1 /* 2 * TCP Westwood+: end-to-end bandwidth estimation for TCP 3 * 4 * Angelo Dell'Aera: author of the first version of TCP Westwood+ in Linux 2.4 5 * 6 * Support at http://c3lab.poliba.it/index.php/Westwood 7 * Main references in literature: 8 * 9 * - Mascolo S, Casetti, M. Gerla et al. 10 * "TCP Westwood: bandwidth estimation for TCP" Proc. ACM Mobicom 2001 11 * 12 * - A. Grieco, s. Mascolo 13 * "Performance evaluation of New Reno, Vegas, Westwood+ TCP" ACM Computer 14 * Comm. Review, 2004 15 * 16 * - A. Dell'Aera, L. Grieco, S. Mascolo. 17 * "Linux 2.4 Implementation of Westwood+ TCP with Rate-Halving : 18 * A Performance Evaluation Over the Internet" (ICC 2004), Paris, June 2004 19 * 20 * Westwood+ employs end-to-end bandwidth measurement to set cwnd and 21 * ssthresh after packet loss. The probing phase is as the original Reno. 22 */ 23 24 #include <linux/mm.h> 25 #include <linux/module.h> 26 #include <linux/skbuff.h> 27 #include <linux/inet_diag.h> 28 #include <net/tcp.h> 29 30 /* TCP Westwood structure */ 31 struct westwood { 32 u32 bw_ns_est; /* first bandwidth estimation..not too smoothed 8) */ 33 u32 bw_est; /* bandwidth estimate */ 34 u32 rtt_win_sx; /* here starts a new evaluation... */ 35 u32 bk; 36 u32 snd_una; /* used for evaluating the number of acked bytes */ 37 u32 cumul_ack; 38 u32 accounted; 39 u32 rtt; 40 u32 rtt_min; /* minimum observed RTT */ 41 u8 first_ack; /* flag which infers that this is the first ack */ 42 u8 reset_rtt_min; /* Reset RTT min to next RTT sample*/ 43 }; 44 45 46 /* TCP Westwood functions and constants */ 47 #define TCP_WESTWOOD_RTT_MIN (HZ/20) /* 50ms */ 48 #define TCP_WESTWOOD_INIT_RTT (20*HZ) /* maybe too conservative?! */ 49 50 /* 51 * @tcp_westwood_create 52 * This function initializes fields used in TCP Westwood+, 53 * it is called after the initial SYN, so the sequence numbers 54 * are correct but new passive connections we have no 55 * information about RTTmin at this time so we simply set it to 56 * TCP_WESTWOOD_INIT_RTT. This value was chosen to be too conservative 57 * since in this way we're sure it will be updated in a consistent 58 * way as soon as possible. It will reasonably happen within the first 59 * RTT period of the connection lifetime. 60 */ 61 static void tcp_westwood_init(struct sock *sk) 62 { 63 struct westwood *w = inet_csk_ca(sk); 64 65 w->bk = 0; 66 w->bw_ns_est = 0; 67 w->bw_est = 0; 68 w->accounted = 0; 69 w->cumul_ack = 0; 70 w->reset_rtt_min = 1; 71 w->rtt_min = w->rtt = TCP_WESTWOOD_INIT_RTT; 72 w->rtt_win_sx = tcp_time_stamp; 73 w->snd_una = tcp_sk(sk)->snd_una; 74 w->first_ack = 1; 75 } 76 77 /* 78 * @westwood_do_filter 79 * Low-pass filter. Implemented using constant coefficients. 80 */ 81 static inline u32 westwood_do_filter(u32 a, u32 b) 82 { 83 return ((7 * a) + b) >> 3; 84 } 85 86 static void westwood_filter(struct westwood *w, u32 delta) 87 { 88 /* If the filter is empty fill it with the first sample of bandwidth */ 89 if (w->bw_ns_est == 0 && w->bw_est == 0) { 90 w->bw_ns_est = w->bk / delta; 91 w->bw_est = w->bw_ns_est; 92 } else { 93 w->bw_ns_est = westwood_do_filter(w->bw_ns_est, w->bk / delta); 94 w->bw_est = westwood_do_filter(w->bw_est, w->bw_ns_est); 95 } 96 } 97 98 /* 99 * @westwood_pkts_acked 100 * Called after processing group of packets. 101 * but all westwood needs is the last sample of srtt. 102 */ 103 static void tcp_westwood_pkts_acked(struct sock *sk, u32 cnt, s32 rtt) 104 { 105 struct westwood *w = inet_csk_ca(sk); 106 107 if (rtt > 0) 108 w->rtt = usecs_to_jiffies(rtt); 109 } 110 111 /* 112 * @westwood_update_window 113 * It updates RTT evaluation window if it is the right moment to do 114 * it. If so it calls filter for evaluating bandwidth. 115 */ 116 static void westwood_update_window(struct sock *sk) 117 { 118 struct westwood *w = inet_csk_ca(sk); 119 s32 delta = tcp_time_stamp - w->rtt_win_sx; 120 121 /* Initialize w->snd_una with the first acked sequence number in order 122 * to fix mismatch between tp->snd_una and w->snd_una for the first 123 * bandwidth sample 124 */ 125 if (w->first_ack) { 126 w->snd_una = tcp_sk(sk)->snd_una; 127 w->first_ack = 0; 128 } 129 130 /* 131 * See if a RTT-window has passed. 132 * Be careful since if RTT is less than 133 * 50ms we don't filter but we continue 'building the sample'. 134 * This minimum limit was chosen since an estimation on small 135 * time intervals is better to avoid... 136 * Obviously on a LAN we reasonably will always have 137 * right_bound = left_bound + WESTWOOD_RTT_MIN 138 */ 139 if (w->rtt && delta > max_t(u32, w->rtt, TCP_WESTWOOD_RTT_MIN)) { 140 westwood_filter(w, delta); 141 142 w->bk = 0; 143 w->rtt_win_sx = tcp_time_stamp; 144 } 145 } 146 147 static inline void update_rtt_min(struct westwood *w) 148 { 149 if (w->reset_rtt_min) { 150 w->rtt_min = w->rtt; 151 w->reset_rtt_min = 0; 152 } else 153 w->rtt_min = min(w->rtt, w->rtt_min); 154 } 155 156 157 /* 158 * @westwood_fast_bw 159 * It is called when we are in fast path. In particular it is called when 160 * header prediction is successful. In such case in fact update is 161 * straight forward and doesn't need any particular care. 162 */ 163 static inline void westwood_fast_bw(struct sock *sk) 164 { 165 const struct tcp_sock *tp = tcp_sk(sk); 166 struct westwood *w = inet_csk_ca(sk); 167 168 westwood_update_window(sk); 169 170 w->bk += tp->snd_una - w->snd_una; 171 w->snd_una = tp->snd_una; 172 update_rtt_min(w); 173 } 174 175 /* 176 * @westwood_acked_count 177 * This function evaluates cumul_ack for evaluating bk in case of 178 * delayed or partial acks. 179 */ 180 static inline u32 westwood_acked_count(struct sock *sk) 181 { 182 const struct tcp_sock *tp = tcp_sk(sk); 183 struct westwood *w = inet_csk_ca(sk); 184 185 w->cumul_ack = tp->snd_una - w->snd_una; 186 187 /* If cumul_ack is 0 this is a dupack since it's not moving 188 * tp->snd_una. 189 */ 190 if (!w->cumul_ack) { 191 w->accounted += tp->mss_cache; 192 w->cumul_ack = tp->mss_cache; 193 } 194 195 if (w->cumul_ack > tp->mss_cache) { 196 /* Partial or delayed ack */ 197 if (w->accounted >= w->cumul_ack) { 198 w->accounted -= w->cumul_ack; 199 w->cumul_ack = tp->mss_cache; 200 } else { 201 w->cumul_ack -= w->accounted; 202 w->accounted = 0; 203 } 204 } 205 206 w->snd_una = tp->snd_una; 207 208 return w->cumul_ack; 209 } 210 211 212 /* 213 * TCP Westwood 214 * Here limit is evaluated as Bw estimation*RTTmin (for obtaining it 215 * in packets we use mss_cache). Rttmin is guaranteed to be >= 2 216 * so avoids ever returning 0. 217 */ 218 static u32 tcp_westwood_bw_rttmin(const struct sock *sk) 219 { 220 const struct tcp_sock *tp = tcp_sk(sk); 221 const struct westwood *w = inet_csk_ca(sk); 222 return max_t(u32, (w->bw_est * w->rtt_min) / tp->mss_cache, 2); 223 } 224 225 static void tcp_westwood_event(struct sock *sk, enum tcp_ca_event event) 226 { 227 struct tcp_sock *tp = tcp_sk(sk); 228 struct westwood *w = inet_csk_ca(sk); 229 230 switch (event) { 231 case CA_EVENT_FAST_ACK: 232 westwood_fast_bw(sk); 233 break; 234 235 case CA_EVENT_COMPLETE_CWR: 236 tp->snd_cwnd = tp->snd_ssthresh = tcp_westwood_bw_rttmin(sk); 237 break; 238 239 case CA_EVENT_LOSS: 240 tp->snd_ssthresh = tcp_westwood_bw_rttmin(sk); 241 /* Update RTT_min when next ack arrives */ 242 w->reset_rtt_min = 1; 243 break; 244 245 case CA_EVENT_SLOW_ACK: 246 westwood_update_window(sk); 247 w->bk += westwood_acked_count(sk); 248 update_rtt_min(w); 249 break; 250 251 default: 252 /* don't care */ 253 break; 254 } 255 } 256 257 258 /* Extract info for Tcp socket info provided via netlink. */ 259 static void tcp_westwood_info(struct sock *sk, u32 ext, 260 struct sk_buff *skb) 261 { 262 const struct westwood *ca = inet_csk_ca(sk); 263 if (ext & (1 << (INET_DIAG_VEGASINFO - 1))) { 264 struct tcpvegas_info info = { 265 .tcpv_enabled = 1, 266 .tcpv_rtt = jiffies_to_usecs(ca->rtt), 267 .tcpv_minrtt = jiffies_to_usecs(ca->rtt_min), 268 }; 269 270 nla_put(skb, INET_DIAG_VEGASINFO, sizeof(info), &info); 271 } 272 } 273 274 275 static struct tcp_congestion_ops tcp_westwood __read_mostly = { 276 .init = tcp_westwood_init, 277 .ssthresh = tcp_reno_ssthresh, 278 .cong_avoid = tcp_reno_cong_avoid, 279 .cwnd_event = tcp_westwood_event, 280 .get_info = tcp_westwood_info, 281 .pkts_acked = tcp_westwood_pkts_acked, 282 283 .owner = THIS_MODULE, 284 .name = "westwood" 285 }; 286 287 static int __init tcp_westwood_register(void) 288 { 289 BUILD_BUG_ON(sizeof(struct westwood) > ICSK_CA_PRIV_SIZE); 290 return tcp_register_congestion_control(&tcp_westwood); 291 } 292 293 static void __exit tcp_westwood_unregister(void) 294 { 295 tcp_unregister_congestion_control(&tcp_westwood); 296 } 297 298 module_init(tcp_westwood_register); 299 module_exit(tcp_westwood_unregister); 300 301 MODULE_AUTHOR("Stephen Hemminger, Angelo Dell'Aera"); 302 MODULE_LICENSE("GPL"); 303 MODULE_DESCRIPTION("TCP Westwood+"); 304