Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> writes:
Hello Geert and Thomas,
> Hi Thomas,
>
> On Mon, Jul 17, 2023 at 10:48 AM Thomas Zimmermann <tzimmermann@xxxxxxx> wrote:
[...]
>>
>> After some discussion on IRC, I'd suggest to allocate the buffer
>> somewhere within probe. So it will always be there when the plane code runs.
>>
>> A full fix would be to allocate the buffer memory as part of the plane
>> state and/or the plane's atomic_check. That's a bit more complicated if
>> you want to shared the buffer memory across plane updates.
>
> Note that actually two buffers are involved: data_array (monochrome,
> needed for each update), and buffer (R8, only needed when converting
> from XR24 to R1).
>
> For the former, I agree, as it's always needed.
> For the latter, I'm afraid it would set a bad example: while allocating
> a possibly-unused buffer doesn't hurt for small displays, it would
> mean wasting 1 MiB in e.g. the repaper driver (once it has gained
> support for R1 ;^).
>
Maybe another option is to allocate on the struct drm_mode_config_funcs
.fb_create callback? That way, we can get the mode_cmd->pixel_format and
determine if only "data_array" buffer must be allocated or also "buffer".
> Gr{oetje,eeting}s,
>
> Geert
>
> --
> Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx
>
> In personal conversations with technical people, I call myself a hacker. But
> when I'm talking to journalists I just say "programmer" or something like that.
> -- Linus Torvalds
>
--
Best regards,
Javier Martinez Canillas
Core Platforms
Red Hat
