Lenny installer hanging?

[Adam Thornton]

I don't seem to be able to boot (under z/VM 5.2 on a Flex-ES machine)
the Lenny RC1 installer.

Sometimes the session hangs just after memory detection. Sometimes I
get to device detection.

Here's the spooled console:

Ready; T=0.01/0.01 15:21:14
access 19f d
Ready; T=0.01/0.01 15:21:23
listf * * d
DEBIAN EXEC D1
INITRD DEBIAN D1
KERNEL DEBIAN D1
PARMFILE DEBIAN D1
Ready; T=0.01/0.01 15:21:32
debian
0000003 FILES PURGED
RDR FILE 0009 SENT FROM LENNY PUN WAS 0009 RECS 042K CPY 001 A
NOHOLD NOKEEP

RDR FILE 0010 SENT FROM LENNY PUN WAS 0010 RECS 0001 CPY 001 A
NOHOLD NOKEEP

RDR FILE 0011 SENT FROM LENNY PUN WAS 0011 RECS 035K CPY 001 A
NOHOLD NOKEEP

0000003 FILES CHANGED
0000003 FILES CHANGED
Ö 0.000000Ü Initializing cgroup subsys cpuset
Ö 0.000000Ü Initializing cgroup subsys cpu
Ö 0.000000Ü Linux version 2.6.26-1-s390 (Debian 2.6.26-8) (***@debian.org
)
(gcc version 4.1.3 20080623 (prerelease) (Debian 4.1
.2-23)) #1 SMP Thu Oct 9 14:15:49 UTC 2008
Ö 0.000000Ü We are running under VM (31 bit mode)
Ö 0.000000Ü This machine has an IEEE fpu
Ö 0.000000Ü Zone PFN ranges:
Ö 0.000000Ü Normal 0 -> 32768
Ö 0.000000Ü Movable zone start PFN for each node
Ö 0.000000Ü early_node_mapÖ1Ü active PFN ranges
Ö 0.000000Ü 0: 0 -> 32767
Ö 0.000000Ü Built 1 zonelists in Zone order, mobility grouping on.
Total pag
es: 32511
Ö 0.000000Ü Kernel command line: ro locale=C
Ö 0.000000Ü
Ö 0.000000Ü PID hash table entries: 512 (order: 9, 2048 bytes)
Ö17179569.187036Ü console ÖttyS0Ü enabled
Ö17179569.188091Ü Dentry cache hash table entries: 16384 (order: 4,
65536 bytes)

Ö17179568.189257Ü Inode-cache hash table entries: 8192 (order: 3,
32768 bytes)
Ö17179568.198595Ü Memory: 122536k/131072k available (2270k kernel
code, 0k reser
ved, 842k data, 148k init)

===NOTE FROM ADAM: this hung for a long time, so I restarted it and
upped the storage from 128 to 512M===

CP IPL CMS
z/VM V5.2.0 2008-02-12 13:23

DMSACR1240E You are not authorized to connect to file pool USER01
Ready; T=0.03/0.05 15:24:15
def stor 512M
STORAGE = 512M
Storage cleared - system reset.
IPL CMS
z/VM V5.2.0 2008-02-12 13:23

DMSACR1240E You are not authorized to connect to file pool USER01
Ready; T=0.03/0.04 15:24:26
acc 19f D
Ready; T=0.01/0.01 15:24:37
Ready; T=0.01/0.01 15:26:34
debian
0000003 FILES PURGED
RDR FILE 0012 SENT FROM LENNY PUN WAS 0012 RECS 042K CPY 001 A
NOHOLD NOKEEP

RDR FILE 0013 SENT FROM LENNY PUN WAS 0013 RECS 0001 CPY 001 A
NOHOLD NOKEEP

RDR FILE 0014 SENT FROM LENNY PUN WAS 0014 RECS 035K CPY 001 A
NOHOLD NOKEEP

0000003 FILES CHANGED
0000003 FILES CHANGED
Ö 0.000000Ü Initializing cgroup subsys cpuset
Ö 0.000000Ü Initializing cgroup subsys cpu
Ö 0.000000Ü Linux version 2.6.26-1-s390 (Debian 2.6.26-8) (***@debian.org
)
(gcc version 4.1.3 20080623 (prerelease) (Debian 4.1
.2-23)) #1 SMP Thu Oct 9 14:15:49 UTC 2008
Ö 0.000000Ü We are running under VM (31 bit mode)
Ö 0.000000Ü This machine has an IEEE fpu
Ö 0.000000Ü Zone PFN ranges:
Ö 0.000000Ü Normal 0 -> 131072
Ö 0.000000Ü Movable zone start PFN for each node
Ö 0.000000Ü early_node_mapÖ1Ü active PFN ranges
Ö 0.000000Ü 0: 0 -> 131071
Ö 0.000000Ü Built 1 zonelists in Zone order, mobility grouping on.
Total pag
es: 130047
Ö 0.000000Ü Kernel command line: ro locale=C
Ö 0.000000Ü
Ö 0.000000Ü PID hash table entries: 2048 (order: 11, 8192 bytes)
Ö17179568.189070Ü console ÖttyS0Ü enabled
Ö17179568.197213Ü Dentry cache hash table entries: 65536 (order: 6,
262144 bytes
)
Ö17179569.214577Ü Inode-cache hash table entries: 32768 (order: 5,
131072 bytes)

Ö17179569.247170Ü Memory: 511872k/524288k available (2270k kernel
code, 0k reser
ved, 842k data, 148k init)

Ö17179568.251718Ü Write protected kernel read-only data: 0x12000 -
0x2edfff
Ö17179568.254542Ü Security Framework initialized
Ö17179568.254634Ü SELinux: Disabled at boot.
Ö17179569.254734Ü Capability LSM initialized
Ö17179569.256117Ü Mount-cache hash table entries: 512
Ö17179568.262898Ü Initializing cgroup subsys ns
Ö17179569.263033Ü Initializing cgroup subsys cpuacct
Ö17179568.263242Ü Initializing cgroup subsys devices

CP SPOOL CONS CLOSE


(that's where I gave up waiting and sent the console back to my CMS
userid....the guest is still there but I have little hope that it will
continue)

In addition to the hang, the control characters, which doubtless look
fine on a VT220 or ANSI terminal, are nasty on a 3270. This would
seem to be a kernel decision, though, so doubt we can change it.

So, is there something I'm doing wrong? This is a perfectly normal
(if slow) z/VM development system (yes, it's actually emulated, but
for all that, IBM still recognizes Flex boxes as being legitimate
s390x machines until the last dongle has expired!). It runs Etch,
SLES10, and CentOS 4.4 just fine, and it runs z/VM 5.2 and z/OS 1.9
fine too.

Is there something in newer Linux kernels that requires z9 or later
functionality?

Adam

[Bastian Blank]

Post by Adam Thornton
So, is there something I'm doing wrong?

Not that I spot anything.

Post by Adam Thornton
This is a perfectly normal (if
slow) z/VM development system (yes, it's actually emulated, but for all
that, IBM still recognizes Flex boxes as being legitimate s390x machines
until the last dongle has expired!). It runs Etch, SLES10, and CentOS
4.4 just fine, and it runs z/VM 5.2 and z/OS 1.9 fine too.

Etch, SLES10 and CentOS 4.4 are much older.

Post by Adam Thornton
Is there something in newer Linux kernels that requires z9 or later
functionality?

No. They work fine on z900. Can you try if a new kernel will work at all
on this system?

Bastian

[Adam Thornton]

Post by Bastian Blank

Post by Adam Thornton
So, is there something I'm doing wrong?

Not that I spot anything.

Post by Adam Thornton
This is a perfectly normal (if
slow) z/VM development system (yes, it's actually emulated, but for all
that, IBM still recognizes Flex boxes as being legitimate s390x machines
until the last dongle has expired!). It runs Etch, SLES10, and CentOS
4.4 just fine, and it runs z/VM 5.2 and z/OS 1.9 fine too.

Etch, SLES10 and CentOS 4.4 are much older.

Post by Adam Thornton
Is there something in newer Linux kernels that requires z9 or later
functionality?

No. They work fine on z900. Can you try if a new kernel will work at all
on this system?

...maybe. Would installing etchnhalf 2.6.24 probably trigger the bug,
if bug it is?

I can reasonably easily install a new etch machine and then attempt an
upgrade.

Adam

[Dan Dart]

It seems that 2.6.26-2.6.27 have a regression that hangs my machine
during startup of the kernel, that has something to do with HPET. I
can stop it by using "hpet=disable". It seems other people have
trouble with other chipsets, too. 2.6.28 seems to be fine.
Maybe it's the same bug? Though my machine is x86_64, it might be a
valid speculation.

Cheers
Dan

[Adam Thornton]

Post by Dan Dart
It seems that 2.6.26-2.6.27 have a regression that hangs my machine
during startup of the kernel, that has something to do with HPET. I
can stop it by using "hpet=disable". It seems other people have
trouble with other chipsets, too. 2.6.28 seems to be fine.
Maybe it's the same bug? Though my machine is x86_64, it might be a
valid speculation.

Tried with parmline

ro locale=C hpet=disable

But same symptoms. Hangs after "initializing cgroup subsys devices"

Adam

[Adam Thornton]

This is filed as Debian bug 511334.

I'm in the middle of updating first to the 2.6.24 etchnhalf kernel,
and if that succeeds, I'll update this new Etch system to Lenny. I
had no trouble with etch module mismatches, but I did have to run
fdasd by hand because parted failed to create /dev/dasda1.

Etchnhalf works OK.

About to try online update to Lenny.

Adam

[Frans Pop]

The question is : would it be possible to have a more recent DI build
for s390 (more recent than October 27th 2008) - and especially with a
newer kernel ?

You could try a daily built image from [1], which uses 2.6.26-12 (current
unstable/testing kernel based on upstream 2.6.26.8).

note that 2.6.25 might be an issue.

Hmmm? The Lenny RC1 image also uses 2.6.26, though based on a somewhat
older upstream stable update. So .25 cannot be an issue here.

PSF/PRSSD issue that's of some concern to me) 2.6.26 version updated by
Frans Pop should be fine.. and 2.6.27 shouldn't have any problem)

2.6.27 is likely to get skipped for Debian as the kernel will not be
updated until after Lenny is released and that will certainly be with
2.6.26.

Cheers,
FJP

[1] http://www.debian.org/devel/debian-installer/

[Sebastian Welton]

I have also come across the same problem as Adam on a FLEX-ES system
running z/VM 5.2 where it hangs in exactly the same place. A quick test
of just IPL'ing under Hercules from tape though works just fine so I
just need to test using tape on FLEX-ES under z/VM.

One thing that might be a problem is that FLEX-ES does not support QDIO
and this version of Debian only supports CTC, QETH and IUCV so instead
of defining OSA in the FLEX-ES configuration file you might want to
try CTC instead (which is what I did under Hercules.)
--
MfG / Best Regards

Sebastian