Unable to boot dom0 with 2.6.16-rc1 and xen-3.0-testing

Unable to boot dom0 with 2.6.16-rc1 and xen-3.0-testing

[Travis Cross]

I am unable to boot dom0 on a x86_64 (Pentium D dual core) machine
using xen-3.0-testing at changeset 8738 and linux-2.6-xen at
changeset 19827.  Booting dom0 using the latest xen hypervisor and
linux-2.6.12.6 seems to work.

I'm afraid that I'm unable to provide much debugging information, as
the machine reboots immediately, seemingly without the kernel
printing any messages.  The last thing I see is:

(XEN) Scrubbing Free RAM: ........................

This is followed quickly by a blank screen, and the machine restarts.

Just wanted to provide a data point here.  Let me know if anyone
would like to see more system information, or if anyone has ideas on
what further steps I can take.

Cheers,

-- Travis

Re: Unable to boot dom0 with 2.6.16-rc1 and xen-3.0-testing

[Keir Fraser]

On 4 Feb 2006, at 21:34, Travis Cross wrote:

> I am unable to boot dom0 on a x86_64 (Pentium D dual core) machine
> using xen-3.0-testing at changeset 8738 and linux-2.6-xen at
> changeset 19827.  Booting dom0 using the latest xen hypervisor and
> linux-2.6.12.6 seems to work.

Older guest on newer Xen should always work, but the same isn't true 
for newer guest on older Xen. linux-2.6-xen.hg tracks xen-unstable, so 
it's newer than 3.0-testing.

It ought to work when rolled back to a suitable version, or just use 
Linux 2.6.12 in the unstable repo.

  -- Keir

Re: Unable to boot dom0 with 2.6.16-rc1 and xen-3.0-testing

[Travis Cross]

> linux-2.6-xen.hg tracks xen-unstable, so it's newer than
> 3.0-testing.

Thanks for pointing that out.  I started to suspect that might be
the case after diff'ing both
{xen-unstable,xen-3.0-testing}/linux-2.6-xen-sparse against
linux-2.6-xen.

So upon your suggestion, I installed xen and xen-tools from
xen-unstable at 8758, to see if that might make the 2.6.16-rc1
kernel fly.  The result was the same: 2.6.16-rc1 crashes promptly,
2.6.12.6 starts without a hitch.

To check the sanity of my build options for the new kernel on this
box, I built a vanilla 2.6.16-rc2 kernel without Xen, and started it
without the hypervisor running (it worked).

> It ought to work when rolled back to a suitable version, or just
> use Linux 2.6.12 in the unstable repo.

I can start the 2.6.12.6 kernel on any Xen build from xen-3.0.0 to
xen-3.0-testing to the latest xen-unstable.  I can't seem to get a
late-model linux kernel to boot though (either as a dom0 or domU,
actually).

Is there a revision after 2.6.12.6 and before the tip that it might
be useful to test?

Cheers,

-- Travis

Keir Fraser wrote:
> 
> On 4 Feb 2006, at 21:34, Travis Cross wrote:
> 
>> I am unable to boot dom0 on a x86_64 (Pentium D dual core)
>> machine using xen-3.0-testing at changeset 8738 and
>> linux-2.6-xen at changeset 19827.  Booting dom0 using the
>> latest xen hypervisor and linux-2.6.12.6 seems to work.
> 
> 
> Older guest on newer Xen should always work, but the same isn't
> true for newer guest on older Xen. linux-2.6-xen.hg tracks
> xen-unstable, so it's newer than 3.0-testing.
> 
> It ought to work when rolled back to a suitable version, or just
> use Linux 2.6.12 in the unstable repo.
> 
> -- Keir

Re: Unable to boot dom0 with 2.6.16-rc1 and xen-3.0-testing

[Keir Fraser]

On 5 Feb 2006, at 19:11, Travis Cross wrote:

> So upon your suggestion, I installed xen and xen-tools from
> xen-unstable at 8758, to see if that might make the 2.6.16-rc1
> kernel fly.  The result was the same: 2.6.16-rc1 crashes promptly,
> 2.6.12.6 starts without a hitch.
>
> To check the sanity of my build options for the new kernel on this
> box, I built a vanilla 2.6.16-rc2 kernel without Xen, and started it
> without the hypervisor running (it worked).

Have you tried using the tip of unstable for everything (Xen, dom0, 
tools, domU)? There's not much point in pulling kernels from other 
repositories now, if your aim is to get a more bleeding-edge kernel, as 
the unstable repo now contains our 2.6.16-rc2 tree (which is as 
bleeding edge as we get right now). If that fails to work for you, it's 
because 'unstable' is living up to its name right now).

  -- Keir

Re: Unable to boot dom0 with 2.6.16-rc1 and xen-3.0-testing

[Jan Beulich]

Try booting the kernel as uni-processor only; I would suspect the problem of booting SMP kernels (version 2.6.14 and
later, cpu_possible_map getting set up too late) still hasn't been addressed ... Jan

>>> Travis Cross <travis@crosswirecorp.com> 05.02.06 20:11:18 >>>
> linux-2.6-xen.hg tracks xen-unstable, so it's newer than
> 3.0-testing.

Thanks for pointing that out.  I started to suspect that might be
the case after diff'ing both
{xen-unstable,xen-3.0-testing}/linux-2.6-xen-sparse against
linux-2.6-xen.

So upon your suggestion, I installed xen and xen-tools from
xen-unstable at 8758, to see if that might make the 2.6.16-rc1
kernel fly.  The result was the same: 2.6.16-rc1 crashes promptly,
2.6.12.6 starts without a hitch.

To check the sanity of my build options for the new kernel on this
box, I built a vanilla 2.6.16-rc2 kernel without Xen, and started it
without the hypervisor running (it worked).

> It ought to work when rolled back to a suitable version, or just
> use Linux 2.6.12 in the unstable repo.

I can start the 2.6.12.6 kernel on any Xen build from xen-3.0.0 to
xen-3.0-testing to the latest xen-unstable.  I can't seem to get a
late-model linux kernel to boot though (either as a dom0 or domU,
actually).

Is there a revision after 2.6.12.6 and before the tip that it might
be useful to test?

Cheers,

-- Travis

Keir Fraser wrote:
> 
> On 4 Feb 2006, at 21:34, Travis Cross wrote:
> 
>> I am unable to boot dom0 on a x86_64 (Pentium D dual core)
>> machine using xen-3.0-testing at changeset 8738 and
>> linux-2.6-xen at changeset 19827.  Booting dom0 using the
>> latest xen hypervisor and linux-2.6.12.6 seems to work.
> 
> 
> Older guest on newer Xen should always work, but the same isn't
> true for newer guest on older Xen. linux-2.6-xen.hg tracks
> xen-unstable, so it's newer than 3.0-testing.
> 
> It ought to work when rolled back to a suitable version, or just
> use Linux 2.6.12 in the unstable repo.
> 
> -- Keir

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com 
http://lists.xensource.com/xen-devel

Re: Unable to boot dom0 with 2.6.16-rc1 and xen-3.0-testing

[Saku Ytti]

On (2006-02-06 09:37 +0100), Jan Beulich wrote:

> Try booting the kernel as uni-processor only; I would suspect the problem of booting SMP kernels (version 2.6.14 and
> later, cpu_possible_map getting set up too late) still hasn't been addressed ... Jan

 I have this SMP issue too, but this appears to be separate issue, as I have
same results in dom0 with later than 2.6.15, without SMP support compiled.
 The SMP issue results in much further output, even output by dom0, this
particular issue doesn't display single line from dom0 kernel.

> >>> Travis Cross <travis@crosswirecorp.com> 05.02.06 20:11:18 >>>
> > linux-2.6-xen.hg tracks xen-unstable, so it's newer than
> > 3.0-testing.
> 
> Thanks for pointing that out.  I started to suspect that might be
> the case after diff'ing both
> {xen-unstable,xen-3.0-testing}/linux-2.6-xen-sparse against
> linux-2.6-xen.
> 
> So upon your suggestion, I installed xen and xen-tools from
> xen-unstable at 8758, to see if that might make the 2.6.16-rc1
> kernel fly.  The result was the same: 2.6.16-rc1 crashes promptly,
> 2.6.12.6 starts without a hitch.
> 
> To check the sanity of my build options for the new kernel on this
> box, I built a vanilla 2.6.16-rc2 kernel without Xen, and started it
> without the hypervisor running (it worked).
> 
> > It ought to work when rolled back to a suitable version, or just
> > use Linux 2.6.12 in the unstable repo.
> 
> I can start the 2.6.12.6 kernel on any Xen build from xen-3.0.0 to
> xen-3.0-testing to the latest xen-unstable.  I can't seem to get a
> late-model linux kernel to boot though (either as a dom0 or domU,
> actually).
> 
> Is there a revision after 2.6.12.6 and before the tip that it might
> be useful to test?
> 
> Cheers,
> 
> -- Travis
> 
> Keir Fraser wrote:
> > 
> > On 4 Feb 2006, at 21:34, Travis Cross wrote:
> > 
> >> I am unable to boot dom0 on a x86_64 (Pentium D dual core)
> >> machine using xen-3.0-testing at changeset 8738 and
> >> linux-2.6-xen at changeset 19827.  Booting dom0 using the
> >> latest xen hypervisor and linux-2.6.12.6 seems to work.
> > 
> > 
> > Older guest on newer Xen should always work, but the same isn't
> > true for newer guest on older Xen. linux-2.6-xen.hg tracks
> > xen-unstable, so it's newer than 3.0-testing.
> > 
> > It ought to work when rolled back to a suitable version, or just
> > use Linux 2.6.12 in the unstable repo.
> > 
> > -- Keir
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com 
> http://lists.xensource.com/xen-devel
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel

-- 
  ++ytti

Re: Unable to boot dom0 with 2.6.16-rc1 and xen-3.0-testing

[Travis Cross]

Keir Fraser wrote:
> Have you tried using the tip of unstable for everything (Xen, dom0, tools, domU)?

I had tried unstable for Xen and tools, but was using the
linux-2.6-xen tree for the kernel builds.

I just pulled the latest changes from the unstable tree (8793) and
tried again, this time building the kernels from unstable as well.
That worked nicely :)

For kicks, I pulled and rebuilt kernels from the linux-2.6-xen tree
(19848).  Those kernels seemed to work as well.

It appears that changes merged in the last few days have solved this
issue (thanks!).

> There's not much point in pulling kernels from other repositories

Just curious, what is the branch philosophy on the linux-2.6-xen.hg
tree?  It appears to pull more recent changes than the testing tree,
but not pull all the latest changes in the unstable tree.

Cheers,

-- Travis

Re: Unable to boot dom0 with 2.6.16-rc1 and xen-3.0-testing

[Christian Limpach]

On 2/8/06, Travis Cross <travis@crosswirecorp.com> wrote:
> Just curious, what is the branch philosophy on the linux-2.6-xen.hg
> tree?  It appears to pull more recent changes than the testing tree,
> but not pull all the latest changes in the unstable tree.

linux-2.6-xen.hg should be ahead of xen-unstable.hg as far as tracking
linux changes is concerned -- this wasn't the case since the end of
last week because the mercurial repository on kernel.org wasn't being
updated.  I'm planing to update linux-2.6-xen.hg whenever there is a
new -rc version and then update the sparse tree if the -rc version
tests ok.

Changes we make get checked into xen-unstable first and then get
pulled across to linux-2.6-xen.hg within a day or two.

    christian