Commit Graph

13 Commits

Author SHA1 Message Date
Daniel Müller
a41e6ef325 Stop running l4lb_all test on 5.5.0
The l4lb_all/l4lb_noinline_dynptr test no does not run on kernel 5.5.0,
because functionality is missing there. Do not allow running it.

Signed-off-by: Daniel Müller <deso@posteo.net>
2023-03-06 09:47:37 -08:00
Song Liu
ea28429902 ci: Remove xdp_info from ALLOWLIST-5.5.0
xdp_info now depends on newer functionalities. Let's skip it for 5.5.0
kernel.

Signed-off-by: Song Liu <song@kernel.org>
2023-02-17 17:17:27 -08:00
Andrii Nakryiko
c2e797c8de ci: temporarily denylist decap_sanity test
It is mysteriously fails in CI, for now don't run it.

Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
2023-01-25 20:44:09 -08:00
Andrii Nakryiko
1e479aec4f ci: don't run test_maps in libbpf CI
It crashes often, it doesn't really test libbpf much.

Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
2022-12-07 09:28:07 -08:00
Matteo Croce
1c17672353 shellcheck: fix errors
Signed-off-by: Matteo Croce <teknoraver@meta.com>
2022-10-27 16:46:38 -07:00
Daniel Müller
abde7fb314 Remove lru_bug from DENYLIST-latest.s390x
The comment associated with the entry is a bit confusing. It stemmed
from the test being denylisted on bpf, but not bpf-next in the past.
Regardless, by now said change has propagated to both trees, so we no
longer need to carry around this deny list entry here.

Signed-off-by: Daniel Müller <deso@posteo.net>
2022-10-12 09:29:08 -07:00
Daniel Müller
10a32130e7 Clean up local allow/deny lists
Now that we are including the upstream allow/deny lists we can remove
any duplicates from our local lists. While at it, we also add some usdt
tests to the denylist, which are currently failing. This is the same
step we took in the vmtest repository [0].

[0] https://github.com/kernel-patches/vmtest/pull/133

Signed-off-by: Daniel Müller <deso@posteo.net>
2022-09-06 15:01:05 -07:00
Daniel Müller
fad270918d Use deny/allow lists from upstream
So far we have relied on allow/deny lists maintained in this repository
to decide which tests to explicitly include/exclude from running in CI.
With recent changes [0] this information is now available in upstream
Linux.
As such, this change switches us over to using the upstream allow/deny
lists in addition to the local ones. We unconditionally honor the
upstream lists for all kernel versions.

[0] https://lore.kernel.org/bpf/165893461358.29339.11641967418379627671.git-patchwork-notify@kernel.org/T/#m2a97b0ea9ef0ddee7a53bbf7919e3f324b233937

Signed-off-by: Daniel Müller <deso@posteo.net>
2022-09-06 15:01:05 -07:00
Daniel Müller
c091b07808 Fix comment: WHITELIST -> ALLOWLIST
Commit 693de729d0 ("Rename blacklists and whitelists") renamed the
black and white lists but missed the adjustment of a comment,
referencing a file name. Update it accordingly.

Signed-off-by: Daniel Müller <deso@posteo.net>
2022-09-06 14:07:51 -07:00
Daniel Müller
a7a525d47a Rename test_progs_noalu function to test_progs_no_alu32
As a follow up to 66b788c1a4 ("Factor out test_progs_noalu function")
and taking into account feedback [0], this change renames the
test_progs_noalu function to test_progs_no_alu32, to stay closer to the
name of the binary being invoked.

[0] https://github.com/kernel-patches/vmtest/pull/124#discussion_r953175641

Signed-off-by: Daniel Müller <deso@posteo.net>
2022-08-24 08:08:21 -07:00
Daniel Müller
7984737fbf Support running of individual tests
This change adjusts the run_selftests.sh script to accept an optional
list of arguments specifying the tests to run. We will make use of it
once we run selftests in parallel.

Signed-off-by: Daniel Müller <deso@posteo.net>
2022-08-18 15:31:52 -07:00
Andrii Nakryiko
a0d1e22c77 ci: blacklist lru_bug selftest on s390x
Make sure we don't fail on lru_bug selftests as it relies of BPF
trampoline, not supported by s390x.

Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
2022-08-18 15:29:04 -07:00
Daniel Müller
bfdf7653e0 Rename travis-ci/ directory to ci/
We are no longer using Travis. As such, we should move away from a lot
of CI functionality located in a folder called travis-ci/. This change
renames the travis-ci/ directory to the more generic ci/.
To preserve backwards compatibility until all "consumers" have
transitioned, we add a symbolic link called travis-ci back. It will be
removed in the near term future.

Signed-off-by: Daniel Müller <deso@posteo.net>
2022-08-18 09:02:13 -07:00