Re: git: 901df07a4768 - main - Code refactoring for existing rk_gpio driver. It supports gpio type checking. Depending on gpio type some register addresses are different.

From: Peter Jeremy <peterj_at_freebsd.org>
Date: Sun, 21 Aug 2022 04:05:05 UTC 
On 2022-Aug-19 13:22:57 +0000, Ganbold Tsagaankhuu <ganbold@FreeBSD.org> wrote:
>The branch main has been updated by ganbold:
>
>URL: https://cgit.FreeBSD.org/src/commit/?id=901df07a47684dca7b06f60d838a56456d751a23
>
>commit 901df07a47684dca7b06f60d838a56456d751a23
>Author:     Søren Schmidt <sos@FreeBSD.org>
>AuthorDate: 2022-08-19 13:22:01 +0000
>Commit:     Ganbold Tsagaankhuu <ganbold@FreeBSD.org>
>CommitDate: 2022-08-19 13:22:01 +0000
>
>    Code refactoring for existing rk_gpio driver.
>    It supports gpio type checking. Depending on gpio type some
>    register addresses are different.
>    
>    Reviewed by:    manu
>    Differential Revision:  https://reviews.freebsd.org/D36262

My Rock64 is now hanging on boot as follows:
rk_pinctrl0: <RockChip Pinctrl controller> on ofwbus0
gpio0: <RockChip GPIO Bank controller> mem 0xff210000-0xff2100ff irq 53 on rk_pinctrl0
gpio0: Unknown gpio version 48000000
<<hang>>

>@@ -170,6 +221,43 @@ rk_gpio_attach(device_t dev)
> 		rk_gpio_detach(dev);
> 		return (ENXIO);
> 	}
>+	RK_GPIO_LOCK(sc);
>+	sc->version = rk_gpio_read_4(sc, RK_GPIO_VERSION);
>+	RK_GPIO_UNLOCK(sc);

This call to rk_gpio_read_4() looks wrong:
a) rk_gpio_read_4() tests sc->version which this call is setting.
b) The second argument to rk_gpio_read_4() is a "enum gpio_regs", not
   an actual offset - RK_GPIO_VERSION (=0x78) is way outside the
   sc->regs array.
c) sc->regs is also uninitialised at this point.

Maybe this should call RK_GPIO_READ() instead, but neither my RK3328
TRM (revision 1.2 from July 2017) nor my RK3399 TRM (revision 1.4 from
April 2017) document a GPIO register at offset 0x78 - both only go to
0x60.  (If you have a later TRM for either chip, I would be interested
in a copy).

-- 
Peter Jeremy