Re: BUG: soft lockup when recording audio on MX28EVK with ASoC sgtl5000

Re: BUG: soft lockup when recording audio on MX28EVK with ASoC sgtl5000

[Fabio Estevam]

[-- Attachment #1: Type: text/plain, Size: 802 bytes --]

Hi Hector,

On Mon, Mar 25, 2013 at 7:52 AM, Hector Palacios
<hector.palacios@digi.com> wrote:
> Hello,
>
> I just tried recording audio on Freescale's MX28EVK that uses ASoC sgtl5000
> (kernel v3.8) with:
>
>         arecord -M -f cd sound.wav --duration 10
>
> and got a scheduler message:
>
> [  789.041847] [sched_delayed] sched: RT throttling activated

I have just tested arecord on 3.8,4 and I haven't seen the 'soft lockup' bug.

However, I see that mxs saif gets busy after the record and any
attempt to do an 'aplay' after the first arecord fails.

With the attached patch applied I am able to do arecord/aplay sequence
several times.

I don't have a cable handy here to actually test if the sound is
recorded properly or not.

Could you please test it against 3.8.4?

Thanks,

Fabio Estevam

[-- Attachment #2: 0001-Do-not-check-busy.patch --]
[-- Type: application/octet-stream, Size: 2747 bytes --]

From dde98beec0ffdd84d497b63923cf26e8c9b0a5c7 Mon Sep 17 00:00:00 2001
From: Fabio Estevam <fabio.estevam@freescale.com>
Date: Mon, 25 Mar 2013 16:23:04 -0300
Subject: [PATCH] ASoC: mxs-saif: Do not check for BM_SAIF_STAT_BUSY

Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
---
 sound/soc/mxs/mxs-saif.c |   30 ++----------------------------
 1 file changed, 2 insertions(+), 28 deletions(-)

diff --git a/sound/soc/mxs/mxs-saif.c b/sound/soc/mxs/mxs-saif.c
index 365d9d2..1986deb 100644
--- a/sound/soc/mxs/mxs-saif.c
+++ b/sound/soc/mxs/mxs-saif.c
@@ -207,17 +207,10 @@ static int mxs_saif_set_clk(struct mxs_saif *saif,
 int mxs_saif_put_mclk(unsigned int saif_id)
 {
 	struct mxs_saif *saif = mxs_saif[saif_id];
-	u32 stat;
 
 	if (!saif)
 		return -EINVAL;
 
-	stat = __raw_readl(saif->base + SAIF_STAT);
-	if (stat & BM_SAIF_STAT_BUSY) {
-		dev_err(saif->dev, "error: busy\n");
-		return -EBUSY;
-	}
-
 	clk_disable_unprepare(saif->clk);
 
 	/* disable MCLK output */
@@ -241,7 +234,6 @@ int mxs_saif_get_mclk(unsigned int saif_id, unsigned int mclk,
 					unsigned int rate)
 {
 	struct mxs_saif *saif = mxs_saif[saif_id];
-	u32 stat;
 	int ret;
 	struct mxs_saif *master_saif;
 
@@ -262,12 +254,6 @@ int mxs_saif_get_mclk(unsigned int saif_id, unsigned int mclk,
 		return -EINVAL;
 	}
 
-	stat = __raw_readl(saif->base + SAIF_STAT);
-	if (stat & BM_SAIF_STAT_BUSY) {
-		dev_err(saif->dev, "error: busy\n");
-		return -EBUSY;
-	}
-
 	saif->mclk_in_use = 1;
 	ret = mxs_saif_set_clk(saif, mclk, rate);
 	if (ret)
@@ -291,16 +277,10 @@ EXPORT_SYMBOL_GPL(mxs_saif_get_mclk);
  */
 static int mxs_saif_set_dai_fmt(struct snd_soc_dai *cpu_dai, unsigned int fmt)
 {
-	u32 scr, stat;
+	u32 scr;
 	u32 scr0;
 	struct mxs_saif *saif = snd_soc_dai_get_drvdata(cpu_dai);
 
-	stat = __raw_readl(saif->base + SAIF_STAT);
-	if (stat & BM_SAIF_STAT_BUSY) {
-		dev_err(cpu_dai->dev, "error: busy\n");
-		return -EBUSY;
-	}
-
 	scr0 = __raw_readl(saif->base + SAIF_CTRL);
 	scr0 = scr0 & ~BM_SAIF_CTRL_BITCLK_EDGE & ~BM_SAIF_CTRL_LRCLK_POLARITY \
 		& ~BM_SAIF_CTRL_JUSTIFY & ~BM_SAIF_CTRL_DELAY;
@@ -397,7 +377,7 @@ static int mxs_saif_hw_params(struct snd_pcm_substream *substream,
 {
 	struct mxs_saif *saif = snd_soc_dai_get_drvdata(cpu_dai);
 	struct mxs_saif *master_saif;
-	u32 scr, stat;
+	u32 scr;
 	int ret;
 
 	master_saif = mxs_saif_get_master(saif);
@@ -410,12 +390,6 @@ static int mxs_saif_hw_params(struct snd_pcm_substream *substream,
 		return -EINVAL;
 	}
 
-	stat = __raw_readl(saif->base + SAIF_STAT);
-	if (stat & BM_SAIF_STAT_BUSY) {
-		dev_err(cpu_dai->dev, "error: busy\n");
-		return -EBUSY;
-	}
-
 	/*
 	 * Set saif clk based on sample rate.
 	 * If mclk is used, we also set mclk, if not, saif->mclk is
-- 
1.7.9.5


[-- Attachment #3: Type: text/plain, Size: 0 bytes --]

Re: BUG: soft lockup when recording audio on MX28EVK with ASoC sgtl5000

[Fabio Estevam]

Hector,

On Mon, Mar 25, 2013 at 4:31 PM, Fabio Estevam <festevam@gmail.com> wrote:
> Hi Hector,
>
> On Mon, Mar 25, 2013 at 7:52 AM, Hector Palacios
> <hector.palacios@digi.com> wrote:
>> Hello,
>>
>> I just tried recording audio on Freescale's MX28EVK that uses ASoC sgtl5000
>> (kernel v3.8) with:
>>
>>         arecord -M -f cd sound.wav --duration 10
>>
>> and got a scheduler message:
>>
>> [  789.041847] [sched_delayed] sched: RT throttling activated
>
> I have just tested arecord on 3.8,4 and I haven't seen the 'soft lockup' bug.
>
> However, I see that mxs saif gets busy after the record and any
> attempt to do an 'aplay' after the first arecord fails.
>
> With the attached patch applied I am able to do arecord/aplay sequence
> several times.
>
> I don't have a cable handy here to actually test if the sound is
> recorded properly or not.
>
> Could you please test it against 3.8.4?

I managed to test arecord and it works with the following as follows:

1. Run alsamixer and select LINE_IN in the Capture element

2. Record with the following command line:

arecord -D hw:0,1 -d 10 -f S16_LE -r 44100 -c  test.wav

3. Then listen test.wav on your PC or in your board itself.

The sequence above works fine.

Anyway, I will submit the patch I sent earlier as it prevents the
system to get unresponsive.

Regards,

Fabio Estevam

Re: BUG: soft lockup when recording audio on MX28EVK with ASoC sgtl5000

[Hector Palacios]

Hi Fabio,

On 03/26/2013 12:15 AM, Fabio Estevam wrote:
> Hector,
>
> On Mon, Mar 25, 2013 at 4:31 PM, Fabio Estevam <festevam@gmail.com> wrote:
>> Hi Hector,
>>
>> On Mon, Mar 25, 2013 at 7:52 AM, Hector Palacios
>> <hector.palacios@digi.com> wrote:
>>> Hello,
>>>
>>> I just tried recording audio on Freescale's MX28EVK that uses ASoC sgtl5000
>>> (kernel v3.8) with:
>>>
>>>          arecord -M -f cd sound.wav --duration 10
>>>
>>> and got a scheduler message:
>>>
>>> [  789.041847] [sched_delayed] sched: RT throttling activated
>>
>> I have just tested arecord on 3.8,4 and I haven't seen the 'soft lockup' bug.
>>
>> However, I see that mxs saif gets busy after the record and any
>> attempt to do an 'aplay' after the first arecord fails.
>>
>> With the attached patch applied I am able to do arecord/aplay sequence
>> several times.
>>
>> I don't have a cable handy here to actually test if the sound is
>> recorded properly or not.
>>
>> Could you please test it against 3.8.4?
>
> I managed to test arecord and it works with the following as follows:
>
> 1. Run alsamixer and select LINE_IN in the Capture element
>
> 2. Record with the following command line:
>
> arecord -D hw:0,1 -d 10 -f S16_LE -r 44100 -c  test.wav

This command works on v3.8 even without your patch. I can arecord/aplay several times 
without problems.

With my original command 'arecord [-M] -f cd sound.wav --duration 10', the soft lockup 
appears both in v3.8 and v3.8.4 with and without your patch. So it looks like the 
patch does not help here.

Regards,
-- 
Héctor Palacios