upstream: start ClientAliveInterval bookkeeping before first pass

through select() loop; fixed theoretical case where busy sshd may ignore
timeouts from client; inspired by and ok dtucker

OpenBSD-Commit-ID: 96bfc4b1f86c7da313882a84755b2b47eb31957f

1 files changed, 3 insertions, 1 deletions

diff --git a/serverloop.c b/serverloop.c
index 340b19a5a..48d936d2e 100644
--- a/serverloop.c
+++ b/serverloop.c
@@ -1,4 +1,4 @@
-/* $OpenBSD: serverloop.c,v 1.222 2020/01/30 07:21:38 djm Exp $ */
+/* $OpenBSD: serverloop.c,v 1.223 2020/07/03 06:29:57 djm Exp $ */
 /*
  * Author: Tatu Ylonen <ylo@cs.hut.fi>
  * Copyright (c) 1995 Tatu Ylonen <ylo@cs.hut.fi>, Espoo, Finland
@@ -253,6 +253,8 @@ wait_until_can_do_something(struct ssh *ssh,
                         max_time_ms = keepalive_ms;
                         client_alive_scheduled = 1;
                 }
+                if (last_client_time == 0)
+                        last_client_time = monotime();
         }
 
 #if 0