1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause 3 * 4 * Copyright (c) 2009-2010 5 * Swinburne University of Technology, Melbourne, Australia 6 * Copyright (c) 2010 Lawrence Stewart <lstewart@freebsd.org> 7 * Copyright (c) 2010-2011 The FreeBSD Foundation 8 * All rights reserved. 9 * 10 * This software was developed at the Centre for Advanced Internet 11 * Architectures, Swinburne University of Technology, by David Hayes and 12 * Lawrence Stewart, made possible in part by a grant from the Cisco University 13 * Research Program Fund at Community Foundation Silicon Valley. 14 * 15 * Portions of this software were developed at the Centre for Advanced Internet 16 * Architectures, Swinburne University of Technology, Melbourne, Australia by 17 * David Hayes under sponsorship from the FreeBSD Foundation. 18 * 19 * Redistribution and use in source and binary forms, with or without 20 * modification, are permitted provided that the following conditions 21 * are met: 22 * 1. Redistributions of source code must retain the above copyright 23 * notice, this list of conditions and the following disclaimer. 24 * 2. Redistributions in binary form must reproduce the above copyright 25 * notice, this list of conditions and the following disclaimer in the 26 * documentation and/or other materials provided with the distribution. 27 * 28 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 29 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 30 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 31 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 32 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 33 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 34 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 35 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 36 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 37 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 38 * SUCH DAMAGE. 39 */ 40 41 /* 42 * An implementation of the Hamilton Institute's delay-based congestion control 43 * algorithm for FreeBSD, based on "A strategy for fair coexistence of loss and 44 * delay-based congestion control algorithms," by L. Budzisz, R. Stanojevic, R. 45 * Shorten, and F. Baker, IEEE Commun. Lett., vol. 13, no. 7, pp. 555--557, Jul. 46 * 2009. 47 * 48 * Originally released as part of the NewTCP research project at Swinburne 49 * University of Technology's Centre for Advanced Internet Architectures, 50 * Melbourne, Australia, which was made possible in part by a grant from the 51 * Cisco University Research Program Fund at Community Foundation Silicon 52 * Valley. More details are available at: 53 * http://caia.swin.edu.au/urp/newtcp/ 54 */ 55 56 #include <sys/cdefs.h> 57 #include <sys/param.h> 58 #include <sys/kernel.h> 59 #include <sys/khelp.h> 60 #include <sys/limits.h> 61 #include <sys/malloc.h> 62 #include <sys/module.h> 63 #include <sys/queue.h> 64 #include <sys/socket.h> 65 #include <sys/socketvar.h> 66 #include <sys/sysctl.h> 67 #include <sys/systm.h> 68 69 #include <net/vnet.h> 70 71 #include <netinet/in.h> 72 #include <netinet/in_pcb.h> 73 #include <netinet/tcp_seq.h> 74 #include <netinet/tcp_timer.h> 75 #include <netinet/tcp_var.h> 76 #include <netinet/cc/cc.h> 77 #include <netinet/cc/cc_module.h> 78 79 #include <netinet/khelp/h_ertt.h> 80 81 /* Largest possible number returned by random(). */ 82 #define RANDOM_MAX INT_MAX 83 84 static void hd_ack_received(struct cc_var *ccv, uint16_t ack_type); 85 static int hd_mod_init(void); 86 static size_t hd_data_sz(void); 87 88 static int ertt_id; 89 90 VNET_DEFINE_STATIC(uint32_t, hd_qthresh) = 20; 91 VNET_DEFINE_STATIC(uint32_t, hd_qmin) = 5; 92 VNET_DEFINE_STATIC(uint32_t, hd_pmax) = 5; 93 #define V_hd_qthresh VNET(hd_qthresh) 94 #define V_hd_qmin VNET(hd_qmin) 95 #define V_hd_pmax VNET(hd_pmax) 96 97 struct cc_algo hd_cc_algo = { 98 .name = "hd", 99 .ack_received = hd_ack_received, 100 .mod_init = hd_mod_init, 101 .cc_data_sz = hd_data_sz, 102 .after_idle = newreno_cc_after_idle, 103 .cong_signal = newreno_cc_cong_signal, 104 .post_recovery = newreno_cc_post_recovery, 105 }; 106 107 static size_t 108 hd_data_sz(void) 109 { 110 return (0); 111 } 112 113 /* 114 * Hamilton backoff function. Returns 1 if we should backoff or 0 otherwise. 115 */ 116 static __inline int 117 should_backoff(int qdly, int maxqdly) 118 { 119 unsigned long p; 120 121 if (qdly < V_hd_qthresh) { 122 p = (((RANDOM_MAX / 100) * V_hd_pmax) / 123 (V_hd_qthresh - V_hd_qmin)) * (qdly - V_hd_qmin); 124 } else { 125 if (qdly > V_hd_qthresh) 126 p = (((RANDOM_MAX / 100) * V_hd_pmax) / 127 (maxqdly - V_hd_qthresh)) * (maxqdly - qdly); 128 else 129 p = (RANDOM_MAX / 100) * V_hd_pmax; 130 } 131 132 return (random() < p); 133 } 134 135 /* 136 * If the ack type is CC_ACK, and the inferred queueing delay is greater than 137 * the Qmin threshold, cwnd is reduced probabilistically. When backing off due 138 * to delay, HD behaves like NewReno when an ECN signal is received. HD behaves 139 * as NewReno in all other circumstances. 140 */ 141 static void 142 hd_ack_received(struct cc_var *ccv, uint16_t ack_type) 143 { 144 struct ertt *e_t; 145 int qdly; 146 147 if (ack_type == CC_ACK) { 148 e_t = khelp_get_osd(&CCV(ccv, t_osd), ertt_id); 149 150 if (e_t->rtt && e_t->minrtt && V_hd_qthresh > 0) { 151 qdly = e_t->rtt - e_t->minrtt; 152 153 if (qdly > V_hd_qmin && 154 !IN_RECOVERY(CCV(ccv, t_flags))) { 155 /* Probabilistic backoff of cwnd. */ 156 if (should_backoff(qdly, 157 e_t->maxrtt - e_t->minrtt)) { 158 /* 159 * Update cwnd and ssthresh update to 160 * half cwnd and behave like an ECN (ie 161 * not a packet loss). 162 */ 163 newreno_cc_cong_signal(ccv, 164 CC_ECN); 165 return; 166 } 167 } 168 } 169 } 170 newreno_cc_ack_received(ccv, ack_type); 171 } 172 173 static int 174 hd_mod_init(void) 175 { 176 177 ertt_id = khelp_get_id("ertt"); 178 if (ertt_id <= 0) { 179 printf("%s: h_ertt module not found\n", __func__); 180 return (ENOENT); 181 } 182 return (0); 183 } 184 185 static int 186 hd_pmax_handler(SYSCTL_HANDLER_ARGS) 187 { 188 int error; 189 uint32_t new; 190 191 new = V_hd_pmax; 192 error = sysctl_handle_int(oidp, &new, 0, req); 193 if (error == 0 && req->newptr != NULL) { 194 if (new == 0 || new > 100) 195 error = EINVAL; 196 else 197 V_hd_pmax = new; 198 } 199 200 return (error); 201 } 202 203 static int 204 hd_qmin_handler(SYSCTL_HANDLER_ARGS) 205 { 206 int error; 207 uint32_t new; 208 209 new = V_hd_qmin; 210 error = sysctl_handle_int(oidp, &new, 0, req); 211 if (error == 0 && req->newptr != NULL) { 212 if (new > V_hd_qthresh) 213 error = EINVAL; 214 else 215 V_hd_qmin = new; 216 } 217 218 return (error); 219 } 220 221 static int 222 hd_qthresh_handler(SYSCTL_HANDLER_ARGS) 223 { 224 int error; 225 uint32_t new; 226 227 new = V_hd_qthresh; 228 error = sysctl_handle_int(oidp, &new, 0, req); 229 if (error == 0 && req->newptr != NULL) { 230 if (new == 0 || new < V_hd_qmin) 231 error = EINVAL; 232 else 233 V_hd_qthresh = new; 234 } 235 236 return (error); 237 } 238 239 SYSCTL_DECL(_net_inet_tcp_cc_hd); 240 SYSCTL_NODE(_net_inet_tcp_cc, OID_AUTO, hd, CTLFLAG_RW | CTLFLAG_MPSAFE, NULL, 241 "Hamilton delay-based congestion control related settings"); 242 243 SYSCTL_PROC(_net_inet_tcp_cc_hd, OID_AUTO, queue_threshold, 244 CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, 245 &VNET_NAME(hd_qthresh), 20, &hd_qthresh_handler, "IU", 246 "queueing congestion threshold (qth) in ticks"); 247 248 SYSCTL_PROC(_net_inet_tcp_cc_hd, OID_AUTO, pmax, 249 CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, 250 &VNET_NAME(hd_pmax), 5, &hd_pmax_handler, "IU", 251 "per packet maximum backoff probability as a percentage"); 252 253 SYSCTL_PROC(_net_inet_tcp_cc_hd, OID_AUTO, queue_min, 254 CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, 255 &VNET_NAME(hd_qmin), 5, &hd_qmin_handler, "IU", 256 "minimum queueing delay threshold (qmin) in ticks"); 257 258 DECLARE_CC_MODULE(hd, &hd_cc_algo); 259 MODULE_VERSION(hd, 2); 260 MODULE_DEPEND(hd, ertt, 1, 1, 1); 261