Il 08/12/22 04:31, Trevor Wu ha scritto:
Add mt8188 etdm dai driver support. Signed-off-by: Trevor Wu <trevor.wu@xxxxxxxxxxxx> --- I don't add Reviewed-by tag because one new header file is included in the patch to resolve compiling issue found by kernel test robot. Additionally, I re-layout the code for better understanding of the follow-up patch. --- sound/soc/mediatek/mt8188/mt8188-dai-etdm.c | 2591 +++++++++++++++++++ 1 file changed, 2591 insertions(+) create mode 100644 sound/soc/mediatek/mt8188/mt8188-dai-etdm.c diff --git a/sound/soc/mediatek/mt8188/mt8188-dai-etdm.c b/sound/soc/mediatek/mt8188/mt8188-dai-etdm.c new file mode 100644 index 000000000000..c653fa5e3f85 --- /dev/null +++ b/sound/soc/mediatek/mt8188/mt8188-dai-etdm.c
..snip..
+ +static void mt8188_dai_etdm_parse_of(struct mtk_base_afe *afe) +{ + const struct device_node *of_node = afe->dev->of_node; + struct mt8188_afe_private *afe_priv = afe->platform_priv; + struct mtk_dai_etdm_priv *etdm_data; + int i, j; + char prop[48]; + u8 disable_chn[MT8188_ETDM_MAX_CHANNELS]; + int max_chn = MT8188_ETDM_MAX_CHANNELS; + u32 sel; + int ret; + int dai_id; + unsigned int sync_id; + struct { + const char *name; + const unsigned int sync_id; + } of_afe_etdms[MT8188_AFE_IO_ETDM_NUM] = { + {"etdm-in1", ETDM_SYNC_FROM_IN1}, + {"etdm-in2", ETDM_SYNC_FROM_IN2}, + {"etdm-out1", ETDM_SYNC_FROM_OUT1}, + {"etdm-out2", ETDM_SYNC_FROM_OUT2}, + {"etdm-out3", ETDM_SYNC_FROM_OUT3}, + }; + + for (i = 0; i < MT8188_AFE_IO_ETDM_NUM; i++) { + dai_id = ETDM_TO_DAI_ID(i); + etdm_data = afe_priv->dai_priv[dai_id]; + + ret = snprintf(prop, sizeof(prop), + "mediatek,%s-mclk-always-on-rate", + of_afe_etdms[i].name); + if (ret < 0) { + dev_info(afe->dev, "%s snprintf err=%d\n",
Is this property optional? If yes, this dev_info() must be a dev_dbg(), otherwise, this must be a dev_err(). Please fix all prints to use the right message level.
+ __func__, ret); + return;
Also, is it possible to specify this property only on selected eTDMs? As it is right now, if anyone wants to specify this only on, for example, etdm-out1 and out2, that won't work. In that case, you should replace that return with a `continue`. P.S.: I'm raising this question because you're not forcing "all or nothing" in commit [10/12] where you introduce the bindings for this driver, so I suppose that returning (hence stopping to parse) is a mistake.
+ } + ret = of_property_read_u32(of_node, prop, &sel); + if (ret == 0) { + etdm_data->mclk_dir = SND_SOC_CLOCK_OUT; + if (mtk_dai_etdm_cal_mclk(afe, sel, dai_id)) + dev_info(afe->dev, "%s unsupported mclk %uHz\n", + __func__, sel); + } +
Regards, Angelo