leverage fixed fifo implementation (rev 1.56 of

src/sys/fs/fifofs/fifo_vnops.c) to serve locks better, my previous
workaround for lack of decent fifo system wasn't cutting it,
particularly the kernel would send a message down the fifo and
immediately close it, this would lead to delayed unlock requests
being seen by the lockd causing all sorts of badness.

Basically, don't reopen the fifo, just select(2) on it.

1 files changed, 3 insertions, 8 deletions

diff --git a/usr.sbin/rpc.lockd/kern.c b/usr.sbin/rpc.lockd/kern.c
index 46cd378..76d1831 100644
--- a/usr.sbin/rpc.lockd/kern.c
+++ b/usr.sbin/rpc.lockd/kern.c
@@ -138,21 +138,19 @@ client_request(void)
 	owner.pid = getpid();
 	(void)gethostname(hostname, sizeof(hostname) - 1);
 
-reopen:
 	/* Open the fifo for reading. */
-	if ((fd = open(_PATH_LCKFIFO, O_RDONLY /* | O_NONBLOCK */)) < 0)
+	if ((fd = open(_PATH_LCKFIFO, O_RDONLY | O_NONBLOCK)) < 0)
 		syslog(LOG_ERR, "open: %s: %m", _PATH_LCKFIFO);
 
 	/* drop our root priviledges */
 	(void)lockd_seteuid(daemon_uid);
 
-	/* Set up the select. */
-	FD_ZERO(&rdset);
-
 	for (;;) {
 		/* Wait for contact... fifo's return EAGAIN when read with 
 		 * no data
 		 */
+		/* Set up the select. */
+		FD_ZERO(&rdset);
 		FD_SET(fd, &rdset);
 		(void)select(fd + 1, &rdset, NULL, NULL, NULL);
 
@@ -206,9 +204,6 @@ reopen:
 		} else if (nr != 0) {
 			syslog(LOG_ERR,
 			    "%s: discard %d bytes", _PATH_LCKFIFO, nr);
-		} if (nr == 0) {
-			close (fd);
-			goto reopen;
 		}
 	}