Currently we clear BH_New bit in case of error and also in the standard
ext4_write_end() handler (in block_commit_write()). However
ext4_journalled_write_end() misses this clearing and thus we are leaving
stale BH_New bits behind. Generally ext4_block_write_begin() clears
these bits before any harm can be done but in case blocksize < pagesize
and we hit some error when processing a page with these stale bits,
we'll try to zero buffers with these stale BH_New bits and jbd2 will
complain (as buffers were not prepared for writing in this transaction).
Fix the problem by clearing BH_New bits in ext4_journalled_write_end()
and WARN if ext4_block_write_begin() sees stale BH_New bits.
Reported-by: Baolin Liu <liubaolin12138@xxxxxxx>
Reported-by: Zhi Long <longzhi@xxxxxxxxxxxxxx>
Fixes: 3910b513fcdf ("ext4: persist the new uptodate buffers in ext4_journalled_zero_new_buffers")
Signed-off-by: Jan Kara <jack@xxxxxxx>
---
fs/ext4/inline.c | 2 ++
fs/ext4/inode.c | 3 ++-
2 files changed, 4 insertions(+), 1 deletion(-)
Changes since v3 (https://lore.kernel.org/all/20241113175550.GA462442@xxxxxxx):
* Clear stale BH_New flags resulting from inline->extent data conversion
diff --git a/fs/ext4/inline.c b/fs/ext4/inline.c
index 3536ca7e4fcc..0d3cf0ca5c2a 100644
--- a/fs/ext4/inline.c
+++ b/fs/ext4/inline.c
@@ -606,6 +606,7 @@ static int ext4_convert_inline_data_to_extent(struct address_space *mapping,
} else
ret = ext4_block_write_begin(handle, folio, from, to,
ext4_get_block);
+ clear_buffer_new(folio_buffers(folio));
if (!ret && ext4_should_journal_data(inode)) {
ret = ext4_walk_page_buffers(handle, inode,
@@ -867,6 +868,7 @@ static int ext4_da_convert_inline_data_to_extent(struct address_space *mapping,
return ret;
}
+ clear_buffer_new(folio_buffers(folio));
folio_mark_dirty(folio);
folio_mark_uptodate(folio);
ext4_clear_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA);
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 54bdd4884fe6..aa56af4a92ad 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -1049,7 +1049,7 @@ int ext4_block_write_begin(handle_t *handle, struct folio *folio,
}
continue;
}
- if (buffer_new(bh))
+ if (WARN_ON_ONCE(buffer_new(bh)))
clear_buffer_new(bh);
if (!buffer_mapped(bh)) {
WARN_ON(bh->b_size != blocksize);
@@ -1265,6 +1265,7 @@ static int write_end_fn(handle_t *handle, struct inode *inode,
ret = ext4_dirty_journalled_data(handle, bh);
clear_buffer_meta(bh);
clear_buffer_prio(bh);
+ clear_buffer_new(bh);
return ret;
}
--
2.35.3
