Re: [PATCH 1/5] irq_work: Architecture support for remote irq work raise

[Peter Zijlstra]

On Mon, May 12, 2014 at 06:26:49PM +0200, Frederic Weisbecker wrote:
> On Mon, May 12, 2014 at 09:56:50AM +0200, Peter Zijlstra wrote:
> > On Mon, May 12, 2014 at 01:33:53AM +0200, Frederic Weisbecker wrote:
> > > We are going to extend irq work to support remote queuing.
> > > 
> > > So lets add a cpu argument to arch_irq_work_raise(). The architectures
> > > willing to support that must then provide the backend to raise irq work
> > > IPIs remotely.
> > > 
> > > Initial support is provided for x86 and ARM since they are easily
> > > extended. The other archs that overwrite arch_irq_work_raise() seem
> > > to use local clock interrupts and therefore need deeper rewrite of their
> > > irq work support to implement remote raising.
> > > 
> > 
> > > Signed-off-by: Frederic Weisbecker <fweisbec@xxxxxxxxx>
> > 
> > Why not borrow the smp_call_function IPI for the remote bits? We could
> > limit the 'safe from NMI' to the local works. And we validate this by
> > putting a WARN_ON(in_nmi()) in irq_work_queue_on().
> 
> Right, but although I don't need it to be safe from NMI, I need it
> to be callable concurrently and when irqs are disabled.
> 
> So we can't use smp_call_function_single() for that. But we can use the async
> version in which case we must keep the irq work claim. But that's
> about the same than smp_queue_function_single() we had previously
> and we are back with our csd_lock issue.

Who said anything about using smp_call_function_single()?


---
diff --git a/kernel/irq_work.c b/kernel/irq_work.c
index a82170e2fa78..2fc9d8ece05a 100644
--- a/kernel/irq_work.c
+++ b/kernel/irq_work.c
@@ -61,7 +61,8 @@ void __weak arch_irq_work_raise(void)
  *
  * Can be re-enqueued while the callback is still in progress.
  */
-bool irq_work_queue(struct irq_work *work)
+static __always_inline bool
+__irq_work_queue_on(struct irq_work *work, int cpu)
 {
 	/* Only queue if not already pending */
 	if (!irq_work_claim(work))
@@ -78,16 +79,31 @@ bool irq_work_queue(struct irq_work *work)
 	 * for the next tick.
 	 */
 	if (!(work->flags & IRQ_WORK_LAZY) || tick_nohz_tick_stopped()) {
-		if (!this_cpu_cmpxchg(irq_work_raised, 0, 1))
-			arch_irq_work_raise();
+		if (cmpxchg(&__get_cpu_var(irq_work_raised, 0, 1) == 0)) {
+			if (cpu == smp_processor_id() || cpu == -1)
+				arch_irq_work_raise();
+			else
+				arch_send_call_function_single_ipi();
+		}
 	}
 
 	preempt_enable();
 
 	return true;
 }
+
+bool irq_work_queue(struct irq_work *work)
+{
+	return __irq_work_queue_on(work, -1);
+}
 EXPORT_SYMBOL_GPL(irq_work_queue);
 
+bool irq_work_queue_on(struct irq_work *work, int cpu)
+{
+	WARN_ON_ONCE(in_nmi());
+	return __irq_work_queue_on(work, cpu);
+}
+
 bool irq_work_needs_cpu(void)
 {
 	struct llist_head *this_list;
diff --git a/kernel/smp.c b/kernel/smp.c
index 06d574e42c72..0fd53963c4fb 100644
--- a/kernel/smp.c
+++ b/kernel/smp.c
@@ -198,6 +198,12 @@ void generic_smp_call_function_single_interrupt(void)
 		csd->func(csd->info);
 		csd_unlock(csd);
 	}
+
+	/*
+	 * First run the synchronous callbacks, people are waiting on them;
+	 * then run the async ones.
+	 */
+	irq_work_run();
 }
 
 /*
--
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/