sched/clock: expire timer in hardirq context
PATCH 1/1 sched/clock: expire timer in hardirq context
To minimize latency, PREEMPT_RT kernels expires hrtimes in preemptible softirq context by default. This can be overriden by marking the timer's expiry with HRTIMER_MODE_HARD.
sched_clock_timer is missing this annotation: if its callback is preempted and the duration of the preemption exceeds the wrap around time of the underlying clocksource, sched clock will get out of sync.
Mark the sched_clock_timer for expiry in hard interrupt context.
--- kernel/time/sched_clock.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-)
kernel/time/sched_clock.c +5 -4
@@ -208,7 +208,8 @@ sched_clock_register(u64 (*read)(void), int bits, unsigned long rate)if (sched_clock_timer.function !=NULL ) {/* update timeout for clock wrap */ -hrtimer_start (&sched_clock_timer, cd.wrap_kt,HRTIMER_MODE_REL ); +hrtimer_start (&sched_clock_timer, cd.wrap_kt, +HRTIMER_MODE_REL_HARD ); } r = rate;
@@ -254,9 +255,9 @@ void __init generic_sched_clock_init(void)* Start the timer to keep sched_clock() properly updated and * sets the initial epoch. */ -hrtimer_init (&sched_clock_timer,CLOCK_MONOTONIC ,HRTIMER_MODE_REL ); +hrtimer_init (&sched_clock_timer,CLOCK_MONOTONIC ,HRTIMER_MODE_REL_HARD ); sched_clock_timer.function = sched_clock_poll; -hrtimer_start (&sched_clock_timer, cd.wrap_kt,HRTIMER_MODE_REL ); +hrtimer_start (&sched_clock_timer, cd.wrap_kt,HRTIMER_MODE_REL_HARD ); }/*
@@ -293,7 +294,7 @@ void sched_clock_resume(void)struct clock_read_data *rd = &cd.read_data[0 ]; rd->epoch_cyc = cd.actual_read_sched_clock (); -hrtimer_start (&sched_clock_timer, cd.wrap_kt,HRTIMER_MODE_REL ); +hrtimer_start (&sched_clock_timer, cd.wrap_kt,HRTIMER_MODE_REL_HARD ); rd->read_sched_clock = cd.actual_read_sched_clock; }
Signature · 1 line
2.20.1
[tip: timers/core] time/sched_clock: Expire timer in hardirq context
The following commit has been merged into the timers/core branch of tip:
The following commit has been merged into the timers/core branch of tip:
Commit-ID: 2c8bd58812ee3dbf0d78b566822f7eacd34bdd7b Gitweb: https://git.kernel.org/tip/2c8bd58812ee3dbf0d78b566822f7eacd34bdd7b Author: Ahmed S. Darwish <[email protected]> AuthorDate: Mon, 09 Mar 2020 18:15:29 Committer: Thomas Gleixner <[email protected]> CommitterDate: Thu, 19 Mar 2020 09:45:08 +01:00
time/sched_clock: Expire timer in hardirq context
To minimize latency, PREEMPT_RT kernels expires hrtimers in preemptible softirq context by default. This can be overriden by marking the timer's expiry with HRTIMER_MODE_HARD.
sched_clock_timer is missing this annotation: if its callback is preempted and the duration of the preemption exceeds the wrap around time of the underlying clocksource, sched clock will get out of sync.
Mark the sched_clock_timer for expiry in hard interrupt context.
--- kernel/time/sched_clock.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-)
kernel/time/sched_clock.c +5 -4
@@ -208,7 +208,8 @@ sched_clock_register(u64 (*read)(void), int bits, unsigned long rate)if (sched_clock_timer.function !=NULL ) {/* update timeout for clock wrap */ -hrtimer_start (&sched_clock_timer, cd.wrap_kt,HRTIMER_MODE_REL ); +hrtimer_start (&sched_clock_timer, cd.wrap_kt, +HRTIMER_MODE_REL_HARD ); } r = rate;
@@ -254,9 +255,9 @@ void __init generic_sched_clock_init(void)* Start the timer to keep sched_clock() properly updated and * sets the initial epoch. */ -hrtimer_init (&sched_clock_timer,CLOCK_MONOTONIC ,HRTIMER_MODE_REL ); +hrtimer_init (&sched_clock_timer,CLOCK_MONOTONIC ,HRTIMER_MODE_REL_HARD ); sched_clock_timer.function = sched_clock_poll; -hrtimer_start (&sched_clock_timer, cd.wrap_kt,HRTIMER_MODE_REL ); +hrtimer_start (&sched_clock_timer, cd.wrap_kt,HRTIMER_MODE_REL_HARD ); }/*
@@ -293,7 +294,7 @@ void sched_clock_resume(void)struct clock_read_data *rd = &cd.read_data[0 ]; rd->epoch_cyc = cd.actual_read_sched_clock (); -hrtimer_start (&sched_clock_timer, cd.wrap_kt,HRTIMER_MODE_REL ); +hrtimer_start (&sched_clock_timer, cd.wrap_kt,HRTIMER_MODE_REL_HARD ); rd->read_sched_clock = cd.actual_read_sched_clock; }