Blame view
net/ipv4/tcp_vegas.c
9.59 KB
b87d8561d [TCP]: Add TCP Ve... |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 |
/* * TCP Vegas congestion control * * This is based on the congestion detection/avoidance scheme described in * Lawrence S. Brakmo and Larry L. Peterson. * "TCP Vegas: End to end congestion avoidance on a global internet." * IEEE Journal on Selected Areas in Communication, 13(8):1465--1480, * October 1995. Available from: * ftp://ftp.cs.arizona.edu/xkernel/Papers/jsac.ps * * See http://www.cs.arizona.edu/xkernel/ for their implementation. * The main aspects that distinguish this implementation from the * Arizona Vegas implementation are: * o We do not change the loss detection or recovery mechanisms of * Linux in any way. Linux already recovers from losses quite well, * using fine-grained timers, NewReno, and FACK. * o To avoid the performance penalty imposed by increasing cwnd * only every-other RTT during slow start, we increase during * every RTT during slow start, just like Reno. * o Largely to allow continuous cwnd growth during slow start, * we use the rate at which ACKs come back as the "actual" * rate, rather than the rate at which data is sent. * o To speed convergence to the right rate, we set the cwnd * to achieve the right ("actual") rate when we exit slow start. * o To filter out the noise caused by delayed ACKs, we use the * minimum RTT sample observed during the last RTT to calculate * the actual rate. * o When the sender re-starts from idle, it waits until it has * received ACKs for an entire flight of new data before making * a cwnd adjustment decision. The original Vegas implementation * assumed senders never went idle. */ |
b87d8561d [TCP]: Add TCP Ve... |
33 34 35 |
#include <linux/mm.h> #include <linux/module.h> #include <linux/skbuff.h> |
a8c2190ee [INET_DIAG]: Rena... |
36 |
#include <linux/inet_diag.h> |
b87d8561d [TCP]: Add TCP Ve... |
37 38 |
#include <net/tcp.h> |
7752237e9 [TCP] TCP YEAH: U... |
39 |
#include "tcp_vegas.h" |
8d3a564da tcp: tcp_vegas co... |
40 41 42 |
static int alpha = 2; static int beta = 4; static int gamma = 1; |
b87d8561d [TCP]: Add TCP Ve... |
43 44 |
module_param(alpha, int, 0644); |
8d3a564da tcp: tcp_vegas co... |
45 |
MODULE_PARM_DESC(alpha, "lower bound of packets in network"); |
b87d8561d [TCP]: Add TCP Ve... |
46 |
module_param(beta, int, 0644); |
8d3a564da tcp: tcp_vegas co... |
47 |
MODULE_PARM_DESC(beta, "upper bound of packets in network"); |
b87d8561d [TCP]: Add TCP Ve... |
48 49 |
module_param(gamma, int, 0644); MODULE_PARM_DESC(gamma, "limit on increase (scale by 2)"); |
b87d8561d [TCP]: Add TCP Ve... |
50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 |
/* There are several situations when we must "re-start" Vegas: * * o when a connection is established * o after an RTO * o after fast recovery * o when we send a packet and there is no outstanding * unacknowledged data (restarting an idle connection) * * In these circumstances we cannot do a Vegas calculation at the * end of the first RTT, because any calculation we do is using * stale info -- both the saved cwnd and congestion feedback are * stale. * * Instead we must wait until the completion of an RTT during * which we actually receive ACKs. */ |
7752237e9 [TCP] TCP YEAH: U... |
66 |
static void vegas_enable(struct sock *sk) |
b87d8561d [TCP]: Add TCP Ve... |
67 |
{ |
6687e988d [ICSK]: Move TCP ... |
68 69 |
const struct tcp_sock *tp = tcp_sk(sk); struct vegas *vegas = inet_csk_ca(sk); |
b87d8561d [TCP]: Add TCP Ve... |
70 71 72 73 74 75 76 77 78 79 80 81 |
/* Begin taking Vegas samples next time we send something. */ vegas->doing_vegas_now = 1; /* Set the beginning of the next send window. */ vegas->beg_snd_nxt = tp->snd_nxt; vegas->cntRTT = 0; vegas->minRTT = 0x7fffffff; } /* Stop taking Vegas samples for now. */ |
6687e988d [ICSK]: Move TCP ... |
82 |
static inline void vegas_disable(struct sock *sk) |
b87d8561d [TCP]: Add TCP Ve... |
83 |
{ |
6687e988d [ICSK]: Move TCP ... |
84 |
struct vegas *vegas = inet_csk_ca(sk); |
b87d8561d [TCP]: Add TCP Ve... |
85 86 87 |
vegas->doing_vegas_now = 0; } |
7752237e9 [TCP] TCP YEAH: U... |
88 |
void tcp_vegas_init(struct sock *sk) |
b87d8561d [TCP]: Add TCP Ve... |
89 |
{ |
6687e988d [ICSK]: Move TCP ... |
90 |
struct vegas *vegas = inet_csk_ca(sk); |
b87d8561d [TCP]: Add TCP Ve... |
91 92 |
vegas->baseRTT = 0x7fffffff; |
6687e988d [ICSK]: Move TCP ... |
93 |
vegas_enable(sk); |
b87d8561d [TCP]: Add TCP Ve... |
94 |
} |
7752237e9 [TCP] TCP YEAH: U... |
95 |
EXPORT_SYMBOL_GPL(tcp_vegas_init); |
b87d8561d [TCP]: Add TCP Ve... |
96 97 98 99 100 101 102 103 104 |
/* Do RTT sampling needed for Vegas. * Basically we: * o min-filter RTT samples from within an RTT to get the current * propagation delay + queuing delay (we are min-filtering to try to * avoid the effects of delayed ACKs) * o min-filter RTT samples from a much longer window (forever for now) * to find the propagation delay (baseRTT) */ |
30cfd0baf [TCP]: congestion... |
105 |
void tcp_vegas_pkts_acked(struct sock *sk, u32 cnt, s32 rtt_us) |
b87d8561d [TCP]: Add TCP Ve... |
106 |
{ |
6687e988d [ICSK]: Move TCP ... |
107 |
struct vegas *vegas = inet_csk_ca(sk); |
164891aad [TCP]: Congestion... |
108 |
u32 vrtt; |
30cfd0baf [TCP]: congestion... |
109 |
if (rtt_us < 0) |
b9ce204f0 [TCP]: Congestion... |
110 |
return; |
164891aad [TCP]: Congestion... |
111 |
/* Never allow zero rtt or baseRTT */ |
30cfd0baf [TCP]: congestion... |
112 |
vrtt = rtt_us + 1; |
b87d8561d [TCP]: Add TCP Ve... |
113 114 115 116 117 118 119 120 121 122 123 |
/* Filter to find propagation delay: */ if (vrtt < vegas->baseRTT) vegas->baseRTT = vrtt; /* Find the min RTT during the last RTT to find * the current prop. delay + queuing delay: */ vegas->minRTT = min(vegas->minRTT, vrtt); vegas->cntRTT++; } |
7752237e9 [TCP] TCP YEAH: U... |
124 |
EXPORT_SYMBOL_GPL(tcp_vegas_pkts_acked); |
b87d8561d [TCP]: Add TCP Ve... |
125 |
|
7752237e9 [TCP] TCP YEAH: U... |
126 |
void tcp_vegas_state(struct sock *sk, u8 ca_state) |
b87d8561d [TCP]: Add TCP Ve... |
127 128 129 |
{ if (ca_state == TCP_CA_Open) |
6687e988d [ICSK]: Move TCP ... |
130 |
vegas_enable(sk); |
b87d8561d [TCP]: Add TCP Ve... |
131 |
else |
6687e988d [ICSK]: Move TCP ... |
132 |
vegas_disable(sk); |
b87d8561d [TCP]: Add TCP Ve... |
133 |
} |
7752237e9 [TCP] TCP YEAH: U... |
134 |
EXPORT_SYMBOL_GPL(tcp_vegas_state); |
b87d8561d [TCP]: Add TCP Ve... |
135 136 137 138 139 140 141 142 143 144 |
/* * If the connection is idle and we are restarting, * then we don't want to do any Vegas calculations * until we get fresh RTT samples. So when we * restart, we reset our Vegas state to a clean * slate. After we get acks for this flight of * packets, _then_ we can make Vegas calculations * again. */ |
7752237e9 [TCP] TCP YEAH: U... |
145 |
void tcp_vegas_cwnd_event(struct sock *sk, enum tcp_ca_event event) |
b87d8561d [TCP]: Add TCP Ve... |
146 147 148 |
{ if (event == CA_EVENT_CWND_RESTART || event == CA_EVENT_TX_START) |
6687e988d [ICSK]: Move TCP ... |
149 |
tcp_vegas_init(sk); |
b87d8561d [TCP]: Add TCP Ve... |
150 |
} |
7752237e9 [TCP] TCP YEAH: U... |
151 |
EXPORT_SYMBOL_GPL(tcp_vegas_cwnd_event); |
b87d8561d [TCP]: Add TCP Ve... |
152 |
|
c80a5cdfc tcp: tcp_vegas ss... |
153 154 155 156 |
static inline u32 tcp_vegas_ssthresh(struct tcp_sock *tp) { return min(tp->snd_ssthresh, tp->snd_cwnd-1); } |
c3a05c605 [TCP]: Cong.ctrl ... |
157 |
static void tcp_vegas_cong_avoid(struct sock *sk, u32 ack, u32 in_flight) |
b87d8561d [TCP]: Add TCP Ve... |
158 |
{ |
6687e988d [ICSK]: Move TCP ... |
159 160 |
struct tcp_sock *tp = tcp_sk(sk); struct vegas *vegas = inet_csk_ca(sk); |
b87d8561d [TCP]: Add TCP Ve... |
161 |
|
ab59859de net: fix returnin... |
162 163 164 165 |
if (!vegas->doing_vegas_now) { tcp_reno_cong_avoid(sk, ack, in_flight); return; } |
b87d8561d [TCP]: Add TCP Ve... |
166 |
|
b87d8561d [TCP]: Add TCP Ve... |
167 168 |
if (after(ack, vegas->beg_snd_nxt)) { /* Do the Vegas once-per-RTT cwnd adjustment. */ |
b87d8561d [TCP]: Add TCP Ve... |
169 170 171 172 |
/* Save the extent of the current window so we can use this * at the end of the next RTT. */ |
b87d8561d [TCP]: Add TCP Ve... |
173 |
vegas->beg_snd_nxt = tp->snd_nxt; |
b87d8561d [TCP]: Add TCP Ve... |
174 |
|
b87d8561d [TCP]: Add TCP Ve... |
175 176 177 178 179 180 181 182 183 184 185 186 187 |
/* We do the Vegas calculations only if we got enough RTT * samples that we can be reasonably sure that we got * at least one RTT sample that wasn't from a delayed ACK. * If we only had 2 samples total, * then that means we're getting only 1 ACK per RTT, which * means they're almost certainly delayed ACKs. * If we have 3 samples, we should be OK. */ if (vegas->cntRTT <= 2) { /* We don't have enough RTT samples to do the Vegas * calculation, so we'll behave like Reno. */ |
c3a05c605 [TCP]: Cong.ctrl ... |
188 |
tcp_reno_cong_avoid(sk, ack, in_flight); |
b87d8561d [TCP]: Add TCP Ve... |
189 |
} else { |
159131149 tcp: Overflow bug... |
190 191 |
u32 rtt, diff; u64 target_cwnd; |
b87d8561d [TCP]: Add TCP Ve... |
192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 |
/* We have enough RTT samples, so, using the Vegas * algorithm, we determine if we should increase or * decrease cwnd, and by how much. */ /* Pluck out the RTT we are using for the Vegas * calculations. This is the min RTT seen during the * last RTT. Taking the min filters out the effects * of delayed ACKs, at the cost of noticing congestion * a bit later. */ rtt = vegas->minRTT; /* Calculate the cwnd we should have, if we weren't * going too fast. * * This is: * (actual rate in segments) * baseRTT |
b87d8561d [TCP]: Add TCP Ve... |
211 |
*/ |
8d3a564da tcp: tcp_vegas co... |
212 |
target_cwnd = tp->snd_cwnd * vegas->baseRTT / rtt; |
b87d8561d [TCP]: Add TCP Ve... |
213 214 215 216 |
/* Calculate the difference between the window we had, * and the window we would like to have. This quantity * is the "Diff" from the Arizona Vegas papers. |
b87d8561d [TCP]: Add TCP Ve... |
217 |
*/ |
8d3a564da tcp: tcp_vegas co... |
218 |
diff = tp->snd_cwnd * (rtt-vegas->baseRTT) / vegas->baseRTT; |
b87d8561d [TCP]: Add TCP Ve... |
219 |
|
c80a5cdfc tcp: tcp_vegas ss... |
220 |
if (diff > gamma && tp->snd_cwnd <= tp->snd_ssthresh) { |
c940587bf [TCP] vegas: Fix ... |
221 222 223 |
/* Going too fast. Time to slow down * and switch to congestion avoidance. */ |
c940587bf [TCP] vegas: Fix ... |
224 225 226 227 228 229 230 231 |
/* Set cwnd to match the actual rate * exactly: * cwnd = (actual rate) * baseRTT * Then we add 1 because the integer * truncation robs us of full link * utilization. */ |
8d3a564da tcp: tcp_vegas co... |
232 |
tp->snd_cwnd = min(tp->snd_cwnd, (u32)target_cwnd+1); |
c80a5cdfc tcp: tcp_vegas ss... |
233 |
tp->snd_ssthresh = tcp_vegas_ssthresh(tp); |
b87d8561d [TCP]: Add TCP Ve... |
234 |
|
c940587bf [TCP] vegas: Fix ... |
235 236 |
} else if (tp->snd_cwnd <= tp->snd_ssthresh) { /* Slow start. */ |
7faffa1c7 [TCP]: add tcp_sl... |
237 |
tcp_slow_start(tp); |
b87d8561d [TCP]: Add TCP Ve... |
238 239 |
} else { /* Congestion avoidance. */ |
b87d8561d [TCP]: Add TCP Ve... |
240 241 242 243 244 245 246 247 |
/* Figure out where we would like cwnd * to be. */ if (diff > beta) { /* The old window was too fast, so * we slow down. */ |
8d3a564da tcp: tcp_vegas co... |
248 |
tp->snd_cwnd--; |
c80a5cdfc tcp: tcp_vegas ss... |
249 250 |
tp->snd_ssthresh = tcp_vegas_ssthresh(tp); |
b87d8561d [TCP]: Add TCP Ve... |
251 252 253 254 |
} else if (diff < alpha) { /* We don't have enough extra packets * in the network, so speed up. */ |
8d3a564da tcp: tcp_vegas co... |
255 |
tp->snd_cwnd++; |
b87d8561d [TCP]: Add TCP Ve... |
256 257 258 259 |
} else { /* Sending just as fast as we * should be. */ |
b87d8561d [TCP]: Add TCP Ve... |
260 |
} |
b87d8561d [TCP]: Add TCP Ve... |
261 |
} |
b87d8561d [TCP]: Add TCP Ve... |
262 |
|
7faffa1c7 [TCP]: add tcp_sl... |
263 264 265 266 |
if (tp->snd_cwnd < 2) tp->snd_cwnd = 2; else if (tp->snd_cwnd > tp->snd_cwnd_clamp) tp->snd_cwnd = tp->snd_cwnd_clamp; |
a6af2d6ba tcp: tcp_vegas ss... |
267 268 |
tp->snd_ssthresh = tcp_current_ssthresh(sk); |
7faffa1c7 [TCP]: add tcp_sl... |
269 |
} |
b87d8561d [TCP]: Add TCP Ve... |
270 |
|
5b4956138 [TCP] Vegas: stop... |
271 272 273 274 |
/* Wipe the slate clean for the next RTT. */ vegas->cntRTT = 0; vegas->minRTT = 0x7fffffff; } |
74cb87982 [TCP] tcp_vegas: ... |
275 |
/* Use normal slow start */ |
e905a9eda [NET] IPV4: Fix w... |
276 |
else if (tp->snd_cwnd <= tp->snd_ssthresh) |
74cb87982 [TCP] tcp_vegas: ... |
277 |
tcp_slow_start(tp); |
e905a9eda [NET] IPV4: Fix w... |
278 |
|
b87d8561d [TCP]: Add TCP Ve... |
279 280 281 |
} /* Extract info for Tcp socket info provided via netlink. */ |
7752237e9 [TCP] TCP YEAH: U... |
282 |
void tcp_vegas_get_info(struct sock *sk, u32 ext, struct sk_buff *skb) |
b87d8561d [TCP]: Add TCP Ve... |
283 |
{ |
6687e988d [ICSK]: Move TCP ... |
284 |
const struct vegas *ca = inet_csk_ca(sk); |
73c1f4a03 [TCPDIAG]: Just r... |
285 |
if (ext & (1 << (INET_DIAG_VEGASINFO - 1))) { |
e9195d677 [TCP] vegas: Use ... |
286 287 288 289 290 291 292 293 |
struct tcpvegas_info info = { .tcpv_enabled = ca->doing_vegas_now, .tcpv_rttcnt = ca->cntRTT, .tcpv_rtt = ca->baseRTT, .tcpv_minrtt = ca->minRTT, }; nla_put(skb, INET_DIAG_VEGASINFO, sizeof(info), &info); |
b87d8561d [TCP]: Add TCP Ve... |
294 295 |
} } |
7752237e9 [TCP] TCP YEAH: U... |
296 |
EXPORT_SYMBOL_GPL(tcp_vegas_get_info); |
b87d8561d [TCP]: Add TCP Ve... |
297 |
|
a252bebe2 tcp: mark tcp_con... |
298 |
static struct tcp_congestion_ops tcp_vegas __read_mostly = { |
164891aad [TCP]: Congestion... |
299 |
.flags = TCP_CONG_RTT_STAMP, |
b87d8561d [TCP]: Add TCP Ve... |
300 301 302 303 |
.init = tcp_vegas_init, .ssthresh = tcp_reno_ssthresh, .cong_avoid = tcp_vegas_cong_avoid, .min_cwnd = tcp_reno_min_cwnd, |
164891aad [TCP]: Congestion... |
304 |
.pkts_acked = tcp_vegas_pkts_acked, |
b87d8561d [TCP]: Add TCP Ve... |
305 306 307 308 309 310 311 312 313 314 |
.set_state = tcp_vegas_state, .cwnd_event = tcp_vegas_cwnd_event, .get_info = tcp_vegas_get_info, .owner = THIS_MODULE, .name = "vegas", }; static int __init tcp_vegas_register(void) { |
74975d40b [TCP] Congestion ... |
315 |
BUILD_BUG_ON(sizeof(struct vegas) > ICSK_CA_PRIV_SIZE); |
b87d8561d [TCP]: Add TCP Ve... |
316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 |
tcp_register_congestion_control(&tcp_vegas); return 0; } static void __exit tcp_vegas_unregister(void) { tcp_unregister_congestion_control(&tcp_vegas); } module_init(tcp_vegas_register); module_exit(tcp_vegas_unregister); MODULE_AUTHOR("Stephen Hemminger"); MODULE_LICENSE("GPL"); MODULE_DESCRIPTION("TCP Vegas"); |