Usually userspace sends SND_BELL and SND_TONE events, and by the time pwm_beeper_suspend() runs userpsace is already frozen, but theoretically in-kernel users may send these events too, and that may cause pwm_beeper_event() scheduling another work after we canceled it.
Let's introduce a "suspended" flag and check it in pwm_beeper_event() to avoid this race. Signed-off-by: Dmitry Torokhov <[email protected]> --- drivers/input/misc/pwm-beeper.c | 21 ++++++++++++++++++--- 1 file changed, 18 insertions(+), 3 deletions(-) diff --git a/drivers/input/misc/pwm-beeper.c b/drivers/input/misc/pwm-beeper.c index ce6eec48ec5f..04c8ad3827d9 100644 --- a/drivers/input/misc/pwm-beeper.c +++ b/drivers/input/misc/pwm-beeper.c @@ -27,6 +27,7 @@ struct pwm_beeper { struct pwm_device *pwm; struct work_struct work; unsigned long period; + bool suspended; }; #define HZ_TO_NANOSECONDS(x) (1000000000UL/(x)) @@ -73,7 +74,8 @@ static int pwm_beeper_event(struct input_dev *input, else beeper->period = HZ_TO_NANOSECONDS(value); - schedule_work(&beeper->work); + if (!beeper->suspended) + schedule_work(&beeper->work); return 0; } @@ -154,6 +156,15 @@ static int __maybe_unused pwm_beeper_suspend(struct device *dev) { struct pwm_beeper *beeper = dev_get_drvdata(dev); + /* + * Spinlock is taken here is not to protect write to + * beeper->suspended, but to ensure that pwm_beeper_event + * does not re-submit work once flag is set. + */ + spin_lock_irq(&beeper->input->event_lock); + beeper->suspended = true; + spin_unlock_irq(&beeper->input->event_lock); + pwm_beeper_stop(beeper); return 0; @@ -163,8 +174,12 @@ static int __maybe_unused pwm_beeper_resume(struct device *dev) { struct pwm_beeper *beeper = dev_get_drvdata(dev); - if (beeper->period) - __pwm_beeper_set(beeper); + spin_lock_irq(&beeper->input->event_lock); + beeper->suspended = false; + spin_unlock_irq(&beeper->input->event_lock); + + /* Let worker figure out if we should resume beeping */ + schedule_work(&beeper->work); return 0; } -- 2.11.0.483.g087da7b7c-goog
