Re: [RFC] Thread Migration Preemption

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

Matt Mackall wrote:
> On Fri, Jul 06, 2007 at 04:12:10PM +1000, Nick Piggin wrote:
> 
>>Mathieu Desnoyers wrote:
>>
>>>Thread Migration Preemption
>>>
>>>This patch adds the ability to protect critical sections from migration to
>>>another CPU without disabling preemption.
>>>
>>>This will be useful to minimize the amount of preemption disabling for the 
>>>-rt
>>>patch. It will help leveraging improvements brought by the local_t types in
>>>asm/local.h (see Documentation/local_ops.txt). Note that the updates done 
>>>to
>>>variables protected by migration_disable must be either atomic or 
>>>protected from
>>>concurrent updates done by other threads.
>>>
>>>Typical use:
>>>
>>>migration_disable();
>>>local_inc(&__get_cpu_var(&my_local_t_var));
>>>migration_enable();
>>>
>>>Which will increment the variable atomically wrt the local CPU.
>>>
>>>Comments (such as how to integrate this in the already almost full
>>>preempt_count) are welcome.
>>
>>This seems like way too much stuff to add just for this type of thing. Why
>>not just disable and reenable preempt? Surely local_inc is not going to take
>>so long that disabling preemption matters.
> 
> 
> I like this patch a lot. Even if we don't add the underlying mechanism
> right now, adding migration_disable as an alias for preempt_disable
> will much better document quite a number of the users.

I'd have no problem with that, and it might make it easier in future to
justify a more complex scheme.


>>The task struct is not something we should just be carefree putting crap
>>into because it is seemingly free :(
> 
> 
> Sadly, it is free at the moment. We can only fit 3 task_structs in an order-1 SLAB,
> with lots of slop.

Well apart from more cacheline access, that's why I say seemingly free. But
actually it uses up space we may like for something else in future and/or
makes it harder to shrink down if any effort ever goes into that.

-- 
SUSE Labs, Novell Inc.