linux-stable.git/fs/jbd2, branch v5.14.2

ext4: inline jbd2_journal_[un]register_shrinker()

2021-07-08T12:37:31+00:00

The function jbd2_journal_unregister_shrinker() was getting called
twice when the file system was getting unmounted.  On Power and ARM
platforms this was causing kernel crash when unmounting the file
system, when a percpu_counter was destroyed twice.

Fix this by removing jbd2_journal_[un]register_shrinker() functions,
and inlining the shrinker setup and teardown into
journal_init_common() and jbd2_journal_destroy().  This means that
ext4 and ocfs2 now no longer need to know about registering and
unregistering jbd2's shrinker.

Also, while we're at it, rename the percpu counter from
j_jh_shrink_count to j_checkpoint_jh_count, since this makes it
clearer what this counter is intended to track.

Link: https://lore.kernel.org/r/20210705145025.3363130-1-tytso@mit.edu
Fixes: 4ba3fcdde7e3 ("jbd2,ext4: add a shrinker to release checkpointed buffers")
Reported-by: Jon Hunter 
Reported-by: Sachin Sant 
Tested-by: Sachin Sant 
Tested-by: Jon Hunter 
Reviewed-by: Jan Kara 
Signed-off-by: Theodore Ts'o

jbd2: export jbd2_journal_[un]register_shrinker()

2021-06-30T15:05:00+00:00

Export jbd2_journal_[un]register_shrinker() to fix this error when
ext4 is built as a module:

  ERROR: modpost: "jbd2_journal_unregister_shrinker" undefined!
  ERROR: modpost: "jbd2_journal_register_shrinker" undefined!

Fixes: 4ba3fcdde7e3 ("jbd2,ext4: add a shrinker to release checkpointed buffers")
Signed-off-by: Zhang Yi 
Reviewed-by: Jan Kara 
Link: https://lore.kernel.org/r/20210630083638.140218-1-yi.zhang@huawei.com
Signed-off-by: Theodore Ts'o

jbd2: simplify journal_clean_one_cp_list()

2021-06-24T14:55:39+00:00

Now that __try_to_free_cp_buf() remove checkpointed buffer or transaction
when the buffer is not 'busy', which is only called by
journal_clean_one_cp_list(). This patch simplify this function by remove
__try_to_free_cp_buf() and invoke __cp_buffer_busy() directly.

Signed-off-by: Zhang Yi 
Reviewed-by: Jan Kara 
Link: https://lore.kernel.org/r/20210610112440.3438139-7-yi.zhang@huawei.com
Signed-off-by: Theodore Ts'o

jbd2,ext4: add a shrinker to release checkpointed buffers

2021-06-24T14:54:49+00:00

Current metadata buffer release logic in bdev_try_to_free_page() have
a lot of use-after-free issues when umount filesystem concurrently, and
it is difficult to fix directly because ext4 is the only user of
s_op->bdev_try_to_free_page callback and we may have to add more special
refcount or lock that is only used by ext4 into the common vfs layer,
which is unacceptable.

One better solution is remove the bdev_try_to_free_page callback, but
the real problem is we cannot easily release journal_head on the
checkpointed buffer, so try_to_free_buffers() cannot release buffers and
page under memory pressure, which is more likely to trigger
out-of-memory. So we cannot remove the callback directly before we find
another way to release journal_head.

This patch introduce a shrinker to free journal_head on the checkpointed
transaction. After the journal_head got freed, try_to_free_buffers()
could free buffer properly.

Signed-off-by: Zhang Yi 
Suggested-by: Jan Kara 
Reviewed-by: Jan Kara 
Link: https://lore.kernel.org/r/20210610112440.3438139-6-yi.zhang@huawei.com
Signed-off-by: Theodore Ts'o

jbd2: remove redundant buffer io error checks

2021-06-24T14:33:50+00:00

Now that __jbd2_journal_remove_checkpoint() can detect buffer io error
and mark journal checkpoint error, then we abort the journal later
before updating log tail to ensure the filesystem works consistently.
So we could remove other redundant buffer io error checkes.

Signed-off-by: Zhang Yi 
Reviewed-by: Jan Kara 
Link: https://lore.kernel.org/r/20210610112440.3438139-5-yi.zhang@huawei.com
Signed-off-by: Theodore Ts'o

jbd2: don't abort the journal when freeing buffers

2021-06-24T14:33:50+00:00

Now that we can be sure the journal is aborted once a buffer has failed
to be written back to disk, we can remove the journal abort logic in
jbd2_journal_try_to_free_buffers() which was introduced in
commit c044f3d8360d ("jbd2: abort journal if free a async write error
metadata buffer"), because it may cost and propably is not safe.

Signed-off-by: Zhang Yi 
Reviewed-by: Jan Kara 
Link: https://lore.kernel.org/r/20210610112440.3438139-4-yi.zhang@huawei.com
Signed-off-by: Theodore Ts'o

jbd2: ensure abort the journal if detect IO error when writing original buffer back

2021-06-24T14:33:49+00:00

Although we merged c044f3d8360 ("jbd2: abort journal if free a async
write error metadata buffer"), there is a race between
jbd2_journal_try_to_free_buffers() and jbd2_journal_destroy(), so the
jbd2_log_do_checkpoint() may still fail to detect the buffer write
io error flag which may lead to filesystem inconsistency.

jbd2_journal_try_to_free_buffers()     ext4_put_super()
                                        jbd2_journal_destroy()
  __jbd2_journal_remove_checkpoint()
  detect buffer write error              jbd2_log_do_checkpoint()
                                         jbd2_cleanup_journal_tail()
                                           <--- lead to inconsistency
  jbd2_journal_abort()

Fix this issue by introducing a new atomic flag which only have one
JBD2_CHECKPOINT_IO_ERROR bit now, and set it in
__jbd2_journal_remove_checkpoint() when freeing a checkpoint buffer
which has write_io_error flag. Then jbd2_journal_destroy() will detect
this mark and abort the journal to prevent updating log tail.

Signed-off-by: Zhang Yi 
Reviewed-by: Jan Kara 
Link: https://lore.kernel.org/r/20210610112440.3438139-3-yi.zhang@huawei.com
Signed-off-by: Theodore Ts'o

jbd2: remove the out label in __jbd2_journal_remove_checkpoint()

2021-06-24T14:33:06+00:00

The 'out' lable just return the 'ret' value and seems not required, so
remove this label and switch to return appropriate value immediately.
This patch also do some minor cleanup, no logical change.

Signed-off-by: Zhang Yi 
Reviewed-by: Jan Kara 
Link: https://lore.kernel.org/r/20210610112440.3438139-2-yi.zhang@huawei.com
Signed-off-by: Theodore Ts'o

jbd2: clean up misleading comments for jbd2_fc_release_bufs

2021-06-24T14:27:32+00:00

This comments was for jbd2_fc_wait_bufs, not for jbd2_fc_release_bufs.
Remove this misleading comments.

Signed-off-by: yangerkun 
Reviewed-by: Jan Kara 
Link: https://lore.kernel.org/r/20210608141236.459441-1-yangerkun@huawei.com
Signed-off-by: Theodore Ts'o

ext4: add discard/zeroout flags to journal flush

2021-06-22T23:27:10+00:00

Add a flags argument to jbd2_journal_flush to enable discarding or
zero-filling the journal blocks while flushing the journal.

Signed-off-by: Leah Rumancik 
Link: https://lore.kernel.org/r/20210518151327.130198-1-leah.rumancik@gmail.com
Signed-off-by: Theodore Ts'o