Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit 7752237e authored by Stephen Hemminger's avatar Stephen Hemminger Committed by David S. Miller
Browse files

[TCP] TCP YEAH: Use vegas dont copy it. 



Rather than using a copy of vegas code, the YEAH code should just have
it exported so there is common code.

Signed-off-by: default avatarStephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parent 164891aa

net/ipv4/tcp_vegas.c

+13 −18
Original line number Diff line number Diff line
@@ -38,6 +38,8 @@ 


#include <net/tcp.h>



#include "tcp_vegas.h"



/* Default values of the Vegas variables, in fixed-point representation

 * with V_PARAM_SHIFT bits to the right of the binary point.

 */
@@ -54,17 +56,6 @@ module_param(gamma, int, 0644); 
MODULE_PARM_DESC(gamma, "limit on increase (scale by 2)");





/* Vegas variables */

struct 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) */

};



/* There are several situations when we must "re-start" Vegas:

 *

 *  o when a connection is established
@@ -81,7 +72,7 @@ struct vegas { 
 * Instead we must wait until the completion of an RTT during

 * which we actually receive ACKs.

 */

static inline void vegas_enable(struct sock *sk)

static void vegas_enable(struct sock *sk)

{

	const struct tcp_sock *tp = tcp_sk(sk);

	struct vegas *vegas = inet_csk_ca(sk);
@@ -104,13 +95,14 @@ static inline void vegas_disable(struct sock *sk) 
	vegas->doing_vegas_now = 0;

}



static void tcp_vegas_init(struct sock *sk)

void tcp_vegas_init(struct sock *sk)

{

	struct vegas *vegas = inet_csk_ca(sk);



	vegas->baseRTT = 0x7fffffff;

	vegas_enable(sk);

}

EXPORT_SYMBOL_GPL(tcp_vegas_init);



/* Do RTT sampling needed for Vegas.

 * Basically we:
@@ -120,7 +112,7 @@ static void tcp_vegas_init(struct sock *sk) 
 *   o min-filter RTT samples from a much longer window (forever for now)

 *     to find the propagation delay (baseRTT)

 */

static void tcp_vegas_pkts_acked(struct sock *sk, u32 cnt, ktime_t last)

void tcp_vegas_pkts_acked(struct sock *sk, u32 cnt, ktime_t last)

{

	struct vegas *vegas = inet_csk_ca(sk);

	u32 vrtt;
@@ -138,8 +130,9 @@ static void tcp_vegas_pkts_acked(struct sock *sk, u32 cnt, ktime_t last) 
	vegas->minRTT = min(vegas->minRTT, vrtt);

	vegas->cntRTT++;

}

EXPORT_SYMBOL_GPL(tcp_vegas_pkts_acked);



static void tcp_vegas_state(struct sock *sk, u8 ca_state)

void tcp_vegas_state(struct sock *sk, u8 ca_state)

{



	if (ca_state == TCP_CA_Open)
@@ -147,6 +140,7 @@ static void tcp_vegas_state(struct sock *sk, u8 ca_state) 
	else

		vegas_disable(sk);

}

EXPORT_SYMBOL_GPL(tcp_vegas_state);



/*

 * If the connection is idle and we are restarting,
@@ -157,12 +151,13 @@ static void tcp_vegas_state(struct sock *sk, u8 ca_state) 
 * packets, _then_ we can make Vegas calculations

 * again.

 */

static void tcp_vegas_cwnd_event(struct sock *sk, enum tcp_ca_event event)

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);

}

EXPORT_SYMBOL_GPL(tcp_vegas_cwnd_event);



static void tcp_vegas_cong_avoid(struct sock *sk, u32 ack,

				 u32 seq_rtt, u32 in_flight, int flag)
@@ -339,8 +334,7 @@ static void tcp_vegas_cong_avoid(struct sock *sk, u32 ack, 
}



/* Extract info for Tcp socket info provided via netlink. */

static void tcp_vegas_get_info(struct sock *sk, u32 ext,

			       struct sk_buff *skb)

void tcp_vegas_get_info(struct sock *sk, u32 ext, struct sk_buff *skb)

{

	const struct vegas *ca = inet_csk_ca(sk);

	if (ext & (1 << (INET_DIAG_VEGASINFO - 1))) {
@@ -354,6 +348,7 @@ static void tcp_vegas_get_info(struct sock *sk, u32 ext, 
		nla_put(skb, INET_DIAG_VEGASINFO, sizeof(info), &info);

	}

}

EXPORT_SYMBOL_GPL(tcp_vegas_get_info);



static struct tcp_congestion_ops tcp_vegas = {

	.flags		= TCP_CONG_RTT_STAMP,

net/ipv4/tcp_vegas.h

0 → 100644
+24 −0
Original line number Diff line number Diff line 
/*

 * TCP Vegas congestion control interface

 */

#ifndef __TCP_VEGAS_H

#define __TCP_VEGAS_H 1



/* Vegas variables */

struct 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) */

};



extern void tcp_vegas_init(struct sock *sk);

extern void tcp_vegas_state(struct sock *sk, u8 ca_state);

extern void tcp_vegas_pkts_acked(struct sock *sk, u32 cnt, ktime_t last);

extern void tcp_vegas_cwnd_event(struct sock *sk, enum tcp_ca_event event);

extern void tcp_vegas_get_info(struct sock *sk, u32 ext, struct sk_buff *skb);



#endif	/* __TCP_VEGAS_H */

net/ipv4/tcp_yeah.c

+24 −29
Original line number Diff line number Diff line
@@ -6,13 +6,14 @@ 
 *    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
@@ -26,14 +27,7 @@ 


/* 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;
@@ -84,9 +78,10 @@ static void tcp_yeah_cong_avoid(struct sock *sk, u32 ack, 
	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;
@@ -110,19 +105,19 @@ static void tcp_yeah_cong_avoid(struct sock *sk, u32 ack, 
		}

	}



	/* 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.

	 *
@@ -132,7 +127,7 @@ static void tcp_yeah_cong_avoid(struct sock *sk, u32 ack, 
	 * 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
@@ -143,7 +138,7 @@ static void tcp_yeah_cong_avoid(struct sock *sk, u32 ack, 
		 * If  we have 3 samples, we should be OK.

		 */



		if (yeah->cntRTT > 2) {

		if (yeah->vegas.cntRTT > 2) {

			u32 rtt, queue;

			u64 bw;
@@ -158,18 +153,18 @@ static void tcp_yeah_cong_avoid(struct sock *sk, u32 ack, 
			 * of delayed ACKs, at the cost of noticing congestion

			 * a bit later.

			 */

			rtt = yeah->minRTT;

			rtt = yeah->vegas.minRTT;



			/* Compute excess number of packets above bandwidth

			 * Avoid doing full 64 bit divide.

			 */

			bw = tp->snd_cwnd;

			bw *= rtt - yeah->baseRTT;

			bw *= rtt - yeah->vegas.baseRTT;

			do_div(bw, rtt);

			queue = bw;



			if (queue > TCP_YEAH_ALPHA ||

			    rtt - yeah->baseRTT > (yeah->baseRTT / TCP_YEAH_PHY)) {

			    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 ,
@@ -208,13 +203,13 @@ static void tcp_yeah_cong_avoid(struct sock *sk, u32 ack, 
		/* 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;

	}

}

net/ipv4/tcp_yeah.h

+0 −131
Original line number Diff line number Diff line
@@ -5,134 +5,3 @@ 
#include <asm/div64.h>



#include <net/tcp.h>
 


/* Vegas variables */

struct 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) */

};



/* 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 inline 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;

}



/* 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;

}



static void tcp_vegas_init(struct sock *sk)

{

	struct vegas *vegas = inet_csk_ca(sk);



	vegas->baseRTT = 0x7fffffff;

	vegas_enable(sk);

}



static void tcp_vegas_state(struct sock *sk, u8 ca_state)

{



	if (ca_state == TCP_CA_Open)

		vegas_enable(sk);

	else

		vegas_disable(sk);

}



/* 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)

 */

static void tcp_vegas_pkts_acked(struct sock *sk, u32 cnt, ktime_t last)

{

	struct vegas *vegas = inet_csk_ca(sk);

	u32 vrtt;



	/* Never allow zero rtt or baseRTT */

	vrtt = (ktime_to_ns(net_timedelta(last)) / NSEC_PER_USEC) + 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++;

}



/*

 * 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.

 */

static 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);

}



/* Extract info for Tcp socket info provided via netlink. */

static void tcp_vegas_get_info(struct sock *sk, u32 ext,

			       struct sk_buff *skb)

{

	const struct vegas *ca = inet_csk_ca(sk);

	if (ext & (1 << (INET_DIAG_VEGASINFO - 1))) {

		struct tcpvegas_info *info;



		info = RTA_DATA(__RTA_PUT(skb, INET_DIAG_VEGASINFO,

					  sizeof(*info)));



		info->tcpv_enabled = ca->doing_vegas_now;

		info->tcpv_rttcnt = ca->cntRTT;

		info->tcpv_rtt = ca->baseRTT;

		info->tcpv_minrtt = ca->minRTT;

	rtattr_failure:	;

	}

}