linux下valgrind检测跨线程访问

code:  

#include<pthread.h>
#include<unistd.h>
static int __val = 0;
 static void* thread_fun1(void *__arg)
 {
    while(true)
    {
        ++__val;
    }
    return NULL;
 }

  static void* thread_fun2(void *__arg)
 {
    while(true)
    {
        ++__val;
    }
    return NULL;
 }

int main()
{
     pthread_t __thread_id1;
     pthread_t __thread_id2;
    int __res1 = pthread_create(&__thread_id1,NULL,&thread_fun1,NULL);
    int __res2 = pthread_create(&__thread_id2,NULL,&thread_fun2,NULL);
    sleep(1);
    return 0;
}

usage:

valgrind --tool=helgrind --log-file=helgrind.log ./helgrind_test

output:

==2268== Helgrind, a thread error detector
==2268== Copyright (C) 2007-2010, and GNU GPL'd, by OpenWorks LLP et al.
==2268== Using Valgrind-3.6.1-Debian and LibVEX; rerun with -h for copyright info
==2268== Command: ./helgrind_test
==2268== Parent PID: 1974
==2268==
==2268== Thread #3 was created
==2268== at 0x5131F4E: clone (clone.S:77)
==2268== by 0x4E36E7F: do_clone.constprop.3 (createthread.c:75)
==2268== by 0x4E38604: pthread_create@@GLIBC_2.2.5 (createthread.c:256)
==2268== by 0x4C29B23: pthread_create_WRK (hg_intercepts.c:257)
==2268== by 0x4C29CA7: pthread_create@* (hg_intercepts.c:288)
==2268== by 0x400616: main (helgrind_test.cc:27)
==2268==
==2268== Thread #2 was created
==2268== at 0x5131F4E: clone (clone.S:77)
==2268== by 0x4E36E7F: do_clone.constprop.3 (createthread.c:75)
==2268== by 0x4E38604: pthread_create@@GLIBC_2.2.5 (createthread.c:256)
==2268== by 0x4C29B23: pthread_create_WRK (hg_intercepts.c:257)
==2268== by 0x4C29CA7: pthread_create@* (hg_intercepts.c:288)
==2268== by 0x4005F8: main (helgrind_test.cc:26)
==2268==
==2268== Possible data race during read of size 4 at 0x601038 by thread #3
==2268== at 0x4005C5: thread_fun2(void*) (helgrind_test.cc:17)
==2268== by 0x4C29C90: mythread_wrapper (hg_intercepts.c:221)
==2268== by 0x4E37EFB: start_thread (pthread_create.c:304)
==2268== by 0x5131F8C: clone (clone.S:112)
==2268== This conflicts with a previous write of size 4 by thread #2
==2268== at 0x4005B5: thread_fun1(void*) (helgrind_test.cc:8)
==2268== by 0x4C29C90: mythread_wrapper (hg_intercepts.c:221)
==2268== by 0x4E37EFB: start_thread (pthread_create.c:304)
==2268== by 0x5131F8C: clone (clone.S:112)
==2268==
==2268== Possible data race during write of size 4 at 0x601038 by thread #3
==2268== at 0x4005CE: thread_fun2(void*) (helgrind_test.cc:17)
==2268== by 0x4C29C90: mythread_wrapper (hg_intercepts.c:221)
==2268== by 0x4E37EFB: start_thread (pthread_create.c:304)
==2268== by 0x5131F8C: clone (clone.S:112)
==2268== This conflicts with a previous write of size 4 by thread #2
==2268== at 0x4005B5: thread_fun1(void*) (helgrind_test.cc:8)
==2268== by 0x4C29C90: mythread_wrapper (hg_intercepts.c:221)
==2268== by 0x4E37EFB: start_thread (pthread_create.c:304)
==2268== by 0x5131F8C: clone (clone.S:112)
==2268==
==2268== Possible data race during read of size 4 at 0x601038 by thread #2
==2268== at 0x4005AC: thread_fun1(void*) (helgrind_test.cc:8)
==2268== by 0x4C29C90: mythread_wrapper (hg_intercepts.c:221)
==2268== by 0x4E37EFB: start_thread (pthread_create.c:304)
==2268== by 0x5131F8C: clone (clone.S:112)
==2268== This conflicts with a previous write of size 4 by thread #3
==2268== at 0x4005CE: thread_fun2(void*) (helgrind_test.cc:17)
==2268== by 0x4C29C90: mythread_wrapper (hg_intercepts.c:221)
==2268== by 0x4E37EFB: start_thread (pthread_create.c:304)
==2268== by 0x5131F8C: clone (clone.S:112)
==2268==
==2268== Possible data race during write of size 4 at 0x601038 by thread #2
==2268== at 0x4005B5: thread_fun1(void*) (helgrind_test.cc:8)
==2268== by 0x4C29C90: mythread_wrapper (hg_intercepts.c:221)
==2268== by 0x4E37EFB: start_thread (pthread_create.c:304)
==2268== by 0x5131F8C: clone (clone.S:112)
==2268== This conflicts with a previous write of size 4 by thread #3
==2268== at 0x4005CE: thread_fun2(void*) (helgrind_test.cc:17)
==2268== by 0x4C29C90: mythread_wrapper (hg_intercepts.c:221)
==2268== by 0x4E37EFB: start_thread (pthread_create.c:304)
==2268== by 0x5131F8C: clone (clone.S:112)
==2268==
==2268==
==2268== For counts of detected and suppressed errors, rerun with: -v
==2268== Use --history-level=approx or =none to gain increased speed, at
==2268== the cost of reduced accuracy of conflicting-access information
==2268== ERROR SUMMARY: 7066 errors from 4 contexts (suppressed: 0 from 0)

anysis：
从输出结果看出，检测到了竞态条件，23~32行，
Possible data race during read of size 4 at 0x601038 by thread #3，
This conflicts with a previous write of size 4 by thread #2
分别对应于代码
helgrind_test.cc:17，helgrind_test.cc:8
__val 为全局变量，同时被thread 2,3进行写操作，不加锁，则可被检测出！