2 * net/dccp/ccids/ccid3.c
4 * Copyright (c) 2005-7 The University of Waikato, Hamilton, New Zealand.
5 * Copyright (c) 2005-7 Ian McDonald <ian.mcdonald@jandi.co.nz>
7 * An implementation of the DCCP protocol
9 * This code has been developed by the University of Waikato WAND
10 * research group. For further information please see http://www.wand.net.nz/
12 * This code also uses code from Lulea University, rereleased as GPL by its
14 * Copyright (c) 2003 Nils-Erik Mattsson, Joacim Haggmark, Magnus Erixzon
16 * Changes to meet Linux coding standards, to make it meet latest ccid3 draft
17 * and to make it work as a loadable module in the DCCP stack written by
18 * Arnaldo Carvalho de Melo <acme@conectiva.com.br>.
20 * Copyright (c) 2005 Arnaldo Carvalho de Melo <acme@conectiva.com.br>
22 * This program is free software; you can redistribute it and/or modify
23 * it under the terms of the GNU General Public License as published by
24 * the Free Software Foundation; either version 2 of the License, or
25 * (at your option) any later version.
27 * This program is distributed in the hope that it will be useful,
28 * but WITHOUT ANY WARRANTY; without even the implied warranty of
29 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
30 * GNU General Public License for more details.
32 * You should have received a copy of the GNU General Public License
33 * along with this program; if not, write to the Free Software
34 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
38 #include "lib/packet_history.h"
39 #include "lib/loss_interval.h"
43 #ifdef CONFIG_IP_DCCP_CCID3_DEBUG
44 static int ccid3_debug;
45 #define ccid3_pr_debug(format, a...) DCCP_PR_DEBUG(ccid3_debug, format, ##a)
47 #define ccid3_pr_debug(format, a...)
50 static struct dccp_tx_hist *ccid3_tx_hist;
51 static struct dccp_rx_hist *ccid3_rx_hist;
54 * Transmitter Half-Connection Routines
56 #ifdef CONFIG_IP_DCCP_CCID3_DEBUG
57 static const char *ccid3_tx_state_name(enum ccid3_hc_tx_states state)
59 static char *ccid3_state_names[] = {
60 [TFRC_SSTATE_NO_SENT] = "NO_SENT",
61 [TFRC_SSTATE_NO_FBACK] = "NO_FBACK",
62 [TFRC_SSTATE_FBACK] = "FBACK",
63 [TFRC_SSTATE_TERM] = "TERM",
66 return ccid3_state_names[state];
70 static void ccid3_hc_tx_set_state(struct sock *sk,
71 enum ccid3_hc_tx_states state)
73 struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
74 enum ccid3_hc_tx_states oldstate = hctx->ccid3hctx_state;
76 ccid3_pr_debug("%s(%p) %-8.8s -> %s\n",
77 dccp_role(sk), sk, ccid3_tx_state_name(oldstate),
78 ccid3_tx_state_name(state));
79 WARN_ON(state == oldstate);
80 hctx->ccid3hctx_state = state;
84 * Compute the initial sending rate X_init according to RFC 3390:
85 * w_init = min(4 * MSS, max(2 * MSS, 4380 bytes))
86 * X_init = w_init / RTT
87 * For consistency with other parts of the code, X_init is scaled by 2^6.
89 static inline u64 rfc3390_initial_rate(struct sock *sk)
91 const struct dccp_sock *dp = dccp_sk(sk);
92 const __u32 w_init = min(4 * dp->dccps_mss_cache,
93 max(2 * dp->dccps_mss_cache, 4380U));
95 return scaled_div(w_init << 6, ccid3_hc_tx_sk(sk)->ccid3hctx_rtt);
99 * Recalculate t_ipi and delta (should be called whenever X changes)
101 static inline void ccid3_update_send_interval(struct ccid3_hc_tx_sock *hctx)
103 /* Calculate new t_ipi = s / X_inst (X_inst is in 64 * bytes/second) */
104 hctx->ccid3hctx_t_ipi = scaled_div32(((u64)hctx->ccid3hctx_s) << 6,
107 /* Calculate new delta by delta = min(t_ipi / 2, t_gran / 2) */
108 hctx->ccid3hctx_delta = min_t(u32, hctx->ccid3hctx_t_ipi / 2,
109 TFRC_OPSYS_HALF_TIME_GRAN);
111 ccid3_pr_debug("t_ipi=%u, delta=%u, s=%u, X=%u\n",
112 hctx->ccid3hctx_t_ipi, hctx->ccid3hctx_delta,
113 hctx->ccid3hctx_s, (unsigned)(hctx->ccid3hctx_x >> 6));
119 * X_calc = calcX(s, R, p);
120 * X = max(min(X_calc, 2 * X_recv), s / t_mbi);
122 * If (now - tld >= R)
123 * X = max(min(2 * X, 2 * X_recv), s / R);
126 * Note: X and X_recv are both stored in units of 64 * bytes/second, to support
127 * fine-grained resolution of sending rates. This requires scaling by 2^6
128 * throughout the code. Only X_calc is unscaled (in bytes/second).
131 static void ccid3_hc_tx_update_x(struct sock *sk, struct timeval *now)
134 struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
135 __u64 min_rate = 2 * hctx->ccid3hctx_x_recv;
136 const __u64 old_x = hctx->ccid3hctx_x;
139 * Handle IDLE periods: do not reduce below RFC3390 initial sending rate
140 * when idling [RFC 4342, 5.1]. See also draft-ietf-dccp-rfc3448bis.
141 * For consistency with X and X_recv, min_rate is also scaled by 2^6.
143 if (unlikely(hctx->ccid3hctx_idle)) {
144 min_rate = rfc3390_initial_rate(sk);
145 min_rate = max(min_rate, 2 * hctx->ccid3hctx_x_recv);
148 if (hctx->ccid3hctx_p > 0) {
150 hctx->ccid3hctx_x = min(((__u64)hctx->ccid3hctx_x_calc) << 6,
152 hctx->ccid3hctx_x = max(hctx->ccid3hctx_x,
153 (((__u64)hctx->ccid3hctx_s) << 6) /
156 } else if (timeval_delta(now, &hctx->ccid3hctx_t_ld) -
157 (suseconds_t)hctx->ccid3hctx_rtt >= 0) {
160 max(min(2 * hctx->ccid3hctx_x, min_rate),
161 scaled_div(((__u64)hctx->ccid3hctx_s) << 6,
162 hctx->ccid3hctx_rtt));
163 hctx->ccid3hctx_t_ld = *now;
166 if (hctx->ccid3hctx_x != old_x) {
167 ccid3_pr_debug("X_prev=%u, X_now=%u, X_calc=%u, "
168 "X_recv=%u\n", (unsigned)(old_x >> 6),
169 (unsigned)(hctx->ccid3hctx_x >> 6),
170 hctx->ccid3hctx_x_calc,
171 (unsigned)(hctx->ccid3hctx_x_recv >> 6));
173 ccid3_update_send_interval(hctx);
178 * Track the mean packet size `s' (cf. RFC 4342, 5.3 and RFC 3448, 4.1)
179 * @len: DCCP packet payload size in bytes
181 static inline void ccid3_hc_tx_update_s(struct ccid3_hc_tx_sock *hctx, int len)
183 const u16 old_s = hctx->ccid3hctx_s;
185 hctx->ccid3hctx_s = old_s == 0 ? len : (9 * old_s + len) / 10;
187 if (hctx->ccid3hctx_s != old_s)
188 ccid3_update_send_interval(hctx);
192 * Update Window Counter using the algorithm from [RFC 4342, 8.1].
193 * The algorithm is not applicable if RTT < 4 microseconds.
195 static inline void ccid3_hc_tx_update_win_count(struct ccid3_hc_tx_sock *hctx,
201 if (unlikely(hctx->ccid3hctx_rtt < 4)) /* avoid divide-by-zero */
204 delta = timeval_delta(now, &hctx->ccid3hctx_t_last_win_count);
205 DCCP_BUG_ON(delta < 0);
207 quarter_rtts = (u32)delta / (hctx->ccid3hctx_rtt / 4);
209 if (quarter_rtts > 0) {
210 hctx->ccid3hctx_t_last_win_count = *now;
211 hctx->ccid3hctx_last_win_count += min_t(u32, quarter_rtts, 5);
212 hctx->ccid3hctx_last_win_count &= 0xF; /* mod 16 */
214 ccid3_pr_debug("now at %#X\n", hctx->ccid3hctx_last_win_count);
218 static void ccid3_hc_tx_no_feedback_timer(unsigned long data)
220 struct sock *sk = (struct sock *)data;
221 struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
223 unsigned long t_nfb = USEC_PER_SEC / 5;
226 if (sock_owned_by_user(sk)) {
227 /* Try again later. */
228 /* XXX: set some sensible MIB */
232 ccid3_pr_debug("%s(%p, state=%s) - entry \n", dccp_role(sk), sk,
233 ccid3_tx_state_name(hctx->ccid3hctx_state));
235 hctx->ccid3hctx_idle = 1;
237 switch (hctx->ccid3hctx_state) {
238 case TFRC_SSTATE_NO_FBACK:
239 /* RFC 3448, 4.4: Halve send rate directly */
240 hctx->ccid3hctx_x = max(hctx->ccid3hctx_x / 2,
241 (((__u64)hctx->ccid3hctx_s) << 6) /
244 ccid3_pr_debug("%s(%p, state=%s), updated tx rate to %u "
245 "bytes/s\n", dccp_role(sk), sk,
246 ccid3_tx_state_name(hctx->ccid3hctx_state),
247 (unsigned)(hctx->ccid3hctx_x >> 6));
248 /* The value of R is still undefined and so we can not recompute
249 * the timout value. Keep initial value as per [RFC 4342, 5]. */
250 t_nfb = TFRC_INITIAL_TIMEOUT;
251 ccid3_update_send_interval(hctx);
253 case TFRC_SSTATE_FBACK:
255 * Modify the cached value of X_recv [RFC 3448, 4.4]
257 * If (p == 0 || X_calc > 2 * X_recv)
258 * X_recv = max(X_recv / 2, s / (2 * t_mbi));
260 * X_recv = X_calc / 4;
262 * Note that X_recv is scaled by 2^6 while X_calc is not
264 BUG_ON(hctx->ccid3hctx_p && !hctx->ccid3hctx_x_calc);
266 if (hctx->ccid3hctx_p == 0 ||
267 (hctx->ccid3hctx_x_calc > (hctx->ccid3hctx_x_recv >> 5))) {
269 hctx->ccid3hctx_x_recv =
270 max(hctx->ccid3hctx_x_recv / 2,
271 (((__u64)hctx->ccid3hctx_s) << 6) /
274 if (hctx->ccid3hctx_p == 0)
275 dccp_timestamp(sk, &now);
277 hctx->ccid3hctx_x_recv = hctx->ccid3hctx_x_calc;
278 hctx->ccid3hctx_x_recv <<= 4;
280 /* Now recalculate X [RFC 3448, 4.3, step (4)] */
281 ccid3_hc_tx_update_x(sk, &now);
283 * Schedule no feedback timer to expire in
284 * max(t_RTO, 2 * s/X) = max(t_RTO, 2 * t_ipi)
285 * See comments in packet_recv() regarding the value of t_RTO.
287 t_nfb = max(hctx->ccid3hctx_t_rto, 2 * hctx->ccid3hctx_t_ipi);
289 case TFRC_SSTATE_NO_SENT:
290 DCCP_BUG("%s(%p) - Illegal state NO_SENT", dccp_role(sk), sk);
292 case TFRC_SSTATE_TERM:
297 sk_reset_timer(sk, &hctx->ccid3hctx_no_feedback_timer,
298 jiffies + usecs_to_jiffies(t_nfb));
306 * > 0: delay (in msecs) that should pass before actually sending
307 * = 0: can send immediately
308 * < 0: error condition; do not send packet
310 static int ccid3_hc_tx_send_packet(struct sock *sk, struct sk_buff *skb)
312 struct dccp_sock *dp = dccp_sk(sk);
313 struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
317 BUG_ON(hctx == NULL);
320 * This function is called only for Data and DataAck packets. Sending
321 * zero-sized Data(Ack)s is theoretically possible, but for congestion
322 * control this case is pathological - ignore it.
324 if (unlikely(skb->len == 0))
327 dccp_timestamp(sk, &now);
329 switch (hctx->ccid3hctx_state) {
330 case TFRC_SSTATE_NO_SENT:
331 sk_reset_timer(sk, &hctx->ccid3hctx_no_feedback_timer,
333 usecs_to_jiffies(TFRC_INITIAL_TIMEOUT)));
334 hctx->ccid3hctx_last_win_count = 0;
335 hctx->ccid3hctx_t_last_win_count = now;
337 /* Set t_0 for initial packet */
338 hctx->ccid3hctx_t_nom = now;
340 hctx->ccid3hctx_s = skb->len;
343 * Use initial RTT sample when available: recommended by erratum
344 * to RFC 4342. This implements the initialisation procedure of
345 * draft rfc3448bis, section 4.2. Remember, X is scaled by 2^6.
347 if (dp->dccps_syn_rtt) {
348 ccid3_pr_debug("SYN RTT = %uus\n", dp->dccps_syn_rtt);
349 hctx->ccid3hctx_rtt = dp->dccps_syn_rtt;
350 hctx->ccid3hctx_x = rfc3390_initial_rate(sk);
351 hctx->ccid3hctx_t_ld = now;
353 /* Sender does not have RTT sample: X = MSS/second */
354 hctx->ccid3hctx_x = dp->dccps_mss_cache;
355 hctx->ccid3hctx_x <<= 6;
357 ccid3_update_send_interval(hctx);
359 ccid3_hc_tx_set_state(sk, TFRC_SSTATE_NO_FBACK);
361 case TFRC_SSTATE_NO_FBACK:
362 case TFRC_SSTATE_FBACK:
363 delay = timeval_delta(&hctx->ccid3hctx_t_nom, &now);
364 ccid3_pr_debug("delay=%ld\n", (long)delay);
366 * Scheduling of packet transmissions [RFC 3448, 4.6]
368 * if (t_now > t_nom - delta)
369 * // send the packet now
371 * // send the packet in (t_nom - t_now) milliseconds.
373 if (delay - (suseconds_t)hctx->ccid3hctx_delta >= 0)
374 return delay / 1000L;
376 ccid3_hc_tx_update_win_count(hctx, &now);
378 case TFRC_SSTATE_TERM:
379 DCCP_BUG("%s(%p) - Illegal state TERM", dccp_role(sk), sk);
383 /* prepare to send now (add options etc.) */
384 dp->dccps_hc_tx_insert_options = 1;
385 DCCP_SKB_CB(skb)->dccpd_ccval = hctx->ccid3hctx_last_win_count;
386 hctx->ccid3hctx_idle = 0;
388 /* set the nominal send time for the next following packet */
389 timeval_add_usecs(&hctx->ccid3hctx_t_nom, hctx->ccid3hctx_t_ipi);
394 static void ccid3_hc_tx_packet_sent(struct sock *sk, int more,
397 struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
399 struct dccp_tx_hist_entry *packet;
401 BUG_ON(hctx == NULL);
403 ccid3_hc_tx_update_s(hctx, len);
405 packet = dccp_tx_hist_entry_new(ccid3_tx_hist, GFP_ATOMIC);
406 if (unlikely(packet == NULL)) {
407 DCCP_CRIT("packet history - out of memory!");
410 dccp_tx_hist_add_entry(&hctx->ccid3hctx_hist, packet);
412 dccp_timestamp(sk, &now);
413 packet->dccphtx_tstamp = now;
414 packet->dccphtx_seqno = dccp_sk(sk)->dccps_gss;
415 packet->dccphtx_rtt = hctx->ccid3hctx_rtt;
416 packet->dccphtx_sent = 1;
419 static void ccid3_hc_tx_packet_recv(struct sock *sk, struct sk_buff *skb)
421 struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
422 struct ccid3_options_received *opt_recv;
423 struct dccp_tx_hist_entry *packet;
428 BUG_ON(hctx == NULL);
430 /* we are only interested in ACKs */
431 if (!(DCCP_SKB_CB(skb)->dccpd_type == DCCP_PKT_ACK ||
432 DCCP_SKB_CB(skb)->dccpd_type == DCCP_PKT_DATAACK))
435 opt_recv = &hctx->ccid3hctx_options_received;
437 switch (hctx->ccid3hctx_state) {
438 case TFRC_SSTATE_NO_FBACK:
439 case TFRC_SSTATE_FBACK:
440 /* get packet from history to look up t_recvdata */
441 packet = dccp_tx_hist_find_entry(&hctx->ccid3hctx_hist,
442 DCCP_SKB_CB(skb)->dccpd_ack_seq);
443 if (unlikely(packet == NULL)) {
444 DCCP_WARN("%s(%p), seqno %llu(%s) doesn't exist "
445 "in history!\n", dccp_role(sk), sk,
446 (unsigned long long)DCCP_SKB_CB(skb)->dccpd_ack_seq,
447 dccp_packet_name(DCCP_SKB_CB(skb)->dccpd_type));
451 /* Update receive rate in units of 64 * bytes/second */
452 hctx->ccid3hctx_x_recv = opt_recv->ccid3or_receive_rate;
453 hctx->ccid3hctx_x_recv <<= 6;
455 /* Update loss event rate */
456 pinv = opt_recv->ccid3or_loss_event_rate;
457 if (pinv == ~0U || pinv == 0) /* see RFC 4342, 8.5 */
458 hctx->ccid3hctx_p = 0;
459 else /* can not exceed 100% */
460 hctx->ccid3hctx_p = 1000000 / pinv;
462 dccp_timestamp(sk, &now);
465 * Calculate new round trip sample as per [RFC 3448, 4.3] by
466 * R_sample = (now - t_recvdata) - t_elapsed
468 r_sample = dccp_sample_rtt(sk, &now, &packet->dccphtx_tstamp);
471 * Update RTT estimate by
472 * If (No feedback recv)
475 * R = q * R + (1 - q) * R_sample;
477 * q is a constant, RFC 3448 recomments 0.9
479 if (hctx->ccid3hctx_state == TFRC_SSTATE_NO_FBACK) {
481 * Larger Initial Windows [RFC 4342, sec. 5]
483 hctx->ccid3hctx_rtt = r_sample;
484 hctx->ccid3hctx_x = rfc3390_initial_rate(sk);
485 hctx->ccid3hctx_t_ld = now;
487 ccid3_update_send_interval(hctx);
489 ccid3_pr_debug("%s(%p), s=%u, MSS=%u, "
490 "R_sample=%uus, X=%u\n", dccp_role(sk),
491 sk, hctx->ccid3hctx_s,
492 dccp_sk(sk)->dccps_mss_cache, r_sample,
493 (unsigned)(hctx->ccid3hctx_x >> 6));
495 ccid3_hc_tx_set_state(sk, TFRC_SSTATE_FBACK);
497 hctx->ccid3hctx_rtt = (9 * hctx->ccid3hctx_rtt +
500 /* Update sending rate (step 4 of [RFC 3448, 4.3]) */
501 if (hctx->ccid3hctx_p > 0)
502 hctx->ccid3hctx_x_calc =
503 tfrc_calc_x(hctx->ccid3hctx_s,
506 ccid3_hc_tx_update_x(sk, &now);
508 ccid3_pr_debug("%s(%p), RTT=%uus (sample=%uus), s=%u, "
509 "p=%u, X_calc=%u, X_recv=%u, X=%u\n",
511 sk, hctx->ccid3hctx_rtt, r_sample,
512 hctx->ccid3hctx_s, hctx->ccid3hctx_p,
513 hctx->ccid3hctx_x_calc,
514 (unsigned)(hctx->ccid3hctx_x_recv >> 6),
515 (unsigned)(hctx->ccid3hctx_x >> 6));
518 /* unschedule no feedback timer */
519 sk_stop_timer(sk, &hctx->ccid3hctx_no_feedback_timer);
521 /* remove all packets older than the one acked from history */
522 dccp_tx_hist_purge_older(ccid3_tx_hist,
523 &hctx->ccid3hctx_hist, packet);
525 * As we have calculated new ipi, delta, t_nom it is possible
526 * that we now can send a packet, so wake up dccp_wait_for_ccid
528 sk->sk_write_space(sk);
531 * Update timeout interval for the nofeedback timer.
532 * We use a configuration option to increase the lower bound.
533 * This can help avoid triggering the nofeedback timer too
534 * often ('spinning') on LANs with small RTTs.
536 hctx->ccid3hctx_t_rto = max_t(u32, 4 * hctx->ccid3hctx_rtt,
537 CONFIG_IP_DCCP_CCID3_RTO *
538 (USEC_PER_SEC/1000));
540 * Schedule no feedback timer to expire in
541 * max(t_RTO, 2 * s/X) = max(t_RTO, 2 * t_ipi)
543 t_nfb = max(hctx->ccid3hctx_t_rto, 2 * hctx->ccid3hctx_t_ipi);
545 ccid3_pr_debug("%s(%p), Scheduled no feedback timer to "
546 "expire in %lu jiffies (%luus)\n",
548 sk, usecs_to_jiffies(t_nfb), t_nfb);
550 sk_reset_timer(sk, &hctx->ccid3hctx_no_feedback_timer,
551 jiffies + usecs_to_jiffies(t_nfb));
554 hctx->ccid3hctx_idle = 1;
556 case TFRC_SSTATE_NO_SENT: /* fall through */
557 case TFRC_SSTATE_TERM: /* ignore feedback when closing */
562 static int ccid3_hc_tx_parse_options(struct sock *sk, unsigned char option,
563 unsigned char len, u16 idx,
564 unsigned char *value)
567 const struct dccp_sock *dp = dccp_sk(sk);
568 struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
569 struct ccid3_options_received *opt_recv;
571 BUG_ON(hctx == NULL);
573 opt_recv = &hctx->ccid3hctx_options_received;
575 if (opt_recv->ccid3or_seqno != dp->dccps_gsr) {
576 opt_recv->ccid3or_seqno = dp->dccps_gsr;
577 opt_recv->ccid3or_loss_event_rate = ~0;
578 opt_recv->ccid3or_loss_intervals_idx = 0;
579 opt_recv->ccid3or_loss_intervals_len = 0;
580 opt_recv->ccid3or_receive_rate = 0;
584 case TFRC_OPT_LOSS_EVENT_RATE:
585 if (unlikely(len != 4)) {
586 DCCP_WARN("%s(%p), invalid len %d "
587 "for TFRC_OPT_LOSS_EVENT_RATE\n",
588 dccp_role(sk), sk, len);
591 opt_recv->ccid3or_loss_event_rate =
592 ntohl(*(__be32 *)value);
593 ccid3_pr_debug("%s(%p), LOSS_EVENT_RATE=%u\n",
595 opt_recv->ccid3or_loss_event_rate);
598 case TFRC_OPT_LOSS_INTERVALS:
599 opt_recv->ccid3or_loss_intervals_idx = idx;
600 opt_recv->ccid3or_loss_intervals_len = len;
601 ccid3_pr_debug("%s(%p), LOSS_INTERVALS=(%u, %u)\n",
603 opt_recv->ccid3or_loss_intervals_idx,
604 opt_recv->ccid3or_loss_intervals_len);
606 case TFRC_OPT_RECEIVE_RATE:
607 if (unlikely(len != 4)) {
608 DCCP_WARN("%s(%p), invalid len %d "
609 "for TFRC_OPT_RECEIVE_RATE\n",
610 dccp_role(sk), sk, len);
613 opt_recv->ccid3or_receive_rate =
614 ntohl(*(__be32 *)value);
615 ccid3_pr_debug("%s(%p), RECEIVE_RATE=%u\n",
617 opt_recv->ccid3or_receive_rate);
625 static int ccid3_hc_tx_init(struct ccid *ccid, struct sock *sk)
627 struct ccid3_hc_tx_sock *hctx = ccid_priv(ccid);
629 hctx->ccid3hctx_s = 0;
630 hctx->ccid3hctx_rtt = 0;
631 hctx->ccid3hctx_state = TFRC_SSTATE_NO_SENT;
632 INIT_LIST_HEAD(&hctx->ccid3hctx_hist);
634 hctx->ccid3hctx_no_feedback_timer.function =
635 ccid3_hc_tx_no_feedback_timer;
636 hctx->ccid3hctx_no_feedback_timer.data = (unsigned long)sk;
637 init_timer(&hctx->ccid3hctx_no_feedback_timer);
642 static void ccid3_hc_tx_exit(struct sock *sk)
644 struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
646 BUG_ON(hctx == NULL);
648 ccid3_hc_tx_set_state(sk, TFRC_SSTATE_TERM);
649 sk_stop_timer(sk, &hctx->ccid3hctx_no_feedback_timer);
651 /* Empty packet history */
652 dccp_tx_hist_purge(ccid3_tx_hist, &hctx->ccid3hctx_hist);
655 static void ccid3_hc_tx_get_info(struct sock *sk, struct tcp_info *info)
657 const struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
659 /* Listen socks doesn't have a private CCID block */
660 if (sk->sk_state == DCCP_LISTEN)
663 BUG_ON(hctx == NULL);
665 info->tcpi_rto = hctx->ccid3hctx_t_rto;
666 info->tcpi_rtt = hctx->ccid3hctx_rtt;
669 static int ccid3_hc_tx_getsockopt(struct sock *sk, const int optname, int len,
670 u32 __user *optval, int __user *optlen)
672 const struct ccid3_hc_tx_sock *hctx = ccid3_hc_tx_sk(sk);
675 /* Listen socks doesn't have a private CCID block */
676 if (sk->sk_state == DCCP_LISTEN)
680 case DCCP_SOCKOPT_CCID_TX_INFO:
681 if (len < sizeof(hctx->ccid3hctx_tfrc))
683 len = sizeof(hctx->ccid3hctx_tfrc);
684 val = &hctx->ccid3hctx_tfrc;
690 if (put_user(len, optlen) || copy_to_user(optval, val, len))
697 * Receiver Half-Connection Routines
699 #ifdef CONFIG_IP_DCCP_CCID3_DEBUG
700 static const char *ccid3_rx_state_name(enum ccid3_hc_rx_states state)
702 static char *ccid3_rx_state_names[] = {
703 [TFRC_RSTATE_NO_DATA] = "NO_DATA",
704 [TFRC_RSTATE_DATA] = "DATA",
705 [TFRC_RSTATE_TERM] = "TERM",
708 return ccid3_rx_state_names[state];
712 static void ccid3_hc_rx_set_state(struct sock *sk,
713 enum ccid3_hc_rx_states state)
715 struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
716 enum ccid3_hc_rx_states oldstate = hcrx->ccid3hcrx_state;
718 ccid3_pr_debug("%s(%p) %-8.8s -> %s\n",
719 dccp_role(sk), sk, ccid3_rx_state_name(oldstate),
720 ccid3_rx_state_name(state));
721 WARN_ON(state == oldstate);
722 hcrx->ccid3hcrx_state = state;
725 static inline void ccid3_hc_rx_update_s(struct ccid3_hc_rx_sock *hcrx, int len)
727 if (unlikely(len == 0)) /* don't update on empty packets (e.g. ACKs) */
728 ccid3_pr_debug("Packet payload length is 0 - not updating\n");
730 hcrx->ccid3hcrx_s = hcrx->ccid3hcrx_s == 0 ? len :
731 (9 * hcrx->ccid3hcrx_s + len) / 10;
734 static void ccid3_hc_rx_send_feedback(struct sock *sk)
736 struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
737 struct dccp_sock *dp = dccp_sk(sk);
738 struct dccp_rx_hist_entry *packet;
742 ccid3_pr_debug("%s(%p) - entry \n", dccp_role(sk), sk);
744 dccp_timestamp(sk, &now);
746 switch (hcrx->ccid3hcrx_state) {
747 case TFRC_RSTATE_NO_DATA:
748 hcrx->ccid3hcrx_x_recv = 0;
750 case TFRC_RSTATE_DATA:
751 delta = timeval_delta(&now,
752 &hcrx->ccid3hcrx_tstamp_last_feedback);
753 DCCP_BUG_ON(delta < 0);
754 hcrx->ccid3hcrx_x_recv =
755 scaled_div32(hcrx->ccid3hcrx_bytes_recv, delta);
757 case TFRC_RSTATE_TERM:
758 DCCP_BUG("%s(%p) - Illegal state TERM", dccp_role(sk), sk);
762 packet = dccp_rx_hist_find_data_packet(&hcrx->ccid3hcrx_hist);
763 if (unlikely(packet == NULL)) {
764 DCCP_WARN("%s(%p), no data packet in history!\n",
769 hcrx->ccid3hcrx_tstamp_last_feedback = now;
770 hcrx->ccid3hcrx_ccval_last_counter = packet->dccphrx_ccval;
771 hcrx->ccid3hcrx_bytes_recv = 0;
773 /* Elapsed time information [RFC 4340, 13.2] in units of 10 * usecs */
774 delta = timeval_delta(&now, &packet->dccphrx_tstamp);
775 DCCP_BUG_ON(delta < 0);
776 hcrx->ccid3hcrx_elapsed_time = delta / 10;
778 if (hcrx->ccid3hcrx_p == 0)
779 hcrx->ccid3hcrx_pinv = ~0U; /* see RFC 4342, 8.5 */
780 else if (hcrx->ccid3hcrx_p > 1000000) {
781 DCCP_WARN("p (%u) > 100%%\n", hcrx->ccid3hcrx_p);
782 hcrx->ccid3hcrx_pinv = 1; /* use 100% in this case */
784 hcrx->ccid3hcrx_pinv = 1000000 / hcrx->ccid3hcrx_p;
786 dp->dccps_hc_rx_insert_options = 1;
790 static int ccid3_hc_rx_insert_options(struct sock *sk, struct sk_buff *skb)
792 const struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
795 BUG_ON(hcrx == NULL);
797 if (!(sk->sk_state == DCCP_OPEN || sk->sk_state == DCCP_PARTOPEN))
800 DCCP_SKB_CB(skb)->dccpd_ccval = hcrx->ccid3hcrx_ccval_last_counter;
802 if (dccp_packet_without_ack(skb))
805 x_recv = htonl(hcrx->ccid3hcrx_x_recv);
806 pinv = htonl(hcrx->ccid3hcrx_pinv);
808 if ((hcrx->ccid3hcrx_elapsed_time != 0 &&
809 dccp_insert_option_elapsed_time(sk, skb,
810 hcrx->ccid3hcrx_elapsed_time)) ||
811 dccp_insert_option_timestamp(sk, skb) ||
812 dccp_insert_option(sk, skb, TFRC_OPT_LOSS_EVENT_RATE,
813 &pinv, sizeof(pinv)) ||
814 dccp_insert_option(sk, skb, TFRC_OPT_RECEIVE_RATE,
815 &x_recv, sizeof(x_recv)))
821 static int ccid3_hc_rx_detect_loss(struct sock *sk,
822 struct dccp_rx_hist_entry *packet)
824 struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
825 struct dccp_rx_hist_entry *rx_hist =
826 dccp_rx_hist_head(&hcrx->ccid3hcrx_hist);
827 u64 seqno = packet->dccphrx_seqno;
833 tmp_seqno = hcrx->ccid3hcrx_seqno_nonloss;
836 follows48(packet->dccphrx_seqno, hcrx->ccid3hcrx_seqno_nonloss)) {
837 hcrx->ccid3hcrx_seqno_nonloss = seqno;
838 hcrx->ccid3hcrx_ccval_nonloss = packet->dccphrx_ccval;
843 while (dccp_delta_seqno(hcrx->ccid3hcrx_seqno_nonloss, seqno)
844 > TFRC_RECV_NUM_LATE_LOSS) {
846 dccp_li_update_li(sk,
847 &hcrx->ccid3hcrx_li_hist,
848 &hcrx->ccid3hcrx_hist,
849 &hcrx->ccid3hcrx_tstamp_last_feedback,
851 hcrx->ccid3hcrx_bytes_recv,
852 hcrx->ccid3hcrx_x_recv,
853 hcrx->ccid3hcrx_seqno_nonloss,
854 hcrx->ccid3hcrx_ccval_nonloss);
855 tmp_seqno = hcrx->ccid3hcrx_seqno_nonloss;
856 dccp_inc_seqno(&tmp_seqno);
857 hcrx->ccid3hcrx_seqno_nonloss = tmp_seqno;
858 dccp_inc_seqno(&tmp_seqno);
859 while (dccp_rx_hist_find_entry(&hcrx->ccid3hcrx_hist,
860 tmp_seqno, &ccval)) {
861 hcrx->ccid3hcrx_seqno_nonloss = tmp_seqno;
862 hcrx->ccid3hcrx_ccval_nonloss = ccval;
863 dccp_inc_seqno(&tmp_seqno);
867 /* FIXME - this code could be simplified with above while */
868 /* but works at moment */
869 if (follows48(packet->dccphrx_seqno, hcrx->ccid3hcrx_seqno_nonloss)) {
870 hcrx->ccid3hcrx_seqno_nonloss = seqno;
871 hcrx->ccid3hcrx_ccval_nonloss = packet->dccphrx_ccval;
875 dccp_rx_hist_add_packet(ccid3_rx_hist, &hcrx->ccid3hcrx_hist,
876 &hcrx->ccid3hcrx_li_hist, packet,
877 hcrx->ccid3hcrx_seqno_nonloss);
881 static void ccid3_hc_rx_packet_recv(struct sock *sk, struct sk_buff *skb)
883 struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
884 const struct dccp_options_received *opt_recv;
885 struct dccp_rx_hist_entry *packet;
887 u32 p_prev, r_sample, rtt_prev;
888 int loss, payload_size;
890 BUG_ON(hcrx == NULL);
892 opt_recv = &dccp_sk(sk)->dccps_options_received;
894 switch (DCCP_SKB_CB(skb)->dccpd_type) {
896 if (hcrx->ccid3hcrx_state == TFRC_RSTATE_NO_DATA)
898 case DCCP_PKT_DATAACK:
899 if (opt_recv->dccpor_timestamp_echo == 0)
901 rtt_prev = hcrx->ccid3hcrx_rtt;
902 dccp_timestamp(sk, &now);
903 r_sample = dccp_sample_rtt(sk, &now, NULL);
905 if (hcrx->ccid3hcrx_state == TFRC_RSTATE_NO_DATA)
906 hcrx->ccid3hcrx_rtt = r_sample;
908 hcrx->ccid3hcrx_rtt = (hcrx->ccid3hcrx_rtt * 9) / 10 +
911 if (rtt_prev != hcrx->ccid3hcrx_rtt)
912 ccid3_pr_debug("%s(%p), New RTT=%uus, elapsed time=%u\n",
913 dccp_role(sk), sk, hcrx->ccid3hcrx_rtt,
914 opt_recv->dccpor_elapsed_time);
918 default: /* We're not interested in other packet types, move along */
922 packet = dccp_rx_hist_entry_new(ccid3_rx_hist, sk, opt_recv->dccpor_ndp,
924 if (unlikely(packet == NULL)) {
925 DCCP_WARN("%s(%p), Not enough mem to add rx packet "
926 "to history, consider it lost!\n", dccp_role(sk), sk);
930 loss = ccid3_hc_rx_detect_loss(sk, packet);
932 if (DCCP_SKB_CB(skb)->dccpd_type == DCCP_PKT_ACK)
935 payload_size = skb->len - dccp_hdr(skb)->dccph_doff * 4;
936 ccid3_hc_rx_update_s(hcrx, payload_size);
938 switch (hcrx->ccid3hcrx_state) {
939 case TFRC_RSTATE_NO_DATA:
940 ccid3_pr_debug("%s(%p, state=%s), skb=%p, sending initial "
941 "feedback\n", dccp_role(sk), sk,
942 dccp_state_name(sk->sk_state), skb);
943 ccid3_hc_rx_send_feedback(sk);
944 ccid3_hc_rx_set_state(sk, TFRC_RSTATE_DATA);
946 case TFRC_RSTATE_DATA:
947 hcrx->ccid3hcrx_bytes_recv += payload_size;
951 dccp_timestamp(sk, &now);
952 if ((timeval_delta(&now, &hcrx->ccid3hcrx_tstamp_last_ack) -
953 (suseconds_t)hcrx->ccid3hcrx_rtt) >= 0) {
954 hcrx->ccid3hcrx_tstamp_last_ack = now;
955 ccid3_hc_rx_send_feedback(sk);
958 case TFRC_RSTATE_TERM:
959 DCCP_BUG("%s(%p) - Illegal state TERM", dccp_role(sk), sk);
963 /* Dealing with packet loss */
964 ccid3_pr_debug("%s(%p, state=%s), data loss! Reacting...\n",
965 dccp_role(sk), sk, dccp_state_name(sk->sk_state));
967 p_prev = hcrx->ccid3hcrx_p;
969 /* Calculate loss event rate */
970 if (!list_empty(&hcrx->ccid3hcrx_li_hist)) {
971 u32 i_mean = dccp_li_hist_calc_i_mean(&hcrx->ccid3hcrx_li_hist);
973 /* Scaling up by 1000000 as fixed decimal */
975 hcrx->ccid3hcrx_p = 1000000 / i_mean;
977 DCCP_BUG("empty loss history");
979 if (hcrx->ccid3hcrx_p > p_prev) {
980 ccid3_hc_rx_send_feedback(sk);
985 static int ccid3_hc_rx_init(struct ccid *ccid, struct sock *sk)
987 struct ccid3_hc_rx_sock *hcrx = ccid_priv(ccid);
989 ccid3_pr_debug("entry\n");
991 hcrx->ccid3hcrx_state = TFRC_RSTATE_NO_DATA;
992 INIT_LIST_HEAD(&hcrx->ccid3hcrx_hist);
993 INIT_LIST_HEAD(&hcrx->ccid3hcrx_li_hist);
994 dccp_timestamp(sk, &hcrx->ccid3hcrx_tstamp_last_ack);
995 hcrx->ccid3hcrx_tstamp_last_feedback = hcrx->ccid3hcrx_tstamp_last_ack;
996 hcrx->ccid3hcrx_s = 0;
997 hcrx->ccid3hcrx_rtt = 0;
1001 static void ccid3_hc_rx_exit(struct sock *sk)
1003 struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
1005 BUG_ON(hcrx == NULL);
1007 ccid3_hc_rx_set_state(sk, TFRC_RSTATE_TERM);
1009 /* Empty packet history */
1010 dccp_rx_hist_purge(ccid3_rx_hist, &hcrx->ccid3hcrx_hist);
1012 /* Empty loss interval history */
1013 dccp_li_hist_purge(&hcrx->ccid3hcrx_li_hist);
1016 static void ccid3_hc_rx_get_info(struct sock *sk, struct tcp_info *info)
1018 const struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
1020 /* Listen socks doesn't have a private CCID block */
1021 if (sk->sk_state == DCCP_LISTEN)
1024 BUG_ON(hcrx == NULL);
1026 info->tcpi_ca_state = hcrx->ccid3hcrx_state;
1027 info->tcpi_options |= TCPI_OPT_TIMESTAMPS;
1028 info->tcpi_rcv_rtt = hcrx->ccid3hcrx_rtt;
1031 static int ccid3_hc_rx_getsockopt(struct sock *sk, const int optname, int len,
1032 u32 __user *optval, int __user *optlen)
1034 const struct ccid3_hc_rx_sock *hcrx = ccid3_hc_rx_sk(sk);
1037 /* Listen socks doesn't have a private CCID block */
1038 if (sk->sk_state == DCCP_LISTEN)
1042 case DCCP_SOCKOPT_CCID_RX_INFO:
1043 if (len < sizeof(hcrx->ccid3hcrx_tfrc))
1045 len = sizeof(hcrx->ccid3hcrx_tfrc);
1046 val = &hcrx->ccid3hcrx_tfrc;
1049 return -ENOPROTOOPT;
1052 if (put_user(len, optlen) || copy_to_user(optval, val, len))
1058 static struct ccid_operations ccid3 = {
1059 .ccid_id = DCCPC_CCID3,
1060 .ccid_name = "ccid3",
1061 .ccid_owner = THIS_MODULE,
1062 .ccid_hc_tx_obj_size = sizeof(struct ccid3_hc_tx_sock),
1063 .ccid_hc_tx_init = ccid3_hc_tx_init,
1064 .ccid_hc_tx_exit = ccid3_hc_tx_exit,
1065 .ccid_hc_tx_send_packet = ccid3_hc_tx_send_packet,
1066 .ccid_hc_tx_packet_sent = ccid3_hc_tx_packet_sent,
1067 .ccid_hc_tx_packet_recv = ccid3_hc_tx_packet_recv,
1068 .ccid_hc_tx_parse_options = ccid3_hc_tx_parse_options,
1069 .ccid_hc_rx_obj_size = sizeof(struct ccid3_hc_rx_sock),
1070 .ccid_hc_rx_init = ccid3_hc_rx_init,
1071 .ccid_hc_rx_exit = ccid3_hc_rx_exit,
1072 .ccid_hc_rx_insert_options = ccid3_hc_rx_insert_options,
1073 .ccid_hc_rx_packet_recv = ccid3_hc_rx_packet_recv,
1074 .ccid_hc_rx_get_info = ccid3_hc_rx_get_info,
1075 .ccid_hc_tx_get_info = ccid3_hc_tx_get_info,
1076 .ccid_hc_rx_getsockopt = ccid3_hc_rx_getsockopt,
1077 .ccid_hc_tx_getsockopt = ccid3_hc_tx_getsockopt,
1080 #ifdef CONFIG_IP_DCCP_CCID3_DEBUG
1081 module_param(ccid3_debug, int, 0444);
1082 MODULE_PARM_DESC(ccid3_debug, "Enable debug messages");
1085 static __init int ccid3_module_init(void)
1089 ccid3_rx_hist = dccp_rx_hist_new("ccid3");
1090 if (ccid3_rx_hist == NULL)
1093 ccid3_tx_hist = dccp_tx_hist_new("ccid3");
1094 if (ccid3_tx_hist == NULL)
1097 rc = ccid_register(&ccid3);
1104 dccp_tx_hist_delete(ccid3_tx_hist);
1105 ccid3_tx_hist = NULL;
1107 dccp_rx_hist_delete(ccid3_rx_hist);
1108 ccid3_rx_hist = NULL;
1111 module_init(ccid3_module_init);
1113 static __exit void ccid3_module_exit(void)
1115 ccid_unregister(&ccid3);
1117 if (ccid3_tx_hist != NULL) {
1118 dccp_tx_hist_delete(ccid3_tx_hist);
1119 ccid3_tx_hist = NULL;
1121 if (ccid3_rx_hist != NULL) {
1122 dccp_rx_hist_delete(ccid3_rx_hist);
1123 ccid3_rx_hist = NULL;
1126 module_exit(ccid3_module_exit);
1128 MODULE_AUTHOR("Ian McDonald <ian.mcdonald@jandi.co.nz>, "
1129 "Arnaldo Carvalho de Melo <acme@ghostprotocols.net>");
1130 MODULE_DESCRIPTION("DCCP TFRC CCID3 CCID");
1131 MODULE_LICENSE("GPL");
1132 MODULE_ALIAS("net-dccp-ccid-3");