| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <20200416171624.GA174353@gerhold.net>
Date: Thu, 16 Apr 2020 19:16:24 +0200
From: Stephan Gerhold <stephan@...hold.net>
To: Srinivas Kandagatla <srinivas.kandagatla@...aro.org>
Cc: broonie@...nel.org, alsa-devel@...a-project.org,
Takahide Higuchi <takahidehiguchi@...il.com>,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH] ASoC: qcom: lpass-cpu: support full duplex operation
On Fri, Mar 06, 2020 at 01:01:47PM +0000, Srinivas Kandagatla wrote:
> From: Takahide Higuchi <takahidehiguchi@...il.com>
>
> This patch fixes a bug where playback on bidirectional I2S interface stops
> when we start recording on the same interface.
>
> We use regmap_update_bits instead of regmap_write so that we will not clear
> SPKEN and SPKMODE bits when we start/stop recording.
>
Thanks for the patch! I don't have suitable hardware to test this
feature, but I checked if playback-only is still working correctly:
With this patch I can only play audio once. Further playback
never starts, it gets stuck immediately. Suggested fix below.
(Also a minor code style suggestion...)
> Signed-off-by: Takahide Higuchi <takahidehiguchi@...il.com>
> Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@...aro.org>
> ---
> sound/soc/qcom/lpass-apq8016.c | 2 ++
> sound/soc/qcom/lpass-cpu.c | 24 ++++++++++++++++++------
> sound/soc/qcom/lpass-lpaif-reg.h | 2 +-
> 3 files changed, 21 insertions(+), 7 deletions(-)
>
> diff --git a/sound/soc/qcom/lpass-apq8016.c b/sound/soc/qcom/lpass-apq8016.c
> index 6575da549237..85079c697faa 100644
> --- a/sound/soc/qcom/lpass-apq8016.c
> +++ b/sound/soc/qcom/lpass-apq8016.c
> @@ -121,6 +121,8 @@ static struct snd_soc_dai_driver apq8016_lpass_cpu_dai_driver[] = {
> },
> .probe = &asoc_qcom_lpass_cpu_dai_probe,
> .ops = &asoc_qcom_lpass_cpu_dai_ops,
> + .symmetric_samplebits = 1,
> + .symmetric_rates = 1,
> },
> };
>
> diff --git a/sound/soc/qcom/lpass-cpu.c b/sound/soc/qcom/lpass-cpu.c
> index dbce7e92baf3..dc8acb380b6f 100644
> --- a/sound/soc/qcom/lpass-cpu.c
> +++ b/sound/soc/qcom/lpass-cpu.c
> @@ -72,6 +72,7 @@ static int lpass_cpu_daiops_hw_params(struct snd_pcm_substream *substream,
> snd_pcm_format_t format = params_format(params);
> unsigned int channels = params_channels(params);
> unsigned int rate = params_rate(params);
> + unsigned int mask;
> unsigned int regval;
> int bitwidth, ret;
>
> @@ -81,6 +82,9 @@ static int lpass_cpu_daiops_hw_params(struct snd_pcm_substream *substream,
> return bitwidth;
> }
>
> + mask = LPAIF_I2SCTL_LOOPBACK_MASK |
> + LPAIF_I2SCTL_WSSRC_MASK |
> + LPAIF_I2SCTL_BITWIDTH_MASK;
> regval = LPAIF_I2SCTL_LOOPBACK_DISABLE |
> LPAIF_I2SCTL_WSSRC_INTERNAL;
>
I understand the space after "mask" here (to align the =) ...
> @@ -100,6 +104,7 @@ static int lpass_cpu_daiops_hw_params(struct snd_pcm_substream *substream,
> }
>
> if (substream->stream == SNDRV_PCM_STREAM_PLAYBACK) {
> + mask |= LPAIF_I2SCTL_SPKMODE_MASK | LPAIF_I2SCTL_SPKMONO_MASK;
> switch (channels) {
> case 1:
> regval |= LPAIF_I2SCTL_SPKMODE_SD0;
> @@ -127,6 +132,7 @@ static int lpass_cpu_daiops_hw_params(struct snd_pcm_substream *substream,
> return -EINVAL;
> }
> } else {
> + mask |= LPAIF_I2SCTL_MICMODE_MASK | LPAIF_I2SCTL_MICMONO_MASK;
> switch (channels) {
> case 1:
> regval |= LPAIF_I2SCTL_MICMODE_SD0;
... but they don't make much sense to me in these two cases.
It looks a bit weird in my opinion :)
> @@ -155,9 +161,9 @@ static int lpass_cpu_daiops_hw_params(struct snd_pcm_substream *substream,
> }
> }
>
> - ret = regmap_write(drvdata->lpaif_map,
> - LPAIF_I2SCTL_REG(drvdata->variant, dai->driver->id),
> - regval);
> + ret = regmap_update_bits(drvdata->lpaif_map,
> + LPAIF_I2SCTL_REG(drvdata->variant, dai->driver->id),
> + mask, regval);
> if (ret) {
> dev_err(dai->dev, "error writing to i2sctl reg: %d\n", ret);
> return ret;
> @@ -178,11 +184,17 @@ static int lpass_cpu_daiops_hw_free(struct snd_pcm_substream *substream,
> struct snd_soc_dai *dai)
> {
> struct lpass_data *drvdata = snd_soc_dai_get_drvdata(dai);
> + unsigned int mask;
> int ret;
>
> - ret = regmap_write(drvdata->lpaif_map,
> - LPAIF_I2SCTL_REG(drvdata->variant, dai->driver->id),
> - 0);
> + if (substream->stream == SNDRV_PCM_STREAM_PLAYBACK)
> + mask = LPAIF_I2SCTL_SPKMODE_MASK;
> + else
> + mask = LPAIF_I2SCTL_MICMODE_MASK;
Same here with the spaces after "mask", but more importantly:
This changes the behavior slightly. Previously we reset the entire
register, now only the SPKMODE/MICMODE. This means that SPKEN/MICEN
does not get reset to 0 when this function is called.
Everything seems to work fine again for me if I reset SPKEN/MICEN too:
diff --git a/sound/soc/qcom/lpass-cpu.c b/sound/soc/qcom/lpass-cpu.c
index dc8acb380b6f..49a36c1f5640 100644
--- a/sound/soc/qcom/lpass-cpu.c
+++ b/sound/soc/qcom/lpass-cpu.c
@@ -188,9 +188,9 @@ static int lpass_cpu_daiops_hw_free(struct snd_pcm_substream *substream,
int ret;
if (substream->stream == SNDRV_PCM_STREAM_PLAYBACK)
- mask = LPAIF_I2SCTL_SPKMODE_MASK;
+ mask = LPAIF_I2SCTL_SPKEN_MASK | LPAIF_I2SCTL_SPKMODE_MASK;
else
- mask = LPAIF_I2SCTL_MICMODE_MASK;
+ mask = LPAIF_I2SCTL_MICEN_MASK | LPAIF_I2SCTL_MICMODE_MASK;
ret = regmap_update_bits(drvdata->lpaif_map,
LPAIF_I2SCTL_REG(drvdata->variant, dai->driver->id),
After looking at the code a bit more I wonder why SPKEN/MICEN is set in
prepare() at all? It is already set/un-set in trigger() on
SNDRV_PCM_TRIGGER_START and SNDRV_PCM_TRIGGER_STOP.
In fact, my problem only happens because prepare() is called again
after the stream was stopped (and therefore sets SPKEN again):
[ 32.809915] lpass_cpu_daiops_hw_params
[ 32.810598] lpass_cpu_daiops_prepare
[ 32.823092] lpass_cpu_daiops_prepare
[ 32.837901] lpass_cpu_daiops_trigger: SNDRV_PCM_TRIGGER_START
[ 37.565261] lpass_cpu_daiops_trigger: SNDRV_PCM_TRIGGER_STOP
[ 37.565402] lpass_cpu_daiops_prepare
[ 37.565854] lpass_cpu_daiops_hw_free
In other words, a much better fix in my opinion would be to
simply remove(!) the entire lpass_cpu_daiops_prepare() method.
I tested it and that also fixes the problem, without my diff above to
unset SPKEN/MICEN in lpass_cpu_daiops_hw_free().
Thanks,
Stephan
> +
> + ret = regmap_update_bits(drvdata->lpaif_map,
> + LPAIF_I2SCTL_REG(drvdata->variant, dai->driver->id),
> + mask, 0);
> if (ret)
> dev_err(dai->dev, "error writing to i2sctl reg: %d\n", ret);
>
> diff --git a/sound/soc/qcom/lpass-lpaif-reg.h b/sound/soc/qcom/lpass-lpaif-reg.h
> index 3d74ae123e9d..7a2b9cf99976 100644
> --- a/sound/soc/qcom/lpass-lpaif-reg.h
> +++ b/sound/soc/qcom/lpass-lpaif-reg.h
> @@ -56,7 +56,7 @@
> #define LPAIF_I2SCTL_MICMODE_6CH (7 << LPAIF_I2SCTL_MICMODE_SHIFT)
> #define LPAIF_I2SCTL_MICMODE_8CH (8 << LPAIF_I2SCTL_MICMODE_SHIFT)
>
> -#define LPAIF_I2SCTL_MIMONO_MASK GENMASK(3, 3)
> +#define LPAIF_I2SCTL_MICMONO_MASK GENMASK(3, 3)
> #define LPAIF_I2SCTL_MICMONO_SHIFT 3
> #define LPAIF_I2SCTL_MICMONO_STEREO (0 << LPAIF_I2SCTL_MICMONO_SHIFT)
> #define LPAIF_I2SCTL_MICMONO_MONO (1 << LPAIF_I2SCTL_MICMONO_SHIFT)
> --
> 2.21.0
>
Powered by blists - more mailing lists