Hi Mark,
On Thu, Jun 02, 2016 at 11:26:51AM +0100, Mark Brown wrote:
> > +static int sun4i_i2s_params_to_sr(struct snd_pcm_hw_params *params)
> > +{
> > + switch (params_width(params)) {
> > + case 16:
> > + return 0;
> > + }
> > +
> > + return -EINVAL;
> > +}
>
> The switch statement here and in the _wss() function look weird because
> they don't have default cases. Since there's only one user of both
> functions it seems better to have the switch statements inline anyway.
I don't know, maybe it's just me, but I really find it cleaner that
way, and the compiler will probably inline it anyway. If you insist,
I'll change it though.
> > + for (i = 0; sun4i_i2s_mclk_div[i].div; i++) {
> > + const struct sun4i_i2s_clk_div *mdiv = sun4i_i2s_mclk_div + i;
>
> Why not just write these as normal array lookups?
By normal, you mean using ARRAY_SIZE()?
> > + /* Enable the first output line */
> > + regmap_update_bits(i2s->regmap, SUN4I_I2S_CTRL_REG,
> > + SUN4I_I2S_CTRL_SDO_EN_MASK,
> > + SUN4I_I2S_CTRL_SDO_EN(0));
> > +
> > + /* Enable the first two channels */
> > + regmap_write(i2s->regmap, SUN4I_I2S_TX_CHAN_SEL_REG,
> > + SUN4I_I2S_TX_CHAN_SEL(2));
> > +
> > + /* Map them to the two first samples coming in */
> > + regmap_write(i2s->regmap, SUN4I_I2S_TX_CHAN_MAP_REG,
> > + SUN4I_I2S_TX_CHAN_MAP(0, 0) | SUN4I_I2S_TX_CHAN_MAP(1, 1));
>
> We don't undo these if setup fails... do them once on probe?
Yep
> > +static int sun4i_i2s_set_fmt(struct snd_soc_dai *dai, unsigned int fmt)
> > +{
> > + struct sun4i_i2s *i2s = snd_soc_dai_get_drvdata(dai);
> > + u32 val;
>
> More tab/space damage, there seems to be quite a bit in the rest of the
> driver.
Indeed, sorry for that.
> > +static int sun4i_i2s_dai_probe(struct snd_soc_dai *dai)
> > +{
> > + struct sun4i_i2s *i2s = snd_soc_dai_get_drvdata(dai);
> > +
> > + /* Enable the whole hardware block */
> > + regmap_write(i2s->regmap, SUN4I_I2S_CTRL_REG,
> > + SUN4I_I2S_CTRL_GL_EN);
>
> Runtime PM? It also seems like this is something that ought to be
> covered in the suspend and resume callbacks.
runtime_pm is supported, and uses the regmap cache to keep those
changes.
suspend and resume is not supported for that SoC yet, so I couldn't
test it properly (and isn't that redundant with runtime_pm?)
Thanks!
Maxime
--
Maxime Ripard, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com
Attachment:
signature.asc
Description: PGP signature
