Bugzilla bug 119340: an inelegant but more reliable way to kill the

multithreaded 'selfserv' process on Linux.
Modified files:
    cmd/selfserv/selfserv.c tests/common/init.sh tests/ssl/ssl.sh

security/nss/cmd/selfserv/selfserv.c

@@ -141,6 +141,15 @@ static int     requestCert;
 static int	verbose;
 static SECItem	bigBuf;
 
+static const char *pidFile;
+/*
+ * On Linux, the first thread we create writes its pid to a file.
+ * (See bug 119340 for why we do this.)
+ */
+#ifdef LINUX
+static PRInt32 threadIndex;
+#endif
+
 static PRThread * acceptorThread;
 
 static PRLogModuleInfo *lm;
@@ -420,6 +429,27 @@ thread_wrapper(void * arg)
 {
     perThread * slot = (perThread *)arg;
 
+#ifdef LINUX  /* bug 119403 */
+    if (pidFile && PR_AtomicIncrement(&threadIndex) == 1) {
+        char *pidFile2;
+        FILE *tmpfile;
+
+        pidFile2 = PORT_Alloc(PORT_Strlen(pidFile) + 3);
+        if (pidFile2 == NULL) {
+            return;  /* failed */
+        }
+        strcpy(pidFile2, pidFile);
+        strcat(pidFile2, ".2");
+        tmpfile=fopen(pidFile2,"w+");
+
+        if (tmpfile) {
+            fprintf(tmpfile,"%d",getpid());
+            fclose(tmpfile);
+        }
+        PORT_Free(pidFile2);
+    }
+#endif /* LINUX */
+
     slot->rv = (* slot->startFunc)(slot->a, slot->b, slot->c);
 
     /* notify the thread exit handler. */
@@ -1384,7 +1414,6 @@ main(int argc, char **argv)
     char *               cipherString= NULL;
     const char *         dir         = ".";
     char *               passwd      = NULL;
-    const char *         pidFile     = NULL;
     char *               tmp;
     char *               envString;
     PRFileDesc *         listen_sock;

security/nss/tests/common/init.sh

@@ -300,11 +300,6 @@ if [ -z "${INIT_SOURCED}" -o "${INIT_SOURCED}" != "TRUE" ]; then
         tee ${LOGFILE}
 
     KILL="kill"
-    if  [ "${OS_ARCH}" = "Linux" ]; then
-#on linux the selfserv needs up to 30 seconds to fully die and free 
-#the socket
-        SLEEP="sleep 5" # taking the chance and trying on the tinderboxes
-    fi
     if [ `uname -s` = "SunOS" ]; then
         PS="/usr/5bin/ps"
     else
@@ -389,7 +384,7 @@ if [ -z "${INIT_SOURCED}" -o "${INIT_SOURCED}" != "TRUE" ]; then
 
     export PATH LD_LIBRARY_PATH SHLIB_PATH LIBPATH
     export DOMSUF HOSTADDR
-    export KILL SLEEP PS
+    export KILL PS
     export MOZILLA_ROOT SECURITY_ROOT DIST TESTDIR OBJDIR HOSTDIR QADIR
     export LOGFILE SCRIPTNAME
 

security/nss/tests/ssl/ssl.sh

@@ -141,13 +141,29 @@ wait_for_selfserv()
 ########################################################################
 kill_selfserv()
 {
-  ${KILL} `cat ${SERVERPID}`
+  # Bug 119340: This is an inelegant but more reliable way to ensure
+  # that a multithreaded process on Linux has been completely killed.
+  # Recall that each thread is a process on Linux.  Instead of killing
+  # the primary thread, we kill the first thread created by selfserv and
+  # let the thread manager take care of terminating the other threads in
+  # selfserv.  The assumption we rely on is that the primary thread is
+  # the last one to go because it's the parent of the thread manager.
+  #
+  # On Linux, the first thread created by selfserv writes its pid to the
+  # {SERVERPID}.2 file.
+  if  [ "${OS_ARCH}" = "Linux" ]; then
+      ${KILL} `cat ${SERVERPID}.2`
+  else
+      ${KILL} `cat ${SERVERPID}`
+  fi
   wait `cat ${SERVERPID}`
   if [ ${fileout} -eq 1 ]; then
       cat ${SERVEROUTFILE}
   fi
-  ${SLEEP}  #FIXME linux waits 30 seconds - find a shorter way (sockets free)
   rm ${SERVERPID}
+  if  [ "${OS_ARCH}" = "Linux" ]; then
+      rm ${SERVERPID}.2
+  fi
 }
 
 ########################### start_selfserv #############################