imx-ssi fixes

imx-ssi fixes

[Sascha Hauer]

Just a resend of the patches I sent yesterday with an updated commit message.

Sascha

[PATCH 1/2] imx-ssi: do not call hrtimer_disable in trigger function

[Sascha Hauer]

Doing so causes a deadlock, so just signal the timer to stop
using an atomic variable.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
 sound/soc/imx/imx-pcm-fiq.c |   10 ++++++++--
 1 files changed, 8 insertions(+), 2 deletions(-)

diff --git a/sound/soc/imx/imx-pcm-fiq.c b/sound/soc/imx/imx-pcm-fiq.c
index 2646e05..24670ee 100644
--- a/sound/soc/imx/imx-pcm-fiq.c
+++ b/sound/soc/imx/imx-pcm-fiq.c
@@ -42,6 +42,7 @@ struct imx_pcm_runtime_data {
 	struct hrtimer hrt;
 	int poll_time_ns;
 	struct snd_pcm_substream *substream;
+	atomic_t running;
 };
 
 static enum hrtimer_restart snd_hrtimer_callback(struct hrtimer *hrt)
@@ -53,6 +54,9 @@ static enum hrtimer_restart snd_hrtimer_callback(struct hrtimer *hrt)
 	struct pt_regs regs;
 	unsigned long delta;
 
+	if (!atomic_read(&iprtd->running))
+		return HRTIMER_NORESTART;
+
 	get_fiq_regs(&regs);
 
 	if (substream->stream == SNDRV_PCM_STREAM_PLAYBACK)
@@ -130,6 +134,7 @@ static int snd_imx_pcm_trigger(struct snd_pcm_substream *substream, int cmd)
 	case SNDRV_PCM_TRIGGER_START:
 	case SNDRV_PCM_TRIGGER_RESUME:
 	case SNDRV_PCM_TRIGGER_PAUSE_RELEASE:
+		atomic_set(&iprtd->running, 1);
 		hrtimer_start(&iprtd->hrt, ns_to_ktime(iprtd->poll_time_ns),
 		      HRTIMER_MODE_REL);
 		if (++fiq_enable == 1)
@@ -140,11 +145,11 @@ static int snd_imx_pcm_trigger(struct snd_pcm_substream *substream, int cmd)
 	case SNDRV_PCM_TRIGGER_STOP:
 	case SNDRV_PCM_TRIGGER_SUSPEND:
 	case SNDRV_PCM_TRIGGER_PAUSE_PUSH:
-		hrtimer_cancel(&iprtd->hrt);
+		atomic_set(&iprtd->running, 0);
+
 		if (--fiq_enable == 0)
 			disable_fiq(imx_pcm_fiq);
 
-
 		break;
 	default:
 		return -EINVAL;
@@ -191,6 +196,7 @@ static int snd_imx_open(struct snd_pcm_substream *substream)
 
 	iprtd->substream = substream;
 
+	atomic_set(&iprtd->running, 0);
 	hrtimer_init(&iprtd->hrt, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
 	iprtd->hrt.function = snd_hrtimer_callback;
 
-- 
1.7.0

Re: [PATCH 1/2] imx-ssi: do not call hrtimer_disable in trigger function

[Liam Girdwood]

On Wed, 2010-04-14 at 09:17 +0200, Sascha Hauer wrote:
> Doing so causes a deadlock, so just signal the timer to stop
> using an atomic variable.
> 
> Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

Acked-by: Liam Girdwood <lrg@slimlogic.co.uk>

-- 
Freelance Developer, SlimLogic Ltd
ASoC and Voltage Regulator Maintainer.
http://www.slimlogic.co.uk

[PATCH 2/2] imx-ssi: increase minimum periods to 4

[Sascha Hauer]

Currently the notification of elapsed periods is not very exact.
Increase minimum periods to 4 as suggested by Liam Girdwood.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Acked-by: Liam Girdwood <lrg@slimlogic.co.uk>
---
 sound/soc/imx/imx-pcm-fiq.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/sound/soc/imx/imx-pcm-fiq.c b/sound/soc/imx/imx-pcm-fiq.c
index 24670ee..6b518e0 100644
--- a/sound/soc/imx/imx-pcm-fiq.c
+++ b/sound/soc/imx/imx-pcm-fiq.c
@@ -180,7 +180,7 @@ static struct snd_pcm_hardware snd_imx_hardware = {
 	.buffer_bytes_max = IMX_SSI_DMABUF_SIZE,
 	.period_bytes_min = 128,
 	.period_bytes_max = 16 * 1024,
-	.periods_min = 2,
+	.periods_min = 4,
 	.periods_max = 255,
 	.fifo_size = 0,
 };
-- 
1.7.0

Re: imx-ssi fixes

[Mark Brown]

On Wed, Apr 14, 2010 at 09:17:29AM +0200, Sascha Hauer wrote:
> Just a resend of the patches I sent yesterday with an updated commit message.

Applied both, thanks.

imx-ssi fixes

[Sascha Hauer]

Hi all,

Here are some bugfix patches for the imx-ssi driver. It should also fix
the problems reported in FIQ mode.

Sascha

Re: imx-ssi fixes

[Mark Brown]

On Thu, Apr 08, 2010 at 11:31:23AM +0200, Sascha Hauer wrote:

> Here are some bugfix patches for the imx-ssi driver. It should also fix
> the problems reported in FIQ mode.

Applied all three, thanks - the hrtimer stuff works an awful lot better
than the vanilla timer stuff did.  It's a shame I hadn't noticed that
hrtimers did manage to give better resolution on i.MX when I looked into
this previously.

It'd be good if you could look into the timer sync issue - doing
something like deferring the timer by a proportion of a period if
the period hasn't elapsed ought to cover it.

Re: imx-ssi fixes

[Sascha Hauer]

On Thu, Apr 08, 2010 at 03:47:50PM +0100, Mark Brown wrote:
> On Thu, Apr 08, 2010 at 11:31:23AM +0200, Sascha Hauer wrote:
> 
> > Here are some bugfix patches for the imx-ssi driver. It should also fix
> > the problems reported in FIQ mode.
> 
> Applied all three, thanks - the hrtimer stuff works an awful lot better
> than the vanilla timer stuff did.  It's a shame I hadn't noticed that
> hrtimers did manage to give better resolution on i.MX when I looked into
> this previously.
> 
> It'd be good if you could look into the timer sync issue - doing
> something like deferring the timer by a proportion of a period if
> the period hasn't elapsed ought to cover it.

I'll look into it. I'm thinking of

- busywaiting if the current period hasn't elapsed (should rarely if
  ever the case since the fiq works reliable and the timer interrupt
  introduces additional delays.
- calculate the next callback time based on the current offset.

If you haven't pushed the patches forward it would be good to delay the
hrtimer patch a bit. I managed to shoot the system down while playing
audio with a hackbench 10. This doesn't happen without the hrtimer
patch. This may be related to the problem Valentin reported.

Sascha

-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

Re: imx-ssi fixes

[Mark Brown]

On Thu, Apr 08, 2010 at 05:51:13PM +0200, Sascha Hauer wrote:

> If you haven't pushed the patches forward it would be good to delay the
> hrtimer patch a bit. I managed to shoot the system down while playing
> audio with a hackbench 10. This doesn't happen without the hrtimer
> patch. This may be related to the problem Valentin reported.

Already pushed to Takashi, but he's not pulled yet.