The main thread could start to send SIG_IPI at any time, even before signal blocked on vcpu thread. Reuse the sem_vcpu_stop to sync on that, so when SIG_IPI is sent the signal will always land correctly as an -EINTR. Without this patch, on very busy cores the dirty_log_test could fail directly on receiving a SIG_USR1 without a handler (when vcpu runs far slower than main). Signed-off-by: Peter Xu <peterx@xxxxxxxxxx> --- tools/testing/selftests/kvm/dirty_log_test.c | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/tools/testing/selftests/kvm/dirty_log_test.c b/tools/testing/selftests/kvm/dirty_log_test.c index 510884f0eab8..25230e799bc4 100644 --- a/tools/testing/selftests/kvm/dirty_log_test.c +++ b/tools/testing/selftests/kvm/dirty_log_test.c @@ -534,6 +534,12 @@ static void *vcpu_worker(void *data) sigemptyset(sigset); sigaddset(sigset, SIG_IPI); + /* + * Tell the main thread that signals are setup already; let's borrow + * sem_vcpu_stop even if it's not for it. + */ + sem_post(&sem_vcpu_stop); + guest_array = addr_gva2hva(vm, (vm_vaddr_t)random_array); while (!READ_ONCE(host_quit)) { @@ -785,6 +791,8 @@ static void run_test(enum vm_guest_mode mode, void *arg) pthread_create(&vcpu_thread, NULL, vcpu_worker, vm); + sem_wait_until(&sem_vcpu_stop); + while (iteration < p->iterations) { /* Give the vcpu thread some time to dirty some pages */ usleep(p->interval * 1000); -- 2.26.2