xen discuss - Aug 2009 - xVM boot on my laptop ==> immediate reset, reboot

Hi,

trying to boot the osol 2009.06 xVM kernel on my laptop for the first time I
find that it
gets some kind of error during boot, immediately resets and reboots.

The message flashes too quickly I cannot read what it wants to say. (Perhaps I
should try
recording with my video camera?).

Is anyone able to boot xVM successfully on a laptop?

Is xVM on intel hardware solid yet, or should I stay away from it?

My hardware is a Dell Precision M64.

--
Regards,
Georg.
-- 
This message posted from opensolaris.org

> trying to boot the osol 2009.06 xVM kernel on my
> laptop for the first time I find that it 
> gets some kind of error during boot, immediately
> resets and reboots.
A kernel panic?

> The message flashes too quickly I cannot read what it
> wants to say.
Try to boot the kernel with options " -kv".  Edit the
grub boot commands for the xVM kernel; append
the options " -kv" at the end of the 
"module$ /platform/i86xpv/kernel/$ISADIR/unix ..."
line. Try to boot using the modified grub boot entry.
Now it should stop in the kernel debugger when it
gets a kernel panic and you can read the console
messages.  You can also try to use commands like
::msgbuf, ::stack, ::status, ::cpuinfo -v
to get more information.
-- 
This message posted from opensolaris.org

> A kernel panic?
Right.  Thanks for your very helpful advice.  I booted with the debugger enabled
several times
and found it always gets a page fault due to null pointer dereference. But it
gets it at different
places each time! One time was from zfs:zap_lookup_norm, one time from
npe:pci_config_rd8,
one time from something that looked like hypervisor lock ... I stopped
recording.

Smells like memory overwrite, doesn''t it?

The one other observation I have is a pair of warnings a little earlier in the
boot, saying "cannot
load platform pm driver acpippm" and "cannot load platform pm driver
ppm". I''m not sure if that
is relevant, however.

I tried different BIOS settings, too, but no success.  One thing I tried was
toggle virtualization
support from off to on (strangely it is off in the default setting).

--
Regards,
Georg.
-- 
This message posted from opensolaris.org

On Sat, Aug 29, 2009 at 03:57:31AM -0700, Georg-W. Koltermann wrote:
> trying to boot the osol 2009.06 xVM kernel on my laptop for the first time
I find that it
> gets some kind of error during boot, immediately resets and reboots.
b121 just came out, can you try creating a new BE with that version, and
repeating your test?

thanks
john

> > A kernel panic?
> 
> Right.  Thanks for your very helpful advice.  I
> booted with the debugger enabled several times
> and found it always gets a page fault due to null
> pointer dereference. But it gets it at different
> places each time! One time was from
> zfs:zap_lookup_norm, one time from
> npe:pci_config_rd8,
> one time from something that looked like hypervisor
> lock ... I stopped recording.
> 
> Smells like memory overwrite, doesn''t it?
May enabling kmem heap checking finds
some problem?
Boot with options " -kdv" and in kmdb 

    kmem_flags/W f
    :c
-- 
This message posted from opensolaris.org

> it always gets a page fault due to null
> pointer dereference. But it gets it at different
> places each time! One time was from
> zfs:zap_lookup_norm, one time from
> npe:pci_config_rd8,
> one time from something that looked like hypervisor
> lock ... I stopped recording.
Could be similar to bug 6838935
Synopsis: "snv113 xvm bits panic on boot"
http://bugs.opensolaris.org/bugdatabase/view_bug.do?bug_id=6838935
-- 
This message posted from opensolaris.org

> b121 just came out, can you try creating a new BE
> with that version, and
> repeating your test?
> 
> thanks
> john
1.2 GB downloads later I must say b121 is a complete looser on my laptop. Not
even the normal (non-xenified) kernel boots.

Booting with -kv I see it stopping right after:

initialized model-specific module ''cpu_m.GenuineIntel'' on chip
0 core 0 strand 0

I''m pretty disappointed, I never imagined this could happen.

Sorry.

--
Regards,
Georg.
-- 
This message posted from opensolaris.org

> > Smells like memory overwrite, doesn''t it?
> 
> May enabling kmem heap checking finds
> some problem?
> Boot with options " -kdv" and in kmdb 
> 
>     kmem_flags/W f
> :c
I tried it and didn''t see anything new printed.

--
Regards,
Georg.
-- 
This message posted from opensolaris.org

> > b121 just came out, can you try creating a new BE
> > with that version, and repeating your test?
> > 
> 
> 1.2 GB downloads later I must say b121 is a complete
> looser on my laptop. Not even the normal
> (non-xenified) kernel boots.
> 
> Booting with -kv I see it stopping right after:
> 
> initialized model-specific module ''cpu_m.GenuineIntel'' on
chip 0 core 0 strand 0
Is this perhaps bug 6867906 ?
Synopsis: "Kernel hangs during boot in startup_bios_disk with snv_120"
http://bugs.opensolaris.org/bugdatabase/view_bug.do?bug_id=6867906

In builds 120 or 120, certain x86/x64 system may hang during
boot.

Work-around: Boot the system under the kernel debugger by
adding the arguments "-kd" to GRUB''s "kernel$"
entry and enter
the following steps
> dobiosdev/W 0
> :c
-- 
This message posted from opensolaris.org

Ok once I conquered that boot problem with b121, too, I am now up with xen:

# uname -a
SunOS hunter 5.11 snv_121 i86pc i386 i86xpv
# 

Btw @jkeil, setting dobiosdev in /etc/system did work.  Now I only need to
remember removing it after the next upgrade ;-)

Thank you all!

--
Regards,
Georg.
-- 
This message posted from opensolaris.org