[alsa-devel] Alsa plugin unable to be loaded by Pulseaudio

[Ng Oon-Ee]

Hello, please find below an issue I had with loading plugins into
pulseaudio. The pulse dev has sent this mail on his list concerning
alsa's interaction with pulse, indicating that there is something buggy
with how it works. Details below. I'd be willing to post something to
the bugtracker, if I only knew what the below means...

-------- Forwarded Message --------
From: Lennart Poettering <lennart at poettering.net>
Reply-to: General PulseAudio Discussion
<pulseaudio-discuss at mail.0pointer.de>
To: pulseaudio-discuss at mail.0pointer.de
Subject: Re: [pulseaudio-discuss] Alsa output devices, able to use alsa
plugin?
Date: Fri, 4 Sep 2009 01:20:15 +0200

On Wed, 26.08.09 12:27, Tanu Kaskinen (tanuk at iki.fi) wrote:

> D: alsa-util.c: Trying test with SND_PCM_NO_AUTO_FORMAT ...
> D: alsa-util.c: Managed to open test
> D: alsa-util.c: Maximum hw buffer size is 371 ms
> I: alsa-util.c: snd_pcm_hw_params_set_periods_near() failed: Tiedostoa tai hakemistoa ei ole
> I: alsa-util.c: snd_pcm_hw_params_set_buffer_size_near() failed: Tiedostoa tai hakemistoa ei ole
> D: alsa-util.c: Trying test without SND_PCM_NO_AUTO_FORMAT ...
> D: alsa-util.c: Managed to open test
> D: alsa-util.c: Maximum hw buffer size is 371 ms
> I: alsa-util.c: snd_pcm_hw_params_set_periods_near() failed: Tiedostoa tai hakemistoa ei ole
> I: alsa-util.c: snd_pcm_hw_params_set_buffer_size_near() failed: Tiedostoa tai hakemistoa ei ole
> D: alsa-util.c: Trying plug:test with SND_PCM_NO_AUTO_FORMAT ...
> D: alsa-util.c: Managed to open plug:test
> D: alsa-util.c: Maximum hw buffer size is 371 ms
> I: alsa-util.c: snd_pcm_hw_params_set_periods_near() failed: Tiedostoa tai hakemistoa ei ole
> I: alsa-util.c: snd_pcm_hw_params_set_buffer_size_near() failed: Tiedostoa tai hakemistoa ei ole
> D: alsa-util.c: Trying plug:test without SND_PCM_NO_AUTO_FORMAT ...
> D: alsa-util.c: Managed to open plug:test
> D: alsa-util.c: Maximum hw buffer size is 371 ms
> I: alsa-util.c: snd_pcm_hw_params_set_periods_near() failed: Tiedostoa tai hakemistoa ei ole
> I: alsa-util.c: snd_pcm_hw_params_set_buffer_size_near() failed: Tiedostoa tai hakemistoa ei ole
> I: alsa-util.c: Failed to set hardware parameters on plug:test: Tiedostoa tai hakemistoa ei ole
> E: module.c: Failed to load  module "module-alsa-sink" (argument: "device=test"): initialization failed.
> 
> That "Tiedostoa tai hakemistoa ei ole" error means "No such file or
> directory". I tried also with device=hw:0,0 and device=plughw:0,0 and
> the first worked while the latter gave the same errors. Since the failed
> function calls pertain to buffering, I also tried with tsched=0. It
> worked. So, maybe this is a bug in the "plug" alsa plugin, ie. it
> doesn't implement the *_near() functions properly.
> 
> Lennart, any insights?

I am pretty sure that ALSA is broken here. We nowadays first try to
set the number of periods, and then set the buffer size in the
hwparams struct. If this fails we print a warning message (the ones
you see above with all the vowels regarding
snd_pcm_hw_params_set_periods_near()
resp. snd_pcm_hw_params_set_buffer_size_near()), but we do not
consider this fatal.

After doing this we call snd_pcm_hw_params() to activate our
settings -- and if this fails then we consider this a fatal problem.

Now the brokeness in ALSA is that after refusing our parameters it
still fails to configure our settings, i.e. calling the two functions
that fail breaks the hwparams setup: if we wouldn't call them, no breakage
happens at all. 

The two functions in question should either succeed or fail, but if
they fail they should not modify hwparams in a way that it
subsequently becomes unusable for snd_pcm_hw_params().

This issue needs to be fixed in ALSA. 

Lennart