6.12-stable review patch. If anyone has any objections, please let me know.
------------------
From: Thomas Hellström thomas.hellstrom@linux.intel.com
commit fe3ccd24138fd391ae8e32289d492c85f67770fc upstream.
When imported dma-bufs are destroyed, TTM is not fully individualizing the dma-resv, but it *is* copying the fences that need to be waited for before declaring idle. So in the case where the bo->resv != bo->_resv we can still drop the preempt-fences, but make sure we do that on bo->_resv which contains the fence-pointer copy.
In the case where the copying fails, bo->_resv will typically not contain any fences pointers at all, so there will be nothing to drop. In that case, TTM would have ensured all fences that would have been copied are signaled, including any remaining preempt fences.
Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs") Fixes: fa0af721bd1f ("drm/ttm: test private resv obj on release/destroy") Cc: Matthew Brost matthew.brost@intel.com Cc: stable@vger.kernel.org # v6.16+ Signed-off-by: Thomas Hellström thomas.hellstrom@linux.intel.com Tested-by: Matthew Brost matthew.brost@intel.com Reviewed-by: Matthew Brost matthew.brost@intel.com Link: https://patch.msgid.link/20251217093441.5073-1-thomas.hellstrom@linux.intel.... (cherry picked from commit 425fe550fb513b567bd6d01f397d274092a9c274) Signed-off-by: Thomas Hellström thomas.hellstrom@linux.intel.com Signed-off-by: Greg Kroah-Hartman gregkh@linuxfoundation.org --- drivers/gpu/drm/xe/xe_bo.c | 15 ++++----------- 1 file changed, 4 insertions(+), 11 deletions(-)
--- a/drivers/gpu/drm/xe/xe_bo.c +++ b/drivers/gpu/drm/xe/xe_bo.c @@ -1041,7 +1041,7 @@ static bool xe_ttm_bo_lock_in_destructor * always succeed here, as long as we hold the lru lock. */ spin_lock(&ttm_bo->bdev->lru_lock); - locked = dma_resv_trylock(ttm_bo->base.resv); + locked = dma_resv_trylock(&ttm_bo->base._resv); spin_unlock(&ttm_bo->bdev->lru_lock); xe_assert(xe, locked);
@@ -1061,13 +1061,6 @@ static void xe_ttm_bo_release_notify(str bo = ttm_to_xe_bo(ttm_bo); xe_assert(xe_bo_device(bo), !(bo->created && kref_read(&ttm_bo->base.refcount)));
- /* - * Corner case where TTM fails to allocate memory and this BOs resv - * still points the VMs resv - */ - if (ttm_bo->base.resv != &ttm_bo->base._resv) - return; - if (!xe_ttm_bo_lock_in_destructor(ttm_bo)) return;
@@ -1077,14 +1070,14 @@ static void xe_ttm_bo_release_notify(str * TODO: Don't do this for external bos once we scrub them after * unbind. */ - dma_resv_for_each_fence(&cursor, ttm_bo->base.resv, + dma_resv_for_each_fence(&cursor, &ttm_bo->base._resv, DMA_RESV_USAGE_BOOKKEEP, fence) { if (xe_fence_is_xe_preempt(fence) && !dma_fence_is_signaled(fence)) { if (!replacement) replacement = dma_fence_get_stub();
- dma_resv_replace_fences(ttm_bo->base.resv, + dma_resv_replace_fences(&ttm_bo->base._resv, fence->context, replacement, DMA_RESV_USAGE_BOOKKEEP); @@ -1092,7 +1085,7 @@ static void xe_ttm_bo_release_notify(str } dma_fence_put(replacement);
- dma_resv_unlock(ttm_bo->base.resv); + dma_resv_unlock(&ttm_bo->base._resv); }
static void xe_ttm_bo_delete_mem_notify(struct ttm_buffer_object *ttm_bo)