alsa-project/alsa-lib issue #51 was edited from bondagit:

Hi all,

I am currently working on an open source project on GitHub [AES67 Linux daemon implementation](https://github.com/bondagit/aes67-linux-daemon). This project uses the [Merging Technologies ALSA RAVENNA/AES67 Linux Driver](https://bitbucket.org/MergingTechnologies/ravenna-alsa-lkm/src/master). This ALSA driver does support the mmap access using the indirect PCM data transfer.

The driver is working properly when used directly but not when it's used via the **dshare** plugin. See issue [No sound using dshare in asound.conf](https://github.com/bondagit/aes67-linux-daemon/issues/7) in the AES67 Linux daemon project.

For all my tests I used the **Linux Kernel 5.3.0-51** and the **alsa-lib v1.1.9** but I get same results with most recent kernel and with the **alsa-lib** on the master branch.

Basically when I playback a sample PCM audio using the ALSA RAVENNA/AES67 HW device directly via mmap I get a proper playout. Ex: `aplay -D hw:1 sample.raw -M` But if I do the same using the dshare plugin I do always get no audio (silence): Ex: `aplay -D mydshare sample.raw -M`

After a number of tests trying to compare the behavior of the **alsa-lib** when using the ALSA driver directly (aplay -D hw:1) and when using the plugin (aplay -D mydshare) I realized that in the second case the **SNDRV_PCM_IOCTL_SYNC_PTR ioctl** is never triggered in the audio loop by the **alsa-lib** . (I could easily capture this by using strace). I would expect the `snd_pcm_dshare_mmap_commit() `function of **alsa-lib** to invoke the above ioctl but this is not happening. The result is that in the ALSA driver the **ack()** callback is not called to indicate that the **appl_ptr** is updated and this causes the problem we have.

In the Linux kernel when the ioctl SNDRV_PCM_IOCTL_SYNC_PTR is called the following functions are called: `snd_pcm_sync_ptr() -> pcm_lib_apply_appl_ptr() -> substream->ops->ack() ` but as stated above this is happening only if the ALSA driver is used directly (aplay -D hw:1) and not when it's used via the plugin (aplay -D mydshare).

I could fix the issue applying the following temporary fix to the `alsa-lib`: ``` diff --git a/src/pcm/pcm_dshare.c b/src/pcm/pcm_dshare.c index f135b5df..1f103414 100644 --- a/src/pcm/pcm_dshare.c +++ b/src/pcm/pcm_dshare.c @@ -562,6 +562,7 @@ static snd_pcm_sframes_t snd_pcm_dshare_mmap_commit(snd_pcm_t *pcm, /* clear timer queue to avoid a bogus return from poll */ if (snd_pcm_mmap_playback_avail(pcm) < pcm->avail_min) snd_pcm_direct_clear_timer_queue(dshare); + dshare->spcm->fast_ops->mmap_commit(dshare->spcm, offset, size); } return size; } ``` The fix basically triggers the call to the `snd_pcm_hw_mmap_commit()` function of the slave pcm of the dshare plugin. This function invokes the ` issue_applptr() ` and then the `sync_ptr1()` that finally calls the SNDRV_PCM_IOCTL_SYNC_PTR ioctl to update the **appl_ptr** in the kernel.

After applying this fix I can properly hear the audio.

I am available to provide any additional information and to do additional tests. Since I don't have much experience with the `alsa-lib` it can be that the problem is related to some misconfiguration but at present I managed to solve the issue only by applying the above fix.

Thanks in advance for your help.

Issue URL : https://github.com/alsa-project/alsa-lib/issues/51 Repository URL: https://github.com/alsa-project/alsa-lib