]> git.ipfire.org Git - thirdparty/linux.git/commit
Merge branch 'selftests-forwarding-vxlan_bridge_1q_mc_ul-fix-flakiness'
authorJakub Kicinski <kuba@kernel.org>
Thu, 11 Dec 2025 08:53:18 +0000 (00:53 -0800)
committerJakub Kicinski <kuba@kernel.org>
Thu, 11 Dec 2025 08:53:19 +0000 (00:53 -0800)
commit237c1e152b4263fc1e6e8a359d95227a78945e6d
treea23d037397cf91083c6dc496d3fb9f212c58b4ef
parent6bcb7727d9e612011b70d64a34401688b986d6ab
parent514520b34ba7d0eb36890f9f9c5c874a7e41544e
Merge branch 'selftests-forwarding-vxlan_bridge_1q_mc_ul-fix-flakiness'

Petr Machata says:

====================
selftests: forwarding: vxlan_bridge_1q_mc_ul: Fix flakiness

The net/forwarding/vxlan_bridge_1q_mc_ul selftest runs an overlay traffic,
forwarded over a multicast-routed VXLAN underlay. In order to determine
whether packets reach their intended destination, it uses a TC match. For
convenience, it uses a flower match, which however does not allow matching
on the encapsulated packet. So various service traffic ends up being
indistinguishable from the test packets, and ends up confusing the test. To
alleviate the problem, the test uses sleep to allow the necessary service
traffic to run and clear the channel, before running the test traffic. This
worked for a while, but lately we have nevertheless seen flakiness of the
test in the CI.

In this patchset, first generalize tc_rule_stats_get() to support u32 in
patch #1, then in patch #2 convert the test to use u32 to allow parsing
deeper into the packet, and in #3 drop the now-unnecessary sleep.
====================

Link: https://patch.msgid.link/cover.1765289566.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>