Hello,
I have a C multi-thread process that periodically forks.
Sometimes child process hangs and WDB output says:
c32dabd0 in __ksleep+0x10 () from /usr/lib/libc.2
#1 0xc3b002a4 in __sleep_1x1+0x610 () from /usr/lib/libpthread.1
#2 0xc3af8f98 in __mxn_sleep+0xdd8 () from /usr/lib/libpthread.1
#3 0xc3ac501c in + 0x3800 () from /usr/lib/libpthread.1
#4 0xc3ac59b0 in pthread_mutex_lock+0xd0 () from /usr/lib/libpthread.1
#5 0xc3211540 in __thread_mutex_lock+0x70 () from /usr/lib/libc.2
#6 0xc32164f0 in printf+0x118 () from /usr/lib/libc.2
#7 0x156c0 in Print (TrcLevel=4, LineNum=748, Format=0x241a8 "CHILD PROCESS - ExecAction - CloseSock_3(%d)") at CommFunc.c:1015
#8 0x8f50 in ExecAction (action=0x40016064, loc_phase=0x40001820 "-p") at HAReplicationManager.c:748
#9 0x74b0 in main (arc=1, argv=0x64101000) at HAReplicationManager.c:506
The routine is:
void Print (int TrcLevel, int LineNum, const char * Format, ...)
{
char loc_buff [BF];
va_list UnNamedParams;
va_start (UnNamedParams, Format);
if (trc_level >= TrcLevel)
{
sprintf (loc_buff, "%d %05d %s %d - ", mypid, LineNum, print_date(), trc_level);
vsprintf (loc_buff + (strlen (loc_buff)), Format, UnNamedParams);
printf ("%s\n", loc_buff);
fflush (NULL);
}
va_end (UnNamedParams);
return;
}
and line 1015 is:
printf ("%s\n", loc_buff);
Now I have checked all points in which I use pthread_mutex_lock but I cannot find any relation with the above line.
A further info: "mypid" and "trc_level" are variables defined in father process.
Finally, fork env managed through:
pthread_atfork_err = pthread_atfork(prepare, parent, child);
In which direction can I proceed to investigate?
Many thanks
Giuseppe
