test: cross arch build/boot smoke test for ppc64le,s390x by mvo5 · Pull Request #2069 · osbuild/images

mvo5 · 2025-12-04T16:29:23Z

test: run cross arch smoke test on ppc64le, s390x

This commit adds a minimal smoke test for centos-10/qcow2 that
will create a qcow2 and boot it. It demos how we could do more
cross arch testing and keeps the qemu code tested.

We need to decide if we want to do more tests like this, its
a trade-off. OTOH its nice to have (some) assurance that our
images boot (we did break s390/ppc64 partition tables in the
past by accident). OTOH its a bit of a pain when something fails
to figure out of it is failing because of some qemu incompatibilites
or because there is a real issue in the image.

Having this at least for local testing/validation is probably
useful though.

Note that the s390x test needs qemu-user >= 7.2-rc2 so this will
not run currently (I did run it on my local machine and it works
but the kdump service takes > 5min to startup).

vmtest: add cross arch boot support for s390x/ppc64le

This adds basic support to boot ppc64le/s390x machine. This
allows to do:

$ ./test/scripts/build-image --arch ppc64le centos-10 qcow2 cfgpath
$ ./test/scripts/boot-image ./build/centos_10-ppc64le-qcow2-empty

to easily boot test machines that are not easily available otherwise.

build-image: add support for (optional) --arch to build

This is useful to do basic testing for architectures like s390x
or ppc64le that are not available easily (or even for a quick
local aarch64 test).

vmtest/vm.py

test/scripts/build-image

bcl

Other than that one question this looks good.

thozza

Thanks for this; I think that it makes sense.

That said, the default behavior of some of the scripts changes AFAICT, and I am not convinced that it is necessarily intentional. Requesting changes until this is sorted out.

test/scripts/build-image

cmd/build/main.go

.github/workflows/tests.yml

This is useful to do basic testing for architectures like s390x or ppc64le that are not available easily (or even for a quick local aarch64 test).

This adds basic support to boot ppc64le/s390x machine. This allows to do: ``` $ ./test/scripts/build-image --arch ppc64le centos-10 qcow2 cfgpath $ ./test/scripts/boot-image ./build/centos_10-ppc64le-qcow2-empty ``` to easily boot test machines that are not easily available otherwise.

This commit adds a minimal smoke test for centos-10/qcow2 that will create a qcow2 and boot it. It demos how we could do more cross arch testing and keeps the qemu code tested. We need to decide if we want to do more tests like this, its a trade-off. OTOH its nice to have (some) assurance that our images boot (we did break s390/ppc64 partition tables in the past by accident). OTOH its a bit of a pain when something fails to figure out of it is failing because of some qemu incompatibilites or because there is a real issue in the image. Having this at least for local testing/validation is probably useful though. Note that the s390x test needs qemu-user >= 7.2-rc2 so this will not run currently (I did run it on my local machine and it works but the kdump service takes > 5min to startup).

.github/workflows/tests.yml

vmtest/vm.py

thozza

Amazing, thank you a lot!

thozza · 2025-12-17T08:52:39Z

cmd/build/main.go

+	if archName == "" {
+		archName = arch.Current().String()
+	}
+	archi, err := distribution.GetArch(archName)


Nitpick: I don't have any problem with the rename of arch -> archi, just wanted to point it out in case this was not intentional.

Its needed because we also import arch so if we have arch there the arch module is shadowed (and a few lines below iirc this code access the arch module now).

lzap · 2025-12-18T09:34:28Z

Somehow, in my PR I rebased on top of this commit CICD started erroring out with:

error obtaining VCS status: exit status 128
Use -buildvcs=false to disable VCS stamping.

My theory is that an environmental variable GOFLAGS is set for that test, but then in my PR I actually override that variable with some other value (tags) so I shadow it. Therefore I only see the problem in my own PR. I guess you can disregard this comment then :-)

#2100

lzap · 2025-12-19T20:56:17Z

vmtest/vm.py

-            "-device", "e1000,netdev=net.0",
            "-qmp", f"unix:{self._qmp_socket},server,nowait",
+            # boot
+            "-drive", f"file={self._img},if=none,id=disk0,format=qcow2",


This hunk sneaked into main somehow and this does not appear to be correct. It renders as:

['qemu-system-x86_64', '-M', 'accel=kvm', '-cpu', 'host', '-device', 'virtio-scsi-pci,id=scsi', '-device', 'scsi-hd,drive=disk0', '-bios', '/usr/share/OVMF/OVMF_CODE.fd', '-m', '2048', '-serial', 'stdio', '-monitor', 'none', '-device', 'virtio-net-pci,netdev=net.0,id=net.0', '-netdev', 'user,id=net.0,hostfwd=tcp::49925-:22', '-qmp', 'unix:/tmp/vmtest-v614uhg7-disk.img/qmp.socket,server,nowait', '-drive', 'file=/var/tmp/tmpp40bxujp/disk.img,if=none,id=disk0,format=qcow2', '-nographic', '-cdrom', PosixPath('/var/tmp/tmpdk5gg6b6/netinst.iso'), PosixPath('/var/tmp/tmpp40bxujp/disk.img')]

Which ends with QEMU complaining about raw image not being qcow2. That can be fixed, but then the issue is the image is passed twice on the command line - once via -device and once via argument. When I remove the argument, then QEMU does not boot there is some problem with SCSI setup.

Another problem is few lines below when -cdrom is appended with Path object, but that is an easy fix with str.

I will look into this later next week. Actually, let me try to do a quick fix and see how many tests are fixed: #2103

Changes with 0.231.0 ---------------- - Drop iommu.strict=0 from aarch64 EC2 images (osbuild/images#2090) - Author: Achilleas Koutsou, Reviewers: Lukáš Zapletal, Simon de Vlieger, Tomáš Hozza - bootc: Fix selinux labeling when using separate build container (osbuild/images#2092) - Author: Alexander Larsson, Reviewers: Lukáš Zapletal, Simon de Vlieger, Tomáš Hozza - ci: resolve dubious ownership for git (osbuild/images#2100) - Author: Lukáš Zapletal, Reviewers: Achilleas Koutsou, Michael Vogt - data: import RH v4 key on rhel-10.1+ only (osbuild/images#2097) - Author: Michael Vogt, Reviewers: Lukáš Zapletal, Simon de Vlieger - distrodefs: drop `use_syslinux` as it has no effect (osbuild/images#2088) - Author: Michael Vogt, Reviewers: Lukáš Zapletal, Simon de Vlieger - fedora: /boot on btrfs for Fedora Cloud 44 (HMS-9737) (osbuild/images#1960) - Author: Simon de Vlieger, Reviewers: Lukáš Zapletal, Neal Gompa (ニール・ゴンパ), Tomáš Hozza - fedora: add `server-network-installer` (osbuild/images#2094) - Author: Simon de Vlieger, Reviewers: Lukáš Zapletal, Tomáš Hozza - installer: only install `syslinux` when needed (osbuild/images#2089) - Author: Simon de Vlieger, Reviewers: Lukáš Zapletal, Tomáš Hozza - manifest: add `set -e` to `bootc switch...` kickstart %post (osbuild/images#2093) - Author: Michael Vogt, Reviewers: Lukáš Zapletal, Simon de Vlieger, Tomáš Hozza - many: include legal and license files in ISO (osbuild/images#2099) - Author: Simon de Vlieger, Reviewers: Lukáš Zapletal, Tomáš Hozza - osbuild: drop `valueIn` helper (osbuild/images#2086) - Author: Michael Vogt, Reviewers: Brian C. Lane, Simon de Vlieger - readme: update link to image definitions (osbuild/images#2070) - Author: Anna Vítová, Reviewers: Achilleas Koutsou, Lukáš Zapletal, Simon de Vlieger, Tomáš Hozza - test: cross arch build/boot smoke test for ppc64le,s390x (osbuild/images#2069) - Author: Michael Vogt, Reviewers: Lukáš Zapletal, Simon de Vlieger, Tomáš Hozza

mvo5 requested a review from supakeen December 4, 2025 16:29

mvo5 force-pushed the vmtest-add-ppc64-s390x branch from f05edae to b85084f Compare December 8, 2025 11:28

mvo5 changed the title ~~test: allow cross arch build/test~~ test: allow cross arch build/test for ppc64le Dec 8, 2025

mvo5 force-pushed the vmtest-add-ppc64-s390x branch from b85084f to f07486e Compare December 8, 2025 11:42

lzap reviewed Dec 8, 2025

View reviewed changes

vmtest/vm.py Outdated Show resolved Hide resolved

mvo5 force-pushed the vmtest-add-ppc64-s390x branch from f07486e to 456beba Compare December 10, 2025 11:30

mvo5 changed the title ~~test: allow cross arch build/test for ppc64le~~ test: allow cross arch build/test for ppc64le,s390x Dec 10, 2025

mvo5 force-pushed the vmtest-add-ppc64-s390x branch 6 times, most recently from 0f50022 to f27edc2 Compare December 10, 2025 12:28

mvo5 mentioned this pull request Dec 10, 2025

test: add --keep-booted option to boot-image #2079

Merged

mvo5 force-pushed the vmtest-add-ppc64-s390x branch 5 times, most recently from af57b3a to 52e8aba Compare December 10, 2025 13:39

mvo5 changed the title ~~test: allow cross arch build/test for ppc64le,s390x~~ [RFC] test: allow cross arch build/test for ppc64le,s390x Dec 10, 2025

mvo5 force-pushed the vmtest-add-ppc64-s390x branch from 52e8aba to 099532d Compare December 10, 2025 14:26

mvo5 marked this pull request as ready for review December 10, 2025 15:14

mvo5 requested review from a team, achilleas-k and thozza as code owners December 10, 2025 15:14

bcl reviewed Dec 11, 2025

View reviewed changes

test/scripts/build-image Show resolved Hide resolved

bcl previously approved these changes Dec 11, 2025

View reviewed changes

thozza requested changes Dec 11, 2025

View reviewed changes

test/scripts/build-image Outdated Show resolved Hide resolved

cmd/build/main.go Outdated Show resolved Hide resolved

.github/workflows/tests.yml Outdated Show resolved Hide resolved

.github/workflows/tests.yml Show resolved Hide resolved

mvo5 dismissed bcl’s stale review via f5418be December 12, 2025 09:48

mvo5 force-pushed the vmtest-add-ppc64-s390x branch from 099532d to f5418be Compare December 12, 2025 09:48

mvo5 added 2 commits December 12, 2025 10:51

build-image: add support for (optional) --arch to build

fc79dd6

This is useful to do basic testing for architectures like s390x or ppc64le that are not available easily (or even for a quick local aarch64 test).

mvo5 force-pushed the vmtest-add-ppc64-s390x branch 3 times, most recently from 718ed8e to c57a64e Compare December 12, 2025 09:59

mvo5 changed the title ~~[RFC] test: allow cross arch build/test for ppc64le,s390x~~ test: cross arch build/boot smoke test for ppc64le,s390x Dec 12, 2025

mvo5 force-pushed the vmtest-add-ppc64-s390x branch 2 times, most recently from 433584b to 8aa85f8 Compare December 12, 2025 10:09

mvo5 mentioned this pull request Dec 12, 2025

test: add some tips around {build,boot}-image for manual tests #2085

Merged

mvo5 force-pushed the vmtest-add-ppc64-s390x branch from 8aa85f8 to 63263ad Compare December 15, 2025 08:01

lzap approved these changes Dec 15, 2025

View reviewed changes

.github/workflows/tests.yml Show resolved Hide resolved

vmtest/vm.py Show resolved Hide resolved

supakeen approved these changes Dec 17, 2025

View reviewed changes

supakeen enabled auto-merge December 17, 2025 08:47

thozza approved these changes Dec 17, 2025

View reviewed changes

supakeen added this pull request to the merge queue Dec 17, 2025

Merged via the queue into osbuild:main with commit 240ce62 Dec 17, 2025
25 checks passed

lzap reviewed Dec 19, 2025

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

test: cross arch build/boot smoke test for ppc64le,s390x#2069

test: cross arch build/boot smoke test for ppc64le,s390x#2069
supakeen merged 3 commits intoosbuild:mainfrom
mvo5:vmtest-add-ppc64-s390x

mvo5 commented Dec 4, 2025 •

edited

Loading

Uh oh!

Uh oh!

Uh oh!

bcl left a comment

Uh oh!

thozza left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

thozza left a comment

Uh oh!

thozza Dec 17, 2025

Uh oh!

mvo5 Dec 17, 2025

Uh oh!

Uh oh!

lzap commented Dec 18, 2025 •

edited

Loading

Uh oh!

lzap Dec 19, 2025 •

edited

Loading

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Conversation

mvo5 commented Dec 4, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

bcl left a comment

Choose a reason for hiding this comment

Uh oh!

thozza left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

thozza left a comment

Choose a reason for hiding this comment

Uh oh!

thozza Dec 17, 2025

Choose a reason for hiding this comment

Uh oh!

mvo5 Dec 17, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

lzap commented Dec 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

lzap Dec 19, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

mvo5 commented Dec 4, 2025 •

edited

Loading

lzap commented Dec 18, 2025 •

edited

Loading

lzap Dec 19, 2025 •

edited

Loading