[GitHub] [nifi-minifi-cpp] msharee9 commented on a change in pull request #743: Minificpp 1169 - Simplify C2 metrics collection and reporting

[GitBox <gi...@apache.org>]

msharee9 commented on a change in pull request #743: Minificpp 1169 - Simplify C2 metrics collection and reporting
URL: https://github.com/apache/nifi-minifi-cpp/pull/743#discussion_r401721628
 
 

 ##########
 File path: libminifi/src/utils/BackTrace.cpp
 ##########
 @@ -80,25 +80,21 @@ BackTrace TraceResolver::getBackTrace(std::string thread_name, std::thread::nati
   // lock so that we only perform one backtrace at a time.
 #ifdef HAS_EXECINFO
   std::lock_guard<std::mutex> lock(mutex_);
-
-  caller_handle_ = pthread_self();
-  thread_handle_ = thread_handle;
   trace_ = BackTrace(std::move(thread_name));
 
-  if (0 == thread_handle_ || pthread_equal(caller_handle_, thread_handle)) {
+  if (0 == thread_handle || pthread_equal(pthread_self(), thread_handle)) {
     pull_trace();
   } else {
-    if (thread_handle_ == 0) {
+    if (thread_handle == 0) {
       return std::move(trace_);
     }
     emplace_handler();
-    if (pthread_kill(thread_handle_, SIGUSR2) != 0) {
+    std::unique_lock<std::mutex> ulock(trace_mutex_);
+    if (pthread_kill(thread_handle, SIGUSR2) != 0) {
       return std::move(trace_);
     }
-    sigset_t mask;
-    sigfillset(&mask);
-    sigdelset(&mask, SIGUSR2);
-    sigsuspend(&mask);
 
 Review comment:
   There was a race condition here which lead to a deadlock.
   Caller thread (Thread that is interested in pulling backtrace of other thread) acquired lock on a mutex (worker_queue_mutex_), that is outside this function and sent a SIGUSR2 signal to another thread (thread_handle_) but before the caller thread goes into a suspended state or may be even before it removes the SIGUSR2 from its blocking signal set (sigdelset) , the callee while executing its thread handler, sent a SIGUSR2 signal back to the caller and at this time the caller thread will run through its signal handler and after returning from signal handler it goes back to suspended state. From this state there is no way of waking it up.
   The callee was waiting on a condition variable tied to the mutex caller was holding thus creating a deadlock.

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services