Re: 2.6.33.5 rt23: machine lockup (nfs/autofs related?)

[john stultz <johnstul@us.ibm.com>]

On Wed, 2010-07-14 at 14:32 -0700, Fernando Lopez-Lezcano wrote:
> On Mon, 2010-07-12 at 20:06 -0700, Fernando Lopez-Lezcano wrote:
> > > > --------
> > > > BUG: unable to handle kernel NULL pointer dereference at 0000008c
> > > > IP: [<c045e50a>] rt_spin_lock_fastunlock.clone.2+0x6/0x3e
> > > ...
> > > > Pid: 2855, comm: nautilus Not tainted
> > > > 2.6.33.6-147.rt23.3.fc12.ccrma.i686.rt #3 P5K/EPU/P5K/EPU
> > > > EIP: 0060:[<c045e50a>] EFLAGS: 00210246 CPU: 0
> > > > EIP is at rt_spin_lock_fastunlock.clone.2+0x6/0x3e
> > > > EAX: 00000078 EBX: ef45393c ECX: 00000000 EDX: 00000078
> > > > ESI: ef716edc EDI: 00000000 EBP: f1977c8c ESP: f1977c88
> > > >  DS: 007b ES: 007b FS: 00d8 GS: 00e0 SS: 0068 preempt:00000000
> > > > Process nautilus (pid: 2855, ti=f1976000 task=f2347130 task.ti=f1976000)
> > > > Stack:
> > > >  ef45393c f1977c94 c0781206 f1977cc8 c04d842e 00000000 ef703e54 faadd5bc
> > > > <0> f1977cdc 126bc87a ef703ddc 00000000 ef45393c ef716edc ef6f5494
> > > > faafdc6c
> > > > <0> f1977df8 faad9041 c3604b5c faafdc6c ef452bfc 00007e7f f5eb41e8
> > > > 00000007
> > > > Call Trace:
> > > >  [<c0781206>] ? rt_spin_unlock+0x8/0xa
> > > >  [<c04d842e>] ? d_materialise_unique+0x210/0x2aa
> > > 
> > > Can you gdb list *0xc04d842e ?
> > 
> > (gdb) list *0xc04d842e
> > 0xc04d842e is in d_materialise_unique (fs/dcache.c:2073).
> > 2068	out_unalias:
> > 2069		d_move_locked(alias, dentry);
> > 2070		ret = alias;
> > 2071	out_err:
> > 2072		spin_unlock(&inode->i_lock);
> > 2073		if (m2)
> > 2074			mutex_unlock(m2);
> > 2075		if (m1)
> > 2076			mutex_unlock(m1);
> > 2077		return ret;
> > 
> > > Thanks again for all the testing here! Its really appreciated!
> 
> I just tried building 2.6.33.6 rt26 (saw it this morning on the site)
> and it does not exhibit this problem. Woohoo!

Yea. The vfs-scalability patches were dropped. Sort of a bummer, but I'm
glad that fixes it for you.

We may run into the issue again when the patches can get beaten back
into shape, but I'll try to hunt down the issue from your log output
prior to that.

Thanks again for the testing and great feedback!
-john