Hi Chen-Yu,
On Mon, 28 Sep 2020 at 07:00, Chen-Yu Tsai wens@csie.org wrote:
On Mon, Sep 28, 2020 at 3:29 AM Clément Péron peron.clem@gmail.com wrote:
As slots and slot_width can be overwritter in case set_tdm() is called. Avoid to have this logic in set_chan_cfg().
It doesn't seem that set_tdm_slot() would get called concurrently with hw_params(), at least not for the simple-card family. If so then we'd have more problems like mismatched slots/slot-width due to no serialization when interacting with those values.
Sorry maybe the commit log is not clear. I was not talking about a concurrent effect but more if the slot_width is setted then we don't properly use it later.
Instead pass the required values as params to set_chan_cfg().
This will also avoid a bug when we will enable 20/24bits support, i2s->slot_width is not actually used in the lrck_period computation.
Suggested-by: Samuel Holland samuel@sholland.org Signed-off-by: Clément Péron peron.clem@gmail.com
sound/soc/sunxi/sun4i-i2s.c | 36 ++++++++++++++---------------------- 1 file changed, 14 insertions(+), 22 deletions(-)
diff --git a/sound/soc/sunxi/sun4i-i2s.c b/sound/soc/sunxi/sun4i-i2s.c index 2baf6c276280..0633b9fba3d7 100644 --- a/sound/soc/sunxi/sun4i-i2s.c +++ b/sound/soc/sunxi/sun4i-i2s.c @@ -177,8 +177,9 @@ struct sun4i_i2s_quirks { unsigned long (*get_bclk_parent_rate)(const struct sun4i_i2s *); s8 (*get_sr)(const struct sun4i_i2s *, int); s8 (*get_wss)(const struct sun4i_i2s *, int);
int (*set_chan_cfg)(const struct sun4i_i2s *,const struct snd_pcm_hw_params *);
int (*set_chan_cfg)(const struct sun4i_i2s *i2s,unsigned int channels, unsigned int slots,unsigned int slot_width); int (*set_fmt)(const struct sun4i_i2s *, unsigned int);};
@@ -414,10 +415,9 @@ static s8 sun8i_i2s_get_sr_wss(const struct sun4i_i2s *i2s, int width) }
static int sun4i_i2s_set_chan_cfg(const struct sun4i_i2s *i2s,
const struct snd_pcm_hw_params *params)
unsigned int channels, unsigned int slots,unsigned int slot_width){
unsigned int channels = params_channels(params);/* Map the channels for playback and capture */ regmap_write(i2s->regmap, SUN4I_I2S_TX_CHAN_MAP_REG, 0x76543210); regmap_write(i2s->regmap, SUN4I_I2S_RX_CHAN_MAP_REG, 0x00003210);@@ -434,15 +434,11 @@ static int sun4i_i2s_set_chan_cfg(const struct sun4i_i2s *i2s, }
static int sun8i_i2s_set_chan_cfg(const struct sun4i_i2s *i2s,
const struct snd_pcm_hw_params *params)
unsigned int channels, unsigned int slots,unsigned int slot_width){
unsigned int channels = params_channels(params);unsigned int slots = channels; unsigned int lrck_period;if (i2s->slots)slots = i2s->slots;Based on the bug you highlighted in your expanded commit log, the simpler fix would be to look at i2s->slot_width in addition to i2s->slots, and replace params_physical_width(params) accordingly.
That's what I did in the first version but was pointed out by Samuel that the same logic was done several times. As I can avoid it by passing the correct params it's simpler.
Also, I would put fixes (even preparatory ones) before patches that introduce support for new features and hardware. That makes them easier to backport if needed.
I will wait for Maxime to comment on this. Regarding the doc fixes tag should only be used to fix a previous commit. This will make the fixes commit a bit more complicated for stable kernel team.
Thanks for your review :)
Regards Clement
ChenYu
/* Map the channels for playback and capture */ regmap_write(i2s->regmap, SUN8I_I2S_TX_CHAN_MAP_REG, 0x76543210); regmap_write(i2s->regmap, SUN8I_I2S_RX_CHAN_MAP_REG, 0x76543210);@@ -467,11 +463,11 @@ static int sun8i_i2s_set_chan_cfg(const struct sun4i_i2s *i2s, case SND_SOC_DAIFMT_DSP_B: case SND_SOC_DAIFMT_LEFT_J: case SND_SOC_DAIFMT_RIGHT_J:
lrck_period = params_physical_width(params) * slots;
lrck_period = slot_width * slots; break; case SND_SOC_DAIFMT_I2S:
lrck_period = params_physical_width(params);
lrck_period = slot_width; break; default: