Merge r5116: Kill child process if CLI gets out of sync

Whenever the CLI comms with the child process runs of the rails, we
have to kill the child:  We cannot get the CLI pipes back in sync,
even if the child process was just preoccupied with something for 
much longer than cli_timeout.

Fixes: #744



git-svn-id: http://www.varnish-cache.org/svn/branches/2.1@5327 d4fa192b-c00b-0410-8231-f00ffab90ce4

bin/varnishd/mgt.h

@@ -47,6 +47,7 @@ extern pid_t child_pid;
 void MGT_Run(void);
 void mgt_stop_child(void);
 void mgt_got_fd(int fd);
+void MGT_Child_Cli_Fail(void);
 
 /* mgt_cli.c */
 

bin/varnishd/mgt_child.c

@@ -200,14 +200,33 @@ child_poker(const struct vev *e, int what)
 		return (0);
 	if (!mgt_cli_askchild(NULL, NULL, "ping\n"))
 		return (0);
-	REPORT(LOG_ERR,
-	    "Child (%jd) not responding to ping, killing it.",
+	return (0);
+}
+
+/*--------------------------------------------------------------------
+ * If CLI communications with the child process fails, there is nothing
+ * for us to do but to drag it behind the barn and get it over with.
+ *
+ * The typical case is where the child process fails to return a reply
+ * before the cli_timeout expires.  This invalidates the CLI pipes for
+ * all future use, as we don't know if the child was just slow and the
+ * result gets piped later on, or if the child is catatonic.
+ */
+
+void
+MGT_Child_Cli_Fail(void)
+{
+
+	if (child_state != CH_RUNNING)
+		return;
+	if (child_pid < 0)
+		return;
+	REPORT(LOG_ERR, "Child (%jd) not responding to CLI, killing it.",
 	    (intmax_t)child_pid);
 	if (params->diag_bitmap & 0x1000)
 		(void)kill(child_pid, SIGKILL);
 	else
 		(void)kill(child_pid, SIGQUIT);
-	return (0);
 }
 
 /*--------------------------------------------------------------------*/

bin/varnishd/mgt_cli.c

@@ -178,16 +178,16 @@ mcf_askchild(struct cli *cli, const char * const *av, void *priv)
 	if (i != strlen(cli->cmd)) {
 		cli_result(cli, CLIS_COMMS);
 		cli_out(cli, "CLI communication error");
+		MGT_Child_Cli_Fail();
 		return;
 	}
 	i = write(cli_o, "\n", 1);
 	if (i != 1) {
 		cli_result(cli, CLIS_COMMS);
 		cli_out(cli, "CLI communication error");
+		MGT_Child_Cli_Fail();
 		return;
 	}
-
-	assert(i == 1 || errno == EPIPE);
 	(void)cli_readres(cli_i, &u, &q, params->cli_timeout);
 	cli_result(cli, u);
 	cli_out(cli, "%s", q);
@@ -233,12 +233,15 @@ mgt_cli_askchild(unsigned *status, char **resp, const char *fmt, ...) {
 			*status = CLIS_COMMS;
 		if (resp != NULL)
 			*resp = strdup("CLI communication error");
+		MGT_Child_Cli_Fail();
 		return (CLIS_COMMS);
 	}
 
 	(void)cli_readres(cli_i, &u, resp, params->cli_timeout);
 	if (status != NULL)
 		*status = u;
+	if (u == CLIS_COMMS)
+		MGT_Child_Cli_Fail();
 	return (u == CLIS_OK ? 0 : u);
 }