Re: [RFC][PATCH 17/18] sched: Move the second half of ttwu() tothe remote cpu
From: Oleg Nesterov
Date: Fri Jan 07 2011 - 10:29:53 EST
On 01/06, Peter Zijlstra wrote:
> On Wed, 2011-01-05 at 22:07 +0100, Oleg Nesterov wrote:
> > I'll try to read it once again with the fresh head, though ;)
> > I also have a couple of very minor nits... In particular, perhaps
> > TASK_WAKING can die...
> I think it might.. I'll do a patch at the end removing it, lets see what
Yes, ttwu can just set TASK_RUNNING. But, otoh, perhaps the
special state makes sense anyway, say, it can help to debug
the problems. We can even have TASK_WAKING_CONTRIBUTES_TO_LOAD
insetad of ->sched_contributes_to_load. But this all is very
A couple of questions...
Why sched_fork() does set_task_cpu() ? Just curious, it seems
that wake_up_new_task() does all we need.
ttwu_queue_remote() does "struct task_struct *next = NULL".
Probably "next = rq->wake_list" makes more sens. Otherwise the
first cmpxchg() always fails if rq->wake_list != NULL.
Doesn't __migrate_task() need pi_lock? Consider:
1. A task T runs on CPU_0, it does set_current_state(TASK_INTERRUBTIBLE)
2. some CPU does set_cpus_allowed_ptr(T, new_mask), new_mask doesn't
T is running, cpumask_any_and() picks CPU_1, set_cpus_allowed_ptr()
drops pi_lock and rq->lock before stop_one_cpu().
3. T calls schedule() and becomes deactivated.
4. CPU_2 does try_to_wake_up(T, TASK_INTERRUPTIBLE), takes pi_lock
and sees on_rq == F.
5. set_cpus_allowed_ptr() resumes and calls stop_one_cpu(cpu => 1).
6. cpu_stopper_thread() runs on CPU_1 and calls ____migrate_task().
It locks CPU_0 and CPU_1 rq's and checks task_cpu() == src_cpu.
7. CPU_2 calls select_task_rq(), it returns (to simplify) 2.
Now try_to_wake_up() does set_task_cpu(T, 2), and calls
8. __migrate_task() on CPU_1 sees p->on_rq and starts the
deactivate/activate dance racing with ttwu_do_activate()
And a final question. This is really, really minor, but
activate_task/deactivate_task are not symmetric, the former
always sets p->on_rq. Looks correct, but imho a bit confusing and
can complicate the understanding. Since p->on_rq is cleared
explicitly by schedule(), perhaps it can be set explicitly to
in try_to_wake_up_*. Or, perhaps, activate/deactivate can check
ENQUEUE_WAKEUP/DEQUEUE_SLEEP and set/clear p->on_rq. Once again,
this is purely cosmetic issue.
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/