1/*
2 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994, 1995
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * @(#)tcp_sack.c 8.12 (Berkeley) 5/24/95
30 * $FreeBSD: head/sys/netinet/tcp_sack.c 130989 2004-06-23 21:04:37Z ps $
31 */
32
33/*
34 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994
35 * The Regents of the University of California. All rights reserved.
36 *
37 * Redistribution and use in source and binary forms, with or without
38 * modification, are permitted provided that the following conditions
39 * are met:
40 * 1. Redistributions of source code must retain the above copyright
41 * notice, this list of conditions and the following disclaimer.
42 * 2. Redistributions in binary form must reproduce the above copyright
43 * notice, this list of conditions and the following disclaimer in the
44 * documentation and/or other materials provided with the distribution.
45 * 3. Neither the name of the University nor the names of its contributors
46 * may be used to endorse or promote products derived from this software
47 * without specific prior written permission.
48 *
49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59 * SUCH DAMAGE.
60 *
61 * @@(#)COPYRIGHT 1.1 (NRL) 17 January 1995
62 *
63 * NRL grants permission for redistribution and use in source and binary
64 * forms, with or without modification, of the software and documentation
65 * created at NRL provided that the following conditions are met:
66 *
67 * 1. Redistributions of source code must retain the above copyright
68 * notice, this list of conditions and the following disclaimer.
69 * 2. Redistributions in binary form must reproduce the above copyright
70 * notice, this list of conditions and the following disclaimer in the
71 * documentation and/or other materials provided with the distribution.
72 * 3. All advertising materials mentioning features or use of this software
73 * must display the following acknowledgements:
74 * This product includes software developed by the University of
75 * California, Berkeley and its contributors.
76 * This product includes software developed at the Information
77 * Technology Division, US Naval Research Laboratory.
78 * 4. Neither the name of the NRL nor the names of its contributors
79 * may be used to endorse or promote products derived from this software
80 * without specific prior written permission.
81 *
82 * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS
83 * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
84 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
85 * PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL NRL OR
86 * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
87 * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
88 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
89 * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
90 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
91 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
92 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
93 *
94 * The views and conclusions contained in the software and documentation
95 * are those of the authors and should not be interpreted as representing
96 * official policies, either expressed or implied, of the US Naval
97 * Research Laboratory (NRL).
98 */
99#include "opt_ipfw.h" /* for ipfw_fwd */
100#include "opt_inet.h"
101#include "opt_inet6.h"
102#include "opt_ipsec.h"
103#include "opt_tcpdebug.h"
104#include "opt_tcp_input.h"
105#include "opt_tcp_sack.h"
106
107#include <sys/param.h>
108#include <sys/systm.h>
109#include <sys/kernel.h>
110#include <sys/sysctl.h>
111#include <sys/malloc.h>
112#include <sys/mbuf.h>
113#include <sys/proc.h> /* for proc0 declaration */
114#include <sys/protosw.h>
115#include <sys/socket.h>
116#include <sys/socketvar.h>
117#include <sys/syslog.h>
118#include <sys/systm.h>
119
120#include <machine/cpu.h> /* before tcp_seq.h, for tcp_random18() */
121
122#include <vm/uma.h>
123
124#include <net/if.h>
125#include <net/route.h>
126
127#include <netinet/in.h>
128#include <netinet/in_systm.h>
129#include <netinet/ip.h>
130#include <netinet/ip_icmp.h> /* for ICMP_BANDLIM */
131#include <netinet/in_var.h>
132#include <netinet/icmp_var.h> /* for ICMP_BANDLIM */
133#include <netinet/in_pcb.h>
134#include <netinet/ip_var.h>
135#include <netinet/ip6.h>
136#include <netinet/icmp6.h>
137#include <netinet6/nd6.h>
138#include <netinet6/ip6_var.h>
139#include <netinet6/in6_pcb.h>
140#include <netinet/tcp.h>
141#include <netinet/tcp_fsm.h>
142#include <netinet/tcp_seq.h>
143#include <netinet/tcp_timer.h>
144#include <netinet/tcp_var.h>
145#include <netinet6/tcp6_var.h>
146#include <netinet/tcpip.h>
147#ifdef TCPDEBUG
148#include <netinet/tcp_debug.h>
149
150u_char tcp_saveipgen[40]; /* the size must be of max ip header, now IPv6 */
151struct tcphdr tcp_savetcp;
152#endif /* TCPDEBUG */
153
154#ifdef FAST_IPSEC
155#include <netipsec/ipsec.h>
156#include <netipsec/ipsec6.h>
157#endif
158
159#ifdef IPSEC
160#include <netinet6/ipsec.h>
161#include <netinet6/ipsec6.h>
162#include <netkey/key.h>
163#endif /*IPSEC*/
164#include <machine/in_cksum.h>
165
166extern struct uma_zone *sack_hole_zone;
167
168/*
169 * This function is called upon receipt of new valid data (while not in header
170 * prediction mode), and it updates the ordered list of sacks.
171 */
172void
173tcp_update_sack_list(tp)
174 struct tcpcb *tp;
175{
176 /*
177 * First reported block MUST be the most recent one. Subsequent
178 * blocks SHOULD be in the order in which they arrived at the
179 * receiver. These two conditions make the implementation fully
180 * compliant with RFC 2018.
181 */
182 int i, j = 0, count = 0, lastpos = -1;
183 struct sackblk sack, firstsack, temp[MAX_SACK_BLKS];
184
185 INP_LOCK_ASSERT(tp->t_inpcb);
186 /* First clean up current list of sacks */
187 for (i = 0; i < tp->rcv_numsacks; i++) {
188 sack = tp->sackblks[i];
189 if (sack.start == 0 && sack.end == 0) {
190 count++; /* count = number of blocks to be discarded */
191 continue;
192 }
193 if (SEQ_LEQ(sack.end, tp->rcv_nxt)) {
194 tp->sackblks[i].start = tp->sackblks[i].end = 0;
195 count++;
196 } else {
197 temp[j].start = tp->sackblks[i].start;
198 temp[j++].end = tp->sackblks[i].end;
199 }
200 }
201 tp->rcv_numsacks -= count;
202 if (tp->rcv_numsacks == 0) { /* no sack blocks currently (fast path) */
203 tcp_clean_sackreport(tp);
204 if (SEQ_LT(tp->rcv_nxt, tp->rcv_laststart)) {
205 /* ==> need first sack block */
206 tp->sackblks[0].start = tp->rcv_laststart;
207 tp->sackblks[0].end = tp->rcv_lastend;
208 tp->rcv_numsacks = 1;
209 }
210 return;
211 }
212 /* Otherwise, sack blocks are already present. */
213 for (i = 0; i < tp->rcv_numsacks; i++)
214 tp->sackblks[i] = temp[i]; /* first copy back sack list */
215 if (SEQ_GEQ(tp->rcv_nxt, tp->rcv_lastend))
216 return; /* sack list remains unchanged */
217 /*
218 * From here, segment just received should be (part of) the 1st sack.
219 * Go through list, possibly coalescing sack block entries.
220 */
221 firstsack.start = tp->rcv_laststart;
222 firstsack.end = tp->rcv_lastend;
223 for (i = 0; i < tp->rcv_numsacks; i++) {
224 sack = tp->sackblks[i];
225 if (SEQ_LT(sack.end, firstsack.start) ||
226 SEQ_GT(sack.start, firstsack.end))
227 continue; /* no overlap */
228 if (sack.start == firstsack.start && sack.end == firstsack.end){
229 /*
230 * identical block; delete it here since we will
231 * move it to the front of the list.
232 */
233 tp->sackblks[i].start = tp->sackblks[i].end = 0;
234 lastpos = i; /* last posn with a zero entry */
235 continue;
236 }
237 if (SEQ_LEQ(sack.start, firstsack.start))
238 firstsack.start = sack.start; /* merge blocks */
239 if (SEQ_GEQ(sack.end, firstsack.end))
240 firstsack.end = sack.end; /* merge blocks */
241 tp->sackblks[i].start = tp->sackblks[i].end = 0;
242 lastpos = i; /* last posn with a zero entry */
243 }
244 if (lastpos != -1) { /* at least one merge */
245 for (i = 0, j = 1; i < tp->rcv_numsacks; i++) {
246 sack = tp->sackblks[i];
247 if (sack.start == 0 && sack.end == 0)
248 continue;
249 temp[j++] = sack;
250 }
251 tp->rcv_numsacks = j; /* including first blk (added later) */
252 for (i = 1; i < tp->rcv_numsacks; i++) /* now copy back */
253 tp->sackblks[i] = temp[i];
254 } else { /* no merges -- shift sacks by 1 */
255 if (tp->rcv_numsacks < MAX_SACK_BLKS)
256 tp->rcv_numsacks++;
257 for (i = tp->rcv_numsacks-1; i > 0; i--)
258 tp->sackblks[i] = tp->sackblks[i-1];
259 }
260 tp->sackblks[0] = firstsack;
261 return;
262}
263
264/*
265 * Delete all receiver-side SACK information.
266 */
267void
268tcp_clean_sackreport(tp)
269 struct tcpcb *tp;
270{
271 int i;
272
273 INP_LOCK_ASSERT(tp->t_inpcb);
274 tp->rcv_numsacks = 0;
275 for (i = 0; i < MAX_SACK_BLKS; i++)
276 tp->sackblks[i].start = tp->sackblks[i].end=0;
277}
278
279/*
280 * Process the TCP SACK option. Returns 1 if tcp_dooptions() should continue,
281 * and 0 otherwise, if the option was fine. tp->snd_holes is an ordered list
282 * of holes (oldest to newest, in terms of the sequence space).
283 */
284int
285tcp_sack_option(struct tcpcb *tp, struct tcphdr *th, u_char *cp, int optlen)
286{
287 int tmp_olen;
288 u_char *tmp_cp;
289 struct sackhole *cur, *p, *temp;
290
291 INP_LOCK_ASSERT(tp->t_inpcb);
292 if (!tp->sack_enable)
293 return (1);
294
295 /* Note: TCPOLEN_SACK must be 2*sizeof(tcp_seq) */
296 if (optlen <= 2 || (optlen - 2) % TCPOLEN_SACK != 0)
297 return (1);
298 tmp_cp = cp + 2;
299 tmp_olen = optlen - 2;
300 tcpstat.tcps_sack_rcv_blocks++;
301 if (tp->snd_numholes < 0)
302 tp->snd_numholes = 0;
303 if (tp->t_maxseg == 0)
304 panic("tcp_sack_option"); /* Should never happen */
305 while (tmp_olen > 0) {
306 struct sackblk sack;
307
308 bcopy(tmp_cp, (char *) &(sack.start), sizeof(tcp_seq));
309 sack.start = ntohl(sack.start);
310 bcopy(tmp_cp + sizeof(tcp_seq),
311 (char *) &(sack.end), sizeof(tcp_seq));
312 sack.end = ntohl(sack.end);
313 tmp_olen -= TCPOLEN_SACK;
314 tmp_cp += TCPOLEN_SACK;
315 if (SEQ_LEQ(sack.end, sack.start))
316 continue; /* bad SACK fields */
317 if (SEQ_LEQ(sack.end, tp->snd_una))
318 continue; /* old block */
319 if (SEQ_GT(th->th_ack, tp->snd_una)) {
320 if (SEQ_LT(sack.start, th->th_ack))
321 continue;
322 }
323 if (SEQ_GT(sack.end, tp->snd_max))
324 continue;
325 if (tp->snd_holes == NULL) { /* first hole */
326 tp->snd_holes = (struct sackhole *)
327 uma_zalloc(sack_hole_zone,M_NOWAIT);
328 if (tp->snd_holes == NULL) {
329 /* ENOBUFS, so ignore SACKed block for now*/
330 continue;
331 }
332 cur = tp->snd_holes;
333 cur->start = th->th_ack;
334 cur->end = sack.start;
335 cur->rxmit = cur->start;
336 cur->next = NULL;
337 tp->snd_numholes = 1;
338 tp->rcv_lastsack = sack.end;
339 continue; /* with next sack block */
340 }
341 /* Go thru list of holes: p = previous, cur = current */
342 p = cur = tp->snd_holes;
343 while (cur) {
344 if (SEQ_LEQ(sack.end, cur->start))
345 /* SACKs data before the current hole */
346 break; /* no use going through more holes */
347 if (SEQ_GEQ(sack.start, cur->end)) {
348 /* SACKs data beyond the current hole */
349 p = cur;
350 cur = cur->next;
351 continue;
352 }
353 if (SEQ_LEQ(sack.start, cur->start)) {
354 /* Data acks at least the beginning of hole */
355 if (SEQ_GEQ(sack.end, cur->end)) {
356 /* Acks entire hole, so delete hole */
357 if (p != cur) {
358 p->next = cur->next;
359 uma_zfree(sack_hole_zone, cur);
360 cur = p->next;
361 } else {
362 cur = cur->next;
363 uma_zfree(sack_hole_zone, p);
364 p = cur;
365 tp->snd_holes = p;
366 }
367 tp->snd_numholes--;
368 continue;
369 }
370 /* otherwise, move start of hole forward */
371 cur->start = sack.end;
372 cur->rxmit = SEQ_MAX(cur->rxmit, cur->start);
373 p = cur;
374 cur = cur->next;
375 continue;
376 }
377 /* move end of hole backward */
378 if (SEQ_GEQ(sack.end, cur->end)) {
379 cur->end = sack.start;
380 cur->rxmit = SEQ_MIN(cur->rxmit, cur->end);
381 p = cur;
382 cur = cur->next;
383 continue;
384 }
385 if (SEQ_LT(cur->start, sack.start) &&
386 SEQ_GT(cur->end, sack.end)) {
387 /*
388 * ACKs some data in middle of a hole; need to
389 * split current hole
390 */
391 temp = (struct sackhole *)
392 uma_zalloc(sack_hole_zone,M_NOWAIT);
393 if (temp == NULL)
394 continue; /* ENOBUFS */
395 temp->next = cur->next;
396 temp->start = sack.end;
397 temp->end = cur->end;
398 temp->rxmit = SEQ_MAX(cur->rxmit, temp->start);
399 cur->end = sack.start;
400 cur->rxmit = SEQ_MIN(cur->rxmit, cur->end);
401 cur->next = temp;
402 p = temp;
403 cur = p->next;
404 tp->snd_numholes++;
405 }
406 }
407 /* At this point, p points to the last hole on the list */
408 if (SEQ_LT(tp->rcv_lastsack, sack.start)) {
409 /*
410 * Need to append new hole at end.
411 * Last hole is p (and it's not NULL).
412 */
413 temp = (struct sackhole *)
414 uma_zalloc(sack_hole_zone,M_NOWAIT);
415 if (temp == NULL)
416 continue; /* ENOBUFS */
417 temp->start = tp->rcv_lastsack;
418 temp->end = sack.start;
419 temp->rxmit = temp->start;
420 temp->next = 0;
421 p->next = temp;
422 tp->rcv_lastsack = sack.end;
423 tp->snd_numholes++;
424 }
425 }
426 return (0);
427}
428
429/*
430 * Delete stale (i.e, cumulatively ack'd) holes. Hole is deleted only if
431 * it is completely acked; otherwise, tcp_sack_option(), called from
432 * tcp_dooptions(), will fix up the hole.
433 */
434void
435tcp_del_sackholes(tp, th)
436 struct tcpcb *tp;
437 struct tcphdr *th;
438{
439 INP_LOCK_ASSERT(tp->t_inpcb);
440 if (tp->sack_enable && tp->t_state != TCPS_LISTEN) {
441 /* max because this could be an older ack just arrived */
442 tcp_seq lastack = SEQ_GT(th->th_ack, tp->snd_una) ?
443 th->th_ack : tp->snd_una;
444 struct sackhole *cur = tp->snd_holes;
445 struct sackhole *prev;
446 while (cur)
447 if (SEQ_LEQ(cur->end, lastack)) {
448 prev = cur;
449 cur = cur->next;
450 uma_zfree(sack_hole_zone, prev);
451 tp->snd_numholes--;
452 } else if (SEQ_LT(cur->start, lastack)) {
453 cur->start = lastack;
454 if (SEQ_LT(cur->rxmit, cur->start))
455 cur->rxmit = cur->start;
456 break;
457 } else
458 break;
459 tp->snd_holes = cur;
460 }
461}
462
463void
464tcp_free_sackholes(struct tcpcb *tp)
465{
466 struct sackhole *p, *q;
467
468 INP_LOCK_ASSERT(tp->t_inpcb);
469 q = tp->snd_holes;
470 while (q != NULL) {
471 p = q;
472 q = q->next;
473 uma_zfree(sack_hole_zone, p);
474 }
475 tp->snd_holes = 0;
476}
477
478/*
479 * Checks for partial ack. If partial ack arrives, turn off retransmission
480 * timer, deflate the window, do not clear tp->t_dupacks, and return 1.
481 * If the ack advances at least to tp->snd_recover, return 0.
482 */
483void
484tcp_sack_partialack(tp, th)
485 struct tcpcb *tp;
486 struct tcphdr *th;
487{
488 INP_LOCK_ASSERT(tp->t_inpcb);
489 u_long ocwnd = tp->snd_cwnd;
490
491 callout_stop(tp->tt_rexmt);
492 tp->t_rtttime = 0;
493 /*
494 * Set snd_cwnd to one segment beyond acknowledged offset
495 * (tp->snd_una has not yet been updated when this function is called.)
496 */
497 /*
498 * Should really be
499 * min(tp->snd_cwnd, tp->t_maxseg + (th->th_ack - tp->snd_una))
500 */
501 tp->snd_cwnd = tp->t_maxseg + (th->th_ack - tp->snd_una);
502 tp->t_flags |= TF_ACKNOW;
503 (void) tcp_output(tp);
504 tp->snd_cwnd = ocwnd;
505 /*
506 * Partial window deflation. Relies on fact that tp->snd_una
507 * not updated yet.
508 */
509 tp->snd_cwnd -= (th->th_ack - tp->snd_una - tp->t_maxseg);
510}
511
512#ifdef TCP_SACK_DEBUG
513void
514tcp_print_holes(struct tcpcb *tp)
515{
516 struct sackhole *p = tp->snd_holes;
517 if (p == 0)
518 return;
519 printf("Hole report: start--end dups rxmit\n");
520 while (p) {
521 printf("%x--%x r %x\n", p->start, p->end, p->rxmit);
522 p = p->next;
523 }
524 printf("\n");
525}
526#endif /* TCP_SACK_DEBUG */
527
528/*
529 * Returns pointer to a sackhole if there are any pending retransmissions;
530 * NULL otherwise.
531 */
532struct sackhole *
533tcp_sack_output(struct tcpcb *tp)
534{
535 struct sackhole *p;
536
537 INP_LOCK_ASSERT(tp->t_inpcb);
538 if (!tp->sack_enable)
539 return (NULL);
540 p = tp->snd_holes;
541 while (p) {
542 if (SEQ_LT(p->rxmit, p->end)) {
543 if (SEQ_LT(p->rxmit, tp->snd_una)) {/* old SACK hole */
544 p = p->next;
545 continue;
546 }
547#ifdef TCP_SACK_DEBUG
548 if (p)
549 tcp_print_holes(tp);
550#endif
551 return (p);
552 }
553 p = p->next;
554 }
555 return (NULL);
556}
557
558/*
559 * After a timeout, the SACK list may be rebuilt. This SACK information
560 * should be used to avoid retransmitting SACKed data. This function
561 * traverses the SACK list to see if snd_nxt should be moved forward.
562 */
563void
564tcp_sack_adjust(struct tcpcb *tp)
565{
566 INP_LOCK_ASSERT(tp->t_inpcb);
567 struct sackhole *cur = tp->snd_holes;
568 if (cur == NULL)
569 return; /* No holes */
570 if (SEQ_GEQ(tp->snd_nxt, tp->rcv_lastsack))
571 return; /* We're already beyond any SACKed blocks */
572 /*
573 * Two cases for which we want to advance snd_nxt:
574 * i) snd_nxt lies between end of one hole and beginning of another
575 * ii) snd_nxt lies between end of last hole and rcv_lastsack
576 */
577 while (cur->next) {
578 if (SEQ_LT(tp->snd_nxt, cur->end))
579 return;
580 if (SEQ_GEQ(tp->snd_nxt, cur->next->start))
581 cur = cur->next;
582 else {
583 tp->snd_nxt = cur->next->start;
584 return;
585 }
586 }
587 if (SEQ_LT(tp->snd_nxt, cur->end))
588 return;
589 tp->snd_nxt = tp->rcv_lastsack;
590 return;
591}
592
2 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994, 1995
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * @(#)tcp_sack.c 8.12 (Berkeley) 5/24/95
30 * $FreeBSD: head/sys/netinet/tcp_sack.c 130989 2004-06-23 21:04:37Z ps $
31 */
32
33/*
34 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994
35 * The Regents of the University of California. All rights reserved.
36 *
37 * Redistribution and use in source and binary forms, with or without
38 * modification, are permitted provided that the following conditions
39 * are met:
40 * 1. Redistributions of source code must retain the above copyright
41 * notice, this list of conditions and the following disclaimer.
42 * 2. Redistributions in binary form must reproduce the above copyright
43 * notice, this list of conditions and the following disclaimer in the
44 * documentation and/or other materials provided with the distribution.
45 * 3. Neither the name of the University nor the names of its contributors
46 * may be used to endorse or promote products derived from this software
47 * without specific prior written permission.
48 *
49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59 * SUCH DAMAGE.
60 *
61 * @@(#)COPYRIGHT 1.1 (NRL) 17 January 1995
62 *
63 * NRL grants permission for redistribution and use in source and binary
64 * forms, with or without modification, of the software and documentation
65 * created at NRL provided that the following conditions are met:
66 *
67 * 1. Redistributions of source code must retain the above copyright
68 * notice, this list of conditions and the following disclaimer.
69 * 2. Redistributions in binary form must reproduce the above copyright
70 * notice, this list of conditions and the following disclaimer in the
71 * documentation and/or other materials provided with the distribution.
72 * 3. All advertising materials mentioning features or use of this software
73 * must display the following acknowledgements:
74 * This product includes software developed by the University of
75 * California, Berkeley and its contributors.
76 * This product includes software developed at the Information
77 * Technology Division, US Naval Research Laboratory.
78 * 4. Neither the name of the NRL nor the names of its contributors
79 * may be used to endorse or promote products derived from this software
80 * without specific prior written permission.
81 *
82 * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS
83 * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
84 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
85 * PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL NRL OR
86 * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
87 * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
88 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
89 * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
90 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
91 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
92 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
93 *
94 * The views and conclusions contained in the software and documentation
95 * are those of the authors and should not be interpreted as representing
96 * official policies, either expressed or implied, of the US Naval
97 * Research Laboratory (NRL).
98 */
99#include "opt_ipfw.h" /* for ipfw_fwd */
100#include "opt_inet.h"
101#include "opt_inet6.h"
102#include "opt_ipsec.h"
103#include "opt_tcpdebug.h"
104#include "opt_tcp_input.h"
105#include "opt_tcp_sack.h"
106
107#include <sys/param.h>
108#include <sys/systm.h>
109#include <sys/kernel.h>
110#include <sys/sysctl.h>
111#include <sys/malloc.h>
112#include <sys/mbuf.h>
113#include <sys/proc.h> /* for proc0 declaration */
114#include <sys/protosw.h>
115#include <sys/socket.h>
116#include <sys/socketvar.h>
117#include <sys/syslog.h>
118#include <sys/systm.h>
119
120#include <machine/cpu.h> /* before tcp_seq.h, for tcp_random18() */
121
122#include <vm/uma.h>
123
124#include <net/if.h>
125#include <net/route.h>
126
127#include <netinet/in.h>
128#include <netinet/in_systm.h>
129#include <netinet/ip.h>
130#include <netinet/ip_icmp.h> /* for ICMP_BANDLIM */
131#include <netinet/in_var.h>
132#include <netinet/icmp_var.h> /* for ICMP_BANDLIM */
133#include <netinet/in_pcb.h>
134#include <netinet/ip_var.h>
135#include <netinet/ip6.h>
136#include <netinet/icmp6.h>
137#include <netinet6/nd6.h>
138#include <netinet6/ip6_var.h>
139#include <netinet6/in6_pcb.h>
140#include <netinet/tcp.h>
141#include <netinet/tcp_fsm.h>
142#include <netinet/tcp_seq.h>
143#include <netinet/tcp_timer.h>
144#include <netinet/tcp_var.h>
145#include <netinet6/tcp6_var.h>
146#include <netinet/tcpip.h>
147#ifdef TCPDEBUG
148#include <netinet/tcp_debug.h>
149
150u_char tcp_saveipgen[40]; /* the size must be of max ip header, now IPv6 */
151struct tcphdr tcp_savetcp;
152#endif /* TCPDEBUG */
153
154#ifdef FAST_IPSEC
155#include <netipsec/ipsec.h>
156#include <netipsec/ipsec6.h>
157#endif
158
159#ifdef IPSEC
160#include <netinet6/ipsec.h>
161#include <netinet6/ipsec6.h>
162#include <netkey/key.h>
163#endif /*IPSEC*/
164#include <machine/in_cksum.h>
165
166extern struct uma_zone *sack_hole_zone;
167
168/*
169 * This function is called upon receipt of new valid data (while not in header
170 * prediction mode), and it updates the ordered list of sacks.
171 */
172void
173tcp_update_sack_list(tp)
174 struct tcpcb *tp;
175{
176 /*
177 * First reported block MUST be the most recent one. Subsequent
178 * blocks SHOULD be in the order in which they arrived at the
179 * receiver. These two conditions make the implementation fully
180 * compliant with RFC 2018.
181 */
182 int i, j = 0, count = 0, lastpos = -1;
183 struct sackblk sack, firstsack, temp[MAX_SACK_BLKS];
184
185 INP_LOCK_ASSERT(tp->t_inpcb);
186 /* First clean up current list of sacks */
187 for (i = 0; i < tp->rcv_numsacks; i++) {
188 sack = tp->sackblks[i];
189 if (sack.start == 0 && sack.end == 0) {
190 count++; /* count = number of blocks to be discarded */
191 continue;
192 }
193 if (SEQ_LEQ(sack.end, tp->rcv_nxt)) {
194 tp->sackblks[i].start = tp->sackblks[i].end = 0;
195 count++;
196 } else {
197 temp[j].start = tp->sackblks[i].start;
198 temp[j++].end = tp->sackblks[i].end;
199 }
200 }
201 tp->rcv_numsacks -= count;
202 if (tp->rcv_numsacks == 0) { /* no sack blocks currently (fast path) */
203 tcp_clean_sackreport(tp);
204 if (SEQ_LT(tp->rcv_nxt, tp->rcv_laststart)) {
205 /* ==> need first sack block */
206 tp->sackblks[0].start = tp->rcv_laststart;
207 tp->sackblks[0].end = tp->rcv_lastend;
208 tp->rcv_numsacks = 1;
209 }
210 return;
211 }
212 /* Otherwise, sack blocks are already present. */
213 for (i = 0; i < tp->rcv_numsacks; i++)
214 tp->sackblks[i] = temp[i]; /* first copy back sack list */
215 if (SEQ_GEQ(tp->rcv_nxt, tp->rcv_lastend))
216 return; /* sack list remains unchanged */
217 /*
218 * From here, segment just received should be (part of) the 1st sack.
219 * Go through list, possibly coalescing sack block entries.
220 */
221 firstsack.start = tp->rcv_laststart;
222 firstsack.end = tp->rcv_lastend;
223 for (i = 0; i < tp->rcv_numsacks; i++) {
224 sack = tp->sackblks[i];
225 if (SEQ_LT(sack.end, firstsack.start) ||
226 SEQ_GT(sack.start, firstsack.end))
227 continue; /* no overlap */
228 if (sack.start == firstsack.start && sack.end == firstsack.end){
229 /*
230 * identical block; delete it here since we will
231 * move it to the front of the list.
232 */
233 tp->sackblks[i].start = tp->sackblks[i].end = 0;
234 lastpos = i; /* last posn with a zero entry */
235 continue;
236 }
237 if (SEQ_LEQ(sack.start, firstsack.start))
238 firstsack.start = sack.start; /* merge blocks */
239 if (SEQ_GEQ(sack.end, firstsack.end))
240 firstsack.end = sack.end; /* merge blocks */
241 tp->sackblks[i].start = tp->sackblks[i].end = 0;
242 lastpos = i; /* last posn with a zero entry */
243 }
244 if (lastpos != -1) { /* at least one merge */
245 for (i = 0, j = 1; i < tp->rcv_numsacks; i++) {
246 sack = tp->sackblks[i];
247 if (sack.start == 0 && sack.end == 0)
248 continue;
249 temp[j++] = sack;
250 }
251 tp->rcv_numsacks = j; /* including first blk (added later) */
252 for (i = 1; i < tp->rcv_numsacks; i++) /* now copy back */
253 tp->sackblks[i] = temp[i];
254 } else { /* no merges -- shift sacks by 1 */
255 if (tp->rcv_numsacks < MAX_SACK_BLKS)
256 tp->rcv_numsacks++;
257 for (i = tp->rcv_numsacks-1; i > 0; i--)
258 tp->sackblks[i] = tp->sackblks[i-1];
259 }
260 tp->sackblks[0] = firstsack;
261 return;
262}
263
264/*
265 * Delete all receiver-side SACK information.
266 */
267void
268tcp_clean_sackreport(tp)
269 struct tcpcb *tp;
270{
271 int i;
272
273 INP_LOCK_ASSERT(tp->t_inpcb);
274 tp->rcv_numsacks = 0;
275 for (i = 0; i < MAX_SACK_BLKS; i++)
276 tp->sackblks[i].start = tp->sackblks[i].end=0;
277}
278
279/*
280 * Process the TCP SACK option. Returns 1 if tcp_dooptions() should continue,
281 * and 0 otherwise, if the option was fine. tp->snd_holes is an ordered list
282 * of holes (oldest to newest, in terms of the sequence space).
283 */
284int
285tcp_sack_option(struct tcpcb *tp, struct tcphdr *th, u_char *cp, int optlen)
286{
287 int tmp_olen;
288 u_char *tmp_cp;
289 struct sackhole *cur, *p, *temp;
290
291 INP_LOCK_ASSERT(tp->t_inpcb);
292 if (!tp->sack_enable)
293 return (1);
294
295 /* Note: TCPOLEN_SACK must be 2*sizeof(tcp_seq) */
296 if (optlen <= 2 || (optlen - 2) % TCPOLEN_SACK != 0)
297 return (1);
298 tmp_cp = cp + 2;
299 tmp_olen = optlen - 2;
300 tcpstat.tcps_sack_rcv_blocks++;
301 if (tp->snd_numholes < 0)
302 tp->snd_numholes = 0;
303 if (tp->t_maxseg == 0)
304 panic("tcp_sack_option"); /* Should never happen */
305 while (tmp_olen > 0) {
306 struct sackblk sack;
307
308 bcopy(tmp_cp, (char *) &(sack.start), sizeof(tcp_seq));
309 sack.start = ntohl(sack.start);
310 bcopy(tmp_cp + sizeof(tcp_seq),
311 (char *) &(sack.end), sizeof(tcp_seq));
312 sack.end = ntohl(sack.end);
313 tmp_olen -= TCPOLEN_SACK;
314 tmp_cp += TCPOLEN_SACK;
315 if (SEQ_LEQ(sack.end, sack.start))
316 continue; /* bad SACK fields */
317 if (SEQ_LEQ(sack.end, tp->snd_una))
318 continue; /* old block */
319 if (SEQ_GT(th->th_ack, tp->snd_una)) {
320 if (SEQ_LT(sack.start, th->th_ack))
321 continue;
322 }
323 if (SEQ_GT(sack.end, tp->snd_max))
324 continue;
325 if (tp->snd_holes == NULL) { /* first hole */
326 tp->snd_holes = (struct sackhole *)
327 uma_zalloc(sack_hole_zone,M_NOWAIT);
328 if (tp->snd_holes == NULL) {
329 /* ENOBUFS, so ignore SACKed block for now*/
330 continue;
331 }
332 cur = tp->snd_holes;
333 cur->start = th->th_ack;
334 cur->end = sack.start;
335 cur->rxmit = cur->start;
336 cur->next = NULL;
337 tp->snd_numholes = 1;
338 tp->rcv_lastsack = sack.end;
339 continue; /* with next sack block */
340 }
341 /* Go thru list of holes: p = previous, cur = current */
342 p = cur = tp->snd_holes;
343 while (cur) {
344 if (SEQ_LEQ(sack.end, cur->start))
345 /* SACKs data before the current hole */
346 break; /* no use going through more holes */
347 if (SEQ_GEQ(sack.start, cur->end)) {
348 /* SACKs data beyond the current hole */
349 p = cur;
350 cur = cur->next;
351 continue;
352 }
353 if (SEQ_LEQ(sack.start, cur->start)) {
354 /* Data acks at least the beginning of hole */
355 if (SEQ_GEQ(sack.end, cur->end)) {
356 /* Acks entire hole, so delete hole */
357 if (p != cur) {
358 p->next = cur->next;
359 uma_zfree(sack_hole_zone, cur);
360 cur = p->next;
361 } else {
362 cur = cur->next;
363 uma_zfree(sack_hole_zone, p);
364 p = cur;
365 tp->snd_holes = p;
366 }
367 tp->snd_numholes--;
368 continue;
369 }
370 /* otherwise, move start of hole forward */
371 cur->start = sack.end;
372 cur->rxmit = SEQ_MAX(cur->rxmit, cur->start);
373 p = cur;
374 cur = cur->next;
375 continue;
376 }
377 /* move end of hole backward */
378 if (SEQ_GEQ(sack.end, cur->end)) {
379 cur->end = sack.start;
380 cur->rxmit = SEQ_MIN(cur->rxmit, cur->end);
381 p = cur;
382 cur = cur->next;
383 continue;
384 }
385 if (SEQ_LT(cur->start, sack.start) &&
386 SEQ_GT(cur->end, sack.end)) {
387 /*
388 * ACKs some data in middle of a hole; need to
389 * split current hole
390 */
391 temp = (struct sackhole *)
392 uma_zalloc(sack_hole_zone,M_NOWAIT);
393 if (temp == NULL)
394 continue; /* ENOBUFS */
395 temp->next = cur->next;
396 temp->start = sack.end;
397 temp->end = cur->end;
398 temp->rxmit = SEQ_MAX(cur->rxmit, temp->start);
399 cur->end = sack.start;
400 cur->rxmit = SEQ_MIN(cur->rxmit, cur->end);
401 cur->next = temp;
402 p = temp;
403 cur = p->next;
404 tp->snd_numholes++;
405 }
406 }
407 /* At this point, p points to the last hole on the list */
408 if (SEQ_LT(tp->rcv_lastsack, sack.start)) {
409 /*
410 * Need to append new hole at end.
411 * Last hole is p (and it's not NULL).
412 */
413 temp = (struct sackhole *)
414 uma_zalloc(sack_hole_zone,M_NOWAIT);
415 if (temp == NULL)
416 continue; /* ENOBUFS */
417 temp->start = tp->rcv_lastsack;
418 temp->end = sack.start;
419 temp->rxmit = temp->start;
420 temp->next = 0;
421 p->next = temp;
422 tp->rcv_lastsack = sack.end;
423 tp->snd_numholes++;
424 }
425 }
426 return (0);
427}
428
429/*
430 * Delete stale (i.e, cumulatively ack'd) holes. Hole is deleted only if
431 * it is completely acked; otherwise, tcp_sack_option(), called from
432 * tcp_dooptions(), will fix up the hole.
433 */
434void
435tcp_del_sackholes(tp, th)
436 struct tcpcb *tp;
437 struct tcphdr *th;
438{
439 INP_LOCK_ASSERT(tp->t_inpcb);
440 if (tp->sack_enable && tp->t_state != TCPS_LISTEN) {
441 /* max because this could be an older ack just arrived */
442 tcp_seq lastack = SEQ_GT(th->th_ack, tp->snd_una) ?
443 th->th_ack : tp->snd_una;
444 struct sackhole *cur = tp->snd_holes;
445 struct sackhole *prev;
446 while (cur)
447 if (SEQ_LEQ(cur->end, lastack)) {
448 prev = cur;
449 cur = cur->next;
450 uma_zfree(sack_hole_zone, prev);
451 tp->snd_numholes--;
452 } else if (SEQ_LT(cur->start, lastack)) {
453 cur->start = lastack;
454 if (SEQ_LT(cur->rxmit, cur->start))
455 cur->rxmit = cur->start;
456 break;
457 } else
458 break;
459 tp->snd_holes = cur;
460 }
461}
462
463void
464tcp_free_sackholes(struct tcpcb *tp)
465{
466 struct sackhole *p, *q;
467
468 INP_LOCK_ASSERT(tp->t_inpcb);
469 q = tp->snd_holes;
470 while (q != NULL) {
471 p = q;
472 q = q->next;
473 uma_zfree(sack_hole_zone, p);
474 }
475 tp->snd_holes = 0;
476}
477
478/*
479 * Checks for partial ack. If partial ack arrives, turn off retransmission
480 * timer, deflate the window, do not clear tp->t_dupacks, and return 1.
481 * If the ack advances at least to tp->snd_recover, return 0.
482 */
483void
484tcp_sack_partialack(tp, th)
485 struct tcpcb *tp;
486 struct tcphdr *th;
487{
488 INP_LOCK_ASSERT(tp->t_inpcb);
489 u_long ocwnd = tp->snd_cwnd;
490
491 callout_stop(tp->tt_rexmt);
492 tp->t_rtttime = 0;
493 /*
494 * Set snd_cwnd to one segment beyond acknowledged offset
495 * (tp->snd_una has not yet been updated when this function is called.)
496 */
497 /*
498 * Should really be
499 * min(tp->snd_cwnd, tp->t_maxseg + (th->th_ack - tp->snd_una))
500 */
501 tp->snd_cwnd = tp->t_maxseg + (th->th_ack - tp->snd_una);
502 tp->t_flags |= TF_ACKNOW;
503 (void) tcp_output(tp);
504 tp->snd_cwnd = ocwnd;
505 /*
506 * Partial window deflation. Relies on fact that tp->snd_una
507 * not updated yet.
508 */
509 tp->snd_cwnd -= (th->th_ack - tp->snd_una - tp->t_maxseg);
510}
511
512#ifdef TCP_SACK_DEBUG
513void
514tcp_print_holes(struct tcpcb *tp)
515{
516 struct sackhole *p = tp->snd_holes;
517 if (p == 0)
518 return;
519 printf("Hole report: start--end dups rxmit\n");
520 while (p) {
521 printf("%x--%x r %x\n", p->start, p->end, p->rxmit);
522 p = p->next;
523 }
524 printf("\n");
525}
526#endif /* TCP_SACK_DEBUG */
527
528/*
529 * Returns pointer to a sackhole if there are any pending retransmissions;
530 * NULL otherwise.
531 */
532struct sackhole *
533tcp_sack_output(struct tcpcb *tp)
534{
535 struct sackhole *p;
536
537 INP_LOCK_ASSERT(tp->t_inpcb);
538 if (!tp->sack_enable)
539 return (NULL);
540 p = tp->snd_holes;
541 while (p) {
542 if (SEQ_LT(p->rxmit, p->end)) {
543 if (SEQ_LT(p->rxmit, tp->snd_una)) {/* old SACK hole */
544 p = p->next;
545 continue;
546 }
547#ifdef TCP_SACK_DEBUG
548 if (p)
549 tcp_print_holes(tp);
550#endif
551 return (p);
552 }
553 p = p->next;
554 }
555 return (NULL);
556}
557
558/*
559 * After a timeout, the SACK list may be rebuilt. This SACK information
560 * should be used to avoid retransmitting SACKed data. This function
561 * traverses the SACK list to see if snd_nxt should be moved forward.
562 */
563void
564tcp_sack_adjust(struct tcpcb *tp)
565{
566 INP_LOCK_ASSERT(tp->t_inpcb);
567 struct sackhole *cur = tp->snd_holes;
568 if (cur == NULL)
569 return; /* No holes */
570 if (SEQ_GEQ(tp->snd_nxt, tp->rcv_lastsack))
571 return; /* We're already beyond any SACKed blocks */
572 /*
573 * Two cases for which we want to advance snd_nxt:
574 * i) snd_nxt lies between end of one hole and beginning of another
575 * ii) snd_nxt lies between end of last hole and rcv_lastsack
576 */
577 while (cur->next) {
578 if (SEQ_LT(tp->snd_nxt, cur->end))
579 return;
580 if (SEQ_GEQ(tp->snd_nxt, cur->next->start))
581 cur = cur->next;
582 else {
583 tp->snd_nxt = cur->next->start;
584 return;
585 }
586 }
587 if (SEQ_LT(tp->snd_nxt, cur->end))
588 return;
589 tp->snd_nxt = tp->rcv_lastsack;
590 return;
591}
592