Re: [PATCH v3 1/7] gpio: gpio-ep93xx: fix BUG_ON port F usage

From: Alexander Sverdlin <alexander.sverdlin@gmail.com>
To: "nikita.shubin@maquefel.me" <nikita.shubin@maquefel.me>,
	Andy Shevchenko <andy.shevchenko@gmail.com>
Cc: Linus Walleij <linus.walleij@linaro.org>,
	Bartosz Golaszewski <bgolaszewski@baylibre.com>,
	"open list:GPIO SUBSYSTEM" <linux-gpio@vger.kernel.org>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH v3 1/7] gpio: gpio-ep93xx: fix BUG_ON port F usage
Date: Thu, 04 Feb 2021 14:36:45 +0100	[thread overview]
Message-ID: <164d778019f68dd024cb42f869e7d967618514cb.camel@gmail.com> (raw)
In-Reply-To: <28201612442592@mail.yandex.ru>

Hi Nikita,

On Thu, 2021-02-04 at 15:55 +0300, nikita.shubin@maquefel.me wrote:
> I considered your offer of using array with holes.
>  
> It looks pretty ugly to me, couse it leads to bloated arrays:
>  
> static unsigned char gpio_int_unmasked[EP93XX_GPIO_CHIP_NUM];
> static unsigned char gpio_int_enabled[EP93XX_GPIO_CHIP_NUM];
> static unsigned char gpio_int_type1[EP93XX_GPIO_CHIP_NUM];
> static unsigned char gpio_int_type2[EP93XX_GPIO_CHIP_NUM];
> static unsigned char gpio_int_debounce[EP93XX_GPIO_CHIP_NUM];
>  
> /* Port ordering is: A B F */
> static const u8 int_type1_register_offset[EP93XX_GPIO_CHIP_NUM]    = { 0x90, 0xac, 0x0, 0x0, 0x0, 0x4c };
> static const u8 int_type2_register_offset[EP93XX_GPIO_CHIP_NUM]    = { 0x94, 0xb0, 0x0, 0x0, 0x0, 0x50 };
> static const u8 eoi_register_offset[EP93XX_GPIO_CHIP_NUM]    = { 0x98, 0xb4, 0x0, 0x0, 0x0, 0x54 };
> static const u8 int_en_register_offset[EP93XX_GPIO_CHIP_NUM]    = { 0x9c, 0xb8, 0x0, 0x0, 0x0, 0x58 };
> static const u8 int_debounce_register_offset[EP93XX_GPIO_CHIP_NUM]    = { 0xa8, 0xc4, 0x0, 0x0, 0x0, 0x64 };
>  
> Is this really the thing we want ?

Even in this form it's less error-prone than to have two
index-spaces, and hidden conversion from one numbering scheme
to other.

Alternatives that I see are:
1.
https://gcc.gnu.org/onlinedocs/gcc/Designated-Inits.html

2.
Embedd the necessary values into struct ep93xx_gpio_bank.
This option can probably simplify the handling of the names
for irq chips as well.

> 28.01.2021, 19:19, "Alexander Sverdlin" <alexander.sverdlin@gmail.com>:
> > Hello Nikita,
> > 
> > On Thu, 2021-01-28 at 18:11 +0200, Andy Shevchenko wrote:
> > >  > +/*
> > >  > + * F Port index in GPIOCHIP'S array is 5
> > >  > + * but we use index 2 for stored values and offsets
> > >  > + */
> > >  > +#define EP93XX_GPIO_F_PORT_INDEX 5
> > >  
> > >  Hmm... Why not to use an array with holes instead.
> > >  
> > >  ...
> > >  
> > >  > +       if (port == EP93XX_GPIO_F_PORT_INDEX)
> > >  > +               port = 2;
> > >  
> > >  Sorry, but I'm not in favour of this as it adds confusion.
> > >  See above for the potential way to solve.
> > 
> > well, I was thinking the same yesterday. It just adds another
> > level on confusion into the code, which even the author got
> > wrong :)
> > 
> > Array with holes would be more obvious, but one can also embedd
> > the necessary values into struct ep93xx_gpio_bank.
> >  
> > --
> > Alexander Sverdlin.
> > 
> >  

-- 
Alexander Sverdlin.