From: Jeremy Harris <jgh146exb@wizmail.org>
Date: Wed, 27 Dec 2017 23:32:02 +0000 (+0000)
Subject: Fix issue with continued-connections when the DNS shifts unreliably
X-Git-Tag: exim-4.90.0.22~12
X-Git-Url: https://git.exim.org/exim.git/commitdiff_plain/70ba304da5933bcf3a9afb040228b2a37b3beefe?ds=inline

Fix issue with continued-connections when the DNS shifts unreliably
---

diff --git a/doc/doc-txt/ChangeLog b/doc/doc-txt/ChangeLog
index 16c8f759e..03be8996b 100644
--- a/doc/doc-txt/ChangeLog
+++ b/doc/doc-txt/ChangeLog
@@ -29,6 +29,12 @@ JH/05 Bug 2215: Fix crash associated with dnsdb lookup done from DKIM ACL.
       active in the testsuite.
       Problem spotted, and debugging aided, by Wolfgang Breyha.
 
+JH/06 Fix issue with continued-connections when the DNS shifts unreliably.
+      When none of the hosts presented to a transport match an already-open
+      connection, close it and proceed with the list.  Previously we would
+      queue the message.  Spotted by Lena with Yahoo, probably involving
+      round-robin DNS.
+
 
 Exim version 4.90
 -----------------
diff --git a/src/src/transports/smtp.c b/src/src/transports/smtp.c
index 61e8d8a4f..5d351dd70 100644
--- a/src/src/transports/smtp.c
+++ b/src/src/transports/smtp.c
@@ -3924,7 +3924,9 @@ for (cutoff_retry = 0;
   {
   host_item *nexthost = NULL;
   int unexpired_hosts_tried = 0;
+  BOOL continue_host_tried = FALSE;
 
+retry_non_continued:
   for (host = hostlist;
           host
        && unexpired_hosts_tried < ob->hosts_max_try
@@ -4057,14 +4059,16 @@ for (cutoff_retry = 0;
     result of the lookup. Set expired FALSE, to save the outer loop executing
     twice. */
 
-    if (  continue_hostname
-       && (  Ustrcmp(continue_hostname, host->name) != 0
-          || Ustrcmp(continue_host_address, host->address) != 0
-       )  )
-      {
-      expired = FALSE;
-      continue;      /* With next host */
-      }
+    if (continue_hostname)
+      if (  Ustrcmp(continue_hostname, host->name) != 0
+         || Ustrcmp(continue_host_address, host->address) != 0
+	 )
+	{
+	expired = FALSE;
+	continue;      /* With next host */
+	}
+      else
+	continue_host_tried = TRUE;
 
     /* Reset the default next host in case a multihomed host whose addresses
     are not looked up till just above added to the host list. */
@@ -4522,6 +4526,32 @@ for (cutoff_retry = 0;
       }
     }   /* End of loop for trying multiple hosts. */
 
+  /* If we failed to find a matching host in the list, for an already-open
+  connection, just close it and start over with the list.  This can happen
+  for routing that changes from run to run, or big multi-IP sites with
+  round-robin DNS. */
+
+  if (continue_hostname && !continue_host_tried)
+    {
+    int fd = cutthrough.fd >= 0 ? cutthrough.fd : 0;
+
+    DEBUG(D_transport) debug_printf("no hosts match already-open connection\n");
+#ifdef SUPPORT_TLS
+    if (tls_out.active == fd)
+      {
+      (void) tls_write(FALSE, US"QUIT\r\n", 6, FALSE);
+      tls_close(FALSE, TRUE);
+      }
+    else
+#else
+      (void) write(fd, US"QUIT\r\n", 6);
+#endif
+    (void) close(fd);
+    cutthrough.fd = -1;
+    continue_hostname = NULL;
+    goto retry_non_continued;
+    }
+
   /* This is the end of the loop that repeats iff expired is TRUE and
   ob->delay_after_cutoff is FALSE. The second time round we will
   try those hosts that haven't been tried since the message arrived. */