Bug#987441: s390x installation bugs

[Valentin Vidic]

On Sun, Aug 01, 2021 at 09:45:00PM +0200, Valentin Vidic wrote:
> Thanks, that does sound similar to what I was getting there. I will try
> to see if it still happens with the latest installer. And since it
> crashes on start I had no way to access the logs or dmesg of the
> machine. Perhaps there is some installer option to help debug this kind
> of thing?

Just tested the rc3 installation with qemu-system-s390x. Installation
went fairly quickly and without any problems. Great work everyone and
happy release :)))

-- 
Valentin

Bug#987441: s390x installation bugs

[Valentin Vidic]

On Sun, Aug 01, 2021 at 05:10:07PM +0200, Cyril Brulebois wrote:
> Valentin Vidic  (2021-08-01):
> > No problem, I was not able to reproduce this reliably or get a core
> > dump for this crash. It could just be an emulation problem with qemu
> > or some timing issue for the first installer step. If there is no
> > update on this problem I think we can even close it for now.
> 
> Speaking of the first step, did anyone mention #987368 before, now fixed
> in udpkg?

Thanks, that does sound similar to what I was getting there. I will try
to see if it still happens with the latest installer. And since it
crashes on start I had no way to access the logs or dmesg of the
machine. Perhaps there is some installer option to help debug this kind
of thing?

-- 
Valentin

Bug#987441: s390x installation bugs

[Cyril Brulebois]

Valentin Vidic  (2021-08-01):
> No problem, I was not able to reproduce this reliably or get a core
> dump for this crash. It could just be an emulation problem with qemu
> or some timing issue for the first installer step. If there is no
> update on this problem I think we can even close it for now.

Speaking of the first step, did anyone mention #987368 before, now fixed
in udpkg?

(The BTS seems to be sad right now, and I'd like to avoid diving into
bug mail.)


Cheers,
-- 
Cyril Brulebois (k...@debian.org)
D-I release manager -- Release team member -- Freelance Consultant


signature.asc
Description: PGP signature

Bug#987441: s390x installation bugs

[Cyril Brulebois]

Cyril Brulebois  (2021-05-04):
> Valentin Vidic  (2021-05-03):
> > Thanks, here is another one for s390x, should be relatively simple if
> > you wish to link it here:
> > 
> > linux: Debian installation fails in qemu-system-s390x due to missing 
> > virtio_blk module
> > https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=988005
> 
> Thanks, but I think I'll just keep your message as a link to it: we
> don't have to do anything on the installer's side except make sure it's
> considered in a further upload of src:linux (be it for 11.0 or 11.1). :)

Following up on that, the merge happened in the sid branch:

kibi@tokyo:~/debian-kernel/linux.git$ git describe 
fb657acabf6aa129889d53c85ccae246a99007b4
debian/5.10.28-1-36-gfb657acabf6a

and we should have that in the next linux upload, which should land in
bullseye before the release.


Cheers,
-- 
Cyril Brulebois (k...@debian.org)
D-I release manager -- Release team member -- Freelance Consultant


signature.asc
Description: PGP signature

Bug#987441: s390x installation bugs

[Cyril Brulebois]

Valentin Vidic  (2021-05-03):
> Thanks, here is another one for s390x, should be relatively simple if
> you wish to link it here:
> 
> linux: Debian installation fails in qemu-system-s390x due to missing 
> virtio_blk module
> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=988005

Thanks, but I think I'll just keep your message as a link to it: we
don't have to do anything on the installer's side except make sure it's
considered in a further upload of src:linux (be it for 11.0 or 11.1). :)


Cheers,
-- 
Cyril Brulebois (k...@debian.org)
D-I release manager -- Release team member -- Freelance Consultant


signature.asc
Description: PGP signature

Bug#987441: s390x installation bugs

[Valentin Vidic]

On Mon, May 03, 2021 at 08:36:58AM +0200, Cyril Brulebois wrote:
> Also adding to the list of bugs to keep an eye on (again, possibly not
> blocking the release on its being resolved; we could have the issue
> listed in errata, and possibly fixed in a point release).

Thanks, here is another one for s390x, should be relatively simple if
you wish to link it here:

linux: Debian installation fails in qemu-system-s390x due to missing virtio_blk 
module
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=988005

-- 
Valentin

Bug#961056: Bug#926539: Bug#987441: s390x installation bugs

[John Paul Adrian Glaubitz]

Hi!

On 5/3/21 12:21 PM, Valentin Vidic wrote:
>>> I'd suggest at least retitling the bug report to mention s390x (release
>>> arch, affected) instead of sparc64 (port arch, no longer affected), to
>>> lower the chances people could overlook this issue, thinking it's only
>>> about a port arch.
>>
>> We could also unmerge #926539 and #961056 again, then close the former bug
>> which was sparc64-specific.
> 
> I have unmerged the bugs now, so the sparc one can be closed.

Alright, done.

Thanks,
Adrian

-- 
 .''`.  John Paul Adrian Glaubitz
: :' :  Debian Developer - glaub...@debian.org
`. `'   Freie Universitaet Berlin - glaub...@physik.fu-berlin.de
  `-GPG: 62FF 8A75 84E0 2956 9546  0006 7426 3B37 F5B5 F913

Bug#926539: Bug#987441: s390x installation bugs

[Valentin Vidic]

On Mon, May 03, 2021 at 08:58:02AM +0200, John Paul Adrian Glaubitz wrote:
> > The same issue exists on s390x but isn't apparently going to get fixed
> > so we need to have d-i be smarter (hence the merge request)?
> 
> Seems so.

QEMU console might get fixed in the kernel, but it looks like LPAR could
have a similar problem (don't have access to test this). So it seems
better (and future proof) to fix this on the Debian side too. I have
updated the merge request to trigger the new code only on s390x as
suggested:

https://salsa.debian.org/installer-team/rootskel/-/merge_requests/2

> > I'd suggest at least retitling the bug report to mention s390x (release
> > arch, affected) instead of sparc64 (port arch, no longer affected), to
> > lower the chances people could overlook this issue, thinking it's only
> > about a port arch.
> 
> We could also unmerge #926539 and #961056 again, then close the former bug
> which was sparc64-specific.

I have unmerged the bugs now, so the sparc one can be closed.

-- 
Valentin

Bug#926539: Bug#987441: s390x installation bugs

[John Paul Adrian Glaubitz]

Hi!

On 5/3/21 8:36 AM, Cyril Brulebois wrote:
> From skimming through the bug log, it seems it was initially a sparc64
> problem, that was fixed in the kernel (inconsistent naming) eventually.

Correct.

> The same issue exists on s390x but isn't apparently going to get fixed
> so we need to have d-i be smarter (hence the merge request)?

Seems so.

> I'd suggest at least retitling the bug report to mention s390x (release
> arch, affected) instead of sparc64 (port arch, no longer affected), to
> lower the chances people could overlook this issue, thinking it's only
> about a port arch.

We could also unmerge #926539 and #961056 again, then close the former bug
which was sparc64-specific.

Adrian

-- 
 .''`.  John Paul Adrian Glaubitz
: :' :  Debian Developer - glaub...@debian.org
`. `'   Freie Universitaet Berlin - glaub...@physik.fu-berlin.de
  `-GPG: 62FF 8A75 84E0 2956 9546  0006 7426 3B37 F5B5 F913

Bug#926539: Bug#987441: s390x installation bugs

[Cyril Brulebois]

Control: block 987441 by 926539
Control: block 987441 by 987788

Hi Valentin,

Thanks for your suggestions.

Valentin Vidic  (2021-05-02):
> Probably not critical, but maybe these installation bugs on s390x
> could be fixed for the release?
> 
> rootskel: steal-ctty no longer works on at least sparc64
> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=926539

From skimming through the bug log, it seems it was initially a sparc64
problem, that was fixed in the kernel (inconsistent naming) eventually.
The same issue exists on s390x but isn't apparently going to get fixed
so we need to have d-i be smarter (hence the merge request)?

I'd suggest at least retitling the bug report to mention s390x (release
arch, affected) instead of sparc64 (port arch, no longer affected), to
lower the chances people could overlook this issue, thinking it's only
about a port arch.

I know it's been around for a while, but trying to get that merged right
now doesn't feeel right; trying to make the change only kick in on s390x
(in an attempt to fix s390x while not risking regressing elsewhere)
might eat up some time (lots of conditionals there), but it would feel a
little safer.

Would it make sense to have two for loops (one version “after” for s390x
only, and one version “before” for all archs) as a temporary measure for
bullseye? We could then switch all archs to the “after” version at the
beginning of the bookworm release cycle and deal with possible
regressions then?

In any cases, I'm happy to keep an eye on #926539 via #987441, even if
we don't block the release on it.

> debian-installer: qemu-system-s390x installation fails due to segfault in 
> main-menu
> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=987788

Yes, it seems works keeping an eye on. I had asked on tips on how to
test s390x but haven't had a chance to follow up on this yet.

Also adding to the list of bugs to keep an eye on (again, possibly not
blocking the release on its being resolved; we could have the issue
listed in errata, and possibly fixed in a point release).


Cheers,
-- 
Cyril Brulebois (k...@debian.org)
D-I release manager -- Release team member -- Freelance Consultant


signature.asc
Description: PGP signature

Bug#987441: s390x installation bugs

[Valentin Vidic]

Hi,

Probably not critical, but maybe these installation bugs on s390x could
be fixed for the release?

rootskel: steal-ctty no longer works on at least sparc64
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=926539

debian-installer: qemu-system-s390x installation fails due to segfault in 
main-menu
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=987788

-- 
Valentin