topic Re: GStreamer audio pipeline hangs in i.MX Processors

GStreamer audio pipeline hangs

chikigai — Fri, 11 Jul 2014 06:04:35 GMT

Hi all,

I'm using i.MX6Q (with Freescale's BSP 3.0.35) and having issues with a GStreamer audio pipeline hanging in an infinite loop.

I have i.MX6Q as the I2S slave sending out an audio stream to an external device set as the I2S master.

Audio is streamed through the alsasink element without any problems until the external device stops the I2S clock.

When the I2S clock stops, the alsasink element keeps trying to write pcm buffers left in the pipeline (gstalsasink.c: gst_alsasink_write: snd_pcm_writei).

The pcm ring buffer in the kernel eventually becomes full, and returns EAGAIN (sound/core/pcm_lib.c: snd_pcm_lib_write1).

Upon receiving EAGAIN, the alsasink element retries indefinitely thinking that a buffer would free up eventually.

Obviously, a buffer is never going to free up since the kernel cannot write out the remaining pcm buffers when the I2S clock stops.

Once in this state, I cannot stop the pipeline since it is stuck in an infinite loop.

Are there any suggestions in overcoming this?

Currently, the best way I can think of is change the kernel code so that the remaining pcm buffers are thrown away when the I2S clock stops.

I checked the kernel source code contained in BSP 3.10.17, but it looks like the issue will still be present.

Any help would be greatly appreciated.

Re: GStreamer audio pipeline hangs

alejandrolozan1 — Wed, 23 Jul 2014 14:02:24 GMT

Hi,

I will delve into this. weidong.sun sorry to bother you, but do you have a clue about this?

Best Regards,

Alejandro

Re: GStreamer audio pipeline hangs

DarioCosta — Wed, 26 Nov 2014 14:12:20 GMT

Hi,

I'm having the same issue on a custom board with iMX6S.

Any news about this matter? Have you found a workaround/fix?

Thanks in advance

Dario

Re: GStreamer audio pipeline hangs

chikigai — Wed, 26 Nov 2014 23:30:23 GMT

I tried to see if a fix could be applied to the kernel, but unfortunately never managed.

As a temporary workaround, I currently have a standalone process handling the gstreamer pipelines which is forked/killed appropriately.

Not a proper fix by any means...