Bug 1016441 - Switch to using real-time signal in ThreadStackHelper; r=snorp

toolkit/components/telemetry/tests/unit/test_ThreadHangStats.js

@@ -20,6 +20,17 @@ function run_test() {
     return;
   }
 
+  if (Services.appinfo.OS === 'Linux' || Services.appinfo.OS === 'Android') {
+    // We use the rt_tgsigqueueinfo syscall on Linux which requires a
+    // certain kernel version. It's not an error if the system running
+    // the test is older than that.
+    let kernel = Services.sysinfo.kernel_version || Services.sysinfo.version;
+    if (Services.vc.compare(kernel, '2.6.31') < 0) {
+      ok("Hang reporting not supported for old kernel.");
+      return;
+    }
+  }
+
   // Run two events in the event loop:
   // the first event causes a hang;
   // the second event checks results from the first event.

toolkit/components/telemetry/tests/unit/xpcshell.ini

@@ -16,3 +16,5 @@ tail =
 [test_TelemetrySendOldPings.js]
 skip-if = debug == true || os == "android" # Disabled due to intermittent orange on Android
 [test_ThreadHangStats.js]
+# Bug 1032996: test fails consistently in Android x86 emulator
+skip-if = os == "android" && processor == "x86"

xpcom/threads/ThreadStackHelper.cpp

@@ -27,8 +27,12 @@
 #ifndef SYS_gettid
 #define SYS_gettid __NR_gettid
 #endif
-#ifndef SYS_tgkill
-#define SYS_tgkill __NR_tgkill
+#if defined(__arm__) && !defined(__NR_rt_tgsigqueueinfo)
+// Some NDKs don't define this constant even though the kernel supports it.
+#define __NR_rt_tgsigqueueinfo (__NR_SYSCALL_BASE+363)
+#endif
+#ifndef SYS_rt_tgsigqueueinfo
+#define SYS_rt_tgsigqueueinfo __NR_rt_tgsigqueueinfo
 #endif
 #endif
 
@@ -40,7 +44,18 @@ ThreadStackHelper::Startup()
 #if defined(XP_LINUX)
   MOZ_ASSERT(NS_IsMainThread());
   if (!sInitialized) {
-    MOZ_ALWAYS_TRUE(!::sem_init(&sSem, 0, 0));
+    // TODO: centralize signal number allocation
+    sFillStackSignum = SIGRTMIN + 4;
+    if (sFillStackSignum > SIGRTMAX) {
+      // Leave uninitialized
+      MOZ_ASSERT(false);
+      return;
+    }
+    struct sigaction sigact = {};
+    sigact.sa_sigaction = FillStackHandler;
+    sigemptyset(&sigact.sa_mask);
+    sigact.sa_flags = SA_SIGINFO | SA_RESTART;
+    MOZ_ALWAYS_TRUE(!::sigaction(sFillStackSignum, &sigact, nullptr));
   }
   sInitialized++;
 #endif
@@ -52,7 +67,9 @@ ThreadStackHelper::Shutdown()
 #if defined(XP_LINUX)
   MOZ_ASSERT(NS_IsMainThread());
   if (sInitialized == 1) {
-    MOZ_ALWAYS_TRUE(!::sem_destroy(&sSem));
+    struct sigaction sigact = {};
+    sigact.sa_handler = SIG_DFL;
+    MOZ_ALWAYS_TRUE(!::sigaction(sFillStackSignum, &sigact, nullptr));
   }
   sInitialized--;
 #endif
@@ -68,6 +85,7 @@ ThreadStackHelper::ThreadStackHelper()
   , mMaxBufferSize(0)
 {
 #if defined(XP_LINUX)
+  MOZ_ALWAYS_TRUE(!::sem_init(&mSem, 0, 0));
   mThreadID = ::syscall(SYS_gettid);
 #elif defined(XP_WIN)
   mInitialized = !!::DuplicateHandle(
@@ -82,41 +100,15 @@ ThreadStackHelper::ThreadStackHelper()
 
 ThreadStackHelper::~ThreadStackHelper()
 {
-#if defined(XP_WIN)
+#if defined(XP_LINUX)
+  MOZ_ALWAYS_TRUE(!::sem_destroy(&mSem));
+#elif defined(XP_WIN)
   if (mInitialized) {
     MOZ_ALWAYS_TRUE(!!::CloseHandle(mThreadID));
   }
 #endif
 }
 
-#if defined(XP_LINUX) && defined(__arm__)
-// Some (old) Linux kernels on ARM have a bug where a signal handler
-// can be called without clearing the IT bits in CPSR first. The result
-// is that the first few instructions of the handler could be skipped,
-// ultimately resulting in crashes. To workaround this bug, the handler
-// on ARM is a trampoline that starts with enough NOP instructions, so
-// that even if the IT bits are not cleared, only the NOP instructions
-// will be skipped over.
-
-template <void (*H)(int, siginfo_t*, void*)>
-__attribute__((naked)) void
-SignalTrampoline(int aSignal, siginfo_t* aInfo, void* aContext)
-{
-  asm volatile (
-    "nop; nop; nop; nop"
-    : : : "memory");
-
-  // Because the assembler may generate additional insturctions below, we
-  // need to ensure NOPs are inserted first by separating them out above.
-
-  asm volatile (
-    "bx %0"
-    :
-    : "r"(H), "l"(aSignal), "l"(aInfo), "l"(aContext)
-    : "memory");
-}
-#endif // XP_LINUX && __arm__
-
 void
 ThreadStackHelper::GetStack(Stack& aStack)
 {
@@ -127,29 +119,23 @@ ThreadStackHelper::GetStack(Stack& aStack)
   }
 
 #if defined(XP_LINUX)
-  if (profiler_is_active()) {
-    // Profiler can interfere with our Linux signal handling
-    return;
-  }
   if (!sInitialized) {
     MOZ_ASSERT(false);
     return;
   }
-  sCurrent = this;
-  struct sigaction sigact = {};
-#ifdef __arm__
-  sigact.sa_sigaction = SignalTrampoline<SigAction>;
-#else
-  sigact.sa_sigaction = SigAction;
-#endif
-  sigemptyset(&sigact.sa_mask);
-  sigact.sa_flags = SA_SIGINFO | SA_RESTART;
-  if (::sigaction(SIGPROF, &sigact, &sOldSigAction)) {
-    MOZ_ASSERT(false);
+  siginfo_t uinfo = {};
+  uinfo.si_signo = sFillStackSignum;
+  uinfo.si_code = SI_QUEUE;
+  uinfo.si_pid = getpid();
+  uinfo.si_uid = getuid();
+  uinfo.si_value.sival_ptr = this;
+  if (::syscall(SYS_rt_tgsigqueueinfo, uinfo.si_pid,
+                mThreadID, sFillStackSignum, &uinfo)) {
+    // rt_tgsigqueueinfo was added in Linux 2.6.31.
+    // Could have failed because the syscall did not exist.
     return;
   }
-  MOZ_ALWAYS_TRUE(!::syscall(SYS_tgkill, getpid(), mThreadID, SIGPROF));
-  MOZ_ALWAYS_TRUE(!::sem_wait(&sSem));
+  MOZ_ALWAYS_TRUE(!::sem_wait(&mSem));
 
 #elif defined(XP_WIN)
   if (!mInitialized) {
@@ -177,17 +163,16 @@ ThreadStackHelper::GetStack(Stack& aStack)
 #ifdef XP_LINUX
 
 int ThreadStackHelper::sInitialized;
-sem_t ThreadStackHelper::sSem;
-struct sigaction ThreadStackHelper::sOldSigAction;
-ThreadStackHelper* ThreadStackHelper::sCurrent;
+int ThreadStackHelper::sFillStackSignum;
 
 void
-ThreadStackHelper::SigAction(int aSignal, siginfo_t* aInfo, void* aContext)
+ThreadStackHelper::FillStackHandler(int aSignal, siginfo_t* aInfo,
+                                    void* aContext)
 {
-  ::sigaction(SIGPROF, &sOldSigAction, nullptr);
-  sCurrent->FillStackBuffer();
-  sCurrent = nullptr;
-  ::sem_post(&sSem);
+  ThreadStackHelper* const helper =
+    reinterpret_cast<ThreadStackHelper*>(aInfo->si_value.sival_ptr);
+  helper->FillStackBuffer();
+  ::sem_post(&helper->mSem);
 }
 
 #endif // XP_LINUX

xpcom/threads/ThreadStackHelper.h

@@ -84,12 +84,11 @@ public:
 #if defined(XP_LINUX)
 private:
   static int sInitialized;
-  static sem_t sSem;
-  static struct sigaction sOldSigAction;
-  static ThreadStackHelper* sCurrent;
+  static int sFillStackSignum;
 
-  static void SigAction(int aSignal, siginfo_t* aInfo, void* aContext);
+  static void FillStackHandler(int aSignal, siginfo_t* aInfo, void* aContext);
 
+  sem_t mSem;
   pid_t mThreadID;
 
 #elif defined(XP_WIN)