Re: [PATCH v6 07/10] hda: cs35l41: Add support for CS35L41 in HDA systems

[Lucas tanure]

On 1/6/22 12:29, Andy Shevchenko wrote:
> On Fri, Dec 17, 2021 at 5:45 PM Lucas Tanure
> <tanureal@xxxxxxxxxxxxxxxxxxxxx> wrote:
> > Add support for CS35L41 using a new separated driver
> > that can be used in all upcoming designs
>
>
>
> > +config SND_HDA_SCODEC_CS35L41_I2C
> > +       tristate "Build CS35L41 HD-audio side codec support for I2C Bus"
> > +       depends on ACPI
> > +       select SND_HDA_GENERIC
> > +       select SND_SOC_CS35L41_LIB
> > +       select SND_HDA_SCODEC_CS35L41
> > +       help
> > +         Say Y or M here to include CS35L41 I2C HD-audio side codec support
> > +         in snd-hda-intel driver, such as ALC287.
> > +
> > +comment "Set to Y if you want auto-loading the side codec driver"
> > +       depends on SND_HDA=y && SND_HDA_SCODEC_CS35L41_I2C=m
> > +
> > +config SND_HDA_SCODEC_CS35L41_SPI
> > +       tristate "Build CS35L41 HD-audio codec support for SPI Bus"
> > +       depends on ACPI
> > +       select SND_HDA_GENERIC
> > +       select SND_SOC_CS35L41_LIB
> > +       select SND_HDA_SCODEC_CS35L41
> > +       help
> > +         Say Y or M here to include CS35L41 SPI HD-audio side codec support
> > +         in snd-hda-intel driver, such as ALC287.
>
> ...
>
> > +// cs35l41.c -- CS35l41 ALSA HDA audio driver
>
> It's an additional burden in case the file will be renamed. i..o.w.
> drop the names of the files from the files.
>
>
> > +#include <linux/acpi.h>
> > +#include <linux/module.h>
> > +#include <sound/hda_codec.h>
> > +#include "hda_local.h"
> > +#include "hda_auto_parser.h"
> > +#include "hda_jack.h"
> > +#include "hda_generic.h"
> > +#include "hda_component.h"
> > +#include "cs35l41_hda.h"
>
> ...
>
> > +       { CS35L41_PWR_CTRL1,            0x00000001, 2000 }, //GLOBAL_EN = 1
>
> Here and everywhere else, missed space after //
>
> ...
>
> > +static void cs35l41_hda_playback_hook(struct device *dev, int action)
> > +{
> > +       struct cs35l41_hda *cs35l41 = dev_get_drvdata(dev);
> > +       const struct cs35l41_hda_reg_sequence *reg_seq = cs35l41->reg_seq;
> > +       struct regmap *reg = cs35l41->regmap;
> > +       int ret = 0;
> > +
> > +       switch (action) {
> > +       case HDA_GEN_PCM_ACT_OPEN:
> > +               if (reg_seq->open)
> > +                       ret = regmap_multi_reg_write(reg, reg_seq->open, reg_seq->num_open);
> > +               break;
> > +       case HDA_GEN_PCM_ACT_PREPARE:
> > +               if (reg_seq->prepare)
> > +                       ret = regmap_multi_reg_write(reg, reg_seq->prepare, reg_seq->num_prepare);
> > +               break;
> > +       case HDA_GEN_PCM_ACT_CLEANUP:
> > +               if (reg_seq->cleanup)
> > +                       ret = regmap_multi_reg_write(reg, reg_seq->cleanup, reg_seq->num_cleanup);
> > +               break;
> > +       case HDA_GEN_PCM_ACT_CLOSE:
> > +               if (reg_seq->close)
> > +                       ret = regmap_multi_reg_write(reg, reg_seq->close, reg_seq->num_close);
> > +               break;
>
> default case?
>
> > +       }
> > +
> > +       if (ret)
> > +               dev_warn(cs35l41->dev, "Failed to apply multi reg write: %d\n", ret);
>
> > +
>
> Redundant blank line.
>
> > +}
>
>
> ...
>
> > +       if (comps && cs35l41->index >= 0 && cs35l41->index < HDA_MAX_COMPONENTS)
> > +               comps = &comps[cs35l41->index];
> > +       else
> > +               return -EINVAL;
>
> Can you check first? In such a case you won't need the 'else' branch at all.
>
> ...
>
> > +       if (!comps->dev) {
>
> Why not a positive check and standard pattern as per above?
>
>
> > +               comps->dev = dev;
> > +               strscpy(comps->name, dev_name(dev), sizeof(comps->name));
> > +               comps->playback_hook = cs35l41_hda_playback_hook;
> > +               comps->set_channel_map = cs35l41_hda_channel_map;
> > +               return 0;
> > +       }
> > +
> > +       return -EBUSY;
> > +}
>
> ...
>
> > +       switch (hw_cfg->gpio1_func) {
> > +       case CS35l41_VSPK_SWITCH:
> > +               regmap_update_bits(cs35l41->regmap, CS35L41_GPIO_PAD_CONTROL,
> > +                                  CS35L41_GPIO1_CTRL_MASK, 1 << CS35L41_GPIO1_CTRL_SHIFT);
> > +               break;
> > +       case CS35l41_SYNC:
> > +               regmap_update_bits(cs35l41->regmap, CS35L41_GPIO_PAD_CONTROL,
> > +                                  CS35L41_GPIO1_CTRL_MASK, 2 << CS35L41_GPIO1_CTRL_SHIFT);
> > +               break;
>
> default case?
>
> Same for all switch-cases in your code.
>
> > +       }
>
> ...
>
> > +       ret = cs35l41_hda_channel_map(cs35l41->dev, 0, NULL, 1, (unsigned int *)&hw_cfg->spk_pos);
> > +       if (ret)
> > +               return ret;
> > +
> > +       return 0;
>
> return cs35l41_hda_...(...);
>
> ...
>
> > +       property = "cirrus,dev-index";
> > +       ret = device_property_count_u32(acpi_dev, property);
>
> Please, name adev the pointer to ACPI device. Ah, what a mess, you
> have named acpi_dev the pointer to the struct device. Please, find a
> better name, like dev, or physdev or so.
>
> > +       if (ret <= 0)
>
> Shouldn't you override the error code for the 0 case?
>
> > +               goto no_acpi_dsd;
>
> ...
>
> > +       if (ret > ARRAY_SIZE(values)) {
> > +               ret = -EINVAL;
> > +               goto err;
> > +       }
>
> Is it really the issue? I would expect the issue when you have less
> than expected, and not otherwise.
>
> ...
>
> > +       /* No devm_ version as CLSA0100, in no_acpi_dsd case, can't use devm version */
>
> Can you elaborate why devm can't be used?
>
> > +       cs35l41->reset_gpio = fwnode_gpiod_get_index(&adev->fwnode, "reset", cs35l41->index,
>
> Please, do not dereference fwnode pointers.
> Also, why can't you use the device instead of fwnode?
>
> > +                                                    GPIOD_OUT_LOW, "cs35l41-reset");
>
> ...
>
> > +       hw_cfg = kzalloc(sizeof(*hw_cfg), GFP_KERNEL);
>
> Why not devm?
>
> > +       if (!hw_cfg) {
> > +               ret = -ENOMEM;
> > +               goto err;
> > +       }
>
> ...
>
> > +       property = "cirrus,speaker-position";
> > +       ret = device_property_read_u32_array(acpi_dev, property, values, nval);
> > +       if (ret)
> > +               goto err_free;
> > +       hw_cfg->spk_pos = values[cs35l41->index];
>
> This and further is weird. Why do you need to retrieve all values for
> just one? Use indexed APIs for that. If there are none, create them.
>
> ...
>
> > +no_acpi_dsd:
> > +       /*
> > +        * Device CLSA0100 doesn't have _DSD so a gpiod_get by the label reset won't work.
>
> So, you need to add mapping tables and switch to regular APIs, tell
> me, why it won't work.
I will submit a patch series to fix most of the issues you pointed out. 
The part about how the driver access the ACPI table is going to be 
improved later if possible.
The laptop has already shipped and doesn't have a _DSD node, so the 
driver needs to read the reset GPIO from a hard coded index inside a 
node that contains more than one cs35l41.

> > +        * And devices created by i2c-multi-instantiate don't have their device struct pointing to
> > +        * the correct fwnode, so acpi_dev must be used here
> > +        * And devm functions expect that the device requesting the resource has the correct
> > +        * fwnode
>
> You missed grammar periods and what else? Please, update your comments
> to use proper English grammar.
>
> > +        */
> > +       if (strncmp(hid, "CLSA0100", 8) != 0)
> > +               return ERR_PTR(-EINVAL);
> > +
> > +       /* check I2C address to assign the index */
> > +       cs35l41->index = id == 0x40 ? 0 : 1;
> > +       cs35l41->reset_gpio = gpiod_get_index(acpi_dev, NULL, 0, GPIOD_OUT_HIGH);
> > +       cs35l41->vspk_always_on = true;
> > +       put_device(acpi_dev);
> > +
> > +       return NULL;
> > +}
>
> ...
>
> > +int cs35l41_hda_probe(struct device *dev, const char *device_name, int id, int irq,
> > +                     struct regmap *regmap)
>
> > +       if (IS_ERR(regmap))
> > +               return PTR_ERR(regmap);
>
> Why?
It is up to the I2C/SPI module to create the regmap and provide to the 
main module. If that fails the main module can't continue.

> ...
>
> > +       if (IS_ERR(cs35l41->reset_gpio)) {
>
> You should use _optinal variants instead,
>
> > +               ret = PTR_ERR(cs35l41->reset_gpio);
> > +               cs35l41->reset_gpio = NULL;
> > +               if (ret == -EBUSY) {
> > +                       dev_info(cs35l41->dev, "Reset line busy, assuming shared reset\n");
> > +               } else {
> > +                       if (ret != -EPROBE_DEFER)
> > +                               dev_err(cs35l41->dev, "Failed to get reset GPIO: %d\n", ret);
> > +                       goto err;
>
> We have dev_err_probe() for a few releases already.
>
> > +               }
> > +       }
>
> ...
>
> > +       ret = regmap_read(cs35l41->regmap, CS35L41_IRQ1_STATUS3, &int_sts);
> > +       if (ret || (int_sts & CS35L41_OTP_BOOT_ERR)) {
> > +               dev_err(cs35l41->dev, "OTP Boot error\n");
> > +               ret = -EIO;
>
> Why shadowing error code?
> Why not use dev_err_probe()?
>
> > +               goto err;
> > +       }
>
> ...
>
> > +EXPORT_SYMBOL_GPL(cs35l41_hda_probe);
>
> Please, use the namespace variant and avoid polluting the global
> namespace with  your symbols.
>
> ...
>
> > + * cs35l41_hda.h -- CS35L41 ALSA HDA audio driver
>
> No file names in the files.
>
> ...
>
> > +#include <linux/regulator/consumer.h>
> > +#include <linux/gpio/consumer.h>
> > +#include <linux/device.h>
> > +#include <sound/cs35l41.h>
>
> Please revisit this. You need to add here only the headers that you
> are a direct user of (or in some cases their top level ones, like
> types.h for compiler_attributes.h).
>
> ...
>
> > +#ifdef CONFIG_ACPI
>
> Drop this ugliness.
>
> > +static const struct acpi_device_id cs35l41_acpi_hda_match[] = {
> > +       {"CLSA0100", 0 },
> > +       {"CSC3551", 0 },
>
> I believe these IDs are officially allocated by the Cirrus Logic, right?
CLSA010* is not a valid id for Cirrus Logic, but the Bios is already in 
production, so we must support it.
CSC3551 is a valid id for Cirrus Logic.

> > +       { },
>
> No comma for terminator line here and everywhere else where it's the case.
>
> > +};
> > +MODULE_DEVICE_TABLE(acpi, cs35l41_acpi_hda_match);
> > +#endif
> > +
> > +static struct i2c_driver cs35l41_i2c_driver = {
> > +       .driver = {
> > +               .name           = "cs35l41-hda",
> > +               .acpi_match_table = ACPI_PTR(cs35l41_acpi_hda_match),
>
> ACPI_PTR() as well.
>
> > +       },
> > +       .id_table       = cs35l41_hda_i2c_id,
> > +       .probe          = cs35l41_hda_i2c_probe,
> > +       .remove         = cs35l41_hda_i2c_remove,
> > +};
>
> > +
>
> No need to have a blank line here.
>
> > +module_i2c_driver(cs35l41_i2c_driver);
>
> I stopped here, so this code needs more work and can't be applied like this.
>
> I believe that current Cirrus Logic drivers are written in the same
> (semi-) bad style and have to be fixed in the future. Put this to your
> TODO list, please.

Thanks for your comments. I will address the remaining ones in due time.
Lucas