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 __FBSDID("$FreeBSD$"); 58 59 #include <sys/param.h> 60 #include <sys/kernel.h> 61 #include <sys/khelp.h> 62 #include <sys/limits.h> 63 #include <sys/malloc.h> 64 #include <sys/module.h> 65 #include <sys/queue.h> 66 #include <sys/socket.h> 67 #include <sys/socketvar.h> 68 #include <sys/sysctl.h> 69 #include <sys/systm.h> 70 71 #include <net/vnet.h> 72 73 #include <netinet/in.h> 74 #include <netinet/in_pcb.h> 75 #include <netinet/tcp_seq.h> 76 #include <netinet/tcp_timer.h> 77 #include <netinet/tcp_var.h> 78 #include <netinet/cc/cc.h> 79 #include <netinet/cc/cc_module.h> 80 81 #include <netinet/khelp/h_ertt.h> 82 83 /* Largest possible number returned by random(). */ 84 #define RANDOM_MAX INT_MAX 85 86 static void hd_ack_received(struct cc_var *ccv, uint16_t ack_type); 87 static int hd_mod_init(void); 88 static size_t hd_data_sz(void); 89 90 static int ertt_id; 91 92 VNET_DEFINE_STATIC(uint32_t, hd_qthresh) = 20; 93 VNET_DEFINE_STATIC(uint32_t, hd_qmin) = 5; 94 VNET_DEFINE_STATIC(uint32_t, hd_pmax) = 5; 95 #define V_hd_qthresh VNET(hd_qthresh) 96 #define V_hd_qmin VNET(hd_qmin) 97 #define V_hd_pmax VNET(hd_pmax) 98 99 struct cc_algo hd_cc_algo = { 100 .name = "hd", 101 .ack_received = hd_ack_received, 102 .mod_init = hd_mod_init, 103 .cc_data_sz = hd_data_sz, 104 .after_idle = newreno_cc_after_idle, 105 .cong_signal = newreno_cc_cong_signal, 106 .post_recovery = newreno_cc_post_recovery, 107 }; 108 109 static size_t 110 hd_data_sz(void) 111 { 112 return (0); 113 } 114 115 /* 116 * Hamilton backoff function. Returns 1 if we should backoff or 0 otherwise. 117 */ 118 static __inline int 119 should_backoff(int qdly, int maxqdly) 120 { 121 unsigned long p; 122 123 if (qdly < V_hd_qthresh) { 124 p = (((RANDOM_MAX / 100) * V_hd_pmax) / 125 (V_hd_qthresh - V_hd_qmin)) * (qdly - V_hd_qmin); 126 } else { 127 if (qdly > V_hd_qthresh) 128 p = (((RANDOM_MAX / 100) * V_hd_pmax) / 129 (maxqdly - V_hd_qthresh)) * (maxqdly - qdly); 130 else 131 p = (RANDOM_MAX / 100) * V_hd_pmax; 132 } 133 134 return (random() < p); 135 } 136 137 /* 138 * If the ack type is CC_ACK, and the inferred queueing delay is greater than 139 * the Qmin threshold, cwnd is reduced probabilistically. When backing off due 140 * to delay, HD behaves like NewReno when an ECN signal is received. HD behaves 141 * as NewReno in all other circumstances. 142 */ 143 static void 144 hd_ack_received(struct cc_var *ccv, uint16_t ack_type) 145 { 146 struct ertt *e_t; 147 int qdly; 148 149 if (ack_type == CC_ACK) { 150 e_t = khelp_get_osd(&CCV(ccv, t_osd), ertt_id); 151 152 if (e_t->rtt && e_t->minrtt && V_hd_qthresh > 0) { 153 qdly = e_t->rtt - e_t->minrtt; 154 155 if (qdly > V_hd_qmin && 156 !IN_RECOVERY(CCV(ccv, t_flags))) { 157 /* Probabilistic backoff of cwnd. */ 158 if (should_backoff(qdly, 159 e_t->maxrtt - e_t->minrtt)) { 160 /* 161 * Update cwnd and ssthresh update to 162 * half cwnd and behave like an ECN (ie 163 * not a packet loss). 164 */ 165 newreno_cc_cong_signal(ccv, 166 CC_ECN); 167 return; 168 } 169 } 170 } 171 } 172 newreno_cc_ack_received(ccv, ack_type); 173 } 174 175 static int 176 hd_mod_init(void) 177 { 178 179 ertt_id = khelp_get_id("ertt"); 180 if (ertt_id <= 0) { 181 printf("%s: h_ertt module not found\n", __func__); 182 return (ENOENT); 183 } 184 return (0); 185 } 186 187 static int 188 hd_pmax_handler(SYSCTL_HANDLER_ARGS) 189 { 190 int error; 191 uint32_t new; 192 193 new = V_hd_pmax; 194 error = sysctl_handle_int(oidp, &new, 0, req); 195 if (error == 0 && req->newptr != NULL) { 196 if (new == 0 || new > 100) 197 error = EINVAL; 198 else 199 V_hd_pmax = new; 200 } 201 202 return (error); 203 } 204 205 static int 206 hd_qmin_handler(SYSCTL_HANDLER_ARGS) 207 { 208 int error; 209 uint32_t new; 210 211 new = V_hd_qmin; 212 error = sysctl_handle_int(oidp, &new, 0, req); 213 if (error == 0 && req->newptr != NULL) { 214 if (new > V_hd_qthresh) 215 error = EINVAL; 216 else 217 V_hd_qmin = new; 218 } 219 220 return (error); 221 } 222 223 static int 224 hd_qthresh_handler(SYSCTL_HANDLER_ARGS) 225 { 226 int error; 227 uint32_t new; 228 229 new = V_hd_qthresh; 230 error = sysctl_handle_int(oidp, &new, 0, req); 231 if (error == 0 && req->newptr != NULL) { 232 if (new == 0 || new < V_hd_qmin) 233 error = EINVAL; 234 else 235 V_hd_qthresh = new; 236 } 237 238 return (error); 239 } 240 241 SYSCTL_DECL(_net_inet_tcp_cc_hd); 242 SYSCTL_NODE(_net_inet_tcp_cc, OID_AUTO, hd, CTLFLAG_RW | CTLFLAG_MPSAFE, NULL, 243 "Hamilton delay-based congestion control related settings"); 244 245 SYSCTL_PROC(_net_inet_tcp_cc_hd, OID_AUTO, queue_threshold, 246 CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, 247 &VNET_NAME(hd_qthresh), 20, &hd_qthresh_handler, "IU", 248 "queueing congestion threshold (qth) in ticks"); 249 250 SYSCTL_PROC(_net_inet_tcp_cc_hd, OID_AUTO, pmax, 251 CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, 252 &VNET_NAME(hd_pmax), 5, &hd_pmax_handler, "IU", 253 "per packet maximum backoff probability as a percentage"); 254 255 SYSCTL_PROC(_net_inet_tcp_cc_hd, OID_AUTO, queue_min, 256 CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, 257 &VNET_NAME(hd_qmin), 5, &hd_qmin_handler, "IU", 258 "minimum queueing delay threshold (qmin) in ticks"); 259 260 DECLARE_CC_MODULE(hd, &hd_cc_algo); 261 MODULE_VERSION(hd, 2); 262 MODULE_DEPEND(hd, ertt, 1, 1, 1); 263