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));
118 * ccid3_hc_tx_update_x - Update allowed sending rate X
119 * @stamp: most recent time if available - can be left NULL.
120 * This function tracks draft rfc3448bis, check there for latest details.
122 * Note: X and X_recv are both stored in units of 64 * bytes/second, to support
123 * fine-grained resolution of sending rates. This requires scaling by 2^6
124 * throughout the code. Only X_calc is unscaled (in bytes/second).
127 static void ccid3_hc_tx_update_x(struct sock
*sk
, ktime_t
*stamp
)
130 struct ccid3_hc_tx_sock
*hctx
= ccid3_hc_tx_sk(sk
);
131 __u64 min_rate
= 2 * hctx
->ccid3hctx_x_recv
;
132 const __u64 old_x
= hctx
->ccid3hctx_x
;
133 ktime_t now
= stamp
? *stamp
: ktime_get_real();
136 * Handle IDLE periods: do not reduce below RFC3390 initial sending rate
137 * when idling [RFC 4342, 5.1]. See also draft-ietf-dccp-rfc3448bis.
138 * For consistency with X and X_recv, min_rate is also scaled by 2^6.
140 if (unlikely(hctx
->ccid3hctx_idle
)) {
141 min_rate
= rfc3390_initial_rate(sk
);
142 min_rate
= max(min_rate
, 2 * hctx
->ccid3hctx_x_recv
);
145 if (hctx
->ccid3hctx_p
> 0) {
147 hctx
->ccid3hctx_x
= min(((__u64
)hctx
->ccid3hctx_x_calc
) << 6,
149 hctx
->ccid3hctx_x
= max(hctx
->ccid3hctx_x
,
150 (((__u64
)hctx
->ccid3hctx_s
) << 6) /
153 } else if (ktime_us_delta(now
, hctx
->ccid3hctx_t_ld
)
154 - (s64
)hctx
->ccid3hctx_rtt
>= 0) {
157 max(min(2 * hctx
->ccid3hctx_x
, min_rate
),
158 scaled_div(((__u64
)hctx
->ccid3hctx_s
) << 6,
159 hctx
->ccid3hctx_rtt
));
160 hctx
->ccid3hctx_t_ld
= now
;
163 if (hctx
->ccid3hctx_x
!= old_x
) {
164 ccid3_pr_debug("X_prev=%u, X_now=%u, X_calc=%u, "
165 "X_recv=%u\n", (unsigned)(old_x
>> 6),
166 (unsigned)(hctx
->ccid3hctx_x
>> 6),
167 hctx
->ccid3hctx_x_calc
,
168 (unsigned)(hctx
->ccid3hctx_x_recv
>> 6));
170 ccid3_update_send_interval(hctx
);
175 * Track the mean packet size `s' (cf. RFC 4342, 5.3 and RFC 3448, 4.1)
176 * @len: DCCP packet payload size in bytes
178 static inline void ccid3_hc_tx_update_s(struct ccid3_hc_tx_sock
*hctx
, int len
)
180 const u16 old_s
= hctx
->ccid3hctx_s
;
182 hctx
->ccid3hctx_s
= old_s
== 0 ? len
: (9 * old_s
+ len
) / 10;
184 if (hctx
->ccid3hctx_s
!= old_s
)
185 ccid3_update_send_interval(hctx
);
189 * Update Window Counter using the algorithm from [RFC 4342, 8.1].
190 * The algorithm is not applicable if RTT < 4 microseconds.
192 static inline void ccid3_hc_tx_update_win_count(struct ccid3_hc_tx_sock
*hctx
,
197 if (unlikely(hctx
->ccid3hctx_rtt
< 4)) /* avoid divide-by-zero */
200 quarter_rtts
= ktime_us_delta(now
, hctx
->ccid3hctx_t_last_win_count
);
201 quarter_rtts
/= hctx
->ccid3hctx_rtt
/ 4;
203 if (quarter_rtts
> 0) {
204 hctx
->ccid3hctx_t_last_win_count
= now
;
205 hctx
->ccid3hctx_last_win_count
+= min_t(u32
, quarter_rtts
, 5);
206 hctx
->ccid3hctx_last_win_count
&= 0xF; /* mod 16 */
210 static void ccid3_hc_tx_no_feedback_timer(unsigned long data
)
212 struct sock
*sk
= (struct sock
*)data
;
213 struct ccid3_hc_tx_sock
*hctx
= ccid3_hc_tx_sk(sk
);
214 unsigned long t_nfb
= USEC_PER_SEC
/ 5;
217 if (sock_owned_by_user(sk
)) {
218 /* Try again later. */
219 /* XXX: set some sensible MIB */
223 ccid3_pr_debug("%s(%p, state=%s) - entry \n", dccp_role(sk
), sk
,
224 ccid3_tx_state_name(hctx
->ccid3hctx_state
));
226 hctx
->ccid3hctx_idle
= 1;
228 switch (hctx
->ccid3hctx_state
) {
229 case TFRC_SSTATE_NO_FBACK
:
230 /* RFC 3448, 4.4: Halve send rate directly */
231 hctx
->ccid3hctx_x
= max(hctx
->ccid3hctx_x
/ 2,
232 (((__u64
)hctx
->ccid3hctx_s
) << 6) /
235 ccid3_pr_debug("%s(%p, state=%s), updated tx rate to %u "
236 "bytes/s\n", dccp_role(sk
), sk
,
237 ccid3_tx_state_name(hctx
->ccid3hctx_state
),
238 (unsigned)(hctx
->ccid3hctx_x
>> 6));
239 /* The value of R is still undefined and so we can not recompute
240 * the timout value. Keep initial value as per [RFC 4342, 5]. */
241 t_nfb
= TFRC_INITIAL_TIMEOUT
;
242 ccid3_update_send_interval(hctx
);
244 case TFRC_SSTATE_FBACK
:
246 * Modify the cached value of X_recv [RFC 3448, 4.4]
248 * If (p == 0 || X_calc > 2 * X_recv)
249 * X_recv = max(X_recv / 2, s / (2 * t_mbi));
251 * X_recv = X_calc / 4;
253 * Note that X_recv is scaled by 2^6 while X_calc is not
255 BUG_ON(hctx
->ccid3hctx_p
&& !hctx
->ccid3hctx_x_calc
);
257 if (hctx
->ccid3hctx_p
== 0 ||
258 (hctx
->ccid3hctx_x_calc
> (hctx
->ccid3hctx_x_recv
>> 5))) {
260 hctx
->ccid3hctx_x_recv
=
261 max(hctx
->ccid3hctx_x_recv
/ 2,
262 (((__u64
)hctx
->ccid3hctx_s
) << 6) /
265 hctx
->ccid3hctx_x_recv
= hctx
->ccid3hctx_x_calc
;
266 hctx
->ccid3hctx_x_recv
<<= 4;
268 /* Now recalculate X [RFC 3448, 4.3, step (4)] */
269 ccid3_hc_tx_update_x(sk
, NULL
);
271 * Schedule no feedback timer to expire in
272 * max(t_RTO, 2 * s/X) = max(t_RTO, 2 * t_ipi)
273 * See comments in packet_recv() regarding the value of t_RTO.
275 t_nfb
= max(hctx
->ccid3hctx_t_rto
, 2 * hctx
->ccid3hctx_t_ipi
);
277 case TFRC_SSTATE_NO_SENT
:
278 DCCP_BUG("%s(%p) - Illegal state NO_SENT", dccp_role(sk
), sk
);
280 case TFRC_SSTATE_TERM
:
285 sk_reset_timer(sk
, &hctx
->ccid3hctx_no_feedback_timer
,
286 jiffies
+ usecs_to_jiffies(t_nfb
));
294 * > 0: delay (in msecs) that should pass before actually sending
295 * = 0: can send immediately
296 * < 0: error condition; do not send packet
298 static int ccid3_hc_tx_send_packet(struct sock
*sk
, struct sk_buff
*skb
)
300 struct dccp_sock
*dp
= dccp_sk(sk
);
301 struct ccid3_hc_tx_sock
*hctx
= ccid3_hc_tx_sk(sk
);
302 ktime_t now
= ktime_get_real();
306 * This function is called only for Data and DataAck packets. Sending
307 * zero-sized Data(Ack)s is theoretically possible, but for congestion
308 * control this case is pathological - ignore it.
310 if (unlikely(skb
->len
== 0))
313 switch (hctx
->ccid3hctx_state
) {
314 case TFRC_SSTATE_NO_SENT
:
315 sk_reset_timer(sk
, &hctx
->ccid3hctx_no_feedback_timer
,
317 usecs_to_jiffies(TFRC_INITIAL_TIMEOUT
)));
318 hctx
->ccid3hctx_last_win_count
= 0;
319 hctx
->ccid3hctx_t_last_win_count
= now
;
321 /* Set t_0 for initial packet */
322 hctx
->ccid3hctx_t_nom
= now
;
324 hctx
->ccid3hctx_s
= skb
->len
;
327 * Use initial RTT sample when available: recommended by erratum
328 * to RFC 4342. This implements the initialisation procedure of
329 * draft rfc3448bis, section 4.2. Remember, X is scaled by 2^6.
331 if (dp
->dccps_syn_rtt
) {
332 ccid3_pr_debug("SYN RTT = %uus\n", dp
->dccps_syn_rtt
);
333 hctx
->ccid3hctx_rtt
= dp
->dccps_syn_rtt
;
334 hctx
->ccid3hctx_x
= rfc3390_initial_rate(sk
);
335 hctx
->ccid3hctx_t_ld
= now
;
337 /* Sender does not have RTT sample: X = MSS/second */
338 hctx
->ccid3hctx_x
= dp
->dccps_mss_cache
;
339 hctx
->ccid3hctx_x
<<= 6;
341 ccid3_update_send_interval(hctx
);
343 ccid3_hc_tx_set_state(sk
, TFRC_SSTATE_NO_FBACK
);
345 case TFRC_SSTATE_NO_FBACK
:
346 case TFRC_SSTATE_FBACK
:
347 delay
= ktime_us_delta(hctx
->ccid3hctx_t_nom
, now
);
348 ccid3_pr_debug("delay=%ld\n", (long)delay
);
350 * Scheduling of packet transmissions [RFC 3448, 4.6]
352 * if (t_now > t_nom - delta)
353 * // send the packet now
355 * // send the packet in (t_nom - t_now) milliseconds.
357 if (delay
- (s64
)hctx
->ccid3hctx_delta
>= 1000)
358 return (u32
)delay
/ 1000L;
360 ccid3_hc_tx_update_win_count(hctx
, now
);
362 case TFRC_SSTATE_TERM
:
363 DCCP_BUG("%s(%p) - Illegal state TERM", dccp_role(sk
), sk
);
367 /* prepare to send now (add options etc.) */
368 dp
->dccps_hc_tx_insert_options
= 1;
369 DCCP_SKB_CB(skb
)->dccpd_ccval
= hctx
->ccid3hctx_last_win_count
;
370 hctx
->ccid3hctx_idle
= 0;
372 /* set the nominal send time for the next following packet */
373 hctx
->ccid3hctx_t_nom
= ktime_add_us(hctx
->ccid3hctx_t_nom
,
374 hctx
->ccid3hctx_t_ipi
);
378 static void ccid3_hc_tx_packet_sent(struct sock
*sk
, int more
,
381 struct ccid3_hc_tx_sock
*hctx
= ccid3_hc_tx_sk(sk
);
382 struct dccp_tx_hist_entry
*packet
;
384 ccid3_hc_tx_update_s(hctx
, len
);
386 packet
= dccp_tx_hist_entry_new(ccid3_tx_hist
, GFP_ATOMIC
);
387 if (unlikely(packet
== NULL
)) {
388 DCCP_CRIT("packet history - out of memory!");
391 dccp_tx_hist_add_entry(&hctx
->ccid3hctx_hist
, packet
);
393 packet
->dccphtx_tstamp
= ktime_get_real();
394 packet
->dccphtx_seqno
= dccp_sk(sk
)->dccps_gss
;
395 packet
->dccphtx_rtt
= hctx
->ccid3hctx_rtt
;
396 packet
->dccphtx_sent
= 1;
399 static void ccid3_hc_tx_packet_recv(struct sock
*sk
, struct sk_buff
*skb
)
401 struct ccid3_hc_tx_sock
*hctx
= ccid3_hc_tx_sk(sk
);
402 struct ccid3_options_received
*opt_recv
;
403 struct dccp_tx_hist_entry
*packet
;
408 /* we are only interested in ACKs */
409 if (!(DCCP_SKB_CB(skb
)->dccpd_type
== DCCP_PKT_ACK
||
410 DCCP_SKB_CB(skb
)->dccpd_type
== DCCP_PKT_DATAACK
))
413 opt_recv
= &hctx
->ccid3hctx_options_received
;
415 switch (hctx
->ccid3hctx_state
) {
416 case TFRC_SSTATE_NO_FBACK
:
417 case TFRC_SSTATE_FBACK
:
418 /* get packet from history to look up t_recvdata */
419 packet
= dccp_tx_hist_find_entry(&hctx
->ccid3hctx_hist
,
420 DCCP_SKB_CB(skb
)->dccpd_ack_seq
);
421 if (unlikely(packet
== NULL
)) {
422 DCCP_WARN("%s(%p), seqno %llu(%s) doesn't exist "
423 "in history!\n", dccp_role(sk
), sk
,
424 (unsigned long long)DCCP_SKB_CB(skb
)->dccpd_ack_seq
,
425 dccp_packet_name(DCCP_SKB_CB(skb
)->dccpd_type
));
429 /* Update receive rate in units of 64 * bytes/second */
430 hctx
->ccid3hctx_x_recv
= opt_recv
->ccid3or_receive_rate
;
431 hctx
->ccid3hctx_x_recv
<<= 6;
433 /* Update loss event rate */
434 pinv
= opt_recv
->ccid3or_loss_event_rate
;
435 if (pinv
== ~0U || pinv
== 0) /* see RFC 4342, 8.5 */
436 hctx
->ccid3hctx_p
= 0;
437 else /* can not exceed 100% */
438 hctx
->ccid3hctx_p
= 1000000 / pinv
;
440 now
= ktime_get_real();
442 * Calculate new round trip sample as per [RFC 3448, 4.3] by
443 * R_sample = (now - t_recvdata) - t_elapsed
445 r_sample
= dccp_sample_rtt(sk
, ktime_us_delta(now
, packet
->dccphtx_tstamp
));
448 * Update RTT estimate by
449 * If (No feedback recv)
452 * R = q * R + (1 - q) * R_sample;
454 * q is a constant, RFC 3448 recomments 0.9
456 if (hctx
->ccid3hctx_state
== TFRC_SSTATE_NO_FBACK
) {
458 * Larger Initial Windows [RFC 4342, sec. 5]
460 hctx
->ccid3hctx_rtt
= r_sample
;
461 hctx
->ccid3hctx_x
= rfc3390_initial_rate(sk
);
462 hctx
->ccid3hctx_t_ld
= now
;
464 ccid3_update_send_interval(hctx
);
466 ccid3_pr_debug("%s(%p), s=%u, MSS=%u, "
467 "R_sample=%uus, X=%u\n", dccp_role(sk
),
468 sk
, hctx
->ccid3hctx_s
,
469 dccp_sk(sk
)->dccps_mss_cache
, r_sample
,
470 (unsigned)(hctx
->ccid3hctx_x
>> 6));
472 ccid3_hc_tx_set_state(sk
, TFRC_SSTATE_FBACK
);
474 hctx
->ccid3hctx_rtt
= (9 * hctx
->ccid3hctx_rtt
+
477 /* Update sending rate (step 4 of [RFC 3448, 4.3]) */
478 if (hctx
->ccid3hctx_p
> 0)
479 hctx
->ccid3hctx_x_calc
=
480 tfrc_calc_x(hctx
->ccid3hctx_s
,
483 ccid3_hc_tx_update_x(sk
, &now
);
485 ccid3_pr_debug("%s(%p), RTT=%uus (sample=%uus), s=%u, "
486 "p=%u, X_calc=%u, X_recv=%u, X=%u\n",
488 sk
, hctx
->ccid3hctx_rtt
, r_sample
,
489 hctx
->ccid3hctx_s
, hctx
->ccid3hctx_p
,
490 hctx
->ccid3hctx_x_calc
,
491 (unsigned)(hctx
->ccid3hctx_x_recv
>> 6),
492 (unsigned)(hctx
->ccid3hctx_x
>> 6));
495 /* unschedule no feedback timer */
496 sk_stop_timer(sk
, &hctx
->ccid3hctx_no_feedback_timer
);
498 /* remove all packets older than the one acked from history */
499 dccp_tx_hist_purge_older(ccid3_tx_hist
,
500 &hctx
->ccid3hctx_hist
, packet
);
502 * As we have calculated new ipi, delta, t_nom it is possible
503 * that we now can send a packet, so wake up dccp_wait_for_ccid
505 sk
->sk_write_space(sk
);
508 * Update timeout interval for the nofeedback timer.
509 * We use a configuration option to increase the lower bound.
510 * This can help avoid triggering the nofeedback timer too
511 * often ('spinning') on LANs with small RTTs.
513 hctx
->ccid3hctx_t_rto
= max_t(u32
, 4 * hctx
->ccid3hctx_rtt
,
514 CONFIG_IP_DCCP_CCID3_RTO
*
515 (USEC_PER_SEC
/1000));
517 * Schedule no feedback timer to expire in
518 * max(t_RTO, 2 * s/X) = max(t_RTO, 2 * t_ipi)
520 t_nfb
= max(hctx
->ccid3hctx_t_rto
, 2 * hctx
->ccid3hctx_t_ipi
);
522 ccid3_pr_debug("%s(%p), Scheduled no feedback timer to "
523 "expire in %lu jiffies (%luus)\n",
525 sk
, usecs_to_jiffies(t_nfb
), t_nfb
);
527 sk_reset_timer(sk
, &hctx
->ccid3hctx_no_feedback_timer
,
528 jiffies
+ usecs_to_jiffies(t_nfb
));
531 hctx
->ccid3hctx_idle
= 1;
533 case TFRC_SSTATE_NO_SENT
: /* fall through */
534 case TFRC_SSTATE_TERM
: /* ignore feedback when closing */
539 static int ccid3_hc_tx_parse_options(struct sock
*sk
, unsigned char option
,
540 unsigned char len
, u16 idx
,
541 unsigned char *value
)
544 const struct dccp_sock
*dp
= dccp_sk(sk
);
545 struct ccid3_hc_tx_sock
*hctx
= ccid3_hc_tx_sk(sk
);
546 struct ccid3_options_received
*opt_recv
;
548 opt_recv
= &hctx
->ccid3hctx_options_received
;
550 if (opt_recv
->ccid3or_seqno
!= dp
->dccps_gsr
) {
551 opt_recv
->ccid3or_seqno
= dp
->dccps_gsr
;
552 opt_recv
->ccid3or_loss_event_rate
= ~0;
553 opt_recv
->ccid3or_loss_intervals_idx
= 0;
554 opt_recv
->ccid3or_loss_intervals_len
= 0;
555 opt_recv
->ccid3or_receive_rate
= 0;
559 case TFRC_OPT_LOSS_EVENT_RATE
:
560 if (unlikely(len
!= 4)) {
561 DCCP_WARN("%s(%p), invalid len %d "
562 "for TFRC_OPT_LOSS_EVENT_RATE\n",
563 dccp_role(sk
), sk
, len
);
566 opt_recv
->ccid3or_loss_event_rate
=
567 ntohl(*(__be32
*)value
);
568 ccid3_pr_debug("%s(%p), LOSS_EVENT_RATE=%u\n",
570 opt_recv
->ccid3or_loss_event_rate
);
573 case TFRC_OPT_LOSS_INTERVALS
:
574 opt_recv
->ccid3or_loss_intervals_idx
= idx
;
575 opt_recv
->ccid3or_loss_intervals_len
= len
;
576 ccid3_pr_debug("%s(%p), LOSS_INTERVALS=(%u, %u)\n",
578 opt_recv
->ccid3or_loss_intervals_idx
,
579 opt_recv
->ccid3or_loss_intervals_len
);
581 case TFRC_OPT_RECEIVE_RATE
:
582 if (unlikely(len
!= 4)) {
583 DCCP_WARN("%s(%p), invalid len %d "
584 "for TFRC_OPT_RECEIVE_RATE\n",
585 dccp_role(sk
), sk
, len
);
588 opt_recv
->ccid3or_receive_rate
=
589 ntohl(*(__be32
*)value
);
590 ccid3_pr_debug("%s(%p), RECEIVE_RATE=%u\n",
592 opt_recv
->ccid3or_receive_rate
);
600 static int ccid3_hc_tx_init(struct ccid
*ccid
, struct sock
*sk
)
602 struct ccid3_hc_tx_sock
*hctx
= ccid_priv(ccid
);
604 hctx
->ccid3hctx_s
= 0;
605 hctx
->ccid3hctx_rtt
= 0;
606 hctx
->ccid3hctx_state
= TFRC_SSTATE_NO_SENT
;
607 INIT_LIST_HEAD(&hctx
->ccid3hctx_hist
);
609 hctx
->ccid3hctx_no_feedback_timer
.function
=
610 ccid3_hc_tx_no_feedback_timer
;
611 hctx
->ccid3hctx_no_feedback_timer
.data
= (unsigned long)sk
;
612 init_timer(&hctx
->ccid3hctx_no_feedback_timer
);
617 static void ccid3_hc_tx_exit(struct sock
*sk
)
619 struct ccid3_hc_tx_sock
*hctx
= ccid3_hc_tx_sk(sk
);
621 ccid3_hc_tx_set_state(sk
, TFRC_SSTATE_TERM
);
622 sk_stop_timer(sk
, &hctx
->ccid3hctx_no_feedback_timer
);
624 /* Empty packet history */
625 dccp_tx_hist_purge(ccid3_tx_hist
, &hctx
->ccid3hctx_hist
);
628 static void ccid3_hc_tx_get_info(struct sock
*sk
, struct tcp_info
*info
)
630 struct ccid3_hc_tx_sock
*hctx
;
632 /* Listen socks doesn't have a private CCID block */
633 if (sk
->sk_state
== DCCP_LISTEN
)
636 hctx
= ccid3_hc_tx_sk(sk
);
637 info
->tcpi_rto
= hctx
->ccid3hctx_t_rto
;
638 info
->tcpi_rtt
= hctx
->ccid3hctx_rtt
;
641 static int ccid3_hc_tx_getsockopt(struct sock
*sk
, const int optname
, int len
,
642 u32 __user
*optval
, int __user
*optlen
)
644 const struct ccid3_hc_tx_sock
*hctx
;
647 /* Listen socks doesn't have a private CCID block */
648 if (sk
->sk_state
== DCCP_LISTEN
)
651 hctx
= ccid3_hc_tx_sk(sk
);
653 case DCCP_SOCKOPT_CCID_TX_INFO
:
654 if (len
< sizeof(hctx
->ccid3hctx_tfrc
))
656 len
= sizeof(hctx
->ccid3hctx_tfrc
);
657 val
= &hctx
->ccid3hctx_tfrc
;
663 if (put_user(len
, optlen
) || copy_to_user(optval
, val
, len
))
670 * Receiver Half-Connection Routines
672 #ifdef CONFIG_IP_DCCP_CCID3_DEBUG
673 static const char *ccid3_rx_state_name(enum ccid3_hc_rx_states state
)
675 static char *ccid3_rx_state_names
[] = {
676 [TFRC_RSTATE_NO_DATA
] = "NO_DATA",
677 [TFRC_RSTATE_DATA
] = "DATA",
678 [TFRC_RSTATE_TERM
] = "TERM",
681 return ccid3_rx_state_names
[state
];
685 static void ccid3_hc_rx_set_state(struct sock
*sk
,
686 enum ccid3_hc_rx_states state
)
688 struct ccid3_hc_rx_sock
*hcrx
= ccid3_hc_rx_sk(sk
);
689 enum ccid3_hc_rx_states oldstate
= hcrx
->ccid3hcrx_state
;
691 ccid3_pr_debug("%s(%p) %-8.8s -> %s\n",
692 dccp_role(sk
), sk
, ccid3_rx_state_name(oldstate
),
693 ccid3_rx_state_name(state
));
694 WARN_ON(state
== oldstate
);
695 hcrx
->ccid3hcrx_state
= state
;
698 static inline void ccid3_hc_rx_update_s(struct ccid3_hc_rx_sock
*hcrx
, int len
)
700 if (unlikely(len
== 0)) /* don't update on empty packets (e.g. ACKs) */
701 ccid3_pr_debug("Packet payload length is 0 - not updating\n");
703 hcrx
->ccid3hcrx_s
= hcrx
->ccid3hcrx_s
== 0 ? len
:
704 (9 * hcrx
->ccid3hcrx_s
+ len
) / 10;
707 static void ccid3_hc_rx_send_feedback(struct sock
*sk
)
709 struct ccid3_hc_rx_sock
*hcrx
= ccid3_hc_rx_sk(sk
);
710 struct dccp_sock
*dp
= dccp_sk(sk
);
711 struct dccp_rx_hist_entry
*packet
;
715 ccid3_pr_debug("%s(%p) - entry \n", dccp_role(sk
), sk
);
717 now
= ktime_get_real();
719 switch (hcrx
->ccid3hcrx_state
) {
720 case TFRC_RSTATE_NO_DATA
:
721 hcrx
->ccid3hcrx_x_recv
= 0;
723 case TFRC_RSTATE_DATA
:
724 delta
= ktime_us_delta(now
,
725 hcrx
->ccid3hcrx_tstamp_last_feedback
);
726 DCCP_BUG_ON(delta
< 0);
727 hcrx
->ccid3hcrx_x_recv
=
728 scaled_div32(hcrx
->ccid3hcrx_bytes_recv
, delta
);
730 case TFRC_RSTATE_TERM
:
731 DCCP_BUG("%s(%p) - Illegal state TERM", dccp_role(sk
), sk
);
735 packet
= dccp_rx_hist_find_data_packet(&hcrx
->ccid3hcrx_hist
);
736 if (unlikely(packet
== NULL
)) {
737 DCCP_WARN("%s(%p), no data packet in history!\n",
742 hcrx
->ccid3hcrx_tstamp_last_feedback
= now
;
743 hcrx
->ccid3hcrx_ccval_last_counter
= packet
->dccphrx_ccval
;
744 hcrx
->ccid3hcrx_bytes_recv
= 0;
746 /* Elapsed time information [RFC 4340, 13.2] in units of 10 * usecs */
747 delta
= ktime_us_delta(now
, packet
->dccphrx_tstamp
);
748 DCCP_BUG_ON(delta
< 0);
749 hcrx
->ccid3hcrx_elapsed_time
= delta
/ 10;
751 if (hcrx
->ccid3hcrx_p
== 0)
752 hcrx
->ccid3hcrx_pinv
= ~0U; /* see RFC 4342, 8.5 */
753 else if (hcrx
->ccid3hcrx_p
> 1000000) {
754 DCCP_WARN("p (%u) > 100%%\n", hcrx
->ccid3hcrx_p
);
755 hcrx
->ccid3hcrx_pinv
= 1; /* use 100% in this case */
757 hcrx
->ccid3hcrx_pinv
= 1000000 / hcrx
->ccid3hcrx_p
;
759 dp
->dccps_hc_rx_insert_options
= 1;
763 static int ccid3_hc_rx_insert_options(struct sock
*sk
, struct sk_buff
*skb
)
765 const struct ccid3_hc_rx_sock
*hcrx
;
768 if (!(sk
->sk_state
== DCCP_OPEN
|| sk
->sk_state
== DCCP_PARTOPEN
))
771 hcrx
= ccid3_hc_rx_sk(sk
);
772 DCCP_SKB_CB(skb
)->dccpd_ccval
= hcrx
->ccid3hcrx_ccval_last_counter
;
774 if (dccp_packet_without_ack(skb
))
777 x_recv
= htonl(hcrx
->ccid3hcrx_x_recv
);
778 pinv
= htonl(hcrx
->ccid3hcrx_pinv
);
780 if ((hcrx
->ccid3hcrx_elapsed_time
!= 0 &&
781 dccp_insert_option_elapsed_time(sk
, skb
,
782 hcrx
->ccid3hcrx_elapsed_time
)) ||
783 dccp_insert_option_timestamp(sk
, skb
) ||
784 dccp_insert_option(sk
, skb
, TFRC_OPT_LOSS_EVENT_RATE
,
785 &pinv
, sizeof(pinv
)) ||
786 dccp_insert_option(sk
, skb
, TFRC_OPT_RECEIVE_RATE
,
787 &x_recv
, sizeof(x_recv
)))
793 static int ccid3_hc_rx_detect_loss(struct sock
*sk
,
794 struct dccp_rx_hist_entry
*packet
)
796 struct ccid3_hc_rx_sock
*hcrx
= ccid3_hc_rx_sk(sk
);
797 struct dccp_rx_hist_entry
*rx_hist
=
798 dccp_rx_hist_head(&hcrx
->ccid3hcrx_hist
);
799 u64 seqno
= packet
->dccphrx_seqno
;
805 tmp_seqno
= hcrx
->ccid3hcrx_seqno_nonloss
;
808 follows48(packet
->dccphrx_seqno
, hcrx
->ccid3hcrx_seqno_nonloss
)) {
809 hcrx
->ccid3hcrx_seqno_nonloss
= seqno
;
810 hcrx
->ccid3hcrx_ccval_nonloss
= packet
->dccphrx_ccval
;
815 while (dccp_delta_seqno(hcrx
->ccid3hcrx_seqno_nonloss
, seqno
)
816 > TFRC_RECV_NUM_LATE_LOSS
) {
818 dccp_li_update_li(sk
,
819 &hcrx
->ccid3hcrx_li_hist
,
820 &hcrx
->ccid3hcrx_hist
,
821 hcrx
->ccid3hcrx_tstamp_last_feedback
,
823 hcrx
->ccid3hcrx_bytes_recv
,
824 hcrx
->ccid3hcrx_x_recv
,
825 hcrx
->ccid3hcrx_seqno_nonloss
,
826 hcrx
->ccid3hcrx_ccval_nonloss
);
827 tmp_seqno
= hcrx
->ccid3hcrx_seqno_nonloss
;
828 dccp_inc_seqno(&tmp_seqno
);
829 hcrx
->ccid3hcrx_seqno_nonloss
= tmp_seqno
;
830 dccp_inc_seqno(&tmp_seqno
);
831 while (dccp_rx_hist_find_entry(&hcrx
->ccid3hcrx_hist
,
832 tmp_seqno
, &ccval
)) {
833 hcrx
->ccid3hcrx_seqno_nonloss
= tmp_seqno
;
834 hcrx
->ccid3hcrx_ccval_nonloss
= ccval
;
835 dccp_inc_seqno(&tmp_seqno
);
839 /* FIXME - this code could be simplified with above while */
840 /* but works at moment */
841 if (follows48(packet
->dccphrx_seqno
, hcrx
->ccid3hcrx_seqno_nonloss
)) {
842 hcrx
->ccid3hcrx_seqno_nonloss
= seqno
;
843 hcrx
->ccid3hcrx_ccval_nonloss
= packet
->dccphrx_ccval
;
847 dccp_rx_hist_add_packet(ccid3_rx_hist
, &hcrx
->ccid3hcrx_hist
,
848 &hcrx
->ccid3hcrx_li_hist
, packet
,
849 hcrx
->ccid3hcrx_seqno_nonloss
);
853 static void ccid3_hc_rx_packet_recv(struct sock
*sk
, struct sk_buff
*skb
)
855 struct ccid3_hc_rx_sock
*hcrx
= ccid3_hc_rx_sk(sk
);
856 const struct dccp_options_received
*opt_recv
;
857 struct dccp_rx_hist_entry
*packet
;
858 u32 p_prev
, r_sample
, rtt_prev
;
859 int loss
, payload_size
;
862 opt_recv
= &dccp_sk(sk
)->dccps_options_received
;
864 switch (DCCP_SKB_CB(skb
)->dccpd_type
) {
866 if (hcrx
->ccid3hcrx_state
== TFRC_RSTATE_NO_DATA
)
868 case DCCP_PKT_DATAACK
:
869 if (opt_recv
->dccpor_timestamp_echo
== 0)
871 r_sample
= dccp_timestamp() - opt_recv
->dccpor_timestamp_echo
;
872 rtt_prev
= hcrx
->ccid3hcrx_rtt
;
873 r_sample
= dccp_sample_rtt(sk
, 10 * r_sample
);
875 if (hcrx
->ccid3hcrx_state
== TFRC_RSTATE_NO_DATA
)
876 hcrx
->ccid3hcrx_rtt
= r_sample
;
878 hcrx
->ccid3hcrx_rtt
= (hcrx
->ccid3hcrx_rtt
* 9) / 10 +
881 if (rtt_prev
!= hcrx
->ccid3hcrx_rtt
)
882 ccid3_pr_debug("%s(%p), New RTT=%uus, elapsed time=%u\n",
883 dccp_role(sk
), sk
, hcrx
->ccid3hcrx_rtt
,
884 opt_recv
->dccpor_elapsed_time
);
888 default: /* We're not interested in other packet types, move along */
892 packet
= dccp_rx_hist_entry_new(ccid3_rx_hist
, opt_recv
->dccpor_ndp
,
894 if (unlikely(packet
== NULL
)) {
895 DCCP_WARN("%s(%p), Not enough mem to add rx packet "
896 "to history, consider it lost!\n", dccp_role(sk
), sk
);
900 loss
= ccid3_hc_rx_detect_loss(sk
, packet
);
902 if (DCCP_SKB_CB(skb
)->dccpd_type
== DCCP_PKT_ACK
)
905 payload_size
= skb
->len
- dccp_hdr(skb
)->dccph_doff
* 4;
906 ccid3_hc_rx_update_s(hcrx
, payload_size
);
908 switch (hcrx
->ccid3hcrx_state
) {
909 case TFRC_RSTATE_NO_DATA
:
910 ccid3_pr_debug("%s(%p, state=%s), skb=%p, sending initial "
911 "feedback\n", dccp_role(sk
), sk
,
912 dccp_state_name(sk
->sk_state
), skb
);
913 ccid3_hc_rx_send_feedback(sk
);
914 ccid3_hc_rx_set_state(sk
, TFRC_RSTATE_DATA
);
916 case TFRC_RSTATE_DATA
:
917 hcrx
->ccid3hcrx_bytes_recv
+= payload_size
;
921 now
= ktime_get_real();
922 if ((ktime_us_delta(now
, hcrx
->ccid3hcrx_tstamp_last_ack
) -
923 (s64
)hcrx
->ccid3hcrx_rtt
) >= 0) {
924 hcrx
->ccid3hcrx_tstamp_last_ack
= now
;
925 ccid3_hc_rx_send_feedback(sk
);
928 case TFRC_RSTATE_TERM
:
929 DCCP_BUG("%s(%p) - Illegal state TERM", dccp_role(sk
), sk
);
933 /* Dealing with packet loss */
934 ccid3_pr_debug("%s(%p, state=%s), data loss! Reacting...\n",
935 dccp_role(sk
), sk
, dccp_state_name(sk
->sk_state
));
937 p_prev
= hcrx
->ccid3hcrx_p
;
939 /* Calculate loss event rate */
940 if (!list_empty(&hcrx
->ccid3hcrx_li_hist
)) {
941 u32 i_mean
= dccp_li_hist_calc_i_mean(&hcrx
->ccid3hcrx_li_hist
);
943 /* Scaling up by 1000000 as fixed decimal */
945 hcrx
->ccid3hcrx_p
= 1000000 / i_mean
;
947 DCCP_BUG("empty loss history");
949 if (hcrx
->ccid3hcrx_p
> p_prev
) {
950 ccid3_hc_rx_send_feedback(sk
);
955 static int ccid3_hc_rx_init(struct ccid
*ccid
, struct sock
*sk
)
957 struct ccid3_hc_rx_sock
*hcrx
= ccid_priv(ccid
);
959 ccid3_pr_debug("entry\n");
961 hcrx
->ccid3hcrx_state
= TFRC_RSTATE_NO_DATA
;
962 INIT_LIST_HEAD(&hcrx
->ccid3hcrx_hist
);
963 INIT_LIST_HEAD(&hcrx
->ccid3hcrx_li_hist
);
964 hcrx
->ccid3hcrx_tstamp_last_feedback
=
965 hcrx
->ccid3hcrx_tstamp_last_ack
= ktime_get_real();
966 hcrx
->ccid3hcrx_s
= 0;
967 hcrx
->ccid3hcrx_rtt
= 0;
971 static void ccid3_hc_rx_exit(struct sock
*sk
)
973 struct ccid3_hc_rx_sock
*hcrx
= ccid3_hc_rx_sk(sk
);
975 ccid3_hc_rx_set_state(sk
, TFRC_RSTATE_TERM
);
977 /* Empty packet history */
978 dccp_rx_hist_purge(ccid3_rx_hist
, &hcrx
->ccid3hcrx_hist
);
980 /* Empty loss interval history */
981 dccp_li_hist_purge(&hcrx
->ccid3hcrx_li_hist
);
984 static void ccid3_hc_rx_get_info(struct sock
*sk
, struct tcp_info
*info
)
986 const struct ccid3_hc_rx_sock
*hcrx
;
988 /* Listen socks doesn't have a private CCID block */
989 if (sk
->sk_state
== DCCP_LISTEN
)
992 hcrx
= ccid3_hc_rx_sk(sk
);
993 info
->tcpi_ca_state
= hcrx
->ccid3hcrx_state
;
994 info
->tcpi_options
|= TCPI_OPT_TIMESTAMPS
;
995 info
->tcpi_rcv_rtt
= hcrx
->ccid3hcrx_rtt
;
998 static int ccid3_hc_rx_getsockopt(struct sock
*sk
, const int optname
, int len
,
999 u32 __user
*optval
, int __user
*optlen
)
1001 const struct ccid3_hc_rx_sock
*hcrx
;
1004 /* Listen socks doesn't have a private CCID block */
1005 if (sk
->sk_state
== DCCP_LISTEN
)
1008 hcrx
= ccid3_hc_rx_sk(sk
);
1010 case DCCP_SOCKOPT_CCID_RX_INFO
:
1011 if (len
< sizeof(hcrx
->ccid3hcrx_tfrc
))
1013 len
= sizeof(hcrx
->ccid3hcrx_tfrc
);
1014 val
= &hcrx
->ccid3hcrx_tfrc
;
1017 return -ENOPROTOOPT
;
1020 if (put_user(len
, optlen
) || copy_to_user(optval
, val
, len
))
1026 static struct ccid_operations ccid3
= {
1027 .ccid_id
= DCCPC_CCID3
,
1028 .ccid_name
= "ccid3",
1029 .ccid_owner
= THIS_MODULE
,
1030 .ccid_hc_tx_obj_size
= sizeof(struct ccid3_hc_tx_sock
),
1031 .ccid_hc_tx_init
= ccid3_hc_tx_init
,
1032 .ccid_hc_tx_exit
= ccid3_hc_tx_exit
,
1033 .ccid_hc_tx_send_packet
= ccid3_hc_tx_send_packet
,
1034 .ccid_hc_tx_packet_sent
= ccid3_hc_tx_packet_sent
,
1035 .ccid_hc_tx_packet_recv
= ccid3_hc_tx_packet_recv
,
1036 .ccid_hc_tx_parse_options
= ccid3_hc_tx_parse_options
,
1037 .ccid_hc_rx_obj_size
= sizeof(struct ccid3_hc_rx_sock
),
1038 .ccid_hc_rx_init
= ccid3_hc_rx_init
,
1039 .ccid_hc_rx_exit
= ccid3_hc_rx_exit
,
1040 .ccid_hc_rx_insert_options
= ccid3_hc_rx_insert_options
,
1041 .ccid_hc_rx_packet_recv
= ccid3_hc_rx_packet_recv
,
1042 .ccid_hc_rx_get_info
= ccid3_hc_rx_get_info
,
1043 .ccid_hc_tx_get_info
= ccid3_hc_tx_get_info
,
1044 .ccid_hc_rx_getsockopt
= ccid3_hc_rx_getsockopt
,
1045 .ccid_hc_tx_getsockopt
= ccid3_hc_tx_getsockopt
,
1048 #ifdef CONFIG_IP_DCCP_CCID3_DEBUG
1049 module_param(ccid3_debug
, bool, 0444);
1050 MODULE_PARM_DESC(ccid3_debug
, "Enable debug messages");
1053 static __init
int ccid3_module_init(void)
1057 ccid3_rx_hist
= dccp_rx_hist_new("ccid3");
1058 if (ccid3_rx_hist
== NULL
)
1061 ccid3_tx_hist
= dccp_tx_hist_new("ccid3");
1062 if (ccid3_tx_hist
== NULL
)
1065 rc
= ccid_register(&ccid3
);
1072 dccp_tx_hist_delete(ccid3_tx_hist
);
1073 ccid3_tx_hist
= NULL
;
1075 dccp_rx_hist_delete(ccid3_rx_hist
);
1076 ccid3_rx_hist
= NULL
;
1079 module_init(ccid3_module_init
);
1081 static __exit
void ccid3_module_exit(void)
1083 ccid_unregister(&ccid3
);
1085 if (ccid3_tx_hist
!= NULL
) {
1086 dccp_tx_hist_delete(ccid3_tx_hist
);
1087 ccid3_tx_hist
= NULL
;
1089 if (ccid3_rx_hist
!= NULL
) {
1090 dccp_rx_hist_delete(ccid3_rx_hist
);
1091 ccid3_rx_hist
= NULL
;
1094 module_exit(ccid3_module_exit
);
1096 MODULE_AUTHOR("Ian McDonald <ian.mcdonald@jandi.co.nz>, "
1097 "Arnaldo Carvalho de Melo <acme@ghostprotocols.net>");
1098 MODULE_DESCRIPTION("DCCP TFRC CCID3 CCID");
1099 MODULE_LICENSE("GPL");
1100 MODULE_ALIAS("net-dccp-ccid-3");