1 /*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License, Version 1.0 only
6 * (the "License"). You may not use this file except in compliance
7 * with the License.
8 *
9 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
10 * or http://www.opensolaris.org/os/licensing.
11 * See the License for the specific language governing permissions
12 * and limitations under the License.
13 *
14 * When distributing Covered Code, include this CDDL HEADER in each
15 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
16 * If applicable, add the following below this CDDL HEADER, with the
17 * fields enclosed by brackets "[]" replaced with your own identifying
18 * information: Portions Copyright [yyyy] [name of copyright owner]
19 *
20 * CDDL HEADER END
21 */
22 /*
23 * Copyright 2005 Sun Microsystems, Inc. All rights reserved.
24 * Use is subject to license terms.
25 */
26
27 #include <sys/types.h>
28 #include <sys/kmem.h>
29 #include <sys/random.h>
30 #include <netinet/in.h>
31 #include <netinet/in_systm.h>
32 #include <netinet/ip6.h>
33 #include <inet/common.h>
34 #include <inet/ip.h>
35 #include <inet/ip6.h>
36 #include <ipp/meters/meter_impl.h>
37
38 /*
39 * Module : Time Sliding Window meter - tswtclmtr
40 * Description
41 * This module implements the metering part of RFC 2859. It accepts the
42 * committed rate, peak rate and the window for a flow and determines
43 * if the flow is within the committed/peak rate and assigns the appropriate
44 * next action.
45 * The meter provides an estimate of the running average bandwidth for the
46 * flow over the specified window. It uses probability to benefit TCP flows
47 * as it reduces the likelihood of dropping multiple packets within a TCP
48 * window without adversely effecting UDP flows.
49 */
50
51 int tswtcl_debug = 0;
52
53 /*
54 * Given a packet and the tswtcl_data it belongs to, this routine meters the
55 * ToS or DSCP for IPv4 and IPv6 resp. with the values configured for
56 * the tswtcl_data.
57 */
58 /* ARGSUSED */
59 int
60 tswtcl_process(mblk_t **mpp, tswtcl_data_t *tswtcl_data,
61 ipp_action_id_t *next_action)
62 {
63 ipha_t *ipha;
64 hrtime_t now;
65 ip6_t *ip6_hdr;
66 uint32_t pkt_len;
67 mblk_t *mp = *mpp;
68 hrtime_t deltaT;
69 uint64_t bitsinwin;
70 uint32_t min = 0, additive, rnd;
71 tswtcl_cfg_t *cfg_parms = tswtcl_data->cfg_parms;
72
73 if (mp == NULL) {
74 tswtcl0dbg(("tswtcl_process: null mp!\n"));
75 atomic_inc_64(&tswtcl_data->epackets);
76 return (EINVAL);
77 }
78
79 if (mp->b_datap->db_type != M_DATA) {
80 if ((mp->b_cont != NULL) &&
81 (mp->b_cont->b_datap->db_type == M_DATA)) {
82 mp = mp->b_cont;
83 } else {
84 tswtcl0dbg(("tswtcl_process: no data\n"));
85 atomic_inc_64(&tswtcl_data->epackets);
86 return (EINVAL);
87 }
88 }
89
90 /* Figure out the ToS/Traffic Class and length from the message */
91 if ((mp->b_wptr - mp->b_rptr) < IP_SIMPLE_HDR_LENGTH) {
92 if (!pullupmsg(mp, IP_SIMPLE_HDR_LENGTH)) {
93 tswtcl0dbg(("tswtcl_process: pullup error\n"));
94 atomic_inc_64(&tswtcl_data->epackets);
95 return (EINVAL);
96 }
97 }
98 ipha = (ipha_t *)mp->b_rptr;
99 if (IPH_HDR_VERSION(ipha) == IPV4_VERSION) {
100 pkt_len = ntohs(ipha->ipha_length);
101 } else {
102 ip6_hdr = (ip6_t *)mp->b_rptr;
103 pkt_len = ntohs(ip6_hdr->ip6_plen) +
104 ip_hdr_length_v6(mp, ip6_hdr);
105 }
106
107 /* Convert into bits */
108 pkt_len <<= 3;
109
110 /* Get current time */
111 now = gethrtime();
112
113 /* Update the avg_rate and win_front tswtcl_data */
114 mutex_enter(&tswtcl_data->tswtcl_lock);
115
116 /* avg_rate = bits/sec and window in msec */
117 bitsinwin = ((uint64_t)tswtcl_data->avg_rate * cfg_parms->window /
118 1000) + pkt_len;
119
120 deltaT = now - tswtcl_data->win_front + cfg_parms->nsecwindow;
121
122 tswtcl_data->avg_rate = (uint64_t)bitsinwin * METER_SEC_TO_NSEC /
123 deltaT;
124 tswtcl_data->win_front = now;
125
126 if (tswtcl_data->avg_rate <= cfg_parms->committed_rate) {
127 *next_action = cfg_parms->green_action;
128 } else if (tswtcl_data->avg_rate <= cfg_parms->peak_rate) {
129 /*
130 * Compute the probability:
131 *
132 * p0 = (avg_rate - committed_rate) / avg_rate
133 *
134 * Yellow with probability p0
135 * Green with probability (1 - p0)
136 *
137 */
138 uint32_t aminusc;
139
140 /* Get a random no. betweeen 0 and avg_rate */
141 (void) random_get_pseudo_bytes((uint8_t *)&additive,
142 sizeof (additive));
143 rnd = min + (additive % (tswtcl_data->avg_rate - min + 1));
144
145 aminusc = tswtcl_data->avg_rate - cfg_parms->committed_rate;
146 if (aminusc >= rnd) {
147 *next_action = cfg_parms->yellow_action;
148 } else {
149 *next_action = cfg_parms->green_action;
150 }
151 } else {
152 /*
153 * Compute the probability:
154 *
155 * p1 = (avg_rate - peak_rate) / avg_rate
156 * p2 = (peak_rate - committed_rate) / avg_rate
157 *
158 * Red with probability p1
159 * Yellow with probability p2
160 * Green with probability (1 - (p1 + p2))
161 *
162 */
163 uint32_t aminusp;
164
165 /* Get a random no. betweeen 0 and avg_rate */
166 (void) random_get_pseudo_bytes((uint8_t *)&additive,
167 sizeof (additive));
168 rnd = min + (additive % (tswtcl_data->avg_rate - min + 1));
169
170 aminusp = tswtcl_data->avg_rate - cfg_parms->peak_rate;
171
172 if (aminusp >= rnd) {
173 *next_action = cfg_parms->red_action;
174 } else if ((cfg_parms->pminusc + aminusp) >= rnd) {
175 *next_action = cfg_parms->yellow_action;
176 } else {
177 *next_action = cfg_parms->green_action;
178 }
179
180 }
181 mutex_exit(&tswtcl_data->tswtcl_lock);
182
183 /* Update Stats */
184 if (*next_action == cfg_parms->green_action) {
185 atomic_inc_64(&tswtcl_data->green_packets);
186 atomic_add_64(&tswtcl_data->green_bits, pkt_len);
187 } else if (*next_action == cfg_parms->yellow_action) {
188 atomic_inc_64(&tswtcl_data->yellow_packets);
189 atomic_add_64(&tswtcl_data->yellow_bits, pkt_len);
190 } else {
191 ASSERT(*next_action == cfg_parms->red_action);
192 atomic_inc_64(&tswtcl_data->red_packets);
193 atomic_add_64(&tswtcl_data->red_bits, pkt_len);
194 }
195 return (0);
196 }