Intel-XE Archive on lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH] gpu/buddy: do not pass caller list to __alloc_contig_try_harder()
@ 2026-10-06 11:07 Matthew Auld
  2026-10-06 11:15 ` ✓ CI.KUnit: success for " Patchwork
                   ` (2 more replies)
  0 siblings, 3 replies; 4+ messages in thread
From: Matthew Auld @ 2026-10-06 11:07 UTC (permalink / raw)
  To: intel-xe
  Cc: dri-devel, Shashiko, Arunpravin Paneer Selvam, Joel Fernandes,
	stable

If the caller passes a non-empty blocks list, __alloc_contig_try_harder()
unconditionally frees it on failure or during retry.

Pass the local allocated list instead and only splice it onto the caller's
list upon success.

Assisted-by: LLM
Reported-by: Shashiko <sashiko-bot@kernel.org>
Closes: https://sashiko.dev/#/bug/linux-fa721daf-7333-48b5-a26b-624c7f5791d7
Fixes: 0a1844bf0b53 ("drm/buddy: Improve contiguous memory allocation")
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Cc: Joel Fernandes <joelagnelf@nvidia.com>
Cc: <stable@vger.kernel.org>
---
 drivers/gpu/buddy.c                | 20 ++++++++++++++------
 drivers/gpu/tests/gpu_buddy_test.c |  9 +++++++++
 2 files changed, 23 insertions(+), 6 deletions(-)

diff --git a/drivers/gpu/buddy.c b/drivers/gpu/buddy.c
index 2f2aaadafe35..a86f454fbb40 100644
--- a/drivers/gpu/buddy.c
+++ b/drivers/gpu/buddy.c
@@ -2022,10 +2022,16 @@ int gpu_buddy_alloc_blocks(struct gpu_buddy *mm,
 
 	if (order > mm->max_order || size > mm->size) {
 		if ((flags & GPU_BUDDY_CONTIGUOUS_ALLOCATION) &&
-		    !(flags & GPU_BUDDY_RANGE_ALLOCATION))
-			return __alloc_contig_try_harder(mm, original_size,
-							 original_min_size,
-							 flags, blocks);
+		    !(flags & GPU_BUDDY_RANGE_ALLOCATION)) {
+			err = __alloc_contig_try_harder(mm, original_size,
+							original_min_size,
+							flags, &allocated);
+			if (err)
+				return err;
+
+			list_splice_tail(&allocated, blocks);
+			return 0;
+		}
 
 		return -EINVAL;
 	}
@@ -2064,9 +2070,11 @@ int gpu_buddy_alloc_blocks(struct gpu_buddy *mm,
 								original_size,
 								original_min_size,
 								flags,
-								blocks);
-				if (!err)
+								&allocated);
+				if (!err) {
+					list_splice_tail(&allocated, blocks);
 					return 0;
+				}
 				if (err != -ENOSPC)
 					return err;
 				goto err_free;
diff --git a/drivers/gpu/tests/gpu_buddy_test.c b/drivers/gpu/tests/gpu_buddy_test.c
index b75d32ca6ca0..2c6f1da1a6d1 100644
--- a/drivers/gpu/tests/gpu_buddy_test.c
+++ b/drivers/gpu/tests/gpu_buddy_test.c
@@ -1235,6 +1235,15 @@ static void gpu_test_buddy_alloc_contiguous(struct kunit *test)
 							    GPU_BUDDY_CONTIGUOUS_ALLOCATION),
 			       "buddy_alloc hit an error size=%lu\n", 2 * ps);
 
+	/*
+	 * A failed contiguous allocation with a non-empty blocks list
+	 * must NOT corrupt or free pre-existing blocks.
+	 */
+	KUNIT_EXPECT_TRUE(test, gpu_buddy_alloc_blocks(&mm, 0, mm_size,
+						       3 * ps, ps, &allocated,
+						       GPU_BUDDY_CONTIGUOUS_ALLOCATION));
+	KUNIT_EXPECT_FALSE(test, list_empty(&allocated));
+
 	gpu_buddy_free_list(&mm, &left, 0);
 	KUNIT_ASSERT_FALSE_MSG(test, gpu_buddy_alloc_blocks(&mm, 0, mm_size,
 							    3 * ps, ps, &allocated,
-- 
2.55.0


^ permalink raw reply related	[flat|nested] 4+ messages in thread
end of thread, other threads:[~2026-10-06 16:58 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-10-06 11:07 [PATCH] gpu/buddy: do not pass caller list to __alloc_contig_try_harder() Matthew Auld
2026-10-06 11:15 ` ✓ CI.KUnit: success for " Patchwork
2026-10-06 11:53 ` ✓ Xe.CI.BAT: " Patchwork
2026-10-06 16:58 ` ✗ 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