Re: question about ruby's keyboard.c & keybdev.c

[Franz S. <Fra...@la...>]

On Samstag, 6. Juli 2002 23:59, James Simmons wrote:
> > > > for maintenance reasons. keybdev.c provides bridge code between input
> > > > layer and old-style keyboard.c. In ruby this is no longer necessary
> > > > as the new-style keyboard.c registers itself directly to the input
> > > > layer.
>
> You still need keybdev.c with 2.5.25. It will be a few weeks before we
> will migrate the console system to the input api.

No problem. The main problem isn't keybdev.c but pc_keyb.c. On PPC I just 
disabled most of pc_keyb.c by adding this to asm/keyboard.h:

static inline void kbd_init_hw(void)
{
}

and removing the corresponding #define. Then atkbd.c and i8042.c can be used 
without problems.

> > > That's right, I'm working on the merge right now.
> >
> > And now part#1 is in 2.5.25!! Yaayyy! Thanks Vojtech!
>
> Just a few more steps to go. I submited the console patches to Linus. Now
> to wait for them to go in.

Well, what are a few weeks more, I'm waiting for 2 years now for this code to 
go in.

> > While preparing the PPC part I noticed one thing though, how do you
> > intend to handle the platforms (2 PPC embedded boards and MIPS AFAICS)
> > that redefined
> > kbd_read_input/kbd_read_status/kbd_write_output/kbd_write_command so far
> > (using ioremap/readb/writeb and sometimes special access sequences)?
>
> Why we break them. One thing I learned the hard way with the fbdev layer
> is if you want to change the api you need to change the upper layer first
> and break every driver. Its the only way it will work.

No complaint here, just searching for a solution :-).

> > Do you want #ifdefs in i8042.[ch]? Or separated includes and/or drivers
> > like i8042-ppc.[ch]?
> >
> :-(

Exactly. i8042.c has quite a few assumptions in it. It uses inb/out instead of 
ioremap/readb/writeb, it assumes control and data register are 4 addresses 
apart and it assumes no special sequences are necessary to access these 
registers.

> > One more thing, CONFIG_SERIO_I8042 currently depends on CONFIG_ISA, does
> > this really make sense?
>
> ??

On PPC quite a lot of embedded boards exist that have PS2 HW which is remapped 
to the standard IO addresses and thus i8042.c is useable out of the box. 
Nevertheless none of these boards have ISA slots (shudder!) and thus 
CONFIG_ISA isn't defined, preventing enabling of CONFIG_SERIO_I8042.

Franz.