Re: New Install
On Sun, Sep 17, 2000 at 01:18:44PM -0500, Neal H Walfield wrote:
> > Well, I had a first go at installing The Hurd Friday night. It appeared
> > to go well enough. Trying to boot it for native-install, grub found
> > the root partition, gnumach, and serverboot (BTW, using grub to "find
> > /boot/gnumach.gz" is a good way to figure out which drive and partition
> > to use - this could go in the instructions I think), however, on running
>
> That is a good idea.
>
> > boot, the kernel couldn't find the root partition (yes, I used root=hd0s2
> > arg to kernel). It turns out HPT366 ATA66 chipset can be handled by grub,
> > but not by gnumach.
>
> That is to say by the bios and not gnumach; All i/o that grub does is via
> the BIOS functions.
ACK.
>
> > Next round, moved disk to an onboard IDE controller (and sacrificed ATA66
> > access speed, and reworked linux fstab, yadayada, so I could dual boot),
> > and got gnumach to boot (grub's really cool BTW). Now, I can't get the
> > network card working. I've been following the new instructions up till
> > now, but they petered out at this point. From discussions on this list
>
> Where exactly? And in what regard?
Network setup, links to troubleshooting info. Oops. I see the
network setup info is there. I must have been distracted by the disk
and other stuff. It was fine otherwise. I used it because I saw a
request on the list (from you?) in the last week or so for someone to
see if the instructions were good enough to be used without having to
refer elsewhere and I've been meaning to get around to installing it
for awhile, so.... Until I ran into problems, they were good enough.
>
> > and from the "Easy Guide", I knew how to do the settrans (settrans -fg
> > /servers/socket/2 /hurd/pfinet --interface=eth0 --address=192.168.56.12
> > --netmask=255.255.255.0) and I could ping localhost and 192.168.56.12
> > afterwards. However, I can't get to any other host on 192.168.56.x.
> > I see from the boot messages that gnumach is detecting the card
> > with the via_rhine driver which is what it uses in linux. Ping to a
> > non-local address hangs the system if I don't background it. Settrans on
> > /servers/socket/2 also hangs it after an unsucessfull ping (which I did
> > a few times playing with args to /hurd/pfinet), so I had to power-down
> > and fsck the filesystem (from linux). Did this a few times, and decided
>
> Try screen
I don't see it in /bin or /sbin. What is it? All I see is a terminfo
entry. I guess I'm missing something or don't understand.
>
> > to reinstall from a clean partition. Watching for errors during the
> > native-install is really hard since they go by so fast and I couldn't
> > find a way to capture them, but I'm pretty sure there were none.
>
> Try redirecting them to a file.
Tried "./native-install 2>&1 | tee native-install-pass1.log" - got an
error from a translator I think. Apparently pipes aren't available this
early. Couldn't use less for the same reason. Tried "./native-install >
native-install-pass1.log 2>&1". Hung. I realized on the next pass that
I forgot about the timezone prompts. I think those are the only prompts
for input so I could do them blind.
This doesn't sound right - I may be remembering this wrong. I had no
trouble logging the second pass which had no errors except while setting
up tcpd: "/dev/fd/5: ALL:: command not found" which I ignored since
/dev/fd seemed irrelevant and /var/lib/dpkg/status has tcpd marked as
"Status: install ok installed" so it looked like an ignorable error.
It was the first pass that I had trouble logging. /me goes to check
where libc is set up - yes, it's in the first pass so I guess that's
where the prompts are so commenting them out probably won't work unless
I want to fiddle with the libc postinst. BTW, I was using the Debian
packages to install instead of the tarball method.
>
> > Tried again from clean install, still no network. The card BTW is a
> > D-Link DFE-530-TX (a 10/100 PCI card). Went back to linux and looked at
> > some files in /proc for a clue and found that linux assigns interrupt
> > 5 to both the network card and the the USB controller. I had a device
> > plugged into the USB port, so I unplugged it. The USB controller
> > itself is onboard so there isn't much I can do to actually disable it.
> > Maybe interrupts are assigned differently by gnumach. Anyway, it still
> > doesn't work.
>
> Shared irqs do not work at all. Maybe you can disable it in the bios?
> Or change the settings on the NIC card?
I don't remember seeing any USB or NIC setup in the BIOS. I'll check
again. The NIC has a utility (in Windows, ugh.) to change its IRQ.
I don't know if the change will persist, but that gives me a couple of
things to try.
>
> > Without ppp, it's going to be somewhat of a PITA to install anything
> > else past the base system without a network. Any clues for something
> > else to try. I've looked at the gnumach source a bit and the via_rhine
> > driver looks like one from the linux source tree, fairly old but otherwise
> > pretty much unmodified from the linux version. Could I just drop in
> > the driver file from a newer linux source tree and recompile gnumach?
> > Any other ideas? OS-Kit?
>
> See about. As for OS-Kit, it is worth a shot, however, it is much less
> tested.
OK, I have some other leads than OS-Kit to pursue when I have some more
time. I'll keep working with gnumach for now. Do you think rebuilding
with the newer via_rhine driver is worth the effort to try?
>
> > PS, I almost forgot, documentation feedback: the New Instructions need
> > network setup information. I can't remember where I found links to
> > hardware compatibility information, but both guides should have links
> > to it.
>
> Will take under advice, thanks.
OK, well I already retracted the network setup comment. For a normal
install your doc has quite enough info. Links to troubleshooting or
compatibility pages would be good, but they're just links. The content
is fine. /me scratches head for sec... I just remembered one thing.
The partitioning section addresses everything except that it doesn't
say how to type the partition. I made it type 83 (linux) based on
information I found elsewhere.
All in all, I think the interrupt business is where the problem lies.
The behavior is just like what I'd expect if I knew there were interrupt
problems.
Thanks much,
Steve
--
Steve Bowman <sbowman@frostwork.net> (preferred)
Buckeye, AZ <sbowman@goodnet.com> <bowmanc@acm.org>
<http://www.goodnet.com/~sbowman/>
Powered by Debian GNU/Linux <http://www.debian.org>
Reply to: