Re: Keyboard lockup with NetBSD5

[Alexander Coers <Alexander.Coers%gmx.de@localhost>]

Hello Frank,


Am 27.10.2009 um 22:28 schrieb Frank Wille:

> Alexander Coers wrote:
>
> > > Unfortunately you cannot use "vmstat -i", but I would bet that you
> > > have more level 2 interrupts than just from the keyboard.
> >
> > Since this kernel heals itself, I will try to install a full NetBSD5
> > system.
> > Maybe then I can use vmstat...
>
> I'm not sure. Maybe you need a NetBSD-current (5.99) userland.
>
> I could upload the files, if you want. This would be a good basis for
> further test-kernels.

Please do so.

> > > What you and Chris Edwards, who reported the problem before, have in
> > > common
> > > is a PCMCIA card plugged in. Does such a card generate level 6 or
> > > level 2
> > > interupts? Did you already try without it? Or is there any other
> > > additional
> > > interrupt source (SCSI? Should be level 6, though).
> >
> > There is no difference between network card being plugged in or not.
>
> Hm. But why are there some A1200-owners without this problem, like  
> John
> Klos? What is the difference?

Different hardware (board) revisions?
I have the PCMCIA reset fix applied to Gayle. Maybe this is a problem?
I also noticed that my card is not reseted between reboots, I have to  
unplug it before the kernel boots.

> > How can I check what IRQ level is used?
>
> > From the source (gayle_pcmcia.c) I would guess that IRQ level 6 is  
> > used for
> inserting/removing a card and IRQ level 2 for data.
>
>
> > I will test whatever you throw at me....
>
> A few days ago I checked the differences between the source of the  
> last
> working kernel (4.0 or 4.0.1?) and 5.0. In the Amiga-specific source  
> there
> are *no* differences at all!
>
> The kern source is used by A3000s and A4000s as well. So maybe there  
> are
> changes in the PCMCIA code causing this behaviour?
>
> For another test I compiled a minimal kernel with just a few devices  
> to make
> the system start. There is no PCMCIA code in this kernel. Please  
> have a look
> if it makes any difference:
>  ftp://ftp.netbsd.org/pub/NetBSD/misc/phx/kbdtest1.tar.gz

With this kernel the system stops right after /sbin/initd (initd died)  
in kernel debugger land.
I can enter exactly one char before the keyboard locks up.