Jakub Kicinski wrote:
> On Mon, 4 May 2026 13:38:31 -0400 Willem de Bruijn wrote:
> > From: Willem de Bruijn <[email protected]>
> >
> > In preparation for extending to pacing hardware offload, convert the
> > so_txtime.sh test to a drv-net test that can be run against netdevsim
> > and real hardware.
> >
> > Two preparatory patches
> > 1. support negative tests, where tests are expected to fail
> > 2. add a tc helper
> >
> > See individual patches for details and detailed changelog
>
> Hi Willem! Looks like we have some flakiness here:
>
> https://netdev.bots.linux.dev/contest.html?executor=vmksft-net-drv-dbg&test=so-txtime-py
Argh sorry. I should have run dbg mode, but hadn't.
The previous version skips all receiver errors if KSFT_MACHINE_SLOW is
set, as of commit a7ee79b9c455 ("selftests: net: cope with slow env in
so_txtime.sh test").
This achieves the same in the new version and seems to indeed mask all
the timing (and other) errors in debug mode.
@@ -24,8 +25,12 @@ def test_so_txtime(cfg, clockid, ipver, args_tx,
args_rx, expect_success):
cmd_rx = f"{cmd_base} {args_rx} -r"
cmd_tx = f"{cmd_base} {args_tx}"
+ expect_fail = not expect_success
+ if os.environ.get('KSFT_MACHINE_SLOW'):
+ expect_success = False
+
with bkg(cmd_rx, host=cfg.remote, fail=expect_success,
- expect_fail=(not expect_success), exit_wait=True):
+ expect_fail=expect_fail, exit_wait=True):
cmd(cmd_tx)
