Intel-XE Archive on lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH] drm/ttm: Fix ttm_bo_swapout() infinite LRU walk on swapout failure
@ 2026-04-28  9:44 Thomas Hellström
  2026-04-28 10:43 ` ✗ CI.checkpatch: warning for " Patchwork
                   ` (3 more replies)
  0 siblings, 4 replies; 5+ messages in thread
From: Thomas Hellström @ 2026-04-28  9:44 UTC (permalink / raw)
  To: intel-xe
  Cc: Thomas Hellström, Jatin Kataria, Christian König,
	Matthew Brost, dri-devel, stable, Boqun Feng

When ttm_tt_swapout() fails, the current code calls
ttm_resource_add_bulk_move() followed by ttm_resource_move_to_lru_tail()
to restore the resource's bulk_move membership.

However, ttm_resource_move_to_lru_tail() places the resource at the tail
of the LRU list which, relative to the walk cursor's hitch node (placed
immediately after the resource when it was yielded), puts the resource
*in front of the* the hitch. The next list_for_each_entry_continue() from
the hitch finds the same resource again, causing an infinite loop.

Fix by deferring del_bulk_move to the success path only.

On the success path, TTM_TT_FLAG_SWAPPED has just been set by
ttm_tt_swapout() but the resource is still tracked in the bulk_move range,
so ttm_resource_del_bulk_move()'s !ttm_resource_unevictable() guard would
incorrectly skip the removal. Introduce
ttm_resource_del_bulk_move_unevictable() which bypasses that guard.

Reported-by: Jatin Kataria <jkataria@netflix.com>
Fixes: fc5d96670eb2 ("drm/ttm: Move swapped objects off the manager's LRU list")
Cc: Christian König <christian.koenig@amd.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Cc: <dri-devel@lists.freedesktop.org>
Cc: <stable@vger.kernel.org> # v6.13+
Assisted-by: GitHub_Copilot:claude-sonnet-4.6
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Tested-by: Boqun Feng <boqun@kernel.org>
---
 drivers/gpu/drm/ttm/ttm_bo.c       | 16 ++++++----------
 drivers/gpu/drm/ttm/ttm_resource.c | 13 +++++++++++++
 include/drm/ttm/ttm_resource.h     |  2 ++
 3 files changed, 21 insertions(+), 10 deletions(-)

diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c
index d85f0a37ac35..293401705542 100644
--- a/drivers/gpu/drm/ttm/ttm_bo.c
+++ b/drivers/gpu/drm/ttm/ttm_bo.c
@@ -1177,17 +1177,13 @@ ttm_bo_swapout_cb(struct ttm_lru_walk *walk, struct ttm_buffer_object *bo)
 		bdev->funcs->swap_notify(bo);
 
 	if (ttm_tt_is_populated(tt)) {
-		spin_lock(&bdev->lru_lock);
-		ttm_resource_del_bulk_move(bo->resource, bo);
-		spin_unlock(&bdev->lru_lock);
-
 		ret = ttm_tt_swapout(bdev, tt, swapout_walk->gfp_flags);
-
-		spin_lock(&bdev->lru_lock);
-		if (ret)
-			ttm_resource_add_bulk_move(bo->resource, bo);
-		ttm_resource_move_to_lru_tail(bo->resource);
-		spin_unlock(&bdev->lru_lock);
+		if (!ret) {
+			spin_lock(&bdev->lru_lock);
+			ttm_resource_del_bulk_move_unevictable(bo->resource, bo);
+			ttm_resource_move_to_lru_tail(bo->resource);
+			spin_unlock(&bdev->lru_lock);
+		}
 	}
 
 out:
diff --git a/drivers/gpu/drm/ttm/ttm_resource.c b/drivers/gpu/drm/ttm/ttm_resource.c
index 9f36631d48b6..0e5f1582f13d 100644
--- a/drivers/gpu/drm/ttm/ttm_resource.c
+++ b/drivers/gpu/drm/ttm/ttm_resource.c
@@ -292,6 +292,19 @@ void ttm_resource_del_bulk_move(struct ttm_resource *res,
 		ttm_lru_bulk_move_del(bo->bulk_move, res);
 }
 
+/*
+ * Remove a resource from its bulk_move, bypassing the unevictable check.
+ * Use only when the resource is known to still be tracked in the range despite
+ * the BO having just become unevictable; asserts that this is the case.
+ */
+void ttm_resource_del_bulk_move_unevictable(struct ttm_resource *res,
+					    struct ttm_buffer_object *bo)
+{
+	WARN_ON_ONCE(!ttm_resource_unevictable(res, bo));
+	if (bo->bulk_move)
+		ttm_lru_bulk_move_del(bo->bulk_move, res);
+}
+
 /* Move a resource to the LRU or bulk tail */
 void ttm_resource_move_to_lru_tail(struct ttm_resource *res)
 {
diff --git a/include/drm/ttm/ttm_resource.h b/include/drm/ttm/ttm_resource.h
index 33e80f30b8b8..a5d386583fb6 100644
--- a/include/drm/ttm/ttm_resource.h
+++ b/include/drm/ttm/ttm_resource.h
@@ -448,6 +448,8 @@ void ttm_resource_add_bulk_move(struct ttm_resource *res,
 				struct ttm_buffer_object *bo);
 void ttm_resource_del_bulk_move(struct ttm_resource *res,
 				struct ttm_buffer_object *bo);
+void ttm_resource_del_bulk_move_unevictable(struct ttm_resource *res,
+					    struct ttm_buffer_object *bo);
 void ttm_resource_move_to_lru_tail(struct ttm_resource *res);
 
 void ttm_resource_init(struct ttm_buffer_object *bo,
-- 
2.53.0


^ permalink raw reply related	[flat|nested] 5+ messages in thread
end of thread, other threads:[~2026-04-28 21:00 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-04-28  9:44 [PATCH] drm/ttm: Fix ttm_bo_swapout() infinite LRU walk on swapout failure Thomas Hellström
2026-04-28 10:43 ` ✗ CI.checkpatch: warning for " Patchwork
2026-04-28 10:46 ` ✓ CI.KUnit: success " Patchwork
2026-04-28 11:58 ` ✓ Xe.CI.BAT: " Patchwork
2026-04-28 21:00 ` ✗ Xe.CI.FULL: failure " Patchwork

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox