Hi Thomas,
On Tue, Apr 16, 2024 at 2:07 PM Thomas Zimmermann <tzimmermann@xxxxxxx> wrote:
> Am 16.04.24 um 13:25 schrieb Javier Martinez Canillas:
> > Thomas Zimmermann <tzimmermann@xxxxxxx> writes:
> > Do I understand correctly that info->fix.smem_start doesn't have to be set
> > because that's only used for I/O memory?
>
> It's the start of the framebuffer memory in physical memory. Setting
> smem_start only makes sense if the framebuffer is physically continuous,
> which isn't the case here.
Nothing really needs fix.smem_start, it's mainly for informative use.
However, if smem_start is not page-aligned, userspace does need to
know the start offset inside the page (see below).
> > This also made me think why info->fix.smem_len is really needed. Can't we
> > make the fbdev core to only look at that if info->screen_size is not set ?
>
> The fbdev core doesn't use smem_len AFAICT. But smem_len is part of the
> fbdev UAPI, so I set it. I assume that programs use it to go to the end
> of the framebuffer memory.
On fbdev drivers also exporting MMIO to userspace, /dev/fbX contains
two parts: first the frame buffer, followed by the MMIO registers.
Both parts are an integral number of pages, based on fix.smem_{start,len}
resp. fix.mmio_{start,len}.
Old XFree86 used the MMIO part to implement hardware acceleration
when running on top of fbdev.
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
