Question

I have a common getter method with rdlock and unlock:

e_state get_state   ( )
{
    rdlock();
    e_state state= m_e_state;
    unlock();
    return state;
}

The setter with writelock and unlock:

void set_state (e_state a_new_e_state)  
{
    wrlock();
    m_e_state = a_new_e_state;
    unlock();
}

The rdlock(), wrlock() and unlock() methods are the followings:

bool rdlock ( ){return pthread_rwlock_rdlock (&m_rwlock) == 0;}
bool wrlock ( ){return pthread_rwlock_wrlock (&m_rwlock) == 0;}
bool unlock ( ){return pthread_rwlock_unlock (&m_rwlock) == 0;}

With Valgrind I got the following errors:

Conflicting load by thread 3 at 0x04328fa0 size 4
==16557==    at 0x8068712: Sample_Namespace::get_state() (CClassManagerXY.h:94)
==16557==    by 0x807FA9B: Sample_Namespace::xy_method(std::string const&,                          Sample_Namespace::IClassX&, int) (IClassY.cpp:773)
==16557==    by 0x48CC8CA: Sources::SourceClassX::run() (SourceClassX.cpp:156)
==16557==    by 0x8065BEF: run(void*) (CClassManagerXY.cpp:120)
==16557==    by 0x402A31B: ??? (in /usr/lib/valgrind/vgpreload_drd-x86-linux.so)
==16557==    by 0x4062B24: start_thread (in /lib/libpthread-2.11.2.so)
==16557==    by 0x428446D: clone (in /lib/libc-2.11.2.so)
==16557== Address 0x4328fa0 is at offset 8 from 0x4328f98. Allocation context:
==16557==    at 0x4029BD3: operator new(unsigned int) (in       /usr/lib/valgrind/vgpreload_drd-x86-linux.so)
==16557==    by 0x807E9FD: Sample_Namespace::loads(std::string const&, bool)    (CClassY.cpp:460)
==16557==    by 0x807EE9A: Sample_Namespace::load() (CClassY.cpp:542)
==16557==    by 0x807D7D4: Sample_Namespace::init() (CClassY.cpp:171)
==16557==    by 0x808D57C: main (main.cpp:203)
==16557== Other segment start (thread 2)
==16557==    at 0x4284458: clone (in /lib/libc-2.11.2.so)
==16557== Other segment end (thread 2)
==16557==    at 0x42925E6: __lll_unlock_wake_private (in /lib/libc-2.11.2.so)
==16557==    by 0x41F45DD: _L_unlock_13308 (in /lib/libc-2.11.2.so)
==16557==    by 0x41F4527: buffered_vfprintf (in /lib/libc-2.11.2.so)
==16557==    by 0x41EF2FA: vfprintf (in /lib/libc-2.11.2.so)
==16557==    by 0x41F9BAE: fprintf (in /lib/libc-2.11.2.so)
==16557==    by 0x40395F7: Sample_Namespace::write_trace_record(int, unsigned int, char const*, int, char const*, char const*, ...) (CTrace.cpp:106)
==16557==    by 0x8065C84: run_adapter(void*) (CClassManagerXY.cpp:124)
==16557==    by 0x402A31B: ??? (in /usr/lib/valgrind/vgpreload_drd-x86-linux.so)
==16557==    by 0x4062B24: start_thread (in /lib/libpthread-2.11.2.so)
==16557==    by 0x428446D: clone (in /lib/libc-2.11.2.so)

I don’t know why can Data Race be near the state variable, I managed there the locks. I dont understand well what is the meaning of the “Address is at offset 8 from…” row. Have you got any tip what is the problem?

Answer 1

Thank You PlasmaHH!
I have found there are one more method use the variable. It was my mistake.
However I don’t understand well, is conflicting load really dangerous? Not only conflicting store?