Re: [lhcs-devel] Re: CPU Hotplug broken -mm5 onwards

[Nick Piggin <nickpiggin@yahoo.com.au>]

Srivatsa Vaddagiri wrote:
> On Mon, Apr 19, 2004 at 01:34:14PM +1000, Nick Piggin wrote:
> 
>>I think a rwsem might be a good idea anyway, because
>>sched_migrate_task can end up being called pretty often with
>>balance on exec and balance on clone. The semaphore could easily
>>place undue serialisation on that path.
> 
> 
> I found that r/w sem does not help here ..It can still lead to deadlocks.
> One example I hit is :
> 
> cpu_up takes write lock, sends out CPU_UP_PREPARE notification. As part
> of it, many do kthread_create, which uses workqueue. The work function
> is never processed because keventd would be blocked on a previous 
> work function, waiting for hotplug sem in exec path.
> 
> So, as Rusty said, I think we really need to consider removing
> lock_cpu_hotplug from sched_migrate_task. AFAICS that lock
> was needed to prevent adding tasks to dead cpus. The same 
> can be accomplished by removing lock_cpu_hotplug from sched_migrate_task
> and adding a cpu_is_offline check in __migrate_task.
> This will eliminate all the deadlocks I have been hitting.
> 

Yes this would be a better idea. Care to send Andrew a patch
against -mm?

> 
> 
>>Can we arrange some of these checks to disappear when HOTPLUG_CPU
>>is not set? For example, make cpu_is_offline only valid to call for
>>CPUs that have been online sometime, and can evaluate to 0 if
>>HOTPLUG_CPU is not set?
> 
> 
> I think this is already being done in include/linux/cpu.h
> 

Yes I see. I didn't realise the first one was under an ifdef :P
Sorry.