b083d24fcf
Selftests here check not only that connect()/accept() for
TCP-AO/TCP-MD5/non-signed-TCP combinations do/don't establish
connections, but also counters: those are per-AO-key, per-socket and
per-netns.
The counters are checked on the server's side, as the server listener
has TCP-AO/TCP-MD5/no keys for different peers. All tests run in
the same namespaces with the same veth pair, created in test_init().
After close() in both client and server, the sides go through
the regular FIN/ACK + FIN/ACK sequence, which goes in the background.
If the selftest has already started a new testing scenario, read
per-netns counters - it may fail in the end iff it doesn't expect
the TCPAOGood per-netns counters go up during the test.
Let's just kill both TCP-AO sides - that will avoid any asynchronous
background TCP-AO segments going to either sides.
Reported-by: Jakub Kicinski <kuba@kernel.org>
Closes: https://lore.kernel.org/all/20240201132153.4d68f45e@kernel.org/T/#u
Fixes:
|
||
|---|---|---|
| .. | ||
| lib | ||
| .gitignore | ||
| Makefile | ||
| bench-lookups.c | ||
| config | ||
| connect-deny.c | ||
| connect.c | ||
| icmps-accept.c | ||
| icmps-discard.c | ||
| key-management.c | ||
| restore.c | ||
| rst.c | ||
| self-connect.c | ||
| seq-ext.c | ||
| setsockopt-closed.c | ||
| settings | ||
| unsigned-md5.c | ||