* http://wil.cs.caltech.edu/pfldnet2007/paper/YeAH_TCP.pdf
*
*/
+#include <linux/mm.h>
+#include <linux/module.h>
+#include <linux/skbuff.h>
+#include <linux/inet_diag.h>
-#include "tcp_yeah.h"
+#include <net/tcp.h>
-/* Default values of the Vegas variables, in fixed-point representation
- * with V_PARAM_SHIFT bits to the right of the binary point.
- */
-#define V_PARAM_SHIFT 1
+#include "tcp_vegas.h"
#define TCP_YEAH_ALPHA 80 //lin number of packets queued at the bottleneck
#define TCP_YEAH_GAMMA 1 //lin fraction of queue to be removed per rtt
/* YeAH variables */
struct yeah {
- /* Vegas */
- u32 beg_snd_nxt; /* right edge during last RTT */
- u32 beg_snd_una; /* left edge during last RTT */
- u32 beg_snd_cwnd; /* saves the size of the cwnd */
- u8 doing_vegas_now;/* if true, do vegas for this RTT */
- u16 cntRTT; /* # of RTTs measured within last RTT */
- u32 minRTT; /* min of RTTs measured within last RTT (in usec) */
- u32 baseRTT; /* the min of all Vegas RTT measurements seen (in usec) */
+ struct vegas vegas; /* must be first */
/* YeAH */
u32 lastQ;
}
-static void tcp_yeah_pkts_acked(struct sock *sk, u32 pkts_acked)
+static void tcp_yeah_pkts_acked(struct sock *sk, u32 pkts_acked, s32 rtt_us)
{
const struct inet_connection_sock *icsk = inet_csk(sk);
struct yeah *yeah = inet_csk_ca(sk);
if (icsk->icsk_ca_state == TCP_CA_Open)
yeah->pkts_acked = pkts_acked;
-}
-
-/* 64bit divisor, dividend and result. dynamic precision */
-static inline u64 div64_64(u64 dividend, u64 divisor)
-{
- u32 d = divisor;
-
- if (divisor > 0xffffffffULL) {
- unsigned int shift = fls(divisor >> 32);
- d = divisor >> shift;
- dividend >>= shift;
- }
-
- /* avoid 64 bit division if possible */
- if (dividend >> 32)
- do_div(dividend, d);
- else
- dividend = (u32) dividend / d;
-
- return dividend;
+ tcp_vegas_pkts_acked(sk, pkts_acked, rtt_us);
}
-static void tcp_yeah_cong_avoid(struct sock *sk, u32 ack,
- u32 seq_rtt, u32 in_flight, int flag)
+static void tcp_yeah_cong_avoid(struct sock *sk, u32 ack, u32 in_flight)
{
struct tcp_sock *tp = tcp_sk(sk);
struct yeah *yeah = inet_csk_ca(sk);
if (!tcp_is_cwnd_limited(sk, in_flight))
return;
- if (tp->snd_cwnd <= tp->snd_ssthresh) {
+ if (tp->snd_cwnd <= tp->snd_ssthresh)
tcp_slow_start(tp);
- } else if (!yeah->doing_reno_now) {
+
+ else if (!yeah->doing_reno_now) {
/* Scalable */
- tp->snd_cwnd_cnt+=yeah->pkts_acked;
+ tp->snd_cwnd_cnt += yeah->pkts_acked;
if (tp->snd_cwnd_cnt > min(tp->snd_cwnd, TCP_SCALABLE_AI_CNT)){
if (tp->snd_cwnd < tp->snd_cwnd_clamp)
tp->snd_cwnd++;
} else {
/* Reno */
-
- if (tp->snd_cwnd_cnt < tp->snd_cwnd)
- tp->snd_cwnd_cnt++;
-
- if (tp->snd_cwnd_cnt >= tp->snd_cwnd) {
- tp->snd_cwnd++;
- tp->snd_cwnd_cnt = 0;
- }
+ tcp_cong_avoid_ai(tp, tp->snd_cwnd);
}
- /* The key players are v_beg_snd_una and v_beg_snd_nxt.
+ /* The key players are v_vegas.beg_snd_una and v_beg_snd_nxt.
*
* These are so named because they represent the approximate values
* of snd_una and snd_nxt at the beginning of the current RTT. More
* precisely, they represent the amount of data sent during the RTT.
* At the end of the RTT, when we receive an ACK for v_beg_snd_nxt,
- * we will calculate that (v_beg_snd_nxt - v_beg_snd_una) outstanding
+ * we will calculate that (v_beg_snd_nxt - v_vegas.beg_snd_una) outstanding
* bytes of data have been ACKed during the course of the RTT, giving
* an "actual" rate of:
*
- * (v_beg_snd_nxt - v_beg_snd_una) / (rtt duration)
+ * (v_beg_snd_nxt - v_vegas.beg_snd_una) / (rtt duration)
*
- * Unfortunately, v_beg_snd_una is not exactly equal to snd_una,
+ * Unfortunately, v_vegas.beg_snd_una is not exactly equal to snd_una,
* because delayed ACKs can cover more than one segment, so they
* don't line up yeahly with the boundaries of RTTs.
*
* So we keep track of our cwnd separately, in v_beg_snd_cwnd.
*/
- if (after(ack, yeah->beg_snd_nxt)) {
+ if (after(ack, yeah->vegas.beg_snd_nxt)) {
/* We do the Vegas calculations only if we got enough RTT
* samples that we can be reasonably sure that we got
* If we have 3 samples, we should be OK.
*/
- if (yeah->cntRTT > 2) {
- u32 rtt;
- u32 queue, maxqueue;
+ if (yeah->vegas.cntRTT > 2) {
+ u32 rtt, queue;
+ u64 bw;
/* We have enough RTT samples, so, using the Vegas
* algorithm, we determine if we should increase or
* of delayed ACKs, at the cost of noticing congestion
* a bit later.
*/
- rtt = yeah->minRTT;
-
- queue = (u32)div64_64((u64)tp->snd_cwnd * (rtt - yeah->baseRTT), rtt);
-
- maxqueue = TCP_YEAH_ALPHA;
+ rtt = yeah->vegas.minRTT;
- if (queue > maxqueue ||
- rtt - yeah->baseRTT > (yeah->baseRTT / TCP_YEAH_PHY)) {
-
- if (queue > maxqueue && tp->snd_cwnd > yeah->reno_count) {
- u32 reduction = min( queue / TCP_YEAH_GAMMA ,
- tp->snd_cwnd >> TCP_YEAH_EPSILON );
+ /* Compute excess number of packets above bandwidth
+ * Avoid doing full 64 bit divide.
+ */
+ bw = tp->snd_cwnd;
+ bw *= rtt - yeah->vegas.baseRTT;
+ do_div(bw, rtt);
+ queue = bw;
+
+ if (queue > TCP_YEAH_ALPHA ||
+ rtt - yeah->vegas.baseRTT > (yeah->vegas.baseRTT / TCP_YEAH_PHY)) {
+ if (queue > TCP_YEAH_ALPHA &&
+ tp->snd_cwnd > yeah->reno_count) {
+ u32 reduction = min(queue / TCP_YEAH_GAMMA ,
+ tp->snd_cwnd >> TCP_YEAH_EPSILON);
tp->snd_cwnd -= reduction;
- tp->snd_cwnd = max( tp->snd_cwnd, yeah->reno_count);
+ tp->snd_cwnd = max(tp->snd_cwnd,
+ yeah->reno_count);
tp->snd_ssthresh = tp->snd_cwnd;
- }
+ }
if (yeah->reno_count <= 2)
- yeah->reno_count = max( tp->snd_cwnd>>1, 2U);
+ yeah->reno_count = max(tp->snd_cwnd>>1, 2U);
else
yeah->reno_count++;
- yeah->doing_reno_now =
- min_t( u32, yeah->doing_reno_now + 1 , 0xffffff);
-
+ yeah->doing_reno_now = min(yeah->doing_reno_now + 1,
+ 0xffffffU);
} else {
yeah->fast_count++;
/* Save the extent of the current window so we can use this
* at the end of the next RTT.
*/
- yeah->beg_snd_una = yeah->beg_snd_nxt;
- yeah->beg_snd_nxt = tp->snd_nxt;
- yeah->beg_snd_cwnd = tp->snd_cwnd;
+ yeah->vegas.beg_snd_una = yeah->vegas.beg_snd_nxt;
+ yeah->vegas.beg_snd_nxt = tp->snd_nxt;
+ yeah->vegas.beg_snd_cwnd = tp->snd_cwnd;
/* Wipe the slate clean for the next RTT. */
- yeah->cntRTT = 0;
- yeah->minRTT = 0x7fffffff;
+ yeah->vegas.cntRTT = 0;
+ yeah->vegas.minRTT = 0x7fffffff;
}
}
reduction = max( reduction, tp->snd_cwnd >> TCP_YEAH_DELTA);
} else
- reduction = max(tp->snd_cwnd>>1,2U);
+ reduction = max(tp->snd_cwnd>>1, 2U);
yeah->fast_count = 0;
yeah->reno_count = max(yeah->reno_count>>1, 2U);
}
static struct tcp_congestion_ops tcp_yeah = {
+ .flags = TCP_CONG_RTT_STAMP,
.init = tcp_yeah_init,
.ssthresh = tcp_yeah_ssthresh,
.cong_avoid = tcp_yeah_cong_avoid,
.min_cwnd = tcp_reno_min_cwnd,
- .rtt_sample = tcp_vegas_rtt_calc,
.set_state = tcp_vegas_state,
.cwnd_event = tcp_vegas_cwnd_event,
.get_info = tcp_vegas_get_info,
git://ftp.safe.ca / safe / jmp / linux-2.6 / blobdiff