atkbd_init lockup with 2.6.11-rc1

[Dmitry Torokhov]

On Fri, 28 Jan 2005 15:55:11 +0100, Olaf Hering <olh at suse.de> wrote:
> On Fri, Jan 28, Dmitry Torokhov wrote:
> 
> > On Fri, 28 Jan 2005 14:58:27 +0100, Olaf Hering <olh at suse.de> wrote:
> > > On Fri, Jan 28, Olaf Hering wrote:
> > >
> > > >
> > > > My IBM RS/6000 B50 locks up with 2.6.11rc1, it dies in atkbd_init():
> > >
> > > It fails also on PReP, not only on CHRP. 2.6.10 looks like this:
> > >
> > > Calling initcall 0xc03bc430: atkbd_init+0x0/0x2c()
> > > atkbd.c: keyboard reset failed on isa0060/serio1
> > > atkbd.c: keyboard reset failed on isa0060/serio0
> > >
> >
> > So it could not reset it even before, but it was not getting stuch
> > tough... What about passing atkbd.reset=0?
> 
> I will try that.
> Adding a printk after the outb() fixes it as well.

Fixes as in "it reports that reset fails" again or it resets the
keyboard cleanly and works fine?

> Do you have a version of that i8042 delay patch for 2.6.11-rc2-bk6?
> Maybe it will help.
> 

No I don't, and I don't think you need all of it. What happens if you
edit drivers/input/serio/i8042.c manually and stick udelay(7); in
front of calls to i8042_write_data() in i8042_kbd_write() and
i8042_aux_write()?

-- 
Dmitry