From 064347afdf6206df73571684a3a75dde63584ad5 Mon Sep 17 00:00:00 2001 From: Jamie Liu Date: Tue, 19 May 2020 11:51:28 -0700 Subject: Skip itimer "fairness" tests on ptrace. With additional logging, the issue described by the new comment looks like: D0518 21:28:08.416810 6777 task_signals.go:459] [ 8] Notified of signal 27 D0518 21:28:08.416852 6777 task_block.go:223] [ 8] Interrupt queued D0518 21:28:08.417013 6777 task_run.go:250] [ 8] Switching to sentry D0518 21:28:08.417033 6777 task_signals.go:220] [ 8] Signal 27: delivering to handler D0518 21:28:08.417127 6777 task_run.go:248] [ 8] Switching to app D0518 21:28:08.443765 6777 task_signals.go:519] [ 8] Refusing masked signal 27 // ED: note the ~26ms elapsed since TID 8 "switched to app" D0518 21:28:08.443814 6777 task_signals.go:465] [ 6] Notified of group signal 27 D0518 21:28:08.443832 6777 task_block.go:223] [ 6] Interrupt queued D0518 21:28:08.443914 6777 task_block.go:223] [ 6] Interrupt queued D0518 21:28:08.443859 6777 task_run.go:250] [ 8] Switching to sentry I0518 21:28:08.443936 6777 strace.go:576] [ 8] exe E rt_sigreturn() Slow context switches on ptrace are probably due to kernel scheduling delays. Slow context switches on KVM are less clear, so leave that bug and TODO open. PiperOrigin-RevId: 312322782 --- test/syscalls/linux/itimer.cc | 21 +++++++++++++++++---- 1 file changed, 17 insertions(+), 4 deletions(-) (limited to 'test') diff --git a/test/syscalls/linux/itimer.cc b/test/syscalls/linux/itimer.cc index dd981a278..e397d5f57 100644 --- a/test/syscalls/linux/itimer.cc +++ b/test/syscalls/linux/itimer.cc @@ -267,8 +267,19 @@ int TestSIGPROFFairness(absl::Duration sleep) { // Random save/restore is disabled as it introduces additional latency and // unpredictable distribution patterns. TEST(ItimerTest, DeliversSIGPROFToThreadsRoughlyFairlyActive_NoRandomSave) { - // TODO(b/143247272): CPU time accounting is inaccurate for the KVM platform. - SKIP_IF(GvisorPlatform() == Platform::kKVM); + // On the KVM and ptrace platforms, switches between sentry and application + // context are sometimes extremely slow, causing the itimer to send SIGPROF to + // a thread that either already has one pending or has had SIGPROF delivered, + // but hasn't handled it yet (and thus therefore still has SIGPROF masked). In + // either case, since itimer signals are group-directed, signal sending falls + // back to notifying the thread group leader. ItimerSignalTest() fails if "too + // many" signals are delivered to the thread group leader, so these tests are + // flaky on these platforms. + // + // TODO(b/143247272): Clarify why context switches are so slow on KVM. + const auto gvisor_platform = GvisorPlatform(); + SKIP_IF(gvisor_platform == Platform::kKVM || + gvisor_platform == Platform::kPtrace); pid_t child; int execve_errno; @@ -291,8 +302,10 @@ TEST(ItimerTest, DeliversSIGPROFToThreadsRoughlyFairlyActive_NoRandomSave) { // Random save/restore is disabled as it introduces additional latency and // unpredictable distribution patterns. TEST(ItimerTest, DeliversSIGPROFToThreadsRoughlyFairlyIdle_NoRandomSave) { - // TODO(b/143247272): CPU time accounting is inaccurate for the KVM platform. - SKIP_IF(GvisorPlatform() == Platform::kKVM); + // See comment in DeliversSIGPROFToThreadsRoughlyFairlyActive. + const auto gvisor_platform = GvisorPlatform(); + SKIP_IF(gvisor_platform == Platform::kKVM || + gvisor_platform == Platform::kPtrace); pid_t child; int execve_errno; -- cgit v1.2.3