Re: [PATCH v4 1/4] ASoC: rockchip: add support for i2s-tdm controller
flat view
From: Mark Brown <broonie@kernel.org>
Date: 2021-09-15 14:05:46
Also in:
alsa-devel, linux-rockchip, lkml
On Sat, Sep 04, 2021 at 01:15:33AM +0200, Nicolas Frattaroli wrote: A few fairly small issues here, nothing too major:
quoted hunk ↗ jump to hunk
@@ -0,0 +1,1832 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * ALSA SoC Audio Layer - Rockchip I2S/TDM Controller driver
Please write the entire comment as a C++ one so it looks more itentional.
quoted hunk ↗ jump to hunk
+ xfer_mask = (tx ? I2S_XFER_TXS_START : 0) | + (rx ? I2S_XFER_RXS_START : 0); + xfer_val = (tx ? I2S_XFER_TXS_STOP : 0) | + (rx ? I2S_XFER_RXS_STOP : 0);
Please write normal conditional statements to improve legibility.
quoted hunk ↗ jump to hunk
+ spin_lock_irqsave(&i2s_tdm->lock, flags); + if (on) { + if (substream->stream == SNDRV_PCM_STREAM_PLAYBACK) + rockchip_enable_tde(i2s_tdm->regmap); + else + rockchip_enable_rde(i2s_tdm->regmap); + + if (atomic_inc_return(&i2s_tdm->refcount) == 1) {
Why do we need to use atomics here given that we're inside a spinlock? Surely the spinlock is already providing adequate concurrency protection. I can't see any other points where we don't have the spinlock already, and I'd be worried if we did. This looks like it could just be regular variables.
+ switch (fmt & SND_SOC_DAIFMT_MASTER_MASK) {
+ case SND_SOC_DAIFMT_CBS_CFS:Please use the modern SOF_DAI_FMT_CBC_CFC defines.
+ delta = (ppm < 0) ? -1 : 1;
Again, please write normal condiditional statements for legibility.
quoted hunk ↗ jump to hunk
+static int rockchip_i2s_tdm_clk_compensation_put(struct snd_kcontrol *kcontrol, + struct snd_ctl_elem_value *ucontrol) +{
+ return ret; +}
This should return 1 if the value changed.
quoted hunk ↗ jump to hunk
+static int __maybe_unused rockchip_i2s_tdm_resume(struct device *dev) +{ + struct rk_i2s_tdm_dev *i2s_tdm = dev_get_drvdata(dev); + int ret; + + ret = pm_runtime_get_sync(dev); + if (ret < 0) + return ret; + ret = regcache_sync(i2s_tdm->regmap); + pm_runtime_put(dev);
Runtime resume also does a regcache sync so why are we doing another one here?