hurd installation feedback and RE: new tarfile
hello,
i want to give feedback for my first hurd installation and problems i still
have not solved.
first, i mirrored
alpha.gnu.org/gnu/hurd/contrib/marcus
alpha.gnu.org/gnu/hurd/debian
ftp.debian.org/pub/debian/dists/potato/main/binary-all/base
ftp.debian.org/pub/debian/dists/sid/main/binary-hurd-i386
ftp.debian.org/pub/debian/dists/slink/main/binary-all/base
the installation scripts from marcus were great. and together with
the "easy giode to installing Hurd on a Linux box" it was not
a great problem to get the system run.
i cd'd to ftp.debian.org/pub/debian/dists/sid/main/binary-hurd-i386/base
and started the script. then needed to link
../xxx/{inetutils_1.3.2-8.deb,perl-5.005_5.005.03-4.deb} to
this directory, because the install script searched for it.
some of the base-all symlinks pointed to nowhere, that's why i needed
to get ftp.debian.org/pub/debian/dists/{potato,slink}/main/binary-all/base.
then i booted the system from marcus' boot-19990907.disk.
i used seperate hard disk on the secondary hd-controller (/dev/hdc under
linux).
to get the system booted was a pain. the name of the device changes
during the boot process. but i cannot be more precise because i do not
remember exactely. ..ah, yes (mounted the boot floppy):
hd1 and hd2s1 refer to the same hard disk:
root= (hd1,0)
kernel= /boot/gnumach.gz root=hd2s1
problems:
it would be useful to have more/less on the system after install,
because reading the settrans parameters (an important new utility)
from settrans --help is more than 24 lines.
df does not find the correct partitions.
i found with strings(1) [what i missed too, because it's great for
debuging] that it looks for /var/run/mtab, which was a null file.
after
echo /dev/hd2s1 / ext2 rw 0 0 >/var/run/mtab
df worked fine.
df ond no other tool i found showed the actual settrans actual state of the
system.
i have definded a swap partition (/dev/hd2s2 $(add-linux-paging-file) in
/boot/servers.boot) but i could not check if it is used. there is no free
command (maybe it's not active, because i have this -unmodified- file on my
boot disk, too?).
the free(1) command is not available.
/proc may be used (the directory is present, and i found /hurd/proc), and
i tried to "mount" it via entry in fstab and settrans, but i was not
successful.
i had a chance to mount (settrans) my linux partiton as described in the
installation guide: hda1. but hda5 (logical linux partition
in hda3) caused the system to freeze/hang (successful settrans, but i could
not read anything).
network is not functional.
my card is (from the linux kernel dmesg):
eth0: RealTek RTL8139 Fast Ethernet at 0x6100, IRQ 10, 00:00:e8:5e:85:b6.
if i remember correctly, the gnumach kernel dectets this card, too. it
also detects the MAC address.
but the settrans for the network has no effect. debuging it is quite
difficult, because the system lacks of route/netstat/ifconfig/..
ping complains that it can't allocate enough memory.
ping from a lan-host to the hurd-box: no response.
then i tried to set up a loopback device, but ping to localhost
still complained the memory failure.
inetd hasn't started, because it searched for files in /etc/inetd.conf
wich are not installed.
the keymaps in marcus' keymap.tar.gz worked fine.
screen (important for a one-console-system :) is still missing.
now, i have installed these further packages:
apt_0.1.9.deb less_340-5.deb nvi_1.79-8.deb
keymap.tar.gz man-db_2.3.10-69s.deb
strings and netcat would be great tools for debuging.
the presence of telnet/telnetd after a minimal install would be fine until
ssh will be ported.
man(1)'s default path to the man pages is wrong. needed to set the
environment variable MANPATH.
and now, some comments to stefan weil's experiences:
> 1. Installation.
[..]
> I tried the installation on two different PCs. With one, the
> keyboard problem occured (shell sees only every second character)
> as you write in the installation script.
i had that bash failure, too; i tried to get a clean shutdown,
but after some tries, the system went in a deep froze. perhaps a memory
failure in the kernel?
> Great: reboot unmounts the root filesystem cleanly, so fsck
> is no longer needed after a reboot.
for me, reboot leads the root-filesystem in an unchecked state.
> /root/.profile calls mesg which is not in the tar file.
> This is not a big problem, but it might be fixed some day?
i have seen it, too, and i wondered where it came from ;)
> 3. Strange things.
> One of my PCs has 128 MB RAM and runs VMWARE with adjustable RAM
> for the virtual machine which runs The Hurd. Hurd boots with 32 MB RAM,
> but not with 48 MB or 64 MB. The boot failed sometimes with 92 MB.
> When it fails, the error message is "bootstrap/default-pager panic: openi
> 5000".
my maschine has 48MB of RAM. no problems at all.
cheers,
- thomas <thomas@x-berg.in-berlin.de>
Reply to: