ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/gvpe/doc/gvpe.protocol.7.pod
(Generate patch)

Comparing gvpe/doc/gvpe.protocol.7.pod (file contents):
Revision 1.2 by pcg, Tue Mar 15 19:23:33 2005 UTC vs.
Revision 1.7 by pcg, Mon Sep 1 05:31:28 2008 UTC

… …
8document. 8document.
9 9
10The first part of this document describes the transport protocols which 10The first part of this document describes the transport protocols which
11are used by GVPE to send it's data packets over the network. 11are used by GVPE to send it's data packets over the network.
12 12
13=head1 PART 1: Tansport protocols 13=head1 PART 1: Transport protocols
14
15GVPE offers a wide range of transport protocols that can be used to
16interchange data between nodes. Protocols differ in their overhead, speed,
17reliability, and robustness.
18
19The following sections describe each transport protocol in more
20detail. They are sorted by overhead/efficiency, the most efficient
21transport is listed first:
14 22
15=head2 RAW IP 23=head2 RAW IP
16 24
25This protocol is the best choice, performance-wise, as the minimum
26overhead per packet is only 38 bytes.
27
28It works by sending the VPN payload using raw IP frames (using the
29protocol set by C<ip-proto>).
30
31Using raw IP frames has the drawback that many firewalls block "unknown"
32protocols, so this transport only works if you have full IP connectivity
33between nodes.
34
17=head2 ICMP 35=head2 ICMP
18 36
37This protocol offers very low overhead (minimum 42 bytes), and can
38sometimes tunnel through firewalls when other protocols can not.
39
40It works by prepending an ICMP header with type C<icmp-type> and a code
41of C<255>. The default C<icmp-type> is C<echo-reply>, so the resulting
42packets look like echo replies, which looks rather strange to network
43administrators.
44
45This transport should only be used if other transports (i.e. raw IP) are
46not available or undesirable (due to their overhead).
47
19=head2 UDP 48=head2 UDP
20 49
50This is a good general choice for the transport protocol as UDP packets
51tunnel well through most firewalls and routers, and the overhead per
52packet is moderate (minimum 58 bytes).
53
54It should be used if RAW IP is not available.
55
21=head2 TCP 56=head2 TCP
22 57
58This protocol is a very bad choice, as it not only has high overhead (more
59than 60 bytes), but the transport also retries on it's own, which leads
60to congestion when the link has moderate packet loss (as both the TCP
61transport and the tunneled traffic will retry, increasing congestion more
62and more). It also has high latency and is quite inefficient.
63
64It's only useful when tunneling through firewalls that block better
65protocols. If a node doesn't have direct internet access but a HTTP proxy
66that supports the CONNECT method it can be used to tunnel through a web
67proxy. For this to work, the C<tcp-port> should be C<443> (C<https>), as
68most proxies do not allow connections to other ports.
69
70It is an abuse of the usage a proxy was designed for, so make sure you are
71allowed to use it for GVPE.
72
73This protocol also has server and client sides. If the C<tcp-port> is
74set to zero, other nodes cannot connect to this node directly. If the
75C<tcp-port> is non-zero, the node can act both as a client as well as a
76server.
77
23=head2 DNS 78=head2 DNS
79
80B<WARNING:> Parsing and generating DNS packets is rather tricky. The code
81almost certainly contains buffer overflows and other, likely exploitable,
82bugs. You have been warned.
83
84This is the worst choice of transport protocol with respect to overhead
85(overhead can be 2-3 times higher than the transferred data), and latency
86(which can be many seconds). Some DNS servers might not be prepared to
87handle the traffic and drop or corrupt packets. The client also has to
88constantly poll the server for data, so the client will constantly create
89traffic even if it doesn't need to transport packets.
90
91In addition, the same problems as the TCP transport also plague this
92protocol.
93
94It's only use is to tunnel through firewalls that do not allow direct
95internet access. Similar to using a HTTP proxy (as the TCP transport
96does), it uses a local DNS server/forwarder (given by the C<dns-forw-host>
97configuration value) as a proxy to send and receive data as a client,
98and an C<NS> record pointing to the GVPE server (as given by the
99C<dns-hostname> directive).
100
101The only good side of this protocol is that it can tunnel through most
102firewalls mostly undetected, iff the local DNS server/forwarder is sane
103(which is true for most routers, wireless LAN gateways and nameservers).
104
105Fine-tuning needs to be done by editing C<src/vpn_dns.C> directly.
24 106
25=head1 PART 2: The GNU VPE protocol 107=head1 PART 2: The GNU VPE protocol
26 108
27This section, unfortunately, is not yet finished, although the protocol 109This section, unfortunately, is not yet finished, although the protocol
28is stable (until bugs in the cryptography are found, which will likely 110is stable (until bugs in the cryptography are found, which will likely
… …
30you some overview over the protocol. 112you some overview over the protocol.
31 113
32=head2 Anatomy of a VPN packet 114=head2 Anatomy of a VPN packet
33 115
34The exact layout and field lengths of a VPN packet is determined at 116The exact layout and field lengths of a VPN packet is determined at
35compiletime and doesn't change. The same structure is used for all 117compile time and doesn't change. The same structure is used for all
36transort protocols, be it RAWIP or TCP. 118transport protocols, be it RAWIP or TCP.
37 119
38 +------+------+--------+------+ 120 +------+------+--------+------+
39 | HMAC | TYPE | SRCDST | DATA | 121 | HMAC | TYPE | SRCDST | DATA |
40 +------+------+--------+------+ 122 +------+------+--------+------+
41 123
… …
46The TYPE field is a single byte and determines the purpose of the packet 128The TYPE field is a single byte and determines the purpose of the packet
47(e.g. RESET, COMPRESSED/UNCOMPRESSED DATA, PING, AUTH REQUEST/RESPONSE, 129(e.g. RESET, COMPRESSED/UNCOMPRESSED DATA, PING, AUTH REQUEST/RESPONSE,
48CONNECT REQUEST/INFO etc.). 130CONNECT REQUEST/INFO etc.).
49 131
50SRCDST is a three byte field which contains the source and destination 132SRCDST is a three byte field which contains the source and destination
51node ids (12 bits each). The protocol does not yet scale well beyond 30+ 133node IDs (12 bits each).
52hosts, since all hosts must connect to each other once on startup. But if
53restarts are rare or tolerable and most connections are on demand, much
54larger networks are feasible.
55 134
56The DATA portion differs between each packet type, naturally, and is the 135The DATA portion differs between each packet type, naturally, and is the
57only part that can be encrypted. Data packets contain more fields, as 136only part that can be encrypted. Data packets contain more fields, as
58shown: 137shown:
59 138
… …
65the data for encryption purposes. 144the data for encryption purposes.
66 145
67SEQNO is a 32-bit sequence number. It is negotiated at every connection 146SEQNO is a 32-bit sequence number. It is negotiated at every connection
68initialization and starts at some random 31 bit value. VPE currently uses 147initialization and starts at some random 31 bit value. VPE currently uses
69a sliding window of 512 packets/sequence numbers to detect reordering, 148a sliding window of 512 packets/sequence numbers to detect reordering,
70duplication and reply attacks. 149duplication and replay attacks.
71 150
72=head2 The authentification protocol 151=head2 The authentication protocol
73 152
74Before hosts can exchange packets, they need to establish authenticity of 153Before nodes can exchange packets, they need to establish authenticity of
75the other side and a key. Every host has a private RSA key and the public 154the other side and a key. Every node has a private RSA key and the public
76RSA keys of all other hosts. 155RSA keys of all other nodes.
77 156
78A host establishes a simplex connection by sending the other host a 157A host establishes a simplex connection by sending the other node an
79RSA encrypted challenge containing a random challenge (consisting of 158RSA encrypted challenge containing a random challenge (consisting of
80the encryption key to use when sending packets, more random data and 159the encryption key to use when sending packets, more random data and
81PKCS1_OAEP padding) and a random 16 byte "challenge-id" (used to detect 160PKCS1_OAEP padding) and a random 16 byte "challenge-id" (used to detect
82duplicate auth packets). The destination host will respond by replying 161duplicate auth packets). The destination node will respond by replying
83with an (unencrypted) RIPEMD160 hash of the decrypted challenge, which 162with an (unencrypted) RIPEMD160 hash of the decrypted challenge, which
84will authentify that host. The destination host will also set the outgoing 163will authenticate that node. The destination node will also set the
85encryption parameters as given in the packet. 164outgoing encryption parameters as given in the packet.
86 165
87When the source host receives a correct auth reply (by verifying the 166When the source node receives a correct auth reply (by verifying the
88hash and the id, which will expire after 120 seconds), it will start to 167hash and the id, which will expire after 120 seconds), it will start to
89accept data packets from the destination host. 168accept data packets from the destination node.
90 169
91This means that a host can only initate a simplex connection, telling the 170This means that a node can only initiate a simplex connection, telling the
92other side the key it has to use when it sends packets. The challenge 171other side the key it has to use when it sends packets. The challenge
93reply is only used to set the current IP address of the other side and 172reply is only used to set the current IP address of the other side and
94protocol parameters. 173protocol parameters.
95 174
96This protocol is completely symmetric, so to be able to send packets the 175This protocol is completely symmetric, so to be able to send packets the
97destination host must send a challenge in the exact same way as already 176destination node must send a challenge in the exact same way as already
98described (so, in essence, two simplex connections are created per host 177described (so, in essence, two simplex connections are created per node
99pair). 178pair).
100 179
101=head2 Retrying 180=head2 Retrying
102 181
103When there is no response to an auth request, the host will send auth 182When there is no response to an auth request, the node will send auth
104requests in bursts with an exponential backoff. After some time it will 183requests in bursts with an exponential back-off. After some time it will
105resort to PING packets, which are very small (8 bytes) and lightweight 184resort to PING packets, which are very small (8 bytes + protocol header)
106(no RSA operations required). A host that receives ping requests from an 185and lightweight (no RSA operations required). A node that receives ping
107unconnected peer will respond by trying to create a connection. 186requests from an unconnected peer will respond by trying to create a
187connection.
108 188
109In addition to the exponential backoff, there is a global rate-limit on 189In addition to the exponential back-off, there is a global rate-limit on
110a per-IP base. It allows long bursts but will limit total packet rate to 190a per-IP base. It allows long bursts but will limit total packet rate to
111something like one control packet every ten seconds, to avoid accidental 191something like one control packet every ten seconds, to avoid accidental
112floods due to protocol problems (like a RSA key file mismatch between two 192floods due to protocol problems (like a RSA key file mismatch between two
113hosts). 193nodes).
194
195The intervals between retries are limited by the C<max-retry>
196configuration value. A node with C<connect> = C<always> will always retry,
197a node with C<connect> = C<ondemand> will only try (and re-try) to connect
198as long as there are packets in the queue, usually this limits the retry
199period to C<max-ttl> seconds.
200
201Sending packets over the VPN will reset the retry intervals as well, which
202means as long as somebody is trying to send packets to a given node, GVPE
203will try to connect every few seconds.
114 204
115=head2 Routing and Protocol translation 205=head2 Routing and Protocol translation
116 206
117The gvpe routing algorithm is easy: there isn't any routing. GVPE always 207The GVPE routing algorithm is easy: there isn't much routing to speak
118tries to establish direct connections, if the protocol abilities of the 208of: When routing packets to another node, GVPE trues the following
119two hosts allow it. 209options, in order:
120 210
211=over 4
212
121If the two hosts should be able to reach each other (common protocol, ip 213=item If the two nodes should be able to reach each other directly (common
122and port all known), but cannot (network down), then there will be no 214protocol, port known), then GVPE will send the packet directly to the
123connection, point. 215other node.
216
217=item If this isn't possible (e.g. because the node doesn't have a
218C<hostname> or known port), but the nodes speak a common protocol and a
219router is available, then GVPE will ask a router to "mediate" between both
220nodes (see below).
221
222=item If a direct connection isn't possible (no common protocols) or
223forbidden (C<deny-direct>) and there are any routers, then GVPE will try
224to send packets to the router with the highest priority that is connected
225already I<and> is able (as specified by the config file) to connect
226directly to the target node.
227
228=item If no such router exists, then GVPE will simply send the packet to
229the node with the highest priority available.
230
231=item Failing all that, the packet will be dropped.
232
233=back
124 234
125A host can usually declare itself unreachable directly by setting it's 235A host can usually declare itself unreachable directly by setting it's
126port number(s) to zero. It can declare other hosts as unreachable by using 236port number(s) to zero. It can declare other hosts as unreachable by using
127a config-file that disables all protocols for these other hosts. 237a config-file that disables all protocols for these other hosts. Another
238option is to disable all protocols on that host in the other config files.
128 239
129If two hosts cannot connect to each other because their IP address(es) 240If two hosts cannot connect to each other because their IP address(es)
130are not known (such as dialup hosts), one side will send a connection 241are not known (such as dial-up hosts), one side will send a I<mediated>
131request to a router (routers must be configured to act as routers!), which 242connection request to a router (routers must be configured to act as
132will send both the originating and the destination host a connection info 243routers!), which will send both the originating and the destination host
133request with protocol information and IP address of the other host (if 244a connection info request with protocol information and IP address of the
134known). Both hosts will then try to establish a connection to the other 245other host (if known). Both hosts will then try to establish a direct
135peer, which is usually possible even when both hosts are behind a NAT 246connection to the other peer, which is usually possible even when both
136gateway. 247hosts are behind a NAT gateway.
137 248
138If the hosts cannot reach each other because they have no common protocol, 249Routing via other nodes works because the SRCDST field is not encrypted,
139the originator instead use the router with highest priority and matching
140protocol as peer. Since the SRCDST field is not encrypted, the router host
141can just forward the packet to the destination host. Since each host uses 250so the router can just forward the packet to the destination host. Since
142it's own private key, the router will not be able to decrypt or encrypt 251each host uses it's own private key, the router will not be able to
143packets, it will just act as a simple router and protocol translator. 252decrypt or encrypt packets, it will just act as a simple router and
253protocol translator.
144 254
145When no router is connected, the host will aggressively try to connect to
146all routers, and if a router is asked for an unconnected host it will try
147to ask another router to establish the connection.
148 255
149... more not yet written about the details of the routing, please bug me
150...
151

Diff Legend

– Removed lines
+ Added lines
< Changed lines
> Changed lines