Comment # 23 on bug 1204315 from Andreas F���rber

(In reply to Takashi Iwai from comment #22)
> The kernel should avoid setting such a known bad config combo, or at least,
> it should give more useful information at runtime if the known bad kconfig
> combo must be still applicable.

Wouldn't the best solution be to implement efifb as drm driver?
I assume it uses runtime callback hooks beyond just accessing the memory area
like simpledrm does.

Also, I would ask to avoid upstream fixes as depending on "ARM" just because we
happened to encounter it there - please make it !X86 or whatever applicable.
RISC-V, C-Sky or MIPS may well run into similar simple-framebuffer issues.