Discussion:
Bug#340508: missing modules on s390/s390x
(too old to reply)
maximilian attems
2006-03-15 15:50:05 UTC
Permalink
waldi added s390 support to initramfs-tools as of version 0.54.
0.55 is in incoming.debian.org and should be available in unstable soon.
klibc-1.2.4-1 has an important s390 fix.

so testing those releases are highly welcome.
you can rebuild your initramfs after installation of aboves packages with
update-initramfs -u -k <kernel-version>
--
maks
--
To UNSUBSCRIBE, email to debian-bugs-dist-***@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact ***@lists.debian.org
Ivan Warren
2006-03-16 16:00:18 UTC
Permalink
Only problem was that the network interface did not come up. Trying to
# ifup eth0
Ignoring unknown interface eth0=eth0.
This looks like an issue related to udev persistent device naming.
# lsmod
Module Size Used by
ctc 79648 0
fsm 12544 1 ctc
cu3088 12292 1 ctc
ccwgroup 15104 1 cu3088
dcssblk 20504 0
dasd_fba_mod 17408 0
dasd_eckd_mod 109824 6
dasd_mod 79924 8 dasd_fba_mod,dasd_eckd_mod
- Network config is a basic static configuration; from
auto ctc0
iface ctc0 inet static
address 10.19.92.1
[...]
... For Info ....

I had to add a file such as this to /etc/network/if-pre-up.d :

ctc0 :
**********
#!/bin/sh
exec > /dev/null 2>&1
echo "0.0.0400,0.0.0401" > /sys/bus/ccwgroup/drivers/ctc/group
echo "1" > /sys/bus/ccwgroup/devices/0.0.0400/online
**********

(it's a dual CTC link at CCUU 0400 & 0401)

Apparently this needs to be done so as to indicate the grouping
(CCUU/CCUU+1 was probably implicit in 2.4).

I had to do this some time ago when I switched to 2.6 and I think it was
Bastian who told me this.

Note that for the 1st time, since the update to klibc 1.2.4-1, I've been
able to complete a 2.6.15 kernel IPL (I suspect the issue was a problem
with pivot_root). Now also I had to change the parm line and /etc/fstab
due to the change in the DASD naming convention from the devfs
(/dev/dasd/0.LCSS.CCUU/partn) to the udev naming convention (/dev/dasdXY))

--Ivan
--
To UNSUBSCRIBE, email to debian-bugs-dist-***@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact ***@lists.debian.org
Hugo Monteiro
2006-03-16 18:30:22 UTC
Permalink
Post by Ivan Warren
Only problem was that the network interface did not come up. Trying to
# ifup eth0
Ignoring unknown interface eth0=eth0.
This looks like an issue related to udev persistent device naming.
# lsmod
Module Size Used by
ctc 79648 0
fsm 12544 1 ctc
cu3088 12292 1 ctc
ccwgroup 15104 1 cu3088
dcssblk 20504 0
dasd_fba_mod 17408 0
dasd_eckd_mod 109824 6
dasd_mod 79924 8 dasd_fba_mod,dasd_eckd_mod
- Network config is a basic static configuration; from
auto ctc0
iface ctc0 inet static
address 10.19.92.1
[...]
... For Info ....
**********
#!/bin/sh
exec > /dev/null 2>&1
echo "0.0.0400,0.0.0401" > /sys/bus/ccwgroup/drivers/ctc/group
echo "1" > /sys/bus/ccwgroup/devices/0.0.0400/online
**********
(it's a dual CTC link at CCUU 0400 & 0401)
Apparently this needs to be done so as to indicate the grouping
(CCUU/CCUU+1 was probably implicit in 2.4).
I had to do this some time ago when I switched to 2.6 and I think it was
Bastian who told me this.
Note that for the 1st time, since the update to klibc 1.2.4-1, I've been
able to complete a 2.6.15 kernel IPL (I suspect the issue was a problem
with pivot_root). Now also I had to change the parm line and /etc/fstab
due to the change in the DASD naming convention from the devfs
(/dev/dasd/0.LCSS.CCUU/partn) to the udev naming convention (/dev/dasdXY))
--Ivan
I used the pre-up directive in /etc/network/interfaces as follows:

iface eth0 inet static
address 10.10.10.10
netmask 255.255.255.0
network 10.10.10.0
broadcast 10.10.10.255
gateway 10.10.10.1
pre-up echo 0.0.0700,0.0.0701,0.0.0702
Post by Ivan Warren
/sys/bus/ccwgroup/drivers/qeth/group
pre-up echo 1 > /sys/devices/qeth/0.0.0700/online


Hugo Monteiro.
--
ci.fct.unl.pt:~# cat .signature

Hugo Monteiro
Email : ***@fct.unl.pt
Telefone : +351 212948300 Ext.15307

Centro de Informática
Faculdade de Ciências e Tecnologia da
Universidade Nova de Lisboa
Quinta da Torre 2829-516 Caparica Portugal
Telefone: +351 212948596 Fax: +351 212948548
www.ci.fct.unl.pt ***@fct.unl.pt

ci.fct.unl.pt:~# _
--
To UNSUBSCRIBE, email to debian-bugs-dist-***@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact ***@lists.debian.org
Frans Pop
2006-03-16 16:00:27 UTC
Permalink
Post by maximilian attems
waldi added s390 support to initramfs-tools as of version 0.54.
0.55 is in incoming.debian.org and should be available in unstable
soon. klibc-1.2.4-1 has an important s390 fix.
so testing those releases are highly welcome.
you can rebuild your initramfs after installation of aboves packages
with update-initramfs -u -k <kernel-version>
I've just upgraded to latest versions and see the following (harmless)
error during upgrade:
Setting up udev (0.087-2) ...
Installing new version of config file /etc/udev/persistent.rules ...
ln: creating symbolic link `/tmp/mkinitramfs_lsxTkx//etc/modprobe.d/dasd'
to `/tmp/initramfs_dasd': File exists

Setting up initramfs-tools (0.55b) ...
ln: creating symbolic link `/tmp/mkinitramfs_kRtv8k//etc/modprobe.d/dasd'
to `/tmp/initramfs_dasd': File exists

Also, is it really necessary to generate the initrd two times? Hope this
will be resolved in the future...


I also noticed that after the installation a file /tmp/initramfs_dasd is
left on the system. It would probably be a good idea to use a real
tempfile for this (and to clean up after it is no longer needed).


After all this I modified my bootloader config (/etc/zipl.conf) to have:
[debian_26]
target = /boot
image = /boot/vmlinuz-2.6.15-1-s390
ramdisk = /boot/initrd.img-2.6.15-1-s390
parameters = ro vmpoff="LOGOFF" root=/dev/dasda1

(Note: I removed the dasd parameter from the last line.)

The system rebooted OK using the 2.6 kernel: disks (both root and other
partitions) were mounted OK :-)

Only problem was that the network interface did not come up. Trying to
manually start it resulted in:
# ifup eth0
Ignoring unknown interface eth0=eth0.

This looks like an issue related to udev persistent device naming.

Further info:
- The ctc module was loaded:
# lsmod
Module Size Used by
ctc 79648 0
fsm 12544 1 ctc
cu3088 12292 1 ctc
ccwgroup 15104 1 cu3088
dcssblk 20504 0
dasd_fba_mod 17408 0
dasd_eckd_mod 109824 6
dasd_mod 79924 8 dasd_fba_mod,dasd_eckd_mod

- Network config is a basic static configuration; from
/etc/network/interfaces:
auto ctc0
iface ctc0 inet static
address 10.19.92.1
[...]

Thanks for getting us this far. I intend to set up a branch for d-i to try
for the switch to 2.6 for the installer (unless Bastian prefers to work
on that; I will probably need his help anyway).

Cheers,
FJP
Bastian Blank
2006-03-16 17:10:27 UTC
Permalink
Only problem was that the network interface did not come up. Trying to
You need a 2.6.16-rc5 kernel and sysconfig-hardware.

The new kernel properly loads the network modules, the later is used to
configure the hardware.

Bastian
--
You're dead, Jim.
-- McCoy, "Amok Time", stardate 3372.7
Frans Pop
2006-03-16 18:30:16 UTC
Permalink
Thanks Bastian.
Post by Bastian Blank
Only problem was that the network interface did not come up. Trying
You need a 2.6.16-rc5 kernel and sysconfig-hardware.
Yes, I remembered sysconfig-hardware and already installed it.
Post by Bastian Blank
The new kernel properly loads the network modules, the later is used to
configure the hardware.
I see that the experimental kernel image does not depend on
sysconfig-hardware. Should it (to ensure that users upgrading from Sarge
will have it installed)?
Bastian Blank
2006-03-16 18:40:17 UTC
Permalink
Post by Frans Pop
I see that the experimental kernel image does not depend on
sysconfig-hardware. Should it (to ensure that users upgrading from Sarge
will have it installed)?
Hmm. It is no dependency as the kernel can work without. Also it needs a
converter for chandev to the new config format.

Bastian
--
The heart is not a logical organ.
-- Dr. Janet Wallace, "The Deadly Years", stardate 3479.4
Bastian Blank
2006-03-16 22:00:07 UTC
Permalink
Post by Bastian Blank
Hmm. It is no dependency as the kernel can work without. Also it needs a
converter for chandev to the new config format.
Which will only work automaticaly in the easiest config: only one
device/type as there is not longer any order. And ifupdown can't
handle device-ids instead of names.

Bastian
--
Lots of people drink from the wrong bottle sometimes.
-- Edith Keeler, "The City on the Edge of Forever",
stardate unknown
Loading...