Member Discussions

Ok, this one has me stumped. I have done DNS via threading with Jobo's threading snipet, this is for a mud which is FAR removed from Godwars but was still close enough for me to be able to get the threading in with a little work, however now we have a crash that I just can't find.

Here is the GDB session for it, including the BT:
[New Thread 7176 (LWP 20029)]

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 7176 (LWP 20029)]
0x400e41a7 in malloc () from /lib/libc.so.6
(gdb) bt
#0 0x400e41a7 in malloc () from /lib/libc.so.6
#1 0x400e3f98 in malloc () from /lib/libc.so.6
#2 0x400d59be in fopen () from /lib/libc.so.6
#3 0x40bf0735 in _nss_files_endservent () from /lib/libnss_files.so.2
#4 0x40bf1870 in _nss_files_gethostbyaddr_r () from /lib/libnss_files.so.2
#5 0x40164b99 in gethostbyaddr_r () from /lib/libc.so.6
#6 0x080e963a in lookup_address (darg=0x40c17248) at comm.c:4827
#7 0x40053e24 in pthread_start_thread_event (arg=0xbebffc00) at manager.c:274

The line of code in question is:
gethostbyaddr_r( darg->buf, sizeof(darg->buf), AF_INET, &ent, buf, 16384, &from, &err);

But, more interesting things, it crashes occasionaly in other places where an fopen is called in the code.

My only thought is that I need to do a pthread_mutex_lock and pthread_mutex_unlock before and after all instances where an fopen would be called, but I'm not sure if that's the solution. Any help on this?

Using mutexes could very well solve this problem (if it is caused by the threads).
Do a pthread_mutex_lock before each fopen() and don't unlock it until after the fclose().

Good luck