Re: kdump on non-boot cpu

[Vivek Goyal <vgoyal@in.ibm.com>]

On Thu, 2005-02-03 at 10:42, Itsuro Oda wrote:
> Hi,
> 
> I found the following in an old mail:
> 
> >From vgoyal at in.ibm.com  Thu Jan  6 07:20:43 2005
> ...
> >2. Kdump can possibly fail on SMP machines if crash occurs on non-boot
> >cpu. Hari is finalizing the stop gap patch to handle this problem.
> 
> Is this finished ?  (It seems it is not in 2.6.11-rc2-mm1.)

Not yet. For the time being focus got shifted to other kdump issues. I
am not even sure if this is a problem. See below a clip from discussions
on fastboot.

> Hi Eric,
> > 
> > I had a quick look at kexec3. Had some queries.
> > 
> > 1. Code for relocating to boot cpu or enabling boot from non-boot cpu is
> > required.
> 
> Actually I just looked and it appears this snippet from smp_boot_cpus
> already handles that case.
> 
>         boot_cpu_physical_apicid = GET_APIC_ID(apic_read(APIC_ID));
>         boot_cpu_logical_apicid = logical_smp_processor_id();
>         x86_cpu_to_apicid[0] = boot_cpu_physical_apicid;
> 
> While looking I certainly did not see anything still in the
> kernel that would complain if we get this wrong.
> 
> Although I  am not really comfortable with a capture kernel using
> multiprocessors. 
> 
> Eric


Do you see a problem in the code flow somewhere?

Vivek