[deliverable/linux.git] / Documentation / networking / dccp.txt

DCCP protocol
============


Contents
========

- Introduction
- Missing features
- Socket options
- Notes

Introduction
============

Datagram Congestion Control Protocol (DCCP) is an unreliable, connection
oriented protocol designed to solve issues present in UDP and TCP, particularly
for real-time and multimedia (streaming) traffic.
It divides into a base protocol (RFC 4340) and plugable congestion control
modules called CCIDs. Like plugable TCP congestion control, at least one CCID
needs to be enabled in order for the protocol to function properly. In the Linux
implementation, this is the TCP-like CCID2 (RFC 4341). Additional CCIDs, such as
the TCP-friendly CCID3 (RFC 4342), are optional.
For a brief introduction to CCIDs and suggestions for choosing a CCID to match
given applications, see section 10 of RFC 4340.

It has a base protocol and pluggable congestion control IDs (CCIDs).

DCCP is a Proposed Standard (RFC 2026), and the homepage for DCCP as a protocol
is at http://www.ietf.org/html.charters/dccp-charter.html

Missing features
================

The Linux DCCP implementation does not currently support all the features that are
specified in RFCs 4340...42.

The known bugs are at:
	http://linux-net.osdl.org/index.php/TODO#DCCP

For more up-to-date versions of the DCCP implementation, please consider using
the experimental DCCP test tree; instructions for checking this out are on:
http://linux-net.osdl.org/index.php/DCCP_Testing#Experimental_DCCP_source_tree


Socket options
==============

DCCP_SOCKOPT_SERVICE sets the service. The specification mandates use of
service codes (RFC 4340, sec. 8.1.2); if this socket option is not set,
the socket will fall back to 0 (which means that no meaningful service code
is present). On active sockets this is set before connect(); specifying more
than one code has no effect (all subsequent service codes are ignored). The
case is different for passive sockets, where multiple service codes (up to 32)
can be set before calling bind().

DCCP_SOCKOPT_GET_CUR_MPS is read-only and retrieves the current maximum packet
size (application payload size) in bytes, see RFC 4340, section 14.

DCCP_SOCKOPT_SERVER_TIMEWAIT enables the server (listening socket) to hold
timewait state when closing the connection (RFC 4340, 8.3). The usual case is
that the closing server sends a CloseReq, whereupon the client holds timewait
state. When this boolean socket option is on, the server sends a Close instead
and will enter TIMEWAIT. This option must be set after accept() returns.

DCCP_SOCKOPT_SEND_CSCOV and DCCP_SOCKOPT_RECV_CSCOV are used for setting the
partial checksum coverage (RFC 4340, sec. 9.2). The default is that checksums
always cover the entire packet and that only fully covered application data is
accepted by the receiver. Hence, when using this feature on the sender, it must
be enabled at the receiver, too with suitable choice of CsCov.

DCCP_SOCKOPT_SEND_CSCOV sets the sender checksum coverage. Values in the
	range 0..15 are acceptable. The default setting is 0 (full coverage),
	values between 1..15 indicate partial coverage.
DCCP_SOCKOPT_RECV_CSCOV is for the receiver and has a different meaning: it
	sets a threshold, where again values 0..15 are acceptable. The default
	of 0 means that all packets with a partial coverage will be discarded.
	Values in the range 1..15 indicate that packets with minimally such a
	coverage value are also acceptable. The higher the number, the more
	restrictive this setting (see [RFC 4340, sec. 9.2.1]). Partial coverage
	settings are inherited to the child socket after accept().

The following two options apply to CCID 3 exclusively and are getsockopt()-only.
In either case, a TFRC info struct (defined in <linux/tfrc.h>) is returned.
DCCP_SOCKOPT_CCID_RX_INFO
	Returns a `struct tfrc_rx_info' in optval; the buffer for optval and
	optlen must be set to at least sizeof(struct tfrc_rx_info).
DCCP_SOCKOPT_CCID_TX_INFO
	Returns a `struct tfrc_tx_info' in optval; the buffer for optval and
	optlen must be set to at least sizeof(struct tfrc_tx_info).

On unidirectional connections it is useful to close the unused half-connection
via shutdown (SHUT_WR or SHUT_RD): this will reduce per-packet processing costs.

Sysctl variables
================
Several DCCP default parameters can be managed by the following sysctls
(sysctl net.dccp.default or /proc/sys/net/dccp/default):

request_retries
	The number of active connection initiation retries (the number of
	Requests minus one) before timing out. In addition, it also governs
	the behaviour of the other, passive side: this variable also sets
	the number of times DCCP repeats sending a Response when the initial
	handshake does not progress from RESPOND to OPEN (i.e. when no Ack
	is received after the initial Request).  This value should be greater
	than 0, suggested is less than 10. Analogue of tcp_syn_retries.

retries1
	How often a DCCP Response is retransmitted until the listening DCCP
	side considers its connecting peer dead. Analogue of tcp_retries1.

retries2
	The number of times a general DCCP packet is retransmitted. This has
	importance for retransmitted acknowledgments and feature negotiation,
	data packets are never retransmitted. Analogue of tcp_retries2.

send_ndp = 1
	Whether or not to send NDP count options (sec. 7.7.2).

send_ackvec = 1
	Whether or not to send Ack Vector options (sec. 11.5).

ack_ratio = 2
	The default Ack Ratio (sec. 11.3) to use.

tx_ccid = 2
	Default CCID for the sender-receiver half-connection.

rx_ccid = 2
	Default CCID for the receiver-sender half-connection.

seq_window = 100
	The initial sequence window (sec. 7.5.2).

tx_qlen = 5
	The size of the transmit buffer in packets. A value of 0 corresponds
	to an unbounded transmit buffer.

sync_ratelimit = 125 ms
	The timeout between subsequent DCCP-Sync packets sent in response to
	sequence-invalid packets on the same socket (RFC 4340, 7.5.4). The unit
	of this parameter is milliseconds; a value of 0 disables rate-limiting.

IOCTLS
======
FIONREAD
	Works as in udp(7): returns in the `int' argument pointer the size of
	the next pending datagram in bytes, or 0 when no datagram is pending.

Notes
=====

DCCP does not travel through NAT successfully at present on many boxes. This is
because the checksum covers the pseudo-header as per TCP and UDP. Linux NAT
support for DCCP has been added.
Commit	Line	Data
98069ff4 IM	1	DCCP protocol
	2	============
	3
98069ff4 IM	4
	5	Contents
	6	========
	7
	8	- Introduction
	9	- Missing features
	10	- Socket options
	11	- Notes
	12
	13	Introduction
	14	============
	15
	16	Datagram Congestion Control Protocol (DCCP) is an unreliable, connection
e333b3ed GR	17	oriented protocol designed to solve issues present in UDP and TCP, particularly
	18	for real-time and multimedia (streaming) traffic.
	19	It divides into a base protocol (RFC 4340) and plugable congestion control
	20	modules called CCIDs. Like plugable TCP congestion control, at least one CCID
	21	needs to be enabled in order for the protocol to function properly. In the Linux
	22	implementation, this is the TCP-like CCID2 (RFC 4341). Additional CCIDs, such as
	23	the TCP-friendly CCID3 (RFC 4342), are optional.
	24	For a brief introduction to CCIDs and suggestions for choosing a CCID to match
	25	given applications, see section 10 of RFC 4340.
98069ff4 IM	26
	27	It has a base protocol and pluggable congestion control IDs (CCIDs).
	28
ebe6f7e7 GR	29	DCCP is a Proposed Standard (RFC 2026), and the homepage for DCCP as a protocol
ebe6f7e7 GR	30	is at http://www.ietf.org/html.charters/dccp-charter.html
98069ff4 IM	31
	32	Missing features
	33	================
	34
ebe6f7e7 GR	35	The Linux DCCP implementation does not currently support all the features that are
ebe6f7e7 GR	36	specified in RFCs 4340...42.
98069ff4	37
ddfe10b8 IM	38	The known bugs are at:
ddfe10b8 IM	39	http://linux-net.osdl.org/index.php/TODO#DCCP
98069ff4	40
ebe6f7e7 GR	41	For more up-to-date versions of the DCCP implementation, please consider using
	42	the experimental DCCP test tree; instructions for checking this out are on:
	43	http://linux-net.osdl.org/index.php/DCCP_Testing#Experimental_DCCP_source_tree
	44
	45
98069ff4 IM	46	Socket options
	47	==============
	48
00e4d116 GR	49	DCCP_SOCKOPT_SERVICE sets the service. The specification mandates use of
	50	service codes (RFC 4340, sec. 8.1.2); if this socket option is not set,
	51	the socket will fall back to 0 (which means that no meaningful service code
126acd5b GR	52	is present). On active sockets this is set before connect(); specifying more
	53	than one code has no effect (all subsequent service codes are ignored). The
	54	case is different for passive sockets, where multiple service codes (up to 32)
	55	can be set before calling bind().
98069ff4	56
7c559a9e GR	57	DCCP_SOCKOPT_GET_CUR_MPS is read-only and retrieves the current maximum packet
	58	size (application payload size) in bytes, see RFC 4340, section 14.
	59
b8599d20 GR	60	DCCP_SOCKOPT_SERVER_TIMEWAIT enables the server (listening socket) to hold
	61	timewait state when closing the connection (RFC 4340, 8.3). The usual case is
	62	that the closing server sends a CloseReq, whereupon the client holds timewait
	63	state. When this boolean socket option is on, the server sends a Close instead
	64	and will enter TIMEWAIT. This option must be set after accept() returns.
	65
6f4e5fff GR	66	DCCP_SOCKOPT_SEND_CSCOV and DCCP_SOCKOPT_RECV_CSCOV are used for setting the
	67	partial checksum coverage (RFC 4340, sec. 9.2). The default is that checksums
	68	always cover the entire packet and that only fully covered application data is
	69	accepted by the receiver. Hence, when using this feature on the sender, it must
	70	be enabled at the receiver, too with suitable choice of CsCov.
	71
	72	DCCP_SOCKOPT_SEND_CSCOV sets the sender checksum coverage. Values in the
	73	range 0..15 are acceptable. The default setting is 0 (full coverage),
	74	values between 1..15 indicate partial coverage.
2bfd754d	75	DCCP_SOCKOPT_RECV_CSCOV is for the receiver and has a different meaning: it
6f4e5fff GR	76	sets a threshold, where again values 0..15 are acceptable. The default
	77	of 0 means that all packets with a partial coverage will be discarded.
	78	Values in the range 1..15 indicate that packets with minimally such a
	79	coverage value are also acceptable. The higher the number, the more
2bfd754d GR	80	restrictive this setting (see [RFC 4340, sec. 9.2.1]). Partial coverage
2bfd754d GR	81	settings are inherited to the child socket after accept().
6f4e5fff	82
f2645101 GR	83	The following two options apply to CCID 3 exclusively and are getsockopt()-only.
	84	In either case, a TFRC info struct (defined in <linux/tfrc.h>) is returned.
	85	DCCP_SOCKOPT_CCID_RX_INFO
	86	Returns a `struct tfrc_rx_info' in optval; the buffer for optval and
	87	optlen must be set to at least sizeof(struct tfrc_rx_info).
	88	DCCP_SOCKOPT_CCID_TX_INFO
	89	Returns a `struct tfrc_tx_info' in optval; the buffer for optval and
	90	optlen must be set to at least sizeof(struct tfrc_tx_info).
	91
8e8c71f1 GR	92	On unidirectional connections it is useful to close the unused half-connection
8e8c71f1 GR	93	via shutdown (SHUT_WR or SHUT_RD): this will reduce per-packet processing costs.
f2645101	94
2e2e9e92 GR	95	Sysctl variables
	96	================
	97	Several DCCP default parameters can be managed by the following sysctls
	98	(sysctl net.dccp.default or /proc/sys/net/dccp/default):
	99
	100	request_retries
	101	The number of active connection initiation retries (the number of
	102	Requests minus one) before timing out. In addition, it also governs
	103	the behaviour of the other, passive side: this variable also sets
	104	the number of times DCCP repeats sending a Response when the initial
	105	handshake does not progress from RESPOND to OPEN (i.e. when no Ack
	106	is received after the initial Request). This value should be greater
	107	than 0, suggested is less than 10. Analogue of tcp_syn_retries.
	108
	109	retries1
	110	How often a DCCP Response is retransmitted until the listening DCCP
	111	side considers its connecting peer dead. Analogue of tcp_retries1.
	112
	113	retries2
	114	The number of times a general DCCP packet is retransmitted. This has
	115	importance for retransmitted acknowledgments and feature negotiation,
	116	data packets are never retransmitted. Analogue of tcp_retries2.
	117
	118	send_ndp = 1
	119	Whether or not to send NDP count options (sec. 7.7.2).
	120
	121	send_ackvec = 1
	122	Whether or not to send Ack Vector options (sec. 11.5).
	123
	124	ack_ratio = 2
	125	The default Ack Ratio (sec. 11.3) to use.
	126
	127	tx_ccid = 2
	128	Default CCID for the sender-receiver half-connection.
	129
	130	rx_ccid = 2
	131	Default CCID for the receiver-sender half-connection.
	132
	133	seq_window = 100
	134	The initial sequence window (sec. 7.5.2).
	135
82e3ab9d IM	136	tx_qlen = 5
	137	The size of the transmit buffer in packets. A value of 0 corresponds
	138	to an unbounded transmit buffer.
	139
a94f0f97 GR	140	sync_ratelimit = 125 ms
	141	The timeout between subsequent DCCP-Sync packets sent in response to
	142	sequence-invalid packets on the same socket (RFC 4340, 7.5.4). The unit
	143	of this parameter is milliseconds; a value of 0 disables rate-limiting.
	144
c2814901 GR	145	IOCTLS
	146	======
	147	FIONREAD
	148	Works as in udp(7): returns in the `int' argument pointer the size of
	149	the next pending datagram in bytes, or 0 when no datagram is pending.
	150
98069ff4 IM	151	Notes
	152	=====
	153
ddfe10b8	154	DCCP does not travel through NAT successfully at present on many boxes. This is
126acd5b	155	because the checksum covers the pseudo-header as per TCP and UDP. Linux NAT
ddfe10b8	156	support for DCCP has been added.