Add UDP datagram relay support to SPTPS.

This commit changes the layout of UDP datagrams to include a 6-byte
destination node ID at the very beginning of the datagram (i.e. before
the source node ID and the seqno). Note that this only applies to SPTPS.

Thanks to this new field, it is now possible to send SPTPS datagrams to
nodes that are not the final recipient of the packets, thereby using
these nodes as relay nodes. Previously SPTPS was unable to relay packets
using UDP, and required a fallback to TCP if the final recipient could
not be contacted directly using UDP. In that sense it fixes a regression
that SPTPS introduced with regard to the legacy protocol.

This change also updates tinc's low-level routing logic (i.e.
send_sptps_data()) to automatically use this relaying facility if at all
possible. Specifically, it will relay packets if we don't have a
confirmed UDP link to the final recipient (but we have one with the next
hop node), or if IndirectData is specified. This is similar to how the
legacy protocol forwards packets.

When sending packets directly without any relaying, the sender node uses
a special value for the destination node ID: instead of setting the
field to the ID of the recipient node, it writes a zero ID instead. This
allows the recipient node to distinguish between a relayed packet and a
direct packet, which is important when determining the UDP address of
the sending node.

On the relay side, relay nodes will happily relay packets that have a
destination ID which is non-zero *and* is different from their own,
provided that the source IP address of the packet is known. This is to
prevent abuse by random strangers, since a node can't authenticate the
packets that are being relayed through it.

This change keeps the protocol number from the previous datagram format
change (source IDs), 17.4. Compatibility is still preserved with 1.0 and
with pre-1.1 releases. Note, however, that nodes running this code won't
understand datagrams sent from nodes that only use source IDs and
vice-versa (not that we really care).

There is one caveat: in the current state, there is no way for the
original sender to know what the PMTU is beyond the first hop, and
contrary to the legacy protocol, relay nodes can't apply MSS clamping
because they can't decrypt the relayed packets. This leads to
inefficient scenarios where a reduced PMTU over some link that's part of
the relay path will result in relays falling back to TCP to send packets
to their final destinations.

Another caveat is that once a packet gets sent over TCP, it will use
TCP over the entire path, even if it is technically possible to use UDP
beyond the TCP-only link(s).

Arguably, these two caveats can be fixed by improving the
metaconnection protocol, but that's out of scope for this change. TODOs
are added instead. In any case, this is no worse than before.

In addition, this change increases SPTPS datagram overhead by another
6 bytes for the destination ID, on top of the existing 6-byte overhead
from the source ID.

diff --git a/src/net.h b/src/net.h
index adb65f30023431e259c2ff10817e1e0f7212ad3d..b570d5e1c584e7dcfa17a8a0fc892f89bd3fde06 100644 (file)
--- a/src/net.h
+++ b/src/net.h
@@ -32,8 +32,8 @@
 #define MTU 1518        /* 1500 bytes payload + 14 bytes ethernet header + 4 bytes VLAN tag */
 #endif
 
-/* MAXSIZE is the maximum size of an encapsulated packet: MTU + seqno + srcid + padding + HMAC + compressor overhead */
-#define MAXSIZE (MTU + 4 + sizeof(node_id_t) + CIPHER_MAX_BLOCK_SIZE + DIGEST_MAX_SIZE + MTU/64 + 20)
+/* MAXSIZE is the maximum size of an encapsulated packet: MTU + seqno + srcid + dstid + padding + HMAC + compressor overhead */
+#define MAXSIZE (MTU + 4 + sizeof(node_id_t) + sizeof(node_id_t) + CIPHER_MAX_BLOCK_SIZE + DIGEST_MAX_SIZE + MTU/64 + 20)
 
 /* MAXBUFSIZE is the maximum size of a request: enough for a MAXSIZEd packet or a 8192 bits RSA key */
 #define MAXBUFSIZE ((MAXSIZE > 2048 ? MAXSIZE : 2048) + 128)
@@ -87,6 +87,7 @@ typedef union sockaddr_t {
 typedef struct vpn_packet_t {
        length_t len;           /* the actual number of bytes in the `data' field */
        int priority;           /* priority or TOS */
+       node_id_t dstid;        /* node ID of the final recipient */
        node_id_t srcid;        /* node ID of the original sender */
        uint8_t seqno[4];       /* 32 bits sequence number (network byte order of course) */
        uint8_t data[MAXSIZE];

diff --git a/src/net_packet.c b/src/net_packet.c
index 6b7c3b94558455a89cca189ced9ef36467908bd4..a1b174e977dca5b65ca0aa56b3f2212e274878b5 100644 (file)
--- a/src/net_packet.c
+++ b/src/net_packet.c
@@ -751,54 +751,62 @@ end:
        origpkt->len = origlen;
 }
 
-bool send_sptps_data(void *handle, uint8_t type, const char *data, size_t len) {
-       node_t *to = handle;
+static bool send_sptps_data_priv(node_t *to, node_t *from, int type, const char *data, size_t len) {
+       node_t *relay = (to->via != myself && (type == PKT_PROBE || (len - SPTPS_DATAGRAM_OVERHEAD) <= to->via->minmtu)) ? to->via : to->nexthop;
+       bool direct = from == myself && to == relay;
+       bool relay_supported = (relay->options >> 24) >= 4;
 
-       /* Send it via TCP if it is a handshake packet, TCPOnly is in use, or this packet is larger than the MTU. */
+       /* Send it via TCP if it is a handshake packet, TCPOnly is in use, this is a relay packet that the other node cannot understand, or this packet is larger than the MTU.
+          TODO: When relaying, the original sender does not know the end-to-end PMTU (it only knows the PMTU of the first hop).
+                This can lead to scenarios where large packets are sent over UDP to relay, but then relay has no choice but fall back to TCP. */
 
-       if(type >= SPTPS_HANDSHAKE || ((myself->options | to->options) & OPTION_TCPONLY) || (type != PKT_PROBE && (len - SPTPS_DATAGRAM_OVERHEAD) > to->minmtu)) {
+       if(type == SPTPS_HANDSHAKE || ((myself->options | relay->options) & OPTION_TCPONLY) || (!direct && !relay_supported) || (type != PKT_PROBE && (len - SPTPS_DATAGRAM_OVERHEAD) > relay->minmtu)) {
                char buf[len * 4 / 3 + 5];
                b64encode(data, buf, len);
                /* If no valid key is known yet, send the packets using ANS_KEY requests,
                   to ensure we get to learn the reflexive UDP address. */
-               if(!to->status.validkey) {
+               if(from == myself && !to->status.validkey) {
                        to->incompression = myself->incompression;
-                       return send_request(to->nexthop->connection, "%d %s %s %s -1 -1 -1 %d", ANS_KEY, myself->name, to->name, buf, to->incompression);
+                       return send_request(to->nexthop->connection, "%d %s %s %s -1 -1 -1 %d", ANS_KEY, from->name, to->name, buf, to->incompression);
                } else {
-                       return send_request(to->nexthop->connection, "%d %s %s %d %s", REQ_KEY, myself->name, to->name, REQ_SPTPS, buf);
+                       return send_request(to->nexthop->connection, "%d %s %s %d %s", REQ_KEY, from->name, to->name, REQ_SPTPS, buf);
                }
        }
 
-       /* Otherwise, send the packet via UDP */
-
-       const sockaddr_t *sa = NULL;
-       int sock;
-
-       if(to->status.send_locally)
-               choose_local_address(to, &sa, &sock);
-       if(!sa)
-               choose_udp_address(to, &sa, &sock);
-
-       bool add_srcid = (to->options >> 24) >= 4;
        size_t overhead = 0;
-       if (add_srcid) overhead += sizeof myself->id;
+       if(relay_supported) overhead += sizeof to->id + sizeof from->id;
        char buf[len + overhead]; char* buf_ptr = buf;
-       if(add_srcid) {
-               memcpy(buf_ptr, &myself->id, sizeof myself->id); buf_ptr += sizeof myself->id;
+       if(relay_supported) {
+               if(direct) {
+                       /* Inform the recipient that this packet was sent directly. */
+                       node_id_t nullid = {0};
+                       memcpy(buf_ptr, &nullid, sizeof nullid); buf_ptr += sizeof nullid;
+               } else {
+                       memcpy(buf_ptr, &to->id, sizeof to->id); buf_ptr += sizeof to->id;
+               }
+               memcpy(buf_ptr, &from->id, sizeof from->id); buf_ptr += sizeof from->id;
+
        }
        /* TODO: if this copy turns out to be a performance concern, change sptps_send_record() to add some "pre-padding" to the buffer and use that instead */
        memcpy(buf_ptr, data, len); buf_ptr += len;
 
+       const sockaddr_t *sa = NULL;
+       int sock;
+       if(relay->status.send_locally)
+               choose_local_address(relay, &sa, &sock);
+       if(!sa)
+               choose_udp_address(relay, &sa, &sock);
+       logger(DEBUG_TRAFFIC, LOG_INFO, "Sending packet from %s (%s) to %s (%s) via %s (%s)", from->name, from->hostname, to->name, to->hostname, relay->name, relay->hostname);
        if(sendto(listen_socket[sock].udp.fd, buf, buf_ptr - buf, 0, &sa->sa, SALEN(sa->sa)) < 0 && !sockwouldblock(sockerrno)) {
                if(sockmsgsize(sockerrno)) {
                        // Compensate for SPTPS overhead
                        len -= SPTPS_DATAGRAM_OVERHEAD;
-                       if(to->maxmtu >= len)
-                               to->maxmtu = len - 1;
-                       if(to->mtu >= len)
-                               to->mtu = len - 1;
+                       if(relay->maxmtu >= len)
+                               relay->maxmtu = len - 1;
+                       if(relay->mtu >= len)
+                               relay->mtu = len - 1;
                } else {
-                       logger(DEBUG_TRAFFIC, LOG_WARNING, "Error sending UDP SPTPS packet to %s (%s): %s", to->name, to->hostname, sockstrerror(sockerrno));
+                       logger(DEBUG_TRAFFIC, LOG_WARNING, "Error sending UDP SPTPS packet to %s (%s): %s", relay->name, relay->hostname, sockstrerror(sockerrno));
                        return false;
                }
        }
@@ -806,6 +814,10 @@ bool send_sptps_data(void *handle, uint8_t type, const char *data, size_t len) {
        return true;
 }
 
+bool send_sptps_data(void *handle, uint8_t type, const char *data, size_t len) {
+       return send_sptps_data_priv(handle, myself, type, data, len);
+}
+
 bool receive_sptps_record(void *handle, uint8_t type, const char *data, uint16_t len) {
        node_t *from = handle;
 
@@ -1006,9 +1018,10 @@ void handle_incoming_vpn_data(void *data, int flags) {
        sockaddr_t from = {{0}};
        socklen_t fromlen = sizeof from;
        node_t *n = NULL;
+       node_t *to = myself;
        int len;
 
-       len = recvfrom(ls->udp.fd, &pkt.srcid, MAXSIZE, 0, &from.sa, &fromlen);
+       len = recvfrom(ls->udp.fd, &pkt.dstid, MAXSIZE, 0, &from.sa, &fromlen);
 
        if(len <= 0 || len > MAXSIZE) {
                if(!sockwouldblock(sockerrno))
@@ -1020,13 +1033,42 @@ void handle_incoming_vpn_data(void *data, int flags) {
 
        sockaddrunmap(&from); /* Some braindead IPv6 implementations do stupid things. */
 
-       if(len >= sizeof pkt.srcid)
+       bool direct = false;
+       if(len >= sizeof pkt.dstid + sizeof pkt.srcid) {
                n = lookup_node_id(&pkt.srcid);
-       if(n)
-               pkt.len -= sizeof pkt.srcid;
-       else {
-               /* Most likely an old-style packet without a source ID. */
-               memmove(pkt.seqno, &pkt.srcid, sizeof pkt - offsetof(vpn_packet_t, seqno));
+               if(n) {
+                       node_id_t nullid = {0};
+                       if(memcmp(&pkt.dstid, &nullid, sizeof nullid) == 0) {
+                               /* A zero dstid is used to indicate a direct, non-relayed packet. */
+                               direct = true;
+                       } else {
+                               to = lookup_node_id(&pkt.dstid);
+                               if(!to) {
+                                       logger(DEBUG_PROTOCOL, LOG_WARNING, "Received UDP packet presumably sent by %s (%s) but with unknown destination ID", n->name, n->hostname);
+                                       return;
+                               }
+                       }
+                       pkt.len -= sizeof pkt.dstid + sizeof pkt.srcid;
+               }
+       }
+
+       if(to != myself) {
+               /* We are being asked to relay this packet. */
+
+               /* Don't allow random strangers to relay through us. Note that we check for *any* known address since we are not necessarily the first relay. */
+               if (!lookup_node_udp(&from)) {
+                       logger(DEBUG_PROTOCOL, LOG_WARNING, "Refusing to relay packet from (presumably) %s (%s) to (presumably) %s (%s) because the packet comes from an unknown address", n->name, n->hostname, to->name, to->hostname);
+                       return;
+               }
+
+               send_sptps_data_priv(to, n, 0, pkt.seqno, pkt.len);
+               return;
+       }
+
+       if(!n) {
+               /* Most likely an old-style packet without node IDs. */
+               direct = true;
+               memmove(pkt.seqno, &pkt.dstid, sizeof pkt - offsetof(vpn_packet_t, seqno));
                n = lookup_node_udp(&from);
        }
 
@@ -1046,7 +1088,7 @@ void handle_incoming_vpn_data(void *data, int flags) {
                return;
 
        n->sock = ls - listen_socket;
-       if(sockaddrcmp(&from, &n->address))
+       if(direct && sockaddrcmp(&from, &n->address))
                update_node_udp(n, &from);
 }
 

diff --git a/src/protocol_key.c b/src/protocol_key.c
index abde7772c40aedcd8f8dcb050ae3755b354cba01..a4d97aeb7f3b39b77b0b48916ce7564a2592bacc 100644 (file)
--- a/src/protocol_key.c
+++ b/src/protocol_key.c
@@ -255,6 +255,7 @@ bool req_key_h(connection_t *c, const char *request) {
                        return true;
                }
 
+               /* TODO: forwarding SPTPS packets in this way is inefficient because we send them over TCP without checking for UDP connectivity */
                send_request(to->nexthop->connection, "%s", request);
        }