[linux-sunxi] Re: [PATCH 3/6] mfd: axp20x: Add support for RSB based AXP223 PMIC

[Chen-Yu Tsai]

On Tue, Oct 20, 2015 at 2:48 AM, Maxime Ripard
<maxime.ripard at free-electrons.com> wrote:
> On Mon, Oct 19, 2015 at 02:20:29PM +0800, Chen-Yu Tsai wrote:
>> On Mon, Oct 19, 2015 at 2:02 PM, Maxime Ripard
>> <maxime.ripard at free-electrons.com> wrote:
>> > On Fri, Oct 16, 2015 at 02:46:23PM +0800, Chen-Yu Tsai wrote:
>> >> On Fri, Oct 16, 2015 at 2:41 PM, Maxime Ripard
>> >> <maxime.ripard at free-electrons.com> wrote:
>> >> > On Thu, Oct 15, 2015 at 12:32:19AM +0800, Chen-Yu Tsai wrote:
>> >> >> The AXP223 is a new PMIC commonly paired with Allwinner A23/A33 SoCs.
>> >> >> It is functionally identical to AXP221; only the regulator default
>> >> >> voltage/status and the external host interface are different.
>> >> >>
>> >> >> Signed-off-by: Chen-Yu Tsai <wens at csie.org>
>> >> >> ---
>> >> >>  drivers/mfd/Kconfig        | 12 ++++++
>> >> >>  drivers/mfd/Makefile       |  1 +
>> >> >>  drivers/mfd/axp20x-core.c  |  2 +
>> >> >>  drivers/mfd/axp20x-rsb.c   | 93 ++++++++++++++++++++++++++++++++++++++++++++++
>> >> >>  include/linux/mfd/axp20x.h |  1 +
>> >> >>  5 files changed, 109 insertions(+)
>> >> >>  create mode 100644 drivers/mfd/axp20x-rsb.c
>> >> >>
>> >> >> diff --git a/drivers/mfd/Kconfig b/drivers/mfd/Kconfig
>> >> >> index 9ba3feb3f2fc..6e5edb61d42e 100644
>> >> >> --- a/drivers/mfd/Kconfig
>> >> >> +++ b/drivers/mfd/Kconfig
>> >> >> @@ -84,6 +84,7 @@ config MFD_BCM590XX
>> >> >>  config MFD_AXP20X
>> >> >>       bool "X-Powers AXP series PMICs"
>> >> >>       select MFD_AXP20X_I2C
>> >> >> +     select MFD_AXP20X_RSB
>> >> >>
>> >> >>  config MFD_AXP20X_CORE
>> >> >>       bool
>> >> >> @@ -102,6 +103,17 @@ config MFD_AXP20X_I2C
>> >> >>         components like regulators or the PEK (Power Enable Key) under the
>> >> >>         corresponding menus.
>> >> >>
>> >> >> +config MFD_AXP20X_RSB
>> >> >> +     bool "X-Powers AXP series RSB PMICs"
>> >> >> +     select MFD_AXP20X_CORE
>> >> >> +     depends on SUNXI_RSB=y
>> >> >
>> >> > Do we need that? Even if the bus is compiled as a module, the driver
>> >> > will not be probed before that, will it?
>> >>
>> >> There's a compile/link dependency on the __devm_regmap_init_sunxi_rsb().
>> >
>> > If it's exported, everything should be fine, no?
>> >
>> >> And both drivers are bool, i.e. can't be compiled as a module. What we
>> >> don't want is enabling MFD_AXP20X_RSB without SUNXI_RSB.
>> >
>> > What would really be the issue here? The driver wouldn't be probed,
>> > and that's it. Or am I missing something?
>>
>> The RSB bus / slave device functions have been merged into the RSB driver
>> itself. Enabling MFD_AXP20X_RSB without enabling SUNXI_RSB means that RSB
>> bus/device related functions are not compiled, i.e. link error:
>>
>> drivers/built-in.o: In function `axp20x_rsb_probe':
>> /home/wens/sunxi/linux/drivers/mfd/axp20x-rsb.c:64: undefined
>> reference to `__devm_regmap_init_sunxi_rsb'
>> drivers/built-in.o: In function `axp20x_rsb_driver_init':
>> /home/wens/sunxi/linux/drivers/mfd/axp20x-rsb.c:89: undefined
>> reference to `sunxi_rsb_driver_register'
>> Makefile:927: recipe for target 'vmlinux' failed
>>
>> The dependency is like "depends on I2C=y" for the I2C version.
>>
>> If you're asking about why "=y", I guess it's because MFD_AXP20X_RSB is bool,
>> and if the depended on symbol is a tristate, which it actually is for I2c,
>> we'd want it to be compiled in, and not built as a module, or again we'd get
>> a undefined reference link error.
>
> Yeah, but my point was more why not have both the RSB driver and MFD
> as a module? The part where RSB is a module and the driver is
> statically built doesn't make sense (and I don't think a depends on
> allow that), but having both make sense.

Ok. I have no problem with building them as modules. I was just following
what the original driver did.

It seems half the mfd driver can be built as modules, while the other
half can only be built-in. I don't know what the criteria is here.

>> Would it make sense to have SUNXI_RSB as a tristate symbol, i.e. can be built
>> as a module? I'm nore sure. For multi-platform kernels, probably? Currently it
>> isn't.
>
> Yes, it's better for multi-platform / distro kernels.

I guess I'll do a follow up patch for sunxi-rsb?

Regards
ChenYu