Re: Debian 11 installer crashed and reboot (in Xen DomU)

[Andy Smith]

Hi Elmar,

On Thu, Jul 27, 2023 at 03:03:20AM +0200, Dr. Nagy Elemér Károly wrote:
> Two years later, all Debian 11.0-12.1 installers still crash in Xen as DomU,
> it is a known bug:
> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=983357
> 
> In PV mode, /boot/grub/grub.cfg is the relevant config file, but in HVM
> mode, isolinux/txt.cfg is.

Are you actually seeing this with PV mode Xen guests? My
understanding is that as it's about the keyboard driver, it can only
affect HVM guests. I have installed several hundred Debian 10/11/12
Xen VMs in PV and PVH mode and never seen this issue.

Also if not doing HVM, it is better to do PVH with Debian 10+,
rather than PV.

I'm afraid I have no suggestion on getting this bug fixed in Debian.
A fix is obviously known as it is mentioned on the (linked)
https://bugzilla.kernel.org/show_bug.cgi?id=207695 but I seem to
recall reading somewhere that this is controversial.

Cheers,
Andy

-- 
https://bitfolk.com/ -- No-nonsense VPS hosting

Re: Debian 11 installer crashed and reboot (in Xen DomU)

[Dr . Nagy Elemér Károly]

Greetings,

Two years later, all Debian 11.0-12.1 installers still crash in Xen as 
DomU, it is a known bug:

https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=983357

In PV mode, /boot/grub/grub.cfg is the relevant config file, but in HVM 
mode, isolinux/txt.cfg is.


The images under install.amd/xen suffer from the same problem at the moment.

Installing Debian 10 and upgrading to Debian 11-12 also works as a 
low-tech solution.


Best wishes,
Elmar

Re: Debian 11 installer crashed and reboot

[zithro]

Hello,

sorry to necro, but did you get ANY news from ANYwhere ?

I also wanted to say thank you very much for your instructions.
You were the only one on which we could rely, in TWO YEARS ! ^^

Have a nice evening !

References (none solved, everyone saying "not me") :

https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=983357
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1005308
https://github.com/systemd/systemd/issues/15784
https://bugzilla.kernel.org/show_bug.cgi?id=207695
(I've lost the link to the report on Xen bugtracker)

Re: Debian 11 installer crashed and reboot

[Chuck Zmudzinski]

On 8/24/2021 2:51 AM, Thomas Schmitt wrote:

Hi,

Chuck Zmudzinski wrote:

I don't know when an official fix will come , but I have come up with
a workaround that works for me:

Congrats. Great investigation work, as far as i can judge as bystander.

So it's only the initrd and not the kernel which spoils booting.
(I wonder whether Secure Boot would detect and refuse on a manipulated
kernel or initrd.)


Have a nice day :)

Thomas



I suppose secure boot might fail but I am not using it in my Xen HVM.
I think the kernel and initrd would need to be digitally signed for
secure boot to be able to check them, and I think in Debian secure boot
only the grub boot loader is digitally signed, and my hack does not
change that. So it is possible Debian's secure boot would also work with
my hack to the initrd.

Cheers,

Chuck

Re: Debian 11 installer crashed and reboot

[Thomas Schmitt]

Hi,

Chuck Zmudzinski wrote:
> I don't know when an official fix will come , but I have come up with
> a workaround that works for me:

Congrats. Great investigation work, as far as i can judge as bystander.

So it's only the initrd and not the kernel which spoils booting.
(I wonder whether Secure Boot would detect and refuse on a manipulated
kernel or initrd.)


Have a nice day :)

Thomas

Re: Debian 11 installer crashed and reboot

[Chuck Zmudzinski]

On 8/19/2021 3:11 PM, Andy Smith wrote:

Hi Chuck,

On Tue, Aug 17, 2021 at 08:04:43AM -0400, Chuck Zmudzinski wrote:

After some testing of the Debian 11 installer on Xen
(using the debian-11.0.0-amd64-netinst.iso), I find that
this image only supports installation into a Xen PV guest,
the guest always crashes and reboots for either a BIOS
or OVMF boot into an HVM Xen guest.

Could you report this to Debian's Xen team as a bug? Perhaps it is
as simple as needing different kernel options in the netinst
installer kernel given that the full install works under HVM?

The Debian Xen team is very under-resourced for human help and it
has been a long time since they have managed to keep the version in
stable to a recent and supported one upstream. If you run a Xen dom0
on Debian I think really you need to be building your own packages
or using the Debian Xen team's packages from sid.

The stable packaged 4.11 hypervisor is out of even security support
upstream so it's not really suitable for production use. I don't
think the Debian Xen team would recommend using it but would instead
suggest using their newer package that;s in sid (on stable) and
test/report bigs against that. But let's get this reported.

I'm not skilled enough in Debian package building to help the team
but I do still report bigs sometimes; for production use I am
building packages from newer upstream source.

For this problem I can't help as I don't run HVM guests (only PV and
PVH).

The Debian Xen team mailing list is at:
https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/pkg-xen-devel

Cheers,
Andy



I just discovered this already has been reported as a bug:

https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=983357

I don't know when an official fix will come , but I have come up with
a workaround that works for me:

This fix might also work on the live ISO images that are also reported
to crash and reboot in Xen HVM domUs.

Prerequisites: xorriso and initramfs-tools packages installed

Download/copy debian-11.0.0-amd64-netinst.iso into current working 
directory:


Step 1: Mount the original iso:

mkdir CD
sudo mount -o ro -t iso9660 debian-11.0.0-amd64-netinst.iso CD

Step 2: Copy the initrd.gz file we need to modify:

cp -p CD/install.amd/initrd.gz .

Step 4: Unmount the original iso:

sudo umount CD

Step 5: As root, extract the initramfs into a working directory:

sudo unmkinitramfs initrd.gz initramfs

Step 6: As root, edit the start-udev script that causes the crash and 
reboot:


sudo vi initramfs/lib/debian-installer/start-udev

In the start-udev script change this line:

udevadm trigger --action=add

to

dmesg | grep DMI: | grep 'Xen HVM domU' || udevadm trigger --action=add

and save the file

This change should only disable the udevadm trigger in Xen HVM DomUs.

Step 7: Re-pack the initrd into a new initrd file:

cd initramfs

find . | sort | sudo cpio --reproducible --quiet -o -H newc > ../newinitrd

cd ..

gzip -v9f newinitrd

Step 8: Create the new installation ISO:

cp -p debian-11.0.0-amd64-netinst.iso debian-11.0.0-amd64-xenhvm-netinst.iso

xorriso -dev debian-11.0.0-amd64-xenhvm-netinst.iso -boot_image any keep \
-map newinitrd.gz /install.amd/initrd.gz

Optional Step 9: Clean up

sudo rm -r initramfs newinitrd.gz initrd.gz

The ISO image debian-11.0.0-amd64-xenhvm-netinst.iso can now be used to 
install
Debian bullseye in a Xen HVM Domu. On the main menu, choose the Install 
option,

not the graphical or expert options, in order to boot using the modified
initrd.gz ramdisk.

Cheers,

Chuck Zmudzinski

Re: Debian 11 installer crashed and reboot

[Chuck Zmudzinski]

On 8/21/2021 7:06 PM, Chuck Zmudzinski wrote:

On 8/21/2021 5:52 AM, Thomas Schmitt wrote:

Hi,

there might be an easier way (still depending on whether such simple
file replacements are tolerated by the booting system):

   # Choose the desired new files
new_kernel=...path.to.desired.new.kernel.in.some.mounted.filesystem...
new_initrd=...path.to.desired.new.initrd.in.some.mounted.filesystem...

   # Choose the files in the ISO to be replaced by the new files.
   # The ISO needs not to be mounted. (Better it should not be mounted.)
   # Naively, i assume that it's about these two:
   old_kernel=/install.amd/xen/vmlinuz
   old_initrd=/install.amd/xen/initrd.gz

   # Copy netinst ISO to playground ISO
   cp debian-11.0.0-amd64-netinst.iso test.iso

   # Use xorriso multi-session to override the undesired files
   xorriso -dev test.iso \
   -boot_image any keep \
   -map "$new_kernel" "$old_kernel" \
   -map "$new_initrd" "$old_initrd"

This will result in a quite short run of xorriso, which will report
something like

   ISO image produced: 8108 sectors
   Written to medium : 8288 sectors at LBA 193024
   Writing to 'test.iso' completed successfully.

(The numbers 8108 and 8288 are result of the sizes of the two dummy 
files

which i used as $new_kernel #and new_initrd. Yours will be substantially
larger due to the typical size of an initrd.)

Inspection of the boot equipment of test.iso shows no differences,
except the new storage position of the El Torito boot catalog and the
enlarged ISO image size.

When mounted by Linux, the paths
   /mnt/iso/install.amd/xen/vmlinuz
   /mnt/iso/install.amd/xen/initrd.gz
lead to the content of $new_kernel and $new_initrd.

Then it depends on the inner doings of boot loader and starting system,
whether this perky change is tolerated.


Have a nice day :)

Thomas



Hi Thomas,

Thank you for these tips. I will use them to try to get a bullseye
Debian installer ISO working in a Xen HVM DomU next week
when I have some spare time and post the results here.

All the best,

Chuck



This approach (using xorriso with -dev, -boot_image, and -map options)
works to change the kernel and ramdisk on the ISO, and the new
ISO does boot into a Xen HVM without crashing and rebooting, but I
still have not succeeded in getiing the ramdisk configured correctly to
boot the Debian installer. Using the ramdisk of a full install on the
ISO causes init to drop to an initramfs shell because that ramdisk
is not configured to start the Debian installer but instead expects a
root filesystem to be passed to it on the linux kernel command line.

I expect it is possible to create a ramdisk that will successfully boot
the Debian installer in a Xen HVM. I think some components from
the ramdisk of a full install are needed to prevent the crash and
reboot, and it is a matter of identifying what components from
the ramdisk of a full install are needed. I would suspect some
kernel module that a Xen HVM needs is missing from the ramdisks
on the installation ISO.

I will continue to work on it this week in my spare time.

Cheers,

Chuck

Re: Debian 11 installer crashed and reboot

[Chuck Zmudzinski]

On 8/16/2021 2:30 AM, John Mok wrote:

Hi all,

Tried to install Debian 11 guest using netinst, but the installer
crashed and reboot automatically.

  Host: Xen 4.11.4 on Debian 10
Guest: Debian 11 (kernel 5.10)

Here is the steps to reproduce the problem:-
1) Either guest BIOS or OVMF boot
2) Select "Expert install" on installation menu

Then, the guest crashed and reboot.

Another try with the installer on Debian Live 11, it immediately
crashed and reboot.

Is it a installer bug or something ?

Thanks a lot.

John Mok



This problem has been reported by Philip Susi on 2021-04-27 in a
comment on linux kernel bug 207695:

https://bugzilla.kernel.org/show_bug.cgi?id=207695#c3

The kernel panic is caused by the Xen virtual keyboard.

I was able to capture a screenshot of the crash by setting
panic=60 and removing quiet from the kernel command line
on the grub menu (the kernel parameters for a grub boot
entry can be adjusted by pressing e when that grub boot
entry is highlighted, when booting with ovmf firmware).

Error message  reported by the kernel is:

failed to write 'add' to '/devices/virtual/input/input2/uevent': cannot 
allocate memory


This problem only occurs using the kernel on the installation
media, a full-featured bullseye kernel boots fine in a Xen HVM
guest.

The installer bullseye kernel also boots fine from the installer
media in a Xen PV guest.

I think the Xen HVM virtual keyboard is one provided by
Qemu. It apparently needs a larger buffer than the one
provided by the installation kernel.

Any ideas how to fix this in the installation kernel?

Chuck

Re: Debian 11 installer crashed and reboot

[Chuck Zmudzinski]

On 8/21/2021 5:52 AM, Thomas Schmitt wrote:

Hi,

there might be an easier way (still depending on whether such simple
file replacements are tolerated by the booting system):

   # Choose the desired new files
   new_kernel=...path.to.desired.new.kernel.in.some.mounted.filesystem...
   new_initrd=...path.to.desired.new.initrd.in.some.mounted.filesystem...

   # Choose the files in the ISO to be replaced by the new files.
   # The ISO needs not to be mounted. (Better it should not be mounted.)
   # Naively, i assume that it's about these two:
   old_kernel=/install.amd/xen/vmlinuz
   old_initrd=/install.amd/xen/initrd.gz

   # Copy netinst ISO to playground ISO
   cp debian-11.0.0-amd64-netinst.iso test.iso

   # Use xorriso multi-session to override the undesired files
   xorriso -dev test.iso \
   -boot_image any keep \
   -map "$new_kernel" "$old_kernel" \
   -map "$new_initrd" "$old_initrd"

This will result in a quite short run of xorriso, which will report
something like

   ISO image produced: 8108 sectors
   Written to medium : 8288 sectors at LBA 193024
   Writing to 'test.iso' completed successfully.

(The numbers 8108 and 8288 are result of the sizes of the two dummy files
which i used as $new_kernel #and new_initrd. Yours will be substantially
larger due to the typical size of an initrd.)

Inspection of the boot equipment of test.iso shows no differences,
except the new storage position of the El Torito boot catalog and the
enlarged ISO image size.

When mounted by Linux, the paths
   /mnt/iso/install.amd/xen/vmlinuz
   /mnt/iso/install.amd/xen/initrd.gz
lead to the content of $new_kernel and $new_initrd.

Then it depends on the inner doings of boot loader and starting system,
whether this perky change is tolerated.


Have a nice day :)

Thomas



Hi Thomas,

Thank you for these tips. I will use them to try to get a bullseye
Debian installer ISO working in a Xen HVM DomU next week
when I have some spare time and post the results here.

All the best,

Chuck

Re: Debian 11 installer crashed and reboot

[Chuck Zmudzinski]

On 8/19/2021 3:11 PM, Andy Smith wrote:

Hi Chuck,

On Tue, Aug 17, 2021 at 08:04:43AM -0400, Chuck Zmudzinski wrote:

After some testing of the Debian 11 installer on Xen
(using the debian-11.0.0-amd64-netinst.iso), I find that
this image only supports installation into a Xen PV guest,
the guest always crashes and reboots for either a BIOS
or OVMF boot into an HVM Xen guest.

Could you report this to Debian's Xen team as a bug? Perhaps it is
as simple as needing different kernel options in the netinst
installer kernel given that the full install works under HVM?


Would you also consider the fact that the OP also noted that the
live ISO also crashes and reboots in a Xen HVM guest a bug? My
thought is that both netinst and live amd64 ISOs should work on
a Xen HVM out of the box (OOTB). If I report a bug, I am inclined
to consider the bug to be that both netinst and live amd64 ISOs
crash in Xen HVM guests and the bug should not be closed until
both ISOs work OOTB in a Xen HVM guest on both a default
Xen/Dom0 stable and oldstable installation, so that users of
oldstable can use oldstable to quickly and easily test stable in a
Xen HVM guest. This would require, as I understand it, close
cooperation between the Xen team, the Qemu team, and the
Debian installer developers so those key parts of the Xen HVM
environment play nicely together with the Debian installer.

I also wonder if the bug should be reported to Debian installer
developers instead of to the Xen team.

Cheers,

Chuck

Re: Debian 11 installer crashed and reboot

[Chuck Zmudzinski]

On 8/21/2021 4:34 AM, Thomas Schmitt wrote:

Hi,

John Mok wrote:

As a quick fix on the make installer media work, how can I remake my own
installation ISO by recompil ing the kernel and initramfs ?

Chuck Zmudzinski wrote:

A quick solution would not even require any recompilation
of a kernel, since we know the kernel and ramdisk of
an installed bullseye system does work in a Xen HVM guest.
[...]
I am not familiar with the process that the debian installer team
uses to generate the installation ISOs, but if I wanted to learn
I would start by looking at
https://wiki.debian.org/DebianInstaller/CheckOut

If there are no checksums in the ISO which would cover the initially
booted kernel and the ramdisk, then you could also go the path of

   https://wiki.debian.org/RepackBootableISO

This article mainly shows how to determine the necessary arguments for
the ISO 9660 producing program. Nevertheless, in
   
https://wiki.debian.org/RepackBootableISO#Methods_to_overwrite.2C_delete.2C_or_add_files
it shows some ways how to do the desired manipulations of the ISO's
file tree.

I assume that questions about the entrails of a Debian installation ISO
are best asked at debian-cd mailing list. (Although two of the experts
there happen to be also our friendly list hosts here.)


Have a nice day :)

Thomas



This is how it should be - I am happy that Debian installer developers are
following the user list. There is no more important Debian user to support
than the user who tries to install Debian for the first time. It is the 
Debian

installer's job to try to ensure, as far as possible, that a new Debian user
will be able to quickly and easily install Debian on all supported 
platforms.

I presume that would include at least the bare hardware of the supported
architectures, things like docker/Linux containers and supported virtual
environments such as qemu/kvm and Xen. So the Debian installer developers
should always monitor the Debian user lists for problems that users are 
having

installing Debian in the supported computing environments.

Cheers,

Chuck

Re: Debian 11 installer crashed and reboot

[Thomas Schmitt]

Hi,

there might be an easier way (still depending on whether such simple
file replacements are tolerated by the booting system):

  # Choose the desired new files
  new_kernel=...path.to.desired.new.kernel.in.some.mounted.filesystem...
  new_initrd=...path.to.desired.new.initrd.in.some.mounted.filesystem...

  # Choose the files in the ISO to be replaced by the new files.
  # The ISO needs not to be mounted. (Better it should not be mounted.)
  # Naively, i assume that it's about these two:
  old_kernel=/install.amd/xen/vmlinuz
  old_initrd=/install.amd/xen/initrd.gz

  # Copy netinst ISO to playground ISO
  cp debian-11.0.0-amd64-netinst.iso test.iso

  # Use xorriso multi-session to override the undesired files
  xorriso -dev test.iso \
  -boot_image any keep \
  -map "$new_kernel" "$old_kernel" \
  -map "$new_initrd" "$old_initrd"

This will result in a quite short run of xorriso, which will report
something like

  ISO image produced: 8108 sectors
  Written to medium : 8288 sectors at LBA 193024
  Writing to 'test.iso' completed successfully.

(The numbers 8108 and 8288 are result of the sizes of the two dummy files
which i used as $new_kernel #and new_initrd. Yours will be substantially
larger due to the typical size of an initrd.)

Inspection of the boot equipment of test.iso shows no differences,
except the new storage position of the El Torito boot catalog and the
enlarged ISO image size.

When mounted by Linux, the paths
  /mnt/iso/install.amd/xen/vmlinuz
  /mnt/iso/install.amd/xen/initrd.gz
lead to the content of $new_kernel and $new_initrd.

Then it depends on the inner doings of boot loader and starting system,
whether this perky change is tolerated.


Have a nice day :)

Thomas

Re: Debian 11 installer crashed and reboot

[Thomas Schmitt]

Hi,

John Mok wrote:
> > As a quick fix on the make installer media work, how can I remake my own
> > installation ISO by recompil ing the kernel and initramfs ?

Chuck Zmudzinski wrote:
> A quick solution would not even require any recompilation
> of a kernel, since we know the kernel and ramdisk of
> an installed bullseye system does work in a Xen HVM guest.
> [...]
> I am not familiar with the process that the debian installer team
> uses to generate the installation ISOs, but if I wanted to learn
> I would start by looking at
> https://wiki.debian.org/DebianInstaller/CheckOut

If there are no checksums in the ISO which would cover the initially
booted kernel and the ramdisk, then you could also go the path of

  https://wiki.debian.org/RepackBootableISO

This article mainly shows how to determine the necessary arguments for
the ISO 9660 producing program. Nevertheless, in
  
https://wiki.debian.org/RepackBootableISO#Methods_to_overwrite.2C_delete.2C_or_add_files
it shows some ways how to do the desired manipulations of the ISO's
file tree.

I assume that questions about the entrails of a Debian installation ISO
are best asked at debian-cd mailing list. (Although two of the experts
there happen to be also our friendly list hosts here.)


Have a nice day :)

Thomas

Re: Debian 11 installer crashed and reboot

[Chuck Zmudzinski]

On 8/21/2021 12:08 AM, Chuck Zmudzinski wrote:

On 8/19/2021 3:11 PM, Andy Smith wrote:

Hi Chuck,

On Tue, Aug 17, 2021 at 08:04:43AM -0400, Chuck Zmudzinski wrote:

After some testing of the Debian 11 installer on Xen
(using the debian-11.0.0-amd64-netinst.iso), I find that
this image only supports installation into a Xen PV guest,
the guest always crashes and reboots for either a BIOS
or OVMF boot into an HVM Xen guest.

Could you report this to Debian's Xen team as a bug? Perhaps it is
as simple as needing different kernel options in the netinst
installer kernel given that the full install works under HVM?


Hi Andy,

That is exactly what I think the problem is. I will look into it and 
see if

there is a simple solution like adding the correct kernel configuration
options and kernel modules to the kernel and ramdisk on the default
debian installer iso. The generic default amd64 iso should work in
Xen HVM guests.


The Debian Xen team is very under-resourced for human help and it
has been a long time since they have managed to keep the version in
stable to a recent and supported one upstream. If you run a Xen dom0
on Debian I think really you need to be building your own packages
or using the Debian Xen team's packages from sid.

The stable packaged 4.11 hypervisor is out of even security support
upstream so it's not really suitable for production use.


With bullseye released, that is oldstable now. Xen 4.14 is on bullseye
and I think Xen 4.15 is the latest release upstream, so it is not
too far behind now. I presume Xen 4.14 is still getting security patches
upstream, but I cannot find a good explanation of Xen's support
cycle on their website and I don't know if Debian can expect upstream
to support 4.14 until bullseye becomes oldstable in two years or so.


I found this page: https://xenbits.xen.org/docs/unstable/support-matrix.html

Xen 4.14 will have security support until 2023-07-24, which should cover 
most

of the time bullseye will be the stable version.

Chuck



I don't
think the Debian Xen team would recommend using it but would instead
suggest using their newer package that;s in sid (on stable) and
test/report bigs against that. But let's get this reported.


If/when I find the solution, I will post a bug report and see if the Xen
team can get the solution into the installer media for future bullseye
point releases. I already know the same problem exists on Xen 4.14
on bullseye, and AFAIK even sid has not bumped to 4.15 yet.


I'm not skilled enough in Debian package building to help the team
but I do still report bigs sometimes; for production use I am
building packages from newer upstream source.

For this problem I can't help as I don't run HVM guests (only PV and
PVH).


I always had difficulty with pygrub/pvgrub on PV domains, and using
bullseye's Xen-4.14 version, I could not boot the debian installer iso
with pygrub but had to extract the xen-enabled kernel and ramdisk to
Dom0 and boot them from within Dom0. I think this is another Xen
bug in the Xen-4.14 package that I also will investigate next week.
Probably some tweaks to the pygrub script are needed there.

I have always wanted to try out PVH domains, but have not done so
yet.


The Debian Xen team mailing list is at:
https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/pkg-xen-devel


I will do some testing on this next week. I do want to help the Xen team
make Debian more Xen-friendly and will report these bugs, hopefully
sometime next week.

Cheers,

Chuck


Cheers,
Andy

Re: Debian 11 installer crashed and reboot

[Chuck Zmudzinski]

On 8/21/2021 12:42 AM, John Mok wrote:

Hi Chuck,

As a quick fix on the make installer media work, how can I remake my 
own installation ISO by recompil ing the kernel and initramfs ?


Can give me some directions ?


Not yet regarding what kernel options are needed, because
I have not done any tests yet to see which kernel configuration
options or modules might be missing that are needed by the
Xen HVM.

A quick solution would not even require any recompilation
of a kernel, since we know the kernel and ramdisk of
an installed bullseye system does work in a Xen HVM guest.

Therefore...

if you can build an iso that is the same as the installation
iso in every way except that it boots using the default Debian kernel
and ramdisk on the final installed system instead of the stripped
down kernels/ramdisks on the installation media, you should get an
iso that can boot the debian installer in a Xen HVM guest.

I am not familiar with the process that the debian installer team
uses to generate the installation ISOs, but if I wanted to learn
I would start by looking at

https://wiki.debian.org/DebianInstaller/CheckOut 



and follow the instructions in the wiki checkout the source code
for the debian installer and from there you might be able to learn
how to build and modify a Debian installation ISO.

Regards,

Chuck

Chuck



Thanks a lot.

John Mok


On Sat, Aug 21, 2021, 12:08 Chuck Zmudzinski > wrote:


On 8/19/2021 3:11 PM, Andy Smith wrote:
> Hi Chuck,
>
> On Tue, Aug 17, 2021 at 08:04:43AM -0400, Chuck Zmudzinski wrote:
>> After some testing of the Debian 11 installer on Xen
>> (using the debian-11.0.0-amd64-netinst.iso), I find that
>> this image only supports installation into a Xen PV guest,
>> the guest always crashes and reboots for either a BIOS
>> or OVMF boot into an HVM Xen guest.
> Could you report this to Debian's Xen team as a bug? Perhaps it is
> as simple as needing different kernel options in the netinst
> installer kernel given that the full install works under HVM?
Hi Andy,

That is exactly what I think the problem is. I will look into it
and see if
there is a simple solution like adding the correct kernel
configuration
options and kernel modules to the kernel and ramdisk on the default
debian installer iso. The generic default amd64 iso should work in
Xen HVM guests.
>
> The Debian Xen team is very under-resourced for human help and it
> has been a long time since they have managed to keep the version in
> stable to a recent and supported one upstream. If you run a Xen dom0
> on Debian I think really you need to be building your own packages
> or using the Debian Xen team's packages from sid.
>
> The stable packaged 4.11 hypervisor is out of even security support
> upstream so it's not really suitable for production use.
With bullseye released, that is oldstable now. Xen 4.14 is on bullseye
and I think Xen 4.15 is the latest release upstream, so it is not
too far behind now. I presume Xen 4.14 is still getting security
patches
upstream, but I cannot find a good explanation of Xen's support
cycle on their website and I don't know if Debian can expect upstream
to support 4.14 until bullseye becomes oldstable in two years or so.
> I don't
> think the Debian Xen team would recommend using it but would instead
> suggest using their newer package that;s in sid (on stable) and
> test/report bigs against that. But let's get this reported.
If/when I find the solution, I will post a bug report and see if
the Xen
team can get the solution into the installer media for future bullseye
point releases. I already know the same problem exists on Xen 4.14
on bullseye, and AFAIK even sid has not bumped to 4.15 yet.
>
> I'm not skilled enough in Debian package building to help the team
> but I do still report bigs sometimes; for production use I am
> building packages from newer upstream source.
>
> For this problem I can't help as I don't run HVM guests (only PV and
> PVH).
I always had difficulty with pygrub/pvgrub on PV domains, and using
bullseye's Xen-4.14 version, I could not boot the debian installer iso
with pygrub but had to extract the xen-enabled kernel and ramdisk to
Dom0 and boot them from within Dom0. I think this is another Xen
bug in the Xen-4.14 package that I also will investigate next week.
Probably some tweaks to the pygrub script are needed there.

I have always wanted to try out PVH domains, but have not done so
yet.
>
> The Debian Xen team mailing list is at:
>
https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/pkg-xen-devel

I will do some testing on this next week. I 

Re: Debian 11 installer crashed and reboot

[John Mok]

Hi Chuck,

As a quick fix on the make installer media work, how can I remake my own
installation ISO by recompil ing the kernel and initramfs ?

Can give me some directions ?

Thanks a lot.

John Mok


On Sat, Aug 21, 2021, 12:08 Chuck Zmudzinski  wrote:

> On 8/19/2021 3:11 PM, Andy Smith wrote:
> > Hi Chuck,
> >
> > On Tue, Aug 17, 2021 at 08:04:43AM -0400, Chuck Zmudzinski wrote:
> >> After some testing of the Debian 11 installer on Xen
> >> (using the debian-11.0.0-amd64-netinst.iso), I find that
> >> this image only supports installation into a Xen PV guest,
> >> the guest always crashes and reboots for either a BIOS
> >> or OVMF boot into an HVM Xen guest.
> > Could you report this to Debian's Xen team as a bug? Perhaps it is
> > as simple as needing different kernel options in the netinst
> > installer kernel given that the full install works under HVM?
> Hi Andy,
>
> That is exactly what I think the problem is. I will look into it and see if
> there is a simple solution like adding the correct kernel configuration
> options and kernel modules to the kernel and ramdisk on the default
> debian installer iso. The generic default amd64 iso should work in
> Xen HVM guests.
> >
> > The Debian Xen team is very under-resourced for human help and it
> > has been a long time since they have managed to keep the version in
> > stable to a recent and supported one upstream. If you run a Xen dom0
> > on Debian I think really you need to be building your own packages
> > or using the Debian Xen team's packages from sid.
> >
> > The stable packaged 4.11 hypervisor is out of even security support
> > upstream so it's not really suitable for production use.
> With bullseye released, that is oldstable now. Xen 4.14 is on bullseye
> and I think Xen 4.15 is the latest release upstream, so it is not
> too far behind now. I presume Xen 4.14 is still getting security patches
> upstream, but I cannot find a good explanation of Xen's support
> cycle on their website and I don't know if Debian can expect upstream
> to support 4.14 until bullseye becomes oldstable in two years or so.
> > I don't
> > think the Debian Xen team would recommend using it but would instead
> > suggest using their newer package that;s in sid (on stable) and
> > test/report bigs against that. But let's get this reported.
> If/when I find the solution, I will post a bug report and see if the Xen
> team can get the solution into the installer media for future bullseye
> point releases. I already know the same problem exists on Xen 4.14
> on bullseye, and AFAIK even sid has not bumped to 4.15 yet.
> >
> > I'm not skilled enough in Debian package building to help the team
> > but I do still report bigs sometimes; for production use I am
> > building packages from newer upstream source.
> >
> > For this problem I can't help as I don't run HVM guests (only PV and
> > PVH).
> I always had difficulty with pygrub/pvgrub on PV domains, and using
> bullseye's Xen-4.14 version, I could not boot the debian installer iso
> with pygrub but had to extract the xen-enabled kernel and ramdisk to
> Dom0 and boot them from within Dom0. I think this is another Xen
> bug in the Xen-4.14 package that I also will investigate next week.
> Probably some tweaks to the pygrub script are needed there.
>
> I have always wanted to try out PVH domains, but have not done so
> yet.
> >
> > The Debian Xen team mailing list is at:
> > https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/pkg-xen-devel
> I will do some testing on this next week. I do want to help the Xen team
> make Debian more Xen-friendly and will report these bugs, hopefully
> sometime next week.
>
> Cheers,
>
> Chuck
> >
> > Cheers,
> > Andy
> >
>
>

Re: Debian 11 installer crashed and reboot

[Chuck Zmudzinski]

On 8/19/2021 3:11 PM, Andy Smith wrote:

Hi Chuck,

On Tue, Aug 17, 2021 at 08:04:43AM -0400, Chuck Zmudzinski wrote:

After some testing of the Debian 11 installer on Xen
(using the debian-11.0.0-amd64-netinst.iso), I find that
this image only supports installation into a Xen PV guest,
the guest always crashes and reboots for either a BIOS
or OVMF boot into an HVM Xen guest.

Could you report this to Debian's Xen team as a bug? Perhaps it is
as simple as needing different kernel options in the netinst
installer kernel given that the full install works under HVM?

Hi Andy,

That is exactly what I think the problem is. I will look into it and see if
there is a simple solution like adding the correct kernel configuration
options and kernel modules to the kernel and ramdisk on the default
debian installer iso. The generic default amd64 iso should work in
Xen HVM guests.


The Debian Xen team is very under-resourced for human help and it
has been a long time since they have managed to keep the version in
stable to a recent and supported one upstream. If you run a Xen dom0
on Debian I think really you need to be building your own packages
or using the Debian Xen team's packages from sid.

The stable packaged 4.11 hypervisor is out of even security support
upstream so it's not really suitable for production use.

With bullseye released, that is oldstable now. Xen 4.14 is on bullseye
and I think Xen 4.15 is the latest release upstream, so it is not
too far behind now. I presume Xen 4.14 is still getting security patches
upstream, but I cannot find a good explanation of Xen's support
cycle on their website and I don't know if Debian can expect upstream
to support 4.14 until bullseye becomes oldstable in two years or so.

I don't
think the Debian Xen team would recommend using it but would instead
suggest using their newer package that;s in sid (on stable) and
test/report bigs against that. But let's get this reported.

If/when I find the solution, I will post a bug report and see if the Xen
team can get the solution into the installer media for future bullseye
point releases. I already know the same problem exists on Xen 4.14
on bullseye, and AFAIK even sid has not bumped to 4.15 yet.


I'm not skilled enough in Debian package building to help the team
but I do still report bigs sometimes; for production use I am
building packages from newer upstream source.

For this problem I can't help as I don't run HVM guests (only PV and
PVH).

I always had difficulty with pygrub/pvgrub on PV domains, and using
bullseye's Xen-4.14 version, I could not boot the debian installer iso
with pygrub but had to extract the xen-enabled kernel and ramdisk to
Dom0 and boot them from within Dom0. I think this is another Xen
bug in the Xen-4.14 package that I also will investigate next week.
Probably some tweaks to the pygrub script are needed there.

I have always wanted to try out PVH domains, but have not done so
yet.


The Debian Xen team mailing list is at:
https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/pkg-xen-devel

I will do some testing on this next week. I do want to help the Xen team
make Debian more Xen-friendly and will report these bugs, hopefully
sometime next week.

Cheers,

Chuck


Cheers,
Andy

Re: Debian 11 installer crashed and reboot

[Andy Smith]

Hi Chuck,

On Tue, Aug 17, 2021 at 08:04:43AM -0400, Chuck Zmudzinski wrote:
> After some testing of the Debian 11 installer on Xen
> (using the debian-11.0.0-amd64-netinst.iso), I find that
> this image only supports installation into a Xen PV guest,
> the guest always crashes and reboots for either a BIOS
> or OVMF boot into an HVM Xen guest.

Could you report this to Debian's Xen team as a bug? Perhaps it is
as simple as needing different kernel options in the netinst
installer kernel given that the full install works under HVM?

The Debian Xen team is very under-resourced for human help and it
has been a long time since they have managed to keep the version in
stable to a recent and supported one upstream. If you run a Xen dom0
on Debian I think really you need to be building your own packages
or using the Debian Xen team's packages from sid.

The stable packaged 4.11 hypervisor is out of even security support
upstream so it's not really suitable for production use. I don't
think the Debian Xen team would recommend using it but would instead
suggest using their newer package that;s in sid (on stable) and
test/report bigs against that. But let's get this reported.

I'm not skilled enough in Debian package building to help the team
but I do still report bigs sometimes; for production use I am
building packages from newer upstream source.

For this problem I can't help as I don't run HVM guests (only PV and
PVH).

The Debian Xen team mailing list is at:
https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/pkg-xen-devel

Cheers,
Andy

-- 
https://bitfolk.com/ -- No-nonsense VPS hosting

Re: Debian 11 installer crashed and reboot

[Chuck Zmudzinski]

On 8/17/2021 8:04 AM, Chuck Zmudzinski wrote:

On 8/16/2021 4:25 AM, Andrew M.A. Cater wrote:

On Mon, Aug 16, 2021 at 02:30:33PM +0800, John Mok wrote:

Hi all,

Tried to install Debian 11 guest using netinst, but the installer
crashed and reboot automatically.

  Host: Xen 4.11.4 on Debian 10
Guest: Debian 11 (kernel 5.10)

Here is the steps to reproduce the problem:-
1) Either guest BIOS or OVMF boot
2) Select "Expert install" on installation menu

Then, the guest crashed and reboot.

Another try with the installer on Debian Live 11, it immediately
crashed and reboot.

Is it a installer bug or something ?

Thanks a lot.

John Mok

I think people would need many more details as to exactly what 
happens and

when it reboots.

So: You boot the installer - from just a .iso file / from the .iso 
flashed

to a USB? Can you show what works and at what point it fails?

The quick suggestion would be that it's something to do with Xen 
hypervisor
but I'm not in any better position than to guess, since I don't use 
Xen here.


I do know that none of the tests the media team do involve booting on 
Xen:
in most cases we prefer to run on real hardware. Occasionally, some 
of the

tests have to be run on KVM/QEMU. Xen as DomU  is untested then.

All the very best, as ever,

Andy Cater




After some testing of the Debian 11 installer on Xen
(using the debian-11.0.0-amd64-netinst.iso), I find that
this image only supports installation into a Xen PV guest,
the guest always crashes and reboots for either a BIOS
or OVMF boot into an HVM Xen guest.

I also found that the device model fails to start if one
tries to run the xen built-in vnc server for graphical
output. The only solution is to connect to the
Xen console using the xl console command. The -c
switch to xl create did not work to connect to the
Xen console when creating the guest; it could not
find the pty device, so it is necessary to let the guest
complete the startup and then run xl console 
to connect to the console once the VM is running.
This will display the first page of the debian installer
where the language is selected, and from there
Debian 11 can be installed onto a virtual disk.

The following lines in the xl.cfg file enabled the VM
to boot the debian-11.0.0-amd64-netinst.iso as a
PV Xen guest:

bootloader = 'pygrub'
bootloader_args= [ 
'--kernel=install.amd/xen/vmlinuz','--ramdisk=install.amd/xen/initrd.gz' 
]


I forgot one more essential tweak to get the VM to
boot the Debian installer under pygrub and xen: It was necessary
for the first disk to be the iso image, not the hard disk,
in the disk = [ ... ] directive in the xl.cfg file when booting from
the cd image. Otherwise, pygrub could not find the kernel and
initrd images. This is documented somewhere, but I don't remember
where, and the documentation for booting the debian installer
into Xen is really not very good.

Chuck Zmudzinski


Once installed and booting from the virtual disk, the
bootloader_args setting can be commented out.

After booting from the virtual disk, the grub-pc package can
be added to boot into a Xen HVM guest with BIOS booting,
and if space is left for an efi partition, the grub-efi package
can be installed to boot with OVMF after converting
the partition table of the virtual disk from MBR to GPT.

Also, the vnc server can be used for graphical output
after booting from the virtual disk to run a desktop
environment.

All in all, it is a big pain to install Debian 11 on the Xen
hypervisor using the debian installer, as it seems
supporting Xen installations is not a priority for Debian.

But once it is installed, it works very well on Xen, and
I find I can use the same bullseye image to boot it
as a Xen PV guest, a Xen HVM guest using BIOS boot,
or as a Xen HVM guest using OVMF boot just by tweaking
the xl.cfg configuration file after getting the image configured
for such a setup. I can even boot the same image on
the bare hardware by tweaking the initramfs-tools
configuration in the installed bullseye image.

Chuck Zmudzinski

Re: Debian 11 installer crashed and reboot

[Chuck Zmudzinski]

On 8/16/2021 4:25 AM, Andrew M.A. Cater wrote:

On Mon, Aug 16, 2021 at 02:30:33PM +0800, John Mok wrote:

Hi all,

Tried to install Debian 11 guest using netinst, but the installer
crashed and reboot automatically.

  Host: Xen 4.11.4 on Debian 10
Guest: Debian 11 (kernel 5.10)

Here is the steps to reproduce the problem:-
1) Either guest BIOS or OVMF boot
2) Select "Expert install" on installation menu

Then, the guest crashed and reboot.

Another try with the installer on Debian Live 11, it immediately
crashed and reboot.

Is it a installer bug or something ?

Thanks a lot.

John Mok


I think people would need many more details as to exactly what happens and
when it reboots.

So: You boot the installer - from just a .iso file / from the .iso flashed
to a USB? Can you show what works and at what point it fails?

The quick suggestion would be that it's something to do with Xen hypervisor
but I'm not in any better position than to guess, since I don't use Xen here.

I do know that none of the tests the media team do involve booting on Xen:
in most cases we prefer to run on real hardware. Occasionally, some of the
tests have to be run on KVM/QEMU. Xen as DomU  is untested then.

All the very best, as ever,

Andy Cater




After some testing of the Debian 11 installer on Xen
(using the debian-11.0.0-amd64-netinst.iso), I find that
this image only supports installation into a Xen PV guest,
the guest always crashes and reboots for either a BIOS
or OVMF boot into an HVM Xen guest.

I also found that the device model fails to start if one
tries to run the xen built-in vnc server for graphical
output. The only solution is to connect to the
Xen console using the xl console command. The -c
switch to xl create did not work to connect to the
Xen console when creating the guest; it could not
find the pty device, so it is necessary to let the guest
complete the startup and then run xl console 
to connect to the console once the VM is running.
This will display the first page of the debian installer
where the language is selected, and from there
Debian 11 can be installed onto a virtual disk.

The following lines in the xl.cfg file enabled the VM
to boot the debian-11.0.0-amd64-netinst.iso as a
PV Xen guest:

bootloader = 'pygrub'
bootloader_args= [ 
'--kernel=install.amd/xen/vmlinuz','--ramdisk=install.amd/xen/initrd.gz' ]


Once installed and booting from the virtual disk, the
bootloader_args setting can be commented out.

After booting from the virtual disk, the grub-pc package can
be added to boot into a Xen HVM guest with BIOS booting,
and if space is left for an efi partition, the grub-efi package
can be installed to boot with OVMF after converting
the partition table of the virtual disk from MBR to GPT.

Also, the vnc server can be used for graphical output
after booting from the virtual disk to run a desktop
environment.

All in all, it is a big pain to install Debian 11 on the Xen
hypervisor using the debian installer, as it seems
supporting Xen installations is not a priority for Debian.

But once it is installed, it works very well on Xen, and
I find I can use the same bullseye image to boot it
as a Xen PV guest, a Xen HVM guest using BIOS boot,
or as a Xen HVM guest using OVMF boot just by tweaking
the xl.cfg configuration file after getting the image configured
for such a setup. I can even boot the same image on
the bare hardware by tweaking the initramfs-tools
configuration in the installed bullseye image.

Chuck Zmudzinski

Re: Debian 11 installer crashed and reboot

[John Mok]

Many thanks Chuck.

On Tue, Aug 17, 2021 at 6:36 AM Chuck Zmudzinski  wrote:
>
> On 8/16/2021 1:40 PM, Chuck Zmudzinski wrote:
> > On 8/16/2021 12:49 PM, Chuck Zmudzinski wrote:
> >> On 8/16/2021 4:25 AM, Andrew M.A. Cater wrote:
> >>> On Mon, Aug 16, 2021 at 02:30:33PM +0800, John Mok wrote:
>  Hi all,
> 
>  Tried to install Debian 11 guest using netinst, but the installer
>  crashed and reboot automatically.
> 
>    Host: Xen 4.11.4 on Debian 10
>  Guest: Debian 11 (kernel 5.10)
> 
>  Here is the steps to reproduce the problem:-
>  1) Either guest BIOS or OVMF boot
>  2) Select "Expert install" on installation menu
> 
>  Then, the guest crashed and reboot.
> 
>  Another try with the installer on Debian Live 11, it immediately
>  crashed and reboot.
> 
>  Is it a installer bug or something ?
> 
>  Thanks a lot.
> 
>  John Mok
> 
> >>> I think people would need many more details as to exactly what
> >>> happens and
> >>> when it reboots.
> >>>
> >>> So: You boot the installer - from just a .iso file / from the .iso
> >>> flashed
> >>> to a USB? Can you show what works and at what point it fails?
> >>>
> >>> The quick suggestion would be that it's something to do with Xen
> >>> hypervisor
> >>> but I'm not in any better position than to guess, since I don't use
> >>> Xen here.
> >>>
> >>> I do know that none of the tests the media team do involve booting
> >>> on Xen:
> >>> in most cases we prefer to run on real hardware. Occasionally, some
> >>> of the
> >>> tests have to be run on KVM/QEMU. Xen as DomU  is untested then.
> >>>
> >>> All the very best, as ever,
> >>>
> >>> Andy Cater
> >>>
> >>
> >> I can say bullseye runs fine as a Xen DomU on both the buster xen-4.11
> >> hypervisor and the bullseye xen-4.14 hypervisor on my workstation, with
> >> either BIOS or OVMF boot, but mine is a bullseye image upgraded from
> >> buster several months ago. I have not tried the bullseye debian
> >> installer
> >> on the xen hypervisor. Maybe for installing bullseye on xen you can try
> >> installing buster first then upgrade it to bullseye if the debian
> >> bullseye
> >> installer does not work on the xen hypervisor.
> >>
> >> Chuck Zmudzinski
> >>
> >
> > I would also ask if you have successfully booted another linux distro
> > (an earlier Debian or another distribution such as ubuntu) on your
> > Debian 10 xen-4.11.4 hypervisor. If not, the problem may be with your
> > xen configuration files on your Debian 10 host system, not with the
> > Debian 11 installer. Maybe some tweaks to the DomU xl.cfg configuration
> > file for your Debian 11 DomU or to the xen hypervisor boot options
> > (usually set in /etc/default/grub) might get it working.
> >
> > Chuck Zmudzinski
> >
>
> I tested the debian-11.0.0-amd64-netinst.iso image on my Debian 10
> xen-4.11.4 hypervisor using both OVMF and BIOS boot using an HVM
> type Xen guest, and in both cases, I can confirm the same behavior as
> the OP: crash and restart after selecting any of the installation options
> from the grub menu that is displayed after booting. Nevertheless, a
> full installation of Debian 11 boots and runs fine in the same environment
> under Xen.
>
> Browsing the files in the iso image, I can see there is a folder for xen:
> /install.amd/xen with three files in it: vmlinuz, initrd.gz, and debian.cfg.
>
> However, these files in the xen folder are never referenced anywhere from
> the menu or submenu items in the /boot/grub/grub.cfg file on the iso image.
> So a workaround would be to create an iso image with a menu entry in the
> /boot/grub/grub.cfg file for an installation on xen, and that grub menu item
> would be configured to load the vmlinuz kernel and initrd.gz image in the
> /install.amd/xen directory.
>
> It might also be possible to escape to the grub command shell
> instead of selecting one of the installation options after booting and
> manually enter the commands to load the xen-enabled kernel and initrd.gz
> images in the /install.amd/xen folder on the netinst iso, but that is not
> very user friendly, and I am not even sure if those images are mainly for
> a PV type Xen guest rather that an HVM type Xen guest that really should
> not need a special kernel or initrd image, but I could be wrong about that.
>
> Also, the debian.cfg file in /install.amd/xen mentions the need to use an
> iso image that supports installation under xen, and it references things
> like multi-arch iso images which I don't think exist anymore in modern
> Debian. Anyway, I suggest the OP read this debian.cfg file and try using
> the suggestions there to get the debian 11 installer running under xen.
>
> To summarize,  I do not think the Debian installer is designed for easy
> installation under Xen, and it is not clear exactly which iso is needed,
> nor is it
> clear which type of Xen guest should be configured (PV or HVM) to enable
> installation of Debian 11 under Xen.
>
> 

Re: Debian 11 installer crashed and reboot

[Chuck Zmudzinski]

On 8/16/2021 1:40 PM, Chuck Zmudzinski wrote:

On 8/16/2021 12:49 PM, Chuck Zmudzinski wrote:

On 8/16/2021 4:25 AM, Andrew M.A. Cater wrote:

On Mon, Aug 16, 2021 at 02:30:33PM +0800, John Mok wrote:

Hi all,

Tried to install Debian 11 guest using netinst, but the installer
crashed and reboot automatically.

  Host: Xen 4.11.4 on Debian 10
Guest: Debian 11 (kernel 5.10)

Here is the steps to reproduce the problem:-
1) Either guest BIOS or OVMF boot
2) Select "Expert install" on installation menu

Then, the guest crashed and reboot.

Another try with the installer on Debian Live 11, it immediately
crashed and reboot.

Is it a installer bug or something ?

Thanks a lot.

John Mok

I think people would need many more details as to exactly what 
happens and

when it reboots.

So: You boot the installer - from just a .iso file / from the .iso 
flashed

to a USB? Can you show what works and at what point it fails?

The quick suggestion would be that it's something to do with Xen 
hypervisor
but I'm not in any better position than to guess, since I don't use 
Xen here.


I do know that none of the tests the media team do involve booting 
on Xen:
in most cases we prefer to run on real hardware. Occasionally, some 
of the

tests have to be run on KVM/QEMU. Xen as DomU  is untested then.

All the very best, as ever,

Andy Cater



I can say bullseye runs fine as a Xen DomU on both the buster xen-4.11
hypervisor and the bullseye xen-4.14 hypervisor on my workstation, with
either BIOS or OVMF boot, but mine is a bullseye image upgraded from
buster several months ago. I have not tried the bullseye debian 
installer

on the xen hypervisor. Maybe for installing bullseye on xen you can try
installing buster first then upgrade it to bullseye if the debian 
bullseye

installer does not work on the xen hypervisor.

Chuck Zmudzinski



I would also ask if you have successfully booted another linux distro
(an earlier Debian or another distribution such as ubuntu) on your
Debian 10 xen-4.11.4 hypervisor. If not, the problem may be with your
xen configuration files on your Debian 10 host system, not with the
Debian 11 installer. Maybe some tweaks to the DomU xl.cfg configuration
file for your Debian 11 DomU or to the xen hypervisor boot options
(usually set in /etc/default/grub) might get it working.

Chuck Zmudzinski



I tested the debian-11.0.0-amd64-netinst.iso image on my Debian 10
xen-4.11.4 hypervisor using both OVMF and BIOS boot using an HVM
type Xen guest, and in both cases, I can confirm the same behavior as
the OP: crash and restart after selecting any of the installation options
from the grub menu that is displayed after booting. Nevertheless, a
full installation of Debian 11 boots and runs fine in the same environment
under Xen.

Browsing the files in the iso image, I can see there is a folder for xen:
/install.amd/xen with three files in it: vmlinuz, initrd.gz, and debian.cfg.

However, these files in the xen folder are never referenced anywhere from
the menu or submenu items in the /boot/grub/grub.cfg file on the iso image.
So a workaround would be to create an iso image with a menu entry in the
/boot/grub/grub.cfg file for an installation on xen, and that grub menu item
would be configured to load the vmlinuz kernel and initrd.gz image in the
/install.amd/xen directory.

It might also be possible to escape to the grub command shell
instead of selecting one of the installation options after booting and
manually enter the commands to load the xen-enabled kernel and initrd.gz
images in the /install.amd/xen folder on the netinst iso, but that is not
very user friendly, and I am not even sure if those images are mainly for
a PV type Xen guest rather that an HVM type Xen guest that really should
not need a special kernel or initrd image, but I could be wrong about that.

Also, the debian.cfg file in /install.amd/xen mentions the need to use an
iso image that supports installation under xen, and it references things
like multi-arch iso images which I don't think exist anymore in modern
Debian. Anyway, I suggest the OP read this debian.cfg file and try using
the suggestions there to get the debian 11 installer running under xen.

To summarize,  I do not think the Debian installer is designed for easy
installation under Xen, and it is not clear exactly which iso is needed, 
nor is it

clear which type of Xen guest should be configured (PV or HVM) to enable
installation of Debian 11 under Xen.

Chuck Zmudzinski

Re: Debian 11 installer crashed and reboot

[Chuck Zmudzinski]

On 8/16/2021 12:49 PM, Chuck Zmudzinski wrote:

On 8/16/2021 4:25 AM, Andrew M.A. Cater wrote:

On Mon, Aug 16, 2021 at 02:30:33PM +0800, John Mok wrote:

Hi all,

Tried to install Debian 11 guest using netinst, but the installer
crashed and reboot automatically.

  Host: Xen 4.11.4 on Debian 10
Guest: Debian 11 (kernel 5.10)

Here is the steps to reproduce the problem:-
1) Either guest BIOS or OVMF boot
2) Select "Expert install" on installation menu

Then, the guest crashed and reboot.

Another try with the installer on Debian Live 11, it immediately
crashed and reboot.

Is it a installer bug or something ?

Thanks a lot.

John Mok

I think people would need many more details as to exactly what 
happens and

when it reboots.

So: You boot the installer - from just a .iso file / from the .iso 
flashed

to a USB? Can you show what works and at what point it fails?

The quick suggestion would be that it's something to do with Xen 
hypervisor
but I'm not in any better position than to guess, since I don't use 
Xen here.


I do know that none of the tests the media team do involve booting on 
Xen:
in most cases we prefer to run on real hardware. Occasionally, some 
of the

tests have to be run on KVM/QEMU. Xen as DomU  is untested then.

All the very best, as ever,

Andy Cater



I can say bullseye runs fine as a Xen DomU on both the buster xen-4.11
hypervisor and the bullseye xen-4.14 hypervisor on my workstation, with
either BIOS or OVMF boot, but mine is a bullseye image upgraded from
buster several months ago. I have not tried the bullseye debian installer
on the xen hypervisor. Maybe for installing bullseye on xen you can try
installing buster first then upgrade it to bullseye if the debian 
bullseye

installer does not work on the xen hypervisor.

Chuck Zmudzinski



I would also ask if you have successfully booted another linux distro
(an earlier Debian or another distribution such as ubuntu) on your
Debian 10 xen-4.11.4 hypervisor. If not, the problem may be with your
xen configuration files on your Debian 10 host system, not with the
Debian 11 installer. Maybe some tweaks to the DomU xl.cfg configuration
file for your Debian 11 DomU or to the xen hypervisor boot options
(usually set in /etc/default/grub) might get it working.

Chuck Zmudzinski

Re: Debian 11 installer crashed and reboot

[Chuck Zmudzinski]

On 8/16/2021 4:25 AM, Andrew M.A. Cater wrote:

On Mon, Aug 16, 2021 at 02:30:33PM +0800, John Mok wrote:

Hi all,

Tried to install Debian 11 guest using netinst, but the installer
crashed and reboot automatically.

  Host: Xen 4.11.4 on Debian 10
Guest: Debian 11 (kernel 5.10)

Here is the steps to reproduce the problem:-
1) Either guest BIOS or OVMF boot
2) Select "Expert install" on installation menu

Then, the guest crashed and reboot.

Another try with the installer on Debian Live 11, it immediately
crashed and reboot.

Is it a installer bug or something ?

Thanks a lot.

John Mok


I think people would need many more details as to exactly what happens and
when it reboots.

So: You boot the installer - from just a .iso file / from the .iso flashed
to a USB? Can you show what works and at what point it fails?

The quick suggestion would be that it's something to do with Xen hypervisor
but I'm not in any better position than to guess, since I don't use Xen here.

I do know that none of the tests the media team do involve booting on Xen:
in most cases we prefer to run on real hardware. Occasionally, some of the
tests have to be run on KVM/QEMU. Xen as DomU  is untested then.

All the very best, as ever,

Andy Cater



I can say bullseye runs fine as a Xen DomU on both the buster xen-4.11
hypervisor and the bullseye xen-4.14 hypervisor on my workstation, with
either BIOS or OVMF boot, but mine is a bullseye image upgraded from
buster several months ago. I have not tried the bullseye debian installer
on the xen hypervisor. Maybe for installing bullseye on xen you can try
installing buster first then upgrade it to bullseye if the debian bullseye
installer does not work on the xen hypervisor.

Chuck Zmudzinski

Re: Debian 11 installer crashed and reboot

[Andrew M.A. Cater]

On Mon, Aug 16, 2021 at 02:30:33PM +0800, John Mok wrote:
> Hi all,
> 
> Tried to install Debian 11 guest using netinst, but the installer
> crashed and reboot automatically.
> 
>  Host: Xen 4.11.4 on Debian 10
> Guest: Debian 11 (kernel 5.10)
> 
> Here is the steps to reproduce the problem:-
> 1) Either guest BIOS or OVMF boot
> 2) Select "Expert install" on installation menu
> 
> Then, the guest crashed and reboot.
> 
> Another try with the installer on Debian Live 11, it immediately
> crashed and reboot.
> 
> Is it a installer bug or something ?
> 
> Thanks a lot.
> 
> John Mok
> 

I think people would need many more details as to exactly what happens and
when it reboots.

So: You boot the installer - from just a .iso file / from the .iso flashed
to a USB? Can you show what works and at what point it fails?

The quick suggestion would be that it's something to do with Xen hypervisor
but I'm not in any better position than to guess, since I don't use Xen here.

I do know that none of the tests the media team do involve booting on Xen: 
in most cases we prefer to run on real hardware. Occasionally, some of the
tests have to be run on KVM/QEMU. Xen as DomU  is untested then.

All the very best, as ever,

Andy Cater

Debian 11 installer crashed and reboot

[John Mok]

Hi all,

Tried to install Debian 11 guest using netinst, but the installer
crashed and reboot automatically.

 Host: Xen 4.11.4 on Debian 10
Guest: Debian 11 (kernel 5.10)

Here is the steps to reproduce the problem:-
1) Either guest BIOS or OVMF boot
2) Select "Expert install" on installation menu

Then, the guest crashed and reboot.

Another try with the installer on Debian Live 11, it immediately
crashed and reboot.

Is it a installer bug or something ?

Thanks a lot.

John Mok