Re: [PATCH] m68k/mvme147: Don't unregister boot console needlessly

[Geert Uytterhoeven]

Hi Finn,

On Fri, 28 Mar 2025 at 23:25, Finn Thain <fthain@xxxxxxxxxxxxxx> wrote:
> On Fri, 28 Mar 2025, Geert Uytterhoeven wrote:
> > > --- a/arch/m68k/kernel/early_printk.c
> > > +++ b/arch/m68k/kernel/early_printk.c
> > > @@ -60,7 +60,7 @@ early_param("earlyprintk", setup_early_printk);
> > >
> > >  static int __init unregister_early_console(void)
> > >  {
> > > -       if (!early_console || MACH_IS_MVME16x)
> > > +       if (!early_console || MACH_IS_MVME147 || MACH_IS_MVME16x)
> > >                 return 0;
> > >
> > >         return unregister_console(early_console);
> >
> > Perhaps the whole function and the late_initcall() can just be removed?
> >
>
> A comment in arch/m68k/kernel/early_printk.c gives the reason why that
> code exists: debug_cons_nputs() lives in .init.text. Platforms like MVME
> which don't use that function to implement earlyprintk don't have to worry
> about this.

Thanks, I had forgotten about that...

> I suppose MACH_IS_FOO is not a great way to encode that requirement. But
> it don't think it has to be self-documenting. It does have to be
> consistent with the conditionals in head.S.

Alternatively, check if early_console_instance.write() lies within
__init_begin..init_end?
That doesn't work with the current setup, as .write() always points
to the common debug_cons_write().  But it can be made to work
by setting up .write() with the proper function pointer in
setup_early_printk(), getting rid of repeated MACH_IS_*() checks in
the process.

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