Re: 2.6.15-rt16

[Clark Williams <williams@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 1817 bytes --]

On Tue, 2006-01-31 at 09:52 -0800, chris perkins wrote:
> > <snip>
> >
> >> CONFIG_LATENCY_TIMING=y
> >
> > I'm betting this is the same thing I'm seeing. Are you running on a
> > uniprocessor x86_64? And if so are you seeing messages similar to the
> > following?
> >
> > init[1]: segfault at ffffffff8010fadc rip ffffffff8010fadc rsp
> > 00007fffffdacfc8
> >
> > If so, then I suspect that you're getting a segfault in ld.so (at least
> > that's the furthest I've gotten so far). Something about how the kernel
> > sets up the memory map is upsetting dynamically loaded executables. I
> > can boot with init=/sbin/sash, but when I try and run a dynamically
> > linked program, I get segfaults.
> >
> > You might try turning off LATENCY_TRACING and see if that allows you to
> > boot and run (works for me).
> >
> > Meanwhile, I'm going to try and pin this down to something better than
> > "somewhere in ld.so..."
> >
> > Clark
> > -
> > Clark Williams <williams@redhat.com>
> 
> hi,
>    actually i'm running on a dual processor x86_64. with the problem i was 
> having, i never got far enough to see the message you asked about. Steven 
> Rostedt's suggestion to turn off NUMA worked and i am now able to boot. 
> However, if I turn LATENCY_TRACING on, i get an immediate reboot after the 
> kernel is uncompressed. this doesn't sound like the same problem you're 
> having, though.

Man, I saw that LATENCY_TRACING and completely missed the NUMA
parameter. Sorry about that...

I must be the only person running a uniprocessor x86_64 that's working
with the -rt patches and trying to trace latencies. If I turn off
LATENCY_TRACING, I boot just fine. 

Are you getting any output from the kernel before the reboot?

Clark

-- 
Clark Williams <williams@redhat.com>

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]