[PATCH 0/6] timers/nohz: Fixes and cleanups

[PATCH 0/6] timers/nohz: Fixes and cleanups

[Frederic Weisbecker]

Try to (partially) fix the issue reported in https://lore.kernel.org/lkml/20230128020051.2328465-1-liaoyu15@huawei.com/

Frederic Weisbecker (6):
  timers/nohz: Restructure and reshuffle struct tick_sched
  timers/nohz: Only ever update sleeptime from idle exit
  timers/nohz: Protect idle/iowait sleep time under seqcount
  timers/nohz: Add a comment about broken iowait counter update race
  timers/nohz: Remove middle-function __tick_nohz_idle_stop_tick()
  MAINTAINERS: Remove stale email address

 MAINTAINERS              |   2 +-
 kernel/time/tick-sched.c | 146 ++++++++++++++++++---------------------
 kernel/time/tick-sched.h |  67 +++++++++++-------
 3 files changed, 112 insertions(+), 103 deletions(-)

-- 
2.34.1

[PATCH 1/6] timers/nohz: Restructure and reshuffle struct tick_sched

[Frederic Weisbecker]

Restructure and group fields by access in order to optimize cache
layout. While at it, also add missing kernel doc for two fields:
@last_jiffies and @idle_expires.

Reported-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Yu Liao <liaoyu15@huawei.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Wei Li <liwei391@huawei.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Mirsad Goran Todorovac <mirsad.todorovac@alu.unizg.hr>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
---
 kernel/time/tick-sched.h | 66 +++++++++++++++++++++++++---------------
 1 file changed, 41 insertions(+), 25 deletions(-)

diff --git a/kernel/time/tick-sched.h b/kernel/time/tick-sched.h
index 504649513399..c6663254d17d 100644
--- a/kernel/time/tick-sched.h
+++ b/kernel/time/tick-sched.h
@@ -22,65 +22,81 @@ enum tick_nohz_mode {
 
 /**
  * struct tick_sched - sched tick emulation and no idle tick control/stats
- * @sched_timer:	hrtimer to schedule the periodic tick in high
- *			resolution mode
- * @check_clocks:	Notification mechanism about clocksource changes
- * @nohz_mode:		Mode - one state of tick_nohz_mode
+ *
  * @inidle:		Indicator that the CPU is in the tick idle mode
  * @tick_stopped:	Indicator that the idle tick has been stopped
  * @idle_active:	Indicator that the CPU is actively in the tick idle mode;
  *			it is reset during irq handling phases.
- * @do_timer_lst:	CPU was the last one doing do_timer before going idle
+ * @do_timer_last:	CPU was the last one doing do_timer before going idle
  * @got_idle_tick:	Tick timer function has run with @inidle set
+ * @stalled_jiffies:	Number of stalled jiffies detected across ticks
+ * @last_tick_jiffies:	Value of jiffies seen on last tick
+ * @sched_timer:	hrtimer to schedule the periodic tick in high
+ *			resolution mode
  * @last_tick:		Store the last tick expiry time when the tick
  *			timer is modified for nohz sleeps. This is necessary
  *			to resume the tick timer operation in the timeline
  *			when the CPU returns from nohz sleep.
  * @next_tick:		Next tick to be fired when in dynticks mode.
  * @idle_jiffies:	jiffies at the entry to idle for idle time accounting
+ * @idle_waketime:	Time when the idle was interrupted
+ * @idle_entrytime:	Time when the idle call was entered
+ * @nohz_mode:		Mode - one state of tick_nohz_mode
+ * @last_jiffies:	Base jiffies snapshot when next event was last computed
+ * @timer_expires_base:	Base time clock monotonic for @timer_expires
+ * @timer_expires:	Anticipated timer expiration time (in case sched tick is stopped)
+ * @next_timer:		Expiry time of next expiring timer for debugging purpose only
+ * @idle_expires:	Next tick in idle, for debugging purpose only
  * @idle_calls:		Total number of idle calls
  * @idle_sleeps:	Number of idle calls, where the sched tick was stopped
- * @idle_entrytime:	Time when the idle call was entered
- * @idle_waketime:	Time when the idle was interrupted
  * @idle_exittime:	Time when the idle state was left
  * @idle_sleeptime:	Sum of the time slept in idle with sched tick stopped
  * @iowait_sleeptime:	Sum of the time slept in idle with sched tick stopped, with IO outstanding
- * @timer_expires:	Anticipated timer expiration time (in case sched tick is stopped)
- * @timer_expires_base:	Base time clock monotonic for @timer_expires
- * @next_timer:		Expiry time of next expiring timer for debugging purpose only
  * @tick_dep_mask:	Tick dependency mask - is set, if someone needs the tick
- * @last_tick_jiffies:	Value of jiffies seen on last tick
- * @stalled_jiffies:	Number of stalled jiffies detected across ticks
+ * @check_clocks:	Notification mechanism about clocksource changes
  */
 struct tick_sched {
-	struct hrtimer			sched_timer;
-	unsigned long			check_clocks;
-	enum tick_nohz_mode		nohz_mode;
-
+	/* Common flags */
 	unsigned int			inidle		: 1;
 	unsigned int			tick_stopped	: 1;
 	unsigned int			idle_active	: 1;
 	unsigned int			do_timer_last	: 1;
 	unsigned int			got_idle_tick	: 1;
 
+	/* Tick handling: jiffies stall check */
+	unsigned int			stalled_jiffies;
+	unsigned long			last_tick_jiffies;
+
+	/* Tick handling */
+	struct hrtimer			sched_timer;
 	ktime_t				last_tick;
 	ktime_t				next_tick;
 	unsigned long			idle_jiffies;
-	unsigned long			idle_calls;
-	unsigned long			idle_sleeps;
-	ktime_t				idle_entrytime;
 	ktime_t				idle_waketime;
-	ktime_t				idle_exittime;
-	ktime_t				idle_sleeptime;
-	ktime_t				iowait_sleeptime;
+
+	/* Idle entry */
+	ktime_t				idle_entrytime;
+
+	/* Tick stop */
+	enum tick_nohz_mode		nohz_mode;
 	unsigned long			last_jiffies;
-	u64				timer_expires;
 	u64				timer_expires_base;
+	u64				timer_expires;
 	u64				next_timer;
 	ktime_t				idle_expires;
+	unsigned long			idle_calls;
+	unsigned long			idle_sleeps;
+
+	/* Idle exit */
+	ktime_t				idle_exittime;
+	ktime_t				idle_sleeptime;
+	ktime_t				iowait_sleeptime;
+
+	/* Full dynticks handling */
 	atomic_t			tick_dep_mask;
-	unsigned long			last_tick_jiffies;
-	unsigned int			stalled_jiffies;
+
+	/* Clocksource changes */
+	unsigned long			check_clocks;
 };
 
 extern struct tick_sched *tick_get_tick_sched(int cpu);
-- 
2.34.1

[PATCH 2/6] timers/nohz: Only ever update sleeptime from idle exit

[Frederic Weisbecker]

The idle and io sleeptime statistics appearing in /proc/stat can be
currently updated from two sites: locally on idle exit and remotely
by cpufreq. However there is no synchronization mechanism protecting
concurrent updates. It is therefore possible to account the sleeptime
twice, among all the possible broken scenarios.

To prevent from breaking the sleeptime accounting source, restrict the
sleeptime updates to the local idle exit site. If there is a delta to
add since the last update, IO/Idle sleep time readers will now only
compute the delta without actually writing it back to the internal idle
statistic fields.

This fixes a writer VS writer race. Note there are still two known
reader VS writer races to handle. A subsequent patch will fix one.

Reported-by: Yu Liao <liaoyu15@huawei.com>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Yu Liao <liaoyu15@huawei.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Wei Li <liwei391@huawei.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Mirsad Goran Todorovac <mirsad.todorovac@alu.unizg.hr>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
---
 kernel/time/tick-sched.c | 104 ++++++++++++++++-----------------------
 1 file changed, 42 insertions(+), 62 deletions(-)

diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index b0e3c9205946..db22342f8948 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -637,31 +637,21 @@ static void tick_nohz_update_jiffies(ktime_t now)
 	touch_softlockup_watchdog_sched();
 }
 
-/*
- * Updates the per-CPU time idle statistics counters
- */
-static void
-update_ts_time_stats(int cpu, struct tick_sched *ts, ktime_t now, u64 *last_update_time)
-{
-	ktime_t delta;
-
-	if (ts->idle_active) {
-		delta = ktime_sub(now, ts->idle_entrytime);
-		if (nr_iowait_cpu(cpu) > 0)
-			ts->iowait_sleeptime = ktime_add(ts->iowait_sleeptime, delta);
-		else
-			ts->idle_sleeptime = ktime_add(ts->idle_sleeptime, delta);
-		ts->idle_entrytime = now;
-	}
-
-	if (last_update_time)
-		*last_update_time = ktime_to_us(now);
-
-}
-
 static void tick_nohz_stop_idle(struct tick_sched *ts, ktime_t now)
 {
-	update_ts_time_stats(smp_processor_id(), ts, now, NULL);
+	ktime_t delta;
+
+	if (WARN_ON_ONCE(!ts->idle_active))
+		return;
+
+	delta = ktime_sub(now, ts->idle_entrytime);
+
+	if (nr_iowait_cpu(smp_processor_id()) > 0)
+		ts->iowait_sleeptime = ktime_add(ts->iowait_sleeptime, delta);
+	else
+		ts->idle_sleeptime = ktime_add(ts->idle_sleeptime, delta);
+
+	ts->idle_entrytime = now;
 	ts->idle_active = 0;
 
 	sched_clock_idle_wakeup_event();
@@ -674,6 +664,31 @@ static void tick_nohz_start_idle(struct tick_sched *ts)
 	sched_clock_idle_sleep_event();
 }
 
+static u64 get_cpu_sleep_time_us(int cpu, ktime_t *sleeptime,
+				 bool compute_delta, u64 *last_update_time)
+{
+	struct tick_sched *ts = &per_cpu(tick_cpu_sched, cpu);
+	ktime_t now, idle;
+
+	if (!tick_nohz_active)
+		return -1;
+
+	now = ktime_get();
+	if (last_update_time)
+		*last_update_time = ktime_to_us(now);
+
+	if (ts->idle_active && !nr_iowait_cpu(cpu)) {
+		ktime_t delta = ktime_sub(now, ts->idle_entrytime);
+
+		idle = ktime_add(*sleeptime, delta);
+	} else {
+		idle = *sleeptime;
+	}
+
+	return ktime_to_us(idle);
+
+}
+
 /**
  * get_cpu_idle_time_us - get the total idle time of a CPU
  * @cpu: CPU number to query
@@ -691,27 +706,9 @@ static void tick_nohz_start_idle(struct tick_sched *ts)
 u64 get_cpu_idle_time_us(int cpu, u64 *last_update_time)
 {
 	struct tick_sched *ts = &per_cpu(tick_cpu_sched, cpu);
-	ktime_t now, idle;
-
-	if (!tick_nohz_active)
-		return -1;
-
-	now = ktime_get();
-	if (last_update_time) {
-		update_ts_time_stats(cpu, ts, now, last_update_time);
-		idle = ts->idle_sleeptime;
-	} else {
-		if (ts->idle_active && !nr_iowait_cpu(cpu)) {
-			ktime_t delta = ktime_sub(now, ts->idle_entrytime);
-
-			idle = ktime_add(ts->idle_sleeptime, delta);
-		} else {
-			idle = ts->idle_sleeptime;
-		}
-	}
-
-	return ktime_to_us(idle);
 
+	return get_cpu_sleep_time_us(cpu, &ts->idle_sleeptime,
+				     !nr_iowait_cpu(cpu), last_update_time);
 }
 EXPORT_SYMBOL_GPL(get_cpu_idle_time_us);
 
@@ -732,26 +729,9 @@ EXPORT_SYMBOL_GPL(get_cpu_idle_time_us);
 u64 get_cpu_iowait_time_us(int cpu, u64 *last_update_time)
 {
 	struct tick_sched *ts = &per_cpu(tick_cpu_sched, cpu);
-	ktime_t now, iowait;
 
-	if (!tick_nohz_active)
-		return -1;
-
-	now = ktime_get();
-	if (last_update_time) {
-		update_ts_time_stats(cpu, ts, now, last_update_time);
-		iowait = ts->iowait_sleeptime;
-	} else {
-		if (ts->idle_active && nr_iowait_cpu(cpu) > 0) {
-			ktime_t delta = ktime_sub(now, ts->idle_entrytime);
-
-			iowait = ktime_add(ts->iowait_sleeptime, delta);
-		} else {
-			iowait = ts->iowait_sleeptime;
-		}
-	}
-
-	return ktime_to_us(iowait);
+	return get_cpu_sleep_time_us(cpu, &ts->iowait_sleeptime,
+				     nr_iowait_cpu(cpu), last_update_time);
 }
 EXPORT_SYMBOL_GPL(get_cpu_iowait_time_us);
 
-- 
2.34.1

[PATCH 3/6] timers/nohz: Protect idle/iowait sleep time under seqcount

[Frederic Weisbecker]

Reading idle/io sleep time (eg: from /proc/stat) can race with idle exit
updates because the state machine handling the stats is not atomic and
requires a coherent read batch.

As a result reading the sleep time may report irrelevant or backward
values.

Fix this with protecting the simple state machine within a seqcount.
This is expected to be cheap enough not to add measurable performance
impact on the idle path.

Note this only fixes reader VS writer condition partitially. A race
remains that involves remote updates of the CPU iowait task counter. It
can hardly be fixed.

Reported-by: Yu Liao <liaoyu15@huawei.com>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Yu Liao <liaoyu15@huawei.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Wei Li <liwei391@huawei.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Mirsad Goran Todorovac <mirsad.todorovac@alu.unizg.hr>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
---
 kernel/time/tick-sched.c | 22 ++++++++++++++++------
 kernel/time/tick-sched.h |  1 +
 2 files changed, 17 insertions(+), 6 deletions(-)

diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index db22342f8948..757e03ef4409 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -646,6 +646,7 @@ static void tick_nohz_stop_idle(struct tick_sched *ts, ktime_t now)
 
 	delta = ktime_sub(now, ts->idle_entrytime);
 
+	write_seqcount_begin(&ts->idle_sleeptime_seq);
 	if (nr_iowait_cpu(smp_processor_id()) > 0)
 		ts->iowait_sleeptime = ktime_add(ts->iowait_sleeptime, delta);
 	else
@@ -653,14 +654,18 @@ static void tick_nohz_stop_idle(struct tick_sched *ts, ktime_t now)
 
 	ts->idle_entrytime = now;
 	ts->idle_active = 0;
+	write_seqcount_end(&ts->idle_sleeptime_seq);
 
 	sched_clock_idle_wakeup_event();
 }
 
 static void tick_nohz_start_idle(struct tick_sched *ts)
 {
+	write_seqcount_begin(&ts->idle_sleeptime_seq);
 	ts->idle_entrytime = ktime_get();
 	ts->idle_active = 1;
+	write_seqcount_end(&ts->idle_sleeptime_seq);
+
 	sched_clock_idle_sleep_event();
 }
 
@@ -669,6 +674,7 @@ static u64 get_cpu_sleep_time_us(int cpu, ktime_t *sleeptime,
 {
 	struct tick_sched *ts = &per_cpu(tick_cpu_sched, cpu);
 	ktime_t now, idle;
+	unsigned int seq;
 
 	if (!tick_nohz_active)
 		return -1;
@@ -677,13 +683,17 @@ static u64 get_cpu_sleep_time_us(int cpu, ktime_t *sleeptime,
 	if (last_update_time)
 		*last_update_time = ktime_to_us(now);
 
-	if (ts->idle_active && !nr_iowait_cpu(cpu)) {
-		ktime_t delta = ktime_sub(now, ts->idle_entrytime);
+	do {
+		seq = read_seqcount_begin(&ts->idle_sleeptime_seq);
 
-		idle = ktime_add(*sleeptime, delta);
-	} else {
-		idle = *sleeptime;
-	}
+		if (ts->idle_active && !nr_iowait_cpu(cpu)) {
+			ktime_t delta = ktime_sub(now, ts->idle_entrytime);
+
+			idle = ktime_add(*sleeptime, delta);
+		} else {
+			idle = *sleeptime;
+		}
+	} while (read_seqcount_retry(&ts->idle_sleeptime_seq, seq));
 
 	return ktime_to_us(idle);
 
diff --git a/kernel/time/tick-sched.h b/kernel/time/tick-sched.h
index c6663254d17d..5ed5a9d41d5a 100644
--- a/kernel/time/tick-sched.h
+++ b/kernel/time/tick-sched.h
@@ -75,6 +75,7 @@ struct tick_sched {
 	ktime_t				idle_waketime;
 
 	/* Idle entry */
+	seqcount_t			idle_sleeptime_seq;
 	ktime_t				idle_entrytime;
 
 	/* Tick stop */
-- 
2.34.1

[PATCH 4/6] timers/nohz: Add a comment about broken iowait counter update race

[Frederic Weisbecker]

The per-cpu iowait task counter is incremented locally upon sleeping.
But since the task can be woken to (and by) another CPU, the counter may
then be decremented remotely. This is the source of a race involving
readers VS writer of idle/iowait sleeptime.

The following scenario shows an example where a /proc/stat reader
observes a pending sleep time as IO whereas that pending sleep time
later eventually gets accounted as non-IO.

    CPU 0                       CPU  1                    CPU 2
    -----                       -----                     ------
    //io_schedule() TASK A
    current->in_iowait = 1
    rq(0)->nr_iowait++
    //switch to idle
                        // READ /proc/stat
                        // See nr_iowait_cpu(0) == 1
                        return ts->iowait_sleeptime +
                               ktime_sub(ktime_get(), ts->idle_entrytime)

                                                          //try_to_wake_up(TASK A)
                                                          rq(0)->nr_iowait--
    //idle exit
    // See nr_iowait_cpu(0) == 0
    ts->idle_sleeptime += ktime_sub(ktime_get(), ts->idle_entrytime)

As a result subsequent reads on /proc/stat may expose backward progress.

This is unfortunately hardly fixable. Just add a comment about that
condition.

Cc: Hillf Danton <hdanton@sina.com>
Cc: Yu Liao <liaoyu15@huawei.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Wei Li <liwei391@huawei.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Mirsad Goran Todorovac <mirsad.todorovac@alu.unizg.hr>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
---
 kernel/time/tick-sched.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index 757e03ef4409..bd9499fb15b3 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -706,7 +706,10 @@ static u64 get_cpu_sleep_time_us(int cpu, ktime_t *sleeptime,
  * counters if NULL.
  *
  * Return the cumulative idle time (since boot) for a given
- * CPU, in microseconds.
+ * CPU, in microseconds. Note this is partially broken due to
+ * the counter of iowait tasks that can be remotely updated without
+ * any synchronization. Therefore it is possible to observe backward
+ * values within two consecutive reads.
  *
  * This time is measured via accounting rather than sampling,
  * and is as accurate as ktime_get() is.
@@ -729,7 +732,10 @@ EXPORT_SYMBOL_GPL(get_cpu_idle_time_us);
  * counters if NULL.
  *
  * Return the cumulative iowait time (since boot) for a given
- * CPU, in microseconds.
+ * CPU, in microseconds. Note this is partially broken due to
+ * the counter of iowait tasks that can be remotely updated without
+ * any synchronization. Therefore it is possible to observe backward
+ * values within two consecutive reads.
  *
  * This time is measured via accounting rather than sampling,
  * and is as accurate as ktime_get() is.
-- 
2.34.1

[PATCH 5/6] timers/nohz: Remove middle-function __tick_nohz_idle_stop_tick()

[Frederic Weisbecker]

There is no need for the __tick_nohz_idle_stop_tick() function between
tick_nohz_idle_stop_tick() and its implementation. Remove that
unnecessary step.

Cc: Hillf Danton <hdanton@sina.com>
Cc: Yu Liao <liaoyu15@huawei.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Wei Li <liwei391@huawei.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Mirsad Goran Todorovac <mirsad.todorovac@alu.unizg.hr>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
---
 kernel/time/tick-sched.c | 20 ++++++++------------
 1 file changed, 8 insertions(+), 12 deletions(-)

diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index bd9499fb15b3..7065c6e373e9 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -1080,10 +1080,16 @@ static bool can_stop_idle_tick(int cpu, struct tick_sched *ts)
 	return true;
 }
 
-static void __tick_nohz_idle_stop_tick(struct tick_sched *ts)
+/**
+ * tick_nohz_idle_stop_tick - stop the idle tick from the idle task
+ *
+ * When the next event is more than a tick into the future, stop the idle tick
+ */
+void tick_nohz_idle_stop_tick(void)
 {
+	struct tick_sched *ts = this_cpu_ptr(&tick_cpu_sched);
+	int cpu = smp_processor_id();
 	ktime_t expires;
-	int cpu = smp_processor_id();
 
 	/*
 	 * If tick_nohz_get_sleep_length() ran tick_nohz_next_event(), the
@@ -1115,16 +1121,6 @@ static void __tick_nohz_idle_stop_tick(struct tick_sched *ts)
 	}
 }
 
-/**
- * tick_nohz_idle_stop_tick - stop the idle tick from the idle task
- *
- * When the next event is more than a tick into the future, stop the idle tick
- */
-void tick_nohz_idle_stop_tick(void)
-{
-	__tick_nohz_idle_stop_tick(this_cpu_ptr(&tick_cpu_sched));
-}
-
 void tick_nohz_idle_retain_tick(void)
 {
 	tick_nohz_retain_tick(this_cpu_ptr(&tick_cpu_sched));
-- 
2.34.1

[PATCH 6/6] MAINTAINERS: Remove stale email address

[Frederic Weisbecker]

Cc: Hillf Danton <hdanton@sina.com>
Cc: Yu Liao <liaoyu15@huawei.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Wei Li <liwei391@huawei.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Mirsad Goran Todorovac <mirsad.todorovac@alu.unizg.hr>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
---
 MAINTAINERS | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/MAINTAINERS b/MAINTAINERS
index fb1471cb5ed3..300ca61fa0bc 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -14794,7 +14794,7 @@ F:	include/uapi/linux/nitro_enclaves.h
 F:	samples/nitro_enclaves/
 
 NOHZ, DYNTICKS SUPPORT
-M:	Frederic Weisbecker <fweisbec@gmail.com>
+M:	Frederic Weisbecker <frederic@kernel.org>
 M:	Thomas Gleixner <tglx@linutronix.de>
 M:	Ingo Molnar <mingo@kernel.org>
 L:	linux-kernel@vger.kernel.org
-- 
2.34.1

Re: [PATCH 4/6] timers/nohz: Add a comment about broken iowait counter update race

[Peter Zijlstra]

On Fri, Feb 10, 2023 at 03:09:15PM +0100, Frederic Weisbecker wrote:
> The per-cpu iowait task counter is incremented locally upon sleeping.
> But since the task can be woken to (and by) another CPU, the counter may
> then be decremented remotely. This is the source of a race involving
> readers VS writer of idle/iowait sleeptime.
> 
> The following scenario shows an example where a /proc/stat reader
> observes a pending sleep time as IO whereas that pending sleep time
> later eventually gets accounted as non-IO.
> 
>     CPU 0                       CPU  1                    CPU 2
>     -----                       -----                     ------
>     //io_schedule() TASK A
>     current->in_iowait = 1
>     rq(0)->nr_iowait++
>     //switch to idle
>                         // READ /proc/stat
>                         // See nr_iowait_cpu(0) == 1
>                         return ts->iowait_sleeptime +
>                                ktime_sub(ktime_get(), ts->idle_entrytime)
> 
>                                                           //try_to_wake_up(TASK A)
>                                                           rq(0)->nr_iowait--
>     //idle exit
>     // See nr_iowait_cpu(0) == 0
>     ts->idle_sleeptime += ktime_sub(ktime_get(), ts->idle_entrytime)
> 
> As a result subsequent reads on /proc/stat may expose backward progress.
> 
> This is unfortunately hardly fixable. Just add a comment about that
> condition.

It is far worse than that, the whole concept of per-cpu iowait is
absurd. Also see the comment near nr_iowait().

Re: [PATCH 0/6] timers/nohz: Fixes and cleanups

[Peter Zijlstra]

On Fri, Feb 10, 2023 at 03:09:11PM +0100, Frederic Weisbecker wrote:
> Try to (partially) fix the issue reported in https://lore.kernel.org/lkml/20230128020051.2328465-1-liaoyu15@huawei.com/
> 
> Frederic Weisbecker (6):
>   timers/nohz: Restructure and reshuffle struct tick_sched
>   timers/nohz: Only ever update sleeptime from idle exit
>   timers/nohz: Protect idle/iowait sleep time under seqcount
>   timers/nohz: Add a comment about broken iowait counter update race
>   timers/nohz: Remove middle-function __tick_nohz_idle_stop_tick()
>   MAINTAINERS: Remove stale email address

Seem sensible enough.

Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>

Re: [PATCH 4/6] timers/nohz: Add a comment about broken iowait counter update race

[Frederic Weisbecker]

On Fri, Feb 10, 2023 at 03:39:43PM +0100, Peter Zijlstra wrote:
> On Fri, Feb 10, 2023 at 03:09:15PM +0100, Frederic Weisbecker wrote:
> > The per-cpu iowait task counter is incremented locally upon sleeping.
> > But since the task can be woken to (and by) another CPU, the counter may
> > then be decremented remotely. This is the source of a race involving
> > readers VS writer of idle/iowait sleeptime.
> > 
> > The following scenario shows an example where a /proc/stat reader
> > observes a pending sleep time as IO whereas that pending sleep time
> > later eventually gets accounted as non-IO.
> > 
> >     CPU 0                       CPU  1                    CPU 2
> >     -----                       -----                     ------
> >     //io_schedule() TASK A
> >     current->in_iowait = 1
> >     rq(0)->nr_iowait++
> >     //switch to idle
> >                         // READ /proc/stat
> >                         // See nr_iowait_cpu(0) == 1
> >                         return ts->iowait_sleeptime +
> >                                ktime_sub(ktime_get(), ts->idle_entrytime)
> > 
> >                                                           //try_to_wake_up(TASK A)
> >                                                           rq(0)->nr_iowait--
> >     //idle exit
> >     // See nr_iowait_cpu(0) == 0
> >     ts->idle_sleeptime += ktime_sub(ktime_get(), ts->idle_entrytime)
> > 
> > As a result subsequent reads on /proc/stat may expose backward progress.
> > 
> > This is unfortunately hardly fixable. Just add a comment about that
> > condition.
> 
> It is far worse than that, the whole concept of per-cpu iowait is
> absurd. Also see the comment near nr_iowait().

Alas I know :-(

Re: [PATCH 0/6] timers/nohz: Fixes and cleanups

[Thomas Gleixner]

On Fri, Feb 10 2023 at 15:09, Frederic Weisbecker wrote:
> Try to (partially) fix the issue reported in https://lore.kernel.org/lkml/20230128020051.2328465-1-liaoyu15@huawei.com/

Looks about right. Can you please remove that assert in the selftest
which started all this?

Thanks,

        tglx