Re: [PATCH v2 04/13] bus: ts-nbus: use gpiod_multi_set_value_cansleep

[Yury Norov]

On Fri, Feb 07, 2025 at 09:23:56AM -0600, David Lechner wrote:
> On 2/7/25 6:17 AM, Andy Shevchenko wrote:
> > +Yury.
> > 
> > On Fri, Feb 7, 2025 at 2:15 PM Andy Shevchenko
> > <andy.shevchenko@xxxxxxxxx> wrote:
> >> On Fri, Feb 7, 2025 at 12:48 AM David Lechner <dlechner@xxxxxxxxxxxx> wrote:
> > 
> > ...
> > 
> >>>  static void ts_nbus_write_byte(struct ts_nbus *ts_nbus, u8 byte)
> >>>  {
> >>> -       struct gpio_descs *gpios = ts_nbus->data;
> >>>         DECLARE_BITMAP(values, 8);
> >>>
> >>>         values[0] = byte;
> >>>
> >>> -       gpiod_set_array_value_cansleep(8, gpios->desc, gpios->info, values);
> >>> +       gpiod_multi_set_value_cansleep(ts_nbus->data, values);
> >>
> >> As I said before, this is buggy code on BE64. Needs to be fixed.
> > 
> > Or isn't? Do we have a test case in bitmap for such a case?
> > 
> >>>  }
> > 
> > 
> 
> Maybe not the best style, but I don't think it is buggy. Bitmaps are always
> handled in long-sized chunks and not cast to bytes so endianness doesn't affect
> it. I didn't see an explicit test, but bitmap_read() and bitmap_write() use
> array access like this so indirectly it is being tested.

Not a bug, but direct addressing to bitmap elements is discouraged.
I'd suggest using bitmap_write(values, byte, 0, 8) instead.

Thanks,
Yury