Re: [PATCH v1 1/6] mfd: Add driver for Embedded Controller found on Acer Iconia Tab A500

[Dmitry Osipenko <digetx@gmail.com>]

23.08.2020 21:16, Lubomir Rintel пишет:
> Hello,
...
>> +config MFD_ACER_A500_EC
>> +	tristate "Embedded Controller driver for Acer Iconia Tab A500"
>> +	depends on (I2C_TEGRA && ARCH_TEGRA_2x_SOC) || COMPILE_TEST
> 
> This seems to also depend on I2C and OF. Perhaps I2C_TEGRA and
> ARCH_TEGRA_2x_SOC imply that, but it could lead to build failures with
> COMPILE_TEST=y. 

Hello, Lubomir! You're right about the I2C because it could be compiled
as a loadable module, good catch! The OF seems should fine as-is.

...
>> +static struct a500_ec *a500_ec_scratch;
> 
> If this is only used for power_off, please rename it. I've been told to
> do so in my driver: https://lore.kernel.org/lkml/20200519104933.GX271301@dell/

I don't mind to rename the variable, but not sure whether it will be a
worthwhile change since _scratch is also a common naming scheme among
MFD drivers. Please see max77620_scratch for example, which I added
about a year ago.

...
>> +int a500_ec_read_locked(struct a500_ec *ec_chip,
>> +			const struct a500_ec_cmd *cmd_desc)
> 
> Any reason you're exporting these to the cell drivers instead of using
> regmap?
> 
> I think regmap would also help you with the locking if you set .lock() and
> .unlock() callbacks in regmap_config.

Yes, perhaps you're right. Right now I can't recall what stopped me from
using regmap. I'll give it a shot for v2, thank you for the suggestion!

...
>> +static struct notifier_block a500_ec_restart_handler = {
>> +	.notifier_call = a500_ec_restart_notify,
>> +	.priority = 200,
> 
> What would happend if you didn't set priority explicitly?

Then the Tegra's default CPU soft-reset handler will be used for
rebooting [1].

[1]
https://elixir.bootlin.com/linux/v5.9-rc1/source/drivers/soc/tegra/pmc.c#L977

...
>> +MODULE_LICENSE("GPL v2");
> 
> MODULE_LICENSE("GPL");
> 
> Your SPDX tag suggests newer versions of GPL than v2 are okay.

Okay! I'll change this in v2, thanks.

...
>> +#define A500_EC_COMMAND(NAME, CMD, DELAY_MS)				\
>> +static const struct a500_ec_cmd A500_EC_##NAME = {			\
>> +	.cmd = CMD,							\
>> +	.post_delay = DELAY_MS,						\
>> +};									\
> 
> I think that the mfd driver should decide about the delay, not the cell
> drivers.

This should be a good idea, especially combined with the regmap, thanks!