C++ Boost::thread - kernel: traps general protection
My code has worked using an old version of boost 1.49, but it was many years ago. Now i'm using boost 1.67
Edit: My project includes a server/client feature using the same binary. one the server is started, i can send commands who are received to launch custom process. This is the code showed below.
I identified the line causing the kernel trap:
boost::thread th(Temporal::Acquire, transmit, ECONF);
The thread is starting and function in argument is called, but the initiated thread crash instantly. I don't understand the "general protection".
I tried to find more answer from try catch (std::exception &e) but it seem's requiring another catcher between... nothing to output.
Tried to understand the handling for tls_destructor inside libs/thread/src/pthread/thread.cpp, but since i've tested my code by replacing all std to boost only without solve the issue...
Valgrind show no errors at all.
There is a way to understand the direct termination (without calling join or interrupt) ?
Part to initiate the server socket (from another file), using standard thread, but i don't think that is the source of issue: Since i started my project, i never got conflict mixing std / boost.
coex.push_back(std::thread(Temporal::Listener, ECONF));
Server part:
#include "lobe.hpp"
#include <errno.h>
#include <sys/types.h>
#include <sys/socket.h>
#include <sys/un.h>
void Temporal::Acquire(std::string transmit, Json::Value ECONF)
{
syslog(LOG_NOTICE, "aquired");
// Temporal::Transcode p(transmit, ECONF);
}
void Temporal::Listener(Json::Value ECONF)
{
socklen_t t;
std::string transmit(100, 0);
int PIPE_local, PIPE_remote, len;
struct sockaddr_un local, remote;
int reuseaddr = 1;
memset(&local, 0, sizeof(local));
if((PIPE_local = socket(AF_UNIX, SOCK_STREAM, 0)) == -1)
perror("socket");
if(setsockopt(PIPE_local, SOL_SOCKET, SO_REUSEADDR, &reuseaddr, sizeof(reuseaddr)) == -1)
perror(strerror(errno));
local.sun_family = AF_UNIX;
strncpy(local.sun_path, P_SOCK, sizeof(local.sun_path)-1);
unlink(P_SOCK);
len = strlen(local.sun_path) + sizeof(local.sun_family);
if(bind(PIPE_local, (struct sockaddr *)&local, len) == -1)
perror("bind");
if(listen(PIPE_local, 5) == -1)
perror("listen");
for(;;)
{
syslog(LOG_INFO, "inside SOCK");
int done, com_Listen, com_Talk;
t = sizeof(remote);
if((PIPE_remote = accept(PIPE_local, (struct sockaddr *)&remote, &t)) == -1)
perror("accept");
done = 0;
do
{
com_Listen = read(PIPE_remote, &transmit[0], 99);
if(com_Listen <= 0)
{
syslog(LOG_NOTICE, "<<-== %s", transmit.c_str());
if(com_Listen < 0) perror("recv");
done = 1;
syslog(LOG_NOTICE, "received");
boost::thread th(Temporal::Acquire, transmit, ECONF);
}
}while(!done);
close(PIPE_remote);
break;
}
close(PIPE_local);
unlink(P_SOCK);
std::this_thread::sleep_for(std::chrono::milliseconds(1000));
//boost::this_thread::sleep_for(boost::chrono::seconds(1));
Temporal::Listener(ECONF);
}
Client part:
The output of systemd coredump:
Jun 17 22:37:25 bytewild kernel: traps: EIE[8033] general protection ip:44f59c sp:7fd32bffecb0 error:0 in EIE[400000+233000]
Jun 17 22:37:25 bytewild EIE[7699]: aquired
Jun 17 22:37:25 bytewild systemd[1]: Started Process Core Dump (PID 8034/UID 0).
-- Subject: Unit systemd-coredump@49-8034-0.service has finished start-up
-- Defined-By: systemd
-- Support: https://lists.freedesktop.org/mailman/listinfo/systemd-devel
-- Unit systemd-coredump@49-8034-0.service has finished starting up.
--
-- The start-up result is RESULT.
Jun 17 22:37:25 bytewild systemd-coredump[8041]: Failed to get ACL: Operation not supported
Jun 17 22:37:26 bytewild systemd-coredump[8041]: Process 7699 (EIE) of user 1000 dumped core.
Stack trace of thread 8033:
#0 0x000000000044f59c tls_destructor (/data/dev/in/native/projects/eie/build/bin/EIE)
#1 0x000000000045092a thread_proxy (/data/dev/in/native/projects/eie/build/bin/EIE)
#2 0x0000000000500155 start_thread (/data/dev/in/native/projects/eie/build/bin/EIE)
#3 0x00000000005707ff __clone (/data/dev/in/native/projects/eie/build/bin/EIE)
Stack trace of thread 7700:
#0 0x0000000000503623 __pthread_cond_timedwait (/data/dev/in/native/projects/eie/build/bin/EIE)
#1 0x000000000041214b _ZN5boost18condition_variable13do_wait_untilERNS_11unique_lockINS_5mutexEEERKNS_6detail23mono_platform_timepointE (/data/dev/in/native/projects/eie/build/bin/EIE)
#2 0x000000000040ebe4 _ZN8Temporal8ListenerEN4Json5ValueE (/data/dev/in/native/projects/eie/build/bin/EIE)
#3 0x000000000041f58e _ZSt13__invoke_implIvPFvN4Json5ValueEEJS1_EET_St14__invoke_otherOT0_DpOT1_ (/data/dev/in/native/projects/eie/build/bin/EIE)
#4 0x00000000004eb0ef execute_native_thread_routine (/data/dev/in/native/projects/eie/build/bin/EIE)
#5 0x0000000000500155 start_thread (/data/dev/in/native/projects/eie/build/bin/EIE)
#6 0x00000000005707ff __clone (/data/dev/in/native/projects/eie/build/bin/EIE)
Stack trace of thread 7699:
#0 0x0000000000504a21 __nanosleep (/data/dev/in/native/projects/eie/build/bin/EIE)
#1 0x000000000056bfea __sleep (/data/dev/in/native/projects/eie/build/bin/EIE)
#2 0x0000000000406e91 main (/data/dev/in/native/projects/eie/build/bin/EIE)
#3 0x0000000000506dfa __libc_start_main (/data/dev/in/native/projects/eie/build/bin/EIE)
#4 0x000000000040790a _start (/data/dev/in/native/projects/eie/build/bin/EIE)
Here is a schematic preview of my system to understand the problem:
Sorry if this is obvious, but i'm stumped. Any clues ?
Looking at your code I believe the issue is in this block of code.
do
{
com_Listen = read(PIPE_remote, &transmit[0], 99);
if(com_Listen <= 0)
{
syslog(LOG_NOTICE, "<<-== %s", transmit.c_str());
if(com_Listen < 0) perror("recv");
done = 1;
syslog(LOG_NOTICE, "received");
boost::thread th(Temporal::Acquire, transmit, ECONF);
}
}while(!done);
Specifically, the way you create your boost::thread. This is a stack-based variable. As soon as you start your thread its object is destroyed and the DTor is called.
I do not have much experience with boost implementation of thread, but have used std::thread which was largely modeled on boost implementation. Looking at their class documentation there is an effects section which list the impact of destroying a running thread.
Effects: - if defined BOOST_THREAD_DONT_PROVIDE_THREAD_DESTRUCTOR_CALLS_TERMINATE_IF_JOINABLE: If the thread is joinable calls detach(), DEPRECATED - if defined BOOST_THREAD_PROVIDES_THREAD_DESTRUCTOR_CALLS_TERMINATE_IF_JOINABLE: If the thread is joinable calls to std::terminate. Destroys *this.
It looks like the default is to call terminate() like std::thread does for joinable threads that are destroyed. It looks like the old behavior was to auto detatch those.
Edit Read the destructor documentation for current vs. 1.49.
1.49 Effects: If *this has an associated thread of execution, calls detach(). Destroys *this.
Now, go read the same code for current(listed above). It states it now defaults to terminate
Edit2
My suggestion would be to no longer spin up a new thread everytime you receive input from your socket. Instead, I suggest creating a worker queue that has a constant number of background threads. Each time you receive a new event simply add the work object to the queue and one of the background threads will handle the response.
#include <chrono>
#include <condition_variable>
#include <deque>
#include <iostream>
#include <mutex>
#include <thread>
#include <vector>
// Synchrnoized Output
std::mutex stdm;
template <typename T>
void Log(T const& t) {
std::lock_guard<std::mutex> lk(stdm);
std::cout << t << std::endl;
}
// Task Queue
// Has user provided list of task, which are handled on background threads
template <class T, int N = 4>
class TaskQueue {
std::deque<T> work; // Holds work
std::vector<std::thread> threads; // Holds worker threads
std::mutex m; // Holds lock for work container/running
std::condition_variable cv; // Worker threads wait on this
bool running; // Inform the task queue if its running
public:
// Constructor, spins up worker threads and waits for work
TaskQueue() : running{true} {
threads.reserve(N);
for (int i = 0; i < N; ++i) {
// Build worker threads
threads.emplace_back([&]() {
// Normal running before queue destruction
while (running) {
{
std::unique_lock<std::mutex> lk(
m);
cv.wait(lk, [&] {
return !running ||
work.size() != 0;
});
// Extract work && Update work
// Queue
T t = std::move(work.front());
work.pop_front();
// Release lock before
// performing work
lk.unlock();
// Peform work
t();
}
}
// Empty Queue on destrucrtion
bool hasMoreWork = true;
do {
std::unique_lock<std::mutex> lk(m);
if ((hasMoreWork = work.size() > 0)) {
// Manage Work
T t = std::move(work.front());
work.pop_front();
// has more?
hasMoreWork = work.size() > 0;
// release lock
lk.unlock();
// perform work
t();
}
} while (hasMoreWork);
});
}
}
~TaskQueue() {
// Inform queue its closing
{
std::lock_guard<std::mutex> lk(m);
running = false;
}
// Inform all threads of change
cv.notify_all();
// Clear out remainings objects
int workObjects = 0;
bool queueCleared = false;
do {
{
std::lock_guard<std::mutex> lg(m);
queueCleared = (workObjects = work.size()) == 0;
Log("Queue Has Remaining: " +
std::to_string(workObjects));
}
// Give worker threads time to work
std::this_thread::sleep_for(
std::chrono::milliseconds(250));
} while (!queueCleared);
// If any threads are still processing, join them to the current
// thread or else terminate() is called
for (int i = 0; i < N; i++) {
if (threads[i].joinable()) threads[i].join();
}
}
template <class... Args>
void emplace_back(Args&&... args) {
{
std::lock_guard<std::mutex> lk(m);
if (running)
work.emplace_back(std::forward<Args>(args)...);
}
cv.notify_one();
}
};
// The actual task that performs work
struct Task {
std::string transmit;
std::string ECONF;
Task() : transmit(""), ECONF("") {}
Task(std::string&& t, std::string&& e) : transmit(t), ECONF(e) {}
void operator()() {
std::thread::id tid = std::this_thread::get_id();
std::hash<std::thread::id> hasher;
Log(transmit + ':' + ECONF +
" Process Time: 500ms One Thread: " +
std::to_string(hasher(tid)));
// Fake work to consume thread
std::this_thread::sleep_for(std::chrono::milliseconds(500));
}
};
int main() {
TaskQueue<Task> tp;
for (int i = 0; i < 200; ++i) {
// Add work
tp.emplace_back("Transmit: " + std::to_string(i),
"ECONF: " + std::to_string(i));
// Simulate waiting for the next event from socket
std::this_thread::sleep_for(std::chrono::milliseconds(25));
}
return 0;
}
There is a good number of improvments that can be made, but this should give you a rough overview. Hope this helps.
- Why would you use 'extern "C++"'?
- Strange Behavior Compiler Ignoring NULL Check Unless I Print Something in the if Statement
- Fast inverse square root using fixed point instead of floating point
- What is the const qualifier attached to in C: the memory area or the pointer?
- What is the scope of `fesetround()`?
- Is this declaration UB?
- GCC options for strictest C code?
- How to do an explicit fall-through in C
- How do compilers treat CONST qualifier when the pointer points to a memory location obtained with malloc()?
- C: cmocka headers - how to unittest?
- Why in C when I print a double with a one decimal it round it to the next number
- Android C to Java SWIG unable to compile: incompatible types: byte cannot be converted to SWIGTYPE_p_uint8_t
- GNU Make in Ubuntu giving fatal error: rpc/types.h: No such file or directory
- How can I exclude non-numeric keys? CS50 Caesar Pset2
- How change every struct in an array of pointers?
- Optimized 2x2 matrix multiplication: Slow assembly versus fast SIMD
- Simple frame by frame video decoder library
- GCC no longer implements <varargs.h>
- Contents of IO buffer unknown == unsafe?
- Avoiding strcpy overflow destination warning
- Sort program not working, not sure why
- Fast & accurate atan/arctan approximation algorithm
- What's the difference between strtok_r and strtok_s in C?
- How memory address for pointer to arrays is same as an element in 2D array?
- Which is the best way to suppress "unused variable" warning
- How to use ellipsis in c's case statement?
- Fast ceiling of an integer division in C / C++
- Is there an invalid pthread_t id?
- How to Implement Universal Setter/Getter Functions for Interrupt-Driven Variables in Embedded C?
- How does SIMD (avx) processing work? for example, if I want 10 32 bit floats how do i fit in a 256 bit avx vector?