X-Git-Url: http://git.tuebingen.mpg.de/?a=blobdiff_plain;f=server.c;h=da55b99923621975335a86a7dfbbad41e3e6a573;hb=e0acbd356abf4d9f1b693065b1ff23c57c53ab1e;hp=a4115eb65734391128481f6fa421d548aba78e20;hpb=ad81d625737e1c096c154aedd8a1d161d3ee592e;p=paraslash.git

diff --git a/server.c b/server.c
index a4115eb6..da55b999 100644
--- a/server.c
+++ b/server.c
@@ -291,15 +291,32 @@ static void signal_post_select(struct sched *s, struct task *t)
 			if (pid != mmd->afs_pid)
 				continue;
 			PARA_EMERG_LOG("fatal: afs died\n");
-			goto genocide;
+			kill(0, SIGTERM);
+			goto cleanup;
 		}
 		break;
 	/* die on sigint/sigterm. Kill all children too. */
 	case SIGINT:
 	case SIGTERM:
 		PARA_EMERG_LOG("terminating on signal %d\n", st->signum);
-genocide:
 		kill(0, SIGTERM);
+		/*
+		 * We must wait for afs because afs catches SIGINT/SIGTERM.
+		 * Before reacting to the signal, afs might want to use the
+		 * shared memory area and the mmd mutex.  If we destroy this
+		 * mutex too early and afs tries to lock the shared memory
+		 * area, the call to mutex_lock() will fail and terminate the
+		 * afs process. This leads to dirty osl tables.
+		 *
+		 * There's no such problem with the other children of the
+		 * server process (the command handlers) as these reset their
+		 * SIGINT/SIGTERM handlers to the default action, i.e.  these
+		 * processes get killed immediately by the above kill().
+		 */
+		PARA_INFO_LOG("waiting for afs (pid %d) to die\n",
+			(int)mmd->afs_pid);
+		waitpid(mmd->afs_pid, NULL, 0);
+cleanup:
 		free(mmd->afd.afhi.chunk_table);
 		free(mmd->afd.afhi.info_string);
 		close_listed_fds();