ext4: stop providing .writepage hook
authorJan Kara <jack@suse.cz>
Wed, 7 Dec 2022 11:27:15 +0000 (12:27 +0100)
committerTheodore Ts'o <tytso@mit.edu>
Fri, 9 Dec 2022 02:49:25 +0000 (21:49 -0500)
commitdae999602eeb0c5482bb62df7fda903f1edc0ff3
tree99f6b9067b64edafd600db9d961f6b9cabacee12
parente26355e21520e56732f58590b812ca619b88ac72
ext4: stop providing .writepage hook

Now we don't need .writepage hook for anything anymore. Reclaim is
fine with relying on .writepages to clean pages and we often couldn't
do much from the .writepage callback anyway. We only need to provide
.migrate_folio callback for the ext4_journalled_aops - let's use
buffer_migrate_page_norefs() there so that buffers cannot be modified
under jdb2's hands as that can cause data corruption. For example when
commit code does writeout of transaction buffers in
jbd2_journal_write_metadata_buffer(), we don't hold page lock or have
page writeback bit set or have the buffer locked. So page migration
code would go and happily migrate the page elsewhere while the copy is
running thus corrupting data.

Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20221207112722.22220-12-jack@suse.cz
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
fs/ext4/inode.c