Re: two lockups with 2.4.25

[Marcelo Tosatti <marcelo.tosatti@cyclades.com>]

On Wed, Apr 21, 2004 at 08:19:20AM -0700, Chris Stromsoe wrote:
> On Wed, 21 Apr 2004, Oleg Drokin wrote:
> 
> > Chris Stromsoe <cbs@cts.ucla.edu> wrote:
> > CS> Pid: 26885, comm:               sophie
> > >>>EIP; c0110ed7 <flush_tlb_others+9b/bc>   <=====
> > >>>EDX; 01000000 Before first symbol
> > >>>ESI; f4762cc0 <_end+343c85cc/3896e90c>
> > >>>EDI; 081a1d28 Before first symbol
> > >>>EBP; e806fe94 <_end+27cd57a0/3896e90c>
> > CS> Trace; c011100f <flush_tlb_page+6f/7c>
> > CS> Trace; c01259b7 <do_wp_page+223/284>
> > CS> Trace; c01260de <handle_mm_fault+82/b8>
> > CS> Trace; c01132f9 <do_page_fault+1a1/4ed>
> > CS> Trace; c0113158 <do_page_fault+0/4ed>
> > CS> Trace; c0224ce1 <__kfree_skb+129/134>
> > CS> Trace; c01145e3 <schedule+45b/520>
> > CS> Trace; c0106fc4 <error_code+34/3c>
> >
> > This backtrace is suspiciously similar to a backtrace from NMI WD I had
> > not so long ago. Also on 2.5.25 Are your boxes SMP?
> 
> The boxes are SMP (dual cpu p3 1400MHz).

Oleg,

AFAICS this backtrace is fine -- flush_tlb_others() does invplg and thats it. No
locking involved, yes?

I can see very interesting JFS traces in Chris's report. I'll answer his message.