Re: 2.6 input drivers FAQ

[Vojtech Pavlik]

On Sun, Feb 01, 2004 at 04:14:52PM +0100, Andries Brouwer wrote:

> On Sun, Feb 01, 2004 at 02:50:01PM +0100, Vojtech Pavlik wrote:
> 
> > Sorry. I was typing that from memory. I'll fix it. Btw, could you make
> > the kbd package accept scancodes in the 0x80-0xff range (same as e000 to
> > e07f), if it is not yet there? And how about scancodes in the
> > 0x100-0x1ff range? Will those work?
> 
> What is needed for the 2.6 kernel is rather volatile
> (and maybe the present kernel version is not quite final yet).
> But setkeycodes does ioctl(fd,KDSETKEYCODE,&a) where 
> 
>                 a.keycode = atoi(argv[2]);
>                 a.scancode = strtol(argv[1], &ep, 16);
>                 if (a.scancode >= 0xe000) {
>                         a.scancode -= 0xe000;
>                         a.scancode += 128;      /* some kernels needed +256 */
>                 }

I think this is fine. It keeps backwards compatibility with
people's old scripts and still it allows arbitrary scancodes.

> The 2.6.1 kernel is still very messy, with code
> 
>         if (atkbd->emul) {
>                 if (--atkbd->emul)
>                         goto out;
>                 code |= (atkbd->set != 3) ? 0x80 : 0x100;
>         }
> 
> where the representation of the e0 prefix depends on the current scancode mode.
> A bad idea.

Well, I think it's a bad idea to have the userspace tool know about the
e0 thing at all. It should be just opaque numbers to it.

I agree that the above code choosing between 0x80 and 0x100 is not very
nice, but it has two reasons:

	0) e0+(00..7f) is the most common code for extra keys in set2,
	   while (80..df) is the most common code for extra keys in set3
	
	1) it will work in old setkeycodes - although the values you'll
           have to specify with set3 are rather nonsensical, you still
	   will be able to set extra keys on a set3 keyboard.

	2) the scancode tables are much more compact.

I don't have a problem with swapping the set3 table, if setkeycodes
works reasonably now for scancodes above 128.

-- 
Vojtech Pavlik
SuSE Labs, SuSE CR
-
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/