From: William Lallemand <wlallemand@haproxy.com>
Date: Wed, 15 Nov 2017 18:02:56 +0000 (+0100)
Subject: BUG/MEDIUM: mworker: wait again for signals when execvp fail
X-Git-Tag: v1.8-rc4~14
X-Git-Url: http://git.ipfire.org/?a=commitdiff_plain;h=2f8b31c2c6e8cb4a15b57796e252226b3154c1bb;p=thirdparty%2Fhaproxy.git

BUG/MEDIUM: mworker: wait again for signals when execvp fail

After execvp fails, the signals were ignored, preventing to try a reload
again. It is now fixed by reaching the top of the mworker_wait()
function once the execvp failed.
---

diff --git a/src/haproxy.c b/src/haproxy.c
index 28811a84f9..b90e2540ba 100644
--- a/src/haproxy.c
+++ b/src/haproxy.c
@@ -670,6 +670,8 @@ static void mworker_wait()
 	int exitpid = -1;
 	int status = 0;
 
+restart_wait:
+
 	mworker_register_signals();
 	mworker_unblock_signals();
 
@@ -679,6 +681,8 @@ static void mworker_wait()
 			int sig = caught_signal;
 			if (sig == SIGUSR2 || sig == SIGHUP) {
 				mworker_reload();
+				/* should reach there only if it fail */
+				goto restart_wait;
 			} else {
 				Warning("Exiting Master process...\n");
 				mworker_kill(sig);