Re: [PATCH v3] x86: sysfb: remove sysfb when probing real hw

[David Herrmann]

Hi

On Thu, Dec 19, 2013 at 6:09 PM, Ingo Molnar <mingo@xxxxxxxxxx> wrote:
>
> * David Herrmann <dh.herrmann@xxxxxxxxx> wrote:
>
>> Hi
>>
>> On Thu, Dec 19, 2013 at 5:36 PM, Ingo Molnar <mingo@xxxxxxxxxx> wrote:
>> >
>> > * David Herrmann <dh.herrmann@xxxxxxxxx> wrote:
>> >
>> >> --- a/drivers/video/fbmem.c
>> >> +++ b/drivers/video/fbmem.c
>> >> @@ -35,6 +35,9 @@
>> >>
>> >>  #include <asm/fb.h>
>> >>
>> >> +#if IS_ENABLED(CONFIG_X86)
>> >> +#include <asm/sysfb.h>
>> >> +#endif
>> >
>> > I think this can be written as:
>> >
>> > #ifdef CONFIG_X86
>> > # include <asm/sysfb.h>
>> > #endif
>> >
>> > also ... the dependency on a large, unrelated option like CONFIG_X86
>> > looks pretty ugly here - is there no other, more specific CONFIG_
>> > option that can be used here - such as CONFIG_FB_SIMPLE or
>> > CONFIG_X86_SYSFB?
>> >
>> >> +#if IS_ENABLED(CONFIG_X86)
>> >> +     sysfb_unregister(apert, primary);
>> >> +#endif
>> >
>> > Ditto.
>>
>> CONFIG_X86 is probably never 'm'.. will fix that. It was
>> CONFIG_X86_SYSFB before and that works, too, but the broader X86
>> seemed more appropriate as sysfb is available on all x86.
>
> Well, sysfb is available if CONFIG_X86_SYSFB is set, right? So on
> !CONFIG_X86_SYSFB x86 kernels this code should not run, right?

No. The sysfb code is always enabled. It provides platform-devices for
firmware-framebuffers which efifb/vesafb pick up. The X86_SYSFB option
only controls whether generic system-framebuffers should be used
instead of the specific efi/vesa FBs (to be compatible to other
architectures and keep efi/vesa specifics in arch/x86/).

But the system-framebuffer conversion (to a format compatible to
simplefb/simple-framebuffer) is what may break as the legacy fbs don't
reserve resources. Hence, putting it enclosed into X86_SYSFB works for
now. But if we start registering efifb/vesafb with sysfb, too, then we
need to change it to CONFIG_X86 as all fbs are affected then.

I think using X86_SYSFB (as in v1 of this patch) is the way to go. I
will fix it up and resend v4. The broader #ifdef can be used once we
do the more complex cleanups, in which case it will go away entirely.

>> Note that I have patches here locally which move
>> sysfb_register/unregister to drivers/video/sysfb.c and add
>> include/linux/sysfb.h with dummies if CONFIG_SYSFB is not selected
>> to avoid the #ifdef. This will allow other architectures to do
>> gfx-hand-over, too. They seem too big for stable, though. That's why
>> I split them up and added it to x86/kernel/sysfb.c first.
>
> Yeah, it's fine to do those cleanups after the minimal fix. But using
> a sensible config option must still be done - we cannot just slap
> broad CONFIG_X86 dependencies into random code.

Ok.

Thanks
David
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/