FAILED: patch "[PATCH] dm-bufio: fix non-working memory limit" failed to apply to 6.6-stable tree

[Sasha Levin <sashal@kernel.org>]

The patch below does not apply to the 6.6-stable tree.
If someone wants it applied there, or to any other stable or longterm
tree, then please email the backport, including the original git commit
id to <stable@vger.kernel.org>.

To reproduce the conflict and resubmit, you may use the following commands:

git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-6.6.y
git checkout FETCH_HEAD
git cherry-pick -x 1190dbeaeaa026a4879e675b8556a150c36a7ced
# <resolve conflicts, build, test, etc.>
git commit -s
git send-email --to '<stable@vger.kernel.org>' --in-reply-to '20261107053426-failed-6.6-1190dbeaeaa0@kernel.org' --subject-prefix 'PATCH 6.6.y' 'HEAD^..'

Thanks,
Sasha

------------------ original commit in Linus's tree ------------------

From 1190dbeaeaa026a4879e675b8556a150c36a7ced Mon Sep 17 00:00:00 2001
From: Mikulas Patocka <mpatocka@redhat.com>
Date: Mon, 12 Oct 2026 16:07:58 +0200
Subject: [PATCH] dm-bufio: fix non-working memory limit

The function evict_old is supposed to loop and evict buffers while the
allocated size is above the threshold. However if it hits one bufio
client that is empty, it stops and doesn't attempt to scan other
clients.

This commit fixes it, so that it skips empty clients and retries up to
dm_bufio_client_count times.

Fixes: 450e8dee51aa ("dm bufio: improve concurrent IO performance")
Cc: stable@vger.kernel.org
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Assisted-by: Claude:claude-opus-5
---
 drivers/md/dm-bufio.c | 41 +++++++++++++++++++++++++++++++++++++----
 1 file changed, 37 insertions(+), 4 deletions(-)

diff --git a/drivers/md/dm-bufio.c b/drivers/md/dm-bufio.c
index d58eed96a8ff5..edad64654ef68 100644
--- a/drivers/md/dm-bufio.c
+++ b/drivers/md/dm-bufio.c
@@ -2545,6 +2545,13 @@ struct dm_bufio_client *dm_bufio_client_create(struct block_device *bdev, unsign
 	init_waitqueue_head(&c->free_buffer_wait);
 	c->async_write_error = 0;
 
+	/*
+	 * A client with no buffers yet is not holding the oldest buffer in
+	 * the system; leaving this zeroed would claim exactly that and keep
+	 * the client pinned at the front of dm_bufio_all_clients.
+	 */
+	c->oldest_buffer = jiffies;
+
 	c->dm_io = dm_io_client_create();
 	if (IS_ERR(c->dm_io)) {
 		r = PTR_ERR(c->dm_io);
@@ -2779,9 +2786,23 @@ static unsigned long __evict_a_few(unsigned long nr_buffers)
 
 	dm_bufio_unlock(c);
 
-	if (count)
+	if (count) {
 		c->oldest_buffer = oldest_buffer;
-	__insert_client(c);
+		__insert_client(c);
+	} else {
+		/*
+		 * We evicted nothing, so 'oldest_buffer' tells us nothing new
+		 * and the client keeps its old, possibly very stale, value.
+		 * Requeue it at the back rather than by that value, otherwise
+		 * __insert_client() would put it straight back at the front
+		 * and every subsequent call would pop this same client again,
+		 * never reaching any of the others.
+		 *
+		 * This leaves the list slightly out of order, which is fine -
+		 * it is only ever an approximation.
+		 */
+		list_add_tail(&c->client_list, &dm_bufio_all_clients);
+	}
 
 	return count;
 }
@@ -2806,11 +2827,23 @@ static void evict_old(void)
 {
 	unsigned long threshold = dm_bufio_cache_size -
 		dm_bufio_cache_size / DM_BUFIO_LOW_WATERMARK_RATIO;
+	int unproductive = 0;
 
 	mutex_lock(&dm_bufio_clients_lock);
 	while (dm_bufio_current_allocated > threshold) {
-		if (!__evict_a_few(64))
-			break;
+		if (__evict_a_few(64)) {
+			unproductive = 0;
+		} else {
+			/*
+			 * This client had nothing to give.  Try the others
+			 * before giving up: one idle client, or one whose
+			 * buffers are all held, must not stop us evicting
+			 * from everybody else.  Stop once a whole sweep of
+			 * the client list has freed nothing.
+			 */
+			if (++unproductive >= dm_bufio_client_count)
+				break;
+		}
 		cond_resched();
 	}
 	mutex_unlock(&dm_bufio_clients_lock);
-- 
2.53.0