[RFC] [PATCH] ASoC: OMAP: full duplex mode fix
This patch tries to correct the problem of full duplex mode not working over a single McBSP based CPU DAI. Created against linux-2.6.31-rc5. Tested on Amstrad Delta. Signed-off-by: Janusz Krzysztofik <jkrzyszt@tis.icnet.pl> --- --- linux-2.6.31-rc5/sound/soc/omap/omap-mcbsp.c.orig 2009-08-01 02:40:45.000000000 +0200 +++ linux-2.6.31-rc5/sound/soc/omap/omap-mcbsp.c 2009-08-03 03:28:07.000000000 +0200 @@ -183,6 +183,7 @@ static int omap_mcbsp_dai_trigger(struct struct snd_soc_pcm_runtime *rtd = substream->private_data; struct snd_soc_dai *cpu_dai = rtd->dai->cpu_dai; struct omap_mcbsp_data *mcbsp_data = to_mcbsp(cpu_dai->private_data); + u16 buf; int err = 0; switch (cmd) { @@ -191,6 +192,14 @@ static int omap_mcbsp_dai_trigger(struct case SNDRV_PCM_TRIGGER_PAUSE_RELEASE: if (!mcbsp_data->active++) omap_mcbsp_start(mcbsp_data->bus_id); + else if (substream->stream == SNDRV_PCM_STREAM_PLAYBACK) + /* looks like capture already in progress, + * start playback by taking it out of error condition */ + omap_mcbsp_pollwrite(mcbsp_data->bus_id, 0x0); + else + /* looks like playback already in progress, + * start capture by taking it out of error condition */ + omap_mcbsp_pollread(mcbsp_data->bus_id, &buf); break; case SNDRV_PCM_TRIGGER_STOP:
Hi On Mon, 3 Aug 2009 03:32:04 +0200 Janusz Krzysztofik <jkrzyszt@tis.icnet.pl> wrote:
This patch tries to correct the problem of full duplex mode not working over a single McBSP based CPU DAI.
Created against linux-2.6.31-rc5. Tested on Amstrad Delta.
Do you have some specific test case how to trigger this? I haven't seen this on 2420 or 34xx (e.g. with 'arecord -d 1 -f dat |aplay') but I have no doubt that this can happen on 1510. At least this doesn't cause any harm on Beagle so I'm fine with the fix.
@@ -191,6 +192,14 @@ static int omap_mcbsp_dai_trigger(struct case SNDRV_PCM_TRIGGER_PAUSE_RELEASE: if (!mcbsp_data->active++) omap_mcbsp_start(mcbsp_data->bus_id); + else if (substream->stream == SNDRV_PCM_STREAM_PLAYBACK) + /* looks like capture already in progress, + * start playback by taking it out of error condition */ + omap_mcbsp_pollwrite(mcbsp_data->bus_id, 0x0); + else + /* looks like playback already in progress, + * start capture by taking it out of error condition */ + omap_mcbsp_pollread(mcbsp_data->bus_id, &buf); break; Minor note: See preferred style for multi-line comments in Documentation/CodingStyle. I'm not 100 % sure about the braces but I think they are also preferred if there are indented comment lines with the single code line.
-- Jarkko
On Mon, Aug 03, 2009 at 11:29:50AM +0300, Jarkko Nikula wrote:
Janusz Krzysztofik <jkrzyszt@tis.icnet.pl> wrote:
+ else + /* looks like playback already in progress, + * start capture by taking it out of error condition */ + omap_mcbsp_pollread(mcbsp_data->bus_id, &buf);
Minor note: See preferred style for multi-line comments in Documentation/CodingStyle. I'm not 100 % sure about the braces but I think they are also preferred if there are indented comment lines with the single code line.
Indeed; only omit the braces if the *only* thing in the block is a single statement.
Jarkko Nikula wrote:
On Mon, 3 Aug 2009 03:32:04 +0200 Janusz Krzysztofik <jkrzyszt@tis.icnet.pl> wrote:
This patch tries to correct the problem of full duplex mode not working over a single McBSP based CPU DAI.
Created against linux-2.6.31-rc5. Tested on Amstrad Delta.
Do you have some specific test case how to trigger this? I haven't seen this on 2420 or 34xx (e.g. with 'arecord -d 1 -f dat |aplay') but I have no doubt that this can happen on 1510. At least this doesn't cause any harm on Beagle so I'm fine with the fix.
Hi, I made more testing on my OMAP1510 and found out that I could get your example usage working without my patch, but only if started like this: arecord -D hw:0,0 -f S16_LE|aplay -D hw:0,0 If I start the same with "-D hw:0,0" omitted from aplay, it doesn't work any longer, waiting forever. It definitelly doesn't work if I start capture and playback one after another, no matter which one goes first (record while playing or play while recording). So it looks like starting both streams simultaneously can do the job, but a short delay breaks it. With my patch, it seems to work fine for me in all cases. Jarkko, have you ever tried it on your OMAP2/3 with parallel playback and capture started one after another, not simultaneously? Arun, can your snd-soc-osk9512 work on OMAP1610 in full duplex mode without any limitations? If the problem appears to be OMAP1510 or AMS_DELTA specific, I can add a check for a machine or cpu type to avoid braking unaffected machines. Thanks, Janusz
On Mon, 03 Aug 2009 16:00:32 +0200 Janusz Krzysztofik <jkrzyszt@tis.icnet.pl> wrote:
Hi, I made more testing on my OMAP1510 and found out that I could get your example usage working without my patch, but only if started like this:
arecord -D hw:0,0 -f S16_LE|aplay -D hw:0,0
If I start the same with "-D hw:0,0" omitted from aplay, it doesn't work any longer, waiting forever. It definitelly doesn't work if I start capture and playback one after another, no matter which one goes first (record while playing or play while recording). So it looks like starting both streams simultaneously can do the job, but a short delay breaks it.
With my patch, it seems to work fine for me in all cases.
So it looks that McBSP-DMA for another stream cease to work if there is more than certain delay between first stream start and second one but omap_mcbsp_pollwrite or _pollread will clear the error condition. Can you debug is this due clearing the possible XSYNC_ERR, waiting for transmit/receive confirmation or playing with data registers there?
Jarkko, have you ever tried it on your OMAP2/3 with parallel playback and capture started one after another, not simultaneously?
Yes I have. Like recording into file (arecord /dev/shm/foo) and start playing it after some time (aplay /dev/shm/foo) or playing a file and start recording into another.
If the problem appears to be OMAP1510 or AMS_DELTA specific, I can add a check for a machine or cpu type to avoid braking unaffected machines.
I'm thinking can your platform show some existing problem not noted before... but yes, check for 1510 would be bit safer now and is also kind of revisit comment why 1510 needs special handling. -- Jarkko
On Mon, Aug 3, 2009 at 7:00 AM, Janusz Krzysztofik <jkrzyszt@tis.icnet.pl>wrote:
Jarkko Nikula wrote:
On Mon, 3 Aug 2009 03:32:04 +0200 Janusz Krzysztofik <jkrzyszt@tis.icnet.pl> wrote:
This patch tries to correct the problem of full duplex mode not working
over a single McBSP based CPU DAI.
Created against linux-2.6.31-rc5. Tested on Amstrad Delta.
Do you have some specific test case how to trigger this? I haven't seen this on 2420 or 34xx (e.g. with 'arecord -d 1 -f dat |aplay') but I have no doubt that this can happen on 1510. At least this doesn't cause any harm on Beagle so I'm fine with the fix.
Hi, I made more testing on my OMAP1510 and found out that I could get your example usage working without my patch, but only if started like this:
arecord -D hw:0,0 -f S16_LE|aplay -D hw:0,0
If I start the same with "-D hw:0,0" omitted from aplay, it doesn't work any longer, waiting forever. It definitelly doesn't work if I start capture and playback one after another, no matter which one goes first (record while playing or play while recording). So it looks like starting both streams simultaneously can do the job, but a short delay breaks it.
With my patch, it seems to work fine for me in all cases.
Jarkko, have you ever tried it on your OMAP2/3 with parallel playback and capture started one after another, not simultaneously?
Arun, can your snd-soc-osk9512 work on OMAP1610 in full duplex mode without any limitations?
Janusz, Haven't done testing in full duplex mode. I don't have access to osk5912 board now. If someone has got osk and do the testing it ll be good. It ll take at least another 2 more month for me to do the testing on osk. Regards, Arun
If the problem appears to be OMAP1510 or AMS_DELTA specific, I can add a check for a machine or cpu type to avoid braking unaffected machines.
Thanks, Janusz
-- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
CCing Jesslyn Abdul Salam <jesslyn.abdulsalam@gmail.com> Hope he has one osk. On Mon, Aug 3, 2009 at 10:53 AM, Arun KS <arunks@mistralsolutions.com> wrote:
On Mon, Aug 3, 2009 at 7:00 AM, Janusz Krzysztofik <jkrzyszt@tis.icnet.pl> wrote:
Jarkko Nikula wrote:
On Mon, 3 Aug 2009 03:32:04 +0200 Janusz Krzysztofik <jkrzyszt@tis.icnet.pl> wrote:
This patch tries to correct the problem of full duplex mode not working over a single McBSP based CPU DAI.
Created against linux-2.6.31-rc5. Tested on Amstrad Delta.
Do you have some specific test case how to trigger this? I haven't seen this on 2420 or 34xx (e.g. with 'arecord -d 1 -f dat |aplay') but I have no doubt that this can happen on 1510. At least this doesn't cause any harm on Beagle so I'm fine with the fix.
Hi, I made more testing on my OMAP1510 and found out that I could get your example usage working without my patch, but only if started like this:
arecord -D hw:0,0 -f S16_LE|aplay -D hw:0,0
If I start the same with "-D hw:0,0" omitted from aplay, it doesn't work any longer, waiting forever. It definitelly doesn't work if I start capture and playback one after another, no matter which one goes first (record while playing or play while recording). So it looks like starting both streams simultaneously can do the job, but a short delay breaks it.
With my patch, it seems to work fine for me in all cases.
Jarkko, have you ever tried it on your OMAP2/3 with parallel playback and capture started one after another, not simultaneously?
Arun, can your snd-soc-osk9512 work on OMAP1610 in full duplex mode without any limitations?
Janusz,
Haven't done testing in full duplex mode. I don't have access to osk5912 board now. If someone has got osk and do the testing it ll be good. It ll take at least another 2 more month for me to do the testing on osk.
Regards, Arun
If the problem appears to be OMAP1510 or AMS_DELTA specific, I can add a check for a machine or cpu type to avoid braking unaffected machines.
Thanks, Janusz
-- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
participants (4)
-
Arun KS -
Janusz Krzysztofik -
Jarkko Nikula -
Mark Brown