]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blame - net/dccp/ccids/ccid2.c
dccp ccid-2: Remove old infrastructure
[mirror_ubuntu-jammy-kernel.git] / net / dccp / ccids / ccid2.c
CommitLineData
2a91aa39
AB
1/*
2 * net/dccp/ccids/ccid2.c
3 *
4 * Copyright (c) 2005, 2006 Andrea Bittau <a.bittau@cs.ucl.ac.uk>
5 *
6 * Changes to meet Linux coding standards, and DCCP infrastructure fixes.
7 *
8 * Copyright (c) 2006 Arnaldo Carvalho de Melo <acme@conectiva.com.br>
9 *
10 * This program is free software; you can redistribute it and/or modify
11 * it under the terms of the GNU General Public License as published by
12 * the Free Software Foundation; either version 2 of the License, or
13 * (at your option) any later version.
14 *
15 * This program is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 * GNU General Public License for more details.
19 *
20 * You should have received a copy of the GNU General Public License
21 * along with this program; if not, write to the Free Software
22 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
23 */
24
25/*
0e64e94e 26 * This implementation should follow RFC 4341
2a91aa39 27 */
86349c8d 28#include "../feat.h"
2a91aa39
AB
29#include "../ccid.h"
30#include "../dccp.h"
31#include "ccid2.h"
32
2a91aa39 33
8d424f6c 34#ifdef CONFIG_IP_DCCP_CCID2_DEBUG
84116716
GR
35static int ccid2_debug;
36#define ccid2_pr_debug(format, a...) DCCP_PR_DEBUG(ccid2_debug, format, ##a)
2a91aa39 37
2a91aa39
AB
38static void ccid2_hc_tx_check_sanity(const struct ccid2_hc_tx_sock *hctx)
39{
40 int len = 0;
2a91aa39 41 int pipe = 0;
1fb87509 42 struct ccid2_seq *seqp = hctx->seqh;
2a91aa39
AB
43
44 /* there is data in the chain */
1fb87509 45 if (seqp != hctx->seqt) {
2a91aa39
AB
46 seqp = seqp->ccid2s_prev;
47 len++;
48 if (!seqp->ccid2s_acked)
49 pipe++;
50
1fb87509 51 while (seqp != hctx->seqt) {
c0c736db 52 struct ccid2_seq *prev = seqp->ccid2s_prev;
2a91aa39 53
2a91aa39
AB
54 len++;
55 if (!prev->ccid2s_acked)
56 pipe++;
57
58 /* packets are sent sequentially */
5e28599a
GR
59 BUG_ON(dccp_delta_seqno(seqp->ccid2s_seq,
60 prev->ccid2s_seq ) >= 0);
29651cda
AB
61 BUG_ON(time_before(seqp->ccid2s_sent,
62 prev->ccid2s_sent));
2a91aa39
AB
63
64 seqp = prev;
65 }
66 }
67
1fb87509 68 BUG_ON(pipe != hctx->pipe);
2a91aa39
AB
69 ccid2_pr_debug("len of chain=%d\n", len);
70
71 do {
72 seqp = seqp->ccid2s_prev;
73 len++;
1fb87509 74 } while (seqp != hctx->seqh);
2a91aa39 75
2a91aa39 76 ccid2_pr_debug("total len=%d\n", len);
1fb87509 77 BUG_ON(len != hctx->seqbufc * CCID2_SEQBUF_LEN);
2a91aa39
AB
78}
79#else
84116716
GR
80#define ccid2_pr_debug(format, a...)
81#define ccid2_hc_tx_check_sanity(hctx)
2a91aa39
AB
82#endif
83
cd1f7d34 84static int ccid2_hc_tx_alloc_seq(struct ccid2_hc_tx_sock *hctx)
07978aab
AB
85{
86 struct ccid2_seq *seqp;
87 int i;
88
89 /* check if we have space to preserve the pointer to the buffer */
1fb87509 90 if (hctx->seqbufc >= sizeof(hctx->seqbuf) / sizeof(struct ccid2_seq *))
07978aab
AB
91 return -ENOMEM;
92
93 /* allocate buffer and initialize linked list */
cd1f7d34 94 seqp = kmalloc(CCID2_SEQBUF_LEN * sizeof(struct ccid2_seq), gfp_any());
07978aab
AB
95 if (seqp == NULL)
96 return -ENOMEM;
97
cd1f7d34 98 for (i = 0; i < (CCID2_SEQBUF_LEN - 1); i++) {
07978aab
AB
99 seqp[i].ccid2s_next = &seqp[i + 1];
100 seqp[i + 1].ccid2s_prev = &seqp[i];
101 }
cd1f7d34
GR
102 seqp[CCID2_SEQBUF_LEN - 1].ccid2s_next = seqp;
103 seqp->ccid2s_prev = &seqp[CCID2_SEQBUF_LEN - 1];
07978aab
AB
104
105 /* This is the first allocation. Initiate the head and tail. */
1fb87509
GR
106 if (hctx->seqbufc == 0)
107 hctx->seqh = hctx->seqt = seqp;
07978aab
AB
108 else {
109 /* link the existing list with the one we just created */
1fb87509
GR
110 hctx->seqh->ccid2s_next = seqp;
111 seqp->ccid2s_prev = hctx->seqh;
07978aab 112
1fb87509
GR
113 hctx->seqt->ccid2s_prev = &seqp[CCID2_SEQBUF_LEN - 1];
114 seqp[CCID2_SEQBUF_LEN - 1].ccid2s_next = hctx->seqt;
07978aab
AB
115 }
116
117 /* store the original pointer to the buffer so we can free it */
1fb87509
GR
118 hctx->seqbuf[hctx->seqbufc] = seqp;
119 hctx->seqbufc++;
07978aab
AB
120
121 return 0;
122}
123
6b57c93d 124static int ccid2_hc_tx_send_packet(struct sock *sk, struct sk_buff *skb)
2a91aa39 125{
6c583248 126 struct ccid2_hc_tx_sock *hctx = ccid2_hc_tx_sk(sk);
2a91aa39 127
1fb87509 128 if (hctx->pipe < hctx->cwnd)
83399361 129 return 0;
2a91aa39 130
446dec30 131 return 1; /* XXX CCID should dequeue when ready instead of polling */
2a91aa39
AB
132}
133
df054e1d 134static void ccid2_change_l_ack_ratio(struct sock *sk, u32 val)
2a91aa39
AB
135{
136 struct dccp_sock *dp = dccp_sk(sk);
1fb87509 137 u32 max_ratio = DIV_ROUND_UP(ccid2_hc_tx_sk(sk)->cwnd, 2);
d50ad163 138
2a91aa39 139 /*
d50ad163
GR
140 * Ensure that Ack Ratio does not exceed ceil(cwnd/2), which is (2) from
141 * RFC 4341, 6.1.2. We ignore the statement that Ack Ratio 2 is always
142 * acceptable since this causes starvation/deadlock whenever cwnd < 2.
143 * The same problem arises when Ack Ratio is 0 (ie. Ack Ratio disabled).
2a91aa39 144 */
d50ad163
GR
145 if (val == 0 || val > max_ratio) {
146 DCCP_WARN("Limiting Ack Ratio (%u) to %u\n", val, max_ratio);
147 val = max_ratio;
2a91aa39 148 }
86349c8d
GR
149 if (val > DCCPF_ACK_RATIO_MAX)
150 val = DCCPF_ACK_RATIO_MAX;
2a91aa39 151
d50ad163
GR
152 if (val == dp->dccps_l_ack_ratio)
153 return;
154
df054e1d 155 ccid2_pr_debug("changing local ack ratio to %u\n", val);
2a91aa39
AB
156 dp->dccps_l_ack_ratio = val;
157}
158
593f16aa
AB
159static void ccid2_change_srtt(struct ccid2_hc_tx_sock *hctx, long val)
160{
161 ccid2_pr_debug("change SRTT to %ld\n", val);
1fb87509 162 hctx->srtt = val;
593f16aa
AB
163}
164
2a91aa39
AB
165static void ccid2_start_rto_timer(struct sock *sk);
166
167static void ccid2_hc_tx_rto_expire(unsigned long data)
168{
169 struct sock *sk = (struct sock *)data;
170 struct ccid2_hc_tx_sock *hctx = ccid2_hc_tx_sk(sk);
171 long s;
172
2a91aa39
AB
173 bh_lock_sock(sk);
174 if (sock_owned_by_user(sk)) {
1fb87509 175 sk_reset_timer(sk, &hctx->rtotimer, jiffies + HZ / 5);
2a91aa39
AB
176 goto out;
177 }
178
179 ccid2_pr_debug("RTO_EXPIRE\n");
180
181 ccid2_hc_tx_check_sanity(hctx);
182
183 /* back-off timer */
1fb87509 184 hctx->rto <<= 1;
2a91aa39 185
1fb87509 186 s = hctx->rto / HZ;
2a91aa39 187 if (s > 60)
1fb87509 188 hctx->rto = 60 * HZ;
2a91aa39
AB
189
190 ccid2_start_rto_timer(sk);
191
192 /* adjust pipe, cwnd etc */
1fb87509
GR
193 hctx->ssthresh = hctx->cwnd / 2;
194 if (hctx->ssthresh < 2)
195 hctx->ssthresh = 2;
196 hctx->cwnd = 1;
197 hctx->pipe = 0;
2a91aa39
AB
198
199 /* clear state about stuff we sent */
1fb87509
GR
200 hctx->seqt = hctx->seqh;
201 hctx->packets_acked = 0;
2a91aa39
AB
202
203 /* clear ack ratio state. */
1fb87509
GR
204 hctx->rpseq = 0;
205 hctx->rpdupack = -1;
2a91aa39
AB
206 ccid2_change_l_ack_ratio(sk, 1);
207 ccid2_hc_tx_check_sanity(hctx);
208out:
209 bh_unlock_sock(sk);
77ff72d5 210 sock_put(sk);
2a91aa39
AB
211}
212
213static void ccid2_start_rto_timer(struct sock *sk)
214{
215 struct ccid2_hc_tx_sock *hctx = ccid2_hc_tx_sk(sk);
216
1fb87509 217 ccid2_pr_debug("setting RTO timeout=%ld\n", hctx->rto);
2a91aa39 218
1fb87509
GR
219 BUG_ON(timer_pending(&hctx->rtotimer));
220 sk_reset_timer(sk, &hctx->rtotimer,
221 jiffies + hctx->rto);
2a91aa39
AB
222}
223
c506d91d 224static void ccid2_hc_tx_packet_sent(struct sock *sk, unsigned int len)
2a91aa39
AB
225{
226 struct dccp_sock *dp = dccp_sk(sk);
227 struct ccid2_hc_tx_sock *hctx = ccid2_hc_tx_sk(sk);
07978aab 228 struct ccid2_seq *next;
2a91aa39 229
1fb87509 230 hctx->pipe++;
2a91aa39 231
1fb87509
GR
232 hctx->seqh->ccid2s_seq = dp->dccps_gss;
233 hctx->seqh->ccid2s_acked = 0;
234 hctx->seqh->ccid2s_sent = jiffies;
2a91aa39 235
1fb87509 236 next = hctx->seqh->ccid2s_next;
07978aab 237 /* check if we need to alloc more space */
1fb87509 238 if (next == hctx->seqt) {
7d9e8931
GR
239 if (ccid2_hc_tx_alloc_seq(hctx)) {
240 DCCP_CRIT("packet history - out of memory!");
241 /* FIXME: find a more graceful way to bail out */
242 return;
243 }
1fb87509
GR
244 next = hctx->seqh->ccid2s_next;
245 BUG_ON(next == hctx->seqt);
2a91aa39 246 }
1fb87509 247 hctx->seqh = next;
07978aab 248
1fb87509 249 ccid2_pr_debug("cwnd=%d pipe=%d\n", hctx->cwnd, hctx->pipe);
2a91aa39 250
900bfed4
GR
251 /*
252 * FIXME: The code below is broken and the variables have been removed
253 * from the socket struct. The `ackloss' variable was always set to 0,
254 * and with arsent there are several problems:
255 * (i) it doesn't just count the number of Acks, but all sent packets;
256 * (ii) it is expressed in # of packets, not # of windows, so the
257 * comparison below uses the wrong formula: Appendix A of RFC 4341
258 * comes up with the number K = cwnd / (R^2 - R) of consecutive windows
259 * of data with no lost or marked Ack packets. If arsent were the # of
260 * consecutive Acks received without loss, then Ack Ratio needs to be
261 * decreased by 1 when
262 * arsent >= K * cwnd / R = cwnd^2 / (R^3 - R^2)
263 * where cwnd / R is the number of Acks received per window of data
264 * (cf. RFC 4341, App. A). The problems are that
265 * - arsent counts other packets as well;
266 * - the comparison uses a formula different from RFC 4341;
267 * - computing a cubic/quadratic equation each time is too complicated.
268 * Hence a different algorithm is needed.
269 */
270#if 0
2a91aa39 271 /* Ack Ratio. Need to maintain a concept of how many windows we sent */
1fb87509 272 hctx->arsent++;
2a91aa39 273 /* We had an ack loss in this window... */
1fb87509
GR
274 if (hctx->ackloss) {
275 if (hctx->arsent >= hctx->cwnd) {
276 hctx->arsent = 0;
277 hctx->ackloss = 0;
2a91aa39 278 }
c0c736db
ACM
279 } else {
280 /* No acks lost up to now... */
2a91aa39
AB
281 /* decrease ack ratio if enough packets were sent */
282 if (dp->dccps_l_ack_ratio > 1) {
283 /* XXX don't calculate denominator each time */
c0c736db
ACM
284 int denom = dp->dccps_l_ack_ratio * dp->dccps_l_ack_ratio -
285 dp->dccps_l_ack_ratio;
2a91aa39 286
1fb87509 287 denom = hctx->cwnd * hctx->cwnd / denom;
2a91aa39 288
1fb87509 289 if (hctx->arsent >= denom) {
2a91aa39 290 ccid2_change_l_ack_ratio(sk, dp->dccps_l_ack_ratio - 1);
1fb87509 291 hctx->arsent = 0;
2a91aa39 292 }
c0c736db
ACM
293 } else {
294 /* we can't increase ack ratio further [1] */
1fb87509 295 hctx->arsent = 0; /* or maybe set it to cwnd*/
2a91aa39
AB
296 }
297 }
900bfed4 298#endif
2a91aa39
AB
299
300 /* setup RTO timer */
1fb87509 301 if (!timer_pending(&hctx->rtotimer))
2a91aa39 302 ccid2_start_rto_timer(sk);
c0c736db 303
8d424f6c 304#ifdef CONFIG_IP_DCCP_CCID2_DEBUG
2a91aa39 305 do {
1fb87509 306 struct ccid2_seq *seqp = hctx->seqt;
2a91aa39 307
1fb87509 308 while (seqp != hctx->seqh) {
2a91aa39 309 ccid2_pr_debug("out seq=%llu acked=%d time=%lu\n",
8109b02b 310 (unsigned long long)seqp->ccid2s_seq,
234af484 311 seqp->ccid2s_acked, seqp->ccid2s_sent);
2a91aa39
AB
312 seqp = seqp->ccid2s_next;
313 }
c0c736db 314 } while (0);
2a91aa39
AB
315 ccid2_pr_debug("=========\n");
316 ccid2_hc_tx_check_sanity(hctx);
317#endif
318}
319
320/* XXX Lame code duplication!
321 * returns -1 if none was found.
322 * else returns the next offset to use in the function call.
323 */
324static int ccid2_ackvector(struct sock *sk, struct sk_buff *skb, int offset,
325 unsigned char **vec, unsigned char *veclen)
326{
c9eaf173
YH
327 const struct dccp_hdr *dh = dccp_hdr(skb);
328 unsigned char *options = (unsigned char *)dh + dccp_hdr_len(skb);
329 unsigned char *opt_ptr;
330 const unsigned char *opt_end = (unsigned char *)dh +
331 (dh->dccph_doff * 4);
332 unsigned char opt, len;
333 unsigned char *value;
2a91aa39
AB
334
335 BUG_ON(offset < 0);
336 options += offset;
337 opt_ptr = options;
338 if (opt_ptr >= opt_end)
339 return -1;
340
341 while (opt_ptr != opt_end) {
c9eaf173
YH
342 opt = *opt_ptr++;
343 len = 0;
344 value = NULL;
345
346 /* Check if this isn't a single byte option */
347 if (opt > DCCPO_MAX_RESERVED) {
348 if (opt_ptr == opt_end)
349 goto out_invalid_option;
350
351 len = *opt_ptr++;
352 if (len < 3)
353 goto out_invalid_option;
354 /*
355 * Remove the type and len fields, leaving
356 * just the value size
357 */
358 len -= 2;
359 value = opt_ptr;
360 opt_ptr += len;
361
362 if (opt_ptr > opt_end)
363 goto out_invalid_option;
364 }
2a91aa39
AB
365
366 switch (opt) {
367 case DCCPO_ACK_VECTOR_0:
368 case DCCPO_ACK_VECTOR_1:
369 *vec = value;
370 *veclen = len;
371 return offset + (opt_ptr - options);
2a91aa39
AB
372 }
373 }
374
375 return -1;
376
377out_invalid_option:
59348b19 378 DCCP_BUG("Invalid option - this should not happen (previous parsing)!");
2a91aa39
AB
379 return -1;
380}
381
77ff72d5 382static void ccid2_hc_tx_kill_rto_timer(struct sock *sk)
2a91aa39 383{
77ff72d5
AB
384 struct ccid2_hc_tx_sock *hctx = ccid2_hc_tx_sk(sk);
385
1fb87509 386 sk_stop_timer(sk, &hctx->rtotimer);
77ff72d5 387 ccid2_pr_debug("deleted RTO timer\n");
2a91aa39
AB
388}
389
390static inline void ccid2_new_ack(struct sock *sk,
c9eaf173 391 struct ccid2_seq *seqp,
2a91aa39
AB
392 unsigned int *maxincr)
393{
394 struct ccid2_hc_tx_sock *hctx = ccid2_hc_tx_sk(sk);
395
1fb87509
GR
396 if (hctx->cwnd < hctx->ssthresh) {
397 if (*maxincr > 0 && ++hctx->packets_acked == 2) {
398 hctx->cwnd += 1;
399 *maxincr -= 1;
400 hctx->packets_acked = 0;
2a91aa39 401 }
1fb87509
GR
402 } else if (++hctx->packets_acked >= hctx->cwnd) {
403 hctx->cwnd += 1;
404 hctx->packets_acked = 0;
2a91aa39
AB
405 }
406
407 /* update RTO */
1fb87509
GR
408 if (hctx->srtt == -1 ||
409 time_after(jiffies, hctx->lastrtt + hctx->srtt)) {
29651cda 410 unsigned long r = (long)jiffies - (long)seqp->ccid2s_sent;
2a91aa39
AB
411 int s;
412
413 /* first measurement */
1fb87509 414 if (hctx->srtt == -1) {
2a91aa39 415 ccid2_pr_debug("R: %lu Time=%lu seq=%llu\n",
8109b02b 416 r, jiffies,
234af484 417 (unsigned long long)seqp->ccid2s_seq);
593f16aa 418 ccid2_change_srtt(hctx, r);
1fb87509 419 hctx->rttvar = r >> 1;
c0c736db 420 } else {
2a91aa39 421 /* RTTVAR */
1fb87509 422 long tmp = hctx->srtt - r;
593f16aa
AB
423 long srtt;
424
2a91aa39
AB
425 if (tmp < 0)
426 tmp *= -1;
427
428 tmp >>= 2;
1fb87509
GR
429 hctx->rttvar *= 3;
430 hctx->rttvar >>= 2;
431 hctx->rttvar += tmp;
2a91aa39
AB
432
433 /* SRTT */
1fb87509 434 srtt = hctx->srtt;
593f16aa
AB
435 srtt *= 7;
436 srtt >>= 3;
2a91aa39 437 tmp = r >> 3;
593f16aa
AB
438 srtt += tmp;
439 ccid2_change_srtt(hctx, srtt);
2a91aa39 440 }
1fb87509 441 s = hctx->rttvar << 2;
2a91aa39
AB
442 /* clock granularity is 1 when based on jiffies */
443 if (!s)
444 s = 1;
1fb87509 445 hctx->rto = hctx->srtt + s;
2a91aa39
AB
446
447 /* must be at least a second */
1fb87509 448 s = hctx->rto / HZ;
2a91aa39
AB
449 /* DCCP doesn't require this [but I like it cuz my code sux] */
450#if 1
451 if (s < 1)
1fb87509 452 hctx->rto = HZ;
2a91aa39
AB
453#endif
454 /* max 60 seconds */
455 if (s > 60)
1fb87509 456 hctx->rto = HZ * 60;
2a91aa39 457
1fb87509 458 hctx->lastrtt = jiffies;
2a91aa39
AB
459
460 ccid2_pr_debug("srtt: %ld rttvar: %ld rto: %ld (HZ=%d) R=%lu\n",
1fb87509
GR
461 hctx->srtt, hctx->rttvar,
462 hctx->rto, HZ, r);
2a91aa39
AB
463 }
464
465 /* we got a new ack, so re-start RTO timer */
77ff72d5 466 ccid2_hc_tx_kill_rto_timer(sk);
2a91aa39
AB
467 ccid2_start_rto_timer(sk);
468}
469
77ff72d5 470static void ccid2_hc_tx_dec_pipe(struct sock *sk)
2a91aa39 471{
77ff72d5
AB
472 struct ccid2_hc_tx_sock *hctx = ccid2_hc_tx_sk(sk);
473
1fb87509 474 if (hctx->pipe == 0)
95b21d7e
GR
475 DCCP_BUG("pipe == 0");
476 else
1fb87509 477 hctx->pipe--;
2a91aa39 478
1fb87509 479 if (hctx->pipe == 0)
77ff72d5 480 ccid2_hc_tx_kill_rto_timer(sk);
2a91aa39
AB
481}
482
d50ad163 483static void ccid2_congestion_event(struct sock *sk, struct ccid2_seq *seqp)
374bcf32 484{
d50ad163
GR
485 struct ccid2_hc_tx_sock *hctx = ccid2_hc_tx_sk(sk);
486
1fb87509 487 if (time_before(seqp->ccid2s_sent, hctx->last_cong)) {
374bcf32
AB
488 ccid2_pr_debug("Multiple losses in an RTT---treating as one\n");
489 return;
490 }
491
1fb87509 492 hctx->last_cong = jiffies;
374bcf32 493
1fb87509
GR
494 hctx->cwnd = hctx->cwnd / 2 ? : 1U;
495 hctx->ssthresh = max(hctx->cwnd, 2U);
d50ad163
GR
496
497 /* Avoid spurious timeouts resulting from Ack Ratio > cwnd */
1fb87509
GR
498 if (dccp_sk(sk)->dccps_l_ack_ratio > hctx->cwnd)
499 ccid2_change_l_ack_ratio(sk, hctx->cwnd);
374bcf32
AB
500}
501
2a91aa39
AB
502static void ccid2_hc_tx_packet_recv(struct sock *sk, struct sk_buff *skb)
503{
504 struct dccp_sock *dp = dccp_sk(sk);
505 struct ccid2_hc_tx_sock *hctx = ccid2_hc_tx_sk(sk);
506 u64 ackno, seqno;
507 struct ccid2_seq *seqp;
508 unsigned char *vector;
509 unsigned char veclen;
510 int offset = 0;
511 int done = 0;
2a91aa39
AB
512 unsigned int maxincr = 0;
513
514 ccid2_hc_tx_check_sanity(hctx);
515 /* check reverse path congestion */
516 seqno = DCCP_SKB_CB(skb)->dccpd_seq;
517
518 /* XXX this whole "algorithm" is broken. Need to fix it to keep track
519 * of the seqnos of the dupacks so that rpseq and rpdupack are correct
520 * -sorbo.
521 */
522 /* need to bootstrap */
1fb87509
GR
523 if (hctx->rpdupack == -1) {
524 hctx->rpdupack = 0;
525 hctx->rpseq = seqno;
c0c736db 526 } else {
2a91aa39 527 /* check if packet is consecutive */
1fb87509
GR
528 if (dccp_delta_seqno(hctx->rpseq, seqno) == 1)
529 hctx->rpseq = seqno;
2a91aa39 530 /* it's a later packet */
1fb87509
GR
531 else if (after48(seqno, hctx->rpseq)) {
532 hctx->rpdupack++;
2a91aa39
AB
533
534 /* check if we got enough dupacks */
1fb87509
GR
535 if (hctx->rpdupack >= NUMDUPACK) {
536 hctx->rpdupack = -1; /* XXX lame */
537 hctx->rpseq = 0;
2a91aa39 538
df054e1d 539 ccid2_change_l_ack_ratio(sk, 2 * dp->dccps_l_ack_ratio);
2a91aa39
AB
540 }
541 }
542 }
543
544 /* check forward path congestion */
545 /* still didn't send out new data packets */
1fb87509 546 if (hctx->seqh == hctx->seqt)
2a91aa39
AB
547 return;
548
549 switch (DCCP_SKB_CB(skb)->dccpd_type) {
550 case DCCP_PKT_ACK:
551 case DCCP_PKT_DATAACK:
552 break;
2a91aa39
AB
553 default:
554 return;
555 }
556
557 ackno = DCCP_SKB_CB(skb)->dccpd_ack_seq;
1fb87509
GR
558 if (after48(ackno, hctx->high_ack))
559 hctx->high_ack = ackno;
32aac18d 560
1fb87509 561 seqp = hctx->seqt;
32aac18d
AB
562 while (before48(seqp->ccid2s_seq, ackno)) {
563 seqp = seqp->ccid2s_next;
1fb87509
GR
564 if (seqp == hctx->seqh) {
565 seqp = hctx->seqh->ccid2s_prev;
32aac18d
AB
566 break;
567 }
568 }
2a91aa39 569
a3020025
GR
570 /*
571 * In slow-start, cwnd can increase up to a maximum of Ack Ratio/2
572 * packets per acknowledgement. Rounding up avoids that cwnd is not
573 * advanced when Ack Ratio is 1 and gives a slight edge otherwise.
2a91aa39 574 */
1fb87509 575 if (hctx->cwnd < hctx->ssthresh)
a3020025 576 maxincr = DIV_ROUND_UP(dp->dccps_l_ack_ratio, 2);
2a91aa39
AB
577
578 /* go through all ack vectors */
579 while ((offset = ccid2_ackvector(sk, skb, offset,
580 &vector, &veclen)) != -1) {
581 /* go through this ack vector */
582 while (veclen--) {
ff49e270 583 u64 ackno_end_rl = SUB48(ackno, dccp_ackvec_runlen(vector));
2a91aa39 584
234af484
RD
585 ccid2_pr_debug("ackvec start:%llu end:%llu\n",
586 (unsigned long long)ackno,
587 (unsigned long long)ackno_end_rl);
2a91aa39
AB
588 /* if the seqno we are analyzing is larger than the
589 * current ackno, then move towards the tail of our
590 * seqnos.
591 */
592 while (after48(seqp->ccid2s_seq, ackno)) {
1fb87509 593 if (seqp == hctx->seqt) {
2a91aa39
AB
594 done = 1;
595 break;
596 }
597 seqp = seqp->ccid2s_prev;
598 }
599 if (done)
600 break;
601
602 /* check all seqnos in the range of the vector
603 * run length
604 */
605 while (between48(seqp->ccid2s_seq,ackno_end_rl,ackno)) {
ff49e270 606 const u8 state = dccp_ackvec_state(vector);
2a91aa39
AB
607
608 /* new packet received or marked */
ff49e270 609 if (state != DCCPAV_NOT_RECEIVED &&
2a91aa39 610 !seqp->ccid2s_acked) {
ff49e270 611 if (state == DCCPAV_ECN_MARKED)
d50ad163 612 ccid2_congestion_event(sk,
374bcf32 613 seqp);
ff49e270 614 else
2a91aa39
AB
615 ccid2_new_ack(sk, seqp,
616 &maxincr);
2a91aa39
AB
617
618 seqp->ccid2s_acked = 1;
619 ccid2_pr_debug("Got ack for %llu\n",
234af484 620 (unsigned long long)seqp->ccid2s_seq);
77ff72d5 621 ccid2_hc_tx_dec_pipe(sk);
2a91aa39 622 }
1fb87509 623 if (seqp == hctx->seqt) {
2a91aa39
AB
624 done = 1;
625 break;
626 }
3de5489f 627 seqp = seqp->ccid2s_prev;
2a91aa39
AB
628 }
629 if (done)
630 break;
631
cfbbeabc 632 ackno = SUB48(ackno_end_rl, 1);
2a91aa39
AB
633 vector++;
634 }
635 if (done)
636 break;
637 }
638
639 /* The state about what is acked should be correct now
640 * Check for NUMDUPACK
641 */
1fb87509
GR
642 seqp = hctx->seqt;
643 while (before48(seqp->ccid2s_seq, hctx->high_ack)) {
32aac18d 644 seqp = seqp->ccid2s_next;
1fb87509
GR
645 if (seqp == hctx->seqh) {
646 seqp = hctx->seqh->ccid2s_prev;
32aac18d
AB
647 break;
648 }
649 }
2a91aa39
AB
650 done = 0;
651 while (1) {
652 if (seqp->ccid2s_acked) {
653 done++;
63df18ad 654 if (done == NUMDUPACK)
2a91aa39 655 break;
2a91aa39 656 }
1fb87509 657 if (seqp == hctx->seqt)
2a91aa39 658 break;
2a91aa39
AB
659 seqp = seqp->ccid2s_prev;
660 }
661
662 /* If there are at least 3 acknowledgements, anything unacknowledged
663 * below the last sequence number is considered lost
664 */
63df18ad 665 if (done == NUMDUPACK) {
2a91aa39
AB
666 struct ccid2_seq *last_acked = seqp;
667
668 /* check for lost packets */
669 while (1) {
670 if (!seqp->ccid2s_acked) {
374bcf32 671 ccid2_pr_debug("Packet lost: %llu\n",
234af484 672 (unsigned long long)seqp->ccid2s_seq);
374bcf32
AB
673 /* XXX need to traverse from tail -> head in
674 * order to detect multiple congestion events in
675 * one ack vector.
676 */
d50ad163 677 ccid2_congestion_event(sk, seqp);
77ff72d5 678 ccid2_hc_tx_dec_pipe(sk);
2a91aa39 679 }
1fb87509 680 if (seqp == hctx->seqt)
2a91aa39
AB
681 break;
682 seqp = seqp->ccid2s_prev;
683 }
684
1fb87509 685 hctx->seqt = last_acked;
2a91aa39
AB
686 }
687
688 /* trim acked packets in tail */
1fb87509
GR
689 while (hctx->seqt != hctx->seqh) {
690 if (!hctx->seqt->ccid2s_acked)
2a91aa39
AB
691 break;
692
1fb87509 693 hctx->seqt = hctx->seqt->ccid2s_next;
2a91aa39
AB
694 }
695
2a91aa39
AB
696 ccid2_hc_tx_check_sanity(hctx);
697}
698
91f0ebf7 699static int ccid2_hc_tx_init(struct ccid *ccid, struct sock *sk)
2a91aa39 700{
c9eaf173 701 struct ccid2_hc_tx_sock *hctx = ccid_priv(ccid);
b00d2bbc
GR
702 struct dccp_sock *dp = dccp_sk(sk);
703 u32 max_ratio;
2a91aa39 704
b00d2bbc 705 /* RFC 4341, 5: initialise ssthresh to arbitrarily high (max) value */
1fb87509 706 hctx->ssthresh = ~0U;
2a91aa39 707
b00d2bbc
GR
708 /*
709 * RFC 4341, 5: "The cwnd parameter is initialized to at most four
710 * packets for new connections, following the rules from [RFC3390]".
711 * We need to convert the bytes of RFC3390 into the packets of RFC 4341.
712 */
1fb87509 713 hctx->cwnd = clamp(4380U / dp->dccps_mss_cache, 2U, 4U);
b00d2bbc
GR
714
715 /* Make sure that Ack Ratio is enabled and within bounds. */
1fb87509 716 max_ratio = DIV_ROUND_UP(hctx->cwnd, 2);
b00d2bbc
GR
717 if (dp->dccps_l_ack_ratio == 0 || dp->dccps_l_ack_ratio > max_ratio)
718 dp->dccps_l_ack_ratio = max_ratio;
719
2a91aa39 720 /* XXX init ~ to window size... */
cd1f7d34 721 if (ccid2_hc_tx_alloc_seq(hctx))
2a91aa39 722 return -ENOMEM;
91f0ebf7 723
1fb87509 724 hctx->rto = 3 * HZ;
593f16aa 725 ccid2_change_srtt(hctx, -1);
1fb87509
GR
726 hctx->rttvar = -1;
727 hctx->rpdupack = -1;
728 hctx->last_cong = jiffies;
729 setup_timer(&hctx->rtotimer, ccid2_hc_tx_rto_expire, (unsigned long)sk);
2a91aa39
AB
730
731 ccid2_hc_tx_check_sanity(hctx);
732 return 0;
733}
734
735static void ccid2_hc_tx_exit(struct sock *sk)
736{
c9eaf173 737 struct ccid2_hc_tx_sock *hctx = ccid2_hc_tx_sk(sk);
07978aab 738 int i;
2a91aa39 739
77ff72d5 740 ccid2_hc_tx_kill_rto_timer(sk);
07978aab 741
1fb87509
GR
742 for (i = 0; i < hctx->seqbufc; i++)
743 kfree(hctx->seqbuf[i]);
744 hctx->seqbufc = 0;
2a91aa39
AB
745}
746
747static void ccid2_hc_rx_packet_recv(struct sock *sk, struct sk_buff *skb)
748{
749 const struct dccp_sock *dp = dccp_sk(sk);
750 struct ccid2_hc_rx_sock *hcrx = ccid2_hc_rx_sk(sk);
751
752 switch (DCCP_SKB_CB(skb)->dccpd_type) {
753 case DCCP_PKT_DATA:
754 case DCCP_PKT_DATAACK:
1fb87509
GR
755 hcrx->data++;
756 if (hcrx->data >= dp->dccps_r_ack_ratio) {
2a91aa39 757 dccp_send_ack(sk);
1fb87509 758 hcrx->data = 0;
2a91aa39
AB
759 }
760 break;
761 }
762}
763
91f0ebf7 764static struct ccid_operations ccid2 = {
3dd9a7c3 765 .ccid_id = DCCPC_CCID2,
84a97b0a 766 .ccid_name = "TCP-like",
2a91aa39 767 .ccid_owner = THIS_MODULE,
91f0ebf7 768 .ccid_hc_tx_obj_size = sizeof(struct ccid2_hc_tx_sock),
2a91aa39
AB
769 .ccid_hc_tx_init = ccid2_hc_tx_init,
770 .ccid_hc_tx_exit = ccid2_hc_tx_exit,
771 .ccid_hc_tx_send_packet = ccid2_hc_tx_send_packet,
772 .ccid_hc_tx_packet_sent = ccid2_hc_tx_packet_sent,
773 .ccid_hc_tx_packet_recv = ccid2_hc_tx_packet_recv,
91f0ebf7 774 .ccid_hc_rx_obj_size = sizeof(struct ccid2_hc_rx_sock),
2a91aa39
AB
775 .ccid_hc_rx_packet_recv = ccid2_hc_rx_packet_recv,
776};
777
84116716 778#ifdef CONFIG_IP_DCCP_CCID2_DEBUG
43264991 779module_param(ccid2_debug, bool, 0644);
2a91aa39 780MODULE_PARM_DESC(ccid2_debug, "Enable debug messages");
84116716 781#endif
2a91aa39
AB
782
783static __init int ccid2_module_init(void)
784{
785 return ccid_register(&ccid2);
786}
787module_init(ccid2_module_init);
788
789static __exit void ccid2_module_exit(void)
790{
791 ccid_unregister(&ccid2);
792}
793module_exit(ccid2_module_exit);
794
795MODULE_AUTHOR("Andrea Bittau <a.bittau@cs.ucl.ac.uk>");
c0c736db 796MODULE_DESCRIPTION("DCCP TCP-Like (CCID2) CCID");
2a91aa39
AB
797MODULE_LICENSE("GPL");
798MODULE_ALIAS("net-dccp-ccid-2");