]> git.ipfire.org Git - thirdparty/kernel/stable.git/commitdiff
btrfs: don't stop integrity writeback too early
authorChristoph Hellwig <hch@lst.de>
Mon, 24 Jul 2023 13:26:53 +0000 (06:26 -0700)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 16 Aug 2023 16:22:02 +0000 (18:22 +0200)
commit effa24f689ce0948f68c754991a445a8d697d3a8 upstream.

extent_write_cache_pages stops writing pages as soon as nr_to_write hits
zero.  That is the right thing for opportunistic writeback, but incorrect
for data integrity writeback, which needs to ensure that no dirty pages
are left in the range.  Thus only stop the writeback for WB_SYNC_NONE
if nr_to_write hits 0.

This is a port of write_cache_pages changes in commit 05fe478dd04e
("mm: write_cache_pages integrity fix").

Note that I've only trigger the problem with other changes to the btrfs
writeback code, but this condition seems worthwhile fixing anyway.

CC: stable@vger.kernel.org # 4.14+
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: David Sterba <dsterba@suse.com>
[ updated comment ]
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fs/btrfs/extent_io.c

index 7bd704779a99b03c5c4ef7e9b10c8b1238b8ce65..f9f6dfbc86bcd9390421f8e87f0e3f78e96b52b9 100644 (file)
@@ -4844,11 +4844,12 @@ retry:
                        }
 
                        /*
-                        * the filesystem may choose to bump up nr_to_write.
+                        * The filesystem may choose to bump up nr_to_write.
                         * We have to make sure to honor the new nr_to_write
-                        * at any time
+                        * at any time.
                         */
-                       nr_to_write_done = wbc->nr_to_write <= 0;
+                       nr_to_write_done = (wbc->sync_mode == WB_SYNC_NONE &&
+                                           wbc->nr_to_write <= 0);
                }
                pagevec_release(&pvec);
                cond_resched();