Release 4.15 net/ipv4/tcp_vegas.c
/*
* 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.
*/
#include <linux/mm.h>
#include <linux/module.h>
#include <linux/skbuff.h>
#include <linux/inet_diag.h>
#include <net/tcp.h>
#include "tcp_vegas.h"
static int alpha = 2;
static int beta = 4;
static int gamma = 1;
module_param(alpha, int, 0644);
MODULE_PARM_DESC(alpha, "lower bound of packets in network");
module_param(beta, int, 0644);
MODULE_PARM_DESC(beta, "upper bound of packets in network");
module_param(gamma, int, 0644);
MODULE_PARM_DESC(gamma, "limit on increase (scale by 2)");
/* 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.
*/
static void vegas_enable(struct sock *sk)
{
const struct tcp_sock *tp = tcp_sk(sk);
struct vegas *vegas = inet_csk_ca(sk);
/* 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;
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
| Stephen Hemminger | 45 | 75.00% | 1 | 50.00% |
| Arnaldo Carvalho de Melo | 15 | 25.00% | 1 | 50.00% |
| Total | 60 | 100.00% | 2 | 100.00% |
/* Stop taking Vegas samples for now. */
static inline void vegas_disable(struct sock *sk)
{
struct vegas *vegas = inet_csk_ca(sk);
vegas->doing_vegas_now = 0;
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
| Stephen Hemminger | 24 | 85.71% | 1 | 50.00% |
| Arnaldo Carvalho de Melo | 4 | 14.29% | 1 | 50.00% |
| Total | 28 | 100.00% | 2 | 100.00% |
void tcp_vegas_init(struct sock *sk)
{
struct vegas *vegas = inet_csk_ca(sk);
vegas->baseRTT = 0x7fffffff;
vegas_enable(sk);
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
| Stephen Hemminger | 26 | 83.87% | 1 | 50.00% |
| Arnaldo Carvalho de Melo | 5 | 16.13% | 1 | 50.00% |
| Total | 31 | 100.00% | 2 | 100.00% |
EXPORT_SYMBOL_GPL(tcp_vegas_init);
/* 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)
*/
void tcp_vegas_pkts_acked(struct sock *sk, const struct ack_sample *sample)
{
struct vegas *vegas = inet_csk_ca(sk);
u32 vrtt;
if (sample->rtt_us < 0)
return;
/* Never allow zero rtt or baseRTT */
vrtt = sample->rtt_us + 1;
/* 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++;
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
| Stephen Hemminger | 64 | 79.01% | 3 | 50.00% |
| Lawrence Brakmo | 9 | 11.11% | 1 | 16.67% |
| Ilpo Järvinen | 4 | 4.94% | 1 | 16.67% |
| Arnaldo Carvalho de Melo | 4 | 4.94% | 1 | 16.67% |
| Total | 81 | 100.00% | 6 | 100.00% |
EXPORT_SYMBOL_GPL(tcp_vegas_pkts_acked);
void tcp_vegas_state(struct sock *sk, u8 ca_state)
{
if (ca_state == TCP_CA_Open)
vegas_enable(sk);
else
vegas_disable(sk);
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
| Stephen Hemminger | 26 | 86.67% | 1 | 50.00% |
| Arnaldo Carvalho de Melo | 4 | 13.33% | 1 | 50.00% |
| Total | 30 | 100.00% | 2 | 100.00% |
EXPORT_SYMBOL_GPL(tcp_vegas_state);
/*
* 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.
*/
void tcp_vegas_cwnd_event(struct sock *sk, enum tcp_ca_event event)
{
if (event == CA_EVENT_CWND_RESTART ||
event == CA_EVENT_TX_START)
tcp_vegas_init(sk);
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
| Stephen Hemminger | 26 | 89.66% | 1 | 50.00% |
| Arnaldo Carvalho de Melo | 3 | 10.34% | 1 | 50.00% |
| Total | 29 | 100.00% | 2 | 100.00% |
EXPORT_SYMBOL_GPL(tcp_vegas_cwnd_event);
static inline u32 tcp_vegas_ssthresh(struct tcp_sock *tp)
{
return min(tp->snd_ssthresh, tp->snd_cwnd);
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
| Doug Leith | 24 | 100.00% | 1 | 100.00% |
| Total | 24 | 100.00% | 1 | 100.00% |
static void tcp_vegas_cong_avoid(struct sock *sk, u32 ack, u32 acked)
{
struct tcp_sock *tp = tcp_sk(sk);
struct vegas *vegas = inet_csk_ca(sk);
if (!vegas->doing_vegas_now) {
tcp_reno_cong_avoid(sk, ack, acked);
return;
}
if (after(ack, vegas->beg_snd_nxt)) {
/* Do the Vegas once-per-RTT cwnd adjustment. */
/* Save the extent of the current window so we can use this
* at the end of the next RTT.
*/
vegas->beg_snd_nxt = tp->snd_nxt;
/* 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.
*/
tcp_reno_cong_avoid(sk, ack, acked);
} else {
u32 rtt, diff;
u64 target_cwnd;
/* 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
*/
target_cwnd = (u64)tp->snd_cwnd * vegas->baseRTT;
do_div(target_cwnd, rtt);
/* 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.
*/
diff = tp->snd_cwnd * (rtt-vegas->baseRTT) / vegas->baseRTT;
if (diff > gamma && tcp_in_slow_start(tp)) {
/* Going too fast. Time to slow down
* and switch to congestion avoidance.
*/
/* 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.
*/
tp->snd_cwnd = min(tp->snd_cwnd, (u32)target_cwnd+1);
tp->snd_ssthresh = tcp_vegas_ssthresh(tp);
} else if (tcp_in_slow_start(tp)) {
/* Slow start. */
tcp_slow_start(tp, acked);
} else {
/* Congestion avoidance. */
/* Figure out where we would like cwnd
* to be.
*/
if (diff > beta) {
/* The old window was too fast, so
* we slow down.
*/
tp->snd_cwnd--;
tp->snd_ssthresh
= tcp_vegas_ssthresh(tp);
} else if (diff < alpha) {
/* We don't have enough extra packets
* in the network, so speed up.
*/
tp->snd_cwnd++;
} else {
/* Sending just as fast as we
* should be.
*/
}
}
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;
tp->snd_ssthresh = tcp_current_ssthresh(sk);
}
/* Wipe the slate clean for the next RTT. */
vegas->cntRTT = 0;
vegas->minRTT = 0x7fffffff;
}
/* Use normal slow start */
else if (tcp_in_slow_start(tp))
tcp_slow_start(tp, acked);
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
| Stephen Hemminger | 210 | 62.31% | 2 | 14.29% |
| Doug Leith | 53 | 15.73% | 3 | 21.43% |
| Yuchung Cheng | 16 | 4.75% | 2 | 14.29% |
| Arnaldo Carvalho de Melo | 15 | 4.45% | 1 | 7.14% |
| Xiaoliang (David) Wei | 13 | 3.86% | 1 | 7.14% |
| Thomas Young | 12 | 3.56% | 2 | 14.29% |
| Christoph Paasch | 9 | 2.67% | 1 | 7.14% |
| Lachlan Andrew | 5 | 1.48% | 1 | 7.14% |
| Harvey Harrison | 4 | 1.19% | 1 | 7.14% |
| Total | 337 | 100.00% | 14 | 100.00% |
/* Extract info for Tcp socket info provided via netlink. */
size_t tcp_vegas_get_info(struct sock *sk, u32 ext, int *attr,
union tcp_cc_info *info)
{
const struct vegas *ca = inet_csk_ca(sk);
if (ext & (1 << (INET_DIAG_VEGASINFO - 1))) {
info->vegas.tcpv_enabled = ca->doing_vegas_now,
info->vegas.tcpv_rttcnt = ca->cntRTT,
info->vegas.tcpv_rtt = ca->baseRTT,
info->vegas.tcpv_minrtt = ca->minRTT,
*attr = INET_DIAG_VEGASINFO;
return sizeof(struct tcpvegas_info);
}
return 0;
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
| Stephen Hemminger | 57 | 54.81% | 1 | 16.67% |
| Eric Dumazet | 34 | 32.69% | 2 | 33.33% |
| Thomas Graf | 8 | 7.69% | 1 | 16.67% |
| Arnaldo Carvalho de Melo | 5 | 4.81% | 2 | 33.33% |
| Total | 104 | 100.00% | 6 | 100.00% |
EXPORT_SYMBOL_GPL(tcp_vegas_get_info);
static struct tcp_congestion_ops tcp_vegas __read_mostly = {
.init = tcp_vegas_init,
.ssthresh = tcp_reno_ssthresh,
.undo_cwnd = tcp_reno_undo_cwnd,
.cong_avoid = tcp_vegas_cong_avoid,
.pkts_acked = tcp_vegas_pkts_acked,
.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)
{
BUILD_BUG_ON(sizeof(struct vegas) > ICSK_CA_PRIV_SIZE);
tcp_register_congestion_control(&tcp_vegas);
return 0;
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
| Stephen Hemminger | 27 | 93.10% | 1 | 33.33% |
| Arnaldo Carvalho de Melo | 1 | 3.45% | 1 | 33.33% |
| Alexey Dobriyan | 1 | 3.45% | 1 | 33.33% |
| Total | 29 | 100.00% | 3 | 100.00% |
static void __exit tcp_vegas_unregister(void)
{
tcp_unregister_congestion_control(&tcp_vegas);
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
| Stephen Hemminger | 15 | 100.00% | 1 | 100.00% |
| Total | 15 | 100.00% | 1 | 100.00% |
module_init(tcp_vegas_register);
module_exit(tcp_vegas_unregister);
MODULE_AUTHOR("Stephen Hemminger");
MODULE_LICENSE("GPL");
MODULE_DESCRIPTION("TCP Vegas");
Overall Contributors
| Person | Tokens | Prop | Commits | CommitProp |
| Stephen Hemminger | 708 | 73.29% | 6 | 21.43% |
| Doug Leith | 79 | 8.18% | 3 | 10.71% |
| Arnaldo Carvalho de Melo | 57 | 5.90% | 3 | 10.71% |
| Eric Dumazet | 34 | 3.52% | 2 | 7.14% |
| Yuchung Cheng | 16 | 1.66% | 2 | 7.14% |
| Xiaoliang (David) Wei | 13 | 1.35% | 1 | 3.57% |
| Thomas Young | 12 | 1.24% | 2 | 7.14% |
| Lawrence Brakmo | 9 | 0.93% | 1 | 3.57% |
| Christoph Paasch | 9 | 0.93% | 1 | 3.57% |
| Thomas Graf | 8 | 0.83% | 1 | 3.57% |
| Florian Westphal | 5 | 0.52% | 1 | 3.57% |
| Lachlan Andrew | 5 | 0.52% | 1 | 3.57% |
| Harvey Harrison | 4 | 0.41% | 1 | 3.57% |
| Ilpo Järvinen | 4 | 0.41% | 1 | 3.57% |
| David S. Miller | 2 | 0.21% | 1 | 3.57% |
| Alexey Dobriyan | 1 | 0.10% | 1 | 3.57% |
| Total | 966 | 100.00% | 28 | 100.00% |
Information contained on this website is for historical information purposes only and does not indicate or represent copyright ownership.