[patch V2 06/10] timekeeping: Distangle resume and clock-was-set events

From: Thomas Gleixner
Date: Tue Jul 13 2021 - 09:54:44 EST


From: Thomas Gleixner <tglx@xxxxxxxxxxxxx>

Resuming timekeeping is a clock-was-set event and uses the clock-was-set
notification mechanism. This is in the way of making the clock-was-set
update for hrtimers selective so unnecessary IPIs are avoided when a CPU
base does not have timers queued which are affected by the clock setting.

Distangle it by invoking hrtimer_resume() on each unfreezing CPU and invoke
the new timerfd_resume() function from timekeeping_resume() which is the
only place where this is needed.

Rename hrtimer_resume() to hrtimer_resume_local() to reflect the change.

With this the clock_was_set*() functions are not longer required to IPI all
CPUs unconditionally and can get some smarts to avoid them.

Signed-off-by: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
---
include/linux/hrtimer.h | 1 -
kernel/time/hrtimer.c | 15 ++++++---------
kernel/time/tick-common.c | 7 +++++++
kernel/time/tick-internal.h | 2 ++
kernel/time/timekeeping.c | 4 +++-
5 files changed, 18 insertions(+), 11 deletions(-)

--- a/include/linux/hrtimer.h
+++ b/include/linux/hrtimer.h
@@ -354,7 +354,6 @@ extern void timerfd_resume(void);
static inline void timerfd_clock_was_set(void) { }
static inline void timerfd_resume(void) { }
#endif
-extern void hrtimers_resume(void);

DECLARE_PER_CPU(struct tick_device, tick_cpu_device);

--- a/kernel/time/hrtimer.c
+++ b/kernel/time/hrtimer.c
@@ -900,8 +900,8 @@ static void clock_was_set_work(struct wo
static DECLARE_WORK(hrtimer_work, clock_was_set_work);

/*
- * Called from timekeeping and resume code to reprogram the hrtimer
- * interrupt device on all cpus and to notify timerfd.
+ * Called from timekeeping code to reprogram the hrtimer interrupt device
+ * on all cpus and to notify timerfd.
*/
void clock_was_set_delayed(void)
{
@@ -909,18 +909,15 @@ void clock_was_set_delayed(void)
}

/*
- * During resume we might have to reprogram the high resolution timer
- * interrupt on all online CPUs. However, all other CPUs will be
- * stopped with IRQs interrupts disabled so the clock_was_set() call
- * must be deferred.
+ * Called during resume either directly from via timekeeping_resume()
+ * or in the case of s2idle from tick_unfreeze() to ensure that the
+ * hrtimers are up to date.
*/
-void hrtimers_resume(void)
+void hrtimers_resume_local(void)
{
lockdep_assert_irqs_disabled();
/* Retrigger on the local CPU */
retrigger_next_event(NULL);
- /* And schedule a retrigger for all others */
- clock_was_set_delayed();
}

/*
--- a/kernel/time/tick-common.c
+++ b/kernel/time/tick-common.c
@@ -470,6 +470,13 @@ void tick_resume_local(void)
else
tick_resume_oneshot();
}
+
+ /*
+ * Ensure that hrtimers are up to date and the clockevents device
+ * is reprogrammed correctly when high resolution timers are
+ * enabled.
+ */
+ hrtimers_resume_local();
}

/**
--- a/kernel/time/tick-internal.h
+++ b/kernel/time/tick-internal.h
@@ -168,3 +168,5 @@ void timer_clear_idle(void);

void clock_was_set(void);
void clock_was_set_delayed(void);
+
+void hrtimers_resume_local(void);
--- a/kernel/time/timekeeping.c
+++ b/kernel/time/timekeeping.c
@@ -1810,8 +1810,10 @@ void timekeeping_resume(void)

touch_softlockup_watchdog();

+ /* Resume the clockevent device(s) and hrtimers */
tick_resume();
- hrtimers_resume();
+ /* Notify timerfd as resume is equivalent to clock_was_set() */
+ timerfd_resume();
}

int timekeeping_suspend(void)