drm/xe: Convert xe_dma_buf.c for exhaustive eviction

Convert dma-buf migration to XE_PL_TT and dma-buf import to
support exhaustive eviction, using xe_validation_guard().
It seems unlikely that the import would result in an -ENOMEM,
but convert import anyway for completeness.

The dma-buf map_attachment() functionality unfortunately doesn't
support passing a drm_exec, which means that foreign devices
validating a dma-buf that we exported will not, unless they are
xeKMD devices, participate in the exhaustive eviction scheme.

v2:
- Avoid gotos from within xe_validation_guard(). (Matt Brost)
- Adapt to signature change of xe_validation_guard(). (Matt Brost)
- Remove an unneeded (void)ret. (Matt Brost)
- Fix up an error path.

Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Brost <matthew.brost@intel.com>
Link: https://lore.kernel.org/r/20250908101246.65025-9-thomas.hellstrom@linux.intel.com

1 files changed, 41 insertions, 20 deletions

diff --git a/drivers/gpu/drm/xe/xe_dma_buf.c b/drivers/gpu/drm/xe/xe_dma_buf.c
index 7c2793265561..2e5b4b37118f 100644
--- a/drivers/gpu/drm/xe/xe_dma_buf.c
+++ b/drivers/gpu/drm/xe/xe_dma_buf.c
@@ -163,16 +163,26 @@ static int xe_dma_buf_begin_cpu_access(struct dma_buf *dma_buf,
 	struct xe_bo *bo = gem_to_xe_bo(obj);
 	bool reads =  (direction == DMA_BIDIRECTIONAL ||
 		       direction == DMA_FROM_DEVICE);
-	struct drm_exec *exec = XE_VALIDATION_UNIMPLEMENTED;
+	struct xe_validation_ctx ctx;
+	struct drm_exec exec;
+	int ret = 0;
 
 	if (!reads)
 		return 0;
 
 	/* Can we do interruptible lock here? */
-	xe_bo_lock(bo, false);
-	(void)xe_bo_migrate(bo, XE_PL_TT, NULL, exec);
-	xe_bo_unlock(bo);
+	xe_validation_guard(&ctx, &xe_bo_device(bo)->val, &exec, (struct xe_val_flags) {}, ret) {
+		ret = drm_exec_lock_obj(&exec, &bo->ttm.base);
+		drm_exec_retry_on_contention(&exec);
+		if (ret)
+			break;
+
+		ret = xe_bo_migrate(bo, XE_PL_TT, NULL, &exec);
+		drm_exec_retry_on_contention(&exec);
+		xe_validation_retry_on_oom(&ctx, &ret);
+	}
 
+	/* If we failed, cpu-access takes place in current placement. */
 	return 0;
 }
 
@@ -211,25 +221,36 @@ xe_dma_buf_init_obj(struct drm_device *dev, struct xe_bo *storage,
 {
 	struct dma_resv *resv = dma_buf->resv;
 	struct xe_device *xe = to_xe_device(dev);
-	struct drm_exec *exec = XE_VALIDATION_UNIMPLEMENTED;
+	struct xe_validation_ctx ctx;
+	struct drm_gem_object *dummy_obj;
+	struct drm_exec exec;
 	struct xe_bo *bo;
-	int ret;
-
-	dma_resv_lock(resv, NULL);
-	bo = ___xe_bo_create_locked(xe, storage, NULL, resv, NULL, dma_buf->size,
-				    0, /* Will require 1way or 2way for vm_bind */
-				    ttm_bo_type_sg, XE_BO_FLAG_SYSTEM, exec);
-	if (IS_ERR(bo)) {
-		ret = PTR_ERR(bo);
-		goto error;
+	int ret = 0;
+
+	dummy_obj = drm_gpuvm_resv_object_alloc(&xe->drm);
+	if (!dummy_obj)
+		return ERR_PTR(-ENOMEM);
+
+	dummy_obj->resv = resv;
+	xe_validation_guard(&ctx, &xe->val, &exec, (struct xe_val_flags) {}, ret) {
+		ret = drm_exec_lock_obj(&exec, dummy_obj);
+		drm_exec_retry_on_contention(&exec);
+		if (ret)
+			break;
+
+		bo = ___xe_bo_create_locked(xe, storage, NULL, resv, NULL, dma_buf->size,
+					    0, /* Will require 1way or 2way for vm_bind */
+					    ttm_bo_type_sg, XE_BO_FLAG_SYSTEM, &exec);
+		drm_exec_retry_on_contention(&exec);
+		if (IS_ERR(bo)) {
+			ret = PTR_ERR(bo);
+			xe_validation_retry_on_oom(&ctx, &ret);
+			break;
+		}
 	}
-	dma_resv_unlock(resv);
-
-	return &bo->ttm.base;
+	drm_gem_object_put(dummy_obj);
 
-error:
-	dma_resv_unlock(resv);
-	return ERR_PTR(ret);
+	return ret ? ERR_PTR(ret) : &bo->ttm.base;
 }
 
 static void xe_dma_buf_move_notify(struct dma_buf_attachment *attach)