Xen users - Mar 2011 - Can''t boot HVM Linux guests on Debian Squeeze Dom0

I just downloaded Debian Squeeze and installed Xen from packages, following
the instructions here: http://wiki.debian.org/Xen
I am able to get PV guests up and running easily, but HVM guests will not
even boot.

The HVM I am currently trying to run is Debian sarge, which I installed into
a disk image and configured using qemu.
I am running on an HP Proliant ML150G6, which has Intel-VT hardware
virtualization. One strange thing that I noticed is both

grep vmx /proc/cpuinfo
grep svm /proc/cpuinfo

return nothing, although according to the BIOS settings the Intel-VT
extensions are enabled.

On trying to boot my HVM guest (xm create sarge.cfg), several different
things would happen. At I fiddled with the configuration file for the VM a
bit in attempts to get things working, it seems to have settled on the
following. It prints

Using config file "./sarge.cfg"

and then hangs for a looong time, at least one minute. After waiting a bit,
it prints

Error: Domain ''sarge'' does not exist.

and quits.

I would really appreciate any advice or insight into this problem and
possible solutions! I am including the configuration file for the VM in
question, please let me know if more information would be useful:


import os, re
arch = os.uname()[4]
if re.search(''64'', arch):
        arch_libdir = ''lib64''
else:
        arch_libdir = ''lib''

device_model = ''/usr/'' + arch_libdir +
''/xen-4.0/bin/qemu-dm''

name = ''sarge''
kernel = ''/usr/lib/xen-default/boot/hvmloader''
builder = ''hvm''
memory = 256
shadow_memory = 16
vcpus = 1
disk = [
''file:/root/sarge.img,hda1,w''
  ]
vif = [ ''type=ioemu, bridge=xenbr0'' ]
on_poweroff = ''destroy''
on_reboot=''restart''
on_crash=''restart''

boot="cda"

sdl=0
vnc=1
vncconsole=1
vncunused=1
stdvga=0
serial=''pty''


Sincerely,
Garrett Robinson

-- 
Garrett Robinson
(484) 802-3882
Oberlin College ''11


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

Am 09.03.2011 11:30, schrieb Garrett Robinson:

Hello,
> grep vmx /proc/cpuinfo
> grep svm /proc/cpuinfo
> 
"xm info | grep xen_caps" should list "hvm-3.0-x86_32
hvm-3.0-x86_32p".
Otherwise something is going wrong with hardware virtualisation.

Regards,
Stefan Kuhne




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

> I just downloaded Debian Squeeze and installed Xen from packages, 
> following the instructions here: http://wiki.debian.org/Xen
> I am able to get PV guests up and running easily, but HVM guests will 
> not even boot.
>
> The HVM I am currently trying to run is Debian sarge, which I 
> installed into a disk image and configured using qemu.
> I am running on an HP Proliant ML150G6, which has Intel-VT hardware 
> virtualization. One strange thing that I noticed is both
>
> grep vmx /proc/cpuinfo
> grep svm /proc/cpuinfo
>
> return nothing, although according to the BIOS settings the Intel-VT 
> extensions are enabled.
This is normal. If things are working properly, "xm info" should show 
"hvm" under "virt_caps" for you (it will just be masked in
cpuinfo).
> On trying to boot my HVM guest (xm create sarge.cfg), several 
> different things would happen. At I fiddled with the configuration 
> file for the VM a bit in attempts to get things working, it seems to 
> have settled on the following. It prints
>
> Using config file "./sarge.cfg"
>
> and then hangs for a looong time, at least one minute. After waiting a 
> bit, it prints
>
> Error: Domain ''sarge'' does not exist.
>
> and quits.
>
> I would really appreciate any advice or insight into this problem and 
> possible solutions! I am including the configuration file for the VM 
> in question, please let me know if more information would be useful:
>
>
> import os, re
> arch = os.uname()[4]
> if re.search(''64'', arch):
>         arch_libdir = ''lib64''
> else:
>         arch_libdir = ''lib''
>
> device_model = ''/usr/'' + arch_libdir +
''/xen-4.0/bin/qemu-dm''
>
> name = ''sarge''
> kernel = ''/usr/lib/xen-default/boot/hvmloader''
> builder = ''hvm''
> memory = 256
> shadow_memory = 16
> vcpus = 1
> disk = [
> ''file:/root/sarge.img,hda1,w''
>  ]
> vif = [ ''type=ioemu, bridge=xenbr0'' ]
> on_poweroff = ''destroy''
> on_reboot=''restart''
> on_crash=''restart''
>
> boot="cda"
>
> sdl=0
> vnc=1
> vncconsole=1
> vncunused=1
> stdvga=0
> serial=''pty''
Try removing these, at minimum:

sdl=0
serial=''pty''

I''ve had the problem that you''re seeing with
"serial=''pty''", in particular.

If it still won''t boot, try removing:

shadow_memory = 16

If there are still problems, then, try running tail -f 
/var/log/xen/*.log while starting your domain, and see what messages Xen 
is kicking out.

-John


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

Thanks to all for the replies!

I checked xm info and virt_caps, and my machine does display the correct
listings there for hardware virtualization.

Commenting out the sdl, serial and shadow_memory parameters in the
configuration file unfortuantely did not change the problem in any way. If
only life were so simple!

The output from the xend.log file is, of course, enormous. It is attached.

The much more manageable log file is /var/log/xen/qemu-dm-sarge, which has
interesting errors related to both networking and the tap driver.

Output of /var/log/xen/qemu-dm-sarge:
domid: 10
-c config qemu network with xen bridge for
tap10.0 xenbr0
bridge xenbr0 does not exist!
/etc/xen/scripts/qemu-ifup: could not launch network script
Could not initialize device ''tap''

I''d really appreciate any help! I''m not sure why the tap
device even needs
to be initialized, as I''m using file: for disk access. And
isn''t xenbr0 the
default bridge?

On Wed, Mar 9, 2011 at 6:03 AM, John Weekes
<lists.xen@nuclearfallout.net>wrote:
>
> I just downloaded Debian Squeeze and installed Xen from packages, following
> the instructions here: http://wiki.debian.org/Xen
> I am able to get PV guests up and running easily, but HVM guests will not
> even boot.
>
>  The HVM I am currently trying to run is Debian sarge, which I installed
> into a disk image and configured using qemu.
> I am running on an HP Proliant ML150G6, which has Intel-VT hardware
> virtualization. One strange thing that I noticed is both
>
>  grep vmx /proc/cpuinfo
> grep svm /proc/cpuinfo
>
>  return nothing, although according to the BIOS settings the Intel-VT
> extensions are enabled.
>
>
> This is normal. If things are working properly, "xm info" should
show "hvm"
> under "virt_caps" for you (it will just be masked in cpuinfo).
>
>
>  On trying to boot my HVM guest (xm create sarge.cfg), several different
> things would happen. At I fiddled with the configuration file for the VM a
> bit in attempts to get things working, it seems to have settled on the
> following. It prints
>
>  Using config file "./sarge.cfg"
>
>  and then hangs for a looong time, at least one minute. After waiting a
> bit, it prints
>
>  Error: Domain ''sarge'' does not exist.
>
>  and quits.
>
>  I would really appreciate any advice or insight into this problem and
> possible solutions! I am including the configuration file for the VM in
> question, please let me know if more information would be useful:
>
>
>  import os, re
> arch = os.uname()[4]
> if re.search(''64'', arch):
>         arch_libdir = ''lib64''
> else:
>         arch_libdir = ''lib''
>
>  device_model = ''/usr/'' + arch_libdir +
''/xen-4.0/bin/qemu-dm''
>
>  name = ''sarge''
> kernel = ''/usr/lib/xen-default/boot/hvmloader''
>  builder = ''hvm''
> memory = 256
> shadow_memory = 16
> vcpus = 1
> disk = [
>  ''file:/root/sarge.img,hda1,w''
>   ]
> vif = [ ''type=ioemu, bridge=xenbr0'' ]
> on_poweroff = ''destroy''
> on_reboot=''restart''
> on_crash=''restart''
>
>  boot="cda"
>
>  sdl=0
> vnc=1
> vncconsole=1
> vncunused=1
> stdvga=0
> serial=''pty''
>
>
> Try removing these, at minimum:
>
>  sdl=0
> serial=''pty''
>
> I''ve had the problem that you''re seeing with
"serial=''pty''", in particular.
>
> If it still won''t boot, try removing:
>
> shadow_memory = 16
>
> If there are still problems, then, try running tail -f /var/log/xen/*.log
> while starting your domain, and see what messages Xen is kicking out.
>
> -John
>


-- 
Garrett Robinson
(484) 802-3882
Oberlin College ''11



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

Success! I was able to boot my HVM sarge guest after making two changes.

The first repaired a simple mistake - I had partitioned a single disk image
using the installer in qemu with a swap and an everything else partition.
The "Domain not found" error went away as soon as I replaced the
erroneous
"hda1" with "hda" in the disk parameter.

The second is more of a long-standing problem. In order to get the guest to
boot, I need to completely disable networking by commenting out the vif=[]
line. Trying to use networking, even with the defaults, results in the
"xenbr0 not found" error seen earlier in /var/log/xen/qemu-dm-sarge.

Another question I have is this: had anybody had any luck using tap:aio on
Xen4.0/Debian? Simply replacing "file:" with "tap:aio" did
not work. I would
love to take advantage of the efficiency of the aio driver, so any advice
with regards to this would be appreciated!

Garrett Robinson

On Wed, Mar 9, 2011 at 12:13 PM, Garrett Robinson <
Garrett.Robinson@oberlin.edu> wrote:
> Thanks to all for the replies!
>
> I checked xm info and virt_caps, and my machine does display the correct
> listings there for hardware virtualization.
>
> Commenting out the sdl, serial and shadow_memory parameters in the
> configuration file unfortuantely did not change the problem in any way. If
> only life were so simple!
>
> The output from the xend.log file is, of course, enormous. It is attached.
>
> The much more manageable log file is /var/log/xen/qemu-dm-sarge, which has
> interesting errors related to both networking and the tap driver.
>
> Output of /var/log/xen/qemu-dm-sarge:
> domid: 10
> -c config qemu network with xen bridge for
> tap10.0 xenbr0
> bridge xenbr0 does not exist!
> /etc/xen/scripts/qemu-ifup: could not launch network script
> Could not initialize device ''tap''
>
> I''d really appreciate any help! I''m not sure why the tap
device even needs
> to be initialized, as I''m using file: for disk access. And
isn''t xenbr0 the
> default bridge?
>
> On Wed, Mar 9, 2011 at 6:03 AM, John Weekes
<lists.xen@nuclearfallout.net>wrote:
>
>>
>> I just downloaded Debian Squeeze and installed Xen from packages,
>> following the instructions here: http://wiki.debian.org/Xen
>> I am able to get PV guests up and running easily, but HVM guests will
not
>> even boot.
>>
>>  The HVM I am currently trying to run is Debian sarge, which I
installed
>> into a disk image and configured using qemu.
>> I am running on an HP Proliant ML150G6, which has Intel-VT hardware
>> virtualization. One strange thing that I noticed is both
>>
>>  grep vmx /proc/cpuinfo
>> grep svm /proc/cpuinfo
>>
>>  return nothing, although according to the BIOS settings the Intel-VT
>> extensions are enabled.
>>
>>
>> This is normal. If things are working properly, "xm info"
should show
>> "hvm" under "virt_caps" for you (it will just be
masked in cpuinfo).
>>
>>
>>  On trying to boot my HVM guest (xm create sarge.cfg), several
different
>> things would happen. At I fiddled with the configuration file for the
VM a
>> bit in attempts to get things working, it seems to have settled on the
>> following. It prints
>>
>>  Using config file "./sarge.cfg"
>>
>>  and then hangs for a looong time, at least one minute. After waiting a
>> bit, it prints
>>
>>  Error: Domain ''sarge'' does not exist.
>>
>>  and quits.
>>
>>  I would really appreciate any advice or insight into this problem and
>> possible solutions! I am including the configuration file for the VM in
>> question, please let me know if more information would be useful:
>>
>>
>>  import os, re
>> arch = os.uname()[4]
>> if re.search(''64'', arch):
>>         arch_libdir = ''lib64''
>> else:
>>         arch_libdir = ''lib''
>>
>>  device_model = ''/usr/'' + arch_libdir +
''/xen-4.0/bin/qemu-dm''
>>
>>  name = ''sarge''
>> kernel = ''/usr/lib/xen-default/boot/hvmloader''
>>  builder = ''hvm''
>> memory = 256
>> shadow_memory = 16
>> vcpus = 1
>> disk = [
>>  ''file:/root/sarge.img,hda1,w''
>>   ]
>> vif = [ ''type=ioemu, bridge=xenbr0'' ]
>> on_poweroff = ''destroy''
>> on_reboot=''restart''
>> on_crash=''restart''
>>
>>  boot="cda"
>>
>>  sdl=0
>> vnc=1
>> vncconsole=1
>> vncunused=1
>> stdvga=0
>> serial=''pty''
>>
>>
>> Try removing these, at minimum:
>>
>>  sdl=0
>> serial=''pty''
>>
>> I''ve had the problem that you''re seeing with
"serial=''pty''", in
>> particular.
>>
>> If it still won''t boot, try removing:
>>
>> shadow_memory = 16
>>
>> If there are still problems, then, try running tail -f
/var/log/xen/*.log
>> while starting your domain, and see what messages Xen is kicking out.
>>
>> -John
>>
>
>
>
> --
> Garrett Robinson
> (484) 802-3882
> Oberlin College ''11
>
>

-- 
Garrett Robinson
(484) 802-3882
Oberlin College ''11


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

On 3/9/2011 5:20 PM, Garrett Robinson wrote:
> Success! I was able to boot my HVM sarge guest after making two changes.
>
> The first repaired a simple mistake - I had partitioned a single disk 
> image using the installer in qemu with a swap and an everything else 
> partition. The "Domain not found" error went away as soon as I 
> replaced the erroneous "hda1" with "hda" in the disk
parameter.
Shoot, I missed that, too.
> The second is more of a long-standing problem. In order to get the 
> guest to boot, I need to completely disable networking by commenting 
> out the vif=[] line. Trying to use networking, even with the defaults, 
> results in the "xenbr0 not found" error seen earlier in 
> /var/log/xen/qemu-dm-sarge.
That''s one that I still don''t quite get, but must have
something to do
with broken networking scripts.
> Another question I have is this: had anybody had any luck using 
> tap:aio on Xen4.0/Debian? Simply replacing "file:" with
"tap:aio" did
> not work. I would love to take advantage of the efficiency of the aio 
> driver, so any advice with regards to this would be appreciated!
Try "tap2:tapdisk:aio:" -- it should work, in 4.0.1. Failing that, try
"tap:tapdisk:aio".

-John

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

Try using eth0 as the bridge device. On my squeeze box they named the
bridge eth0 and the real eth0 got named peth0.


-----Original Message-----
From: xen-users-bounces@lists.xensource.com
[mailto:xen-users-bounces@lists.xensource.com] On Behalf Of John Weekes
Sent: Wednesday, March 09, 2011 7:50 PM
To: xen-users@lists.xensource.com
Subject: Re: [Xen-users] Can''t boot HVM Linux guests on Debian Squeeze
Dom0

On 3/9/2011 5:20 PM, Garrett Robinson wrote:
> Success! I was able to boot my HVM sarge guest after making two
changes.
>
> The first repaired a simple mistake - I had partitioned a single disk 
> image using the installer in qemu with a swap and an everything else 
> partition. The "Domain not found" error went away as soon as I 
> replaced the erroneous "hda1" with "hda" in the disk
parameter.
Shoot, I missed that, too.
> The second is more of a long-standing problem. In order to get the 
> guest to boot, I need to completely disable networking by commenting 
> out the vif=[] line. Trying to use networking, even with the defaults,
> results in the "xenbr0 not found" error seen earlier in 
> /var/log/xen/qemu-dm-sarge.
That''s one that I still don''t quite get, but must have
something to do
with broken networking scripts.
> Another question I have is this: had anybody had any luck using 
> tap:aio on Xen4.0/Debian? Simply replacing "file:" with
"tap:aio" did
> not work. I would love to take advantage of the efficiency of the aio 
> driver, so any advice with regards to this would be appreciated!
Try "tap2:tapdisk:aio:" -- it should work, in 4.0.1. Failing that, try
"tap:tapdisk:aio".

-John

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

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

Hi
>> Trying to use networking, even with the defaults, results in the
>> "xenbr0 not found" error seen earlier in
/var/log/xen/qemu-dm-sarge.
> Try using eth0 as the bridge device. On my squeeze box they named the
> bridge eth0 and the real eth0 got named peth0.
may be eth0, or bond0, or anything else - check it with brctl show first.

Ervin

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

Thanks for all the replies, this is really a fantastic and supportive
community. I will try out these suggestions this weekend and reply with the
results.

On Fri, Mar 11, 2011 at 4:48 PM, Ervin Novak <enovak@opensuse.hu> wrote:
> Hi
>
> >> Trying to use networking, even with the defaults, results in the
> >> "xenbr0 not found" error seen earlier in
/var/log/xen/qemu-dm-sarge.
>
> > Try using eth0 as the bridge device. On my squeeze box they named the
> > bridge eth0 and the real eth0 got named peth0.
>
> may be eth0, or bond0, or anything else - check it with brctl show first.
>
> Ervin
>
> _______________________________________________
> Xen-users mailing list
> Xen-users@lists.xensource.com
> http://lists.xensource.com/xen-users
>


-- 
Garrett Robinson
(484) 802-3882
Oberlin College ''11


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

On Fri, Mar 11, 2011 at 9:48 PM, Ervin Novak <enovak@opensuse.hu>
wrote:
> Hi
>
>>> Trying to use networking, even with the defaults, results in the
>>> "xenbr0 not found" error seen earlier in
/var/log/xen/qemu-dm-sarge.
>
>> Try using eth0 as the bridge device. On my squeeze box they named the
>> bridge eth0 and the real eth0 got named peth0.
>
> may be eth0, or bond0, or anything else - check it with brctl show first.
>
Changing to these may very well work, but I would recommend that we
start getting users used to disabling the xen bridge scripts and
setting up the network bridges using the distro tools. This is what is
going to be happening going forward (Xen 4.1+) and we might as well
get used to it. It actually makes things cleaner and more
understandable as well.

For reference see:
http://wiki.xensource.com/xenwiki/HostConfiguration/Networking
http://wiki.xensource.com/xenwiki/MigrationGuideToXen4.1%2B


> Ervin
>
> _______________________________________________
> Xen-users mailing list
> Xen-users@lists.xensource.com
> http://lists.xensource.com/xen-users
>


-- 
Todd Deshane
http://www.linkedin.com/in/deshantm
http://www.xen.org/products/cloudxen.html

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

Hey John,

I tried both of your suggestions. Neither of them worked, unfortunately.

With ''tap2:tapdisk:aio:'', I got the following error almost
immediately:

# xm create 343-3-0-0
Using config file "./343-3-0-0".
Error: Failed to create device.
    stdout:
    stderr:
Check that target "/home/xenlabs/labs/cs343/hw3/group0/overflow.img"
exists
and that blktap2 driver installed in dom0.

And with ''tap:tapdisk:aio:'', the system hung for almost a
minute on the xm
create command. Then it returned this error:

# xm create 343-3-0-0
Using config file "./343-3-0-0".
Error: Device 768 (tap) could not be connected. Hotplug scripts not working.

Any insight? I was not able to get tap:aio to work on Debian Lenny Dom0
either, but I would really love to get it to work!

On Wed, Mar 9, 2011 at 8:50 PM, John Weekes
<lists.xen@nuclearfallout.net>wrote:
> On 3/9/2011 5:20 PM, Garrett Robinson wrote:
>
>> Success! I was able to boot my HVM sarge guest after making two
changes.
>>
>> The first repaired a simple mistake - I had partitioned a single disk
>> image using the installer in qemu with a swap and an everything else
>> partition. The "Domain not found" error went away as soon as
I replaced the
>> erroneous "hda1" with "hda" in the disk parameter.
>>
>
> Shoot, I missed that, too.
>
>
>  The second is more of a long-standing problem. In order to get the guest
>> to boot, I need to completely disable networking by commenting out the
>> vif=[] line. Trying to use networking, even with the defaults, results
in
>> the "xenbr0 not found" error seen earlier in
/var/log/xen/qemu-dm-sarge.
>>
>
> That''s one that I still don''t quite get, but must have
something to do with
> broken networking scripts.
>
>
>  Another question I have is this: had anybody had any luck using tap:aio on
>> Xen4.0/Debian? Simply replacing "file:" with
"tap:aio" did not work. I would
>> love to take advantage of the efficiency of the aio driver, so any
advice
>> with regards to this would be appreciated!
>>
>
> Try "tap2:tapdisk:aio:" -- it should work, in 4.0.1. Failing
that, try
> "tap:tapdisk:aio".
>
> -John
>
> _______________________________________________
> Xen-users mailing list
> Xen-users@lists.xensource.com
> http://lists.xensource.com/xen-users
>


-- 
Garrett Robinson
(484) 802-3882
Oberlin College ''11


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

On Tue, Mar 15, 2011 at 9:10 AM, Garrett Robinson
<Garrett.Robinson@oberlin.edu> wrote:
> Hey John,
> I tried both of your suggestions. Neither of them worked, unfortunately.
> With ''tap2:tapdisk:aio:'', I got the following error
almost immediately:
> And with ''tap:tapdisk:aio:'', the system hung for almost a
minute on the xm
> create command. Then it returned this error:
There was a discussion thread some time ago on this list (search list
archive for details), and the info was that Squeeze does not include
either tap or tap2 due to some license problem. A patch fixed this,
but Debian manitainers haven''t include it yet.

So short summary, you could either:
- Dump Debian, use other distro with proper tap:aio support (like
RHEL5. Opensuse should support it as well), OR
- Compile your own version of xen (and probably the kernel as well), OR
- use phy:/, or stuck with file:/

-- 
Fajar

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

Thank you, Fajar. This is what I was suspected - it is nice to have it
confirmed.

On Mon, Mar 14, 2011 at 11:58 PM, Fajar A. Nugraha <list@fajar.net> wrote:
> On Tue, Mar 15, 2011 at 9:10 AM, Garrett Robinson
> <Garrett.Robinson@oberlin.edu> wrote:
> > Hey John,
> > I tried both of your suggestions. Neither of them worked,
unfortunately.
> > With ''tap2:tapdisk:aio:'', I got the following error
almost immediately:
>
> > And with ''tap:tapdisk:aio:'', the system hung for
almost a minute on the
> xm
> > create command. Then it returned this error:
>
> There was a discussion thread some time ago on this list (search list
> archive for details), and the info was that Squeeze does not include
> either tap or tap2 due to some license problem. A patch fixed this,
> but Debian manitainers haven''t include it yet.
>
> So short summary, you could either:
> - Dump Debian, use other distro with proper tap:aio support (like
> RHEL5. Opensuse should support it as well), OR
> - Compile your own version of xen (and probably the kernel as well), OR
> - use phy:/, or stuck with file:/
>
> --
> Fajar
>


-- 
Garrett Robinson
(484) 802-3882
Oberlin College ''11


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