Bug#416208: Installation report HP-715/64 (HIL)
Frans Pop wrote:
> On Thursday 25 December 2008, Helge Deller wrote:
>>> - why doesn't the kernel/udev autoload this module?
>> Probably yes, but as we discussed last time, this needs to be
>> implemented. I already looked into implementing it, but sadly it isn't
>> that easy to understand all the flow between kernel and userspace and
>> how to handle it then.
> Ah, so it's on the parisc bus again. I've seen the patch you sent upstream
> for that. Cool. Please keep me (or the d-hppa list) informed of its
> progress so we can remove the manual loading of these modules in D-I for
Sure, I'll let you know.
My goal is to get module autoloading now into 2.6.29, so for sqeeze the
manual module loading can be removed entirely.
Fortunately no userspace modifications for udev or similiar are needed,
only patches to Linux kernel. This makes it easier for me.
>> My proposal would be to just add the modprobe for hilkbd there blindly
>> as well.
> I've committed a patch for rootskel for that.
>> On the other side, if you would add this modprobe I would try the lenny
>> installer at once on 4 very different machines to rule out any problems
>> due to the modprobe:
>> - 715/64 with HIL only
>> - B160L, no HIL, PS/2 keyboard/mouse
>> - C3000 (32 and 64bit kernel), no HIL, USB keyboard/mouse only
>> - Tadpole parisc laptop, PS/2 keyboard/mouse only
> Please test this image:
It works like a charm :-)
I've used this image via serial console on all machines mentioned above.
The hilkbd module was loaded on every machine, and all of them worked
nicely. Even if the module was loaded, it did not break anything.
Then I booted this image on the 715/64 (the only machine I have with
HIL) the classical way with HIL keyboard/mouse and monitor. The module
was loaded and I could for the first time ever use a D-I image to
install the machine. HIL keyboard was working OK.
> I've tested it on my box (which is headless and does not need/use any of
> the modules we currently load manually) and it seems to work fine.
> Let me know ASAP and we may just be able to squeeze it in for Lenny.
Yes, please. This would be great!
> P.S. hilkbd is a rather nasty module: when I tried to rmmod it after a
> test install, it froze both the SSH session and the serial console
> connection I had open to the box...
I see the same here. Removing the hilkbd module on a machine without HIL
panics the kernel. I've queued up a patch for that now and send it to
linux-input mailing list soon as well. If you think it makes sense to
fix this for lenny-d-i as well, just let me know.