More development on dhcp-split-relay.

To provoke this bug, at least the following must be true.

A reply must be a CNAME.
The target of the CNAME must not exist for the queried RR (a No Data reply).
The No Data reply must include an SOA record in the NS section.
The query must take place over TCP.

The result is the caching of a CNAME whose target is the
name of the SOA record, instead of the No data record.

If there is a RR at this target, then a subsequent query for that
RRtype will get a qrong reply.

Thanks to the testers extraordinaire at Pi-Hole for spotting this
and providing enough information to chase it down.

CHANGELOG

@@ -61,6 +61,15 @@ version 2.92
 
 	Add TFTP options windowsize (RFC 7440) and timeout (RFC 2349).
 	
+	Change the behaviour of the DHVPv6 server when a REBIND message
+	is received but no lease exists. Under these circumstances a new
+	lease is created _only_ when the --dhcp-authoritative option is
+	set. This matches the behavior of the DHCPv4 server.
+
+	Add --dhcp-split-relay option. This makes a DHCPv4 relay which
+	is functional when client and server networks aren't mutually
+	route-able.
+
 	
 version 2.91
 	Fix spurious "resource limit exceeded messages". Thanks to 

doc.html

@@ -95,7 +95,5 @@ a contribution towards my expenses, please use the donation button below.
 <input type="image" src="https://www.paypalobjects.com/en_US/GB/i/btn/btn_donateCC_LG.gif" border="0" name="submit" alt="PayPal – The safer, easier way to pay online.">
 <img alt="" border="0" src="https://www.paypalobjects.com/en_GB/i/scr/pixel.gif" width="1" height="1">
 </form>
-
-
 </BODY>
-
+</HTML>

man/dnsmasq.8

@@ -673,7 +673,7 @@ machine on which dnsmasq is running) for each
 local machine. Local machines are those in /etc/hosts or with DHCP
 leases.
 .TP
-.B \-W, --srv-host=<_service>.<_prot>.[<domain>],[<target>[,<port>[,<priority>[,<weight>]]]]
+.B \-W, --srv-host=<_service>.<_prot>[.<domain>],[<target>[,<port>[,<priority>[,<weight>]]]]
 Return a SRV DNS record. See RFC2782 for details. If not supplied, the
 domain defaults to that given by
 .B --domain.
@@ -1487,7 +1487,31 @@ DHCPv4 to a DHCPv6 server or vice-versa.
 The DHCP relay function for IPv6 includes the ability to snoop
 prefix-delegation from relayed DHCP transactions. See
 .B --dhcp-script
-for details. 
+for details.
+.TP
+.B --dhcp-split-relay=<local address>,[<server address>[#<server port>]],<server-facing-interface>|<server-facing-address>
+A usefully enchanced version of DHCPv4 relay. IPv4 DHCP normally uses a single address
+for two functions; it is used by the DHCP server to determine which network to allocate
+an address on, and it is used as the address of the relay to which the server sends packets.
+
+This version of DHCP relay splits these functions. It uses the address of the server-facing relay
+interface or a directly-specified address as the address that the server talks to. The address of the client-facing interface
+(the first item in the config) is used as to determine the client's subnet. The
+local address is also used as server-ID override so that the client always sends requests
+via the relay. The effect of this is that server doesn't require
+a route to the client network and the clients don't require a route to the server.
+
+The third parameter is mandatory. If it is an interface name it cannot be a wildcard and the same filtering as descibed in
+--dhcp-relay applies; answers from the server must arrve via the specified interface. If the third parameter
+is an IP address it must be an address of a local interface which is routable from the server; In this case no filtering
+is done, the reply packets can arrive via any route.
+
+If setting up a network where the client networks have limited routing, be careful
+about configuring the DHCP server. Dnsmasq, as DHCP server, will set the default route to the
+client-facing relay interface unless explicitly configured: that is a sensible default.
+The normal default DNS server (the same address as the DHCP server)
+will not be appropriate when there is no route bewteen the
+two, so this will have to be explicitly configured. 
 .TP
 .B \-U, --dhcp-vendorclass=set:<tag>,[enterprise:<IANA-enterprise number>,]<vendor-class>
 Map from a vendor-class string to a tag. Most DHCP clients provide a 
@@ -1739,7 +1763,9 @@ For DHCPv4, it changes the behaviour from strict RFC compliance so that DHCP req
 unknown leases from unknown hosts are not ignored. This allows new hosts
 to get a lease without a tedious timeout under all circumstances. It also 
 allows dnsmasq to rebuild its lease database without each client needing to 
-reacquire a lease, if the database is lost. For DHCPv6 it sets the
+reacquire a lease, if the database is lost. For DHCPv6 it controls the same
+behaviour as DHCPv4 with missing leases (except for the RFC uncompliance - the
+DHCPv6 RFCs allow this behaviour if configured). It also sets the
 priority in replies to 255 (the maximum) instead of 0 (the minimum).
 .TP
 .B --dhcp-rapid-commit

setup.html

@@ -229,3 +229,5 @@ contents of the cache to the syslog.
 
 <P>For a complete listing of options please take a look at the manpage
 dnsmasq(8).
+</BODY>
+</HTML>

src/cache.c

@@ -938,15 +938,14 @@ int cache_recv_insert(time_t now, int fd)
 		if (newc)
 		  {
 		    newc->addr.cname.is_name_ptr = 0;
+		    newc->addr.cname.target.cache = crecp;
 		    
-		    if (!crecp)
-		      newc->addr.cname.target.cache = NULL;
-		    else
+		    if (crecp)
 		      {
 			next_uid(crecp);
-			newc->addr.cname.target.cache = crecp;
 			newc->addr.cname.uid = crecp->uid;
 		      }
+		    crecp = newc;
 		  }
 	      }
 	    else

src/config.h

@@ -290,6 +290,7 @@ HAVE_SOCKADDR_SA_LEN
 #define HAVE_BSD_NETWORK
 #define HAVE_GETOPT_LONG
 #define HAVE_SOCKADDR_SA_LEN
+#define NO_IPSET
 /* Define before sys/socket.h is included so we get socklen_t */
 #define _BSD_SOCKLEN_T_
 /* Select the RFC_3542 version of the IPv6 socket API. 

src/dhcp-common.c

@@ -1067,10 +1067,12 @@ void log_relay(int family, struct dhcp_relay *relay)
     {
       if (broadcast)
 	my_syslog(MS_DHCP | LOG_INFO, _("DHCP relay from %s via %s"), daemon->addrbuff, relay->interface);
+      else if (relay->split_mode)
+	my_syslog(MS_DHCP | LOG_INFO, _("DHCP split-relay from %s to %s via %s"), daemon->addrbuff, daemon->namebuff, relay->interface);
       else
 	my_syslog(MS_DHCP | LOG_INFO, _("DHCP relay from %s to %s via %s"), daemon->addrbuff, daemon->namebuff, relay->interface);
     }
-  else
+  else 
     my_syslog(MS_DHCP | LOG_INFO, _("DHCP relay from %s to %s"), daemon->addrbuff, daemon->namebuff);
 }
    

src/dhcp-protocol.h

@@ -73,6 +73,7 @@
 #define SUBOPT_REMOTE_ID         2
 #define SUBOPT_SUBNET_SELECT     5     /* RFC 3527 */
 #define SUBOPT_SUBSCR_ID         6     /* RFC 3393 */
+#define SUBOPT_FLAGS             10    /* RFC 5010 */
 #define SUBOPT_SERVER_OR         11    /* RFC 5107 */
 
 #define SUBOPT_PXE_BOOT_ITEM     71    /* PXE standard */

src/dhcp.c

@@ -32,8 +32,6 @@ static int complete_context(struct in_addr local, int if_index, char *label,
 			    struct in_addr netmask, struct in_addr broadcast, void *vparam);
 static int check_listen_addrs(struct in_addr local, int if_index, char *label,
 			      struct in_addr netmask, struct in_addr broadcast, void *vparam);
-static void relay_upstream4(int iface_index, struct dhcp_packet *mess, size_t sz);
-static struct dhcp_relay *relay_reply4(struct dhcp_packet *mess, char *arrival_interface);
 
 static int make_fd(int port)
 {
@@ -144,13 +142,14 @@ void dhcp_packet(time_t now, int pxe_fd)
   struct iovec iov;
   ssize_t sz; 
   int iface_index = 0, unicast_dest = 0, is_inform = 0, loopback = 0;
-  int rcvd_iface_index;
+  int rcvd_iface_index, relay_index;
   struct in_addr iface_addr;
   struct iface_param parm;
   time_t recvtime = now;
 #ifdef HAVE_LINUX_NETWORK
   struct arpreq arp_req;
   struct timeval tv;
+  struct in_addr dst_addr;
 #endif
   
   union {
@@ -176,13 +175,11 @@ void dhcp_packet(time_t now, int pxe_fd)
       (sz < (ssize_t)(sizeof(*mess) - sizeof(mess->options)))) 
     return;
   
-#ifdef HAVE_DUMPFILE
-  dump_packet_udp(DUMP_DHCP, (void *)daemon->dhcp_packet.iov_base, sz, (union mysockaddr *)&dest, NULL, fd);
-#endif
-  
 #if defined (HAVE_LINUX_NETWORK)
   if (ioctl(fd, SIOCGSTAMP, &tv) == 0)
     recvtime = tv.tv_sec;
+
+  dst_addr.s_addr = 0;
   
   if (msg.msg_controllen >= sizeof(struct cmsghdr))
     for (cmptr = CMSG_FIRSTHDR(&msg); cmptr; cmptr = CMSG_NXTHDR(&msg, cmptr))
@@ -194,7 +191,8 @@ void dhcp_packet(time_t now, int pxe_fd)
 	  } p;
 	  p.c = CMSG_DATA(cmptr);
 	  iface_index = p.p->ipi_ifindex;
-	  if (p.p->ipi_addr.s_addr != INADDR_BROADCAST)
+	  dst_addr = p.p->ipi_addr; 
+	  if (dst_addr.s_addr != INADDR_BROADCAST)
 	    unicast_dest = 1;
 	}
 
@@ -224,7 +222,25 @@ void dhcp_packet(time_t now, int pxe_fd)
 	  iface_index = *(p.i);
 	}
 #endif
-	
+
+#ifdef HAVE_DUMPFILE
+  union mysockaddr *sockp = NULL;
+  
+#  ifdef HAVE_LINUX_NETWORK
+  union mysockaddr tosock;
+  
+  sockp = &tosock;
+  tosock.in.sin_port = htons(daemon->dhcp_server_port);
+  tosock.in.sin_addr = dst_addr;
+  tosock.sa.sa_family = AF_INET;
+#    ifdef HAVE_SOCKADDR_SA_LEN
+  tosock.in.sin_len = sizeof(struct sockaddr_in);
+#    endif
+#  endif
+  
+  dump_packet_udp(DUMP_DHCP, (void *)daemon->dhcp_packet.iov_base, sz, (union mysockaddr *)&dest, sockp, -1);
+#endif
+  	
   if (!indextoname(daemon->dhcpfd, iface_index, ifr.ifr_name) ||
       ioctl(daemon->dhcpfd, SIOCGIFFLAGS, &ifr) != 0)
     return;
@@ -286,10 +302,10 @@ void dhcp_packet(time_t now, int pxe_fd)
     unicast_dest = 1;
 #endif
   
-  if ((relay = relay_reply4((struct dhcp_packet *)daemon->dhcp_packet.iov_base, ifr.ifr_name)))
+  if ((relay_index = relay_reply4((struct dhcp_packet *)daemon->dhcp_packet.iov_base, (size_t)sz, ifr.ifr_name)))
     {
       /* Reply from server, using us as relay. */
-      rcvd_iface_index = relay->iface_index;
+      rcvd_iface_index = relay_index;
       if (!indextoname(daemon->dhcpfd, rcvd_iface_index, ifr.ifr_name))
 	return;
       is_relay_reply = 1; 
@@ -314,9 +330,12 @@ void dhcp_packet(time_t now, int pxe_fd)
 	if (tmp->name && (tmp->flags & INAME_4) && wildcard_match(tmp->name, ifr.ifr_name))
 	  return;
       
-      /* unlinked contexts/relays are marked by context->current == context */
+      /* unlinked contexts are marked by context->current == context */
       for (context = daemon->dhcp; context; context = context->next)
 	context->current = context;
+
+      for (relay = daemon->relay4; relay; relay = relay->next)
+	relay->matchcount = 0;
       
       parm.current = NULL;
       parm.ind = iface_index;
@@ -344,7 +363,7 @@ void dhcp_packet(time_t now, int pxe_fd)
       if (!iface_enumerate(AF_INET, &parm, (callback_t){.af_inet=complete_context}))
 	return;
 
-      relay_upstream4(iface_index, mess, (size_t)sz);
+      relay_upstream4(iface_addr, iface_index, mess, (size_t)sz, unicast_dest);
        
       /* May have configured relay, but not DHCP server */
       if (!daemon->dhcp)
@@ -648,8 +667,19 @@ static int complete_context(struct in_addr local, int if_index, char *label,
     }
 
   for (relay = daemon->relay4; relay; relay = relay->next)
-    if (relay->local.addr4.s_addr == local.s_addr)
-      relay->iface_index = if_index;
+    if (!relay->split_mode && relay->local.addr4.s_addr == local.s_addr)
+      {
+	if (if_index == param->ind)
+	  relay->iface_index = if_index;
+	
+	/* More than one interface with the relay address breaks things. */
+	if (relay->matchcount++ == 1 && !relay->warned)
+	  {
+	    relay->warned = 1;
+	    inet_ntop(AF_INET, &local, daemon->addrbuff, ADDRSTRLEN);
+	    my_syslog(MS_DHCP | LOG_WARNING, _("DHCP relay address %s appears on more than one interface"), daemon->addrbuff);
+	  }
+      }
   
   return 1;
 }
@@ -1091,122 +1121,4 @@ char *host_from_dns(struct in_addr addr)
   return NULL;
 }
 
-static void relay_upstream4(int iface_index, struct dhcp_packet *mess, size_t sz)
-{
-  struct in_addr giaddr = mess->giaddr;
-  u8 hops = mess->hops;
-  struct dhcp_relay *relay;
-
-  if (mess->op != BOOTREQUEST)
-    return;
-
-  for (relay = daemon->relay4; relay; relay = relay->next)
-    if (relay->iface_index != 0 && relay->iface_index == iface_index)
-      break;
-
-  /* No relay config. */
-  if (!relay)
-    return;
-  
-  for (; relay; relay = relay->next)
-    if (relay->iface_index != 0 && relay->iface_index == iface_index)
-      {
-	union mysockaddr to;
-	union all_addr from;
-
-	mess->hops = hops;
-	mess->giaddr = giaddr;
-	
-	if ((mess->hops++) > 20)
-	  continue;
-	
-	/* source address == relay address */
-	from.addr4 = relay->local.addr4;
-
-	/* already gatewayed ? */
-	if (giaddr.s_addr)
-	  {
-	    /* if so check if by us, to stomp on loops. */
-	    if (giaddr.s_addr == relay->local.addr4.s_addr)
-	      continue;
-	  }
-	else
-	  {
-	    /* plug in our address */
-	    mess->giaddr.s_addr = relay->local.addr4.s_addr;
-	  }
-	
-	to.sa.sa_family = AF_INET;
-	to.in.sin_addr = relay->server.addr4;
-	to.in.sin_port = htons(relay->port);
-#ifdef HAVE_SOCKADDR_SA_LEN
-	to.in.sin_len = sizeof(struct sockaddr_in);
-#endif
-	
-	/* Broadcasting to server. */
-	if (relay->server.addr4.s_addr == 0)
-	  {
-	    struct ifreq ifr;
-	    
-	    if (relay->interface)
-	      safe_strncpy(ifr.ifr_name, relay->interface, IF_NAMESIZE);
-	    
-	    if (!relay->interface || strchr(relay->interface, '*') ||
-		ioctl(daemon->dhcpfd, SIOCGIFBRDADDR, &ifr) == -1)
-	      {
-		my_syslog(MS_DHCP | LOG_ERR, _("Cannot broadcast DHCP relay via interface %s"), relay->interface);
-		continue;
-	      }
-	    
-	    to.in.sin_addr = ((struct sockaddr_in *) &ifr.ifr_addr)->sin_addr;
-	  }
-	
-#ifdef HAVE_DUMPFILE
-	{
-	  union mysockaddr fromsock;
-	  fromsock.in.sin_port = htons(daemon->dhcp_server_port);
-	  fromsock.in.sin_addr = from.addr4;
-	  fromsock.sa.sa_family = AF_INET;
-
-	  dump_packet_udp(DUMP_DHCP, (void *)mess, sz, &fromsock, &to, -1);
-	}
-#endif
-	
-	 send_from(daemon->dhcpfd, 0, (char *)mess, sz, &to, &from, 0);
-	 
-	 if (option_bool(OPT_LOG_OPTS))
-	   {
-	     inet_ntop(AF_INET, &relay->local, daemon->addrbuff, ADDRSTRLEN);
-	     if (relay->server.addr4.s_addr == 0)
-	       snprintf(daemon->dhcp_buff2, DHCP_BUFF_SZ, _("broadcast via %s"), relay->interface);
-	     else
-	       inet_ntop(AF_INET, &relay->server.addr4, daemon->dhcp_buff2, DHCP_BUFF_SZ);
-	     my_syslog(MS_DHCP | LOG_INFO, _("DHCP relay at %s -> %s"), daemon->addrbuff, daemon->dhcp_buff2);
-	   }
-      }
-  
-  /* restore in case of a local reply. */
-  mess->giaddr = giaddr;
-}
-
-
-static struct dhcp_relay *relay_reply4(struct dhcp_packet *mess, char *arrival_interface)
-{
-  struct dhcp_relay *relay;
-
-  if (mess->giaddr.s_addr == 0 || mess->op != BOOTREPLY)
-    return NULL;
-
-  for (relay = daemon->relay4; relay; relay = relay->next)
-    {
-      if (mess->giaddr.s_addr == relay->local.addr4.s_addr)
-	{
-	  if (!relay->interface || wildcard_match(relay->interface, arrival_interface))
-	    return relay->iface_index != 0 ? relay : NULL;
-	}
-    }
-  
-  return NULL;	 
-}     
-
 #endif

src/dhcp6.c

@@ -89,6 +89,7 @@ void dhcp6_init(void)
 void dhcp6_packet(time_t now)
 {
   struct dhcp_context *context;
+  struct dhcp_relay *relay;
   struct iface_param parm;
   struct cmsghdr *cmptr;
   struct msghdr msg;
@@ -176,6 +177,7 @@ void dhcp6_packet(time_t now)
   else
     {
       struct dhcp_bridge *bridge, *alias;
+      int multicast_dest = 0;
       
       for (tmp = daemon->if_except; tmp; tmp = tmp->next)
 	if (tmp->name && wildcard_match(tmp->name, ifr.ifr_name))
@@ -230,14 +232,24 @@ void dhcp6_packet(time_t now)
 	    context->current = context;
 	    memset(&context->local6, 0, IN6ADDRSZ);
 	  }
+
+      for (relay = daemon->relay6; relay; relay = relay->next)
+	relay->matchcount = 0;
+
+      inet_pton(AF_INET6, ALL_RELAY_AGENTS_AND_SERVERS, &all_servers);
+      if (IN6_ARE_ADDR_EQUAL(&dst_addr, &all_servers))
+	multicast_dest = 1;
       
-      /* Ignore requests sent to the ALL_SERVERS multicast address for relay when
-	 we're listening there for DHCPv6 server reasons. */
       inet_pton(AF_INET6, ALL_SERVERS, &all_servers);
-      
-      if (!IN6_ARE_ADDR_EQUAL(&dst_addr, &all_servers) &&
-	  relay_upstream6(if_index, (size_t)sz, &from.sin6_addr, from.sin6_scope_id, now))
-	return;
+      if (IN6_ARE_ADDR_EQUAL(&dst_addr, &all_servers))
+	multicast_dest = 1;
+      else
+	{
+	  /* Ignore requests sent to the ALL_SERVERS multicast address for relay when
+	     we're listening there for DHCPv6 server reasons. */
+	  if (relay_upstream6(if_index, (size_t)sz, &from.sin6_addr, from.sin6_scope_id, now))
+	    return;
+	}
       
       if (!iface_enumerate(AF_INET6, &parm, (callback_t){.af_inet6=complete_context6}))
 	return;
@@ -266,7 +278,7 @@ void dhcp6_packet(time_t now)
       
       lease_prune(NULL, now); /* lose any expired leases */
       
-      port = dhcp6_reply(parm.current, if_index, ifr.ifr_name, &parm.fallback, 
+      port = dhcp6_reply(parm.current, multicast_dest, if_index, ifr.ifr_name, &parm.fallback, 
 			 &parm.ll_addr, &parm.ula_addr, sz, &from.sin6_addr, now);
       
       /* The port in the source address of the original request should
@@ -444,7 +456,17 @@ static int complete_context6(struct in6_addr *local,  int prefix,
   if (match)
     for (relay = daemon->relay6; relay; relay = relay->next)
       if (IN6_ARE_ADDR_EQUAL(local, &relay->local.addr6))
-	relay->iface_index = if_index;
+	{
+	  relay->iface_index = if_index;
+
+	  /* More than one interface with the relay address breaks things. */
+	  if (relay->matchcount++ == 1 && !relay->warned)
+	    {
+	      relay->warned = 1;
+	      inet_ntop(AF_INET6, &local, daemon->addrbuff, ADDRSTRLEN);
+	      my_syslog(MS_DHCP | LOG_WARNING, _("DHCP relay address %s appears on more than one interface"), daemon->addrbuff);
+	    }
+	}
   
   return 1;
 }

src/dnsmasq.c

@@ -1777,9 +1777,9 @@ void clear_cache_and_reload(time_t now)
 #ifdef HAVE_DHCP
   if (daemon->dhcp || daemon->doing_dhcp6)
     {
+      reread_dhcp();
       if (option_bool(OPT_ETHERS))
 	dhcp_read_ethers();
-      reread_dhcp();
       dhcp_update_configs(daemon->dhcp_conf);
       lease_update_from_configs(); 
       lease_update_file(now); 

src/dnsmasq.h

@@ -1144,10 +1144,12 @@ struct dhcp_relay {
   union {
     struct in_addr addr4;
     struct in6_addr addr6;
-  } local, server;
+  } local, server, uplink;
   char *interface; /* Allowable interface for replies from server, and dest for IPv6 multicast */
   int iface_index; /* working - interface in which requests arrived, for return */
   int port;        /* Port of relay we forward to. */
+  int split_mode;  /* Split address allocation and relay address. */
+  int warned, matchcount;
 #ifdef HAVE_SCRIPT
   struct snoop_record {
     struct in6_addr client, prefix;
@@ -1666,6 +1668,9 @@ size_t dhcp_reply(struct dhcp_context *context, char *iface_name, int int_index,
 		  time_t recvtime, struct in_addr leasequery_source);
 unsigned char *extended_hwaddr(int hwtype, int hwlen, unsigned char *hwaddr, 
 			       int clid_len, unsigned char *clid, int *len_out);
+void relay_upstream4(struct in_addr iface_addr, int iface_index,
+		     struct dhcp_packet *mess, size_t sz, int unicast);
+unsigned int relay_reply4(struct dhcp_packet *mess, size_t sz, char *arrival_interface);
 #endif
 
 /* dnsmasq.c */
@@ -1801,7 +1806,7 @@ void get_client_mac(struct in6_addr *client, int iface, unsigned char *mac,
   
 /* rfc3315.c */
 #ifdef HAVE_DHCP6
-unsigned short dhcp6_reply(struct dhcp_context *context, int interface, char *iface_name,  
+unsigned short dhcp6_reply(struct dhcp_context *context, int multicast_dest, int interface, char *iface_name,  
 			   struct in6_addr *fallback, struct in6_addr *ll_addr, struct in6_addr *ula_addr,
 			   size_t sz, struct in6_addr *client_addr, time_t now);
 int relay_upstream6(int iface_index, ssize_t sz, struct in6_addr *peer_address, 
@@ -1938,7 +1943,7 @@ size_t make_local_answer(int flags, int gotname, size_t size, struct dns_header
 			 char *name, char *limit, int first, int last, int ede);
 int server_samegroup(struct server *a, struct server *b);
 #ifdef HAVE_DNSSEC
-int dnssec_server(struct server *server, char *keyname, int *firstp, int *lastp);
+int dnssec_server(struct server *server, char *keyname, int is_ds, int *firstp, int *lastp);
 #endif
 void mark_servers(int flag);
 void cleanup_servers(void);

src/domain-match.c

@@ -95,6 +95,7 @@ void build_server_array(void)
 
    A flag of F_SERVER returns an upstream server only.
    A flag of F_DNSSECOK disables NODOTS servers from consideration.
+   A flag of F_DS returns parent domain server.
    A flag of F_DOMAINSRV returns a domain-specific server only.
    A flag of F_CONFIG returns anything that generates a local
    reply of IPv4 or IPV6.
@@ -106,12 +107,23 @@ int lookup_domain(char *domain, int flags, int *lowout, int *highout)
   ssize_t qlen;
   int try, high, low = 0;
   int nlow = 0, nhigh = 0;
-  char *cp, *qdomain = domain;
-
+  char *cp, *qdomain;
+  
   /* may be no configured servers. */
   if (daemon->serverarraysz == 0)
     return 0;
+
+  /* DS records should come from the parent domain. */
+  if (flags & F_DS)
+    {
+      if ((cp = strchr(domain, '.')))
+	domain = cp+1;
+      else
+	domain = "";
+    }
   
+  qdomain = domain;
+
   /* find query length and presence of '.' */
   for (cp = qdomain, nodots = 1, qlen = 0; *cp; qlen++, cp++)
     if (*cp == '.')
@@ -403,7 +415,7 @@ size_t make_local_answer(int flags, int gotname, size_t size, struct dns_header
   
   setup_reply(header, flags, ede);
 
-  gotname &= ~F_QUERY;
+  gotname &= ~(F_QUERY | F_DS);
   
   if (flags & (F_NXDOMAIN | F_NOERR))
     log_query(flags | gotname | F_NEG | F_CONFIG | F_FORWARD, name, NULL, NULL, 0);
@@ -463,14 +475,14 @@ size_t make_local_answer(int flags, int gotname, size_t size, struct dns_header
 }
 
 #ifdef HAVE_DNSSEC
-int dnssec_server(struct server *server, char *keyname, int *firstp, int *lastp)
+int dnssec_server(struct server *server, char *keyname, int is_ds, int *firstp, int *lastp)
 {
   int first, last, index;
-
+  
   /* Find server to send DNSSEC query to. This will normally be the 
      same as for the original query, but may be another if
      servers for domains are involved. */		      
-  if (!lookup_domain(keyname, F_SERVER | F_DNSSECOK, &first, &last))
+  if (!lookup_domain(keyname, F_SERVER | F_DNSSECOK | (is_ds ? F_DS : 0), &first, &last))
     return -1;
 
   for (index = first; index != last; index++)

src/forward.c

@@ -763,6 +763,16 @@ static size_t process_reply(struct dns_header *header, time_t now, struct server
   if (!is_sign && !option_bool(OPT_DNSSEC_PROXY))
      header->hb4 &= ~HB4_AD;
 
+  /* Complain loudly if the upstream server is non-recursive. */
+  if (!(header->hb4 & HB4_RA) && rcode == NOERROR &&
+      server && !(server->flags & SERV_WARNED_RECURSIVE))
+    {
+      (void)prettyprint_addr(&server->addr, daemon->namebuff);
+      my_syslog(LOG_WARNING, _("nameserver %s refused to do a recursive query"), daemon->namebuff);
+      if (!option_bool(OPT_LOG))
+	server->flags |= SERV_WARNED_RECURSIVE;
+    }  
+
   header->hb4 |= HB4_RA; /* recursion if available */
 
   if (OPCODE(header) != QUERY)
@@ -778,16 +788,6 @@ static size_t process_reply(struct dns_header *header, time_t now, struct server
       return resize_packet(header, n, pheader, plen);
     }
   
-  /* Complain loudly if the upstream server is non-recursive. */
-  if (!(header->hb4 & HB4_RA) && rcode == NOERROR &&
-      server && !(server->flags & SERV_WARNED_RECURSIVE))
-    {
-      (void)prettyprint_addr(&server->addr, daemon->namebuff);
-      my_syslog(LOG_WARNING, _("nameserver %s refused to do a recursive query"), daemon->namebuff);
-      if (!option_bool(OPT_LOG))
-	server->flags |= SERV_WARNED_RECURSIVE;
-    }  
-
   if (header->hb3 & HB3_TC)
     log_query(F_UPSTREAM, NULL, NULL, "truncated", 0);
   else if (!bogusanswer || (header->hb4 & HB4_CD))
@@ -1024,7 +1024,7 @@ static void dnssec_validate(struct frec *forward, struct dns_header *header,
 	     	      
 	      /* Make sure we don't expire and free the orig frec during the
 		 allocation of a new one: third arg of get_new_frec() does that. */
-	      if ((serverind = dnssec_server(forward->sentto, daemon->keyname, NULL, NULL)) != -1 &&
+	      if ((serverind = dnssec_server(forward->sentto, daemon->keyname, STAT_ISEQUAL(status, STAT_NEED_DS), NULL, NULL)) != -1 &&
 		  (server = daemon->serverarray[serverind]) &&
 		  (nn = dnssec_generate_query(header, ((unsigned char *) header) + daemon->edns_pktsz,
 					      daemon->keyname, forward->class, get_id(),
@@ -2199,7 +2199,7 @@ int tcp_from_udp(time_t now, int status, struct dns_header *header, ssize_t *ple
   first = start = server->arrayposn;
   last = first + 1;
   
-  if (!STAT_ISEQUAL(status, STAT_OK) && (start = dnssec_server(server, name, &first, &last)) == -1)
+  if (!STAT_ISEQUAL(status, STAT_OK) && (start = dnssec_server(server, name, STAT_ISEQUAL(status, STAT_NEED_DS), &first, &last)) == -1)
     new_status = STAT_ABANDONED;
   else
     {
@@ -2307,7 +2307,7 @@ static int tcp_key_recurse(time_t now, int status, struct dns_header *header, si
       m = dnssec_generate_query(new_header, ((unsigned char *) new_header) + 65536, keyname, class, 0,
 				STAT_ISEQUAL(new_status, STAT_NEED_KEY) ? T_DNSKEY : T_DS);
       
-      if ((start = dnssec_server(server, keyname, &first, &last)) == -1)
+      if ((start = dnssec_server(server, keyname, STAT_ISEQUAL(new_status, STAT_NEED_DS), &first, &last)) == -1)
 	{
 	  new_status = STAT_ABANDONED;
 	  break;

src/option.c

@@ -196,6 +196,7 @@ struct myoption {
 #define LOPT_NO_ENCODE     387
 #define LOPT_DO_ENCODE     388
 #define LOPT_LEASEQUERY    389
+#define LOPT_SPLIT_RELAY   390
 
 #ifdef HAVE_GETOPT_LONG
 static const struct option opts[] =  
@@ -374,6 +375,7 @@ static const struct myoption opts[] =
     { "dnssec-timestamp", 1, 0, LOPT_DNSSEC_STAMP },
     { "dnssec-limits", 1, 0, LOPT_DNSSEC_LIMITS },
     { "dhcp-relay", 1, 0, LOPT_RELAY },
+    { "dhcp-split-relay", 1, 0, LOPT_SPLIT_RELAY },
     { "ra-param", 1, 0, LOPT_RA_PARAM },
     { "quiet-dhcp", 0, 0, LOPT_QUIET_DHCP },
     { "quiet-dhcp6", 0, 0, LOPT_QUIET_DHCP6 },
@@ -542,6 +544,7 @@ static struct {
   { LOPT_GEN_NAMES, ARG_DUP, "[=tag:<tag>]", gettext_noop("Generate hostnames based on MAC address for nameless clients."), NULL},
   { LOPT_PROXY, ARG_DUP, "[=<ipaddr>]...", gettext_noop("Use these DHCP relays as full proxies."), NULL },
   { LOPT_RELAY, ARG_DUP, "<local-addr>,<server>[,<iface>]", gettext_noop("Relay DHCP requests to a remote server"), NULL},
+  { LOPT_SPLIT_RELAY, ARG_DUP, "<local-addr>,<server>,<iface>", gettext_noop("Relay DHCP requests to a remote server"), NULL},
   { LOPT_CNAME, ARG_DUP, "<alias>,<target>[,<ttl>]", gettext_noop("Specify alias name for LOCAL DNS name."), NULL },
   { LOPT_PXE_PROMT, ARG_DUP, "<prompt>,[<timeout>]", gettext_noop("Prompt to send to PXE clients."), NULL },
   { LOPT_PXE_SERV, ARG_DUP, "<service>", gettext_noop("Boot service for PXE menu."), NULL },
@@ -4716,11 +4719,21 @@ static int one_opt(int option, char *arg, char *errstr, char *gen_err, int comma
       break;
       
     case LOPT_RELAY: /* --dhcp-relay */
+    case LOPT_SPLIT_RELAY: /* --dhcp-splt-relay */
       {
 	struct dhcp_relay *new = opt_malloc(sizeof(struct dhcp_relay));
 	char *two = split(arg);
 	char *three = split(two);
-	
+
+	if (option == LOPT_SPLIT_RELAY)
+	  {
+	    new->split_mode = 1;
+	    
+	    /* split mode must have two addresses and a non-wildcard interface name. */
+	    if (!three || strchr(three, '*'))
+	      two = NULL;
+	  }
+		    
 	new->iface_index = 0;
 
 	if (two)
@@ -4743,12 +4756,15 @@ static int one_opt(int option, char *arg, char *errstr, char *gen_err, int comma
 		    else
 		      three = two;
 		  }
+		else if (new->split_mode && inet_pton(AF_INET, three, &new->uplink))
+		  /* Third arg in split mode can be an address. */
+		  three = NULL;
 		
 		new->next = daemon->relay4;
 		daemon->relay4 = new;
 	      }
 #ifdef HAVE_DHCP6
-	    else if (inet_pton(AF_INET6, arg, &new->local))
+	    else if (inet_pton(AF_INET6, arg, &new->local) && !new->split_mode)
 	      {
 		char *hash = split_chr(two, '#');
 
@@ -4769,7 +4785,9 @@ static int one_opt(int option, char *arg, char *errstr, char *gen_err, int comma
 		daemon->relay6 = new;
 	      }
 #endif
-
+	    else
+	      two = NULL;
+	    
 	    new->interface = opt_string_alloc(three);
 	  }
 	
@@ -6262,6 +6280,8 @@ void read_opts(int argc, char **argv, char *compile_opts)
 	    strchr(srv->name, '.') && 
 	    strchr(srv->name, '.') == strrchr(srv->name, '.'))
 	  {
+	    if (strlen(srv->name) + 1 + strlen(daemon->domain_suffix) > MAXDNAME)
+	      die(_("srv-host name %s too long after domain appended"), srv->name, EC_MISC);
 	    strcpy(buff, srv->name);
 	    strcat(buff, ".");
 	    strcat(buff, daemon->domain_suffix);

src/rfc1035.c

@@ -415,6 +415,7 @@ int private_net(struct in_addr addr, int ban_localhost)
     (((ip_addr & 0xFF000000) == 0x7F000000) && ban_localhost)  /* 127.0.0.0/8    (loopback) */ ||
     (((ip_addr & 0xFF000000) == 0x00000000) && ban_localhost) /* RFC 5735 section 3. "here" network */ ||
     ((ip_addr & 0xFF000000) == 0x0A000000)  /* 10.0.0.0/8     (private)  */ ||
+    ((ip_addr & 0xFFC00000) == 0x64400000)  /* 100.64.0.0/10  (CG-NAT) RFC6598/RFC7793*/ ||
     ((ip_addr & 0xFFF00000) == 0xAC100000)  /* 172.16.0.0/12  (private)  */ ||
     ((ip_addr & 0xFFFF0000) == 0xC0A80000)  /* 192.168.0.0/16 (private)  */ ||
     ((ip_addr & 0xFFFF0000) == 0xA9FE0000)  /* 169.254.0.0/16 (zeroconf) */ ||
@@ -515,7 +516,7 @@ int do_doctor(struct dns_header *header, size_t qlen, char *namebuff)
    Cache said SOA and return the difference in length between name and the name of the 
    SOA RR so we can look it up again.
 */
-static int find_soa(struct dns_header *header, size_t qlen, char *name, int *substring, unsigned long *ttlp, int no_cache, time_t now)
+static int find_soa(struct dns_header *header, size_t qlen, char *name, int *substring, unsigned long *ttlp, int cache, time_t now)
 {
   unsigned char *p, *psave;
   int qtype, qclass, rdlen;
@@ -555,7 +556,7 @@ static int find_soa(struct dns_header *header, size_t qlen, char *name, int *sub
 	    {
 	      int prefix = name_len - soa_len;
 	      
-	      if (!no_cache)
+	      if (cache)
 		{
 		  if (!(addr.rrblock.rrdata = blockdata_alloc(NULL, 0)))
 		    return 0;
@@ -567,12 +568,12 @@ static int find_soa(struct dns_header *header, size_t qlen, char *name, int *sub
 		{
 		  if (!extract_name(header, qlen, &p, daemon->workspacename, EXTR_NAME_EXTRACT, 0))
 		    {
-		      if (!no_cache)
+		      if (cache)
 			blockdata_free(addr.rrblock.rrdata);
 		      return 0;
 		    }
 		  
-		  if (!no_cache)
+		  if (cache)
 		    {
 		      len = to_wire(daemon->workspacename);
 		      if (!blockdata_expand(addr.rrblock.rrdata, addr.rrblock.datalen, daemon->workspacename, len))
@@ -587,13 +588,13 @@ static int find_soa(struct dns_header *header, size_t qlen, char *name, int *sub
 
 	      if (!CHECK_LEN(header, p, qlen, 20))
 		{
-		  if (!no_cache)
+		  if (cache)
 		    blockdata_free(addr.rrblock.rrdata);
 		  return 0;
 		}
 	      
 	      /* rest of RR */
-	      if (!no_cache)
+	      if (cache)
 		{
 		  int secflag = 0;
 
@@ -1078,7 +1079,17 @@ int extract_addresses(struct dns_header *header, size_t qlen, char *name, time_t
 	  
 	  if (insert && !option_bool(OPT_NO_NEG))
 	    {
-	      int substring, have_soa = find_soa(header, qlen, name, &substring, &ttl, no_cache_dnssec, now);
+	      /* The order of records going into the cache matters (see  cache_recv_insert()).
+		 The target of a CNAME must immediately follow the CNAME.
+		 (CNAME has already gone into the cache at this point)
+		 Here we call find_soa to get the ttl and substring, but
+		 we DON'T LET IT INSERT the SOA into the cache if our negative record is a CNAME target
+		 so that the SOA doesn't come before the CNAME target.
+
+		 We call find_soa() again after inserting the CNAME target to insert the SOA
+		 if necessary. */
+
+	      int substring, have_soa = find_soa(header, qlen, name, &substring, &ttl, cpp == NULL, now);
 	      
 	      if (have_soa || daemon->neg_ttl)
 		{		  
@@ -1100,6 +1111,10 @@ int extract_addresses(struct dns_header *header, size_t qlen, char *name, time_t
 		      next_uid(newc);
 		      cpp->addr.cname.target.cache = newc;
 		      cpp->addr.cname.uid = newc->uid;
+		    
+		      /* we didn't insert the SOA before a CNAME target above, do it now. */
+		      if (have_soa)
+			find_soa(header, qlen, name, NULL, NULL, 1, now);
 		    }
 		}
 	    }
@@ -1244,12 +1259,10 @@ unsigned int extract_request(struct dns_header *header, size_t qlen, char *name,
 	return  F_IPV4 | F_IPV6;
     }
 
-#ifdef HAVE_DNSSEC
   /* Make the behaviour for DS and DNSKEY queries we forward the same
      as for DS and DNSKEY queries we originate. */
-  if (option_bool(OPT_DNSSEC_VALID) && (qtype == T_DS || qtype == T_DNSKEY))
-    return F_DNSSECOK;
-#endif
+  if (qtype == T_DS || qtype == T_DNSKEY)
+    return F_DNSSECOK | (qtype == T_DS ? F_DS : 0);
   
   return F_QUERY;
 }

src/rfc2131.c

@@ -204,6 +204,10 @@ size_t dhcp_reply(struct dhcp_context *context, char *iface_name, int int_index,
 	      memcpy(agent_id, opt, total);
 	    }
 
+	  /* look for RFC5010 flags sub-option */
+	  if ((sopt = option_find1(option_ptr(opt, 0), option_ptr(opt, option_len(opt)), SUBOPT_FLAGS, INADDRSZ)))
+	    unicast_dest = !!(option_uint(opt, 0, 1) & 0x80);
+	      
 	  /* look for RFC3527 Link selection sub-option */
 	  if ((sopt = option_find1(option_ptr(opt, 0), option_ptr(opt, option_len(opt)), SUBOPT_SUBNET_SELECT, INADDRSZ)))
 	    subnet_addr = option_addr(sopt);
@@ -289,6 +293,8 @@ size_t dhcp_reply(struct dhcp_context *context, char *iface_name, int int_index,
 	{
 	  addr = subnet_addr;
 	  force = 1;
+	  if (mess->giaddr.s_addr)
+	    via_relay = 1;
 	}
       else if (mess->giaddr.s_addr)
 	{
@@ -357,7 +363,12 @@ size_t dhcp_reply(struct dhcp_context *context, char *iface_name, int int_index,
 		  if (context_tmp->local.s_addr == 0)
 		    context_tmp->local = fallback;
 		  if (context_tmp->router.s_addr == 0 && !share)
-		    context_tmp->router = mess->giaddr;
+		    {
+		      if (override.s_addr)
+			context_tmp->router = override;
+		      else
+			context_tmp->router = mess->giaddr;
+		    }
 		  
 		  /* fill in missing broadcast addresses for relayed ranges */
 		  if (!(context_tmp->flags & CONTEXT_BRDCAST) && context_tmp->broadcast.s_addr == 0 )
@@ -3044,4 +3055,207 @@ static void apply_delay(u32 xid, time_t recvtime, struct dhcp_netid *netid)
     }
 }
 
+void relay_upstream4(struct in_addr iface_addr, int iface_index, struct dhcp_packet *mess, size_t sz, int unicast)
+{
+  struct in_addr giaddr = mess->giaddr;
+  u8 hops = mess->hops;
+  struct dhcp_relay *relay;
+  size_t orig_sz = sz;
+  unsigned char *endopt = NULL;
+    
+  if (mess->op != BOOTREQUEST || (mess->hops++) > 20)
+    return;
+  
+  for (relay = daemon->relay4; relay; relay = relay->next)
+    {
+      union mysockaddr to;
+      union all_addr from;
+      struct ifreq ifr;
+
+      /* restore orig packet */
+      mess->giaddr = giaddr;
+      if (endopt)
+	*endopt = OPTION_END;
+      sz = orig_sz;
+
+      if (relay->interface)
+	{
+	  safe_strncpy(ifr.ifr_name, relay->interface, IF_NAMESIZE);
+	  ifr.ifr_addr.sa_family = AF_INET;
+	}
+      
+      if (!relay->split_mode && relay->iface_index && relay->iface_index == iface_index)
+	{
+	  /* already gatewayed ? */
+	  if (giaddr.s_addr)
+	    {
+	      /* if so check if by us, to stomp on loops. */
+	      if (giaddr.s_addr == relay->local.addr4.s_addr)
+		continue;
+	    }
+	  else
+	    /* plug in our address */
+	    mess->giaddr = relay->local.addr4;
+	  
+	  from.addr4 = relay->local.addr4;
+	}
+      else if (relay->split_mode && relay->local.addr4.s_addr == iface_addr.s_addr)
+	{
+	  /* Split mode. We put our address on the server-facing interface
+	     or a directly specified third address into giaddr for the server to talk back to us on.
+	     
+	     Our address on client-facing interface goes into agent-id subnet-selector subopt,
+	     so that the server allocates the correct address. We also send a
+	     remote-id with the interface on which the request arrived,
+	     so that we can send the reply back the same way. */
+	  unsigned int net_index = htonl(iface_index);
+	  
+	  if (relay->interface)
+	    {
+	      /* get our address on the server-facing interface. */
+	      if (ioctl(daemon->dhcpfd, SIOCGIFADDR, &ifr) == -1)
+		continue;
+	      relay->uplink.addr4 = ((struct sockaddr_in *) &ifr.ifr_addr)->sin_addr;
+	    }
+	  
+	  /* already gatewayed ? */
+	  if (giaddr.s_addr)
+	    {
+	      /* if so check if by us, to stomp on loops. */
+	      if (giaddr.s_addr == relay->uplink.addr4.s_addr)
+		continue;
+	    }
+	  else
+	    {
+	      /* giaddr is our address on the outgoing interface in split mode. */
+	      mess->giaddr = relay->uplink.addr4;
+	      
+	      if (!endopt)
+		{
+		  /* Add an RFC3026 relay agent information option (2 bytes) at the very end of the options.
+		     Said option to contain a RFC 3527 link selection sub option (6 bytes) and
+		     RFC 5017 serverid-override option (6 bytes) and RFC5010 flags (3 bytes) and
+		     an RFC3046 remote-id which holds an interface index (6 bytes)
+		     
+		     New END option is a 24th byte, so we need 24 bytes free.
+		     We only need to do this once, and poke the address/interface/flags into the same place each time. */
+		  
+		  if (!(endopt = option_find1((&mess->options[0] + sizeof(u32)), ((unsigned char *)mess) + sz, OPTION_END, 0)) ||
+		      (endopt + 24 > (unsigned char *)(mess + 1)))
+		    continue;
+		  
+		  endopt[1] = 21; /* length */
+		  endopt[2] = SUBOPT_SUBNET_SELECT;
+		  endopt[3] = 4; /* length */
+		  endopt[8] = SUBOPT_SERVER_OR;
+		  endopt[9] = 4;
+		  endopt[14] = SUBOPT_FLAGS;
+		  endopt[15] = 1; /* length */
+		  endopt[17] = SUBOPT_REMOTE_ID;
+		  endopt[18] = 4; /* length */
+		  endopt[23] = OPTION_END;
+		  sz = (endopt - (unsigned char *)mess) + 24;
+		}
+	      
+	      /* IP address is already in network byte order */
+	      memcpy(&endopt[4], &relay->local.addr4.s_addr, INADDRSZ);
+	      memcpy(&endopt[10], &relay->local.addr4.s_addr, INADDRSZ);
+	      endopt[16] = unicast ? 0x80 : 0x00;
+	      memcpy(&endopt[19], &net_index, 4);
+	      endopt[0] = OPTION_AGENT_ID;
+	    }
+	  
+	  from.addr4 = relay->uplink.addr4;
+	}
+      else
+	continue;
+	
+      to.sa.sa_family = AF_INET;
+      to.in.sin_addr = relay->server.addr4;
+      to.in.sin_port = htons(relay->port);
+#ifdef HAVE_SOCKADDR_SA_LEN
+      to.in.sin_len = sizeof(struct sockaddr_in);
+#endif
+      
+      /* Broadcasting to server. */
+      if (relay->server.addr4.s_addr == 0)
+	{
+	  if (ioctl(daemon->dhcpfd, SIOCGIFBRDADDR, &ifr) == -1)
+	    {
+	      my_syslog(MS_DHCP | LOG_ERR, _("Cannot broadcast DHCP relay via interface %s: %s"), relay->interface, strerror(errno));
+	      continue;
+	    }
+	  
+	  to.in.sin_addr = ((struct sockaddr_in *) &ifr.ifr_addr)->sin_addr;
+	}
+      
+#ifdef HAVE_DUMPFILE
+      {
+	union mysockaddr fromsock;
+	fromsock.in.sin_port = htons(daemon->dhcp_server_port);
+	fromsock.in.sin_addr = from.addr4;
+	fromsock.sa.sa_family = AF_INET;
+	
+	dump_packet_udp(DUMP_DHCP, (void *)mess, sz, &fromsock, &to, -1);
+      }
+#endif
+      
+      send_from(daemon->dhcpfd, 0, (char *)mess, sz, &to, &from, 0);
+      
+      if (option_bool(OPT_LOG_OPTS))
+	{
+	  inet_ntop(AF_INET, &relay->local, daemon->addrbuff, ADDRSTRLEN);
+	  if (relay->server.addr4.s_addr == 0)
+	    snprintf(daemon->dhcp_buff2, DHCP_BUFF_SZ, _("broadcast via %s"), relay->interface);
+	  else
+	    inet_ntop(AF_INET, &relay->server.addr4, daemon->dhcp_buff2, DHCP_BUFF_SZ);
+	  my_syslog(MS_DHCP | LOG_INFO, _("DHCP relay at %s -> %s"), daemon->addrbuff, daemon->dhcp_buff2);
+	}
+    }
+  
+  /* restore in case of a local reply. */
+  mess->hops = hops;
+  mess->giaddr = giaddr;
+  if (endopt)
+    *endopt = OPTION_END;
+}
+
+unsigned int relay_reply4(struct dhcp_packet *mess, size_t sz, char *arrival_interface)
+{
+  struct dhcp_relay *relay;
+    
+  if (mess->giaddr.s_addr == 0 || mess->op != BOOTREPLY)
+    return 0;
+
+  for (relay = daemon->relay4; relay; relay = relay->next)
+    {
+      unsigned int return_iface = 0;
+
+      if (relay->split_mode)
+	{
+	  unsigned char *opt, *sopt;
+	  
+	  /* giaddr is our address on the returning interface in split mode. */
+	  if (mess->giaddr.s_addr == relay->uplink.addr4.s_addr &&
+	      (opt = option_find(mess, sz, OPTION_AGENT_ID, 1)))
+	    {
+	      if ((sopt = option_find1(option_ptr(opt, 0), option_ptr(opt, option_len(opt)), SUBOPT_REMOTE_ID, sizeof(unsigned int))))
+		return_iface = option_uint(sopt, 0, sizeof(unsigned int));
+
+	      /* delete agent info before return RFC 3046 para 2.1 */
+	      *opt = OPTION_END;
+	      memset(opt + 1, 0, option_len(opt) + 2);
+	    }
+	}
+      else if (mess->giaddr.s_addr == relay->local.addr4.s_addr)
+	    return_iface = relay->iface_index;
+      
+      if (return_iface && (!relay->interface || wildcard_match(relay->interface, arrival_interface)))
+	return return_iface;
+    }
+  
+  return 0;	 
+}     
+
+
 #endif /* HAVE_DHCP */

src/rfc3315.c

@@ -21,7 +21,7 @@
 
 struct state {
   unsigned char *clid;
-  int clid_len, ia_type, interface, hostname_auth, lease_allocate;
+  int multicast_dest, clid_len, ia_type, interface, hostname_auth, lease_allocate;
   char *client_hostname, *hostname, *domain, *send_domain;
   struct dhcp_context *context;
   struct in6_addr *link_address, *fallback, *ll_addr, *ula_addr;
@@ -68,7 +68,7 @@ static void calculate_times(struct dhcp_context *context, unsigned int *min_time
 #define opt6_user_vendor_next(opt, end) (opt6_next(((uint8_t *) opt) - 2, end))
  
 
-unsigned short dhcp6_reply(struct dhcp_context *context, int interface, char *iface_name,
+unsigned short dhcp6_reply(struct dhcp_context *context, int multicast_dest, int interface, char *iface_name,
 			   struct in6_addr *fallback,  struct in6_addr *ll_addr, struct in6_addr *ula_addr,
 			   size_t sz, struct in6_addr *client_addr, time_t now)
 {
@@ -87,6 +87,7 @@ unsigned short dhcp6_reply(struct dhcp_context *context, int interface, char *if
   
   reset_counter();
   state.context = context;
+  state.multicast_dest = multicast_dest;
   state.interface = interface;
   state.iface_name = iface_name;
   state.fallback = fallback;
@@ -244,7 +245,11 @@ static int dhcp6_maybe_relay(struct state *state, unsigned char *inbuff, size_t
 	      struct in6_addr align;
 	      /* the packet data is unaligned, copy to aligned storage */
 	      memcpy(&align, inbuff + 2, IN6ADDRSZ); 
-	      state->link_address = &align;
+
+
+	      /* RFC6221 para 4 */
+	      if (!IN6_IS_ADDR_UNSPECIFIED(&align))
+		state->link_address = &align;
 	      /* zero is_unicast since that is now known to refer to the 
 		 relayed packet, not the original sent by the client */
 	      if (!dhcp6_maybe_relay(state, opt6_ptr(opt, 0), opt6_len(opt), client_addr, 0, now))
@@ -333,20 +338,24 @@ static int dhcp6_no_relay(struct state *state, int msg_type, unsigned char *inbu
   else if (msg_type != DHCP6IREQ)
     return 0;
 
-  /* server-id must match except for SOLICIT, CONFIRM and REBIND messages, which MUST NOT
-     have a server-id.  3315 para 15.x */
   opt = opt6_find(state->packet_options, state->end, OPTION6_SERVER_ID, 1);
-
-  if (msg_type == DHCP6SOLICIT || msg_type == DHCP6CONFIRM || msg_type == DHCP6REBIND)
+  
+  if (msg_type == DHCP6SOLICIT || msg_type == DHCP6CONFIRM || msg_type == DHCP6REBIND || msg_type == DHCP6IREQ)
     {
-      if (opt)
+      /* Above message types must be multicast 3315 Section 15. */
+      if (!state->multicast_dest)
 	return 0;
-    }
-  else if (msg_type == DHCP6IREQ)
-    {
-      /* If server-id provided, it must match. */
-      if (opt && (opt6_len(opt) != daemon->duid_len ||
-		  memcmp(opt6_ptr(opt, 0), daemon->duid, daemon->duid_len) != 0))
+
+      /* server-id must match except for SOLICIT, CONFIRM and REBIND messages, which MUST NOT
+	 have a server-id.  3315 para 15.x */
+      if (msg_type == DHCP6IREQ)
+	{
+	  /* If server-id provided in IREQ, it must match. */
+	  if (opt && (opt6_len(opt) != daemon->duid_len ||
+		      memcmp(opt6_ptr(opt, 0), daemon->duid, daemon->duid_len) != 0))
+	    return 0;
+	}
+      else if (opt) 
 	return 0;
     }
   else
@@ -634,7 +643,7 @@ static int dhcp6_no_relay(struct state *state, int msg_type, unsigned char *inbu
       
     case DHCP6SOLICIT:
       {
-      	int address_assigned = 0;
+      	int address_assigned;
 	/* tags without all prefix-class tags */
 	struct dhcp_netid *solicit_tags;
 	struct dhcp_context *c;
@@ -653,6 +662,7 @@ static int dhcp6_no_relay(struct state *state, int msg_type, unsigned char *inbu
 
       request_no_address:
 	solicit_tags = tagif;
+	address_assigned = 0;
 	
 	if (ignore)
 	  return 0;
@@ -764,15 +774,6 @@ static int dhcp6_no_relay(struct state *state, int msg_type, unsigned char *inbu
 	    
 	    if (address_assigned != 1)
 	      {
-		/* If the server will not assign any addresses to any IAs in a
-		   subsequent Request from the client, the server MUST send an Advertise
-		   message to the client that doesn't include any IA options. */
-		if (!state->lease_allocate)
-		  {
-		    save_counter(o);
-		    continue;
-		  }
-		
 		/* If the server cannot assign any addresses to an IA in the message
 		   from the client, the server MUST include the IA in the Reply message
 		   with no addresses in the IA and a Status Code option in the IA
@@ -799,7 +800,6 @@ static int dhcp6_no_relay(struct state *state, int msg_type, unsigned char *inbu
 	    o = new_opt6(OPTION6_PREFERENCE);
 	    put_opt6_char(option_bool(OPT_AUTHORITATIVE) ? 255 : 0);
 	    end_opt6(o);
-	    tagif = add_options(state, 0);
 	  }
 	else
 	  { 
@@ -819,7 +819,8 @@ static int dhcp6_no_relay(struct state *state, int msg_type, unsigned char *inbu
 		  break;
 		}
 	  }
-
+	
+	tagif = add_options(state, 0);
 	break;
       }
       
@@ -977,8 +978,10 @@ static int dhcp6_no_relay(struct state *state, int msg_type, unsigned char *inbu
 		  {
 		    if (msg_type == DHCP6REBIND)
 		      {
-			/* When rebinding, we can create a lease if it doesn't exist. */
-			lease = lease6_allocate(&req_addr, state->ia_type == OPTION6_IA_NA ? LEASE_NA : LEASE_TA);
+			/* When rebinding, we can create a lease if it doesn't exist, as long
+			   as --dhcp-authoritative is set. */
+			if (option_bool(OPT_AUTHORITATIVE))
+			  lease = lease6_allocate(&req_addr, state->ia_type == OPTION6_IA_NA ? LEASE_NA : LEASE_TA);
 			if (lease)
 			  lease_set_iaid(lease, state->iaid);
 			else
@@ -2276,64 +2279,61 @@ int relay_reply6(struct sockaddr_in6 *peer, ssize_t sz, char *arrival_interface)
     {
       void *opt, *opts = inbuff + 34;
       void *end = inbuff + sz;
-      for (opt = opts; opt; opt = opt6_next(opt, end))
-	if (opt6_type(opt) == OPTION6_RELAY_MSG && opt6_len(opt) > 0)
-	  {
-	    int encap_type = *((unsigned char *)opt6_ptr(opt, 0));
-	    put_opt6(opt6_ptr(opt, 0), opt6_len(opt));
-	    memcpy(&peer->sin6_addr, &inbuff[18], IN6ADDRSZ); 
-	    peer->sin6_scope_id = relay->iface_index;
-
-	    if (encap_type == DHCP6RELAYREPL)
-	      {
-		peer->sin6_port = ntohs(DHCPV6_SERVER_PORT);
-		return 1;
-	      }
-
-	    peer->sin6_port = ntohs(DHCPV6_CLIENT_PORT);
-	    
-#ifdef HAVE_SCRIPT
-	    if (daemon->lease_change_command && encap_type == DHCP6REPLY)
-	      {
-		/* decapsulate relayed message */
-		opts = opt6_ptr(opt, 4);
-		end = opt6_ptr(opt, opt6_len(opt));
-
-		for (opt = opts; opt; opt = opt6_next(opt, end))
-		  if (opt6_type(opt) == OPTION6_IA_PD && opt6_len(opt) > 12) 
-		    {
-		      void *ia_opts = opt6_ptr(opt, 12);
-		      void *ia_end = opt6_ptr(opt, opt6_len(opt));
-		      void *ia_opt;
-		      
-		      for (ia_opt = ia_opts; ia_opt; ia_opt = opt6_next(ia_opt, ia_end))
-			/* valid lifetime must not be zero. */
-			if (opt6_type(ia_opt) == OPTION6_IAPREFIX && opt6_len(ia_opt) >= 25 && opt6_uint(ia_opt, 4, 4) != 0)
-			  {
-			    if (daemon->free_snoops ||
-				(daemon->free_snoops = whine_malloc(sizeof(struct snoop_record))))
-			      {
-				struct snoop_record *snoop = daemon->free_snoops;
-				
-				daemon->free_snoops = snoop->next;
-				snoop->client = peer->sin6_addr;
-				snoop->prefix_len = opt6_uint(ia_opt, 8, 1); 
-				memcpy(&snoop->prefix, opt6_ptr(ia_opt, 9), IN6ADDRSZ); 
-				snoop->next = relay->snoop_records;
-				relay->snoop_records = snoop;
-			      }
-			  }
-		    }
-	      }
-#endif		
-	    return 1;
-	  }
       
+      if ((opt = opt6_find(opts, end, OPTION6_RELAY_MSG, 4)))
+	{
+	  int encap_type = opt6_uint(opt, 0, 1);
+	  put_opt6(opt6_ptr(opt, 0), opt6_len(opt));
+	  memcpy(&peer->sin6_addr, &inbuff[18], IN6ADDRSZ); 
+	  peer->sin6_scope_id = relay->iface_index;
+
+	  if (encap_type == DHCP6RELAYREPL)
+	    {
+	      peer->sin6_port = ntohs(DHCPV6_SERVER_PORT);
+	      return 1;
+	    }
+	  
+	  peer->sin6_port = ntohs(DHCPV6_CLIENT_PORT);
+	  
+#ifdef HAVE_SCRIPT
+	  if (daemon->lease_change_command && encap_type == DHCP6REPLY)
+	    {
+	      /* skip over message type and transaction-id. to get to options. */
+	      opts = opt6_ptr(opt, 4);
+	      end = opt6_ptr(opt, opt6_len(opt));
+
+	      if ((opt = opt6_find(opts, end, OPTION6_IA_PD, 12)))
+		{
+		  opts = opt6_ptr(opt, 12);
+		  end = opt6_ptr(opt, opt6_len(opt));
+		  
+		  for (opt = opt6_find(opts, end, OPTION6_IAPREFIX, 25); opt; opt = opt6_find(opt6_next(opt, end), end, OPTION6_IAPREFIX, 25))
+		    /* valid lifetime must not be zero. */
+		    if (opt6_uint(opt, 4, 4) != 0)
+		      {
+			if (daemon->free_snoops ||
+			    (daemon->free_snoops = whine_malloc(sizeof(struct snoop_record))))
+			  {
+			    struct snoop_record *snoop = daemon->free_snoops;
+			    
+			    daemon->free_snoops = snoop->next;
+			    snoop->client = peer->sin6_addr;
+			    snoop->prefix_len = opt6_uint(opt, 8, 1); 
+			    memcpy(&snoop->prefix, opt6_ptr(opt, 9), IN6ADDRSZ); 
+			    snoop->next = relay->snoop_records;
+			    relay->snoop_records = snoop;
+			  }
+		      }
+		}	      
+	    }
+#endif
+	  return 1;
+	}
     }
   
   return 0;
 }
-
+  
 #ifdef HAVE_SCRIPT
 int do_snoop_script_run(void)
 {