Re: Linux kernel / USB keyboard scancode problems

[Frank Murphy <murphyf+linux-ppc@f-m.fm>]

On Saturday 30 August 2003 5:20, Gabriel Paubert wrote:
> On Fri, Aug 29, 2003 at 05:28:01PM +0200, Frank Murphy wrote:
> > Thanks for these. Because they're such a mess, perhaps the codes for F13,
> > F14, and F15 should be used instead. Now just to find out what they are,
> > and if all Apple keyboards have these keys as F13, 14, & 15.
>
> From include/linux/input.h (latest Linus'bk repository, lines 195-200):
>
> #define KEY_F13                 85
> #define KEY_102ND               86
> #define KEY_F11                 87
> #define KEY_F12                 88
> #define KEY_F14                 89
> #define KEY_F15                 90
>
> which are mapped to 85/115/119 for emulated raw scancodes
> in drivers/char/keyboard.c, or rather 0x55/0x73/0x77 since
> scancodes ar normally typed in hex.
>
> Of course the problem is that X wants these in 118/119/120.  I've just
> checked out the X CVS tree and my head hurts. I believe that the relevant
> code is in:  programs/Xserver/hw/xfree86/common/xf86Events.c.
>
> But it's so hard to follow :-(
>
> Anyway, if I follow the X code correctly (a big if), 0x55 and 0x77
> won't be remapped, but 0x73 will be, to 0xCB. After that we add 8
> to compute the X keycode which become: 93, 211, and 127.
>
> With some luck, there may be a solution since none of these key
> codes are used under X, neither on the Mac nor on the PC.

Yeah, the keycode/scancode/keysym stuff is a twisty maze. I do have the
hardware to test these changes, if you have a guess to what they should be.

> > > > > The following might be sufficient (just a wild guess, and it will
> > > > > certainly break if you don't use linux keycodes):
> > > >
> > > > Thanks for your guessed fix. I'll have to look at what you've done,
> > > > but, wow, it's pretty cryptic.
> > >
> > > Well, it is really simple. Replace 101 by 127 in the ADB to Linux
> > > keycode mappings. In this way the translation back to "raw" keycodes
> > > (only cooked twice!) gives 349, which is 256 + 93 (0xe0 0x5d).
> >
> > Well, I guess the trick is to know which one to change. :)
>
> Did you actually try the patch. If yes, on which kernel?

I have, and it seems to work fine for me. I patched the debian source from
kernel-source-2.4.21. Other keys seem to work fine, but I havent checked
every single key, but I'm using it succesfully now.

Frank


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/