On Mon, May 06, 2024 at 06:12:54PM -0700, Jakub Kicinski wrote: > On Mon, 6 May 2024 20:53:22 +0000 Joe Damato wrote: > > Add a simple test for the epoll busy poll ioctls, using the kernel > > selftest harness. > > > > This test ensures that the ioctls have the expected return codes and > > that the kernel properly gets and sets epoll busy poll parameters. > > > > The test can be expanded in the future to do real busy polling (provided > > another machine to act as the client is available). > > Hm, we get: > > # timeout set to 3600 > # selftests: net: epoll_busy_poll > # TAP version 13 > # 1..5 > # # Starting 5 tests from 2 test cases. > # # RUN invalid_fd.test_invalid_fd ... > # # OK invalid_fd.test_invalid_fd > # ok 1 invalid_fd.test_invalid_fd > # # RUN epoll_busy_poll.test_get_params ... > # # OK epoll_busy_poll.test_get_params > # ok 2 epoll_busy_poll.test_get_params > # # RUN epoll_busy_poll.test_set_invalid ... > # # epoll_busy_poll.c:204:test_set_invalid:Expected -1 (-1) == ret (0) > # # epoll_busy_poll.c:205:test_set_invalid:EPIOCSPARAMS should error busy_poll_budget > NAPI_POLL_WEIGHT > # # epoll_busy_poll.c:207:test_set_invalid:Expected EPERM (1) == errno (22) > # # epoll_busy_poll.c:208:test_set_invalid:EPIOCSPARAMS errno should be EPERM busy_poll_budget > NAPI_POLL_WEIGHT > # # test_set_invalid: Test failed > # # FAIL epoll_busy_poll.test_set_invalid > # not ok 3 epoll_busy_poll.test_set_invalid > # # RUN epoll_busy_poll.test_set_and_get_valid ... > # # OK epoll_busy_poll.test_set_and_get_valid > # ok 4 epoll_busy_poll.test_set_and_get_valid > # # RUN epoll_busy_poll.test_invalid_ioctl ... > # # OK epoll_busy_poll.test_invalid_ioctl > # ok 5 epoll_busy_poll.test_invalid_ioctl > > https://netdev-3.bots.linux.dev/vmksft-net/results/584001/98-epoll-busy-poll/stdout Ah, sorry -- this is because I had assumed the test would run without CAP_NET_ADMIN, but since: epoll_busy_poll.c:204:test_set_invalid:Expected -1 (-1) == ret (0) succeeds (ret = 0), clearly I am mistaken. Sorry about that. I think I'll spin up a v3 and I'll add a test with and without CAP_NET_ADMIN to check both cases, which would probably be better anyway.
