That's actually a bit of a tricky question. It seems that the existing IGT syncobj_timeline test asserts the incorrect behavior that waiting for an unsubmitted fence with WAIT_AVAILABLE set should fail with EINVAL. I've sent a patch to igt-dev correcting this https://lists.freedesktop.org/archives/igt-dev/2023-August/059858.html however that will cause the test to fail with current kernels. I don't know how big of a problem that would be. Personally I kinda feel like the test *should* fail with current kernels, since current kernels do indeed have a bug. On 8/16/23 09:40, Simon Ser wrote: > BTW, question: do you know if we could add an IGT test to make sure we > don't regress this?
