if_gre: Add a regression test
AcceptedPublic
Actions

Authored by pouria on Feb 19 2026, 9:37 AM.

Details

Reviewers

ae
glebius
markj

Group Reviewers

network

Summary

Add a regression test for gre(4)
to make sure all of the gre capabilities and options are
working as intended.

This test depends on PR275474 to be resolved.
otherwise it's going to trigger kernel panic.

Test Plan

To test, without triggering kernel panic, add the following lines below on top of every cleanup function:

sleep 1
ifconfig -j gretest1 gre1 destroy
sleep 1
ifconfig -j gretest2 gre1 destroy
sleep 1

and then run the tests:

# kyua test -k /usr/tests/sys/net/Kyuafile if_gre
if_gre:gre4_basic  ->  passed  [0.145s]
if_gre:gre6_basic  ->  passed  [0.148s]
if_gre:gre6_csum  ->  passed  [0.149s]
if_gre:gre6_key  ->  passed  [0.153s]
if_gre:gre6_seq  ->  passed  [0.160s]
if_gre:gre6_udpencap  ->  passed  [0.149s]
if_gre:gre_blind_options  -> passed [0.157s]

Diff Detail

Repository

rG FreeBSD src repository

Lint

Lint Passed

Unit

No Test Coverage

Build Status

Buildable 74652
Build 71535: arc lint + arc unit

Event Timeline

pouria created this revision.Feb 19 2026, 9:37 AM

Herald added subscribers: melifaro, imp. · View Herald TranscriptFeb 19 2026, 9:37 AM

pouria requested review of this revision.Feb 19 2026, 9:37 AM

Harbormaster completed remote builds in B70768: Diff 172212.Feb 19 2026, 9:37 AM

pouria edited the summary of this revision. (Show Details)Feb 19 2026, 9:45 AM

pouria edited the test plan for this revision. (Show Details)

pouria mentioned this in D55366: ifconfig: Add gre netlink support.Feb 19 2026, 10:58 AM

pouria added a parent revision: D55398: if_gre: make access to softc's data safe in network epoch.Feb 20 2026, 3:09 PM

pouria mentioned this in D55398: if_gre: make access to softc's data safe in network epoch.Feb 23 2026, 6:27 PM

pouria edited parent revisions, added: D57669: if_gre(4): Fix races by changing initialization order and locks; removed: D55398: if_gre: make access to softc's data safe in network epoch.Fri, Jun 19, 3:40 PM

markj added a subscriber: markj.Mon, Jul 6, 2:32 PM

markj added inline comments.

tests/sys/net/if_gre.sh
50	You can now set `required.kmods` in the ATF header instead.

Address @markj comment.

Harbormaster completed remote builds in B74639: Diff 181571.Wed, Jul 8, 7:54 PM

markj added inline comments.Thu, Jul 9, 2:28 PM

tests/sys/net/if_gre.sh
35	It should be `require.kmods`, my comment had a typo.
37	IMO it is nicer to group test subroutines together without whitespace in between: atf_test_case foo foo_head() { } foo_body() { } foo_cleanup() { } This requires less vertical space, and makes it easier to visually group the test subroutines together.
59	`-s exit:0` is the default, you don't need to write it everywhere.
418	What does "blindly" mean here?

Address @markj comments.
Thank you for all of your great reviews.

Harbormaster completed remote builds in B74652: Diff 181602.Thu, Jul 9, 2:46 PM

pouria added inline comments.Thu, Jul 9, 2:46 PM

tests/sys/net/if_gre.sh
35	It should be `require.kmods`, my comment had a typo. Oops, didn't think/test when I updated the code. :( sorry
418	I don't verify its actual effect on dataplane here. We simply want to make sure this options works and driver updates itself, regardless of using netlink or ioctl. For example, I want to make sure that driver knows I've set the udpport to x to make sure my netlink implementation works or at least it won't break without failing a test. I've done the same thing in if_geneve due its large number of options and it helps A LOT!

markj accepted this revision.Mon, Jul 13, 2:53 PM

markj added inline comments.

tests/sys/net/if_gre.sh
67	Do you actually need most of these `-o ignore`? I suspect not. It would be better to remove them where possible, the default `-o empty` check is stricter.
418	You might add some comment here explaining that. I saw the same test in if_geneve and was confused about the meaning there too.