On Tue, Jun 8, 2021 at 6:49 PM Hans Verkuil hverkuil@xs4all.nl wrote:

On 08/06/2021 18:14, Arnd Bergmann wrote:

Right now it is inherent to the driver. It is probably possible to drop support for video overlay devices if CONFIG_FB=n, but it is not something I have time for. It's just a test driver (albeit a very useful test driver), so it is no big deal if it is disabled when CONFIG_FB=n.

Ok, thanks for the reply, makes sense.

I checked what other consequences there are if we disable CONFIG_FB and CONFIG_DRM_KMS_FB_HELPER=y in all the defconfigs now, as the patch from Kees did.

It appears that the only other arm32 framebuffers that remain are FB_EFI=y, FB_WM8505=y, FB_MX3=m and FB_SIMPLE=y.

On arm64, losing CONFIG_FB=y would disable FB_EFI=y, XEN_FBDEV_FRONTEND=y, and FB_MX3=m

On x86, it's only CONFIG_FB_EFI

It appears that FB_MX3 is orphaned since commit e1324ece2af4 ("ARM: imx: Remove i.MX35 board files") because all Armv6 or newer i.MX now use drivers/gpu/drm/mxsfb/mxsfb_drv.c, and FB_WM8505 is probably unused as well (we discussed removing the platform last winter, but decided to give it another year to see if new users come up, which has not happened).

As long as simplefb, efifb and xenfb are needed though, we probably want CONFIG_FB=y anyway and leaving VIVID=m with the dependency does not cause problems until those are all turned into drm drivers.

Arnd