1 /*- 2 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994, 1995 3 * The Regents of the University of California. 4 * All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 4. Neither the name of the University nor the names of its contributors 15 * may be used to endorse or promote products derived from this software 16 * without specific prior written permission. 17 * 18 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 21 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 22 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 28 * SUCH DAMAGE. 29 * 30 * @(#)tcp_sack.c 8.12 (Berkeley) 5/24/95 31 */ 32 33 /*- 34 * @@(#)COPYRIGHT 1.1 (NRL) 17 January 1995 35 * 36 * NRL grants permission for redistribution and use in source and binary 37 * forms, with or without modification, of the software and documentation 38 * created at NRL provided that the following conditions are met: 39 * 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. All advertising materials mentioning features or use of this software 46 * must display the following acknowledgements: 47 * This product includes software developed by the University of 48 * California, Berkeley and its contributors. 49 * This product includes software developed at the Information 50 * Technology Division, US Naval Research Laboratory. 51 * 4. Neither the name of the NRL nor the names of its contributors 52 * may be used to endorse or promote products derived from this software 53 * without specific prior written permission. 54 * 55 * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS 56 * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 57 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A 58 * PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL NRL OR 59 * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, 60 * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 61 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR 62 * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF 63 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING 64 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 65 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 66 * 67 * The views and conclusions contained in the software and documentation 68 * are those of the authors and should not be interpreted as representing 69 * official policies, either expressed or implied, of the US Naval 70 * Research Laboratory (NRL). 71 */ 72 73 #include <sys/cdefs.h> 74 __FBSDID("$FreeBSD$"); 75 76 #include "opt_inet.h" 77 #include "opt_inet6.h" 78 #include "opt_tcpdebug.h" 79 80 #include <sys/param.h> 81 #include <sys/systm.h> 82 #include <sys/kernel.h> 83 #include <sys/sysctl.h> 84 #include <sys/malloc.h> 85 #include <sys/mbuf.h> 86 #include <sys/proc.h> /* for proc0 declaration */ 87 #include <sys/protosw.h> 88 #include <sys/socket.h> 89 #include <sys/socketvar.h> 90 #include <sys/syslog.h> 91 #include <sys/systm.h> 92 93 #include <machine/cpu.h> /* before tcp_seq.h, for tcp_random18() */ 94 95 #include <vm/uma.h> 96 97 #include <net/if.h> 98 #include <net/if_var.h> 99 #include <net/route.h> 100 #include <net/vnet.h> 101 102 #include <netinet/in.h> 103 #include <netinet/in_systm.h> 104 #include <netinet/ip.h> 105 #include <netinet/in_var.h> 106 #include <netinet/in_pcb.h> 107 #include <netinet/ip_var.h> 108 #include <netinet/ip6.h> 109 #include <netinet/icmp6.h> 110 #include <netinet6/nd6.h> 111 #include <netinet6/ip6_var.h> 112 #include <netinet6/in6_pcb.h> 113 #include <netinet/tcp.h> 114 #include <netinet/tcp_fsm.h> 115 #include <netinet/tcp_seq.h> 116 #include <netinet/tcp_timer.h> 117 #include <netinet/tcp_var.h> 118 #include <netinet6/tcp6_var.h> 119 #include <netinet/tcpip.h> 120 #ifdef TCPDEBUG 121 #include <netinet/tcp_debug.h> 122 #endif /* TCPDEBUG */ 123 124 #include <machine/in_cksum.h> 125 126 VNET_DECLARE(struct uma_zone *, sack_hole_zone); 127 #define V_sack_hole_zone VNET(sack_hole_zone) 128 129 SYSCTL_NODE(_net_inet_tcp, OID_AUTO, sack, CTLFLAG_RW, 0, "TCP SACK"); 130 VNET_DEFINE(int, tcp_do_sack) = 1; 131 #define V_tcp_do_sack VNET(tcp_do_sack) 132 SYSCTL_INT(_net_inet_tcp_sack, OID_AUTO, enable, CTLFLAG_VNET | CTLFLAG_RW, 133 &VNET_NAME(tcp_do_sack), 0, "Enable/Disable TCP SACK support"); 134 135 VNET_DEFINE(int, tcp_sack_maxholes) = 128; 136 #define V_tcp_sack_maxholes VNET(tcp_sack_maxholes) 137 SYSCTL_INT(_net_inet_tcp_sack, OID_AUTO, maxholes, CTLFLAG_VNET | CTLFLAG_RW, 138 &VNET_NAME(tcp_sack_maxholes), 0, 139 "Maximum number of TCP SACK holes allowed per connection"); 140 141 VNET_DEFINE(int, tcp_sack_globalmaxholes) = 65536; 142 #define V_tcp_sack_globalmaxholes VNET(tcp_sack_globalmaxholes) 143 SYSCTL_INT(_net_inet_tcp_sack, OID_AUTO, globalmaxholes, CTLFLAG_VNET | CTLFLAG_RW, 144 &VNET_NAME(tcp_sack_globalmaxholes), 0, 145 "Global maximum number of TCP SACK holes"); 146 147 VNET_DEFINE(int, tcp_sack_globalholes) = 0; 148 #define V_tcp_sack_globalholes VNET(tcp_sack_globalholes) 149 SYSCTL_INT(_net_inet_tcp_sack, OID_AUTO, globalholes, CTLFLAG_VNET | CTLFLAG_RD, 150 &VNET_NAME(tcp_sack_globalholes), 0, 151 "Global number of TCP SACK holes currently allocated"); 152 153 /* 154 * This function is called upon receipt of new valid data (while not in 155 * header prediction mode), and it updates the ordered list of sacks. 156 */ 157 void 158 tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_start, tcp_seq rcv_end) 159 { 160 /* 161 * First reported block MUST be the most recent one. Subsequent 162 * blocks SHOULD be in the order in which they arrived at the 163 * receiver. These two conditions make the implementation fully 164 * compliant with RFC 2018. 165 */ 166 struct sackblk head_blk, saved_blks[MAX_SACK_BLKS]; 167 int num_head, num_saved, i; 168 169 INP_WLOCK_ASSERT(tp->t_inpcb); 170 171 /* Check arguments. */ 172 KASSERT(SEQ_LT(rcv_start, rcv_end), ("rcv_start < rcv_end")); 173 174 /* SACK block for the received segment. */ 175 head_blk.start = rcv_start; 176 head_blk.end = rcv_end; 177 178 /* 179 * Merge updated SACK blocks into head_blk, and save unchanged SACK 180 * blocks into saved_blks[]. num_saved will have the number of the 181 * saved SACK blocks. 182 */ 183 num_saved = 0; 184 for (i = 0; i < tp->rcv_numsacks; i++) { 185 tcp_seq start = tp->sackblks[i].start; 186 tcp_seq end = tp->sackblks[i].end; 187 if (SEQ_GEQ(start, end) || SEQ_LEQ(start, tp->rcv_nxt)) { 188 /* 189 * Discard this SACK block. 190 */ 191 } else if (SEQ_LEQ(head_blk.start, end) && 192 SEQ_GEQ(head_blk.end, start)) { 193 /* 194 * Merge this SACK block into head_blk. This SACK 195 * block itself will be discarded. 196 */ 197 if (SEQ_GT(head_blk.start, start)) 198 head_blk.start = start; 199 if (SEQ_LT(head_blk.end, end)) 200 head_blk.end = end; 201 } else { 202 /* 203 * Save this SACK block. 204 */ 205 saved_blks[num_saved].start = start; 206 saved_blks[num_saved].end = end; 207 num_saved++; 208 } 209 } 210 211 /* 212 * Update SACK list in tp->sackblks[]. 213 */ 214 num_head = 0; 215 if (SEQ_GT(head_blk.start, tp->rcv_nxt)) { 216 /* 217 * The received data segment is an out-of-order segment. Put 218 * head_blk at the top of SACK list. 219 */ 220 tp->sackblks[0] = head_blk; 221 num_head = 1; 222 /* 223 * If the number of saved SACK blocks exceeds its limit, 224 * discard the last SACK block. 225 */ 226 if (num_saved >= MAX_SACK_BLKS) 227 num_saved--; 228 } 229 if (num_saved > 0) { 230 /* 231 * Copy the saved SACK blocks back. 232 */ 233 bcopy(saved_blks, &tp->sackblks[num_head], 234 sizeof(struct sackblk) * num_saved); 235 } 236 237 /* Save the number of SACK blocks. */ 238 tp->rcv_numsacks = num_head + num_saved; 239 } 240 241 /* 242 * Delete all receiver-side SACK information. 243 */ 244 void 245 tcp_clean_sackreport(struct tcpcb *tp) 246 { 247 int i; 248 249 INP_WLOCK_ASSERT(tp->t_inpcb); 250 tp->rcv_numsacks = 0; 251 for (i = 0; i < MAX_SACK_BLKS; i++) 252 tp->sackblks[i].start = tp->sackblks[i].end=0; 253 } 254 255 /* 256 * Allocate struct sackhole. 257 */ 258 static struct sackhole * 259 tcp_sackhole_alloc(struct tcpcb *tp, tcp_seq start, tcp_seq end) 260 { 261 struct sackhole *hole; 262 263 if (tp->snd_numholes >= V_tcp_sack_maxholes || 264 V_tcp_sack_globalholes >= V_tcp_sack_globalmaxholes) { 265 TCPSTAT_INC(tcps_sack_sboverflow); 266 return NULL; 267 } 268 269 hole = (struct sackhole *)uma_zalloc(V_sack_hole_zone, M_NOWAIT); 270 if (hole == NULL) 271 return NULL; 272 273 hole->start = start; 274 hole->end = end; 275 hole->rxmit = start; 276 277 tp->snd_numholes++; 278 atomic_add_int(&V_tcp_sack_globalholes, 1); 279 280 return hole; 281 } 282 283 /* 284 * Free struct sackhole. 285 */ 286 static void 287 tcp_sackhole_free(struct tcpcb *tp, struct sackhole *hole) 288 { 289 290 uma_zfree(V_sack_hole_zone, hole); 291 292 tp->snd_numholes--; 293 atomic_subtract_int(&V_tcp_sack_globalholes, 1); 294 295 KASSERT(tp->snd_numholes >= 0, ("tp->snd_numholes >= 0")); 296 KASSERT(V_tcp_sack_globalholes >= 0, ("tcp_sack_globalholes >= 0")); 297 } 298 299 /* 300 * Insert new SACK hole into scoreboard. 301 */ 302 static struct sackhole * 303 tcp_sackhole_insert(struct tcpcb *tp, tcp_seq start, tcp_seq end, 304 struct sackhole *after) 305 { 306 struct sackhole *hole; 307 308 /* Allocate a new SACK hole. */ 309 hole = tcp_sackhole_alloc(tp, start, end); 310 if (hole == NULL) 311 return NULL; 312 313 /* Insert the new SACK hole into scoreboard. */ 314 if (after != NULL) 315 TAILQ_INSERT_AFTER(&tp->snd_holes, after, hole, scblink); 316 else 317 TAILQ_INSERT_TAIL(&tp->snd_holes, hole, scblink); 318 319 /* Update SACK hint. */ 320 if (tp->sackhint.nexthole == NULL) 321 tp->sackhint.nexthole = hole; 322 323 return hole; 324 } 325 326 /* 327 * Remove SACK hole from scoreboard. 328 */ 329 static void 330 tcp_sackhole_remove(struct tcpcb *tp, struct sackhole *hole) 331 { 332 333 /* Update SACK hint. */ 334 if (tp->sackhint.nexthole == hole) 335 tp->sackhint.nexthole = TAILQ_NEXT(hole, scblink); 336 337 /* Remove this SACK hole. */ 338 TAILQ_REMOVE(&tp->snd_holes, hole, scblink); 339 340 /* Free this SACK hole. */ 341 tcp_sackhole_free(tp, hole); 342 } 343 344 /* 345 * Process cumulative ACK and the TCP SACK option to update the scoreboard. 346 * tp->snd_holes is an ordered list of holes (oldest to newest, in terms of 347 * the sequence space). 348 * Returns 1 if incoming ACK has previously unknown SACK information, 349 * 0 otherwise. Note: We treat (snd_una, th_ack) as a sack block so any changes 350 * to that (i.e. left edge moving) would also be considered a change in SACK 351 * information which is slightly different than rfc6675. 352 */ 353 int 354 tcp_sack_doack(struct tcpcb *tp, struct tcpopt *to, tcp_seq th_ack) 355 { 356 struct sackhole *cur, *temp; 357 struct sackblk sack, sack_blocks[TCP_MAX_SACK + 1], *sblkp; 358 int i, j, num_sack_blks, sack_changed; 359 360 INP_WLOCK_ASSERT(tp->t_inpcb); 361 362 num_sack_blks = 0; 363 sack_changed = 0; 364 /* 365 * If SND.UNA will be advanced by SEG.ACK, and if SACK holes exist, 366 * treat [SND.UNA, SEG.ACK) as if it is a SACK block. 367 */ 368 if (SEQ_LT(tp->snd_una, th_ack) && !TAILQ_EMPTY(&tp->snd_holes)) { 369 sack_blocks[num_sack_blks].start = tp->snd_una; 370 sack_blocks[num_sack_blks++].end = th_ack; 371 } 372 /* 373 * Append received valid SACK blocks to sack_blocks[], but only if we 374 * received new blocks from the other side. 375 */ 376 if (to->to_flags & TOF_SACK) { 377 tp->sackhint.sacked_bytes = 0; /* reset */ 378 for (i = 0; i < to->to_nsacks; i++) { 379 bcopy((to->to_sacks + i * TCPOLEN_SACK), 380 &sack, sizeof(sack)); 381 sack.start = ntohl(sack.start); 382 sack.end = ntohl(sack.end); 383 if (SEQ_GT(sack.end, sack.start) && 384 SEQ_GT(sack.start, tp->snd_una) && 385 SEQ_GT(sack.start, th_ack) && 386 SEQ_LT(sack.start, tp->snd_max) && 387 SEQ_GT(sack.end, tp->snd_una) && 388 SEQ_LEQ(sack.end, tp->snd_max)) { 389 sack_blocks[num_sack_blks++] = sack; 390 tp->sackhint.sacked_bytes += 391 (sack.end-sack.start); 392 } 393 } 394 } 395 /* 396 * Return if SND.UNA is not advanced and no valid SACK block is 397 * received. 398 */ 399 if (num_sack_blks == 0) 400 return (sack_changed); 401 402 /* 403 * Sort the SACK blocks so we can update the scoreboard with just one 404 * pass. The overhead of sorting up to 4+1 elements is less than 405 * making up to 4+1 passes over the scoreboard. 406 */ 407 for (i = 0; i < num_sack_blks; i++) { 408 for (j = i + 1; j < num_sack_blks; j++) { 409 if (SEQ_GT(sack_blocks[i].end, sack_blocks[j].end)) { 410 sack = sack_blocks[i]; 411 sack_blocks[i] = sack_blocks[j]; 412 sack_blocks[j] = sack; 413 } 414 } 415 } 416 if (TAILQ_EMPTY(&tp->snd_holes)) 417 /* 418 * Empty scoreboard. Need to initialize snd_fack (it may be 419 * uninitialized or have a bogus value). Scoreboard holes 420 * (from the sack blocks received) are created later below 421 * (in the logic that adds holes to the tail of the 422 * scoreboard). 423 */ 424 tp->snd_fack = SEQ_MAX(tp->snd_una, th_ack); 425 /* 426 * In the while-loop below, incoming SACK blocks (sack_blocks[]) and 427 * SACK holes (snd_holes) are traversed from their tails with just 428 * one pass in order to reduce the number of compares especially when 429 * the bandwidth-delay product is large. 430 * 431 * Note: Typically, in the first RTT of SACK recovery, the highest 432 * three or four SACK blocks with the same ack number are received. 433 * In the second RTT, if retransmitted data segments are not lost, 434 * the highest three or four SACK blocks with ack number advancing 435 * are received. 436 */ 437 sblkp = &sack_blocks[num_sack_blks - 1]; /* Last SACK block */ 438 tp->sackhint.last_sack_ack = sblkp->end; 439 if (SEQ_LT(tp->snd_fack, sblkp->start)) { 440 /* 441 * The highest SACK block is beyond fack. Append new SACK 442 * hole at the tail. If the second or later highest SACK 443 * blocks are also beyond the current fack, they will be 444 * inserted by way of hole splitting in the while-loop below. 445 */ 446 temp = tcp_sackhole_insert(tp, tp->snd_fack,sblkp->start,NULL); 447 if (temp != NULL) { 448 tp->snd_fack = sblkp->end; 449 /* Go to the previous sack block. */ 450 sblkp--; 451 sack_changed = 1; 452 } else { 453 /* 454 * We failed to add a new hole based on the current 455 * sack block. Skip over all the sack blocks that 456 * fall completely to the right of snd_fack and 457 * proceed to trim the scoreboard based on the 458 * remaining sack blocks. This also trims the 459 * scoreboard for th_ack (which is sack_blocks[0]). 460 */ 461 while (sblkp >= sack_blocks && 462 SEQ_LT(tp->snd_fack, sblkp->start)) 463 sblkp--; 464 if (sblkp >= sack_blocks && 465 SEQ_LT(tp->snd_fack, sblkp->end)) 466 tp->snd_fack = sblkp->end; 467 } 468 } else if (SEQ_LT(tp->snd_fack, sblkp->end)) { 469 /* fack is advanced. */ 470 tp->snd_fack = sblkp->end; 471 sack_changed = 1; 472 } 473 /* We must have at least one SACK hole in scoreboard. */ 474 KASSERT(!TAILQ_EMPTY(&tp->snd_holes), 475 ("SACK scoreboard must not be empty")); 476 cur = TAILQ_LAST(&tp->snd_holes, sackhole_head); /* Last SACK hole. */ 477 /* 478 * Since the incoming sack blocks are sorted, we can process them 479 * making one sweep of the scoreboard. 480 */ 481 while (sblkp >= sack_blocks && cur != NULL) { 482 if (SEQ_GEQ(sblkp->start, cur->end)) { 483 /* 484 * SACKs data beyond the current hole. Go to the 485 * previous sack block. 486 */ 487 sblkp--; 488 continue; 489 } 490 if (SEQ_LEQ(sblkp->end, cur->start)) { 491 /* 492 * SACKs data before the current hole. Go to the 493 * previous hole. 494 */ 495 cur = TAILQ_PREV(cur, sackhole_head, scblink); 496 continue; 497 } 498 tp->sackhint.sack_bytes_rexmit -= (cur->rxmit - cur->start); 499 KASSERT(tp->sackhint.sack_bytes_rexmit >= 0, 500 ("sackhint bytes rtx >= 0")); 501 sack_changed = 1; 502 if (SEQ_LEQ(sblkp->start, cur->start)) { 503 /* Data acks at least the beginning of hole. */ 504 if (SEQ_GEQ(sblkp->end, cur->end)) { 505 /* Acks entire hole, so delete hole. */ 506 temp = cur; 507 cur = TAILQ_PREV(cur, sackhole_head, scblink); 508 tcp_sackhole_remove(tp, temp); 509 /* 510 * The sack block may ack all or part of the 511 * next hole too, so continue onto the next 512 * hole. 513 */ 514 continue; 515 } else { 516 /* Move start of hole forward. */ 517 cur->start = sblkp->end; 518 cur->rxmit = SEQ_MAX(cur->rxmit, cur->start); 519 } 520 } else { 521 /* Data acks at least the end of hole. */ 522 if (SEQ_GEQ(sblkp->end, cur->end)) { 523 /* Move end of hole backward. */ 524 cur->end = sblkp->start; 525 cur->rxmit = SEQ_MIN(cur->rxmit, cur->end); 526 } else { 527 /* 528 * ACKs some data in middle of a hole; need 529 * to split current hole 530 */ 531 temp = tcp_sackhole_insert(tp, sblkp->end, 532 cur->end, cur); 533 if (temp != NULL) { 534 if (SEQ_GT(cur->rxmit, temp->rxmit)) { 535 temp->rxmit = cur->rxmit; 536 tp->sackhint.sack_bytes_rexmit 537 += (temp->rxmit 538 - temp->start); 539 } 540 cur->end = sblkp->start; 541 cur->rxmit = SEQ_MIN(cur->rxmit, 542 cur->end); 543 } 544 } 545 } 546 tp->sackhint.sack_bytes_rexmit += (cur->rxmit - cur->start); 547 /* 548 * Testing sblkp->start against cur->start tells us whether 549 * we're done with the sack block or the sack hole. 550 * Accordingly, we advance one or the other. 551 */ 552 if (SEQ_LEQ(sblkp->start, cur->start)) 553 cur = TAILQ_PREV(cur, sackhole_head, scblink); 554 else 555 sblkp--; 556 } 557 return (sack_changed); 558 } 559 560 /* 561 * Free all SACK holes to clear the scoreboard. 562 */ 563 void 564 tcp_free_sackholes(struct tcpcb *tp) 565 { 566 struct sackhole *q; 567 568 INP_WLOCK_ASSERT(tp->t_inpcb); 569 while ((q = TAILQ_FIRST(&tp->snd_holes)) != NULL) 570 tcp_sackhole_remove(tp, q); 571 tp->sackhint.sack_bytes_rexmit = 0; 572 573 KASSERT(tp->snd_numholes == 0, ("tp->snd_numholes == 0")); 574 KASSERT(tp->sackhint.nexthole == NULL, 575 ("tp->sackhint.nexthole == NULL")); 576 } 577 578 /* 579 * Partial ack handling within a sack recovery episode. Keeping this very 580 * simple for now. When a partial ack is received, force snd_cwnd to a value 581 * that will allow the sender to transmit no more than 2 segments. If 582 * necessary, a better scheme can be adopted at a later point, but for now, 583 * the goal is to prevent the sender from bursting a large amount of data in 584 * the midst of sack recovery. 585 */ 586 void 587 tcp_sack_partialack(struct tcpcb *tp, struct tcphdr *th) 588 { 589 int num_segs = 1; 590 591 INP_WLOCK_ASSERT(tp->t_inpcb); 592 tcp_timer_activate(tp, TT_REXMT, 0); 593 tp->t_rtttime = 0; 594 /* Send one or 2 segments based on how much new data was acked. */ 595 if ((BYTES_THIS_ACK(tp, th) / tp->t_maxseg) >= 2) 596 num_segs = 2; 597 tp->snd_cwnd = (tp->sackhint.sack_bytes_rexmit + 598 (tp->snd_nxt - tp->sack_newdata) + num_segs * tp->t_maxseg); 599 if (tp->snd_cwnd > tp->snd_ssthresh) 600 tp->snd_cwnd = tp->snd_ssthresh; 601 tp->t_flags |= TF_ACKNOW; 602 (void) tp->t_fb->tfb_tcp_output(tp); 603 } 604 605 #if 0 606 /* 607 * Debug version of tcp_sack_output() that walks the scoreboard. Used for 608 * now to sanity check the hint. 609 */ 610 static struct sackhole * 611 tcp_sack_output_debug(struct tcpcb *tp, int *sack_bytes_rexmt) 612 { 613 struct sackhole *p; 614 615 INP_WLOCK_ASSERT(tp->t_inpcb); 616 *sack_bytes_rexmt = 0; 617 TAILQ_FOREACH(p, &tp->snd_holes, scblink) { 618 if (SEQ_LT(p->rxmit, p->end)) { 619 if (SEQ_LT(p->rxmit, tp->snd_una)) {/* old SACK hole */ 620 continue; 621 } 622 *sack_bytes_rexmt += (p->rxmit - p->start); 623 break; 624 } 625 *sack_bytes_rexmt += (p->rxmit - p->start); 626 } 627 return (p); 628 } 629 #endif 630 631 /* 632 * Returns the next hole to retransmit and the number of retransmitted bytes 633 * from the scoreboard. We store both the next hole and the number of 634 * retransmitted bytes as hints (and recompute these on the fly upon SACK/ACK 635 * reception). This avoids scoreboard traversals completely. 636 * 637 * The loop here will traverse *at most* one link. Here's the argument. For 638 * the loop to traverse more than 1 link before finding the next hole to 639 * retransmit, we would need to have at least 1 node following the current 640 * hint with (rxmit == end). But, for all holes following the current hint, 641 * (start == rxmit), since we have not yet retransmitted from them. 642 * Therefore, in order to traverse more 1 link in the loop below, we need to 643 * have at least one node following the current hint with (start == rxmit == 644 * end). But that can't happen, (start == end) means that all the data in 645 * that hole has been sacked, in which case, the hole would have been removed 646 * from the scoreboard. 647 */ 648 struct sackhole * 649 tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt) 650 { 651 struct sackhole *hole = NULL; 652 653 INP_WLOCK_ASSERT(tp->t_inpcb); 654 *sack_bytes_rexmt = tp->sackhint.sack_bytes_rexmit; 655 hole = tp->sackhint.nexthole; 656 if (hole == NULL || SEQ_LT(hole->rxmit, hole->end)) 657 goto out; 658 while ((hole = TAILQ_NEXT(hole, scblink)) != NULL) { 659 if (SEQ_LT(hole->rxmit, hole->end)) { 660 tp->sackhint.nexthole = hole; 661 break; 662 } 663 } 664 out: 665 return (hole); 666 } 667 668 /* 669 * After a timeout, the SACK list may be rebuilt. This SACK information 670 * should be used to avoid retransmitting SACKed data. This function 671 * traverses the SACK list to see if snd_nxt should be moved forward. 672 */ 673 void 674 tcp_sack_adjust(struct tcpcb *tp) 675 { 676 struct sackhole *p, *cur = TAILQ_FIRST(&tp->snd_holes); 677 678 INP_WLOCK_ASSERT(tp->t_inpcb); 679 if (cur == NULL) 680 return; /* No holes */ 681 if (SEQ_GEQ(tp->snd_nxt, tp->snd_fack)) 682 return; /* We're already beyond any SACKed blocks */ 683 /*- 684 * Two cases for which we want to advance snd_nxt: 685 * i) snd_nxt lies between end of one hole and beginning of another 686 * ii) snd_nxt lies between end of last hole and snd_fack 687 */ 688 while ((p = TAILQ_NEXT(cur, scblink)) != NULL) { 689 if (SEQ_LT(tp->snd_nxt, cur->end)) 690 return; 691 if (SEQ_GEQ(tp->snd_nxt, p->start)) 692 cur = p; 693 else { 694 tp->snd_nxt = p->start; 695 return; 696 } 697 } 698 if (SEQ_LT(tp->snd_nxt, cur->end)) 699 return; 700 tp->snd_nxt = tp->snd_fack; 701 } 702