/drm/xe/xe_vm.c
> @@ -1709,10 +1709,16 @@ struct xe_vm *xe_vm_create(struct xe_device *xe, u32 flags)
> xe_pm_runtime_get_noresume(xe);
> }
>
> + if (flags & XE_VM_FLAG_FAULT_MODE) {
> + err = xe_svm_init(vm);
> + if (err)
> + goto err_no_resv;
> + }
> +
> vm_resv_obj = drm_gpuvm_resv_object_alloc(&xe->drm);
> if (!vm_resv_obj) {
> err = -ENOMEM;
> - goto err_no_resv;
> + goto err_svm_fini;
> }
>
> drm_gpuvm_init(&vm->gpuvm, "Xe VM", DRM_GPUVM_RESV_PROTECTED, &xe->drm,
> @@ -1783,12 +1789,6 @@ struct xe_vm *xe_vm_create(struct xe_device *xe, u32 flags)
> }
> }
>
> - if (flags & XE_VM_FLAG_FAULT_MODE) {
> - err = xe_svm_init(vm);
> - if (err)
> - goto err_close;
> - }
> -
> if (number_tiles > 1)
> vm->composite_fence_ctx = dma_fence_context_alloc(1);
>
> @@ -1802,6 +1802,11 @@ struct xe_vm *xe_vm_create(struct xe_device *xe, u32 flags)
> xe_vm_close_and_put(vm);
> return ERR_PTR(err);
>
> +err_svm_fini:
> + if (flags & XE_VM_FLAG_FAULT_MODE) {
> + vm->size = 0; /* close the vm */
> + xe_svm_fini(vm);
> + }
> err_no_resv:
> mutex_destroy(&vm->snap_mutex);
> for_each_tile(tile, xe, id)
> --
> 2.49.0
>
More information about the Intel-xe
mailing list