From: David Hankins <dhankins@isc.org>
Date: Mon, 21 Jan 2008 19:05:20 +0000 (+0000)
Subject: - When a failover server suspects it has encountered a peer running a
X-Git-Tag: v4_1_0a1~18
X-Git-Url: http://git.ipfire.org/gitweb.cgi?a=commitdiff_plain;h=ffdf3c8cb93d22604961558f9857df41d8485143;p=thirdparty%2Fdhcp.git

- When a failover server suspects it has encountered a peer running a
  version 3.0.x failover server, a warning that the failover wire protocol
  is incompatible is printed.  [ISC-Bugs #17129]
- The failover server no longer issues a floating point error if it encounters
  a previously undefined option code.  [ISC-Bugs #17129]
---

diff --git a/RELNOTES b/RELNOTES
index 3133d4b6f..25d5d1c4e 100644
--- a/RELNOTES
+++ b/RELNOTES
@@ -132,6 +132,13 @@ suggested fixes to <dhcp-users@isc.org>.
   allows the client to work in 'captive' network environments, where the
   operator does not want clients to crosstalk directly.
 
+- When a failover server suspects it has encountered a peer running a
+  version 3.0.x failover server, a warning that the failover wire protocol
+  is incompatible is printed.
+
+- The failover server no longer issues a floating point error if it encounters
+  a previously undefined option code.
+
 			Changes since 4.0.0b3
 
 - The reverse dns name for PTR updates on IPv6 addresses has been fixed to
diff --git a/server/failover.c b/server/failover.c
index b96cc855b..a9345d162 100644
--- a/server/failover.c
+++ b/server/failover.c
@@ -645,7 +645,8 @@ static isc_result_t do_a_failover_option (c, link)
 	}
 
 	/* If it's an unknown code, skip over it. */
-	if (option_code > FTO_MAX) {
+	if ((option_code > FTO_MAX) ||
+	    (ft_options[option_code].type == FT_UNDEF)) {
 #if defined (DEBUG_FAILOVER_MESSAGES)
 		log_debug ("  option code %d (%s) len %d (not recognized)",
 			   option_code,
@@ -788,6 +789,35 @@ static isc_result_t do_a_failover_option (c, link)
 	if (op_size == 1 || ft_options [option_code].type == FT_IPADDR) {
 		omapi_connection_copyout ((unsigned char *)op, c, option_len);
 		link -> imsg_count += option_len;
+
+		/*
+		 * As of 3.1.0, many option codes were changed to conform to
+		 * draft revision 12 (which alphabetized, then renumbered all
+		 * the option codes without preserving the version option code
+		 * nor bumping its value).  As it turns out, the message codes
+		 * for CONNECT and CONNECTACK turn out the same, so it tries
+		 * its darndest to connect, and falls short (when TLS_REQUEST
+		 * comes up size 2 rather than size 1 as draft revision 12 also
+		 * mandates).
+		 *
+		 * The VENDOR_CLASS code in 3.0.x was 11, which is now the HBA
+		 * code.  Both work out to be arbitrarily long text-or-byte
+		 * strings, so they pass parsing.
+		 *
+		 * Note that it is possible (or intentional), if highly
+		 * improbable, for the HBA bit array to exactly match
+		 * isc-V3.0.x.  Warning here is not an issue; if it really is
+		 * 3.0.x, there will be a protocol error later on.  If it isn't
+		 * actually 3.0.x, then I guess the lucky user will have to
+		 * live with a weird warning.
+		 */
+		if ((option_code == 11) && (option_len > 9) &&
+		    (strncmp((const char *)op, "isc-V3.0.", 9) == 0)) {
+		        log_error("WARNING: failover as of versions 3.1.0 and "
+				  "on are not reverse compatible with "
+				  "versions 3.0.x.");
+		}
+
 		goto out;
 	}