Re: [PATCH 0/2] Fix regression on NFS in mainline kernel

[Luis Henriques <luis.henriques@canonical.com>]

On Thu, Apr 19, 2012 at 05:29:09PM +0000, Myklebust, Trond wrote:
> On Thu, 2012-04-19 at 14:48 +0100, Luis Henriques wrote:
> > On Thu, Apr 19, 2012 at 10:30:25AM +0100, Luis Henriques wrote:
> > > On Wed, Apr 18, 2012 at 01:36:52PM -0400, Trond Myklebust wrote:
> > > > The first patch addresses the Oops.
> > > > The second will hopefully address the looping.
> > > > 
> > > > Trond Myklebust (2):
> > > >   NFSv4: Ensure that the LOCK code sets exception->inode
> > > >   NFSv4: Ensure that we check lock exclusive/shared type against open
> > > >     modes
> > > > 
> > > >  fs/nfs/nfs4proc.c |   23 +++++++++++++++++++++--
> > > >  1 files changed, 21 insertions(+), 2 deletions(-)
> > > > 
> > > > -- 
> > > > 1.7.7.6
> > > > 
> > > Just to let you know we do have a few successful tests on these patches.
> > > You can check the details here:
> > > 
> > > http://bugs.launchpad.net/bugs/974664
> > > 
> > > Although I haven't tested the patches myself, fell free to add a Tested-by
> > > to these patches.
> > 
> > There's something else a forgot to mention, which is the fact that the bug
> > reports were for kernel 3.2.14.  So you may want to update the
> > "Cc: stable@vger.kernel.org" on commit 487790f27df9bb27d3400486bd021dd59edc7589
> > to include at least this version.
> 
> Yep. I also saw that there is a need for it in 3.0.27, so I'll just
> remove that >= 3.3.1...

Great, thanks.
 
> > There are two another patches we have applied that are present in mainline
> > but haven't made its way into stable:
> >  - 14977489ffdb80d4caf5a184ba41b23b02fbacd9
> >  - 96dcadc2fdd111dca90d559f189a30c65394451a
> 
> I don't plan on sending these 2 commits to stable unless we see some
> specific problems that need to be corrected. I understand that seeing a
> NFS4ERR_OPENMODE at the wrong time could theoretically cause an Oops
> without the 1497748 commit, but broken servers can wreak all sorts of
> havoc anyway. There isn't much you can do to protect against them.

I'm afraid I can't follow you on the protocol-related details, but the bug
report I referred above was caused by an Oops which could be avoided by
improving the robustness of the code (which the first commit does by
checking for a NULL).  Also, I'm not sure the NFS server being used by the
bug reporter would fit into the "broken servers" category, as it seems to
be working OK.  For all of this, I would be glad to see the first commit
on stable.

Anyway, thanks a lot for your help sorting this out.

Cheers,
--
Luis