Re: vfs related crash in 2.6.33-rc2

Re: vfs related crash in 2.6.33-rc2

[Andrew Morton]

On Thu, 31 Dec 2009 05:59:32 +0900
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp> wrote:

> Marvin <marvin24@gmx.de> writes:
> 
> >> Marvin <marvin24@gmx.de> writes:
> >> > Hi,
> >> >
> >> > I'm getting a lot of these:
> >> >
> >> > kernel: general protection fault: 0000 [#1] SMP
> >> > kernel: last sysfs file: /sys/devices/pci0000:00/0000:00:18.3/modalias
> >> > kernel: CPU 0
> >> > kernel: Pid: 12177, comm: packagekitd Not tainted 2.6.33-rc2 #1
> >> > ...
> >> >
> >> > filesystem is ext4 (in case it matters).
> >> 
> >> BTW, are you using nfs client on this machine?
> >>
> >
> > um - yes, now that I think about it... I killed a nfs umount process (because of an 
> > offline server) shortly before the oopses started to fire.
> 
> OK. Probably, this oops would be same with one which happened on my
> machine recently. That path in patch corrupts dcache hash, so it can be
> the cause of strange behavior or oops on dcache hash.
> 
> If so, the attached patch would fix it.
> 
> Thanks.
> -- 
> OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
> 
> 
> Recent change is missing to update "rehash". With that change, it will
> become the cause of adding dentry to hash twice.
> 
> This explains the reason of Oops (dereference the freed dentry in
> __d_lookup()) on my machine.
> 
> Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
> ---
> 
>  fs/nfs/dir.c |    1 +
>  1 file changed, 1 insertion(+)
> 
> diff -puN fs/nfs/dir.c~nfs-d_rehash-fix fs/nfs/dir.c
> --- linux-2.6/fs/nfs/dir.c~nfs-d_rehash-fix	2009-12-28 06:18:09.000000000 +0900
> +++ linux-2.6-hirofumi/fs/nfs/dir.c	2009-12-28 06:18:16.000000000 +0900
> @@ -1615,6 +1615,7 @@ static int nfs_rename(struct inode *old_
>  				goto out;
>  
>  			new_dentry = dentry;
> +			rehash = NULL;
>  			new_inode = NULL;
>  		}
>  	}

Guys, what's the status of this fix?  Did Marvin have a chance to test
it?  Are the NFS developers aware of it?

Thanks.

Re: vfs related crash in 2.6.33-rc2

[Trond Myklebust]

On Wed, 2010-01-06 at 15:41 -0800, Andrew Morton wrote: 
> On Thu, 31 Dec 2009 05:59:32 +0900
> OGAWA Hirofumi <hirofumi@mail.parknet.co.jp> wrote:
> 
> > Marvin <marvin24@gmx.de> writes:
> > 
> > >> Marvin <marvin24@gmx.de> writes:
> > >> > Hi,
> > >> >
> > >> > I'm getting a lot of these:
> > >> >
> > >> > kernel: general protection fault: 0000 [#1] SMP
> > >> > kernel: last sysfs file: /sys/devices/pci0000:00/0000:00:18.3/modalias
> > >> > kernel: CPU 0
> > >> > kernel: Pid: 12177, comm: packagekitd Not tainted 2.6.33-rc2 #1
> > >> > ...
> > >> >
> > >> > filesystem is ext4 (in case it matters).
> > >> 
> > >> BTW, are you using nfs client on this machine?
> > >>
> > >
> > > um - yes, now that I think about it... I killed a nfs umount process (because of an 
> > > offline server) shortly before the oopses started to fire.
> > 
> > OK. Probably, this oops would be same with one which happened on my
> > machine recently. That path in patch corrupts dcache hash, so it can be
> > the cause of strange behavior or oops on dcache hash.
> > 
> > If so, the attached patch would fix it.
> > 
> > Thanks.
> > -- 
> > OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
> > 
> > 
> > Recent change is missing to update "rehash". With that change, it will
> > become the cause of adding dentry to hash twice.
> > 
> > This explains the reason of Oops (dereference the freed dentry in
> > __d_lookup()) on my machine.
> > 
> > Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
> > ---
> > 
> >  fs/nfs/dir.c |    1 +
> >  1 file changed, 1 insertion(+)
> > 
> > diff -puN fs/nfs/dir.c~nfs-d_rehash-fix fs/nfs/dir.c
> > --- linux-2.6/fs/nfs/dir.c~nfs-d_rehash-fix	2009-12-28 06:18:09.000000000 +0900
> > +++ linux-2.6-hirofumi/fs/nfs/dir.c	2009-12-28 06:18:16.000000000 +0900
> > @@ -1615,6 +1615,7 @@ static int nfs_rename(struct inode *old_
> >  				goto out;
> >  
> >  			new_dentry = dentry;
> > +			rehash = NULL;
> >  			new_inode = NULL;
> >  		}
> >  	}
> 
> Guys, what's the status of this fix?  Did Marvin have a chance to test
> it?  Are the NFS developers aware of it?
> 
> Thanks.
> 

Sorry for the delay. The above fix looks correct to me, but I too would
like a confirmation that it fixes the Oops before I push it to Linus.

In the meantime, I've committed it to my linux-next branch.

Cheers
  Trond

Re: vfs related crash in 2.6.33-rc2

[Marvin]

Hi,

> On Wed, 2010-01-06 at 15:41 -0800, Andrew Morton wrote:
> > On Thu, 31 Dec 2009 05:59:32 +0900
> >
> > OGAWA Hirofumi <hirofumi@mail.parknet.co.jp> wrote:
> > > Marvin <marvin24@gmx.de> writes:
> > > >> Marvin <marvin24@gmx.de> writes:
> > > >> > Hi,
> > > >> >
> > > >> > I'm getting a lot of these:
> > > >> >
> > > >> > kernel: general protection fault: 0000 [#1] SMP
> > > >> > kernel: last sysfs file:
> > > >> > /sys/devices/pci0000:00/0000:00:18.3/modalias kernel: CPU 0
> > > >> > kernel: Pid: 12177, comm: packagekitd Not tainted 2.6.33-rc2 #1
> > > >> > ...
> > > >> >
> > > >> > filesystem is ext4 (in case it matters).
> > > >>
> > > >> BTW, are you using nfs client on this machine?
> > > >
> > > > um - yes, now that I think about it... I killed a nfs umount process
> > > > (because of an offline server) shortly before the oopses started to
> > > > fire.
> > >
> > > OK. Probably, this oops would be same with one which happened on my
> > > machine recently. That path in patch corrupts dcache hash, so it can be
> > > the cause of strange behavior or oops on dcache hash.
> > >
> > > If so, the attached patch would fix it.
> > >
> > > Thanks.
> >
> > Guys, what's the status of this fix?  Did Marvin have a chance to test
> > it?  Are the NFS developers aware of it?
> >
> > Thanks.
> 
> Sorry for the delay. The above fix looks correct to me, but I too would
> like a confirmation that it fixes the Oops before I push it to Linus.
> 
> In the meantime, I've committed it to my linux-next branch.

It seems that I send the reply to Hirofumi only, sorry for that. The patch works fine 
- no oops anymore.

Thanks

Marvin

Re: vfs related crash in 2.6.33-rc2

[Trond Myklebust]

On Thu, 2010-01-07 at 10:27 +0100, Marvin wrote: 
> It seems that I send the reply to Hirofumi only, sorry for that. The patch works fine 
> - no oops anymore.

OK. Thanks for testing!

Trond