[RFC][PATCH] ASoC: Suppress restore of default register values for lzo cache sync

[RFC][PATCH] ASoC: Suppress restore of default register values for lzo cache sync

[Axel Lin]

Currently we do suppress restore of default register values for rbtree
and flat cache sync.
A couple users use this trick to avoid writing reset register while resume.
Thus make lzo cache sync suppress restore of default register values,
so we have consistent behavior for all types of cache sync.

Signed-off-by: Axel Lin <axel.lin@gmail.com>
---
 sound/soc/soc-cache.c |    8 +++++++-
 1 files changed, 7 insertions(+), 1 deletions(-)

diff --git a/sound/soc/soc-cache.c b/sound/soc/soc-cache.c
index 9077aa4..eb3c016 100644
--- a/sound/soc/soc-cache.c
+++ b/sound/soc/soc-cache.c
@@ -555,16 +555,22 @@ static inline int snd_soc_lzo_get_blksize(struct snd_soc_codec *codec)
 static int snd_soc_lzo_cache_sync(struct snd_soc_codec *codec)
 {
 	struct snd_soc_lzo_ctx **lzo_blocks;
-	unsigned int val;
+	unsigned int val, def;
+	unsigned int word_size;
 	int i;
 	int ret;
 
 	lzo_blocks = codec->reg_cache;
+	word_size = codec->driver->reg_word_size;
 	for_each_set_bit(i, lzo_blocks[0]->sync_bmp, lzo_blocks[0]->sync_bmp_nbits) {
 		WARN_ON(!snd_soc_codec_writable_register(codec, i));
 		ret = snd_soc_cache_read(codec, i, &val);
 		if (ret)
 			return ret;
+		def = snd_soc_get_cache_val(codec->reg_def_copy, i, word_size);
+		if (val == def)
+			continue;
+
 		codec->cache_bypass = 1;
 		ret = snd_soc_write(codec, i, val);
 		codec->cache_bypass = 0;
-- 
1.7.4.1

Re: [RFC][PATCH] ASoC: Suppress restore of default register values for lzo cache sync

[Mark Brown]

On Thu, Oct 13, 2011 at 05:28:33PM +0800, Axel Lin wrote:
> Currently we do suppress restore of default register values for rbtree
> and flat cache sync.
> A couple users use this trick to avoid writing reset register while resume.
> Thus make lzo cache sync suppress restore of default register values,
> so we have consistent behavior for all types of cache sync.

Given that we never acquired any in tree users for the LZO cache it'd
probably be as well to just work on the regmap version and deprecate
this one - there's no users to move over and it'll save effort.