xref: /freebsd/share/man/man9/mod_cc.9 (revision a812392203d7c4c3f0db9d8a0f3391374c49c71f)
1.\"
2.\" Copyright (c) 2008-2009 Lawrence Stewart <lstewart@FreeBSD.org>
3.\" Copyright (c) 2010-2011 The FreeBSD Foundation
4.\" All rights reserved.
5.\"
6.\" Portions of this documentation were written at the Centre for Advanced
7.\" Internet Architectures, Swinburne University of Technology, Melbourne,
8.\" Australia by David Hayes and Lawrence Stewart under sponsorship from the
9.\" FreeBSD Foundation.
10.\"
11.\" Redistribution and use in source and binary forms, with or without
12.\" modification, are permitted provided that the following conditions
13.\" are met:
14.\" 1. Redistributions of source code must retain the above copyright
15.\"    notice, this list of conditions and the following disclaimer.
16.\" 2. Redistributions in binary form must reproduce the above copyright
17.\"    notice, this list of conditions and the following disclaimer in the
18.\"    documentation and/or other materials provided with the distribution.
19.\"
20.\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
21.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23.\" ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR
24.\" ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30.\" SUCH DAMAGE.
31.\"
32.\" $FreeBSD$
33.\"
34.Dd December 26, 2014
35.Dt MOD_CC 9
36.Os
37.Sh NAME
38.Nm mod_cc ,
39.Nm DECLARE_CC_MODULE ,
40.Nm CCV
41.Nd Modular Congestion Control
42.Sh SYNOPSIS
43.In netinet/cc.h
44.In netinet/cc/cc_module.h
45.Fn DECLARE_CC_MODULE "ccname" "ccalgo"
46.Fn CCV "ccv" "what"
47.Sh DESCRIPTION
48The
49.Nm
50framework allows congestion control algorithms to be implemented as dynamically
51loadable kernel modules via the
52.Xr kld 4
53facility.
54Transport protocols can select from the list of available algorithms on a
55connection-by-connection basis, or use the system default (see
56.Xr mod_cc 4
57for more details).
58.Pp
59.Nm
60modules are identified by an
61.Xr ascii 7
62name and set of hook functions encapsulated in a
63.Vt "struct cc_algo" ,
64which has the following members:
65.Bd -literal -offset indent
66struct cc_algo {
67	char	name[TCP_CA_NAME_MAX];
68	int	(*mod_init) (void);
69	int	(*mod_destroy) (void);
70	int	(*cb_init) (struct cc_var *ccv);
71	void	(*cb_destroy) (struct cc_var *ccv);
72	void	(*conn_init) (struct cc_var *ccv);
73	void	(*ack_received) (struct cc_var *ccv, uint16_t type);
74	void	(*cong_signal) (struct cc_var *ccv, uint32_t type);
75	void	(*post_recovery) (struct cc_var *ccv);
76	void	(*after_idle) (struct cc_var *ccv);
77};
78.Ed
79.Pp
80The
81.Va name
82field identifies the unique name of the algorithm, and should be no longer than
83TCP_CA_NAME_MAX-1 characters in length (the TCP_CA_NAME_MAX define lives in
84.In netinet/tcp.h
85for compatibility reasons).
86.Pp
87The
88.Va mod_init
89function is called when a new module is loaded into the system but before the
90registration process is complete.
91It should be implemented if a module needs to set up some global state prior to
92being available for use by new connections.
93Returning a non-zero value from
94.Va mod_init
95will cause the loading of the module to fail.
96.Pp
97The
98.Va mod_destroy
99function is called prior to unloading an existing module from the kernel.
100It should be implemented if a module needs to clean up any global state before
101being removed from the kernel.
102The return value is currently ignored.
103.Pp
104The
105.Va cb_init
106function is called when a TCP control block
107.Vt struct tcpcb
108is created.
109It should be implemented if a module needs to allocate memory for storing
110private per-connection state.
111Returning a non-zero value from
112.Va cb_init
113will cause the connection set up to be aborted, terminating the connection as a
114result.
115.Pp
116The
117.Va cb_destroy
118function is called when a TCP control block
119.Vt struct tcpcb
120is destroyed.
121It should be implemented if a module needs to free memory allocated in
122.Va cb_init .
123.Pp
124The
125.Va conn_init
126function is called when a new connection has been established and variables are
127being initialised.
128It should be implemented to initialise congestion control algorithm variables
129for the newly established connection.
130.Pp
131The
132.Va ack_received
133function is called when a TCP acknowledgement (ACK) packet is received.
134Modules use the
135.Fa type
136argument as an input to their congestion management algorithms.
137The ACK types currently reported by the stack are CC_ACK and CC_DUPACK.
138CC_ACK indicates the received ACK acknowledges previously unacknowledged data.
139CC_DUPACK indicates the received ACK acknowledges data we have already received
140an ACK for.
141.Pp
142The
143.Va cong_signal
144function is called when a congestion event is detected by the TCP stack.
145Modules use the
146.Fa type
147argument as an input to their congestion management algorithms.
148The congestion event types currently reported by the stack are CC_ECN, CC_RTO,
149CC_RTO_ERR and CC_NDUPACK.
150CC_ECN is reported when the TCP stack receives an explicit congestion notification
151(RFC3168).
152CC_RTO is reported when the retransmission time out timer fires.
153CC_RTO_ERR is reported if the retransmission time out timer fired in error.
154CC_NDUPACK is reported if N duplicate ACKs have been received back-to-back,
155where N is the fast retransmit duplicate ack threshold (N=3 currently as per
156RFC5681).
157.Pp
158The
159.Va post_recovery
160function is called after the TCP connection has recovered from a congestion event.
161It should be implemented to adjust state as required.
162.Pp
163The
164.Va after_idle
165function is called when data transfer resumes after an idle period.
166It should be implemented to adjust state as required.
167.Pp
168The
169.Fn DECLARE_CC_MODULE
170macro provides a convenient wrapper around the
171.Xr DECLARE_MODULE 9
172macro, and is used to register a
173.Nm
174module with the
175.Nm
176framework.
177The
178.Fa ccname
179argument specifies the module's name.
180The
181.Fa ccalgo
182argument points to the module's
183.Vt struct cc_algo .
184.Pp
185.Nm
186modules must instantiate a
187.Vt struct cc_algo ,
188but are only required to set the name field, and optionally any of the function
189pointers.
190The stack will skip calling any function pointer which is NULL, so there is no
191requirement to implement any of the function pointers.
192Using the C99 designated initialiser feature to set fields is encouraged.
193.Pp
194Each function pointer which deals with congestion control state is passed a
195pointer to a
196.Vt struct cc_var ,
197which has the following members:
198.Bd -literal -offset indent
199struct cc_var {
200	void		*cc_data;
201	int		bytes_this_ack;
202	tcp_seq		curack;
203	uint32_t	flags;
204	int		type;
205	union ccv_container {
206		struct tcpcb		*tcp;
207		struct sctp_nets	*sctp;
208	} ccvc;
209};
210.Ed
211.Pp
212.Vt struct cc_var
213groups congestion control related variables into a single, embeddable structure
214and adds a layer of indirection to accessing transport protocol control blocks.
215The eventual goal is to allow a single set of
216.Nm
217modules to be shared between all congestion aware transport protocols, though
218currently only
219.Xr tcp 4
220is supported.
221.Pp
222To aid the eventual transition towards this goal, direct use of variables from
223the transport protocol's data structures is strongly discouraged.
224However, it is inevitable at the current time to require access to some of these
225variables, and so the
226.Fn CCV
227macro exists as a convenience accessor.
228The
229.Fa ccv
230argument points to the
231.Vt struct cc_var
232passed into the function by the
233.Nm
234framework.
235The
236.Fa what
237argument specifies the name of the variable to access.
238.Pp
239Apart from the
240.Va type
241and
242.Va ccv_container
243fields, the remaining fields in
244.Vt struct cc_var
245are for use by
246.Nm
247modules.
248.Pp
249The
250.Va cc_data
251field is available for algorithms requiring additional per-connection state to
252attach a dynamic memory pointer to.
253The memory should be allocated and attached in the module's
254.Va cb_init
255hook function.
256.Pp
257The
258.Va bytes_this_ack
259field specifies the number of new bytes acknowledged by the most recently
260received ACK packet.
261It is only valid in the
262.Va ack_received
263hook function.
264.Pp
265The
266.Va curack
267field specifies the sequence number of the most recently received ACK packet.
268It is only valid in the
269.Va ack_received ,
270.Va cong_signal
271and
272.Va post_recovery
273hook functions.
274.Pp
275The
276.Va flags
277field is used to pass useful information from the stack to a
278.Nm
279module.
280The CCF_ABC_SENTAWND flag is relevant in
281.Va ack_received
282and is set when appropriate byte counting (RFC3465) has counted a window's worth
283of bytes has been sent.
284It is the module's responsibility to clear the flag after it has processed the
285signal.
286The CCF_CWND_LIMITED flag is relevant in
287.Va ack_received
288and is set when the connection's ability to send data is currently constrained
289by the value of the congestion window.
290Algorithms should use the absence of this flag being set to avoid accumulating
291a large difference between the congestion window and send window.
292.Sh SEE ALSO
293.Xr cc_cdg 4 ,
294.Xr cc_chd 4 ,
295.Xr cc_cubic 4 ,
296.Xr cc_hd 4 ,
297.Xr cc_htcp 4 ,
298.Xr cc_newreno 4 ,
299.Xr cc_vegas 4 ,
300.Xr mod_cc 4 ,
301.Xr tcp 4
302.Sh ACKNOWLEDGEMENTS
303Development and testing of this software were made possible in part by grants
304from the FreeBSD Foundation and Cisco University Research Program Fund at
305Community Foundation Silicon Valley.
306.Sh FUTURE WORK
307Integrate with
308.Xr sctp 4 .
309.Sh HISTORY
310The modular Congestion Control (CC) framework first appeared in
311.Fx 9.0 .
312.Pp
313The framework was first released in 2007 by James Healy and Lawrence Stewart
314whilst working on the NewTCP research project at Swinburne University of
315Technology's Centre for Advanced Internet Architectures, Melbourne, Australia,
316which was made possible in part by a grant from the Cisco University Research
317Program Fund at Community Foundation Silicon Valley.
318More details are available at:
319.Pp
320http://caia.swin.edu.au/urp/newtcp/
321.Sh AUTHORS
322.An -nosplit
323The
324.Nm
325framework was written by
326.An Lawrence Stewart Aq Mt lstewart@FreeBSD.org ,
327.An James Healy Aq Mt jimmy@deefa.com
328and
329.An David Hayes Aq Mt david.hayes@ieee.org .
330.Pp
331This manual page was written by
332.An David Hayes Aq Mt david.hayes@ieee.org
333and
334.An Lawrence Stewart Aq Mt lstewart@FreeBSD.org .
335