Re: 2.6.19-rc5-mm1: HPC nx6325 breakage, VESA fb problem, md-raid problem

[Andrew Morton <akpm@osdl.org>]

On Fri, 10 Nov 2006 05:49:08 +0100
Andi Kleen <ak@suse.de> wrote:

> 
> > > > 
> > > > Well, I've got some data from earlyprintk (forgot I needed to boot with
> > > > vga=normal).
> > > > 
> > > > Unfortunately, I had to rewrite the trace manually:
> > > > 
> > > > clear_IO_APIC_pin+0x15/0x6a
> > > > try_apic_pin+0x7a/0x98
> > > > setup_IO_APIC+0x600/0xb7a
> > > > smp_prepare_cpus+0x33a/0x371
> > > > init+0x60/0x32d
> > > > child_rip+0xa/0x12
> > > > 
> > > > [And then the unwinder said it got stuck.]
> > > > 
> > > > RIP is reported to be at ioapic_read_entry+0x33/0x61,
> > > 
> > > This is 100% reproducible on the nx6325 (but not on the other boxes) and
> > > apparently caused by x86_64-mm-try-multiple-timer-pins.patch (doesn't
> > > happen with this patch reverted).
> > 
> > Thanks, dropped.
> 
> can I have details please? 

I think what's in this thread is all you'll get.

It would be nice to see the access address.  I'd be guessing that it's
trying to read the io-apic before we're ready to read it and io_apic_base()
is returning gunk and boom.

> On what system (CPU, motherboard, BIOS version) does the noidlehz stuff break?

nx6325

It's x86_64: no noidlehz.

> And what did you drop exactly?

x86_64-mm-try-multiple-timer-pins.patch