Re: replacement for /sys/kernel/debug/omap_mux in DT/pinctrl land ?

[Koen Kooi <koen@dominion.thruhere.net>]

Op 26 sep. 2012, om 18:15 heeft Matt Porter <mporter@ti.com> het volgende geschreven:

> On Wed, Sep 26, 2012 at 03:03:27PM +0200, Linus Walleij wrote:
>> On Wed, Sep 26, 2012 at 2:56 PM, Matt Porter <mporter@ti.com> wrote:
>>> Adding Linus W. and lkml.
>>> On Wed, Sep 26, 2012 at 01:46:45PM +0200, Koen Kooi wrote:
>>>> Hi,
>>>> 
>>>> With a patched 3.6rc7 on my beaglebone I can set the pinmux for pins using pinctrl and that seems to work. On the 3.2 vendor tree there was the omap_mux driver with an awesome debugfs interface:
>>>> 
>>>> # cat /sys/kernel/debug/omap_mux/lcd_data0
>>>> name: lcd_data0.ehrpwm2A (0x44e108a0/0x8a0 = 0x0003), b NA, t NA
>>>> mode: OMAP_PIN_OUTPUT | OMAP_MUX_MODE3
>>>> signals: lcd_data0 | gpmc_a0 | pr1_mii_mt0_clk | ehrpwm2A | NA | pr1_pru1_pru_r30_0 | pr1_pru1_pru_r31_0 | gpio2_6
>>>> 
>>>> Notice how it tells me that it's muxed the PWM in 2 ways: signal name (ehrpwm2A) and register content (0x0003). Compare to pinctrl:
>>>> 
>>>> root@bone-mainline:/sys/kernel/debug/pinctrl/44e10800.pinmux# grep 8a0 *
>>>> pinconf-pins:pin 40 (44e108a0):
>>>> pingroups:pin 40 (44e108a0)
>>>> pinmux-pins:pin 40 (44e108a0): 4a300000.pruss (GPIO UNCLAIMED) function pinmux_pruss_led_pins group pinmux_pruss_led_pins
>>>> pins:pin 40 (44e108a0)  pinctrl-single
>>>> 
>>>> What is that pin muxed to? It is part of the 'pinmux_pruss_led_pins' in the DT, but debugfs remains mute on how pin 40 is muxed.
>>> 
>>> It does seem like a pretty big gap in the pinctrl/pinmux debugfs
>>> interface when viewed from an OMAP perspective. Ideally there would
>>> be a pinctrl/pinmux hook to the pinmux driver to provide the detailed
>>> h/w specific pin state info.
>> 
>> So add the hooks you need?
> 
> Ok. :)
> 
>> I assume you are using Tony's pinctrl-single driver, so Tony is the one to ask.
> 
> Yes, so roughly for pinctrl-single I have the following...likely broken
> for arbitrary register sizes but a starting point. Tony, any thoughts
> about this?
> 
> Koen: you just need a userspace tool that groks the raw data for human
> consumption. The nice thing is that the old omap_mux implementation had
> plenty of OMAP-isms in the parser that didn't apply to AM33xx. A
> userspace tool can do a better job of parsing on a per-part basis.
> 
> --- a/drivers/pinctrl/pinctrl-single.c
> +++ b/drivers/pinctrl/pinctrl-single.c
> @@ -246,7 +246,15 @@ static void pcs_pin_dbg_show(struct pinctrl_dev *pctldev,
> 					struct seq_file *s,
> 					unsigned offset)
> {
> -	seq_printf(s, " " DRIVER_NAME);
> +	struct pcs_device *pcs;
> +	unsigned val;
> +
> +	pcs = pinctrl_dev_get_drvdata(pctldev);
> +
> +	val = pcs->read(pcs->base + offset);
> +	val &= pcs->fmask;
> +
> +	seq_printf(s, "%08x %s " , val, DRIVER_NAME);
> }
> 
> static void pcs_dt_free_map(struct pinctrl_dev *pctldev,

Much better already:

root@bone-mainline:/sys/kernel/debug/pinctrl/44e10800.pinmux# grep 8a0 pins
pin 40 (44e108a0) 00000027 pinctrl-single 

Now I can write a userspace tool do list the current muxes without resorting to devmem2!