From d3cbc3f7872a47768eb48f1ca1ca23af004fc85d Mon Sep 17 00:00:00 2001 From: Felix Moessbauer Date: Wed, 14 Aug 2024 14:10:32 +0200 Subject: [PATCH] hrtimer: Use and report correct timerslack values for realtime tasks The timerslack_ns setting is used to specify how much the hardware timers should be delayed, to potentially dispatch multiple timers in a single interrupt. This is a performance optimization. Timers of realtime tasks (having a realtime scheduling policy) should not be delayed. This logic was inconsitently applied to the hrtimers, leading to delays of realtime tasks which used timed waits for events (e.g. condition variables). Due to the downstream override of the slack for rt tasks, the procfs reported incorrect (non-zero) timerslack_ns values. This is changed by setting the timer_slack_ns task attribute to 0 for all tasks with a rt policy. By that, downstream users do not need to specially handle rt tasks (w.r.t. the slack), and the procfs entry shows the correct value of "0". Setting non-zero slack values (either via procfs or PR_SET_TIMERSLACK) on tasks with a rt policy is ignored, as stated in "man 2 PR_SET_TIMERSLACK": Timer slack is not applied to threads that are scheduled under a real-time scheduling policy (see sched_setscheduler(2)). The special handling of timerslack on rt tasks in downstream users is removed as well. Signed-off-by: Felix Moessbauer Signed-off-by: Thomas Gleixner Link: https://lore.kernel.org/all/20240814121032.368444-2-felix.moessbauer@siemens.com [Sultan Alsawaf: backport to 6.1] Signed-off-by: Sultan Alsawaf Change-Id: I8b2c81ffdeea181ab729935de71d1b0131c16ffc Signed-off-by: Richard Raya --- fs/proc/base.c | 9 +++++---- fs/select.c | 11 ++++------- kernel/sched/core.c | 8 ++++++++ kernel/sys.c | 2 ++ kernel/time/hrtimer.c | 7 +------ 5 files changed, 20 insertions(+), 17 deletions(-) diff --git a/fs/proc/base.c b/fs/proc/base.c index 97706092ad55..2c5944e90be0 100644 --- a/fs/proc/base.c +++ b/fs/proc/base.c @@ -2540,10 +2540,11 @@ static ssize_t timerslack_ns_write(struct file *file, const char __user *buf, } task_lock(p); - if (slack_ns == 0) - p->timer_slack_ns = p->default_timer_slack_ns; - else - p->timer_slack_ns = slack_ns; + if (task_is_realtime(p)) + slack_ns = 0; + else if (slack_ns == 0) + slack_ns = p->default_timer_slack_ns; + p->timer_slack_ns = slack_ns; task_unlock(p); out: diff --git a/fs/select.c b/fs/select.c index 5290dee97a6c..e9edefd2f636 100644 --- a/fs/select.c +++ b/fs/select.c @@ -76,19 +76,16 @@ u64 select_estimate_accuracy(struct timespec64 *tv) { u64 ret; struct timespec64 now; + u64 slack = current->timer_slack_ns; - /* - * Realtime tasks get a slack of 0 for obvious reasons. - */ - - if (rt_task(current)) + if (slack == 0) return 0; ktime_get_ts64(&now); now = timespec64_sub(*tv, now); ret = __estimate_accuracy(&now); - if (ret < current->timer_slack_ns) - return current->timer_slack_ns; + if (ret < slack) + return slack; return ret; } diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 8e9e6891e9bc..986eeaea0705 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -5195,6 +5195,14 @@ static void __setscheduler_params(struct task_struct *p, else if (fair_policy(policy)) p->static_prio = NICE_TO_PRIO(attr->sched_nice); + /* rt-policy tasks do not have a timerslack */ + if (task_is_realtime(p)) { + p->timer_slack_ns = 0; + } else if (p->timer_slack_ns == 0) { + /* when switching back to non-rt policy, restore timerslack */ + p->timer_slack_ns = p->default_timer_slack_ns; + } + /* * __sched_setscheduler() ensures attr->sched_priority == 0 when * !rt_policy. Always setting this ensures that things like diff --git a/kernel/sys.c b/kernel/sys.c index 9ad067e181d4..cb2c0372f1ca 100644 --- a/kernel/sys.c +++ b/kernel/sys.c @@ -2475,6 +2475,8 @@ SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3, error = current->timer_slack_ns; break; case PR_SET_TIMERSLACK: + if (task_is_realtime(current)) + break; if (arg2 <= 0) current->timer_slack_ns = current->default_timer_slack_ns; diff --git a/kernel/time/hrtimer.c b/kernel/time/hrtimer.c index ab379e3e2345..b1b7b3f4ed6a 100644 --- a/kernel/time/hrtimer.c +++ b/kernel/time/hrtimer.c @@ -1965,14 +1965,9 @@ long hrtimer_nanosleep(const struct timespec64 *rqtp, struct restart_block *restart; struct hrtimer_sleeper t; int ret = 0; - u64 slack; - - slack = current->timer_slack_ns; - if (dl_task(current) || rt_task(current)) - slack = 0; hrtimer_init_sleeper_on_stack(&t, clockid, mode); - hrtimer_set_expires_range_ns(&t.timer, timespec64_to_ktime(*rqtp), slack); + hrtimer_set_expires_range_ns(&t.timer, timespec64_to_ktime(*rqtp), current->timer_slack_ns); ret = do_nanosleep(&t, mode); if (ret != -ERESTART_RESTARTBLOCK) goto out;