Re: [RFC PATCH 07/11] xfs: retry trans reservation on ENOSPC in xfs_inactive()

[Brian Foster <bfoster@redhat.com>]

On 09/05/2013 08:17 PM, Dave Chinner wrote:
> On Thu, Sep 05, 2013 at 12:18:31PM -0400, Brian Foster wrote:
>> On 09/04/2013 09:40 PM, Dave Chinner wrote:
>>> On Tue, Sep 03, 2013 at 02:25:04PM -0400, Brian Foster wrote:
>>>> An ifree data block reservation can fail with ENOSPC. Flush inodes
>>>> to try and free up space or attempt without a data block
>>>> reservation to avoid failing out of xfs_inactive().
>>>>
>>>> Signed-off-by: Brian Foster <bfoster@redhat.com>
>>>> ---
>>>>  fs/xfs/xfs_inode.c | 11 +++++++++++
>>>>  1 file changed, 11 insertions(+)
>>>>
...
> 
>> Subsequent to avoiding that, I
>> believe there were inconsistent fs issues called out due to the unlinked
>> lists being populated after umount.
> 
> That sounds like a recovery failure, not so much an ENOSPC failure.
> i.e. that recovery only looks at the log to see if it's clean, and
> only recovers unlinked lists if it's dirty. There is the
> *possibility* of having a clean log with inodes on the unlinked
> list, and log recovery doesn't run the unlinked list processing in
> that case.
> 

Interesting, I'll have a closer look when I rework the inactive
transaction reservation bits. Thanks.

Brian

> This is one of the issues we'll need to fix for O_TMPFILE support
> as it will actively use inodes on unlinked list for potentially long
> periods of time.
> 
>> Taking a further look, I missed the XFS_TRANS_RESERVE flag and whole
>> m_resblks mechanism. I'll take a closer look at that and see if that
>> works to resolve the problem instead of the flush.
> 
> It should - the only time it won't is if we exhaust the pool, but
> that doesn't happen in normal ENOSPC situations and any blocks we do
> end up freeing will immediately refill the reserve pool...
> 
> Cheers,
> 
> Dave.
> 

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs