tjh.dev
Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
1DCCP protocol
2============
3
4
5Contents
6========
7
8- Introduction
9- Missing features
10- Socket options
11- Notes
12
13Introduction
14============
15
16Datagram Congestion Control Protocol (DCCP) is an unreliable, connection
17oriented protocol designed to solve issues present in UDP and TCP, particularly
18for real-time and multimedia (streaming) traffic.
19It divides into a base protocol (RFC 4340) and plugable congestion control
20modules called CCIDs. Like plugable TCP congestion control, at least one CCID
21needs to be enabled in order for the protocol to function properly. In the Linux
22implementation, this is the TCP-like CCID2 (RFC 4341). Additional CCIDs, such as
23the TCP-friendly CCID3 (RFC 4342), are optional.
24For a brief introduction to CCIDs and suggestions for choosing a CCID to match
25given applications, see section 10 of RFC 4340.
26
27It has a base protocol and pluggable congestion control IDs (CCIDs).
28
29DCCP is a Proposed Standard (RFC 2026), and the homepage for DCCP as a protocol
30is at http://www.ietf.org/html.charters/dccp-charter.html
31
32Missing features
33================
34
35The Linux DCCP implementation does not currently support all the features that are
36specified in RFCs 4340...42.
37
38The known bugs are at:
39 http://linux-net.osdl.org/index.php/TODO#DCCP
40
41For more up-to-date versions of the DCCP implementation, please consider using
42the experimental DCCP test tree; instructions for checking this out are on:
43http://linux-net.osdl.org/index.php/DCCP_Testing#Experimental_DCCP_source_tree
44
45
46Socket options
47==============
48
49DCCP_SOCKOPT_SERVICE sets the service. The specification mandates use of
50service codes (RFC 4340, sec. 8.1.2); if this socket option is not set,
51the socket will fall back to 0 (which means that no meaningful service code
52is present). On active sockets this is set before connect(); specifying more
53than one code has no effect (all subsequent service codes are ignored). The
54case is different for passive sockets, where multiple service codes (up to 32)
55can be set before calling bind().
56
57DCCP_SOCKOPT_GET_CUR_MPS is read-only and retrieves the current maximum packet
58size (application payload size) in bytes, see RFC 4340, section 14.
59
60DCCP_SOCKOPT_AVAILABLE_CCIDS is also read-only and returns the list of CCIDs
61supported by the endpoint (see include/linux/dccp.h for symbolic constants).
62The caller needs to provide a sufficiently large (> 2) array of type uint8_t.
63
64DCCP_SOCKOPT_CCID is write-only and sets both the TX and RX CCIDs at the same
65time, combining the operation of the next two socket options. This option is
66preferrable over the latter two, since often applications will use the same
67type of CCID for both directions; and mixed use of CCIDs is not currently well
68understood. This socket option takes as argument at least one uint8_t value, or
69an array of uint8_t values, which must match available CCIDS (see above). CCIDs
70must be registered on the socket before calling connect() or listen().
71
72DCCP_SOCKOPT_TX_CCID is read/write. It returns the current CCID (if set) or sets
73the preference list for the TX CCID, using the same format as DCCP_SOCKOPT_CCID.
74Please note that the getsockopt argument type here is `int', not uint8_t.
75
76DCCP_SOCKOPT_RX_CCID is analogous to DCCP_SOCKOPT_TX_CCID, but for the RX CCID.
77
78DCCP_SOCKOPT_SERVER_TIMEWAIT enables the server (listening socket) to hold
79timewait state when closing the connection (RFC 4340, 8.3). The usual case is
80that the closing server sends a CloseReq, whereupon the client holds timewait
81state. When this boolean socket option is on, the server sends a Close instead
82and will enter TIMEWAIT. This option must be set after accept() returns.
83
84DCCP_SOCKOPT_SEND_CSCOV and DCCP_SOCKOPT_RECV_CSCOV are used for setting the
85partial checksum coverage (RFC 4340, sec. 9.2). The default is that checksums
86always cover the entire packet and that only fully covered application data is
87accepted by the receiver. Hence, when using this feature on the sender, it must
88be enabled at the receiver, too with suitable choice of CsCov.
89
90DCCP_SOCKOPT_SEND_CSCOV sets the sender checksum coverage. Values in the
91 range 0..15 are acceptable. The default setting is 0 (full coverage),
92 values between 1..15 indicate partial coverage.
93DCCP_SOCKOPT_RECV_CSCOV is for the receiver and has a different meaning: it
94 sets a threshold, where again values 0..15 are acceptable. The default
95 of 0 means that all packets with a partial coverage will be discarded.
96 Values in the range 1..15 indicate that packets with minimally such a
97 coverage value are also acceptable. The higher the number, the more
98 restrictive this setting (see [RFC 4340, sec. 9.2.1]). Partial coverage
99 settings are inherited to the child socket after accept().
100
101The following two options apply to CCID 3 exclusively and are getsockopt()-only.
102In either case, a TFRC info struct (defined in <linux/tfrc.h>) is returned.
103DCCP_SOCKOPT_CCID_RX_INFO
104 Returns a `struct tfrc_rx_info' in optval; the buffer for optval and
105 optlen must be set to at least sizeof(struct tfrc_rx_info).
106DCCP_SOCKOPT_CCID_TX_INFO
107 Returns a `struct tfrc_tx_info' in optval; the buffer for optval and
108 optlen must be set to at least sizeof(struct tfrc_tx_info).
109
110On unidirectional connections it is useful to close the unused half-connection
111via shutdown (SHUT_WR or SHUT_RD): this will reduce per-packet processing costs.
112
113Sysctl variables
114================
115Several DCCP default parameters can be managed by the following sysctls
116(sysctl net.dccp.default or /proc/sys/net/dccp/default):
117
118request_retries
119 The number of active connection initiation retries (the number of
120 Requests minus one) before timing out. In addition, it also governs
121 the behaviour of the other, passive side: this variable also sets
122 the number of times DCCP repeats sending a Response when the initial
123 handshake does not progress from RESPOND to OPEN (i.e. when no Ack
124 is received after the initial Request). This value should be greater
125 than 0, suggested is less than 10. Analogue of tcp_syn_retries.
126
127retries1
128 How often a DCCP Response is retransmitted until the listening DCCP
129 side considers its connecting peer dead. Analogue of tcp_retries1.
130
131retries2
132 The number of times a general DCCP packet is retransmitted. This has
133 importance for retransmitted acknowledgments and feature negotiation,
134 data packets are never retransmitted. Analogue of tcp_retries2.
135
136tx_ccid = 2
137 Default CCID for the sender-receiver half-connection. Depending on the
138 choice of CCID, the Send Ack Vector feature is enabled automatically.
139
140rx_ccid = 2
141 Default CCID for the receiver-sender half-connection; see tx_ccid.
142
143seq_window = 100
144 The initial sequence window (sec. 7.5.2) of the sender. This influences
145 the local ackno validity and the remote seqno validity windows (7.5.1).
146
147tx_qlen = 5
148 The size of the transmit buffer in packets. A value of 0 corresponds
149 to an unbounded transmit buffer.
150
151sync_ratelimit = 125 ms
152 The timeout between subsequent DCCP-Sync packets sent in response to
153 sequence-invalid packets on the same socket (RFC 4340, 7.5.4). The unit
154 of this parameter is milliseconds; a value of 0 disables rate-limiting.
155
156IOCTLS
157======
158FIONREAD
159 Works as in udp(7): returns in the `int' argument pointer the size of
160 the next pending datagram in bytes, or 0 when no datagram is pending.
161
162Notes
163=====
164
165DCCP does not travel through NAT successfully at present on many boxes. This is
166because the checksum covers the pseudo-header as per TCP and UDP. Linux NAT
167support for DCCP has been added.