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 #include <asm/unaligned.h>
45 #ifdef CONFIG_IP_DCCP_CCID3_DEBUG
46 static int ccid3_debug
;
47 #define ccid3_pr_debug(format, a...) DCCP_PR_DEBUG(ccid3_debug, format, ##a)
49 #define ccid3_pr_debug(format, a...)
53 * Transmitter Half-Connection Routines
55 #ifdef CONFIG_IP_DCCP_CCID3_DEBUG
56 static const char *ccid3_tx_state_name(enum ccid3_hc_tx_states state
)
58 static char *ccid3_state_names
[] = {
59 [TFRC_SSTATE_NO_SENT
] = "NO_SENT",
60 [TFRC_SSTATE_NO_FBACK
] = "NO_FBACK",
61 [TFRC_SSTATE_FBACK
] = "FBACK",
62 [TFRC_SSTATE_TERM
] = "TERM",
65 return ccid3_state_names
[state
];
69 static void ccid3_hc_tx_set_state(struct sock
*sk
,
70 enum ccid3_hc_tx_states state
)
72 struct ccid3_hc_tx_sock
*hctx
= ccid3_hc_tx_sk(sk
);
73 enum ccid3_hc_tx_states oldstate
= hctx
->ccid3hctx_state
;
75 ccid3_pr_debug("%s(%p) %-8.8s -> %s\n",
76 dccp_role(sk
), sk
, ccid3_tx_state_name(oldstate
),
77 ccid3_tx_state_name(state
));
78 WARN_ON(state
== oldstate
);
79 hctx
->ccid3hctx_state
= state
;
83 * Compute the initial sending rate X_init in the manner of RFC 3390:
85 * X_init = min(4 * s, max(2 * s, 4380 bytes)) / RTT
87 * Note that RFC 3390 uses MSS, RFC 4342 refers to RFC 3390, and rfc3448bis
88 * (rev-02) clarifies the use of RFC 3390 with regard to the above formula.
89 * For consistency with other parts of the code, X_init is scaled by 2^6.
91 static inline u64
rfc3390_initial_rate(struct sock
*sk
)
93 const struct ccid3_hc_tx_sock
*hctx
= ccid3_hc_tx_sk(sk
);
94 const __u32 w_init
= min_t(__u32
, 4 * hctx
->ccid3hctx_s
,
95 max_t(__u32
, 2 * hctx
->ccid3hctx_s
, 4380));
97 return scaled_div(w_init
<< 6, hctx
->ccid3hctx_rtt
);
101 * Recalculate t_ipi and delta (should be called whenever X changes)
103 static inline void ccid3_update_send_interval(struct ccid3_hc_tx_sock
*hctx
)
105 /* Calculate new t_ipi = s / X_inst (X_inst is in 64 * bytes/second) */
106 hctx
->ccid3hctx_t_ipi
= scaled_div32(((u64
)hctx
->ccid3hctx_s
) << 6,
109 /* Calculate new delta by delta = min(t_ipi / 2, t_gran / 2) */
110 hctx
->ccid3hctx_delta
= min_t(u32
, hctx
->ccid3hctx_t_ipi
/ 2,
111 TFRC_OPSYS_HALF_TIME_GRAN
);
113 ccid3_pr_debug("t_ipi=%u, delta=%u, s=%u, X=%u\n",
114 hctx
->ccid3hctx_t_ipi
, hctx
->ccid3hctx_delta
,
115 hctx
->ccid3hctx_s
, (unsigned)(hctx
->ccid3hctx_x
>> 6));
119 static u32
ccid3_hc_tx_idle_rtt(struct ccid3_hc_tx_sock
*hctx
, ktime_t now
)
121 u32 delta
= ktime_us_delta(now
, hctx
->ccid3hctx_t_last_win_count
);
123 return delta
/ hctx
->ccid3hctx_rtt
;
127 * ccid3_hc_tx_update_x - Update allowed sending rate X
128 * @stamp: most recent time if available - can be left NULL.
129 * This function tracks draft rfc3448bis, check there for latest details.
131 * Note: X and X_recv are both stored in units of 64 * bytes/second, to support
132 * fine-grained resolution of sending rates. This requires scaling by 2^6
133 * throughout the code. Only X_calc is unscaled (in bytes/second).
136 static void ccid3_hc_tx_update_x(struct sock
*sk
, ktime_t
*stamp
)
139 struct ccid3_hc_tx_sock
*hctx
= ccid3_hc_tx_sk(sk
);
140 __u64 min_rate
= 2 * hctx
->ccid3hctx_x_recv
;
141 const __u64 old_x
= hctx
->ccid3hctx_x
;
142 ktime_t now
= stamp
? *stamp
: ktime_get_real();
145 * Handle IDLE periods: do not reduce below RFC3390 initial sending rate
146 * when idling [RFC 4342, 5.1]. Definition of idling is from rfc3448bis:
147 * a sender is idle if it has not sent anything over a 2-RTT-period.
148 * For consistency with X and X_recv, min_rate is also scaled by 2^6.
150 if (ccid3_hc_tx_idle_rtt(hctx
, now
) >= 2) {
151 min_rate
= rfc3390_initial_rate(sk
);
152 min_rate
= max(min_rate
, 2 * hctx
->ccid3hctx_x_recv
);
155 if (hctx
->ccid3hctx_p
> 0) {
157 hctx
->ccid3hctx_x
= min(((__u64
)hctx
->ccid3hctx_x_calc
) << 6,
159 hctx
->ccid3hctx_x
= max(hctx
->ccid3hctx_x
,
160 (((__u64
)hctx
->ccid3hctx_s
) << 6) /
163 } else if (ktime_us_delta(now
, hctx
->ccid3hctx_t_ld
)
164 - (s64
)hctx
->ccid3hctx_rtt
>= 0) {
167 max(min(2 * hctx
->ccid3hctx_x
, min_rate
),
168 scaled_div(((__u64
)hctx
->ccid3hctx_s
) << 6,
169 hctx
->ccid3hctx_rtt
));
170 hctx
->ccid3hctx_t_ld
= now
;
173 if (hctx
->ccid3hctx_x
!= old_x
) {
174 ccid3_pr_debug("X_prev=%u, X_now=%u, X_calc=%u, "
175 "X_recv=%u\n", (unsigned)(old_x
>> 6),
176 (unsigned)(hctx
->ccid3hctx_x
>> 6),
177 hctx
->ccid3hctx_x_calc
,
178 (unsigned)(hctx
->ccid3hctx_x_recv
>> 6));
180 ccid3_update_send_interval(hctx
);
185 * Track the mean packet size `s' (cf. RFC 4342, 5.3 and RFC 3448, 4.1)
186 * @len: DCCP packet payload size in bytes
188 static inline void ccid3_hc_tx_update_s(struct ccid3_hc_tx_sock
*hctx
, int len
)
190 const u16 old_s
= hctx
->ccid3hctx_s
;
192 hctx
->ccid3hctx_s
= tfrc_ewma(hctx
->ccid3hctx_s
, len
, 9);
194 if (hctx
->ccid3hctx_s
!= old_s
)
195 ccid3_update_send_interval(hctx
);
199 * Update Window Counter using the algorithm from [RFC 4342, 8.1].
200 * The algorithm is not applicable if RTT < 4 microseconds.
202 static inline void ccid3_hc_tx_update_win_count(struct ccid3_hc_tx_sock
*hctx
,
207 if (unlikely(hctx
->ccid3hctx_rtt
< 4)) /* avoid divide-by-zero */
210 quarter_rtts
= ktime_us_delta(now
, hctx
->ccid3hctx_t_last_win_count
);
211 quarter_rtts
/= hctx
->ccid3hctx_rtt
/ 4;
213 if (quarter_rtts
> 0) {
214 hctx
->ccid3hctx_t_last_win_count
= now
;
215 hctx
->ccid3hctx_last_win_count
+= min_t(u32
, quarter_rtts
, 5);
216 hctx
->ccid3hctx_last_win_count
&= 0xF; /* mod 16 */
220 static void ccid3_hc_tx_no_feedback_timer(unsigned long data
)
222 struct sock
*sk
= (struct sock
*)data
;
223 struct ccid3_hc_tx_sock
*hctx
= ccid3_hc_tx_sk(sk
);
224 unsigned long t_nfb
= USEC_PER_SEC
/ 5;
227 if (sock_owned_by_user(sk
)) {
228 /* Try again later. */
229 /* XXX: set some sensible MIB */
233 ccid3_pr_debug("%s(%p, state=%s) - entry \n", dccp_role(sk
), sk
,
234 ccid3_tx_state_name(hctx
->ccid3hctx_state
));
236 switch (hctx
->ccid3hctx_state
) {
237 case TFRC_SSTATE_NO_FBACK
:
238 /* RFC 3448, 4.4: Halve send rate directly */
239 hctx
->ccid3hctx_x
= max(hctx
->ccid3hctx_x
/ 2,
240 (((__u64
)hctx
->ccid3hctx_s
) << 6) /
243 ccid3_pr_debug("%s(%p, state=%s), updated tx rate to %u "
244 "bytes/s\n", dccp_role(sk
), sk
,
245 ccid3_tx_state_name(hctx
->ccid3hctx_state
),
246 (unsigned)(hctx
->ccid3hctx_x
>> 6));
247 /* The value of R is still undefined and so we can not recompute
248 * the timeout value. Keep initial value as per [RFC 4342, 5]. */
249 t_nfb
= TFRC_INITIAL_TIMEOUT
;
250 ccid3_update_send_interval(hctx
);
252 case TFRC_SSTATE_FBACK
:
254 * Modify the cached value of X_recv [RFC 3448, 4.4]
256 * If (p == 0 || X_calc > 2 * X_recv)
257 * X_recv = max(X_recv / 2, s / (2 * t_mbi));
259 * X_recv = X_calc / 4;
261 * Note that X_recv is scaled by 2^6 while X_calc is not
263 BUG_ON(hctx
->ccid3hctx_p
&& !hctx
->ccid3hctx_x_calc
);
265 if (hctx
->ccid3hctx_p
== 0 ||
266 (hctx
->ccid3hctx_x_calc
> (hctx
->ccid3hctx_x_recv
>> 5))) {
268 hctx
->ccid3hctx_x_recv
=
269 max(hctx
->ccid3hctx_x_recv
/ 2,
270 (((__u64
)hctx
->ccid3hctx_s
) << 6) /
273 hctx
->ccid3hctx_x_recv
= hctx
->ccid3hctx_x_calc
;
274 hctx
->ccid3hctx_x_recv
<<= 4;
276 /* Now recalculate X [RFC 3448, 4.3, step (4)] */
277 ccid3_hc_tx_update_x(sk
, NULL
);
279 * Schedule no feedback timer to expire in
280 * max(t_RTO, 2 * s/X) = max(t_RTO, 2 * t_ipi)
281 * See comments in packet_recv() regarding the value of t_RTO.
283 t_nfb
= max(hctx
->ccid3hctx_t_rto
, 2 * hctx
->ccid3hctx_t_ipi
);
285 case TFRC_SSTATE_NO_SENT
:
286 DCCP_BUG("%s(%p) - Illegal state NO_SENT", dccp_role(sk
), sk
);
288 case TFRC_SSTATE_TERM
:
293 sk_reset_timer(sk
, &hctx
->ccid3hctx_no_feedback_timer
,
294 jiffies
+ usecs_to_jiffies(t_nfb
));
302 * > 0: delay (in msecs) that should pass before actually sending
303 * = 0: can send immediately
304 * < 0: error condition; do not send packet
306 static int ccid3_hc_tx_send_packet(struct sock
*sk
, struct sk_buff
*skb
)
308 struct dccp_sock
*dp
= dccp_sk(sk
);
309 struct ccid3_hc_tx_sock
*hctx
= ccid3_hc_tx_sk(sk
);
310 ktime_t now
= ktime_get_real();
314 * This function is called only for Data and DataAck packets. Sending
315 * zero-sized Data(Ack)s is theoretically possible, but for congestion
316 * control this case is pathological - ignore it.
318 if (unlikely(skb
->len
== 0))
321 switch (hctx
->ccid3hctx_state
) {
322 case TFRC_SSTATE_NO_SENT
:
323 sk_reset_timer(sk
, &hctx
->ccid3hctx_no_feedback_timer
,
325 usecs_to_jiffies(TFRC_INITIAL_TIMEOUT
)));
326 hctx
->ccid3hctx_last_win_count
= 0;
327 hctx
->ccid3hctx_t_last_win_count
= now
;
329 /* Set t_0 for initial packet */
330 hctx
->ccid3hctx_t_nom
= now
;
332 hctx
->ccid3hctx_s
= skb
->len
;
335 * Use initial RTT sample when available: recommended by erratum
336 * to RFC 4342. This implements the initialisation procedure of
337 * draft rfc3448bis, section 4.2. Remember, X is scaled by 2^6.
339 if (dp
->dccps_syn_rtt
) {
340 ccid3_pr_debug("SYN RTT = %uus\n", dp
->dccps_syn_rtt
);
341 hctx
->ccid3hctx_rtt
= dp
->dccps_syn_rtt
;
342 hctx
->ccid3hctx_x
= rfc3390_initial_rate(sk
);
343 hctx
->ccid3hctx_t_ld
= now
;
345 /* Sender does not have RTT sample: X_pps = 1 pkt/sec */
346 hctx
->ccid3hctx_x
= hctx
->ccid3hctx_s
;
347 hctx
->ccid3hctx_x
<<= 6;
349 ccid3_update_send_interval(hctx
);
351 ccid3_hc_tx_set_state(sk
, TFRC_SSTATE_NO_FBACK
);
353 case TFRC_SSTATE_NO_FBACK
:
354 case TFRC_SSTATE_FBACK
:
355 delay
= ktime_us_delta(hctx
->ccid3hctx_t_nom
, now
);
356 ccid3_pr_debug("delay=%ld\n", (long)delay
);
358 * Scheduling of packet transmissions [RFC 3448, 4.6]
360 * if (t_now > t_nom - delta)
361 * // send the packet now
363 * // send the packet in (t_nom - t_now) milliseconds.
365 if (delay
- (s64
)hctx
->ccid3hctx_delta
>= 1000)
366 return (u32
)delay
/ 1000L;
368 ccid3_hc_tx_update_win_count(hctx
, now
);
370 case TFRC_SSTATE_TERM
:
371 DCCP_BUG("%s(%p) - Illegal state TERM", dccp_role(sk
), sk
);
375 /* prepare to send now (add options etc.) */
376 dp
->dccps_hc_tx_insert_options
= 1;
377 DCCP_SKB_CB(skb
)->dccpd_ccval
= hctx
->ccid3hctx_last_win_count
;
379 /* set the nominal send time for the next following packet */
380 hctx
->ccid3hctx_t_nom
= ktime_add_us(hctx
->ccid3hctx_t_nom
,
381 hctx
->ccid3hctx_t_ipi
);
385 static void ccid3_hc_tx_packet_sent(struct sock
*sk
, int more
,
388 struct ccid3_hc_tx_sock
*hctx
= ccid3_hc_tx_sk(sk
);
390 ccid3_hc_tx_update_s(hctx
, len
);
392 if (tfrc_tx_hist_add(&hctx
->ccid3hctx_hist
, dccp_sk(sk
)->dccps_gss
))
393 DCCP_CRIT("packet history - out of memory!");
396 static void ccid3_hc_tx_packet_recv(struct sock
*sk
, struct sk_buff
*skb
)
398 struct ccid3_hc_tx_sock
*hctx
= ccid3_hc_tx_sk(sk
);
399 struct ccid3_options_received
*opt_recv
;
404 /* we are only interested in ACKs */
405 if (!(DCCP_SKB_CB(skb
)->dccpd_type
== DCCP_PKT_ACK
||
406 DCCP_SKB_CB(skb
)->dccpd_type
== DCCP_PKT_DATAACK
))
409 opt_recv
= &hctx
->ccid3hctx_options_received
;
411 switch (hctx
->ccid3hctx_state
) {
412 case TFRC_SSTATE_NO_FBACK
:
413 case TFRC_SSTATE_FBACK
:
414 now
= ktime_get_real();
416 /* estimate RTT from history if ACK number is valid */
417 r_sample
= tfrc_tx_hist_rtt(hctx
->ccid3hctx_hist
,
418 DCCP_SKB_CB(skb
)->dccpd_ack_seq
, now
);
420 DCCP_WARN("%s(%p): %s with bogus ACK-%llu\n", dccp_role(sk
), sk
,
421 dccp_packet_name(DCCP_SKB_CB(skb
)->dccpd_type
),
422 (unsigned long long)DCCP_SKB_CB(skb
)->dccpd_ack_seq
);
426 /* Update receive rate in units of 64 * bytes/second */
427 hctx
->ccid3hctx_x_recv
= opt_recv
->ccid3or_receive_rate
;
428 hctx
->ccid3hctx_x_recv
<<= 6;
430 /* Update loss event rate */
431 pinv
= opt_recv
->ccid3or_loss_event_rate
;
432 if (pinv
== ~0U || pinv
== 0) /* see RFC 4342, 8.5 */
433 hctx
->ccid3hctx_p
= 0;
434 else /* can not exceed 100% */
435 hctx
->ccid3hctx_p
= 1000000 / pinv
;
437 * Validate new RTT sample and update moving average
439 r_sample
= dccp_sample_rtt(sk
, r_sample
);
440 hctx
->ccid3hctx_rtt
= tfrc_ewma(hctx
->ccid3hctx_rtt
, r_sample
, 9);
442 if (hctx
->ccid3hctx_state
== TFRC_SSTATE_NO_FBACK
) {
444 * Larger Initial Windows [RFC 4342, sec. 5]
446 hctx
->ccid3hctx_x
= rfc3390_initial_rate(sk
);
447 hctx
->ccid3hctx_t_ld
= now
;
449 ccid3_update_send_interval(hctx
);
451 ccid3_pr_debug("%s(%p), s=%u, MSS=%u, "
452 "R_sample=%uus, X=%u\n", dccp_role(sk
),
453 sk
, hctx
->ccid3hctx_s
,
454 dccp_sk(sk
)->dccps_mss_cache
, r_sample
,
455 (unsigned)(hctx
->ccid3hctx_x
>> 6));
457 ccid3_hc_tx_set_state(sk
, TFRC_SSTATE_FBACK
);
460 /* Update sending rate (step 4 of [RFC 3448, 4.3]) */
461 if (hctx
->ccid3hctx_p
> 0)
462 hctx
->ccid3hctx_x_calc
=
463 tfrc_calc_x(hctx
->ccid3hctx_s
,
466 ccid3_hc_tx_update_x(sk
, &now
);
468 ccid3_pr_debug("%s(%p), RTT=%uus (sample=%uus), s=%u, "
469 "p=%u, X_calc=%u, X_recv=%u, X=%u\n",
471 sk
, hctx
->ccid3hctx_rtt
, r_sample
,
472 hctx
->ccid3hctx_s
, hctx
->ccid3hctx_p
,
473 hctx
->ccid3hctx_x_calc
,
474 (unsigned)(hctx
->ccid3hctx_x_recv
>> 6),
475 (unsigned)(hctx
->ccid3hctx_x
>> 6));
478 /* unschedule no feedback timer */
479 sk_stop_timer(sk
, &hctx
->ccid3hctx_no_feedback_timer
);
482 * As we have calculated new ipi, delta, t_nom it is possible
483 * that we now can send a packet, so wake up dccp_wait_for_ccid
485 sk
->sk_write_space(sk
);
488 * Update timeout interval for the nofeedback timer.
489 * We use a configuration option to increase the lower bound.
490 * This can help avoid triggering the nofeedback timer too
491 * often ('spinning') on LANs with small RTTs.
493 hctx
->ccid3hctx_t_rto
= max_t(u32
, 4 * hctx
->ccid3hctx_rtt
,
494 CONFIG_IP_DCCP_CCID3_RTO
*
495 (USEC_PER_SEC
/1000));
497 * Schedule no feedback timer to expire in
498 * max(t_RTO, 2 * s/X) = max(t_RTO, 2 * t_ipi)
500 t_nfb
= max(hctx
->ccid3hctx_t_rto
, 2 * hctx
->ccid3hctx_t_ipi
);
502 ccid3_pr_debug("%s(%p), Scheduled no feedback timer to "
503 "expire in %lu jiffies (%luus)\n",
505 sk
, usecs_to_jiffies(t_nfb
), t_nfb
);
507 sk_reset_timer(sk
, &hctx
->ccid3hctx_no_feedback_timer
,
508 jiffies
+ usecs_to_jiffies(t_nfb
));
510 case TFRC_SSTATE_NO_SENT
: /* fall through */
511 case TFRC_SSTATE_TERM
: /* ignore feedback when closing */
516 static int ccid3_hc_tx_parse_options(struct sock
*sk
, unsigned char option
,
517 unsigned char len
, u16 idx
,
518 unsigned char *value
)
521 const struct dccp_sock
*dp
= dccp_sk(sk
);
522 struct ccid3_hc_tx_sock
*hctx
= ccid3_hc_tx_sk(sk
);
523 struct ccid3_options_received
*opt_recv
;
526 opt_recv
= &hctx
->ccid3hctx_options_received
;
528 if (opt_recv
->ccid3or_seqno
!= dp
->dccps_gsr
) {
529 opt_recv
->ccid3or_seqno
= dp
->dccps_gsr
;
530 opt_recv
->ccid3or_loss_event_rate
= ~0;
531 opt_recv
->ccid3or_loss_intervals_idx
= 0;
532 opt_recv
->ccid3or_loss_intervals_len
= 0;
533 opt_recv
->ccid3or_receive_rate
= 0;
537 case TFRC_OPT_LOSS_EVENT_RATE
:
538 if (unlikely(len
!= 4)) {
539 DCCP_WARN("%s(%p), invalid len %d "
540 "for TFRC_OPT_LOSS_EVENT_RATE\n",
541 dccp_role(sk
), sk
, len
);
544 opt_val
= get_unaligned((__be32
*)value
);
545 opt_recv
->ccid3or_loss_event_rate
= ntohl(opt_val
);
546 ccid3_pr_debug("%s(%p), LOSS_EVENT_RATE=%u\n",
548 opt_recv
->ccid3or_loss_event_rate
);
551 case TFRC_OPT_LOSS_INTERVALS
:
552 opt_recv
->ccid3or_loss_intervals_idx
= idx
;
553 opt_recv
->ccid3or_loss_intervals_len
= len
;
554 ccid3_pr_debug("%s(%p), LOSS_INTERVALS=(%u, %u)\n",
556 opt_recv
->ccid3or_loss_intervals_idx
,
557 opt_recv
->ccid3or_loss_intervals_len
);
559 case TFRC_OPT_RECEIVE_RATE
:
560 if (unlikely(len
!= 4)) {
561 DCCP_WARN("%s(%p), invalid len %d "
562 "for TFRC_OPT_RECEIVE_RATE\n",
563 dccp_role(sk
), sk
, len
);
566 opt_val
= get_unaligned((__be32
*)value
);
567 opt_recv
->ccid3or_receive_rate
= ntohl(opt_val
);
568 ccid3_pr_debug("%s(%p), RECEIVE_RATE=%u\n",
570 opt_recv
->ccid3or_receive_rate
);
578 static int ccid3_hc_tx_init(struct ccid
*ccid
, struct sock
*sk
)
580 struct ccid3_hc_tx_sock
*hctx
= ccid_priv(ccid
);
582 hctx
->ccid3hctx_state
= TFRC_SSTATE_NO_SENT
;
583 hctx
->ccid3hctx_hist
= NULL
;
584 setup_timer(&hctx
->ccid3hctx_no_feedback_timer
,
585 ccid3_hc_tx_no_feedback_timer
, (unsigned long)sk
);
590 static void ccid3_hc_tx_exit(struct sock
*sk
)
592 struct ccid3_hc_tx_sock
*hctx
= ccid3_hc_tx_sk(sk
);
594 ccid3_hc_tx_set_state(sk
, TFRC_SSTATE_TERM
);
595 sk_stop_timer(sk
, &hctx
->ccid3hctx_no_feedback_timer
);
597 tfrc_tx_hist_purge(&hctx
->ccid3hctx_hist
);
600 static void ccid3_hc_tx_get_info(struct sock
*sk
, struct tcp_info
*info
)
602 struct ccid3_hc_tx_sock
*hctx
;
604 /* Listen socks doesn't have a private CCID block */
605 if (sk
->sk_state
== DCCP_LISTEN
)
608 hctx
= ccid3_hc_tx_sk(sk
);
609 info
->tcpi_rto
= hctx
->ccid3hctx_t_rto
;
610 info
->tcpi_rtt
= hctx
->ccid3hctx_rtt
;
613 static int ccid3_hc_tx_getsockopt(struct sock
*sk
, const int optname
, int len
,
614 u32 __user
*optval
, int __user
*optlen
)
616 const struct ccid3_hc_tx_sock
*hctx
;
619 /* Listen socks doesn't have a private CCID block */
620 if (sk
->sk_state
== DCCP_LISTEN
)
623 hctx
= ccid3_hc_tx_sk(sk
);
625 case DCCP_SOCKOPT_CCID_TX_INFO
:
626 if (len
< sizeof(hctx
->ccid3hctx_tfrc
))
628 len
= sizeof(hctx
->ccid3hctx_tfrc
);
629 val
= &hctx
->ccid3hctx_tfrc
;
635 if (put_user(len
, optlen
) || copy_to_user(optval
, val
, len
))
642 * Receiver Half-Connection Routines
645 /* CCID3 feedback types */
646 enum ccid3_fback_type
{
647 CCID3_FBACK_NONE
= 0,
649 CCID3_FBACK_PERIODIC
,
650 CCID3_FBACK_PARAM_CHANGE
653 #ifdef CONFIG_IP_DCCP_CCID3_DEBUG
654 static const char *ccid3_rx_state_name(enum ccid3_hc_rx_states state
)
656 static char *ccid3_rx_state_names
[] = {
657 [TFRC_RSTATE_NO_DATA
] = "NO_DATA",
658 [TFRC_RSTATE_DATA
] = "DATA",
659 [TFRC_RSTATE_TERM
] = "TERM",
662 return ccid3_rx_state_names
[state
];
666 static void ccid3_hc_rx_set_state(struct sock
*sk
,
667 enum ccid3_hc_rx_states state
)
669 struct ccid3_hc_rx_sock
*hcrx
= ccid3_hc_rx_sk(sk
);
670 enum ccid3_hc_rx_states oldstate
= hcrx
->ccid3hcrx_state
;
672 ccid3_pr_debug("%s(%p) %-8.8s -> %s\n",
673 dccp_role(sk
), sk
, ccid3_rx_state_name(oldstate
),
674 ccid3_rx_state_name(state
));
675 WARN_ON(state
== oldstate
);
676 hcrx
->ccid3hcrx_state
= state
;
679 static void ccid3_hc_rx_send_feedback(struct sock
*sk
,
680 const struct sk_buff
*skb
,
681 enum ccid3_fback_type fbtype
)
683 struct ccid3_hc_rx_sock
*hcrx
= ccid3_hc_rx_sk(sk
);
684 struct dccp_sock
*dp
= dccp_sk(sk
);
688 if (unlikely(hcrx
->ccid3hcrx_state
== TFRC_RSTATE_TERM
))
691 now
= ktime_get_real();
694 case CCID3_FBACK_INITIAL
:
695 hcrx
->ccid3hcrx_x_recv
= 0;
696 hcrx
->ccid3hcrx_pinv
= ~0U; /* see RFC 4342, 8.5 */
698 case CCID3_FBACK_PARAM_CHANGE
:
700 * When parameters change (new loss or p > p_prev), we do not
701 * have a reliable estimate for R_m of [RFC 3448, 6.2] and so
702 * need to reuse the previous value of X_recv. However, when
703 * X_recv was 0 (due to early loss), this would kill X down to
704 * s/t_mbi (i.e. one packet in 64 seconds).
705 * To avoid such drastic reduction, we approximate X_recv as
706 * the number of bytes since last feedback.
707 * This is a safe fallback, since X is bounded above by X_calc.
709 if (hcrx
->ccid3hcrx_x_recv
> 0)
712 case CCID3_FBACK_PERIODIC
:
713 delta
= ktime_us_delta(now
, hcrx
->ccid3hcrx_tstamp_last_feedback
);
715 DCCP_BUG("delta (%ld) <= 0", (long)delta
);
717 hcrx
->ccid3hcrx_x_recv
=
718 scaled_div32(hcrx
->ccid3hcrx_bytes_recv
, delta
);
724 ccid3_pr_debug("Interval %ldusec, X_recv=%u, 1/p=%u\n", (long)delta
,
725 hcrx
->ccid3hcrx_x_recv
, hcrx
->ccid3hcrx_pinv
);
727 hcrx
->ccid3hcrx_tstamp_last_feedback
= now
;
728 hcrx
->ccid3hcrx_last_counter
= dccp_hdr(skb
)->dccph_ccval
;
729 hcrx
->ccid3hcrx_bytes_recv
= 0;
731 dp
->dccps_hc_rx_insert_options
= 1;
735 static int ccid3_hc_rx_insert_options(struct sock
*sk
, struct sk_buff
*skb
)
737 const struct ccid3_hc_rx_sock
*hcrx
;
740 if (!(sk
->sk_state
== DCCP_OPEN
|| sk
->sk_state
== DCCP_PARTOPEN
))
743 hcrx
= ccid3_hc_rx_sk(sk
);
745 if (dccp_packet_without_ack(skb
))
748 x_recv
= htonl(hcrx
->ccid3hcrx_x_recv
);
749 pinv
= htonl(hcrx
->ccid3hcrx_pinv
);
751 if (dccp_insert_option(sk
, skb
, TFRC_OPT_LOSS_EVENT_RATE
,
752 &pinv
, sizeof(pinv
)) ||
753 dccp_insert_option(sk
, skb
, TFRC_OPT_RECEIVE_RATE
,
754 &x_recv
, sizeof(x_recv
)))
760 static void ccid3_hc_rx_packet_recv(struct sock
*sk
, struct sk_buff
*skb
)
762 struct ccid3_hc_rx_sock
*hcrx
= ccid3_hc_rx_sk(sk
);
763 enum ccid3_fback_type do_feedback
= CCID3_FBACK_NONE
;
764 const u32 ndp
= dccp_sk(sk
)->dccps_options_received
.dccpor_ndp
;
765 const bool is_data_packet
= dccp_data_packet(skb
);
767 if (unlikely(hcrx
->ccid3hcrx_state
== TFRC_RSTATE_NO_DATA
)) {
768 if (is_data_packet
) {
769 const u32 payload
= skb
->len
- dccp_hdr(skb
)->dccph_doff
* 4;
770 do_feedback
= CCID3_FBACK_INITIAL
;
771 ccid3_hc_rx_set_state(sk
, TFRC_RSTATE_DATA
);
772 hcrx
->ccid3hcrx_s
= payload
;
774 * Not necessary to update ccid3hcrx_bytes_recv here,
775 * since X_recv = 0 for the first feedback packet (cf.
776 * RFC 3448, 6.3) -- gerrit
782 if (tfrc_rx_hist_duplicate(&hcrx
->ccid3hcrx_hist
, skb
))
783 return; /* done receiving */
785 if (is_data_packet
) {
786 const u32 payload
= skb
->len
- dccp_hdr(skb
)->dccph_doff
* 4;
788 * Update moving-average of s and the sum of received payload bytes
790 hcrx
->ccid3hcrx_s
= tfrc_ewma(hcrx
->ccid3hcrx_s
, payload
, 9);
791 hcrx
->ccid3hcrx_bytes_recv
+= payload
;
795 * Handle pending losses and otherwise check for new loss
797 if (tfrc_rx_hist_new_loss_indicated(&hcrx
->ccid3hcrx_hist
, skb
, ndp
))
801 * Handle data packets: RTT sampling and monitoring p
803 if (unlikely(!is_data_packet
))
806 if (list_empty(&hcrx
->ccid3hcrx_li_hist
)) { /* no loss so far: p = 0 */
807 const u32 sample
= tfrc_rx_hist_sample_rtt(&hcrx
->ccid3hcrx_hist
, skb
);
809 * Empty loss history: no loss so far, hence p stays 0.
810 * Sample RTT values, since an RTT estimate is required for the
811 * computation of p when the first loss occurs; RFC 3448, 6.3.1.
814 hcrx
->ccid3hcrx_rtt
= tfrc_ewma(hcrx
->ccid3hcrx_rtt
, sample
, 9);
818 * Check if the periodic once-per-RTT feedback is due; RFC 4342, 10.3
820 if (SUB16(dccp_hdr(skb
)->dccph_ccval
, hcrx
->ccid3hcrx_last_counter
) > 3)
821 do_feedback
= CCID3_FBACK_PERIODIC
;
824 tfrc_rx_hist_add_packet(&hcrx
->ccid3hcrx_hist
, skb
, ndp
);
827 ccid3_hc_rx_send_feedback(sk
, skb
, do_feedback
);
830 static int ccid3_hc_rx_init(struct ccid
*ccid
, struct sock
*sk
)
832 struct ccid3_hc_rx_sock
*hcrx
= ccid_priv(ccid
);
834 ccid3_pr_debug("entry\n");
836 hcrx
->ccid3hcrx_state
= TFRC_RSTATE_NO_DATA
;
837 INIT_LIST_HEAD(&hcrx
->ccid3hcrx_li_hist
);
838 return tfrc_rx_hist_alloc(&hcrx
->ccid3hcrx_hist
);
841 static void ccid3_hc_rx_exit(struct sock
*sk
)
843 struct ccid3_hc_rx_sock
*hcrx
= ccid3_hc_rx_sk(sk
);
845 ccid3_hc_rx_set_state(sk
, TFRC_RSTATE_TERM
);
847 /* Empty packet history */
848 tfrc_rx_hist_purge(&hcrx
->ccid3hcrx_hist
);
850 /* Empty loss interval history */
851 dccp_li_hist_purge(&hcrx
->ccid3hcrx_li_hist
);
854 static void ccid3_hc_rx_get_info(struct sock
*sk
, struct tcp_info
*info
)
856 const struct ccid3_hc_rx_sock
*hcrx
;
858 /* Listen socks doesn't have a private CCID block */
859 if (sk
->sk_state
== DCCP_LISTEN
)
862 hcrx
= ccid3_hc_rx_sk(sk
);
863 info
->tcpi_ca_state
= hcrx
->ccid3hcrx_state
;
864 info
->tcpi_options
|= TCPI_OPT_TIMESTAMPS
;
865 info
->tcpi_rcv_rtt
= hcrx
->ccid3hcrx_rtt
;
868 static int ccid3_hc_rx_getsockopt(struct sock
*sk
, const int optname
, int len
,
869 u32 __user
*optval
, int __user
*optlen
)
871 const struct ccid3_hc_rx_sock
*hcrx
;
874 /* Listen socks doesn't have a private CCID block */
875 if (sk
->sk_state
== DCCP_LISTEN
)
878 hcrx
= ccid3_hc_rx_sk(sk
);
880 case DCCP_SOCKOPT_CCID_RX_INFO
:
881 if (len
< sizeof(hcrx
->ccid3hcrx_tfrc
))
883 len
= sizeof(hcrx
->ccid3hcrx_tfrc
);
884 val
= &hcrx
->ccid3hcrx_tfrc
;
890 if (put_user(len
, optlen
) || copy_to_user(optval
, val
, len
))
896 static struct ccid_operations ccid3
= {
897 .ccid_id
= DCCPC_CCID3
,
898 .ccid_name
= "ccid3",
899 .ccid_owner
= THIS_MODULE
,
900 .ccid_hc_tx_obj_size
= sizeof(struct ccid3_hc_tx_sock
),
901 .ccid_hc_tx_init
= ccid3_hc_tx_init
,
902 .ccid_hc_tx_exit
= ccid3_hc_tx_exit
,
903 .ccid_hc_tx_send_packet
= ccid3_hc_tx_send_packet
,
904 .ccid_hc_tx_packet_sent
= ccid3_hc_tx_packet_sent
,
905 .ccid_hc_tx_packet_recv
= ccid3_hc_tx_packet_recv
,
906 .ccid_hc_tx_parse_options
= ccid3_hc_tx_parse_options
,
907 .ccid_hc_rx_obj_size
= sizeof(struct ccid3_hc_rx_sock
),
908 .ccid_hc_rx_init
= ccid3_hc_rx_init
,
909 .ccid_hc_rx_exit
= ccid3_hc_rx_exit
,
910 .ccid_hc_rx_insert_options
= ccid3_hc_rx_insert_options
,
911 .ccid_hc_rx_packet_recv
= ccid3_hc_rx_packet_recv
,
912 .ccid_hc_rx_get_info
= ccid3_hc_rx_get_info
,
913 .ccid_hc_tx_get_info
= ccid3_hc_tx_get_info
,
914 .ccid_hc_rx_getsockopt
= ccid3_hc_rx_getsockopt
,
915 .ccid_hc_tx_getsockopt
= ccid3_hc_tx_getsockopt
,
918 #ifdef CONFIG_IP_DCCP_CCID3_DEBUG
919 module_param(ccid3_debug
, bool, 0444);
920 MODULE_PARM_DESC(ccid3_debug
, "Enable debug messages");
923 static __init
int ccid3_module_init(void)
925 return ccid_register(&ccid3
);
927 module_init(ccid3_module_init
);
929 static __exit
void ccid3_module_exit(void)
931 ccid_unregister(&ccid3
);
933 module_exit(ccid3_module_exit
);
935 MODULE_AUTHOR("Ian McDonald <ian.mcdonald@jandi.co.nz>, "
936 "Arnaldo Carvalho de Melo <acme@ghostprotocols.net>");
937 MODULE_DESCRIPTION("DCCP TFRC CCID3 CCID");
938 MODULE_LICENSE("GPL");
939 MODULE_ALIAS("net-dccp-ccid-3");