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
)
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) /
157 const ktime_t now
= ktime_get_real();
159 if ((ktime_us_delta(now
, hctx
->ccid3hctx_t_ld
) -
160 (s64
)hctx
->ccid3hctx_rtt
) >= 0) {
163 max(min(2 * hctx
->ccid3hctx_x
, min_rate
),
164 scaled_div(((__u64
)hctx
->ccid3hctx_s
) << 6,
165 hctx
->ccid3hctx_rtt
));
166 hctx
->ccid3hctx_t_ld
= now
;
170 if (hctx
->ccid3hctx_x
!= old_x
) {
171 ccid3_pr_debug("X_prev=%u, X_now=%u, X_calc=%u, "
172 "X_recv=%u\n", (unsigned)(old_x
>> 6),
173 (unsigned)(hctx
->ccid3hctx_x
>> 6),
174 hctx
->ccid3hctx_x_calc
,
175 (unsigned)(hctx
->ccid3hctx_x_recv
>> 6));
177 ccid3_update_send_interval(hctx
);
182 * Track the mean packet size `s' (cf. RFC 4342, 5.3 and RFC 3448, 4.1)
183 * @len: DCCP packet payload size in bytes
185 static inline void ccid3_hc_tx_update_s(struct ccid3_hc_tx_sock
*hctx
, int len
)
187 const u16 old_s
= hctx
->ccid3hctx_s
;
189 hctx
->ccid3hctx_s
= old_s
== 0 ? len
: (9 * old_s
+ len
) / 10;
191 if (hctx
->ccid3hctx_s
!= old_s
)
192 ccid3_update_send_interval(hctx
);
196 * Update Window Counter using the algorithm from [RFC 4342, 8.1].
197 * The algorithm is not applicable if RTT < 4 microseconds.
199 static inline void ccid3_hc_tx_update_win_count(struct ccid3_hc_tx_sock
*hctx
,
204 if (unlikely(hctx
->ccid3hctx_rtt
< 4)) /* avoid divide-by-zero */
207 quarter_rtts
= ktime_us_delta(now
, hctx
->ccid3hctx_t_last_win_count
);
208 quarter_rtts
/= hctx
->ccid3hctx_rtt
/ 4;
210 if (quarter_rtts
> 0) {
211 hctx
->ccid3hctx_t_last_win_count
= now
;
212 hctx
->ccid3hctx_last_win_count
+= min_t(u32
, quarter_rtts
, 5);
213 hctx
->ccid3hctx_last_win_count
&= 0xF; /* mod 16 */
217 static void ccid3_hc_tx_no_feedback_timer(unsigned long data
)
219 struct sock
*sk
= (struct sock
*)data
;
220 struct ccid3_hc_tx_sock
*hctx
= ccid3_hc_tx_sk(sk
);
221 unsigned long t_nfb
= USEC_PER_SEC
/ 5;
224 if (sock_owned_by_user(sk
)) {
225 /* Try again later. */
226 /* XXX: set some sensible MIB */
230 ccid3_pr_debug("%s(%p, state=%s) - entry \n", dccp_role(sk
), sk
,
231 ccid3_tx_state_name(hctx
->ccid3hctx_state
));
233 hctx
->ccid3hctx_idle
= 1;
235 switch (hctx
->ccid3hctx_state
) {
236 case TFRC_SSTATE_NO_FBACK
:
237 /* RFC 3448, 4.4: Halve send rate directly */
238 hctx
->ccid3hctx_x
= max(hctx
->ccid3hctx_x
/ 2,
239 (((__u64
)hctx
->ccid3hctx_s
) << 6) /
242 ccid3_pr_debug("%s(%p, state=%s), updated tx rate to %u "
243 "bytes/s\n", dccp_role(sk
), sk
,
244 ccid3_tx_state_name(hctx
->ccid3hctx_state
),
245 (unsigned)(hctx
->ccid3hctx_x
>> 6));
246 /* The value of R is still undefined and so we can not recompute
247 * the timout value. Keep initial value as per [RFC 4342, 5]. */
248 t_nfb
= TFRC_INITIAL_TIMEOUT
;
249 ccid3_update_send_interval(hctx
);
251 case TFRC_SSTATE_FBACK
:
253 * Modify the cached value of X_recv [RFC 3448, 4.4]
255 * If (p == 0 || X_calc > 2 * X_recv)
256 * X_recv = max(X_recv / 2, s / (2 * t_mbi));
258 * X_recv = X_calc / 4;
260 * Note that X_recv is scaled by 2^6 while X_calc is not
262 BUG_ON(hctx
->ccid3hctx_p
&& !hctx
->ccid3hctx_x_calc
);
264 if (hctx
->ccid3hctx_p
== 0 ||
265 (hctx
->ccid3hctx_x_calc
> (hctx
->ccid3hctx_x_recv
>> 5))) {
267 hctx
->ccid3hctx_x_recv
=
268 max(hctx
->ccid3hctx_x_recv
/ 2,
269 (((__u64
)hctx
->ccid3hctx_s
) << 6) /
272 hctx
->ccid3hctx_x_recv
= hctx
->ccid3hctx_x_calc
;
273 hctx
->ccid3hctx_x_recv
<<= 4;
275 /* Now recalculate X [RFC 3448, 4.3, step (4)] */
276 ccid3_hc_tx_update_x(sk
);
278 * Schedule no feedback timer to expire in
279 * max(t_RTO, 2 * s/X) = max(t_RTO, 2 * t_ipi)
280 * See comments in packet_recv() regarding the value of t_RTO.
282 t_nfb
= max(hctx
->ccid3hctx_t_rto
, 2 * hctx
->ccid3hctx_t_ipi
);
284 case TFRC_SSTATE_NO_SENT
:
285 DCCP_BUG("%s(%p) - Illegal state NO_SENT", dccp_role(sk
), sk
);
287 case TFRC_SSTATE_TERM
:
292 sk_reset_timer(sk
, &hctx
->ccid3hctx_no_feedback_timer
,
293 jiffies
+ usecs_to_jiffies(t_nfb
));
301 * > 0: delay (in msecs) that should pass before actually sending
302 * = 0: can send immediately
303 * < 0: error condition; do not send packet
305 static int ccid3_hc_tx_send_packet(struct sock
*sk
, struct sk_buff
*skb
)
307 struct dccp_sock
*dp
= dccp_sk(sk
);
308 struct ccid3_hc_tx_sock
*hctx
= ccid3_hc_tx_sk(sk
);
309 ktime_t now
= ktime_get_real();
312 BUG_ON(hctx
== NULL
);
315 * This function is called only for Data and DataAck packets. Sending
316 * zero-sized Data(Ack)s is theoretically possible, but for congestion
317 * control this case is pathological - ignore it.
319 if (unlikely(skb
->len
== 0))
322 switch (hctx
->ccid3hctx_state
) {
323 case TFRC_SSTATE_NO_SENT
:
324 sk_reset_timer(sk
, &hctx
->ccid3hctx_no_feedback_timer
,
326 usecs_to_jiffies(TFRC_INITIAL_TIMEOUT
)));
327 hctx
->ccid3hctx_last_win_count
= 0;
328 hctx
->ccid3hctx_t_last_win_count
= now
;
330 /* Set t_0 for initial packet */
331 hctx
->ccid3hctx_t_nom
= now
;
333 hctx
->ccid3hctx_s
= skb
->len
;
336 * Use initial RTT sample when available: recommended by erratum
337 * to RFC 4342. This implements the initialisation procedure of
338 * draft rfc3448bis, section 4.2. Remember, X is scaled by 2^6.
340 if (dp
->dccps_syn_rtt
) {
341 ccid3_pr_debug("SYN RTT = %uus\n", dp
->dccps_syn_rtt
);
342 hctx
->ccid3hctx_rtt
= dp
->dccps_syn_rtt
;
343 hctx
->ccid3hctx_x
= rfc3390_initial_rate(sk
);
344 hctx
->ccid3hctx_t_ld
= now
;
346 /* Sender does not have RTT sample: X = MSS/second */
347 hctx
->ccid3hctx_x
= dp
->dccps_mss_cache
;
348 hctx
->ccid3hctx_x
<<= 6;
350 ccid3_update_send_interval(hctx
);
352 ccid3_hc_tx_set_state(sk
, TFRC_SSTATE_NO_FBACK
);
354 case TFRC_SSTATE_NO_FBACK
:
355 case TFRC_SSTATE_FBACK
:
356 delay
= ktime_us_delta(hctx
->ccid3hctx_t_nom
, now
);
357 ccid3_pr_debug("delay=%ld\n", (long)delay
);
359 * Scheduling of packet transmissions [RFC 3448, 4.6]
361 * if (t_now > t_nom - delta)
362 * // send the packet now
364 * // send the packet in (t_nom - t_now) milliseconds.
366 if (delay
- (s64
)hctx
->ccid3hctx_delta
>= 1000)
367 return (u32
)delay
/ 1000L;
369 ccid3_hc_tx_update_win_count(hctx
, now
);
371 case TFRC_SSTATE_TERM
:
372 DCCP_BUG("%s(%p) - Illegal state TERM", dccp_role(sk
), sk
);
376 /* prepare to send now (add options etc.) */
377 dp
->dccps_hc_tx_insert_options
= 1;
378 DCCP_SKB_CB(skb
)->dccpd_ccval
= hctx
->ccid3hctx_last_win_count
;
379 hctx
->ccid3hctx_idle
= 0;
381 /* set the nominal send time for the next following packet */
382 hctx
->ccid3hctx_t_nom
= ktime_add_us(hctx
->ccid3hctx_t_nom
,
383 hctx
->ccid3hctx_t_ipi
);
387 static void ccid3_hc_tx_packet_sent(struct sock
*sk
, int more
,
390 struct ccid3_hc_tx_sock
*hctx
= ccid3_hc_tx_sk(sk
);
392 struct dccp_tx_hist_entry
*packet
;
394 BUG_ON(hctx
== NULL
);
396 ccid3_hc_tx_update_s(hctx
, len
);
398 packet
= dccp_tx_hist_entry_new(ccid3_tx_hist
, GFP_ATOMIC
);
399 if (unlikely(packet
== NULL
)) {
400 DCCP_CRIT("packet history - out of memory!");
403 dccp_tx_hist_add_entry(&hctx
->ccid3hctx_hist
, packet
);
405 dccp_timestamp(sk
, &now
);
406 packet
->dccphtx_tstamp
= now
;
407 packet
->dccphtx_seqno
= dccp_sk(sk
)->dccps_gss
;
408 packet
->dccphtx_rtt
= hctx
->ccid3hctx_rtt
;
409 packet
->dccphtx_sent
= 1;
412 static void ccid3_hc_tx_packet_recv(struct sock
*sk
, struct sk_buff
*skb
)
414 struct ccid3_hc_tx_sock
*hctx
= ccid3_hc_tx_sk(sk
);
415 struct ccid3_options_received
*opt_recv
;
416 struct dccp_tx_hist_entry
*packet
;
421 BUG_ON(hctx
== NULL
);
423 /* we are only interested in ACKs */
424 if (!(DCCP_SKB_CB(skb
)->dccpd_type
== DCCP_PKT_ACK
||
425 DCCP_SKB_CB(skb
)->dccpd_type
== DCCP_PKT_DATAACK
))
428 opt_recv
= &hctx
->ccid3hctx_options_received
;
430 switch (hctx
->ccid3hctx_state
) {
431 case TFRC_SSTATE_NO_FBACK
:
432 case TFRC_SSTATE_FBACK
:
433 /* get packet from history to look up t_recvdata */
434 packet
= dccp_tx_hist_find_entry(&hctx
->ccid3hctx_hist
,
435 DCCP_SKB_CB(skb
)->dccpd_ack_seq
);
436 if (unlikely(packet
== NULL
)) {
437 DCCP_WARN("%s(%p), seqno %llu(%s) doesn't exist "
438 "in history!\n", dccp_role(sk
), sk
,
439 (unsigned long long)DCCP_SKB_CB(skb
)->dccpd_ack_seq
,
440 dccp_packet_name(DCCP_SKB_CB(skb
)->dccpd_type
));
444 /* Update receive rate in units of 64 * bytes/second */
445 hctx
->ccid3hctx_x_recv
= opt_recv
->ccid3or_receive_rate
;
446 hctx
->ccid3hctx_x_recv
<<= 6;
448 /* Update loss event rate */
449 pinv
= opt_recv
->ccid3or_loss_event_rate
;
450 if (pinv
== ~0U || pinv
== 0) /* see RFC 4342, 8.5 */
451 hctx
->ccid3hctx_p
= 0;
452 else /* can not exceed 100% */
453 hctx
->ccid3hctx_p
= 1000000 / pinv
;
455 dccp_timestamp(sk
, &now
);
458 * Calculate new round trip sample as per [RFC 3448, 4.3] by
459 * R_sample = (now - t_recvdata) - t_elapsed
461 r_sample
= dccp_sample_rtt(sk
, &now
, &packet
->dccphtx_tstamp
);
464 * Update RTT estimate by
465 * If (No feedback recv)
468 * R = q * R + (1 - q) * R_sample;
470 * q is a constant, RFC 3448 recomments 0.9
472 if (hctx
->ccid3hctx_state
== TFRC_SSTATE_NO_FBACK
) {
474 * Larger Initial Windows [RFC 4342, sec. 5]
476 hctx
->ccid3hctx_rtt
= r_sample
;
477 hctx
->ccid3hctx_x
= rfc3390_initial_rate(sk
);
478 hctx
->ccid3hctx_t_ld
= timeval_to_ktime(now
);
480 ccid3_update_send_interval(hctx
);
482 ccid3_pr_debug("%s(%p), s=%u, MSS=%u, "
483 "R_sample=%uus, X=%u\n", dccp_role(sk
),
484 sk
, hctx
->ccid3hctx_s
,
485 dccp_sk(sk
)->dccps_mss_cache
, r_sample
,
486 (unsigned)(hctx
->ccid3hctx_x
>> 6));
488 ccid3_hc_tx_set_state(sk
, TFRC_SSTATE_FBACK
);
490 hctx
->ccid3hctx_rtt
= (9 * hctx
->ccid3hctx_rtt
+
493 /* Update sending rate (step 4 of [RFC 3448, 4.3]) */
494 if (hctx
->ccid3hctx_p
> 0)
495 hctx
->ccid3hctx_x_calc
=
496 tfrc_calc_x(hctx
->ccid3hctx_s
,
499 ccid3_hc_tx_update_x(sk
);
501 ccid3_pr_debug("%s(%p), RTT=%uus (sample=%uus), s=%u, "
502 "p=%u, X_calc=%u, X_recv=%u, X=%u\n",
504 sk
, hctx
->ccid3hctx_rtt
, r_sample
,
505 hctx
->ccid3hctx_s
, hctx
->ccid3hctx_p
,
506 hctx
->ccid3hctx_x_calc
,
507 (unsigned)(hctx
->ccid3hctx_x_recv
>> 6),
508 (unsigned)(hctx
->ccid3hctx_x
>> 6));
511 /* unschedule no feedback timer */
512 sk_stop_timer(sk
, &hctx
->ccid3hctx_no_feedback_timer
);
514 /* remove all packets older than the one acked from history */
515 dccp_tx_hist_purge_older(ccid3_tx_hist
,
516 &hctx
->ccid3hctx_hist
, packet
);
518 * As we have calculated new ipi, delta, t_nom it is possible
519 * that we now can send a packet, so wake up dccp_wait_for_ccid
521 sk
->sk_write_space(sk
);
524 * Update timeout interval for the nofeedback timer.
525 * We use a configuration option to increase the lower bound.
526 * This can help avoid triggering the nofeedback timer too
527 * often ('spinning') on LANs with small RTTs.
529 hctx
->ccid3hctx_t_rto
= max_t(u32
, 4 * hctx
->ccid3hctx_rtt
,
530 CONFIG_IP_DCCP_CCID3_RTO
*
531 (USEC_PER_SEC
/1000));
533 * Schedule no feedback timer to expire in
534 * max(t_RTO, 2 * s/X) = max(t_RTO, 2 * t_ipi)
536 t_nfb
= max(hctx
->ccid3hctx_t_rto
, 2 * hctx
->ccid3hctx_t_ipi
);
538 ccid3_pr_debug("%s(%p), Scheduled no feedback timer to "
539 "expire in %lu jiffies (%luus)\n",
541 sk
, usecs_to_jiffies(t_nfb
), t_nfb
);
543 sk_reset_timer(sk
, &hctx
->ccid3hctx_no_feedback_timer
,
544 jiffies
+ usecs_to_jiffies(t_nfb
));
547 hctx
->ccid3hctx_idle
= 1;
549 case TFRC_SSTATE_NO_SENT
: /* fall through */
550 case TFRC_SSTATE_TERM
: /* ignore feedback when closing */
555 static int ccid3_hc_tx_parse_options(struct sock
*sk
, unsigned char option
,
556 unsigned char len
, u16 idx
,
557 unsigned char *value
)
560 const struct dccp_sock
*dp
= dccp_sk(sk
);
561 struct ccid3_hc_tx_sock
*hctx
= ccid3_hc_tx_sk(sk
);
562 struct ccid3_options_received
*opt_recv
;
564 BUG_ON(hctx
== NULL
);
566 opt_recv
= &hctx
->ccid3hctx_options_received
;
568 if (opt_recv
->ccid3or_seqno
!= dp
->dccps_gsr
) {
569 opt_recv
->ccid3or_seqno
= dp
->dccps_gsr
;
570 opt_recv
->ccid3or_loss_event_rate
= ~0;
571 opt_recv
->ccid3or_loss_intervals_idx
= 0;
572 opt_recv
->ccid3or_loss_intervals_len
= 0;
573 opt_recv
->ccid3or_receive_rate
= 0;
577 case TFRC_OPT_LOSS_EVENT_RATE
:
578 if (unlikely(len
!= 4)) {
579 DCCP_WARN("%s(%p), invalid len %d "
580 "for TFRC_OPT_LOSS_EVENT_RATE\n",
581 dccp_role(sk
), sk
, len
);
584 opt_recv
->ccid3or_loss_event_rate
=
585 ntohl(*(__be32
*)value
);
586 ccid3_pr_debug("%s(%p), LOSS_EVENT_RATE=%u\n",
588 opt_recv
->ccid3or_loss_event_rate
);
591 case TFRC_OPT_LOSS_INTERVALS
:
592 opt_recv
->ccid3or_loss_intervals_idx
= idx
;
593 opt_recv
->ccid3or_loss_intervals_len
= len
;
594 ccid3_pr_debug("%s(%p), LOSS_INTERVALS=(%u, %u)\n",
596 opt_recv
->ccid3or_loss_intervals_idx
,
597 opt_recv
->ccid3or_loss_intervals_len
);
599 case TFRC_OPT_RECEIVE_RATE
:
600 if (unlikely(len
!= 4)) {
601 DCCP_WARN("%s(%p), invalid len %d "
602 "for TFRC_OPT_RECEIVE_RATE\n",
603 dccp_role(sk
), sk
, len
);
606 opt_recv
->ccid3or_receive_rate
=
607 ntohl(*(__be32
*)value
);
608 ccid3_pr_debug("%s(%p), RECEIVE_RATE=%u\n",
610 opt_recv
->ccid3or_receive_rate
);
618 static int ccid3_hc_tx_init(struct ccid
*ccid
, struct sock
*sk
)
620 struct ccid3_hc_tx_sock
*hctx
= ccid_priv(ccid
);
622 hctx
->ccid3hctx_s
= 0;
623 hctx
->ccid3hctx_rtt
= 0;
624 hctx
->ccid3hctx_state
= TFRC_SSTATE_NO_SENT
;
625 INIT_LIST_HEAD(&hctx
->ccid3hctx_hist
);
627 hctx
->ccid3hctx_no_feedback_timer
.function
=
628 ccid3_hc_tx_no_feedback_timer
;
629 hctx
->ccid3hctx_no_feedback_timer
.data
= (unsigned long)sk
;
630 init_timer(&hctx
->ccid3hctx_no_feedback_timer
);
635 static void ccid3_hc_tx_exit(struct sock
*sk
)
637 struct ccid3_hc_tx_sock
*hctx
= ccid3_hc_tx_sk(sk
);
639 BUG_ON(hctx
== NULL
);
641 ccid3_hc_tx_set_state(sk
, TFRC_SSTATE_TERM
);
642 sk_stop_timer(sk
, &hctx
->ccid3hctx_no_feedback_timer
);
644 /* Empty packet history */
645 dccp_tx_hist_purge(ccid3_tx_hist
, &hctx
->ccid3hctx_hist
);
648 static void ccid3_hc_tx_get_info(struct sock
*sk
, struct tcp_info
*info
)
650 const struct ccid3_hc_tx_sock
*hctx
= ccid3_hc_tx_sk(sk
);
652 /* Listen socks doesn't have a private CCID block */
653 if (sk
->sk_state
== DCCP_LISTEN
)
656 BUG_ON(hctx
== NULL
);
658 info
->tcpi_rto
= hctx
->ccid3hctx_t_rto
;
659 info
->tcpi_rtt
= hctx
->ccid3hctx_rtt
;
662 static int ccid3_hc_tx_getsockopt(struct sock
*sk
, const int optname
, int len
,
663 u32 __user
*optval
, int __user
*optlen
)
665 const struct ccid3_hc_tx_sock
*hctx
= ccid3_hc_tx_sk(sk
);
668 /* Listen socks doesn't have a private CCID block */
669 if (sk
->sk_state
== DCCP_LISTEN
)
673 case DCCP_SOCKOPT_CCID_TX_INFO
:
674 if (len
< sizeof(hctx
->ccid3hctx_tfrc
))
676 len
= sizeof(hctx
->ccid3hctx_tfrc
);
677 val
= &hctx
->ccid3hctx_tfrc
;
683 if (put_user(len
, optlen
) || copy_to_user(optval
, val
, len
))
690 * Receiver Half-Connection Routines
692 #ifdef CONFIG_IP_DCCP_CCID3_DEBUG
693 static const char *ccid3_rx_state_name(enum ccid3_hc_rx_states state
)
695 static char *ccid3_rx_state_names
[] = {
696 [TFRC_RSTATE_NO_DATA
] = "NO_DATA",
697 [TFRC_RSTATE_DATA
] = "DATA",
698 [TFRC_RSTATE_TERM
] = "TERM",
701 return ccid3_rx_state_names
[state
];
705 static void ccid3_hc_rx_set_state(struct sock
*sk
,
706 enum ccid3_hc_rx_states state
)
708 struct ccid3_hc_rx_sock
*hcrx
= ccid3_hc_rx_sk(sk
);
709 enum ccid3_hc_rx_states oldstate
= hcrx
->ccid3hcrx_state
;
711 ccid3_pr_debug("%s(%p) %-8.8s -> %s\n",
712 dccp_role(sk
), sk
, ccid3_rx_state_name(oldstate
),
713 ccid3_rx_state_name(state
));
714 WARN_ON(state
== oldstate
);
715 hcrx
->ccid3hcrx_state
= state
;
718 static inline void ccid3_hc_rx_update_s(struct ccid3_hc_rx_sock
*hcrx
, int len
)
720 if (unlikely(len
== 0)) /* don't update on empty packets (e.g. ACKs) */
721 ccid3_pr_debug("Packet payload length is 0 - not updating\n");
723 hcrx
->ccid3hcrx_s
= hcrx
->ccid3hcrx_s
== 0 ? len
:
724 (9 * hcrx
->ccid3hcrx_s
+ len
) / 10;
727 static void ccid3_hc_rx_send_feedback(struct sock
*sk
)
729 struct ccid3_hc_rx_sock
*hcrx
= ccid3_hc_rx_sk(sk
);
730 struct dccp_sock
*dp
= dccp_sk(sk
);
731 struct dccp_rx_hist_entry
*packet
;
735 ccid3_pr_debug("%s(%p) - entry \n", dccp_role(sk
), sk
);
737 dccp_timestamp(sk
, &now
);
739 switch (hcrx
->ccid3hcrx_state
) {
740 case TFRC_RSTATE_NO_DATA
:
741 hcrx
->ccid3hcrx_x_recv
= 0;
743 case TFRC_RSTATE_DATA
:
744 delta
= timeval_delta(&now
,
745 &hcrx
->ccid3hcrx_tstamp_last_feedback
);
746 DCCP_BUG_ON(delta
< 0);
747 hcrx
->ccid3hcrx_x_recv
=
748 scaled_div32(hcrx
->ccid3hcrx_bytes_recv
, delta
);
750 case TFRC_RSTATE_TERM
:
751 DCCP_BUG("%s(%p) - Illegal state TERM", dccp_role(sk
), sk
);
755 packet
= dccp_rx_hist_find_data_packet(&hcrx
->ccid3hcrx_hist
);
756 if (unlikely(packet
== NULL
)) {
757 DCCP_WARN("%s(%p), no data packet in history!\n",
762 hcrx
->ccid3hcrx_tstamp_last_feedback
= now
;
763 hcrx
->ccid3hcrx_ccval_last_counter
= packet
->dccphrx_ccval
;
764 hcrx
->ccid3hcrx_bytes_recv
= 0;
766 /* Elapsed time information [RFC 4340, 13.2] in units of 10 * usecs */
767 delta
= timeval_delta(&now
, &packet
->dccphrx_tstamp
);
768 DCCP_BUG_ON(delta
< 0);
769 hcrx
->ccid3hcrx_elapsed_time
= delta
/ 10;
771 if (hcrx
->ccid3hcrx_p
== 0)
772 hcrx
->ccid3hcrx_pinv
= ~0U; /* see RFC 4342, 8.5 */
773 else if (hcrx
->ccid3hcrx_p
> 1000000) {
774 DCCP_WARN("p (%u) > 100%%\n", hcrx
->ccid3hcrx_p
);
775 hcrx
->ccid3hcrx_pinv
= 1; /* use 100% in this case */
777 hcrx
->ccid3hcrx_pinv
= 1000000 / hcrx
->ccid3hcrx_p
;
779 dp
->dccps_hc_rx_insert_options
= 1;
783 static int ccid3_hc_rx_insert_options(struct sock
*sk
, struct sk_buff
*skb
)
785 const struct ccid3_hc_rx_sock
*hcrx
= ccid3_hc_rx_sk(sk
);
788 BUG_ON(hcrx
== NULL
);
790 if (!(sk
->sk_state
== DCCP_OPEN
|| sk
->sk_state
== DCCP_PARTOPEN
))
793 DCCP_SKB_CB(skb
)->dccpd_ccval
= hcrx
->ccid3hcrx_ccval_last_counter
;
795 if (dccp_packet_without_ack(skb
))
798 x_recv
= htonl(hcrx
->ccid3hcrx_x_recv
);
799 pinv
= htonl(hcrx
->ccid3hcrx_pinv
);
801 if ((hcrx
->ccid3hcrx_elapsed_time
!= 0 &&
802 dccp_insert_option_elapsed_time(sk
, skb
,
803 hcrx
->ccid3hcrx_elapsed_time
)) ||
804 dccp_insert_option_timestamp(sk
, skb
) ||
805 dccp_insert_option(sk
, skb
, TFRC_OPT_LOSS_EVENT_RATE
,
806 &pinv
, sizeof(pinv
)) ||
807 dccp_insert_option(sk
, skb
, TFRC_OPT_RECEIVE_RATE
,
808 &x_recv
, sizeof(x_recv
)))
814 static int ccid3_hc_rx_detect_loss(struct sock
*sk
,
815 struct dccp_rx_hist_entry
*packet
)
817 struct ccid3_hc_rx_sock
*hcrx
= ccid3_hc_rx_sk(sk
);
818 struct dccp_rx_hist_entry
*rx_hist
=
819 dccp_rx_hist_head(&hcrx
->ccid3hcrx_hist
);
820 u64 seqno
= packet
->dccphrx_seqno
;
826 tmp_seqno
= hcrx
->ccid3hcrx_seqno_nonloss
;
829 follows48(packet
->dccphrx_seqno
, hcrx
->ccid3hcrx_seqno_nonloss
)) {
830 hcrx
->ccid3hcrx_seqno_nonloss
= seqno
;
831 hcrx
->ccid3hcrx_ccval_nonloss
= packet
->dccphrx_ccval
;
836 while (dccp_delta_seqno(hcrx
->ccid3hcrx_seqno_nonloss
, seqno
)
837 > TFRC_RECV_NUM_LATE_LOSS
) {
839 dccp_li_update_li(sk
,
840 &hcrx
->ccid3hcrx_li_hist
,
841 &hcrx
->ccid3hcrx_hist
,
842 &hcrx
->ccid3hcrx_tstamp_last_feedback
,
844 hcrx
->ccid3hcrx_bytes_recv
,
845 hcrx
->ccid3hcrx_x_recv
,
846 hcrx
->ccid3hcrx_seqno_nonloss
,
847 hcrx
->ccid3hcrx_ccval_nonloss
);
848 tmp_seqno
= hcrx
->ccid3hcrx_seqno_nonloss
;
849 dccp_inc_seqno(&tmp_seqno
);
850 hcrx
->ccid3hcrx_seqno_nonloss
= tmp_seqno
;
851 dccp_inc_seqno(&tmp_seqno
);
852 while (dccp_rx_hist_find_entry(&hcrx
->ccid3hcrx_hist
,
853 tmp_seqno
, &ccval
)) {
854 hcrx
->ccid3hcrx_seqno_nonloss
= tmp_seqno
;
855 hcrx
->ccid3hcrx_ccval_nonloss
= ccval
;
856 dccp_inc_seqno(&tmp_seqno
);
860 /* FIXME - this code could be simplified with above while */
861 /* but works at moment */
862 if (follows48(packet
->dccphrx_seqno
, hcrx
->ccid3hcrx_seqno_nonloss
)) {
863 hcrx
->ccid3hcrx_seqno_nonloss
= seqno
;
864 hcrx
->ccid3hcrx_ccval_nonloss
= packet
->dccphrx_ccval
;
868 dccp_rx_hist_add_packet(ccid3_rx_hist
, &hcrx
->ccid3hcrx_hist
,
869 &hcrx
->ccid3hcrx_li_hist
, packet
,
870 hcrx
->ccid3hcrx_seqno_nonloss
);
874 static void ccid3_hc_rx_packet_recv(struct sock
*sk
, struct sk_buff
*skb
)
876 struct ccid3_hc_rx_sock
*hcrx
= ccid3_hc_rx_sk(sk
);
877 const struct dccp_options_received
*opt_recv
;
878 struct dccp_rx_hist_entry
*packet
;
880 u32 p_prev
, r_sample
, rtt_prev
;
881 int loss
, payload_size
;
883 BUG_ON(hcrx
== NULL
);
885 opt_recv
= &dccp_sk(sk
)->dccps_options_received
;
887 switch (DCCP_SKB_CB(skb
)->dccpd_type
) {
889 if (hcrx
->ccid3hcrx_state
== TFRC_RSTATE_NO_DATA
)
891 case DCCP_PKT_DATAACK
:
892 if (opt_recv
->dccpor_timestamp_echo
== 0)
894 rtt_prev
= hcrx
->ccid3hcrx_rtt
;
895 dccp_timestamp(sk
, &now
);
896 r_sample
= dccp_sample_rtt(sk
, &now
, NULL
);
898 if (hcrx
->ccid3hcrx_state
== TFRC_RSTATE_NO_DATA
)
899 hcrx
->ccid3hcrx_rtt
= r_sample
;
901 hcrx
->ccid3hcrx_rtt
= (hcrx
->ccid3hcrx_rtt
* 9) / 10 +
904 if (rtt_prev
!= hcrx
->ccid3hcrx_rtt
)
905 ccid3_pr_debug("%s(%p), New RTT=%uus, elapsed time=%u\n",
906 dccp_role(sk
), sk
, hcrx
->ccid3hcrx_rtt
,
907 opt_recv
->dccpor_elapsed_time
);
911 default: /* We're not interested in other packet types, move along */
915 packet
= dccp_rx_hist_entry_new(ccid3_rx_hist
, sk
, opt_recv
->dccpor_ndp
,
917 if (unlikely(packet
== NULL
)) {
918 DCCP_WARN("%s(%p), Not enough mem to add rx packet "
919 "to history, consider it lost!\n", dccp_role(sk
), sk
);
923 loss
= ccid3_hc_rx_detect_loss(sk
, packet
);
925 if (DCCP_SKB_CB(skb
)->dccpd_type
== DCCP_PKT_ACK
)
928 payload_size
= skb
->len
- dccp_hdr(skb
)->dccph_doff
* 4;
929 ccid3_hc_rx_update_s(hcrx
, payload_size
);
931 switch (hcrx
->ccid3hcrx_state
) {
932 case TFRC_RSTATE_NO_DATA
:
933 ccid3_pr_debug("%s(%p, state=%s), skb=%p, sending initial "
934 "feedback\n", dccp_role(sk
), sk
,
935 dccp_state_name(sk
->sk_state
), skb
);
936 ccid3_hc_rx_send_feedback(sk
);
937 ccid3_hc_rx_set_state(sk
, TFRC_RSTATE_DATA
);
939 case TFRC_RSTATE_DATA
:
940 hcrx
->ccid3hcrx_bytes_recv
+= payload_size
;
944 dccp_timestamp(sk
, &now
);
945 if ((timeval_delta(&now
, &hcrx
->ccid3hcrx_tstamp_last_ack
) -
946 (suseconds_t
)hcrx
->ccid3hcrx_rtt
) >= 0) {
947 hcrx
->ccid3hcrx_tstamp_last_ack
= now
;
948 ccid3_hc_rx_send_feedback(sk
);
951 case TFRC_RSTATE_TERM
:
952 DCCP_BUG("%s(%p) - Illegal state TERM", dccp_role(sk
), sk
);
956 /* Dealing with packet loss */
957 ccid3_pr_debug("%s(%p, state=%s), data loss! Reacting...\n",
958 dccp_role(sk
), sk
, dccp_state_name(sk
->sk_state
));
960 p_prev
= hcrx
->ccid3hcrx_p
;
962 /* Calculate loss event rate */
963 if (!list_empty(&hcrx
->ccid3hcrx_li_hist
)) {
964 u32 i_mean
= dccp_li_hist_calc_i_mean(&hcrx
->ccid3hcrx_li_hist
);
966 /* Scaling up by 1000000 as fixed decimal */
968 hcrx
->ccid3hcrx_p
= 1000000 / i_mean
;
970 DCCP_BUG("empty loss history");
972 if (hcrx
->ccid3hcrx_p
> p_prev
) {
973 ccid3_hc_rx_send_feedback(sk
);
978 static int ccid3_hc_rx_init(struct ccid
*ccid
, struct sock
*sk
)
980 struct ccid3_hc_rx_sock
*hcrx
= ccid_priv(ccid
);
982 ccid3_pr_debug("entry\n");
984 hcrx
->ccid3hcrx_state
= TFRC_RSTATE_NO_DATA
;
985 INIT_LIST_HEAD(&hcrx
->ccid3hcrx_hist
);
986 INIT_LIST_HEAD(&hcrx
->ccid3hcrx_li_hist
);
987 dccp_timestamp(sk
, &hcrx
->ccid3hcrx_tstamp_last_ack
);
988 hcrx
->ccid3hcrx_tstamp_last_feedback
= hcrx
->ccid3hcrx_tstamp_last_ack
;
989 hcrx
->ccid3hcrx_s
= 0;
990 hcrx
->ccid3hcrx_rtt
= 0;
994 static void ccid3_hc_rx_exit(struct sock
*sk
)
996 struct ccid3_hc_rx_sock
*hcrx
= ccid3_hc_rx_sk(sk
);
998 BUG_ON(hcrx
== NULL
);
1000 ccid3_hc_rx_set_state(sk
, TFRC_RSTATE_TERM
);
1002 /* Empty packet history */
1003 dccp_rx_hist_purge(ccid3_rx_hist
, &hcrx
->ccid3hcrx_hist
);
1005 /* Empty loss interval history */
1006 dccp_li_hist_purge(&hcrx
->ccid3hcrx_li_hist
);
1009 static void ccid3_hc_rx_get_info(struct sock
*sk
, struct tcp_info
*info
)
1011 const struct ccid3_hc_rx_sock
*hcrx
= ccid3_hc_rx_sk(sk
);
1013 /* Listen socks doesn't have a private CCID block */
1014 if (sk
->sk_state
== DCCP_LISTEN
)
1017 BUG_ON(hcrx
== NULL
);
1019 info
->tcpi_ca_state
= hcrx
->ccid3hcrx_state
;
1020 info
->tcpi_options
|= TCPI_OPT_TIMESTAMPS
;
1021 info
->tcpi_rcv_rtt
= hcrx
->ccid3hcrx_rtt
;
1024 static int ccid3_hc_rx_getsockopt(struct sock
*sk
, const int optname
, int len
,
1025 u32 __user
*optval
, int __user
*optlen
)
1027 const struct ccid3_hc_rx_sock
*hcrx
= ccid3_hc_rx_sk(sk
);
1030 /* Listen socks doesn't have a private CCID block */
1031 if (sk
->sk_state
== DCCP_LISTEN
)
1035 case DCCP_SOCKOPT_CCID_RX_INFO
:
1036 if (len
< sizeof(hcrx
->ccid3hcrx_tfrc
))
1038 len
= sizeof(hcrx
->ccid3hcrx_tfrc
);
1039 val
= &hcrx
->ccid3hcrx_tfrc
;
1042 return -ENOPROTOOPT
;
1045 if (put_user(len
, optlen
) || copy_to_user(optval
, val
, len
))
1051 static struct ccid_operations ccid3
= {
1052 .ccid_id
= DCCPC_CCID3
,
1053 .ccid_name
= "ccid3",
1054 .ccid_owner
= THIS_MODULE
,
1055 .ccid_hc_tx_obj_size
= sizeof(struct ccid3_hc_tx_sock
),
1056 .ccid_hc_tx_init
= ccid3_hc_tx_init
,
1057 .ccid_hc_tx_exit
= ccid3_hc_tx_exit
,
1058 .ccid_hc_tx_send_packet
= ccid3_hc_tx_send_packet
,
1059 .ccid_hc_tx_packet_sent
= ccid3_hc_tx_packet_sent
,
1060 .ccid_hc_tx_packet_recv
= ccid3_hc_tx_packet_recv
,
1061 .ccid_hc_tx_parse_options
= ccid3_hc_tx_parse_options
,
1062 .ccid_hc_rx_obj_size
= sizeof(struct ccid3_hc_rx_sock
),
1063 .ccid_hc_rx_init
= ccid3_hc_rx_init
,
1064 .ccid_hc_rx_exit
= ccid3_hc_rx_exit
,
1065 .ccid_hc_rx_insert_options
= ccid3_hc_rx_insert_options
,
1066 .ccid_hc_rx_packet_recv
= ccid3_hc_rx_packet_recv
,
1067 .ccid_hc_rx_get_info
= ccid3_hc_rx_get_info
,
1068 .ccid_hc_tx_get_info
= ccid3_hc_tx_get_info
,
1069 .ccid_hc_rx_getsockopt
= ccid3_hc_rx_getsockopt
,
1070 .ccid_hc_tx_getsockopt
= ccid3_hc_tx_getsockopt
,
1073 #ifdef CONFIG_IP_DCCP_CCID3_DEBUG
1074 module_param(ccid3_debug
, int, 0444);
1075 MODULE_PARM_DESC(ccid3_debug
, "Enable debug messages");
1078 static __init
int ccid3_module_init(void)
1082 ccid3_rx_hist
= dccp_rx_hist_new("ccid3");
1083 if (ccid3_rx_hist
== NULL
)
1086 ccid3_tx_hist
= dccp_tx_hist_new("ccid3");
1087 if (ccid3_tx_hist
== NULL
)
1090 rc
= ccid_register(&ccid3
);
1097 dccp_tx_hist_delete(ccid3_tx_hist
);
1098 ccid3_tx_hist
= NULL
;
1100 dccp_rx_hist_delete(ccid3_rx_hist
);
1101 ccid3_rx_hist
= NULL
;
1104 module_init(ccid3_module_init
);
1106 static __exit
void ccid3_module_exit(void)
1108 ccid_unregister(&ccid3
);
1110 if (ccid3_tx_hist
!= NULL
) {
1111 dccp_tx_hist_delete(ccid3_tx_hist
);
1112 ccid3_tx_hist
= NULL
;
1114 if (ccid3_rx_hist
!= NULL
) {
1115 dccp_rx_hist_delete(ccid3_rx_hist
);
1116 ccid3_rx_hist
= NULL
;
1119 module_exit(ccid3_module_exit
);
1121 MODULE_AUTHOR("Ian McDonald <ian.mcdonald@jandi.co.nz>, "
1122 "Arnaldo Carvalho de Melo <acme@ghostprotocols.net>");
1123 MODULE_DESCRIPTION("DCCP TFRC CCID3 CCID");
1124 MODULE_LICENSE("GPL");
1125 MODULE_ALIAS("net-dccp-ccid-3");