Re: [PATCH 1/3] cpuset: clean up cpuset hierarchy traversal code

[Max Krasnyansky <maxk@qualcomm.com>]

Paul Jackson wrote:
> Li Zefan wrote:
>> -	q = kfifo_alloc(number_of_cpusets * sizeof(cp), GFP_KERNEL, NULL);
> 
> 
> The block comment for rebuild_sched_domains() states:
> 
>> ... May take callback_mutex during
>> * call due to the kfifo_alloc() and kmalloc() calls.
> 
> I suspect that mention of kfifo_alloc() is no longer correct,
> with your patch.  If so, perhaps you could send a little additional
> fix patch, to remove that mention from the comment.
> 

Paul, please take a look at
	cpuset: Rework sched domains and CPU hotplug handling
patch I sent out last week.
I'd appreciate if we applied that one first. It simplifies lock nesting 
and rearranges the way sched domains are rebuilt. It is IMO a bit higher 
priority than this patch because scheduler depends on the 
rebuild_sched_domains() call and we have to call rebuild_sched_domains() 
from cpu hotplug handlers.

Max