From: Davidlohr Bueso <dave@xxxxxxxxxxxx> Subject: kselftest: introduce new epoll test case Patch series "fs/epoll: restore user-visible behavior upon event ready". This series tries to address a change in user visible behavior, reported in https://bugzilla.kernel.org/show_bug.cgi?id=208943. Epoll does not report an event to all the threads running epoll_wait() on the same epoll descriptor. Unsurprisingly, this was bisected back to 339ddb53d373 (fs/epoll: remove unnecessary wakeups of nested epoll), which has had various problems in the past, beyond only nested epoll usage. This patch (of 2): This incorporates the testcase originally reported in: https://bugzilla.kernel.org/show_bug.cgi?id=208943 Which ensures an event is reported to all threads blocked on the same epoll descriptor, otherwise only a single thread will receive the wakeup once the event become ready. Link: https://lkml.kernel.org/r/20210405231025.33829-1-dave@xxxxxxxxxxxx Link: https://lkml.kernel.org/r/20210405231025.33829-2-dave@xxxxxxxxxxxx Signed-off-by: Davidlohr Bueso <dbueso@xxxxxxx> Cc: Jason Baron <jbaron@xxxxxxxxxx> Cc: Roman Penyaev <rpenyaev@xxxxxxx> Cc: Al Viro <viro@xxxxxxxxxxxxxxxxxx> Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> --- tools/testing/selftests/filesystems/epoll/epoll_wakeup_test.c | 44 ++++++++++ 1 file changed, 44 insertions(+) --- a/tools/testing/selftests/filesystems/epoll/epoll_wakeup_test.c~kselftest-introduce-new-epoll-test-case +++ a/tools/testing/selftests/filesystems/epoll/epoll_wakeup_test.c @@ -3449,4 +3449,48 @@ TEST(epoll63) close(sfd[1]); } +/* + * t0 t1 + * (ew) \ / (ew) + * e0 + * | (lt) + * s0 + */ +TEST(epoll64) +{ + pthread_t waiter[2]; + struct epoll_event e; + struct epoll_mtcontext ctx = { 0 }; + + signal(SIGUSR1, signal_handler); + + ASSERT_EQ(socketpair(AF_UNIX, SOCK_STREAM, 0, ctx.sfd), 0); + + ctx.efd[0] = epoll_create(1); + ASSERT_GE(ctx.efd[0], 0); + + e.events = EPOLLIN; + ASSERT_EQ(epoll_ctl(ctx.efd[0], EPOLL_CTL_ADD, ctx.sfd[0], &e), 0); + + /* + * main will act as the emitter once both waiter threads are + * blocked and expects to both be awoken upon the ready event. + */ + ctx.main = pthread_self(); + ASSERT_EQ(pthread_create(&waiter[0], NULL, waiter_entry1a, &ctx), 0); + ASSERT_EQ(pthread_create(&waiter[1], NULL, waiter_entry1a, &ctx), 0); + + usleep(100000); + ASSERT_EQ(write(ctx.sfd[1], "w", 1), 1); + + ASSERT_EQ(pthread_join(waiter[0], NULL), 0); + ASSERT_EQ(pthread_join(waiter[1], NULL), 0); + + EXPECT_EQ(ctx.count, 2); + + close(ctx.efd[0]); + close(ctx.sfd[0]); + close(ctx.sfd[1]); +} + TEST_HARNESS_MAIN _
