lib: Correctly handle ppoll pfds.events == 0

The frrevent system is spitting out this message in bgpd:
20:40:15 mem1-roc-f2-b1-r5-t2-d4 bgpd[13166]: [XETTR-D5MR0][EC 100663316] Attempting to process an I/O event but for fd: 214(8) no thread to handle this!

This is because as each io event is processed, it is possible that a
.events is set to 0.  This can leave a situation where we ask
ppoll to handle anything that happens on a fd with a .events of 0,
in this situation ppoll can return POLLERR, which indicates that
something bad has happened on the fd.

Let's cleanup the pfds after io events are handled that can be
cleaned up.

Signed-off-by: Donald Sharp <[email protected]>

Author: donaldsharp

lib/event.c

@@ -1690,6 +1690,35 @@ static void thread_process_io(struct event_loop *m, unsigned int num)
 	for (i = 0; i < last_read && ready < num; ++i)
 		thread_process_io_inner_loop(m, num, pfds, &i, &ready);
 
+	/*
+	 * At this point any fd's that were processed
+	 * have possibly had their .events == 0
+	 * so we should remove them now from the list
+	 * in it's entirety.  This is being done
+	 * because if we leave a fd in the poll even
+	 * with a .events of 0, if the fd is closed
+	 * poll will return a .revents of POLLERR
+	 * This is not desirable, as that we are attempting
+	 * to handle something we have absolutely no
+	 * way to handle it at all.
+	 */
+	i = 0;
+	while (i < m->handler.pfdcount) {
+		if (m->handler.pfds[i].events != 0) {
+			i++;
+		} else {
+			memmove(m->handler.pfds + i, m->handler.pfds + i + 1,
+				(m->handler.pfdcount - i - 1) * sizeof(struct pollfd));
+			m->handler.pfdcount--;
+			m->handler.pfds[m->handler.pfdcount].fd = 0;
+			m->handler.pfds[m->handler.pfdcount].events = 0;
+			/*
+			 * Intentionally keeping i the same
+			 * We need to look at it again
+			 */
+		}
+	}
+
 	m->last_read++;
 }