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

1 files changed, 17 insertions, 4 deletions

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;