From: Shida Zhang <zhangshida@xxxxxxxxxx>
Hi all,
On an old kernel version(4.19, ext3, data=journal, pagesize=64k),
an assertion failure will occasionally be triggered by the line below:
---------
jbd2_journal_commit_transaction
{
...
J_ASSERT_BH(bh, !buffer_dirty(bh));
/*
* The buffer on BJ_Forget list and not jbddirty means
...
}
---------
The same condition may also be applied to the lattest kernel version.
This patch set fixes it by:
1.Fix a small bug for ext4_journalled_zero_new_buffers first.(patch 1)
2.Replace the __block_write_begin with the hoisted
ext4_block_write_begin().(patch 2)
3.Trace the user data dirtying in ext4_block_write_begin().(patch 3)
4.Clean up some extra things.(patch 4)
Changes since v5:
- Moved a hunk inside the if (buffer_new(bh)) check in patch 3.
- Add a cleanup in patch 4.
[5] Version 5:
https://lore.kernel.org/linux-ext4/20240829085407.3331490-1-zhangshida@xxxxxxxxxx/
Changes since v4:
- At first, we fix a bug in ext4_journalled_zero_new_buffers, as suggested
by Jan.
- In patch 2, clean up the related comment. And remove the #ifdef in
ext4_block_write_begin(), as suggested by Eric.
- Add some comments in patch 3.
[4] Version 4:
https://lore.kernel.org/linux-ext4/20240823013329.1996741-1-zhangshida@xxxxxxxxxx/
Changes since v3:
- Ditch the patch 3 in v3, because some other code paths can set the
buffer dirty:
ext4_write_begin
ext4_block_write_begin
create_empty_buffers
set_buffer_dirty
[3] Version 3:
https://lore.kernel.org/linux-ext4/20240810082814.3709867-1-zhangshida@xxxxxxxxxx/
Changes since v2:
- Adjust the applied order of patch 1 and patch 2 in v1.
- Reword the commit message.
- Remove some superfluous logic in patch 2 and patch 3.
[2] Version 2:
https://lore.kernel.org/linux-ext4/20240809064606.3490994-2-zhangshida@xxxxxxxxxx/
Changes since v1:
- v1 use a hack into jbd2 to fix the bug while v2 choose to journal
the dirty data in *_block_write_begin.
[1] Version 1:
https://lore.kernel.org/linux-ext4/20240720062356.2522658-1-zhangshida@xxxxxxxxxx/
Shida Zhang (4):
ext4: persist the new uptodate buffers in
ext4_journalled_zero_new_buffers
ext4: hoist ext4_block_write_begin and replace the __block_write_begin
ext4: fix a potential assertion failure due to improperly dirtied
buffer
ext4: remove the special buffer dirty handling in
do_journal_get_write_access
fs/ext4/ext4.h | 3 ++
fs/ext4/inline.c | 11 ++++---
fs/ext4/inode.c | 80 +++++++++++++++++++++++-------------------------
3 files changed, 47 insertions(+), 47 deletions(-)
--
2.33.0
