[PATCH v3 0/3] ALSA: pcm: add snd_pcm_period_elapsed() variant without acquiring lock of PCM substream
Hi,
This patchset is revised version of my previous one: * https://lore.kernel.org/alsa-devel/20210609143145.146680-1-o-takashi@sakamoc...
All of drivers in ALSA firewire stack have two chances to process isochronous packets of any isochronous context; in software IRQ context for 1394 OHCI, and in process context of ALSA PCM application.
In the process context, callbacks of .pointer and .ack are utilized. The callbacks are done by ALSA PCM core under acquiring lock of PCM substream,
In design of ALSA PCM core, call of snd_pcm_period_elapsed() is used for drivers to awaken user processes from waiting for available frames. The function voluntarily acquires lock of PCM substream, therefore it is not called in the process context since it causes dead lock. As a workaround to avoid the dead lock, all of drivers in ALSA firewire stack use workqueue to delegate the call.
This patchset is my attempt for the issue. A variant of 'snd_pcm_period_elapsed()' without lock acquisition is going to be added, named 'snd_pcm_period_elapsed_under_stream_lock()'. The call is available in callbacks of .pointer and .ack of snd_pcm_ops structure.
Changes from v2: * delete context section from kernel API documentation
Takashi Sakamoto (3): ALSA: pcm: add snd_pcm_period_elapsed() variant without acquiring lock of PCM substream ALSA: firewire-lib: operate for period elapse event in process context ALSA: firewire-lib: obsolete workqueue for period update
include/sound/pcm.h | 1 + sound/core/pcm_lib.c | 64 +++++++++++++++++++++++++++-------- sound/firewire/amdtp-stream.c | 46 ++++++++----------------- sound/firewire/amdtp-stream.h | 1 - 4 files changed, 64 insertions(+), 48 deletions(-)
Current implementation of ALSA PCM core has a kernel API, snd_pcm_period_elapsed(), for drivers to queue event to awaken processes from waiting for available frames. The function voluntarily acquires lock of PCM substream, therefore it is not called in process context for any PCM operation since the lock is already acquired.
It is convenient for packet-oriented driver, at least for drivers to audio and music unit in IEEE 1394 bus. The drivers are allowed by Linux FireWire subsystem to process isochronous packets queued till recent isochronous cycle in process context in any time.
This commit adds snd_pcm_period_elapsed() variant, snd_pcm_period_elapsed_without_lock(), for drivers to queue the event in the process context.
Signed-off-by: Takashi Sakamoto o-takashi@sakamocchi.jp --- include/sound/pcm.h | 1 + sound/core/pcm_lib.c | 64 +++++++++++++++++++++++++++++++++----------- 2 files changed, 50 insertions(+), 15 deletions(-)
diff --git a/include/sound/pcm.h b/include/sound/pcm.h index 2e1200d17d0c..bae90696cd06 100644 --- a/include/sound/pcm.h +++ b/include/sound/pcm.h @@ -1066,6 +1066,7 @@ void snd_pcm_set_ops(struct snd_pcm * pcm, int direction, void snd_pcm_set_sync(struct snd_pcm_substream *substream); int snd_pcm_lib_ioctl(struct snd_pcm_substream *substream, unsigned int cmd, void *arg); +void snd_pcm_period_elapsed_under_stream_lock(struct snd_pcm_substream *substream); void snd_pcm_period_elapsed(struct snd_pcm_substream *substream); snd_pcm_sframes_t __snd_pcm_lib_xfer(struct snd_pcm_substream *substream, void *buf, bool interleaved, diff --git a/sound/core/pcm_lib.c b/sound/core/pcm_lib.c index b7e3d8f44511..7d5883432085 100644 --- a/sound/core/pcm_lib.c +++ b/sound/core/pcm_lib.c @@ -1778,27 +1778,38 @@ int snd_pcm_lib_ioctl(struct snd_pcm_substream *substream, EXPORT_SYMBOL(snd_pcm_lib_ioctl);
/** - * snd_pcm_period_elapsed - update the pcm status for the next period - * @substream: the pcm substream instance + * snd_pcm_period_elapsed_under_stream_lock() - update the status of runtime for the next period + * under acquired lock of PCM substream. + * @substream: the instance of pcm substream. + * + * This function is called when the batch of audio data frames as the same size as the period of + * buffer is already processed in audio data transmission. + * + * The call of function updates the status of runtime with the latest position of audio data + * transmission, checks overrun and underrun over buffer, awaken user processes from waiting for + * available audio data frames, sampling audio timestamp, and performs stop or drain the PCM + * substream according to configured threshold. + * + * The function is intended to use for the case that PCM driver operates audio data frames under + * acquired lock of PCM substream; e.g. in callback of any operation of &snd_pcm_ops in process + * context. In any interrupt context, it's preferrable to use ``snd_pcm_period_elapsed()`` instead + * since lock of PCM substream should be acquired in advance. * - * This function is called from the interrupt handler when the - * PCM has processed the period size. It will update the current - * pointer, wake up sleepers, etc. + * Developer should pay enough attention that some callbacks in &snd_pcm_ops are done by the call of + * function: * - * Even if more than one periods have elapsed since the last call, you - * have to call this only once. + * - .pointer - to retrieve current position of audio data transmission by frame count or XRUN state. + * - .trigger - with SNDRV_PCM_TRIGGER_STOP at XRUN or DRAINING state. + * - .get_time_info - to retrieve audio time stamp if needed. + * + * Even if more than one periods have elapsed since the last call, you have to call this only once. */ -void snd_pcm_period_elapsed(struct snd_pcm_substream *substream) +void snd_pcm_period_elapsed_under_stream_lock(struct snd_pcm_substream *substream) { struct snd_pcm_runtime *runtime; - unsigned long flags;
- if (snd_BUG_ON(!substream)) - return; - - snd_pcm_stream_lock_irqsave(substream, flags); if (PCM_RUNTIME_CHECK(substream)) - goto _unlock; + return; runtime = substream->runtime;
if (!snd_pcm_running(substream) || @@ -1811,7 +1822,30 @@ void snd_pcm_period_elapsed(struct snd_pcm_substream *substream) #endif _end: kill_fasync(&runtime->fasync, SIGIO, POLL_IN); - _unlock: +} +EXPORT_SYMBOL(snd_pcm_period_elapsed_under_stream_lock); + +/** + * snd_pcm_period_elapsed() - update the status of runtime for the next period by acquiring lock of + * PCM substream. + * @substream: the instance of PCM substream. + * + * This function is mostly similar to ``snd_pcm_period_elapsed_under_stream_lock()`` except for + * acquiring lock of PCM substream voluntarily. + * + * It's typically called by any type of IRQ handler when hardware IRQ occurs to notify event that + * the batch of audio data frames as the same size as the period of buffer is already processed in + * audio data transmission. + */ +void snd_pcm_period_elapsed(struct snd_pcm_substream *substream) +{ + unsigned long flags; + + if (snd_BUG_ON(!substream)) + return; + + snd_pcm_stream_lock_irqsave(substream, flags); + snd_pcm_period_elapsed_under_stream_lock(substream); snd_pcm_stream_unlock_irqrestore(substream, flags); } EXPORT_SYMBOL(snd_pcm_period_elapsed);
All of drivers in ALSA firewire stack processes two chances to process isochronous packets in any isochronous context; in software IRQ context for 1394 OHCI, and in process context of ALSA PCM application.
In the process context, callbacks of .pointer and .ack are utilized. The callbacks are done by ALSA PCM core under acquiring lock of PCM substream,
In design of ALSA PCM core, call of snd_pcm_period_elapsed() is used for drivers to awaken user processes from waiting for available frames. The function voluntarily acquires lock of PCM substream, therefore it is not called in the process context since it causes dead lock.
As a workaround to avoid the dead lock, all of drivers in ALSA firewire stack uses workqueue to delegate the call. A variant of snd_pcm_period_elapsed() without lock acquisition can obsolete the workqueue.
An extra care is needed for the callback of .pointer since it's called from snd_pcm_period_elapsed(). The isochronous context in Linux FireWire subsystem is safe mostly for nested call except in software IRQ context.
Signed-off-by: Takashi Sakamoto o-takashi@sakamocchi.jp --- sound/firewire/amdtp-stream.c | 31 ++++++++++++++----------------- 1 file changed, 14 insertions(+), 17 deletions(-)
diff --git a/sound/firewire/amdtp-stream.c b/sound/firewire/amdtp-stream.c index 150ee0b9e707..426a85b56cf1 100644 --- a/sound/firewire/amdtp-stream.c +++ b/sound/firewire/amdtp-stream.c @@ -613,8 +613,16 @@ static void update_pcm_pointers(struct amdtp_stream *s, // The program in user process should periodically check the status of intermediate // buffer associated to PCM substream to process PCM frames in the buffer, instead // of receiving notification of period elapsed by poll wait. - if (!pcm->runtime->no_period_wakeup) - queue_work(system_highpri_wq, &s->period_work); + if (!pcm->runtime->no_period_wakeup) { + if (in_interrupt()) { + // In software IRQ context for 1394 OHCI. + snd_pcm_period_elapsed(pcm); + } else { + // In process context of ALSA PCM application under acquired lock of + // PCM substream. + snd_pcm_period_elapsed_under_stream_lock(pcm); + } + } } }
@@ -1740,22 +1748,11 @@ unsigned long amdtp_domain_stream_pcm_pointer(struct amdtp_domain *d, { struct amdtp_stream *irq_target = d->irq_target;
+ // Process isochronous packets queued till recent isochronous cycle to handle PCM frames. if (irq_target && amdtp_stream_running(irq_target)) { - // This function is called in software IRQ context of - // period_work or process context. - // - // When the software IRQ context was scheduled by software IRQ - // context of IT contexts, queued packets were already handled. - // Therefore, no need to flush the queue in buffer furthermore. - // - // When the process context reach here, some packets will be - // already queued in the buffer. These packets should be handled - // immediately to keep better granularity of PCM pointer. - // - // Later, the process context will sometimes schedules software - // IRQ context of the period_work. Then, no need to flush the - // queue by the same reason as described in the above - if (current_work() != &s->period_work) + // In software IRQ context, the call causes dead-lock to disable the tasklet + // synchronously. + if (!in_interrupt()) fw_iso_context_flush_completions(irq_target->context); }
The workqueue to notify PCM period elapse is not used anymore.
Signed-off-by: Takashi Sakamoto o-takashi@sakamocchi.jp --- sound/firewire/amdtp-stream.c | 15 --------------- sound/firewire/amdtp-stream.h | 1 - 2 files changed, 16 deletions(-)
diff --git a/sound/firewire/amdtp-stream.c b/sound/firewire/amdtp-stream.c index 426a85b56cf1..1d9bc7b07df1 100644 --- a/sound/firewire/amdtp-stream.c +++ b/sound/firewire/amdtp-stream.c @@ -77,8 +77,6 @@ // overrun. Actual device can skip more, then this module stops the packet streaming. #define IR_JUMBO_PAYLOAD_MAX_SKIP_CYCLES 5
-static void pcm_period_work(struct work_struct *work); - /** * amdtp_stream_init - initialize an AMDTP stream structure * @s: the AMDTP stream to initialize @@ -107,7 +105,6 @@ int amdtp_stream_init(struct amdtp_stream *s, struct fw_unit *unit, s->flags = flags; s->context = ERR_PTR(-1); mutex_init(&s->mutex); - INIT_WORK(&s->period_work, pcm_period_work); s->packet_index = 0;
init_waitqueue_head(&s->ready_wait); @@ -346,7 +343,6 @@ EXPORT_SYMBOL(amdtp_stream_get_max_payload); */ void amdtp_stream_pcm_prepare(struct amdtp_stream *s) { - cancel_work_sync(&s->period_work); s->pcm_buffer_pointer = 0; s->pcm_period_pointer = 0; } @@ -626,16 +622,6 @@ static void update_pcm_pointers(struct amdtp_stream *s, } }
-static void pcm_period_work(struct work_struct *work) -{ - struct amdtp_stream *s = container_of(work, struct amdtp_stream, - period_work); - struct snd_pcm_substream *pcm = READ_ONCE(s->pcm); - - if (pcm) - snd_pcm_period_elapsed(pcm); -} - static int queue_packet(struct amdtp_stream *s, struct fw_iso_packet *params, bool sched_irq) { @@ -1808,7 +1794,6 @@ static void amdtp_stream_stop(struct amdtp_stream *s) return; }
- cancel_work_sync(&s->period_work); fw_iso_context_stop(s->context); fw_iso_context_destroy(s->context); s->context = ERR_PTR(-1); diff --git a/sound/firewire/amdtp-stream.h b/sound/firewire/amdtp-stream.h index b25592d5f6af..1f957c946c95 100644 --- a/sound/firewire/amdtp-stream.h +++ b/sound/firewire/amdtp-stream.h @@ -186,7 +186,6 @@ struct amdtp_stream {
/* For a PCM substream processing. */ struct snd_pcm_substream *pcm; - struct work_struct period_work; snd_pcm_uframes_t pcm_buffer_pointer; unsigned int pcm_period_pointer;
On Thu, 10 Jun 2021 05:17:30 +0200, Takashi Sakamoto wrote:
Hi,
This patchset is revised version of my previous one:
All of drivers in ALSA firewire stack have two chances to process isochronous packets of any isochronous context; in software IRQ context for 1394 OHCI, and in process context of ALSA PCM application.
In the process context, callbacks of .pointer and .ack are utilized. The callbacks are done by ALSA PCM core under acquiring lock of PCM substream,
In design of ALSA PCM core, call of snd_pcm_period_elapsed() is used for drivers to awaken user processes from waiting for available frames. The function voluntarily acquires lock of PCM substream, therefore it is not called in the process context since it causes dead lock. As a workaround to avoid the dead lock, all of drivers in ALSA firewire stack use workqueue to delegate the call.
This patchset is my attempt for the issue. A variant of 'snd_pcm_period_elapsed()' without lock acquisition is going to be added, named 'snd_pcm_period_elapsed_under_stream_lock()'. The call is available in callbacks of .pointer and .ack of snd_pcm_ops structure.
Changes from v2:
- delete context section from kernel API documentation
Takashi Sakamoto (3): ALSA: pcm: add snd_pcm_period_elapsed() variant without acquiring lock of PCM substream ALSA: firewire-lib: operate for period elapse event in process context ALSA: firewire-lib: obsolete workqueue for period update
Applied all three patches now. Thanks.
Takashi
participants (2)
-
Takashi Iwai
-
Takashi Sakamoto