DeakLock in multithreaded programmming is really an annoying thing. Something if we are lucky, we might get some deadlock information from debugger.

Prerequiste

g++, gdb
C++11
gdb has the permission to attach a process

Write a typical code to generate deadlock

as we know, the mutex is not allow to reenter. if you reenter ,it will cause the deadlock

  
#include <iostream>
#include <map>
#include <string>
#include <chrono>
#include <thread>
#include <mutex>

std::mutex gMutex;

int Reenter(){
   std::lock_guard<std::mutex> lLock(gMutex);
   return 10;	
}

int Callback()
{
   std::lock_guard<std::mutex> lLock(gMutex); 
   return Reenter();
}
int main(int argc, char**argv) {
    // Prints hello message...
    std::cout << "Hello CMake World!" << std::endl;
    std::thread lThread(Callback);
    lThread.join();
    std::cout << "sub thread is gone!" << std::endl;
    return 0;
}

compile with gdb information

g++ test.cpp -ggdb -lpthread -std=c++11 -o test

run it

./test

check the result,it deadlocked as we expected

ethan@ubuntu:~/Desktop$ ./test
Hello CMake World!

Analyze using gdb

Get the pid

ethan@ubuntu:~$ ps -a | grep test
4094 pts/2 00:00:00 test

Attach the pid using another terminal

ethan@ubuntu:~/Desktop$ sudo gdb -q test 4094
[sudo] password for ethan:
Reading symbols from test…done.
Attaching to program: /home/ethan/Desktop/test, process 4094
[New LWP 4095]
[Thread debugging using libthread_db enabled]

Backtrace

(gdb) bt
#0 0x00007f65c7d3e9cd in pthread_join (threadid=140075107923712,
thread_return=0x0) at pthread_join.c:90
#1 0x00007f65c7a6cb97 in std::thread::join() ()
from /usr/lib/x86_64-linux-gnu/libstdc++.so.6
#2 0x00000000004011ea in main (argc=1, argv=0x7fff7a26af28) at test.cpp:27

Info threads

(gdb) info threads
Id Target Id Frame

1 Thread 0x7f65c815a740 (LWP 4094) “test” 0x00007f65c7d3e9cd in pthread_join (threadid=140075107923712, thread_return=0x0) at pthread_join.c:90
2 Thread 0x7f65c70cb700 (LWP 4095) “test” __lll_lock_wait ()
at ../sysdeps/unix/sysv/linux/x86_64/lowlevellock.S:135

Info of mutex

(gdb) p gMutex
$1 = { = {_M_mutex = {__data = {__lock = 2, __count = 0, ***__owner = 4095,*** __nusers = 1, __kind = 0, __spins = 0, __elision = 0, __list = {__prev = 0x0, __next = 0x0}}, __size = "\002\000\000\000\000\000\000\000\377\017\000\000\001", '\000' <repeats 26 times>, __align = 2}}, }

Switch to the thread

(gdb) thread 2
[Switching to thread 2 (Thread 0x7f65c70cb700 (LWP 4095))]
#0 __lll_lock_wait () at ../sysdeps/unix/sysv/linux/x86_64/lowlevellock.S:135
135 ../sysdeps/unix/sysv/linux/x86_64/lowlevellock.S: No such file or directory.

Check the stack of that thread

(gdb) bt
#0 __lll_lock_wait () at ../sysdeps/unix/sysv/linux/x86_64/lowlevellock.S:135
#1 0x00007f65c7d3fdfd in GI_pthread_mutex_lock (mutex=0x6052e0 ) at ../nptl/pthread_mutex_lock.c:80 #2 0x0000000000401007 in __gthread_mutex_lock (__mutex=0x6052e0 ) at /usr/include/x86_64-linux-gnu/c++/5/bits/gthr-default.h:748 #3 0x0000000000401498 in std::mutex::lock (this=0x6052e0 ) at /usr/include/c++/5/mutex:135 #4 0x000000000040151e in std::lock_guard::lock_guard ( this=0x7f65c70cae20, __m=...) at /usr/include/c++/5/mutex:386 #5 0x00000000004010ef in Reenter () at test.cpp:11 #6 0x000000000040114b in Callback () at test.cpp:18 #7 0x0000000000402801 in std::_Bind_simple<int (*())()>::_M_invoke<>(std::_Index_tuple<>) (this=0x2085058) at /usr/include/c++/5/functional:1531 #8 0x000000000040275a in std::_Bind_simple<int (*())()>::operator()() ( this=0x2085058) at /usr/include/c++/5/functional:1520 #9 0x00000000004026ea in std::thread::_Impl<std::_Bind_simple<int (*())()> >::_M_run() (this=0x2085040) at /usr/include/c++/5/thread:115 #10 0x00007f65c7a6cc80 in ?? () from /usr/lib/x86_64-linux-gnu/libstdc++.so.6 #11 0x00007f65c7d3d6fa in start_thread (arg=0x7f65c70cb700) at pthread_create.c:333 #12 0x00007f65c74dbb5d in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:109