Re: [PATCH v2 6/7] platform: x86: Add intel_skl_int3472 driver

[Laurent Pinchart]

Hi Daniel,

On Tue, Jan 19, 2021 at 12:11:40AM +0000, Daniel Scally wrote:
> On 18/01/2021 21:19, Daniel Scally wrote:
> >>>> +static const struct clk_ops skl_int3472_clock_ops = {
> >>>> +	.prepare = skl_int3472_clk_prepare,
> >>>> +	.unprepare = skl_int3472_clk_unprepare,
> >>>> +	.enable = skl_int3472_clk_enable,
> >>>> +	.disable = skl_int3472_clk_disable,
> >>>> +};
> >>
> >> Yeah, sounds like reinventing clk-gpio.c.
> >>
> >> static const struct clk_ops clk_gpio_gate_ops = {
> >> 	.enable = clk_gpio_gate_enable,
> >> 	.disable = clk_gpio_gate_disable,
> >> 	.is_enabled = clk_gpio_gate_is_enabled,
> >> };
> >>
> >> (Or is it mux? It has support there as well.
> >>
> > Hmm, yeah, this looks like it would work actually. So I think I'd need to:
> >
> > 1. Make enabling INTEL_SKL_INT3472 also enable the clk-gpio driver
> >
> > 2. Register a platform device to bind to the clk-gpio driver
> >
> > 3. Register a gpio lookup table so that the clk-gpio driver can find the
> > gpio in question using gpiod_get()
> >
> > And that looks like it will work; I'll try it.
> 
> I'm more and more confident that this will work, but it has some
> knock-on effects:
> 
> The both clk and regulator gpio driver expects to be able to fetch the
> GPIO using devm_gpiod_get(&pdev->dev, "enable", ...). That won't work of
> course, so we need to add another GPIO lookup table so those drivers can
> see the GPIOs. For that, we need to know what dev_name(&pdev->dev) will
> be so we can set the .dev_id member of a gpiod_lookup_table to that
> value, but that isn't set until _after_ the pdev is registered (because
> it has to figure out the id, we can't manually set the IDs because there
> could be more than one instance of int3472-discrete bound to multiple
> PMIC devices, and we don't know which id the current one should have).
> Finally, we can't wait until the device is registered because it
> immediately probes, can't find the GPIO and then fails probe.
> 
> It's similar problem that causes us to need the i2c-acpi name format
> macros, but complicated by the dynamic ID part of dev_name(&pdev->dev)
> 
> Solving it is a bit of a sticky one; perhaps something like moving the
> dev_set_name() part of platform_device_add() [1] to its own function,
> that's called in both platform_device_alloc() and
> platform_device_register(). That way it would be available before the
> device itself was registered, meaning we could create the lookup table
> before it probes the driver.
> 
> (also, Laurent, if we did it this way we wouldn't be able to also handle
> the led-indicator GPIO here without some fairly major rework)

Given the additional complexity I don't think it's worth it, your
implementation is fine and code duplication with clk-gpio is minimal.

> [1]
> https://elixir.bootlin.com/linux/latest/source/drivers/base/platform.c#L563

-- 
Regards,

Laurent Pinchart