[alsa-devel] [PATCH] ASoC: soc-io: Only force use regmap if the codec have it implemented
Some codecs (like twl4030, twl6040) can not have regmap implementation since they are accessing to the chip via MFD core drivers.
Force the use of regmap only in cases when the regmap is actually available for the codec.
Signed-off-by: Peter Ujfalusi peter.ujfalusi@ti.com ---
Hi Mark,
the following commit: 38cbf95 ASoC: core: Try to use regmap if the driver doesn't set up any I/O
broke audio for boards using twl4030 or twl6040 as codec. These devices can not have regmap implementation in the codec driver since they rely on their MFD core to access to the chip.
As it is now big number of OMAP based devices have broken audio. Would it be possible to include this fix for 3.6-rc1 already?
Regards, Peter
sound/soc/soc-io.c | 12 ++++++++---- 1 files changed, 8 insertions(+), 4 deletions(-)
diff --git a/sound/soc/soc-io.c b/sound/soc/soc-io.c index 29183ef..26948be 100644 --- a/sound/soc/soc-io.c +++ b/sound/soc/soc-io.c @@ -117,9 +117,6 @@ int snd_soc_codec_set_cache_io(struct snd_soc_codec *codec, int ret;
memset(&config, 0, sizeof(config)); - codec->write = hw_write; - codec->read = hw_read; - codec->bulk_write_raw = snd_soc_hw_bulk_write_raw;
config.reg_bits = addr_bits; config.val_bits = data_bits; @@ -141,11 +138,11 @@ int snd_soc_codec_set_cache_io(struct snd_soc_codec *codec,
case SND_SOC_REGMAP: /* Device has made its own regmap arrangements */ - codec->using_regmap = true; if (!codec->control_data) codec->control_data = dev_get_regmap(codec->dev, NULL);
if (codec->control_data) { + codec->using_regmap = true; ret = regmap_get_val_bytes(codec->control_data); /* Errors are legitimate for non-integer byte * multiples */ @@ -161,6 +158,13 @@ int snd_soc_codec_set_cache_io(struct snd_soc_codec *codec, if (IS_ERR(codec->control_data)) return PTR_ERR(codec->control_data);
+ /* Only override the read/write callbacks if regmap is available */ + if (codec->control_data) { + codec->write = hw_write; + codec->read = hw_read; + codec->bulk_write_raw = snd_soc_hw_bulk_write_raw; + } + return 0; } EXPORT_SYMBOL_GPL(snd_soc_codec_set_cache_io);
On Wed, Aug 01, 2012 at 03:01:45PM +0300, Peter Ujfalusi wrote:
case SND_SOC_REGMAP: /* Device has made its own regmap arrangements */
codec->using_regmap = true;
No, this is crazy.
On 08/01/2012 03:16 PM, Mark Brown wrote:
On Wed, Aug 01, 2012 at 03:01:45PM +0300, Peter Ujfalusi wrote:
case SND_SOC_REGMAP: /* Device has made its own regmap arrangements */
codec->using_regmap = true;No, this is crazy.
Why? We set it when we have confirmed that the codec actually have regmap of it's own.
On Wed, Aug 01, 2012 at 03:20:29PM +0300, Peter Ujfalusi wrote:
On 08/01/2012 03:16 PM, Mark Brown wrote:
On Wed, Aug 01, 2012 at 03:01:45PM +0300, Peter Ujfalusi wrote:
case SND_SOC_REGMAP: /* Device has made its own regmap arrangements */
codec->using_regmap = true;
No, this is crazy.
Why? We set it when we have confirmed that the codec actually have regmap of it's own.
You're changing the function setting up register I/O to ignore what's being set. This is silly - the soc-io code is doing exactly the right thing here. Abstraction, folks.
On 08/01/2012 03:40 PM, Mark Brown wrote:
You're changing the function setting up register I/O to ignore what's being set. This is silly - the soc-io code is doing exactly the right thing here.
Well, it overrides the codec->read/write/bulk_write_raw callbacks, sets the codec->using_regmap to true regardless if the codec have regmap or not. It should only do all of these if the regmap is there IMHO.
The soc-core calls snd_soc_codec_set_cache_io() if codec->control_data is NULL at the moment.
Reverting commit 38cbf95 ASoC: core: Try to use regmap if the driver doesn't set up any I/O would fix the issue as well without changing the soc-io file.
We can as well return if the codec does not have the regmap and snd_soc_codec_set_cache_io() fails in the SND_SOC_REGMAP case.
On Wed, Aug 01, 2012 at 04:09:30PM +0300, Peter Ujfalusi wrote:
On 08/01/2012 03:40 PM, Mark Brown wrote:
You're changing the function setting up register I/O to ignore what's being set. This is silly - the soc-io code is doing exactly the right thing here.
Well, it overrides the codec->read/write/bulk_write_raw callbacks, sets the codec->using_regmap to true regardless if the codec have regmap or not. It should only do all of these if the regmap is there IMHO.
That's exactly what it's being asked to do!
The soc-core calls snd_soc_codec_set_cache_io() if codec->control_data is NULL at the moment.
Right, so perhaps that's the decision which needs revisiting?
participants (2)
-
Mark Brown -
Peter Ujfalusi