Re: [PATCH] SN2 user-MMIO CPU migration

[Jack Steiner]

On Thu, Jan 19, 2006 at 06:18:43PM -0800, Jesse Barnes wrote:
> On Thursday, January 19, 2006 4:06 pm, Brent Casavant wrote:
> >  #ifndef __ARCH_WANT_UNLOCKED_CTXSW
> >  static inline int task_running(runqueue_t *rq, task_t *p)
> > @@ -936,6 +939,7 @@ static int migrate_task(task_t *p, int d
> >  	 * it is sufficient to simply update the task's cpu field.
> >  	 */
> >  	if (!p->array && !task_running(rq, p)) {
> > +		arch_task_migrate(p);
> >  		set_task_cpu(p, dest_cpu);
> >  		return 0;
> >  	}
> > @@ -1353,6 +1357,7 @@ static int try_to_wake_up(task_t *p, uns
> >  out_set_cpu:
> >  	new_cpu = wake_idle(new_cpu, p);
> >  	if (new_cpu != cpu) {
> > +		arch_task_migrate(p);
> >  		set_task_cpu(p, new_cpu);
> >  		task_rq_unlock(rq, &flags);
> >  		/* might preempt at this point */
> > @@ -1876,6 +1881,7 @@ void pull_task(runqueue_t *src_rq, prio_
> >  {
> >  	dequeue_task(p, src_array);
> >  	dec_nr_running(p, src_rq);
> > +	arch_task_migrate(p);
> >  	set_task_cpu(p, this_cpu);
> >  	inc_nr_running(p, this_rq);
> >  	enqueue_task(p, this_array);
> > @@ -4547,6 +4553,7 @@ static void __migrate_task(struct task_s
> >  	if (!cpu_isset(dest_cpu, p->cpus_allowed))
> >  		goto out;
> >
> > +	arch_task_migrate(p);
> >  	set_task_cpu(p, dest_cpu);
> >  	if (p->array) {
> >  		/*
> 
> Maybe you could just turn the above into mmiowb() calls instead?  That 
> would cover altix, origin, and ppc as well I think.  On other platforms 
> it would be a complete no-op.
> 
> Jesse

I don't think calling mmiob() directly would work. In order to make 
CONFIG_IA64_GENERIC work, the call to mmiob() needs to be underneath a
platform vector. Using ia64_platform_is() would also work but I think
a platform vector is cleaner.

A second reason for an arch_task_migrate() instead of a specific mmiob() is
to provide a hook for a future platform that require additional work
to be done when a task migrates. 


-- 
Thanks

Jack Steiner (steiner@xxxxxxx)          651-683-5302


-
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/