blob: 855d8da57a232bd12f765205c4ef047b51264006 [file] [log] [blame]
Gerrit Renkerba4e58e2006-11-27 11:10:57 -08001 ===========================================================================
2 The UDP-Lite protocol (RFC 3828)
3 ===========================================================================
4
5
6 UDP-Lite is a Standards-Track IETF transport protocol whose characteristic
7 is a variable-length checksum. This has advantages for transport of multimedia
8 (video, VoIP) over wireless networks, as partly damaged packets can still be
9 fed into the codec instead of being discarded due to a failed checksum test.
10
11 This file briefly describes the existing kernel support and the socket API.
12 For in-depth information, you can consult:
13
14 o The UDP-Lite Homepage: http://www.erg.abdn.ac.uk/users/gerrit/udp-lite/
Matt LaPlante01dd2fb2007-10-20 01:34:40 +020015 From here you can also download some example application source code.
Gerrit Renkerba4e58e2006-11-27 11:10:57 -080016
17 o The UDP-Lite HOWTO on
18 http://www.erg.abdn.ac.uk/users/gerrit/udp-lite/files/UDP-Lite-HOWTO.txt
19
20 o The Wireshark UDP-Lite WiKi (with capture files):
21 http://wiki.wireshark.org/Lightweight_User_Datagram_Protocol
22
23 o The Protocol Spec, RFC 3828, http://www.ietf.org/rfc/rfc3828.txt
24
25
26 I) APPLICATIONS
27
28 Several applications have been ported successfully to UDP-Lite. Ethereal
29 (now called wireshark) has UDP-Litev4/v6 support by default. The tarball on
30
31 http://www.erg.abdn.ac.uk/users/gerrit/udp-lite/files/udplite_linux.tar.gz
32
33 has source code for several v4/v6 client-server and network testing examples.
34
35 Porting applications to UDP-Lite is straightforward: only socket level and
36 IPPROTO need to be changed; senders additionally set the checksum coverage
37 length (default = header length = 8). Details are in the next section.
38
39
40 II) PROGRAMMING API
41
42 UDP-Lite provides a connectionless, unreliable datagram service and hence
43 uses the same socket type as UDP. In fact, porting from UDP to UDP-Lite is
44 very easy: simply add `IPPROTO_UDPLITE' as the last argument of the socket(2)
45 call so that the statement looks like:
46
47 s = socket(PF_INET, SOCK_DGRAM, IPPROTO_UDPLITE);
48
49 or, respectively,
50
51 s = socket(PF_INET6, SOCK_DGRAM, IPPROTO_UDPLITE);
52
53 With just the above change you are able to run UDP-Lite services or connect
54 to UDP-Lite servers. The kernel will assume that you are not interested in
55 using partial checksum coverage and so emulate UDP mode (full coverage).
56
57 To make use of the partial checksum coverage facilities requires setting a
58 single socket option, which takes an integer specifying the coverage length:
59
60 * Sender checksum coverage: UDPLITE_SEND_CSCOV
61
62 For example,
63
64 int val = 20;
65 setsockopt(s, SOL_UDPLITE, UDPLITE_SEND_CSCOV, &val, sizeof(int));
66
67 sets the checksum coverage length to 20 bytes (12b data + 8b header).
68 Of each packet only the first 20 bytes (plus the pseudo-header) will be
69 checksummed. This is useful for RTP applications which have a 12-byte
70 base header.
71
72
73 * Receiver checksum coverage: UDPLITE_RECV_CSCOV
74
75 This option is the receiver-side analogue. It is truly optional, i.e. not
76 required to enable traffic with partial checksum coverage. Its function is
77 that of a traffic filter: when enabled, it instructs the kernel to drop
78 all packets which have a coverage _less_ than this value. For example, if
79 RTP and UDP headers are to be protected, a receiver can enforce that only
80 packets with a minimum coverage of 20 are admitted:
81
82 int min = 20;
83 setsockopt(s, SOL_UDPLITE, UDPLITE_RECV_CSCOV, &min, sizeof(int));
84
85 The calls to getsockopt(2) are analogous. Being an extension and not a stand-
86 alone protocol, all socket options known from UDP can be used in exactly the
87 same manner as before, e.g. UDP_CORK or UDP_ENCAP.
88
89 A detailed discussion of UDP-Lite checksum coverage options is in section IV.
90
91
92 III) HEADER FILES
93
94 The socket API requires support through header files in /usr/include:
95
96 * /usr/include/netinet/in.h
97 to define IPPROTO_UDPLITE
98
99 * /usr/include/netinet/udplite.h
100 for UDP-Lite header fields and protocol constants
101
102 For testing purposes, the following can serve as a `mini' header file:
103
104 #define IPPROTO_UDPLITE 136
105 #define SOL_UDPLITE 136
106 #define UDPLITE_SEND_CSCOV 10
107 #define UDPLITE_RECV_CSCOV 11
108
109 Ready-made header files for various distros are in the UDP-Lite tarball.
110
111
112 IV) KERNEL BEHAVIOUR WITH REGARD TO THE VARIOUS SOCKET OPTIONS
113
114 To enable debugging messages, the log level need to be set to 8, as most
115 messages use the KERN_DEBUG level (7).
116
117 1) Sender Socket Options
118
119 If the sender specifies a value of 0 as coverage length, the module
120 assumes full coverage, transmits a packet with coverage length of 0
121 and according checksum. If the sender specifies a coverage < 8 and
122 different from 0, the kernel assumes 8 as default value. Finally,
123 if the specified coverage length exceeds the packet length, the packet
124 length is used instead as coverage length.
125
126 2) Receiver Socket Options
127
128 The receiver specifies the minimum value of the coverage length it
129 is willing to accept. A value of 0 here indicates that the receiver
130 always wants the whole of the packet covered. In this case, all
131 partially covered packets are dropped and an error is logged.
132
133 It is not possible to specify illegal values (<0 and <8); in these
134 cases the default of 8 is assumed.
135
136 All packets arriving with a coverage value less than the specified
137 threshold are discarded, these events are also logged.
138
139 3) Disabling the Checksum Computation
140
141 On both sender and receiver, checksumming will always be performed
Matt LaPlantea982ac02007-05-09 07:35:06 +0200142 and cannot be disabled using SO_NO_CHECK. Thus
Gerrit Renkerba4e58e2006-11-27 11:10:57 -0800143
144 setsockopt(sockfd, SOL_SOCKET, SO_NO_CHECK, ... );
145
146 will always will be ignored, while the value of
147
148 getsockopt(sockfd, SOL_SOCKET, SO_NO_CHECK, &value, ...);
149
150 is meaningless (as in TCP). Packets with a zero checksum field are
Gerrit Renker47112e22008-07-21 13:35:08 -0700151 illegal (cf. RFC 3828, sec. 3.1) and will be silently discarded.
Gerrit Renkerba4e58e2006-11-27 11:10:57 -0800152
153 4) Fragmentation
154
155 The checksum computation respects both buffersize and MTU. The size
156 of UDP-Lite packets is determined by the size of the send buffer. The
157 minimum size of the send buffer is 2048 (defined as SOCK_MIN_SNDBUF
158 in include/net/sock.h), the default value is configurable as
159 net.core.wmem_default or via setting the SO_SNDBUF socket(7)
160 option. The maximum upper bound for the send buffer is determined
161 by net.core.wmem_max.
162
163 Given a payload size larger than the send buffer size, UDP-Lite will
164 split the payload into several individual packets, filling up the
165 send buffer size in each case.
166
167 The precise value also depends on the interface MTU. The interface MTU,
168 in turn, may trigger IP fragmentation. In this case, the generated
169 UDP-Lite packet is split into several IP packets, of which only the
170 first one contains the L4 header.
171
172 The send buffer size has implications on the checksum coverage length.
173 Consider the following example:
174
175 Payload: 1536 bytes Send Buffer: 1024 bytes
176 MTU: 1500 bytes Coverage Length: 856 bytes
177
178 UDP-Lite will ship the 1536 bytes in two separate packets:
179
180 Packet 1: 1024 payload + 8 byte header + 20 byte IP header = 1052 bytes
181 Packet 2: 512 payload + 8 byte header + 20 byte IP header = 540 bytes
182
183 The coverage packet covers the UDP-Lite header and 848 bytes of the
184 payload in the first packet, the second packet is fully covered. Note
185 that for the second packet, the coverage length exceeds the packet
186 length. The kernel always re-adjusts the coverage length to the packet
187 length in such cases.
188
189 As an example of what happens when one UDP-Lite packet is split into
190 several tiny fragments, consider the following example.
191
192 Payload: 1024 bytes Send buffer size: 1024 bytes
193 MTU: 300 bytes Coverage length: 575 bytes
194
195 +-+-----------+--------------+--------------+--------------+
196 |8| 272 | 280 | 280 | 280 |
197 +-+-----------+--------------+--------------+--------------+
198 280 560 840 1032
199 ^
200 *****checksum coverage*************
201
202 The UDP-Lite module generates one 1032 byte packet (1024 + 8 byte
203 header). According to the interface MTU, these are split into 4 IP
204 packets (280 byte IP payload + 20 byte IP header). The kernel module
205 sums the contents of the entire first two packets, plus 15 bytes of
206 the last packet before releasing the fragments to the IP module.
207
208 To see the analogous case for IPv6 fragmentation, consider a link
209 MTU of 1280 bytes and a write buffer of 3356 bytes. If the checksum
210 coverage is less than 1232 bytes (MTU minus IPv6/fragment header
211 lengths), only the first fragment needs to be considered. When using
212 larger checksum coverage lengths, each eligible fragment needs to be
213 checksummed. Suppose we have a checksum coverage of 3062. The buffer
214 of 3356 bytes will be split into the following fragments:
215
216 Fragment 1: 1280 bytes carrying 1232 bytes of UDP-Lite data
217 Fragment 2: 1280 bytes carrying 1232 bytes of UDP-Lite data
218 Fragment 3: 948 bytes carrying 900 bytes of UDP-Lite data
219
220 The first two fragments have to be checksummed in full, of the last
221 fragment only 598 (= 3062 - 2*1232) bytes are checksummed.
222
223 While it is important that such cases are dealt with correctly, they
224 are (annoyingly) rare: UDP-Lite is designed for optimising multimedia
225 performance over wireless (or generally noisy) links and thus smaller
Matt LaPlante01dd2fb2007-10-20 01:34:40 +0200226 coverage lengths are likely to be expected.
Gerrit Renkerba4e58e2006-11-27 11:10:57 -0800227
228
229 V) UDP-LITE RUNTIME STATISTICS AND THEIR MEANING
230
231 Exceptional and error conditions are logged to syslog at the KERN_DEBUG
232 level. Live statistics about UDP-Lite are available in /proc/net/snmp
233 and can (with newer versions of netstat) be viewed using
234
235 netstat -svu
236
237 This displays UDP-Lite statistics variables, whose meaning is as follows.
238
Wang Chencb759942007-12-03 22:33:28 +1100239 InDatagrams: The total number of datagrams delivered to users.
Gerrit Renkerba4e58e2006-11-27 11:10:57 -0800240
241 NoPorts: Number of packets received to an unknown port.
242 These cases are counted separately (not as InErrors).
243
244 InErrors: Number of erroneous UDP-Lite packets. Errors include:
245 * internal socket queue receive errors
246 * packet too short (less than 8 bytes or stated
247 coverage length exceeds received length)
248 * xfrm4_policy_check() returned with error
249 * application has specified larger min. coverage
250 length than that of incoming packet
251 * checksum coverage violated
252 * bad checksum
253
254 OutDatagrams: Total number of sent datagrams.
255
256 These statistics derive from the UDP MIB (RFC 2013).
257
258
259 VI) IPTABLES
260
261 There is packet match support for UDP-Lite as well as support for the LOG target.
Matt LaPlante01dd2fb2007-10-20 01:34:40 +0200262 If you copy and paste the following line into /etc/protocols,
Gerrit Renkerba4e58e2006-11-27 11:10:57 -0800263
264 udplite 136 UDP-Lite # UDP-Lite [RFC 3828]
265
266 then
267 iptables -A INPUT -p udplite -j LOG
268
269 will produce logging output to syslog. Dropping and rejecting packets also works.
270
271
272 VII) MAINTAINER ADDRESS
273
274 The UDP-Lite patch was developed at
275 University of Aberdeen
276 Electronics Research Group
277 Department of Engineering
278 Fraser Noble Building
279 Aberdeen AB24 3UE; UK
280 The current maintainer is Gerrit Renker, <gerrit@erg.abdn.ac.uk>. Initial
281 code was developed by William Stanislaus, <william@erg.abdn.ac.uk>.