fix bug 54: make abrt (without -d) report initialization errors

Example (strace):

10053 17:18:58.384259 close(0)          = 0
10053 17:18:58.384293 open("/dev/null", O_RDWR) = 0
    child is created:
10053 17:18:58.384354 clone(child_stack=0, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD, child_tidptr=0x7f50eea1c880) = 10054
    parent sleeps, child initializes:
10053 17:18:58.384678 nanosleep({0, 100000000},  <unfinished ...>
10054 17:18:58.384751 setsid()          = 10054
10054 17:18:58.384794 close(1)          = 0
10054 17:18:58.384828 close(2)          = 0
10054 17:18:58.384870 dup(0)            = 1
10054 17:18:58.384901 dup(0)            = 2
10054 17:18:58.385099 open("/usr/app/abrt-TEST/var/etc/abrt/abrt.conf", O_RDONLY) = 3
...
10054 17:18:58.449108 inotify_init()    = 4
10054 17:18:58.449148 inotify_add_watch(4, "/usr/app/abrt-TEST/var/cache/abrt", IN_CREATE) = 1
10054 17:18:58.449228 fstat(4, {st_mode=S_IFDIR|0600, st_size=0, ...}) = 0
10054 17:18:58.449288 fcntl(4, F_GETFL) = 0 (flags O_RDONLY)
    child: initialization is done, it signals parent:
10054 17:18:58.449346 getppid()         = 10053
10054 17:18:58.449380 kill(10053, SIGTERM) = 0
10054 17:18:58.449427 write(1, "Debug: Running...\n"..., 18 <unfinished ...>
10053 17:18:58.449447 <... nanosleep resumed> 0) = ? ERESTART_RESTARTBLOCK (To be restarted)
10054 17:18:58.449470 <... write resumed> ) = 18
10054 17:18:58.449494 open("/usr/app/abrt-TEST/var/run/abrt.lock", O_RDWR|O_CREAT, 0640 <unfinished ...>
10053 17:18:58.449517 --- SIGTERM (Terminated) @ 0 (0) ---
10054 17:18:58.449565 <... open resumed> ) = 5
10053 17:18:58.449583 rt_sigreturn(0xf <unfinished ...>
10054 17:18:58.449607 fcntl(5, F_SETLK, {type=F_WRLCK, whence=SEEK_CUR, start=0, len=0} <unfinished ...>
    sleep(100ms) is interrupted:
10053 17:18:58.449629 <... rt_sigreturn resumed> ) = -1 EINTR (Interrupted system call)
10054 17:18:58.449664 <... fcntl resumed> ) = 0
10054 17:18:58.449688 unlink("/usr/app/abrt-TEST/var/run/abrt.pid" <unfinished ...>
    parent exits 0 because it got TERM which means child is ok:
10053 17:18:58.449713 exit_group(0)     = ?
...

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

1 files changed, 20 insertions, 6 deletions

diff --git a/src/Daemon/Daemon.cpp b/src/Daemon/Daemon.cpp
index dff8de6c..0b7393aa 100644
--- a/src/Daemon/Daemon.cpp
+++ b/src/Daemon/Daemon.cpp
@@ -56,25 +56,34 @@ int main(int argc, char** argv)
                 daemonize = 0;
             }
         }
-        if(daemonize)
+        if (daemonize)
         {
             /* Open stdin to /dev/null. We do it before forking
              * in order to emit useful exitcode to the parent
-	     * if open fails */
+             * if open fails */
             close(STDIN_FILENO);
             if (open("/dev/null", O_RDWR))
             {
                 throw CABRTException(EXCEP_FATAL, "Can't open /dev/null");
             }
-    	    /* forking to background */
+            /* forking to background */
             pid_t pid = fork();
             if (pid < 0)
             {
                 throw CABRTException(EXCEP_FATAL, "CCrashWatcher::Daemonize(): Fork error");
             }
-            /* parent exits */
-            if (pid > 0) _exit(0);
-            /* child (daemon) continues */
+            if (pid > 0)
+            {
+                /* Parent */
+                /* Wait for child to notify us via SIGTERM that it feels ok */
+                int i = 20; /* 2 sec */
+                while (sig_caught == 0 && --i)
+                {
+                        usleep(100 * 1000);
+                }
+                _exit(sig_caught != SIGTERM); /* TERM:ok (0), anything else: bad (1) */
+            }
+            /* Child (daemon) continues */
             pid_t sid = setsid();
             if(sid == -1)
             {
@@ -88,6 +97,11 @@ int main(int argc, char** argv)
             dup(0);
         }
         g_pCrashWatcher = new CCrashWatcher(DEBUG_DUMPS_DIR);
+        if (daemonize)
+        {
+            /* Let parent know we initialized ok */
+            kill(getppid(), SIGTERM);
+        }
         g_pCrashWatcher->Run();
     }
     catch(CABRTException& e)