Re: [PATCH 2/6] irq_work: Implement remote queueing

[Peter Zijlstra]

On Wed, Jun 25, 2014 at 07:12:34AM +0200, Peter Zijlstra wrote:
> On Tue, Jun 24, 2014 at 02:33:41PM -0600, Stephen Warren wrote:
> > On 06/10/2014 09:15 AM, Frederic Weisbecker wrote:
> > > irq work currently only supports local callbacks. However its code
> > > is mostly ready to run remote callbacks and we have some potential user.
> > > 
> > > The full nohz subsystem currently open codes its own remote irq work
> > > on top of the scheduler ipi when it wants a CPU to reevaluate its next
> > > tick. However this ad hoc solution bloats the scheduler IPI.
> > > 
> > > Lets just extend the irq work subsystem to support remote queuing on top
> > > of the generic SMP IPI to handle this kind of user. This shouldn't add
> > > noticeable overhead.
> > 
> > I'm running next-20140624 on an ARM system, and this patch causes CPU
> > hot(un)plug to Oops for me; the following fires:
> > 
> > void irq_work_run(void)
> > {
> > 	BUG_ON(!in_irq());
> > 
> > I found that Linus's master (8b8f5d971584 "Merge tag 'compress-3.16-rc3'
> > of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core")
> > works fine. I found that this commit inside the tip(?) tree works fine
> > (478850160636 "irq_work: Implement remote queueing"). However, if I
> > merge the two together, I hit that BUG_ON.
> > 
> > I think the issue is:
> > 
> > This commit adds a call from
> > generic_smp_call_function_single_interrupt() to irq_work_run().
> > 
> > Srivatsa's patch adds a call from hotplug_cfd() to
> > flush_smp_call_function_queue() to, which I imagine happens in
> > non-interrupt context. Note that this patch moves most of the body of
> > generic_smp_call_function_single_interrupt() into
> > flush_smp_call_function_queue().
> 
> Right you are.. I think I'll just remove the BUG_ON(), Frederic?

Something a little so like:

---
Subject: irq_work: Remove BUG_ON in irq_work_run_list()
From: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
Date: Wed Jun 25 07:13:07 CEST 2014

Because of a collision with 8d056c48e486 ("CPU hotplug, smp: flush any
pending IPI callbacks before CPU offline"), which ends up calling
hotplug_cfd()->flush_smp_call_function_queue()->run_irq_work(), which
is not from IRQ context.

Cc: Frederic Weisbecker <fweisbec@xxxxxxxxx>
Reported-by: Stephen Warren <swarren@xxxxxxxxxxxxx>
Signed-off-by: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
---
 kernel/irq_work.c |    2 --
 1 file changed, 2 deletions(-)

--- a/kernel/irq_work.c
+++ b/kernel/irq_work.c
@@ -130,8 +130,6 @@ static void irq_work_run_list(struct lli
 	struct irq_work *work;
 	struct llist_node *llnode;
 
-	BUG_ON(!irqs_disabled());
-
 	if (llist_empty(list))
 		return;

Attachment: pgpIJ6eK_yvQt.pgp
Description: PGP signature