Message-ID: <20200526071754.33819-1-yi.zhang@huawei.com>
Date:   Tue, 26 May 2020 15:17:44 +0800
From:   "zhangyi (F)" <yi.zhang@...wei.com>
To:     <linux-ext4@...r.kernel.org>
CC:     <tytso@....edu>, <jack@...e.cz>, <adilger.kernel@...ger.ca>,
        <yi.zhang@...wei.com>, <zhangxiaoxu5@...wei.com>
Subject: [PATCH 00/10] ext4: fix inconsistency since reading old metadata from disk

Background
==========

This patch set point to fix the inconsistency problem which has been
discussed and partial fixed in [1].

Now, the problem is on the unstable storage which has a flaky transport
(e.g. iSCSI transport may disconnect few seconds and reconnect due to
the bad network environment), if we failed to async write metadata in
background, the end write routine in block layer will clear the buffer's
uptodate flag, but the data in such buffer is actually uptodate. Finally
we may read "old && inconsistent" metadata from the disk when we get the
buffer later because not only the uptodate flag was cleared but also we
do not check the write io error flag, or even worse the buffer has been
freed due to memory presure.

Fortunately, if the jbd2 do checkpoint after async IO error happens,
the checkpoint routine will check the write_io_error flag and abort the
the journal if detect IO error. And in the journal recover case, the
recover code will invoke sync_blockdev() after recover complete, it will
also detect IO error and refuse to mount the filesystem.

Current ext4 have already deal with this problem in __ext4_get_inode_loc()
and commit 7963e5ac90125 ("ext4: treat buffers with write errors as
containing valid data"), but it's not enough.

[1] https://lore.kernel.org/linux-ext4/20190823030207.GC8130@mit.edu/

Description
===========

This patch set add and rework 7 wrapper functions of getting metadata
blocks, replace all sb_bread() / sb_getblk*() / ext4_bread() and
sb_breadahead*(). Add buffer_write_io_error() checking into them, if
the buffer isn't uptodate and write_io_error flag was set, which means
that the buffer has been failed to write out to disk, re-add the
uptodate flag to prevent subsequent read operation.

 - ext4_sb_getblk(): works the same as sb_getblk(), use to replace all
   sb_getblk() used for newly allocated blocks and getting buffers.
 - ext4_sb_getblk_locked(): works the same as sb_getblk() except check &
   fix buffer uotpdate flag, use to replace all sb_getblk() used for
   getting buffers to read.
 - ext4_sb_getblk_gfp(): gfp version of ext4_sb_getblk().
 - ext4_sb_getblk_locked_gfp(): gfp version of ext4_sb_getblk_locked().
 - ext4_sb_bread(): get buffer and submit read bio if buffer is actually
   not uptodate.
 - ext4_sb_bread_unmovable(): unmovable version of ext4_sb_bread().
 - ext4_sb_breadahead_unmovable(): works the same to ext4_sb_bread_unmovable()
   except skip submit read bio if failed to lock the buffer.

Patch 1-2: do some small change in ext4 inode eio simulation and add a
helper in buffer.c, just prepare for below patches.
Patch 3: add the ext4_sb_*() function to deal with the write_io_error
flag in buffer.
Patch 4-8: replace all sb_*() with ext4_sb_*() in ext4.
Patch 9: deal with the buffer shrinking case, abort jbd2/fs when
shrinking a buffer with write_io_error flag.
Patch 10: just do some cleanup.

After this patch set, we need to use above 7 wrapper functions to
get/read metadata block instead of invoke sb_*() functions defined in
fs/buffer.h.

Test
====

This patch set is based on linux-5.7-rc7 and has been tests by xfstests
in auto mode.

Thanks,
Yi.


zhangyi (F) (10):
  ext4: move inode eio simulation behind io completeion
  fs: pick out ll_rw_one_block() helper function
  ext4: add ext4_sb_getblk*() wrapper functions
  ext4: replace sb_getblk() with ext4_sb_getblk_locked()
  ext4: replace sb_bread*() with ext4_sb_bread*()
  ext4: replace sb_getblk() with ext4_sb_getblk()
  ext4: switch to use ext4_sb_getblk_locked() in ext4_getblk()
  ext4: replace sb_breadahead() with ext4_sb_breadahead()
  ext4: abort the filesystem while freeing the write error io buffer
  ext4: remove unused parameter in jbd2_journal_try_to_free_buffers()

 fs/buffer.c                 |  41 ++++++----
 fs/ext4/balloc.c            |   6 +-
 fs/ext4/ext4.h              |  60 ++++++++++++---
 fs/ext4/extents.c           |  13 ++--
 fs/ext4/ialloc.c            |   6 +-
 fs/ext4/indirect.c          |  13 ++--
 fs/ext4/inline.c            |   2 +-
 fs/ext4/inode.c             |  53 +++++--------
 fs/ext4/mmp.c               |   2 +-
 fs/ext4/resize.c            |  24 +++---
 fs/ext4/super.c             | 145 +++++++++++++++++++++++++++++++-----
 fs/ext4/xattr.c             |   4 +-
 fs/jbd2/transaction.c       |  20 +++--
 include/linux/buffer_head.h |   1 +
 include/linux/jbd2.h        |   3 +-
 15 files changed, 277 insertions(+), 116 deletions(-)

-- 
2.21.3

Powered by blists - more mailing lists