Bug#1059765: linux: isolation-machine autopkgtest fails for multiple reasons
Hi On 01-01-2024 22:33, Bastian Blank wrote: Do we have serial of the machines? Do you mean of the system where the VM's run, or of the VM itself? IIRC the qemu backend of autopkgtest is talking to the VM over serial, but if you want to be sure, I'll need to check. Paul OpenPGP_signature.asc Description: OpenPGP digital signature
Bug#1059765: linux: isolation-machine autopkgtest fails for multiple reasons
On Mon, Jan 01, 2024 at 10:33:10PM +0100, Bastian Blank wrote: > Turns out this requires more work. Currently it is not possible to > build some tests. I'll remove the current selftest stuff. This needs much more work, too much work for now. The runner does not support recursive test selection. Many tests fail instead of skipping itself on missing requirements. Bastian -- Immortality consists largely of boredom. -- Zefrem Cochrane, "Metamorphosis", stardate 3219.8
Bug#1059765: linux: isolation-machine autopkgtest fails for multiple reasons
On Sun, Dec 31, 2023 at 06:33:29PM +0100, Bastian Blank wrote: > On Sun, Dec 31, 2023 at 04:47:28PM +0100, Paul Gevers wrote: > > Recently I added some isolation-machine support to ci.debian.net and one of > > the first packages I tried to run the test for is src:linux. > Do you have a handy script available to try this by hand? I was just > looking at this test (to unravel the loop logic and replace it with one > test per kernel), but I'm not sure if this ever worked before. Turns out this requires more work. Currently it is not possible to build some tests. Plan: - get vmlinux.h working - build a linux-selftests package - from there run tests Do we have serial of the machines? Otherwise running those tests is not adviced anyway, as we will have no way to see if something breaks. Bastian -- Lots of people drink from the wrong bottle sometimes. -- Edith Keeler, "The City on the Edge of Forever", stardate unknown
Bug#1059765: linux: isolation-machine autopkgtest fails for multiple reasons
Hi Bastian,
On 31-12-2023 18:33, Bastian Blank wrote:
Do you have a handy script available to try this by hand? I was just
looking at this test (to unravel the loop logic and replace it with one
test per kernel), but I'm not sure if this ever worked before.
It was nearly completely attached at the bottom of the e-mail. Writing
it out and providing a patch (from memory, so untested) as well.
# in path with src:linux unpacked
# Apply patch
export ADT_REBOOT_MARK=step0 # or whatever number matches your kernel
export AUTOPKGTEST_TMP=$(mktemp -d)
chmod a+x debian/tests/selftests
debian/tests/selftests
Paul
diff --git a/debian/tests/selftests b/debian/tests/selftests
index 02cc29372e..ff12a0cd17 100644
--- a/debian/tests/selftests
+++ b/debian/tests/selftests
@@ -1,4 +1,4 @@
-#!/bin/bash -eu
+#!/bin/bash -eux
PATH=/usr/sbin:/sbin:/usr/bin:/bin
@@ -81,8 +81,8 @@ step=$((step + 1))
if [ "$step" -lt "$steps" ]; then
# Load the next kernel
ver=$abiname${localversion[$step]}
-kexec -l /boot/vmlinuz-$ver --initrd /boot/initrd.img-$ver --reuse-cmdline
-/tmp/autopkgtest-reboot step$step
+#kexec -l /boot/vmlinuz-$ver --initrd /boot/initrd.img-$ver --reuse-cmdline
+#/tmp/autopkgtest-reboot step$step
fi
exit 0
OpenPGP_signature.asc
Description: OpenPGP digital signature
Bug#1059765: linux: isolation-machine autopkgtest fails for multiple reasons
On Sun, Dec 31, 2023 at 04:47:28PM +0100, Paul Gevers wrote: > Recently I added some isolation-machine support to ci.debian.net and one of > the first packages I tried to run the test for is src:linux. Do you have a handy script available to try this by hand? I was just looking at this test (to unravel the loop logic and replace it with one test per kernel), but I'm not sure if this ever worked before. Bastian -- Earth -- mother of the most beautiful women in the universe. -- Apollo, "Who Mourns for Adonais?" stardate 3468.1
Bug#1059765: linux: isolation-machine autopkgtest fails for multiple reasons
Source: linux Version: 6.6.8-1 Severity: normal Dear linux maintainers, Recently I added some isolation-machine support to ci.debian.net and one of the first packages I tried to run the test for is src:linux. Unfortunately, the isolation-machine based test (selftests) fails (the reboot that autopgtest schedules just hangs until it times out). I haven't used kexec before in my life, so I'm not totally sure what to expect from $(kexec -l ...) but I *suspect* (after reading the autopkgtest spec [1]) that the call after kexec in selftests is wrong: """ In some cases your test needs to do the reboot by itself, e. g. through kexec, or a reboot command that is hardcoded in the piece of software that you want to test. To support those, you need to call /tmp/autopkgtest-reboot-prepare my_mark at a point as close as possible to the reboot instead; this will merely save the state but not issue the actual reboot by itself """ (so /tmp/autopkgtest-reboot-prepare instead of /tmp/autopkgtest-reboot, and maybe it should be *before* the kexec call) On the other hand, when I run $(uname -a) after a $(manual kexec -l ..) call, it looks like I'm still running the "old" kernel, so I'm not sure if that line does what I would hope it to do either (it's extremely quick too). When I play a bit and finally can get the test to run for the current kernel (manually passing the right step number in ADT_REBOOT_MARK for the current kernel and an appropriate AUTOPKGTEST_TMP [2]) I also notice that $(make headers_install) ends with: """ /bin/sh: 1: rsync: not found """ so that's a missing test dependency. After all that, the test (for the rt kernel) runs nice although it seems to hang on (maybe just a very long running test with no output, I waited several minutes, is that timeout referenced in seconds or minutes?): """ ok 7 selftests: cgroup: test_cpuset # timeout set to 45 # selftests: cgroup: test_zswap """ Second time I ran this was for the regular kernel used by autopkgtest-build-qemu and that passed that stage with ease and seemed to run to the end (with failures, but those you can iron out once we get decent logs). Also, I suspect the total test will timeout (after 2:47) if we test all desired kernel flavors in one test. So instead of squeezing everything in one autopkgtest (stanza) it's probably smarter to generate a stanza per kernel you want to run (because then you're only limited by the overall timeout of 8.5 hours). Paul [1] https://salsa.debian.org/ci-team/autopkgtest/-/blob/master/doc/README.package-tests.rst [2] # emacs debian/tests/selftests # to disable kexec and /tmp/autopkgtest-reboot call and to set -x chmod a+x debian/tests/selftests # export ADT_REBOOT_MARK=step0 # export root@host:/tmp/autopkgtest.Y35OYd/build.rih/src# export AUTOPKGTEST_TMP=$(mktemp -d) # debian/tests/selftests # testing the default installed kernel OpenPGP_signature.asc Description: OpenPGP digital signature
