[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <d8faabe9-1123-3d0e-6b7d-8335cd2b6a44@ise.pw.edu.pl>
Date: Mon, 15 Oct 2018 15:41:20 +0200
From: Wojciech Zabołotny <wzab@....pw.edu.pl>
To: Linus Walleij <linus.walleij@...aro.org>,
wzabolot@...ktron.elka.pw.edu.pl
Cc: Michal Simek <michal.simek@...inx.com>,
Randy Dunlap <rdunlap@...radead.org>,
Masahiro Yamada <yamada.masahiro@...ionext.com>,
"open list:GPIO SUBSYSTEM" <linux-gpio@...r.kernel.org>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: Bug introduced in the of_get_named_gpiod_flags function.
On 13.10.2018 18:23, Linus Walleij wrote:
> On Sat, Oct 13, 2018 at 5:53 PM wzabolot@...ktron.elka.pw.edu.pl
> <wzabolot@...ktron.elka.pw.edu.pl> wrote:
>
>> The question is, if there may be any other in-tree GPIO controller
>> driver that does not initialize those flags?
> So as I said, I looked over them and they all initialize
> their flags.
>
>> Anyway the current situation is somehow dangerous.
> Not very dangerous compared to much other kernel code.
>
>> Maybe the best solution would be to initialize the automatic variable in
>> the of_find_gpio?
> Patches welcom, make sure they are tested with mainline
> code and drivers.
>
> Yours.
> Linus Walleij
OK. So I suggested correction of the Xilinx GPIO driver so that it always sets the flags:
static int xgpio_xlate(struct gpio_chip *gc,
const struct of_phandle_args *gpiospec, u32 *flags)
{
struct of_mm_gpio_chip *mm_gc = to_of_mm_gpio_chip(gc);
struct xgpio_instance *chip = container_of(mm_gc, struct xgpio_instance,
mmchip);
if(flags)
* flags = 0;
if (gpiospec->args[1] == chip->offset)
return gpiospec->args[0];
return -EINVAL;
}
With best regards,
Wojtek
Powered by blists - more mailing lists