On Wed, Mar 27, 2013 at 11:12:48AM -0400, Theodore Ts'o wrote:
> On Wed, Mar 27, 2013 at 11:19:22PM +0800, Zheng Liu wrote:
> > > > Otherwise, couldn't we end up with problems where a failed write calls
> > > > ext4_truncate() without i_data_sem(), and that races with something
> > > > else --- say, a punch or truncate call to that same inode?
> >
> > Let me think about it. I need to take a close look at it.
>
> Note that I'm not so concerned when we are creating symlink --- you
> are quite right in pointing out in that case the inode isn't in the
> namespace yet, so that prevents races --- but also what might happen
> in an ENOSPC write(2) failure racing against a punch/truncate call.
>
> But again, this is why I added the warning --- it was to find these
> edge cases that we might not have considered. :-)
ext4_truncate_failed_write() is called by the following functions:
- ext4_ind_direct_IO
- ext4_convert_inline_data_to_extent
- ext4_da_convert_inline_data_to_extent
- ext4_write_begin
- ext4_write_end
- ext4_journalled_write_end
- ext4_da_write_begin
All these functions are protected by i_mutex. So we can serialize it
with truncate/punch hole.
Regards,
- Zheng
--
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
