I'll give it a quick run this afternoon - Thanks
-----Original Message-----
From: Myklebust, Trond [mailto:[email protected]]
Sent: Thursday, August 16, 2012 1:27 PM
To: Linus Torvalds
Cc: David Howells; Steve Dickson; stable; Shai Lazmi
Subject: Re: kernel 3.6.0-rc1-00312-g3bf671a bug report
On Thu, 2012-08-16 at 15:22 -0400, Trond Myklebust wrote:
> On Thu, 2012-08-16 at 11:55 -0700, Linus Torvalds wrote:
> > Shai - *please* don't send stuff to me personally unless I'm
> > absolutely the only person who can help. Which is not normally the
> > case at all.
> >
> > Trond, this BUG_ON() seems to have been introduced in commit
> > a427b9ec4eda8 ("NFS: Fix a number of bugs in the idmapper") from David
> > Howells.
> >
> > David? Why is there a BUG_ON() there? Killing the machine helps us exactly
> > how?
> >
> > There seems to be some missing locking wrt idmap->idmap_key_cons
> > accesses. Instead, it uses the insane
> >
> > + cons = ACCESS_ONCE(idmap->idmap_key_cons);
> > + idmap->idmap_key_cons = NULL;
> >
> > sequence that makes no sense. What's so special about it that it needs
> > ACCESS_ONCE()? If that access is unlocked, then setting the field to
> > NULL directly afterwards seems singularly unsafe. What's the logic
> > behind such an access pattern?
>
> Hi Linus,
>
> I believe that this is the same issue that William Dauchy hit last week,
> and for which we have already have a diagnosis: it is due to the
> upstream code failing to clear idmap->idmap_key_cons if/when the legacy
> upcall fails.
>
> I do have a couple of patches available to fix this issue in linux-next,
> but we found a problem with one of them during testing earlier today, so
> I'm holding back sending them upstream pending further testing.
OK. I've just sent out a new version of the those 2 patches for testing,
and I made sure to Cc: Shai.
Shai, can you please test if they fix the problem you are seeing?
Cheers
Trond
--
Trond Myklebust
Linux NFS client maintainer
NetApp
[email protected]
www.netapp.com
This message and any attached documents contain information from QLogic
Corporation or its wholly-owned subsidiaries that may be confidential. If you
are not the intended recipient, you may not read, copy, distribute, or use this
information. If you have received this transmission in error, please notify the
sender immediately by reply e-mail and then delete this message.
