xref: /freebsd/sys/netinet/ip_dummynet.h (revision 884ba43116d4456d5900d3c8824153c604f132b8)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (c) 1998-2010 Luigi Rizzo, Universita` di Pisa
5  * Portions Copyright (c) 2000 Akamba Corp.
6  * All rights reserved
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  * $FreeBSD$
30  */
31 
32 #ifndef _IP_DUMMYNET_H
33 #define _IP_DUMMYNET_H
34 #define NEW_AQM
35 /*
36  * Definition of the kernel-userland API for dummynet.
37  *
38  * Setsockopt() and getsockopt() pass a batch of objects, each
39  * of them starting with a "struct dn_id" which should fully identify
40  * the object and its relation with others in the sequence.
41  * The first object in each request should have
42  *	 type= DN_CMD_*, id = DN_API_VERSION.
43  * For other objects, type and subtype specify the object, len indicates
44  * the total length including the header, and 'id' identifies the specific
45  * object.
46  *
47  * Most objects are numbered with an identifier in the range 1..65535.
48  * DN_MAX_ID indicates the first value outside the range.
49  */
50 
51 #define	DN_API_VERSION	12500000
52 #define	DN_MAX_ID	0x10000
53 
54 struct dn_id {
55 	uint16_t	len;	/* total obj len including this header */
56 	uint8_t		type;
57 	uint8_t		subtype;
58 	uint32_t	id;	/* generic id */
59 };
60 
61 /*
62  * These values are in the type field of struct dn_id.
63  * To preserve the ABI, never rearrange the list or delete
64  * entries with the exception of DN_LAST
65  */
66 enum {
67 	DN_NONE = 0,
68 	DN_LINK = 1,
69 	DN_FS,
70 	DN_SCH,
71 	DN_SCH_I,
72 	DN_QUEUE,
73 	DN_DELAY_LINE,
74 	DN_PROFILE,
75 	DN_FLOW,		/* struct dn_flow */
76 	DN_TEXT,		/* opaque text is the object */
77 
78 	DN_CMD_CONFIG = 0x80,	/* objects follow */
79 	DN_CMD_DELETE,		/* subtype + list of entries */
80 	DN_CMD_GET,		/* subtype + list of entries */
81 	DN_CMD_FLUSH,
82 	/* for compatibility with FreeBSD 7.2/8 */
83 	DN_COMPAT_PIPE,
84 	DN_COMPAT_QUEUE,
85 	DN_GET_COMPAT,
86 
87 	/* special commands for emulation of sysctl variables */
88 	DN_SYSCTL_GET,
89 	DN_SYSCTL_SET,
90 #ifdef NEW_AQM
91 	/* subtypes used for setting/getting extra parameters.
92 	 * these subtypes used with IP_DUMMYNET3 command (get)
93 	 * and DN_TEXT (set). */
94 	DN_AQM_PARAMS, /* AQM extra params */
95 	DN_SCH_PARAMS, /* scheduler extra params */
96 #endif
97 	DN_LAST,
98 };
99 
100 enum { /* subtype for schedulers, flowset and the like */
101 	DN_SCHED_UNKNOWN = 0,
102 	DN_SCHED_FIFO = 1,
103 	DN_SCHED_WF2QP = 2,
104 	/* others are in individual modules */
105 };
106 
107 enum {	/* user flags */
108 	DN_HAVE_MASK	= 0x0001,	/* fs or sched has a mask */
109 	DN_NOERROR	= 0x0002,	/* do not report errors */
110 	DN_QHT_HASH	= 0x0004,	/* qht is a hash table */
111 	DN_QSIZE_BYTES	= 0x0008,	/* queue size is in bytes */
112 	DN_HAS_PROFILE	= 0x0010,	/* a link has a profile */
113 	DN_IS_RED	= 0x0020,
114 	DN_IS_GENTLE_RED= 0x0040,
115 	DN_IS_ECN	= 0x0080,
116 	#ifdef NEW_AQM
117 	DN_IS_AQM = 0x0100,     /* AQMs: e.g Codel & PIE */
118 	#endif
119 	DN_PIPE_CMD	= 0x1000,	/* pipe config... */
120 };
121 
122 /*
123  * link template.
124  */
125 struct dn_link {
126 	struct dn_id oid;
127 
128 	/*
129 	 * Userland sets bw and delay in bits/s and milliseconds.
130 	 * The kernel converts this back and forth to bits/tick and ticks.
131 	 * XXX what about burst ?
132 	 */
133 	int32_t		link_nr;
134 	uint32_t	bandwidth;	/* bit/s or bits/tick.   */
135 	int		delay;		/* ms and ticks */
136 	uint64_t	burst;		/* scaled. bits*Hz  XXX */
137 };
138 
139 /*
140  * A flowset, which is a template for flows. Contains parameters
141  * from the command line: id, target scheduler, queue sizes, plr,
142  * flow masks, buckets for the flow hash, and possibly scheduler-
143  * specific parameters (weight, quantum and so on).
144  */
145 struct dn_fs {
146 	struct dn_id oid;
147 	uint32_t fs_nr;		/* the flowset number */
148 	uint32_t flags;		/* userland flags */
149 	int qsize;		/* queue size in slots or bytes */
150 	int32_t plr;		/* PLR, pkt loss rate (2^31-1 means 100%) */
151 	uint32_t buckets;	/* buckets used for the queue hash table */
152 
153 	struct ipfw_flow_id flow_mask;
154 	uint32_t sched_nr;	/* the scheduler we attach to */
155 	/* generic scheduler parameters. Leave them at -1 if unset.
156 	 * Now we use 0: weight, 1: lmax, 2: priority
157 	 */
158 	int par[4];
159 
160 	/* RED/GRED parameters.
161 	 * weight and probabilities are in the range 0..1 represented
162 	 * in fixed point arithmetic with SCALE_RED decimal bits.
163 	 */
164 #define SCALE_RED	16
165 #define SCALE(x)	( (x) << SCALE_RED )
166 #define SCALE_VAL(x)	( (x) >> SCALE_RED )
167 #define SCALE_MUL(x,y)	( ( (x) * (y) ) >> SCALE_RED )
168 	int w_q ;		/* queue weight (scaled) */
169 	int max_th ;		/* maximum threshold for queue (scaled) */
170 	int min_th ;		/* minimum threshold for queue (scaled) */
171 	int max_p ;		/* maximum value for p_b (scaled) */
172 
173 };
174 
175 /*
176  * dn_flow collects flow_id and stats for queues and scheduler
177  * instances, and is used to pass these info to userland.
178  * oid.type/oid.subtype describe the object, oid.id is number
179  * of the parent object.
180  */
181 struct dn_flow {
182 	struct dn_id	oid;
183 	struct ipfw_flow_id fid;
184 	uint64_t	tot_pkts; /* statistics counters  */
185 	uint64_t	tot_bytes;
186 	uint32_t	length; /* Queue length, in packets */
187 	uint32_t	len_bytes; /* Queue length, in bytes */
188 	uint32_t	drops;
189 };
190 
191 /*
192  * Scheduler template, mostly indicating the name, number,
193  * sched_mask and buckets.
194  */
195 struct dn_sch {
196 	struct dn_id	oid;
197 	uint32_t	sched_nr; /* N, scheduler number */
198 	uint32_t	buckets; /* number of buckets for the instances */
199 	uint32_t	flags;	/* have_mask, ... */
200 
201 	char name[16];	/* null terminated */
202 	/* mask to select the appropriate scheduler instance */
203 	struct ipfw_flow_id sched_mask; /* M */
204 };
205 
206 /* A delay profile is attached to a link.
207  * Note that a profile, as any other object, cannot be longer than 2^16
208  */
209 #define	ED_MAX_SAMPLES_NO	1024
210 struct dn_profile {
211 	struct dn_id	oid;
212 	/* fields to simulate a delay profile */
213 #define ED_MAX_NAME_LEN		32
214 	char	name[ED_MAX_NAME_LEN];
215 	int	link_nr;
216 	int	loss_level;
217 	uint32_t	bandwidth;			// XXX use link bandwidth?
218 	int	samples_no;			/* actual len of samples[] */
219 	int	samples[ED_MAX_SAMPLES_NO];	/* may be shorter */
220 };
221 
222 #ifdef NEW_AQM
223 /* Extra parameters for AQM and scheduler.
224  * This struct is used to pass and retrieve parameters (configurations)
225  * to/from AQM and Scheduler.
226  */
227 struct dn_extra_parms {
228 	struct dn_id oid;
229 	char name[16];
230 	uint32_t nr;
231 #define DN_MAX_EXTRA_PARM	10
232 	int64_t par[DN_MAX_EXTRA_PARM];
233 };
234 #endif
235 
236 /*
237  * Overall structure of dummynet
238 
239 In dummynet, packets are selected with the firewall rules, and passed
240 to two different objects: PIPE or QUEUE (bad name).
241 
242 A QUEUE defines a classifier, which groups packets into flows
243 according to a 'mask', puts them into independent queues (one
244 per flow) with configurable size and queue management policy,
245 and passes flows to a scheduler:
246 
247                  (flow_mask|sched_mask)  sched_mask
248 	 +---------+   weight Wx  +-------------+
249          |         |->-[flow]-->--|             |-+
250     -->--| QUEUE x |   ...        |             | |
251          |         |->-[flow]-->--| SCHEDuler N | |
252 	 +---------+              |             | |
253 	     ...                  |             +--[LINK N]-->--
254 	 +---------+   weight Wy  |             | +--[LINK N]-->--
255          |         |->-[flow]-->--|             | |
256     -->--| QUEUE y |   ...        |             | |
257          |         |->-[flow]-->--|             | |
258 	 +---------+              +-------------+ |
259 	                            +-------------+
260 
261 Many QUEUE objects can connect to the same scheduler, each
262 QUEUE object can have its own set of parameters.
263 
264 In turn, the SCHEDuler 'forks' multiple instances according
265 to a 'sched_mask', each instance manages its own set of queues
266 and transmits on a private instance of a configurable LINK.
267 
268 A PIPE is a simplified version of the above, where there
269 is no flow_mask, and each scheduler instance handles a single queue.
270 
271 The following data structures (visible from userland) describe
272 the objects used by dummynet:
273 
274  + dn_link, contains the main configuration parameters related
275    to delay and bandwidth;
276  + dn_profile describes a delay profile;
277  + dn_flow describes the flow status (flow id, statistics)
278 
279  + dn_sch describes a scheduler
280  + dn_fs describes a flowset (msk, weight, queue parameters)
281 
282  *
283  */
284 
285 #endif /* _IP_DUMMYNET_H */
286