Improve postmaster's behavior if an accept() call fails. Because the server

socket is still read-ready, the code was a tight loop, wasting lots of CPU.
We can't do anything to clear the failure, other than wait, but we should give
other processes more chance to finish and release FDs; so insert a small sleep.
Also, avoid bogus "close(-1)" in this case.  Per report from Jim Nasby.

2 files changed, 11 insertions, 2 deletions

diff --git a/src/backend/libpq/pqcomm.c b/src/backend/libpq/pqcomm.c
index c41983326d..3c0c5bbdd2 100644
--- a/src/backend/libpq/pqcomm.c
+++ b/src/backend/libpq/pqcomm.c
@@ -528,7 +528,7 @@ Setup_AF_UNIX(void)
 
 /*
  * StreamConnection -- create a new connection with client using
- *		server port.
+ *		server port.  Set port->sock to the FD of the new connection.
  *
  * ASSUME: that this doesn't need to be non-blocking because
  *		the Postmaster uses select() to tell when the server master
@@ -548,6 +548,14 @@ StreamConnection(int server_fd, Port *port)
 		ereport(LOG,
 				(errcode_for_socket_access(),
 				 errmsg("could not accept new connection: %m")));
+		/*
+		 * If accept() fails then postmaster.c will still see the server
+		 * socket as read-ready, and will immediately try again.  To avoid
+		 * uselessly sucking lots of CPU, delay a bit before trying again.
+		 * (The most likely reason for failure is being out of kernel file
+		 * table slots; we can do little except hope some will get freed up.)
+		 */
+		pg_usleep(100000L);		/* wait 0.1 sec */
 		return STATUS_ERROR;
 	}
 
diff --git a/src/backend/postmaster/postmaster.c b/src/backend/postmaster/postmaster.c
index fdef23bce4..149ad18718 100644
--- a/src/backend/postmaster/postmaster.c
+++ b/src/backend/postmaster/postmaster.c
@@ -1710,7 +1710,8 @@ ConnCreate(int serverFd)
 
 	if (StreamConnection(serverFd, port) != STATUS_OK)
 	{
-		StreamClose(port->sock);
+		if (port->sock >= 0)
+			StreamClose(port->sock);
 		ConnFree(port);
 		port = NULL;
 	}