Re: [BUG] 2.5.63: ESR killed my box!

["Martin J. Bligh" <mbligh@aracnet.com>]

> No, look further down in apic.c: APIC_init_uniprocessor() calls 
> setup_IO_APIC() which then calls setup_ioapic_ids_from_mpc(). This
> happens  after the code quoted above runs.

OK, sounds good - I might be twisting it with the SMP code, which I'm much
more familiar with.
 
>> boot_cpu_physical_apicid = hard_smp_processor_id();
>> phys_cpu_present_map = 1 << boot_cpu_physical_apicid;
> 
> But is it necessarily true that hard_smp_processor_id() equals the APIC
> id?

Well it should be. Except that someone did this for some odd reason
#define hard_smp_processor_id()                 0
on non-SMP rather than non-local-apic.

But we could just substitute the real code:
static __inline int hard_smp_processor_id(void)
{
        /* we don't want to mark this access volatile - bad code generation
*/
        return GET_APIC_ID(*(unsigned long *)(APIC_BASE+APIC_ID));
}
which is in smp.h, but wrapped in #ifdef CONFIG_X86_LOCAL_APIC.
All very twisted. Probably not worth it.

> and that's just about as good as it gets, certainly the CPU knows best 
> what its APIC id is.

Absolutely ;-)

M.