Re: Adding a printk in start_secondary breaks 2.4.10, not 2.4.9 ??

[Martin J. Bligh]

>> FWIW, I still think that means that the console locking changes are
>> broken  - adding a printk shouldn't panic the kernel. I'll go look at
>> the console locking changes (*and* fix my disgusting hack ;-) )
>
> I suspect the panic has nothing to do with adding the printk, but merely
> that the timing patterns of your disgusting hack have changed
>
> Happy logical analysers 8)

OK, you're right ;-) It turned out to be a race between init_idle and
reschedule_idle, that is a generic bug, but is revealed by the timing
changes that my printk introduces (both in standard SMP and NUMA
kernels, if you turn on serial consoles).

See imminent post for a patch. Having fixed the real problem, I can
now axe that disgusting hack ;-)

Thanks for your help,

Martin.