Re: [PATCH v4 0/8] Introduce intel_skl_int3472 module

[Daniel Scally <djrscally@xxxxxxxxx>]

Hi Hans

On 25/05/2021 14:10, Hans de Goede wrote:
> Hi Daniel,
>
> On 5/20/21 4:09 PM, Daniel Scally wrote:
>> Hello all
>>
>> Apologies for the long delay since the last version of this series; the time I
>> had free to work on it became somewhat restrained.
> No worries, thank you for all the work you are putting into this.


My pleasure

> I have not taken a close look at the code yet, but I see that Andy has and
> the amount of remarks which he has on patch 7/8 which is the big one seems
> to be limited, so I believe that we are getting close to this being ready
> for merging.
>
> This touches a lot of subsystems, so we need to come up with a plan to
> merge this. Here is my proposal for how to do this:
>
> 1/8   ACPI: scan: Extend acpi_walk_dep_device_list()
> 2/8   ACPI: scan: Add function to fetch dependent of acpi device
> 3/8   i2c: core: Add a format macro for I2C device names
> 4/8   gpiolib: acpi: Export acpi_get_gpiod()
> 5/8   clkdev: Make clkdev_drop() null aware
> 6/8   gpiolib: acpi: Add acpi_gpio_get_io_resource()
> 7/8   platform/x86: Add intel_skl_int3472 driver
> 8/8   mfd: tps68470: Remove tps68470 MFD driver
>
> Rafael already indicated that he wants to merge 1/8 (and presumably also 2/8)
> through his tree and that he will provide an immutable branch with those
> for merging into the pdx86 tree.


I'll send a v5 with the renames asap, might try and do the other changes
and send the whole series, depends how much time I get to work on it
over the next few days...

> 4/8 and 6/8 are both gpiolib-acpi patches and seem to be ready for merging
> now, perhaps the gpiolib-acpi maintainers can already merge these and also
> provide an immutable branch ?  Andy/Mika ?


So, Andy, you'd prefer I re-order these so they're consecutive...did I
understand that right?

> 3/8 and 5/8 seem to be nice cleanups, but not really necessary. IMHO it
> would be best to park these cleanups for later and for 3/8 add the following
> where necessary for now:
>
> /* FIXME drop this once the I2C_DEV_NAME_FORMAT macro has been added to include/linux/i2c.h */
> #ifndef I2C_DEV_NAME_FORMAT
> #define I2C_DEV_NAME_FORMAT		"i2c-%s"
> #endif
>
> This is not the prettiest but it reduces all the subsys cross-deps and things
> like this have been done before for similar reasons.
>
> Likewise it would be good if you can add if (foo) as condition before any
> clkdev_drop(foo) calls in this patch-set and then merge
> 5/8 "clkdev: Make clkdev_drop() null aware" independently of this and then
> once both are in Linux tree follow-up with a cleanup patch dropping the if (foo)
> guards.


This is fine by me if people are happy for it to go in like that; I'll
just fix it up later.

> So this would leave as deps for 7/8 just the 2 ACPI and 2 gpiolib-acpi patches
> which I can hopefully pull-in via immutable branches and then we are good.
>
> AFAICT patch 8/8 can be merged independently once 7/8 hits for-next (IOW once
> we are sure the next kernel will have 7/8).
>
>
>
> Or alternatively one of the involved subsys maintainers just merges the entire
> set (once it is ready) and then provides an immutable branch with the entire set
> on top of 5.13-rc1 (or 5.14-rc1). But that requires acks from all the other
> subsys maintainers. Note I'm fine with either approach.