public inbox for linux-xfs@vger.kernel.org
 help / color / mirror / Atom feed
* Mystery solved?  files not written for days...
@ 2009-03-30 22:01 Eric Sandeen
  2009-03-30 22:30 ` Felix Blyakher
  0 siblings, 1 reply; 2+ messages in thread
From: Eric Sandeen @ 2009-03-30 22:01 UTC (permalink / raw)
  To: xfs mailing list

On LKML:

[PATCH] writeback: guard against jiffies wraparound on
inode->dirtied_when checks

When a file is continually dirtied, the check in sync_sb_inodes:

        /* Was this inode dirtied after sync_sb_inodes was called? */
        if (time_after(inode->dirtied_when, start))
                break;

may have large windows where this trips on the first
(continually-dirtied) inode on the list when the time_after check has
wrapped, causing the nothing to be written out for that superblock,
potentially for days.

Could this be the reason for those various "hey, my file disappeared
after a crash and I hadn't written to it for *days*" reports we got now
and then, leaving us all scratching our heads?

-Eric

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

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: Mystery solved?  files not written for days...
  2009-03-30 22:01 Mystery solved? files not written for days Eric Sandeen
@ 2009-03-30 22:30 ` Felix Blyakher
  0 siblings, 0 replies; 2+ messages in thread
From: Felix Blyakher @ 2009-03-30 22:30 UTC (permalink / raw)
  To: Eric Sandeen; +Cc: xfs mailing list


On Mar 30, 2009, at 5:01 PM, Eric Sandeen wrote:

> On LKML:
>
> [PATCH] writeback: guard against jiffies wraparound on
> inode->dirtied_when checks

The "jiffies wraparound" sounds very familiar, but I
thought it's been fixed ages ago.

> When a file is continually dirtied, the check in sync_sb_inodes:
>
>        /* Was this inode dirtied after sync_sb_inodes was called? */
>        if (time_after(inode->dirtied_when, start))
>                break;
>
> may have large windows where this trips on the first
> (continually-dirtied) inode on the list when the time_after check has
> wrapped, causing the nothing to be written out for that superblock,
> potentially for days.
>
> Could this be the reason for those various "hey, my file disappeared
> after a crash and I hadn't written to it for *days*" reports we got  
> now
> and then, leaving us all scratching our heads?

The "jiffies wraparound" could definitely cause this.

Felix

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

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

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2009-03-30 22:57 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2009-03-30 22:01 Mystery solved? files not written for days Eric Sandeen
2009-03-30 22:30 ` Felix Blyakher

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox