ALSA: aloop - do not reschedule timer if deleted
PATCH 1/1 ALSA: aloop - do not reschedule timer if deleted
If the timer is deleted while the timer handler is running, it may get rescheduled and an unnecessary period elapsed will be sent.
Add a flag to reschedule the timer only if it has not been stopped.
--- sound/drivers/aloop.c | 21 ++++++++++++++++++--- 1 files changed, 18 insertions(+), 3 deletions(-)
sound/drivers/aloop.c +18 -3
@@ -121,6 +121,7 @@ struct loopback_pcm {unsigned long last_jiffies;struct timer_list timer;spinlock_t timer_lock; +bool timer_stopped; };static struct platform_device *devices[SNDRV_CARDS ];
@@ -166,12 +167,11 @@ static inline unsigned int get_rate_shift(struct loopback_pcm *dpcm)return get_setup (dpcm)->rate_shift; } -static void loopback_timer_start (struct loopback_pcm *dpcm) +static void do_timer_start (struct loopback_pcm *dpcm) {unsigned long tick;unsigned int rate_shift =get_rate_shift (dpcm); -spin_lock (&dpcm->timer_lock);if (rate_shift != dpcm->pcm_rate_shift) { dpcm->pcm_rate_shift = rate_shift; dpcm->period_size_frac =frac_pos (dpcm, dpcm->pcm_period_size);
@@ -184,6 +184,13 @@ static void loopback_timer_start(struct loopback_pcm *dpcm) tick = (tick + dpcm->pcm_bps -1 ) / dpcm->pcm_bps; dpcm->timer.expires = jiffies + tick;add_timer (&dpcm->timer); +} + +static inline void loopback_timer_start (struct loopback_pcm *dpcm) +{ +spin_lock (&dpcm->timer_lock); + dpcm->timer_stopped = false; +do_timer_start (dpcm);spin_unlock (&dpcm->timer_lock); }
@@ -192,6 +199,7 @@ static inline void loopback_timer_stop(struct loopback_pcm *dpcm)spin_lock (&dpcm->timer_lock);del_timer (&dpcm->timer); dpcm->timer.expires =0 ; + dpcm->timer_stopped = true;spin_unlock (&dpcm->timer_lock); }
@@ -540,7 +548,13 @@ static void loopback_timer_function(unsigned long data) running =loopback_pos_update (dpcm->cable);if (running & (1 << dpcm->substream->stream)) { -loopback_timer_start (dpcm); +spin_lock (&dpcm->timer_lock); +if (dpcm->timer_stopped) { +spin_unlock (&dpcm->timer_lock); +return ; + } +do_timer_start (dpcm); +spin_unlock (&dpcm->timer_lock);if (dpcm->period_update_pending) { dpcm->period_update_pending =0 ;snd_pcm_period_elapsed (dpcm->substream);
@@ -670,6 +684,7 @@ static int loopback_open(struct snd_pcm_substream *substream) } dpcm->loopback = loopback; dpcm->substream = substream; + dpcm->timer_stopped = true;setup_timer (&dpcm->timer, loopback_timer_function, (unsigned long )dpcm);spin_lock_init (&dpcm->timer_lock);
Signature · 1 line
1.7.4.1
At Wed, 17 Oct 2012 19:05:46 +0530,
Quoted 5 lines · Omair Mohammed Abdullah
If the timer is deleted while the timer handler is running, it may get rescheduled and an unnecessary period elapsed will be sent.
Add a flag to reschedule the timer only if it has not been stopped.
In which situation is it rescheduled exactly?
Actually there is a small race in the check of running bit and the timer lock. Do you mean this?
If so, wouldn't it be cleaner to get rid of timer_lock and protect the whole start/stop with cable->lock, too, like below?
thanks,
Takashi
---
sound/drivers/aloop.c +19 -17
@@ -120,7 +120,6 @@ struct loopback_pcm {unsigned int last_drift;unsigned long last_jiffies;struct timer_list timer; -spinlock_t timer_lock; };static struct platform_device *devices[SNDRV_CARDS ];
@@ -166,12 +165,12 @@ static inline unsigned int get_rate_shift(struct loopback_pcm *dpcm)return get_setup (dpcm)->rate_shift; } +/* call in cable->lock */ static void loopback_timer_start (struct loopback_pcm *dpcm) {unsigned long tick;unsigned int rate_shift =get_rate_shift (dpcm); -spin_lock (&dpcm->timer_lock);if (rate_shift != dpcm->pcm_rate_shift) { dpcm->pcm_rate_shift = rate_shift; dpcm->period_size_frac =frac_pos (dpcm, dpcm->pcm_period_size);
@@ -184,15 +183,13 @@ static void loopback_timer_start(struct loopback_pcm *dpcm) tick = (tick + dpcm->pcm_bps -1 ) / dpcm->pcm_bps; dpcm->timer.expires = jiffies + tick;add_timer (&dpcm->timer); -spin_unlock (&dpcm->timer_lock); } +/* call in cable->lock */ static inline void loopback_timer_stop (struct loopback_pcm *dpcm) { -spin_lock (&dpcm->timer_lock);del_timer (&dpcm->timer); dpcm->timer.expires =0 ; -spin_unlock (&dpcm->timer_lock); }#define CABLE_VALID_PLAYBACK (1 << SNDRV_PCM_STREAM_PLAYBACK)
@@ -274,8 +271,8 @@ static int loopback_trigger(struct snd_pcm_substream *substream, int cmd)spin_lock (&cable->lock); cable->running |= stream; cable->pause &= ~stream; -spin_unlock (&cable->lock);loopback_timer_start (dpcm); +spin_unlock (&cable->lock);if (substream->stream ==SNDRV_PCM_STREAM_PLAYBACK )loopback_active_notify (dpcm);break ;
@@ -283,23 +280,23 @@ static int loopback_trigger(struct snd_pcm_substream *substream, int cmd)spin_lock (&cable->lock); cable->running &= ~stream; cable->pause &= ~stream; -spin_unlock (&cable->lock);loopback_timer_stop (dpcm); +spin_unlock (&cable->lock);if (substream->stream ==SNDRV_PCM_STREAM_PLAYBACK )loopback_active_notify (dpcm);break ;case SNDRV_PCM_TRIGGER_PAUSE_PUSH :spin_lock (&cable->lock); cable->pause |= stream; -spin_unlock (&cable->lock);loopback_timer_stop (dpcm); +spin_unlock (&cable->lock);break ;case SNDRV_PCM_TRIGGER_PAUSE_RELEASE :spin_lock (&cable->lock); dpcm->last_jiffies = jiffies; cable->pause &= ~stream; -spin_unlock (&cable->lock);loopback_timer_start (dpcm); +spin_unlock (&cable->lock);break ;default :return -EINVAL ;
@@ -477,6 +474,7 @@ static inline void bytepos_finish(struct loopback_pcm *dpcm, dpcm->buf_pos %= dpcm->pcm_buffer_size; } +/* call in cable->lock */ static unsigned int loopback_pos_update (struct loopback_cable *cable) { struct loopback_pcm *dpcm_play =
@@ -485,9 +483,7 @@ static unsigned int loopback_pos_update(struct loopback_cable *cable) cable->streams[SNDRV_PCM_STREAM_CAPTURE ];unsigned long delta_play =0 , delta_capt =0 ;unsigned int running, count1, count2; -unsigned long flags; -spin_lock_irqsave (&cable->lock, flags); running = cable->running ^ cable->pause;if (running & (1 <<SNDRV_PCM_STREAM_PLAYBACK )) { delta_play = jiffies - dpcm_play->last_jiffies;
@@ -529,32 +525,39 @@ static unsigned int loopback_pos_update(struct loopback_cable *cable)bytepos_finish (dpcm_play, count1);bytepos_finish (dpcm_capt, count1); unlock: -spin_unlock_irqrestore (&cable->lock, flags);return running; }static void loopback_timer_function (unsigned long data) {struct loopback_pcm *dpcm = (struct loopback_pcm *)data; -unsigned int running; +unsigned long flags; - running =loopback_pos_update (dpcm->cable); -if (running & (1 << dpcm->substream->stream)) { +spin_lock_irqsave (&dpcm->cable->lock, flags); +if (loopback_pos_update (dpcm->cable) & (1 << dpcm->substream->stream)) {loopback_timer_start (dpcm);if (dpcm->period_update_pending) { dpcm->period_update_pending =0 ; +spin_unlock_irqrestore (&dpcm->cable->lock, flags); +/* need to unlock before calling below */ snd_pcm_period_elapsed (dpcm->substream); +return ; } } +spin_unlock_irqrestore (&dpcm->cable->lock, flags); }static snd_pcm_uframes_t loopback_pointer (struct snd_pcm_substream *substream) {struct snd_pcm_runtime *runtime = substream->runtime;struct loopback_pcm *dpcm = runtime->private_data; +snd_pcm_uframes_t pos; +spin_lock (&dpcm->cable->lock);loopback_pos_update (dpcm->cable); -return bytes_to_frames (runtime, dpcm->buf_pos); + pos = dpcm->buf_pos; +spin_unlock (&dpcm->cable->lock); +return bytes_to_frames (runtime, pos); } static struct snd_pcm_hardware loopback_pcm_hardware =
@@ -672,7 +675,6 @@ static int loopback_open(struct snd_pcm_substream *substream) dpcm->substream = substream;setup_timer (&dpcm->timer, loopback_timer_function, (unsigned long )dpcm); -spin_lock_init (&dpcm->timer_lock); cable = loopback->cables[substream->number][dev];if (!cable) {
Quoted 8 lines
Quoted 6 lines
If the timer is deleted while the timer handler is running, it may get rescheduled and an unnecessary period elapsed will be sent.
Add a flag to reschedule the timer only if it has not been stopped.
In which situation is it rescheduled exactly?
When the loopback_timer_stop() runs on one CPU while loopback_timer_function() runs on the other CPU. This causes the loopback_timer_function() to reschedule the timer beacuse del_timer() will not wait for the timer function to finish.
Quoted 7 lines
Actually there is a small race in the check of running bit and the timer lock. Do you mean this?
If so, wouldn't it be cleaner to get rid of
timer_lockand protect the whole start/stop withcable->lock, too, like below?
Yes, I think this fixes the problem quite well.
Quoted 162 lines
thanks,
Takashi
---
sound/drivers/aloop.c +19
-17@@ -120,7 +120,6 @@ struct loopback_pcm {unsigned int last_drift;unsigned long last_jiffies;struct timer_list timer; -spinlock_t timer_lock; };static struct platform_device *devices[SNDRV_CARDS ];@@ -166,12 +165,12 @@ static inline unsigned int get_rate_shift(struct loopback_pcm *dpcm)return get_setup (dpcm)->rate_shift; } +/* call in cable->lock */ static void loopback_timer_start (struct loopback_pcm *dpcm) {unsigned long tick;unsigned int rate_shift =get_rate_shift (dpcm); -spin_lock (&dpcm->timer_lock);if (rate_shift != dpcm->pcm_rate_shift) { dpcm->pcm_rate_shift = rate_shift; dpcm->period_size_frac =frac_pos (dpcm, dpcm->pcm_period_size);@@ -184,15 +183,13 @@ static void loopback_timer_start(struct loopback_pcm *dpcm) tick = (tick + dpcm->pcm_bps -1 ) / dpcm->pcm_bps; dpcm->timer.expires = jiffies + tick;add_timer (&dpcm->timer); -spin_unlock (&dpcm->timer_lock); } +/* call in cable->lock */ static inline void loopback_timer_stop (struct loopback_pcm *dpcm) { -spin_lock (&dpcm->timer_lock);del_timer (&dpcm->timer); dpcm->timer.expires =0 ; -spin_unlock (&dpcm->timer_lock); }#define CABLE_VALID_PLAYBACK (1 << SNDRV_PCM_STREAM_PLAYBACK)@@ -274,8 +271,8 @@ static int loopback_trigger(struct snd_pcm_substream *substream, int cmd)spin_lock (&cable->lock); cable->running |= stream; cable->pause &= ~stream; -spin_unlock (&cable->lock);loopback_timer_start (dpcm); +spin_unlock (&cable->lock);if (substream->stream ==SNDRV_PCM_STREAM_PLAYBACK )loopback_active_notify (dpcm);break ;@@ -283,23 +280,23 @@ static int loopback_trigger(struct snd_pcm_substream *substream, int cmd)spin_lock (&cable->lock); cable->running &= ~stream; cable->pause &= ~stream; -spin_unlock (&cable->lock);loopback_timer_stop (dpcm); +spin_unlock (&cable->lock);if (substream->stream ==SNDRV_PCM_STREAM_PLAYBACK )loopback_active_notify (dpcm);break ;case SNDRV_PCM_TRIGGER_PAUSE_PUSH :spin_lock (&cable->lock); cable->pause |= stream; -spin_unlock (&cable->lock);loopback_timer_stop (dpcm); +spin_unlock (&cable->lock);break ;case SNDRV_PCM_TRIGGER_PAUSE_RELEASE :spin_lock (&cable->lock); dpcm->last_jiffies = jiffies; cable->pause &= ~stream; -spin_unlock (&cable->lock);loopback_timer_start (dpcm); +spin_unlock (&cable->lock);break ;default :return -EINVAL ;@@ -477,6 +474,7 @@ static inline void bytepos_finish(struct loopback_pcm *dpcm, dpcm->buf_pos %= dpcm->pcm_buffer_size; } +/* call in cable->lock */ static unsigned int loopback_pos_update (struct loopback_cable *cable) { struct loopback_pcm *dpcm_play =@@ -485,9 +483,7 @@ static unsigned int loopback_pos_update(struct loopback_cable *cable) cable->streams[SNDRV_PCM_STREAM_CAPTURE ];unsigned long delta_play =0 , delta_capt =0 ;unsigned int running, count1, count2; -unsigned long flags; -spin_lock_irqsave (&cable->lock, flags); running = cable->running ^ cable->pause;if (running & (1 <<SNDRV_PCM_STREAM_PLAYBACK )) { delta_play = jiffies - dpcm_play->last_jiffies;@@ -529,32 +525,39 @@ static unsigned int loopback_pos_update(struct loopback_cable *cable)bytepos_finish (dpcm_play, count1);bytepos_finish (dpcm_capt, count1); unlock: -spin_unlock_irqrestore (&cable->lock, flags);return running; }static void loopback_timer_function (unsigned long data) {struct loopback_pcm *dpcm = (struct loopback_pcm *)data; -unsigned int running; +unsigned long flags; - running =loopback_pos_update (dpcm->cable); -if (running & (1 << dpcm->substream->stream)) { +spin_lock_irqsave (&dpcm->cable->lock, flags); +if (loopback_pos_update (dpcm->cable) & (1 << dpcm->substream->stream)) {loopback_timer_start (dpcm);if (dpcm->period_update_pending) { dpcm->period_update_pending =0 ; +spin_unlock_irqrestore (&dpcm->cable->lock, flags); +/* need to unlock before calling below */ snd_pcm_period_elapsed (dpcm->substream); +return ; } } +spin_unlock_irqrestore (&dpcm->cable->lock, flags); }static snd_pcm_uframes_t loopback_pointer (struct snd_pcm_substream *substream) {struct snd_pcm_runtime *runtime = substream->runtime;struct loopback_pcm *dpcm = runtime->private_data; +snd_pcm_uframes_t pos; +spin_lock (&dpcm->cable->lock);loopback_pos_update (dpcm->cable); -return bytes_to_frames (runtime, dpcm->buf_pos); + pos = dpcm->buf_pos; +spin_unlock (&dpcm->cable->lock); +return bytes_to_frames (runtime, pos); } static struct snd_pcm_hardware loopback_pcm_hardware =@@ -672,7 +675,6 @@ static int loopback_open(struct snd_pcm_substream *substream) dpcm->substream = substream;setup_timer (&dpcm->timer, loopback_timer_function, (unsigned long )dpcm); -spin_lock_init (&dpcm->timer_lock); cable = loopback->cables[substream->number][dev];if (!cable) {_______________________________________________
Alsa-devel mailing list
[email protected]
http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
At Sat, 20 Oct 2012 18:12:29 +0530,
Quoted 15 lines · Omair Mohammed Abdullah
Quoted 9 lines
Quoted 7 lines
If the timer is deleted while the timer handler is running, it may get rescheduled and an unnecessary period elapsed will be sent.
Add a flag to reschedule the timer only if it has not been stopped.
In which situation is it rescheduled exactly?
When the
loopback_timer_stop()runs on one CPU whileloopback_timer_function()runs on the other CPU. This causes theloopback_timer_function()to reschedule the timer beacusedel_timer()will not wait for the timer function to finish.Quoted 1 line
Quoted 7 lines
Quoted 6 lines
Actually there is a small race in the check of running bit and the timer lock. Do you mean this?
If so, wouldn't it be cleaner to get rid of
timer_lockand protect the whole start/stop withcable->lock, too, like below?Yes, I think this fixes the problem quite well.
OK, now applied my patch. Thanks for checking.
Takashi