Re: [PATCH] ext4: Fix data corruption in inodes with journalled data

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



  Hello Amir,

On Sat 23-07-11 16:21:55, Amir Goldstein wrote:
> On Sat, Jul 23, 2011 at 3:39 AM, Jan Kara <jack@xxxxxxx> wrote:
> > When journalling data for an inode (either because it is a symlink or
> > because the filesystem is mounted in data=journal mode),
> > ext4_evict_inode() can discard unwritten data by calling
> > truncate_inode_pages(). This is because we don't mark the buffer / page
> > dirty when journalling data but only add the buffer to the running
> > transaction and thus mm does not know there are still unwritten data.
> >
> > Fix the problem by carefully tracking transaction containing inode's
> > data, committing this transaction, and writing uncheckpointed buffers
> > when inode should be reaped.
> >
> > Signed-off-by: Jan Kara <jack@xxxxxxx> ---  fs/ext4/inode.c |   29
> > +++++++++++++++++++++++++++++  1 files changed, 29 insertions(+), 0
> > deletions(-)
> >
> >  This is ext4 version of an ext3 fix I sent a while ago. It received
> > only light testing but I figured you might want get the patch earlier
> > rather than later given the merge window is open.
> >
> > diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index e3126c0..019995b
> > 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -190,6 +190,33 @@
> > void ext4_evict_inode(struct inode *inode)
> >
> >        trace_ext4_evict_inode(inode);        if (inode->i_nlink) { +  
> >             /* +                * When journalling data dirty buffers
> > are tracked only in the +                * journal. So although mm
> > thinks everything is clean and +                * ready for reaping the
> > inode might still have some pages to +                * write in the
> > running transaction or waiting to be +                * checkpointed.
> > Thus calling jbd2_journal_invalidatepage() +                * (via
> > truncate_inode_pages()) to discard these buffers can +                *
> > cause data loss. Also even if we did not discard these +              
> >  * buffers, we would have no way to find them after the inode +        
> >        * is reaped and thus user could see stale data if he tries to +
> >                * read them before the transaction is checkpointed. So
> > be +                * careful and force everything to disk here... We
> > use +                * ei->i_datasync_tid to store the newest
> > transaction +                * containing inode's data.  +            
> >    * +                * Note that directories do not have this problem
> > because they +                * don't use page cache.  +              
> >  */ +               if (ext4_should_journal_data(inode) && +          
> >         (S_ISLNK(inode->i_mode) || S_ISREG(inode->i_mode))) { +        
> >               journal_t *journal = EXT4_SB(inode->i_sb)->s_journal; +  
> >                     tid_t commit_tid = EXT4_I(inode)->i_datasync_tid; +
> > +                       jbd2_log_start_commit(journal, commit_tid); +  
> >                     jbd2_log_wait_commit(journal, commit_tid); +      
> >                 filemap_write_and_wait(&inode->i_data); +              
> > }                truncate_inode_pages(&inode->i_data, 0);              
> >  goto no_delete;        } @@ -1863,6 +1890,7 @@ static int
> > ext4_journalled_write_end(struct file *file,        if (new_i_size >
> > inode->i_size)                i_size_write(inode, pos+copied);      
> >  ext4_set_inode_state(inode, EXT4_STATE_JDATA); +      
> > EXT4_I(inode)->i_datasync_tid = handle->h_transaction->t_tid;        if
> > (new_i_size > EXT4_I(inode)->i_disksize) {              
> >  ext4_update_i_disksize(inode, new_i_size);                ret2 =
> > ext4_mark_inode_dirty(handle, inode); @@ -2571,6 +2599,7 @@ static int
> > __ext4_journalled_writepage(struct page *page,                        
> >        write_end_fn);        if (ret == 0)                ret = err; +
> >       EXT4_I(inode)->i_datasync_tid = handle->h_transaction->t_tid;    
> >    err = ext4_journal_stop(handle);        if (!ret)                ret
> > = err; -- 1.7.1
> >
> Patch looks correct to me, but I am uncomfortable with i_datasync_tid
> being treated differently in journalled write - that is, being set on
> different places in the write paths.
> 
> How about setting i_datasync_tid in a more generic place like
> ext4_{,da_}write_begin()?  I know it's a bit redundant to setting dirty
> pages, but at least this way i_datasync_tid can be checked in all journal
> modes and have a consistent meaning.
  Well, I kept the meaning that i_datasync_tid is ID of a transaction that
must be committed for a data of an inode to be safely on disk. It is true
that in data=journal mode, we need to update this number differently than
in other journaling modes but that's not important I think. Currently, we
just force commit in data=journal mode in every case and thus we do not
really care about the value of i_datasync_tid for fsync. In future we could
be more clever and avoid transaction commits for fsync in data=journal mode
in some cases.  So in fact I'd say the code is now *more* consistent than
it used to be.  The only thing that isn't quite consistent is that I didn't
bother with updating i_sync_tid because we currently do not use it. If
people want, that might be a useful cleanup which I can do.

> Perhaps we can even use i_datasync_tid to optimize away things like
> fiemap checks for dirty pages.
  Umm, I'm not sure which checks do you mean...

								Honza
-- 
Jan Kara <jack@xxxxxxx>
SUSE Labs, CR
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html


[Index of Archives]     [Reiser Filesystem Development]     [Ceph FS]     [Kernel Newbies]     [Security]     [Netfilter]     [Bugtraq]     [Linux FS]     [Yosemite National Park]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Samba]     [Device Mapper]     [Linux Media]

  Powered by Linux