[U-Boot] [PATCH 55/60] i2c: tegra: move pinmux setup to board files

[Simon Glass]

Hi Stephen,

On 27 April 2016 at 10:24, Stephen Warren <swarren at wwwdotorg.org> wrote:
> On 04/27/2016 09:12 AM, Simon Glass wrote:
>>
>> Hi Stephen,
>>
>> On 19 April 2016 at 14:59, Stephen Warren <swarren at wwwdotorg.org> wrote:
>>>
>>> From: Stephen Warren <swarren at nvidia.com>
>>>
>>> Remove funcmux calls from the Tegra I2C driver. Knowledge of pinmux
>>> setup must come from either board files or DT; it should not be embedded
>>> into board-agnostic driver code. The DT pinmux bindings do not allow
>>> drivers to derive funcmux-style information, since the DT bindings are
>>> pin-based whereas funcmux is controller-based, so there's no good way to
>>> call the existing funcmux APIs from drivers. Converting drivers to use a
>>> new (as yet non-existent in U-Boot) API that pulls pinmux information
>>> from
>>> DT isn't useful for Tegra, since Tegra's DT files don't contain any
>>> per-device pinmux tables, so this would simply be extra code that has no
>>> effect; doesn't actually set up the pinmux. We are left with moving the
>>> pinmux setup functionality into board files. In theory the board files
>>> could be converted later to use DT, but that would be a separate change.
>>>
>>> Signed-off-by: Stephen Warren <swarren at nvidia.com>
>>> ---
>>>   board/avionic-design/common/tamonten.c  |  5 +++++
>>>   board/nvidia/seaboard/seaboard.c        |  3 +++
>>>   board/nvidia/whistler/whistler.c        |  1 +
>>>   board/toradex/colibri_t20/colibri_t20.c |  3 +++
>>>   drivers/i2c/tegra_i2c.c                 | 19 -------------------
>>>   5 files changed, 12 insertions(+), 19 deletions(-)
>>
>>
>> This should use driver model, which handles pinmux automatically if
>> you have a pinctrl driver.
>
>
> Can you define "automatic"? I don't understand exactly which benefit you're
> describing there.

When you probe a device, its pinmux is set up automatically, so the
explicit funcmux calls can go away.

>
> I'd rather keep this series as simple cleanup of existing code, and handle
> any large-scale DM conversions separately later. That said, as full
> disclosure, I'm certainly not signing up for any more work on pinctrl,
> especially if it involves reading the pinmux tables from DT, as I see no
> benefit in that. Note that in future chips the boot ROM will handle 100% of
> static pinmux setup so there won't be a U-Boot driver for those SoCs going
> forward. I'd rather not invest any more than minimal effort in something
> that's going away.

There's no need for it to read from DT, and it doesn't have to be
comprehensive but I think you should create a pinctrl driver and put
the funcmux stuff in that (along with processing your big tables of
pinmux stuff).

Regards,
Simon