ccid3.c

linux 内核源代码

/*
 *  net/dccp/ccids/ccid3.c
 *
 *  Copyright (c) 2005-7 The University of Waikato, Hamilton, New Zealand.
 *  Copyright (c) 2005-7 Ian McDonald <[e-mail]>
 *
 *  An implementation of the DCCP protocol
 *
 *  This code has been developed by the University of Waikato WAND
 *  research group. For further information please see http://www.wand.net.nz/
 *
 *  This code also uses code from Lulea University, rereleased as GPL by its
 *  authors:
 *  Copyright (c) 2003 Nils-Erik Mattsson, Joacim Haggmark, Magnus Erixzon
 *
 *  Changes to meet Linux coding standards, to make it meet latest ccid3 draft
 *  and to make it work as a loadable module in the DCCP stack written by
 *  Arnaldo Carvalho de Melo <[e-mail]>.
 *
 *  Copyright (c) 2005 Arnaldo Carvalho de Melo <[e-mail]>
 *
 *  This program is free software; you can redistribute it and/or modify
 *  it under the terms of the GNU General Public License as published by
 *  the Free Software Foundation; either version 2 of the License, or
 *  (at your option) any later version.
 *
 *  This program is distributed in the hope that it will be useful,
 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *  GNU General Public License for more details.
 *
 *  You should have received a copy of the GNU General Public License
 *  along with this program; if not, write to the Free Software
 *  Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 */
#include "../ccid.h"
#include "../dccp.h"
#include "lib/packet_history.h"
#include "lib/loss_interval.h"
#include "lib/tfrc.h"
#include "ccid3.h"

#include <asm/unaligned.h>

#ifdef CONFIG_IP_DCCP_CCID3_DEBUG
static int ccid3_debug;
#define ccid3_pr_debug(format, a...)	DCCP_PR_DEBUG(ccid3_debug, format, ##a)
#else
#define ccid3_pr_debug(format, a...)
#endif

static struct dccp_tx_hist *ccid3_tx_hist;
static struct dccp_rx_hist *ccid3_rx_hist;

/*
 *	Transmitter Half-Connection Routines
 */
#ifdef CONFIG_IP_DCCP_CCID3_DEBUG
static const char *ccid3_tx_state_name(enum ccid3_hc_tx_states state)
{
	static char *ccid3_state_names[] = {
	[TFRC_SSTATE_NO_SENT]  = "NO_SENT",
	[TFRC_SSTATE_NO_FBACK] = "NO_FBACK",
	[TFRC_SSTATE_FBACK]    = "FBACK",
	[TFRC_SSTATE_TERM]     = "TERM",
	};

	return ccid3_state_names[state];
}
#endif

static void ccid3_hc_tx_set_state(struct sock *sk,
				  enum ccid3_hc_tx_states state)
{
	struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
	enum ccid3_hc_tx_states oldstate = hctx->ccid3hctx_state;

	ccid3_pr_debug("%s(%p) %-8.8s -> %s\n",
		       dccp_role(sk), sk, ccid3_tx_state_name(oldstate),
		       ccid3_tx_state_name(state));
	WARN_ON(state == oldstate);
	hctx->ccid3hctx_state = state;
}

/*
 * Compute the initial sending rate X_init according to RFC 3390:
 *	w_init   =    min(4 * MSS, max(2 * MSS, 4380 bytes))
 *	X_init   =    w_init / RTT
 * For consistency with other parts of the code, X_init is scaled by 2^6.
 */
static inline u64 rfc3390_initial_rate(struct sock *sk)
{
	const struct dccp_sock *dp = dccp_sk(sk);
	const __u32 w_init = min(4 * dp->dccps_mss_cache,
				 max(2 * dp->dccps_mss_cache, 4380U));

	return scaled_div(w_init << 6, ccid3_hc_tx_sk(sk)->ccid3hctx_rtt);
}

/*
 * Recalculate t_ipi and delta (should be called whenever X changes)
 */
static inline void ccid3_update_send_interval(struct ccid3_hc_tx_sock *hctx)
{
	/* Calculate new t_ipi = s / X_inst (X_inst is in 64 * bytes/second) */
	hctx->ccid3hctx_t_ipi = scaled_div32(((u64)hctx->ccid3hctx_s) << 6,
					     hctx->ccid3hctx_x);

	/* Calculate new delta by delta = min(t_ipi / 2, t_gran / 2) */
	hctx->ccid3hctx_delta = min_t(u32, hctx->ccid3hctx_t_ipi / 2,
					   TFRC_OPSYS_HALF_TIME_GRAN);

	ccid3_pr_debug("t_ipi=%u, delta=%u, s=%u, X=%u\n",
		       hctx->ccid3hctx_t_ipi, hctx->ccid3hctx_delta,
		       hctx->ccid3hctx_s, (unsigned)(hctx->ccid3hctx_x >> 6));

}

/**
 * ccid3_hc_tx_update_x  -  Update allowed sending rate X
 * @stamp: most recent time if available - can be left NULL.
 * This function tracks draft rfc3448bis, check there for latest details.
 *
 * Note: X and X_recv are both stored in units of 64 * bytes/second, to support
 *       fine-grained resolution of sending rates. This requires scaling by 2^6
 *       throughout the code. Only X_calc is unscaled (in bytes/second).
 *
 */
static void ccid3_hc_tx_update_x(struct sock *sk, ktime_t *stamp)

{
	struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
	__u64 min_rate = 2 * hctx->ccid3hctx_x_recv;
	const  __u64 old_x = hctx->ccid3hctx_x;
	ktime_t now = stamp? *stamp : ktime_get_real();

	/*
	 * Handle IDLE periods: do not reduce below RFC3390 initial sending rate
	 * when idling [RFC 4342, 5.1]. See also draft-ietf-dccp-rfc3448bis.
	 * For consistency with X and X_recv, min_rate is also scaled by 2^6.
	 */
	if (unlikely(hctx->ccid3hctx_idle)) {
		min_rate = rfc3390_initial_rate(sk);
		min_rate = max(min_rate, 2 * hctx->ccid3hctx_x_recv);
	}

	if (hctx->ccid3hctx_p > 0) {

		hctx->ccid3hctx_x = min(((__u64)hctx->ccid3hctx_x_calc) << 6,
					min_rate);
		hctx->ccid3hctx_x = max(hctx->ccid3hctx_x,
					(((__u64)hctx->ccid3hctx_s) << 6) /
								TFRC_T_MBI);

	} else if (ktime_us_delta(now, hctx->ccid3hctx_t_ld)
				- (s64)hctx->ccid3hctx_rtt >= 0) {

		hctx->ccid3hctx_x =
			max(min(2 * hctx->ccid3hctx_x, min_rate),
			    scaled_div(((__u64)hctx->ccid3hctx_s) << 6,
				       hctx->ccid3hctx_rtt));
		hctx->ccid3hctx_t_ld = now;
	}

	if (hctx->ccid3hctx_x != old_x) {
		ccid3_pr_debug("X_prev=%u, X_now=%u, X_calc=%u, "
			       "X_recv=%u\n", (unsigned)(old_x >> 6),
			       (unsigned)(hctx->ccid3hctx_x >> 6),
			       hctx->ccid3hctx_x_calc,
			       (unsigned)(hctx->ccid3hctx_x_recv >> 6));

		ccid3_update_send_interval(hctx);
	}
}

/*
 *	Track the mean packet size `s' (cf. RFC 4342, 5.3 and  RFC 3448, 4.1)
 *	@len: DCCP packet payload size in bytes
 */
static inline void ccid3_hc_tx_update_s(struct ccid3_hc_tx_sock *hctx, int len)
{
	const u16 old_s = hctx->ccid3hctx_s;

	hctx->ccid3hctx_s = old_s == 0 ? len : (9 * old_s + len) / 10;

	if (hctx->ccid3hctx_s != old_s)
		ccid3_update_send_interval(hctx);
}

/*
 *	Update Window Counter using the algorithm from [RFC 4342, 8.1].
 *	The algorithm is not applicable if RTT < 4 microseconds.
 */
static inline void ccid3_hc_tx_update_win_count(struct ccid3_hc_tx_sock *hctx,
						ktime_t now)
{
	u32 quarter_rtts;

	if (unlikely(hctx->ccid3hctx_rtt < 4))	/* avoid divide-by-zero */
		return;

	quarter_rtts = ktime_us_delta(now, hctx->ccid3hctx_t_last_win_count);
	quarter_rtts /= hctx->ccid3hctx_rtt / 4;

	if (quarter_rtts > 0) {
		hctx->ccid3hctx_t_last_win_count = now;
		hctx->ccid3hctx_last_win_count	+= min_t(u32, quarter_rtts, 5);
		hctx->ccid3hctx_last_win_count	&= 0xF;		/* mod 16 */
	}
}

static void ccid3_hc_tx_no_feedback_timer(unsigned long data)
{
	struct sock *sk = (struct sock *)data;
	struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
	unsigned long t_nfb = USEC_PER_SEC / 5;

	bh_lock_sock(sk);
	if (sock_owned_by_user(sk)) {
		/* Try again later. */
		/* XXX: set some sensible MIB */
		goto restart_timer;
	}

	ccid3_pr_debug("%s(%p, state=%s) - entry \n", dccp_role(sk), sk,
		       ccid3_tx_state_name(hctx->ccid3hctx_state));

	hctx->ccid3hctx_idle = 1;

	switch (hctx->ccid3hctx_state) {
	case TFRC_SSTATE_NO_FBACK:
		/* RFC 3448, 4.4: Halve send rate directly */
		hctx->ccid3hctx_x = max(hctx->ccid3hctx_x / 2,
					(((__u64)hctx->ccid3hctx_s) << 6) /
								    TFRC_T_MBI);

		ccid3_pr_debug("%s(%p, state=%s), updated tx rate to %u "
			       "bytes/s\n", dccp_role(sk), sk,
			       ccid3_tx_state_name(hctx->ccid3hctx_state),
			       (unsigned)(hctx->ccid3hctx_x >> 6));
		/* The value of R is still undefined and so we can not recompute
		 * the timeout value. Keep initial value as per [RFC 4342, 5]. */
		t_nfb = TFRC_INITIAL_TIMEOUT;
		ccid3_update_send_interval(hctx);
		break;
	case TFRC_SSTATE_FBACK:
		/*
		 *  Modify the cached value of X_recv [RFC 3448, 4.4]
		 *
		 *  If (p == 0 || X_calc > 2 * X_recv)
		 *    X_recv = max(X_recv / 2, s / (2 * t_mbi));
		 *  Else
		 *    X_recv = X_calc / 4;
		 *
		 *  Note that X_recv is scaled by 2^6 while X_calc is not
		 */
		BUG_ON(hctx->ccid3hctx_p && !hctx->ccid3hctx_x_calc);

		if (hctx->ccid3hctx_p == 0 ||
		    (hctx->ccid3hctx_x_calc > (hctx->ccid3hctx_x_recv >> 5))) {

			hctx->ccid3hctx_x_recv =
				max(hctx->ccid3hctx_x_recv / 2,
				    (((__u64)hctx->ccid3hctx_s) << 6) /
							      (2 * TFRC_T_MBI));
		} else {
			hctx->ccid3hctx_x_recv = hctx->ccid3hctx_x_calc;
			hctx->ccid3hctx_x_recv <<= 4;
		}
		/* Now recalculate X [RFC 3448, 4.3, step (4)] */
		ccid3_hc_tx_update_x(sk, NULL);
		/*
		 * Schedule no feedback timer to expire in
		 * max(t_RTO, 2 * s/X)  =  max(t_RTO, 2 * t_ipi)
		 * See comments in packet_recv() regarding the value of t_RTO.
		 */
		t_nfb = max(hctx->ccid3hctx_t_rto, 2 * hctx->ccid3hctx_t_ipi);
		break;
	case TFRC_SSTATE_NO_SENT:
		DCCP_BUG("%s(%p) - Illegal state NO_SENT", dccp_role(sk), sk);
		/* fall through */
	case TFRC_SSTATE_TERM:
		goto out;
	}

restart_timer:
	sk_reset_timer(sk, &hctx->ccid3hctx_no_feedback_timer,
			   jiffies + usecs_to_jiffies(t_nfb));
out:
	bh_unlock_sock(sk);
	sock_put(sk);
}

/*
 * returns
 *   > 0: delay (in msecs) that should pass before actually sending
 *   = 0: can send immediately
 *   < 0: error condition; do not send packet
 */
static int ccid3_hc_tx_send_packet(struct sock *sk, struct sk_buff *skb)
{
	struct dccp_sock *dp = dccp_sk(sk);
	struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
	ktime_t now = ktime_get_real();
	s64 delay;

	/*
	 * This function is called only for Data and DataAck packets. Sending
	 * zero-sized Data(Ack)s is theoretically possible, but for congestion
	 * control this case is pathological - ignore it.
	 */
	if (unlikely(skb->len == 0))
		return -EBADMSG;

	switch (hctx->ccid3hctx_state) {
	case TFRC_SSTATE_NO_SENT:
		sk_reset_timer(sk, &hctx->ccid3hctx_no_feedback_timer,
			       (jiffies +
				usecs_to_jiffies(TFRC_INITIAL_TIMEOUT)));
		hctx->ccid3hctx_last_win_count	 = 0;
		hctx->ccid3hctx_t_last_win_count = now;

		/* Set t_0 for initial packet */
		hctx->ccid3hctx_t_nom = now;

		hctx->ccid3hctx_s = skb->len;

		/*
		 * Use initial RTT sample when available: recommended by erratum
		 * to RFC 4342. This implements the initialisation procedure of
		 * draft rfc3448bis, section 4.2. Remember, X is scaled by 2^6.
		 */
		if (dp->dccps_syn_rtt) {
			ccid3_pr_debug("SYN RTT = %uus\n", dp->dccps_syn_rtt);
			hctx->ccid3hctx_rtt  = dp->dccps_syn_rtt;
			hctx->ccid3hctx_x    = rfc3390_initial_rate(sk);
			hctx->ccid3hctx_t_ld = now;
		} else {
			/* Sender does not have RTT sample: X = MSS/second */
			hctx->ccid3hctx_x = dp->dccps_mss_cache;
			hctx->ccid3hctx_x <<= 6;
		}
		ccid3_update_send_interval(hctx);

		ccid3_hc_tx_set_state(sk, TFRC_SSTATE_NO_FBACK);
		break;
	case TFRC_SSTATE_NO_FBACK:
	case TFRC_SSTATE_FBACK:
		delay = ktime_us_delta(hctx->ccid3hctx_t_nom, now);
		ccid3_pr_debug("delay=%ld\n", (long)delay);
		/*
		 *	Scheduling of packet transmissions [RFC 3448, 4.6]
		 *
		 * if (t_now > t_nom - delta)
		 *       // send the packet now
		 * else
		 *       // send the packet in (t_nom - t_now) milliseconds.
		 */
		if (delay - (s64)hctx->ccid3hctx_delta >= 1000)
			return (u32)delay / 1000L;

		ccid3_hc_tx_update_win_count(hctx, now);
		break;
	case TFRC_SSTATE_TERM:
		DCCP_BUG("%s(%p) - Illegal state TERM", dccp_role(sk), sk);
		return -EINVAL;
	}