[PATCH] ASoC: fsi: Stop using __raw_*() I/O accessors

[Arnd Bergmann]

On Thu, Nov 19, 2020 at 1:53 PM Geert Uytterhoeven
<geert+renesas at glider.be> wrote:
>
> There is no reason to keep on using the __raw_{read,write}l() I/O
> accessors in Renesas ARM or SuperH driver code.  Switch to using the
> plain {read,write}l() I/O accessors, to have a chance that this works on
> big-endian.
>
> Suggested-by: Arnd Bergmann <arnd at kernel.org>
> Signed-off-by: Geert Uytterhoeven <geert+renesas at glider.be>

I don't think this one is correct, as based on

static void fsi_pio_push16(struct fsi_priv *fsi, u8 *_buf, int samples)
{
        int i;

        if (fsi_is_enable_stream(fsi)) {
                /*
                 * stream mode
                 * see
                 *      fsi_pio_push_init()
                 */
                u32 *buf = (u32 *)_buf;

                for (i = 0; i < samples / 2; i++)
                        fsi_reg_write(fsi, DODT, buf[i]);
        } else {
                /* normal mode */
                u16 *buf = (u16 *)_buf;

                for (i = 0; i < samples; i++)
                        fsi_reg_write(fsi, DODT, ((u32)*(buf + i) << 8));
        }
}

the same helper is used for both accessing endianness-sensitive
register values (which need the converrsion in writel()) and
possibly in-memory byte streams that need the __raw_writel()
behavior of copying the input bytes in sequence rather than sets of
native-endian 16-bit or 32-bit samples.

> ---
> Assembler output difference on SuperH checked.

I'm also not sure whether changing this breaks big-endian SuperH,
which defines the accessors differently from Arm and most other
architectures.

Maybe better just mark the driver as 'depends on SH || !CPU_BIG_ENDIAN'
as it is clearly broken on big-endian Arm.

     Arnd