Re: [PATCH 1/4] jbd: abort when failed to log metadata buffers

[Eric Sandeen <sandeen@redhat.com>]

Eric Sandeen wrote:
> Hidehiro Kawai wrote:
>> If we failed to write metadata buffers to the journal space and
>> succeeded to write the commit record, stale data can be written
>> back to the filesystem as metadata in the recovery phase.
>>
>> To avoid this, when we failed to write out metadata buffers,
>> abort the journal before writing the commit record.
>>
>> Signed-off-by: Hidehiro Kawai <hidehiro.kawai.ez@hitachi.com>
>> Acked-by: Jan Kara <jack@suse.cz>
>> ---
>>  fs/jbd/commit.c |    3 +++
>>  1 file changed, 3 insertions(+)
>>
>> Index: linux-2.6.26-rc8-mm1/fs/jbd/commit.c
>> ===================================================================
>> --- linux-2.6.26-rc8-mm1.orig/fs/jbd/commit.c
>> +++ linux-2.6.26-rc8-mm1/fs/jbd/commit.c
>> @@ -762,6 +762,9 @@ wait_for_iobuf:
>>  		/* AKPM: bforget here */
>>  	}
>>  
>> +	if (err)
>> +		journal_abort(journal, err);
>> +
>>  	jbd_debug(3, "JBD: commit phase 6\n");
>>  
>>  	if (journal_write_commit_record(journal, commit_transaction))
> 
> A little late for me to chime in, but why is this a journal_abort() when
> journal aborts before and after this call __journal_abort_hard()?
> 
> I'm just not sure what's different about this one condition that we wish
> to update the error and the superblock here?


I take that back, was looking at an old tree, and Josef pointed out
commit 7a266e75cf5a1efd20d084408a1b7f1a185496dd to me that changes this,
and now it makes much more sense.  :)

http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=7a266e75cf5a1efd20d084408a1b7f1a185496dd

Thanks,
-Eric