timers/nohz: Add a comment about broken iowait counter update race

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.

Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20230222144649.624380-5-frederic@kernel.org

1 files changed, 8 insertions, 2 deletions

diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index 90d9b7b29875..edd6e9f26d16 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -705,7 +705,10 @@ static u64 get_cpu_sleep_time_us(struct tick_sched *ts, 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.
@@ -728,7 +731,10 @@ 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.