The irq work interrupt is optional: If an architecture does not provide
it then the callback will be run from update_process_times(). This does
not work well for nohz_full_kick_work_func() as we learned in commit
76a33061b932 ("irq_work: Force raised irq work to run on irq work
interrupt").
This patch is just a precaution to ensure that that an irq work
interrupt is required and the timer callback is not enough.
Signed-off-by: Sebastian Andrzej Siewior <[email protected]>
---
kernel/time/tick-sched.c | 5 +++++
1 file changed, 5 insertions(+)
diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index c792429e98c6..d9481bebae22 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -176,6 +176,11 @@ static bool can_stop_full_tick(void)
return false;
}
+ if (!arch_irq_work_has_interrupt()) {
+ trace_tick_stop(0, "missing irq work interrupt\n");
+ return false;
+ }
+
/* sched_clock_tick() needs us? */
#ifdef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK
/*
--
2.1.4
On Tue, Jun 23, 2015 at 04:07:43PM +0200, Sebastian Andrzej Siewior wrote:
> The irq work interrupt is optional: If an architecture does not provide
> it then the callback will be run from update_process_times(). This does
> not work well for nohz_full_kick_work_func() as we learned in commit
> 76a33061b932 ("irq_work: Force raised irq work to run on irq work
> interrupt").
>
> This patch is just a precaution to ensure that that an irq work
> interrupt is required and the timer callback is not enough.
>
> Signed-off-by: Sebastian Andrzej Siewior <[email protected]>
But we already check that in tick_nohz_init(). Or am I missing something
else?
Thanks.
> ---
> kernel/time/tick-sched.c | 5 +++++
> 1 file changed, 5 insertions(+)
>
> diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
> index c792429e98c6..d9481bebae22 100644
> --- a/kernel/time/tick-sched.c
> +++ b/kernel/time/tick-sched.c
> @@ -176,6 +176,11 @@ static bool can_stop_full_tick(void)
> return false;
> }
>
> + if (!arch_irq_work_has_interrupt()) {
> + trace_tick_stop(0, "missing irq work interrupt\n");
> + return false;
> + }
> +
> /* sched_clock_tick() needs us? */
> #ifdef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK
> /*
> --
> 2.1.4
>
On 07/07/2015 04:16 PM, Frederic Weisbecker wrote:
> But we already check that in tick_nohz_init(). Or am I missing something
> else?
Ah indeed, thanks. Haven't seen it.
>
> Thanks.
Sebastian