github/workflows: Enable test-in-svsm for NOCC by luigix25 · Pull Request #823 · coconut-svsm/svsm

luigix25 · 2025-09-26T08:40:55Z

Enable test-in-svsm for NOCC and run them in the CI to make sure that a PR doesn't break tests.

Suggested-by: Stefano Garzarella sgarzare@redhat.com

.github/workflows/rust.yml

luigix25 · 2025-09-29T14:52:53Z

I modified this PR, now it's possible to run test-in-svsm for nocc.
I modified the CI so that it executes those tests, unfortunately all tests can be performed:
all tests requiring snp capabilities or mmio have to be skipped.

/cc @osteffenrh who is the author of the original script

luigix25 · 2025-09-29T15:12:01Z

update: if some of the tests doesn't pass (and panics the kernel) the GH action will run until timeout (180s).
Investigating..

stefano-garzarella

Cool, this is nice, maybe it will slow down a bit our CI, but makes sense to me to run tests at least in native, hoping one day we will have SNP/TDX machines for our CI.

Makefile

kernel/src/cpu/vc.rs

.github/workflows/qemu.yml

stefano-garzarella · 2025-09-29T15:27:46Z

update: if some of the tests doesn't pass (and panics the kernel) the GH action will run until timeout (180s). Investigating..

@luigix25 IIRC @osteffenrh added the timeout in scripts/test-qemu-nocc-boot.sh because QEMU was never exiting with just the boot. But with test-in-svsm IIRC, QEMU will exit at the end of the tests, so maybe we don't need that script anymore. (And if we need, maybe we need to update the name)

luigix25 · 2025-09-29T15:32:33Z

update: if some of the tests doesn't pass (and panics the kernel) the GH action will run until timeout (180s). Investigating..

@luigix25 IIRC @osteffenrh added the timeout in scripts/test-qemu-nocc-boot.sh because QEMU was never exiting with just the boot. But with test-in-svsm IIRC, QEMU will exit at the end of the tests, so maybe we don't need that script anymore. (And if we need, maybe we need to update the name)

If everything goes well, then test-in-svsm exits correctly, but if something goes wrong it panics and qemu remains open. I'm modifying Oliver's script to detect a panic.

stefano-garzarella · 2025-09-29T15:47:53Z

update: if some of the tests doesn't pass (and panics the kernel) the GH action will run until timeout (180s). Investigating..

@luigix25 IIRC @osteffenrh added the timeout in scripts/test-qemu-nocc-boot.sh because QEMU was never exiting with just the boot. But with test-in-svsm IIRC, QEMU will exit at the end of the tests, so maybe we don't need that script anymore. (And if we need, maybe we need to update the name)

If everything goes well, then test-in-svsm exits correctly, but if something goes wrong it panics and qemu remains open. I'm modifying Oliver's script to detect a panic.

Strange, in the panic handler we have platform::halt(); (

svsm/kernel/src/svsm.rs

Line 422 in 579de9f

platform::halt();

), so I'm expecting QEMU to exit on hlt, no?

stefano-garzarella · 2025-09-29T15:52:25Z

update: if some of the tests doesn't pass (and panics the kernel) the GH action will run until timeout (180s). Investigating..

@luigix25 IIRC @osteffenrh added the timeout in scripts/test-qemu-nocc-boot.sh because QEMU was never exiting with just the boot. But with test-in-svsm IIRC, QEMU will exit at the end of the tests, so maybe we don't need that script anymore. (And if we need, maybe we need to update the name)

If everything goes well, then test-in-svsm exits correctly, but if something goes wrong it panics and qemu remains open. I'm modifying Oliver's script to detect a panic.

Strange, in the panic handler we have platform::halt(); (

svsm/kernel/src/svsm.rs

Line 422 in 579de9f

platform::halt();

), so I'm expecting QEMU to exit on hlt, no?

no, so maybe we should call testing:exit(), we need to write in the QEMU_EXIT_PORT

luigix25 · 2025-09-29T15:56:44Z

Strange, in the panic handler we have platform::halt(); (

svsm/kernel/src/svsm.rs

Line 422 in 579de9f

platform::halt();

), so I'm expecting QEMU to exit on hlt, no?

no, so maybe we should call testing:exit(), we need to write in the QEMU_EXIT_PORT

Exactly, you have to write in the QEMU_EXIT_PORT. You mean calling it every time the kernel panics?

stefano-garzarella · 2025-09-29T16:00:37Z

no, so maybe we should call testing:exit(), we need to write in the QEMU_EXIT_PORT

Exactly, you have to write in the QEMU_EXIT_PORT. You mean calling it every time the kernel panics?

IMO we should do something similar to debug_break(), so add a new function to call there, where we do what we have in kernel/src/testing.rs::exit(). (Maybe you can factor out the write to QEMU_EXIT_PORT in a pub function and call it on both exit() and panic handler.

kernel/src/testing.rs

osteffenrh · 2025-09-30T08:12:29Z

@luigix25 IIRC @osteffenrh added the timeout in scripts/test-qemu-nocc-boot.sh because QEMU was never exiting with just the boot.

Also the tests (or SVSM) might run into an infinite loop or never terminate for some other reason.

stefano-garzarella · 2025-10-01T08:43:11Z

@luigix25 we just merged #775 so I think we can now rebase and enable also virtio-drivers feature, right?

luigix25 · 2025-10-01T08:45:53Z

@luigix25 we just merged #775 so I think we can now rebase and enable also virtio-drivers feature, right?

Thanks! yep, I am already working on this and some minor fixes

luigix25 · 2025-10-01T09:24:34Z

Rebased on latest upstream, picking up MMIO native PR.

A write to the QEMU exit port is performed whenever the test exits, successfully or panic.
This is to allow QEMU to close.

kernel/src/utils/util.rs

Makefile

scripts/test-in-svsm.sh

.github/workflows/qemu.yml

kernel/src/testing.rs

kernel/src/utils/util.rs

scripts/test-in-svsm.sh

luigix25 · 2025-10-06T16:15:41Z

Addressed Stefano's comments:

I moved qemu_write_exit to kernel/src/testing.rs
Removed the stub, it is not necessary anymore.
Refactored qemu_write_exit: there is no need to differentiate per platform anymore.
Added comment explaining the QEMU value
Introduced a commit for QEMU exit value

kernel/src/testing.rs

kernel/src/svsm.rs

scripts/test-in-svsm.sh

.github/workflows/qemu.yml

luigix25 · 2025-10-06T18:19:19Z

Reordered commits according to Stefano's suggestions.
Reworded a commit message
Very minor changes to bash and yaml

stefano-garzarella

LGTM, just a minor comment.

kernel/src/testing.rs

…uid` Test `test_has_memory_encryption_info_cpuid` can't be run on `native` platform. Add a check to prevent its execution. Signed-off-by: Luigi Leonardi <leonardi@redhat.com>

`exit` function, that writes to the QEMU port, uses GHCB unconditionally that is SNP only. Use `SVSM_PLATFORM` abstraction to perform the io operation according to the platform used. Tested only on native and SNP. Signed-off-by: Luigi Leonardi <leonardi@redhat.com>

QEMU, when a value is written in the exit port, returns `(value << 1) | 1`. This means that writing 0, it will return 1. This would create an ambiguity because 1 could also mean a failure in QEMU launch (i.e. error in the parameters). Change success value to 0x10. Suggested-by: Oliver Steffen <osteffen@redhat.com> Signed-off-by: Luigi Leonardi <leonardi@redhat.com>

…c handler When an assert is hit during testing, it can be convenient to close QEMU immediately. This allows test failures to be detected quickly using specific return values. This requires writing to the QEMU exit port. Factor out `qemu_write_exit` from the `exit` function, allowing it to be reused in both the panic handler and the `exit` function. Them, add a call to `qemu_write_exit` in the panic handler to terminate QEMU with a failure code when a panic occurs during tests. Signed-off-by: Luigi Leonardi <leonardi@redhat.com>

Add necessary parameters to Makefile and `test-in-svsm` script to run tests in a nocc environment. Usage: `TEST_ARGS=--nocc make test-in-svsm` Signed-off-by: Luigi Leonardi <leonardi@redhat.com>

Build and run test images in the CI to make sure that a PR doesn't break tests. Suggested-by: Stefano Garzarella <sgarzare@redhat.com> Signed-off-by: Luigi Leonardi <leonardi@redhat.com>

luigix25 · 2025-10-07T08:20:18Z

Moved the enum to the next commit
All commit messages are not longer than 75 columns

stefano-garzarella

@joergroedel I think we can merge this, but I'll wait for your review.

Note the CI time is not affected by this PR, since running tests is just 2 sec (see Run test-in-svsm in QEMU in https://github.com/coconut-svsm/svsm/actions/runs/18306479849/job/52124795547?pr=823)

luigix25 · 2025-10-07T08:37:08Z

Thanks for the review!!

stefano-garzarella reviewed Sep 26, 2025

View reviewed changes

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

luigix25 force-pushed the ci_test branch from 5af9015 to ee1986e Compare September 29, 2025 14:49

luigix25 changed the title ~~github/workflows: Build test images~~ github/workflows: Enable test-in-svsm for NOCC Sep 29, 2025

stefano-garzarella reviewed Sep 29, 2025

View reviewed changes

Makefile Outdated Show resolved Hide resolved

kernel/src/cpu/vc.rs Outdated Show resolved Hide resolved

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

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

osteffenrh reviewed Sep 30, 2025

View reviewed changes

kernel/src/testing.rs Outdated Show resolved Hide resolved

osteffenrh mentioned this pull request Sep 30, 2025

kernel/platform: Add mmio support for native platform #775

Merged

joergroedel self-requested a review September 30, 2025 15:08

joergroedel added the in-review PR is under active review and not yet approved label Sep 30, 2025

stefano-garzarella added the needs-rebase The PR needs to be rebased to the latest upstream branch label Oct 1, 2025

luigix25 force-pushed the ci_test branch from ee1986e to abb96a9 Compare October 1, 2025 09:22

luigix25 removed the needs-rebase The PR needs to be rebased to the latest upstream branch label Oct 1, 2025

osteffenrh reviewed Oct 1, 2025

View reviewed changes

kernel/src/utils/util.rs Outdated Show resolved Hide resolved

luigix25 force-pushed the ci_test branch from abb96a9 to 82d2277 Compare October 1, 2025 10:05

stefano-garzarella reviewed Oct 1, 2025

View reviewed changes

luigix25 force-pushed the ci_test branch 2 times, most recently from 2e8f39d to 5b45026 Compare October 1, 2025 12:42

stefano-garzarella reviewed Oct 6, 2025

View reviewed changes

kernel/src/testing.rs Show resolved Hide resolved

kernel/src/utils/util.rs Outdated Show resolved Hide resolved

kernel/src/utils/util.rs Outdated Show resolved Hide resolved

scripts/test-in-svsm.sh Outdated Show resolved Hide resolved

luigix25 force-pushed the ci_test branch from 5b45026 to 6f8ee9c Compare October 6, 2025 16:10

stefano-garzarella reviewed Oct 6, 2025

View reviewed changes

kernel/src/testing.rs Show resolved Hide resolved

kernel/src/testing.rs Show resolved Hide resolved

kernel/src/svsm.rs Show resolved Hide resolved

scripts/test-in-svsm.sh Outdated Show resolved Hide resolved

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

luigix25 force-pushed the ci_test branch from 6f8ee9c to 67b1e9d Compare October 6, 2025 18:17

stefano-garzarella reviewed Oct 7, 2025

View reviewed changes

kernel/src/testing.rs Show resolved Hide resolved

luigix25 added 6 commits October 7, 2025 04:17

kernel/cpu: Add platform check to `test_has_memory_encryption_info_cp…

e0ace32

…uid` Test `test_has_memory_encryption_info_cpuid` can't be run on `native` platform. Add a check to prevent its execution. Signed-off-by: Luigi Leonardi <leonardi@redhat.com>

scripts: enable nocc for test-in-svsm

1b78f37

Add necessary parameters to Makefile and `test-in-svsm` script to run tests in a nocc environment. Usage: `TEST_ARGS=--nocc make test-in-svsm` Signed-off-by: Luigi Leonardi <leonardi@redhat.com>

github/workflows: Run test-in-svsm

96f8eea

Build and run test images in the CI to make sure that a PR doesn't break tests. Suggested-by: Stefano Garzarella <sgarzare@redhat.com> Signed-off-by: Luigi Leonardi <leonardi@redhat.com>

luigix25 force-pushed the ci_test branch from 67b1e9d to 96f8eea Compare October 7, 2025 08:18

stefano-garzarella approved these changes Oct 7, 2025

View reviewed changes

stefano-garzarella added ready-to-merge PR is ready for merging into main branch and removed in-review PR is under active review and not yet approved labels Oct 7, 2025

joergroedel approved these changes Oct 16, 2025

View reviewed changes

joergroedel merged commit 32c1e6f into coconut-svsm:main Oct 16, 2025
4 checks passed

luigix25 deleted the ci_test branch October 16, 2025 09:46

Conversation

luigix25 commented Sep 26, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

luigix25 commented Sep 29, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

luigix25 commented Sep 29, 2025

Uh oh!

stefano-garzarella left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

stefano-garzarella commented Sep 29, 2025

Uh oh!

luigix25 commented Sep 29, 2025

Uh oh!

stefano-garzarella commented Sep 29, 2025

Uh oh!

stefano-garzarella commented Sep 29, 2025

Uh oh!

luigix25 commented Sep 29, 2025

Uh oh!

stefano-garzarella commented Sep 29, 2025

Uh oh!

Uh oh!

osteffenrh commented Sep 30, 2025

Uh oh!

stefano-garzarella commented Oct 1, 2025

Uh oh!

luigix25 commented Oct 1, 2025

Uh oh!

luigix25 commented Oct 1, 2025

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

luigix25 commented Oct 6, 2025

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

luigix25 commented Oct 6, 2025

Uh oh!

stefano-garzarella left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

luigix25 commented Oct 7, 2025

Uh oh!

stefano-garzarella left a comment

Choose a reason for hiding this comment

Uh oh!

luigix25 commented Oct 7, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

luigix25 commented Sep 26, 2025 •

edited

Loading

luigix25 commented Sep 29, 2025 •

edited

Loading