Re: 2.6.16-rt10 crash on ppc

[Kumar Gala <galak@kernel.crashing.org>]

On Mar 30, 2006, at 1:13 AM, Ingo Molnar wrote:

>
> * Nick Piggin <nickpiggin@yahoo.com.au> wrote:
>
>> Yes, that patch is basically what I had in mind.
>>
>> Is -rt ever allocating memory from really-hard-don't-preempt-me
>> context? I guess not, unless the zone->lock is one of those locks  
>> too,
>> right?
>
> no. zone->lock (and all the slab locks, and all the other MM locks)  
> are
> fully preemptible too.
>
>> Should you add a
>>
>>  #else
>>     BUG_ON(_really_dont_preempt_me());
>>  #endif
>>
>> just for safety, or will such misusage get caught elsewhere (eg. when
>> attempting to take zone->lock).
>
> it should be caught immediately, by the cond_resched().

The issue me actually be a driver interrupt locking bug.  The driver  
supports three distinct interrupts for TX, RX, Error.  I asked Emin  
to try changing the driver to use SA_INTERRUPT in the request_irq()  
to see what happens.  I believe that when he did that it worked but  
hurts performance.

- kumar