Re: [PATCH 2/2] pinctrl: qcom: ipq4019: add remaining pin definitions

[Bjorn Andersson <bjorn.andersson@linaro.org>]

On Thu 18 May 11:38 PDT 2017, Christian Lamparter wrote:

> On Wednesday, May 17, 2017 1:07:29 PM CEST Bjorn Andersson wrote:
> > On Wed 10 May 04:27 PDT 2017, Christian Lamparter wrote:
> > > diff --git a/drivers/pinctrl/qcom/pinctrl-ipq4019.c b/drivers/pinctrl/qcom/pinctrl-ipq4019.c
> > > index 743d1f458205..7219d1e33c71 100644
> > > --- a/drivers/pinctrl/qcom/pinctrl-ipq4019.c
> > > +++ b/drivers/pinctrl/qcom/pinctrl-ipq4019.c
> > > +	qca_mux_rmii0_refclk,
> > > +	qca_mux_wifi0_rfsilient0,
> > > +	qca_mux_wifi1_rfsilient0,
> > > +	qca_mux_smart2,
> > > +	qca_mux_led4,
> > 
> > What drives ledX? Is it 11 different LED controllers or is it a single
> > LED controller with 11 outputs.
> 
> The latter. The IPQ40xx have one LED controller @ 0x1937000. 
> According to the driver (leds-ipq40xx.c in the SDK), it 
> does control up to 11 LEDs. A LED can either be muxed to
> one of the hardware sources (wifi, lan or wan-ports activity/linkspeed),
> or it can be operated by one of four software-programmable "blink"
> sources (each with a variable blink rate and duty cycle). 
> The driver labels each LED as "ipq40xx::led%d".
> 

As they all stem from the same hardware block I suggest we name the
function "led".

> That said: ASUS, Cisco, Compex, Netgear, Zyxel... opted to either
> 1. export the individual GPIOs with sysfs
> 2. gpio-leds
> 

For any software-driven LED control it makes sense to configure the pins
as gpio and just rely on the gpio-leds driver.

But as you have both hardware driven control and hardware blink support
it sounds like there's good reason to have a specific LED driver as
well.

> > [..]
> > > +	qca_mux_wifi01,
> > 
> > Please make these "wifi0" and include all "wifi0XY", rather than having
> > a group per pin.
> > 
> > > +	qca_mux_wifi11,
> > 
> > "wifi1"
> 
> Ok. Can I leave wifi1_cal, _wci, uartX... the way they are?
>  

Yes, that sounds reasonable

> > > +	qca_mux_atest_char3,
> > > +	qca_mux_pmu0,
> > > +	qca_mux_boot8,
> > > +	qca_mux_tm1,
> > > +	qca_mux_atest_char2,
> > > +	qca_mux_pmu1,
> > > +	qca_mux_boot9,
> > > +	qca_mux_tm2,
> > > +	qca_mux_atest_char1,
> > > +	qca_mux_tm_ack,
> > > +	qca_mux_wifi03,
> > > +	qca_mux_wifi13,
> > > +	qca_mux_qpic_pad4,
> > 
> > Please keep an eye on the ipq8074 patch from Varadarajan and make this
> > follow the same scheme.
> Ok, I'll wait for how qca8074 plays out then.
> By the way, can you ask if the QCA8074 follows the IPQ40XX SoC's
> GPIO Pull-up config?
> 
> I'm asking because Varadarajan was also involved in the
> pinctrl-ipq4019 back in 2015: <https://patchwork.kernel.org/patch/7662241/>
> 
> Back then, this wasn't mentioned anywhere. In fact, the special pull-up
> configuration was only discovered due to an issue with the NAND on the
> Cisco Meraki MR33. So I think it is better ask them now, when the devs
> are actually present/responding. 
> 

I do unfortunately not have access to the specification or either
platform, I would appreciate if you could post a reply on Varadarajan's
IPQ8074 patch asking him to verify the layout of the config register.

[..]
> > > +	qca_mux_boot18,
> > 
> > Do you know what the "boot" function is and what 2, 4, 5, 7, 8, 9 11,
> > 14, 18, 19 and 20 means?
> Sadly no. That said, neither the u-boot nor the linux kernel
> sources set any pin to bootX. I think I'll remove it for now.
> 

Better leave them out for now then.

> > [..]
> > > +	qca_mux_sdio0,
> > 
> > There are 8 of these, so that's more likely the 8 data pins in a single
> > function. Please squash them into "sdio_data".
> Yes, this is very likely. We know that this is the case for the qpic_padX.
> 

Sounds good.

Regards,
Bjorn