1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994, 1995 5 * The Regents of the University of California. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the University nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 * 32 * @(#)tcp_sack.c 8.12 (Berkeley) 5/24/95 33 */ 34 35 /*- 36 * @@(#)COPYRIGHT 1.1 (NRL) 17 January 1995 37 * 38 * NRL grants permission for redistribution and use in source and binary 39 * forms, with or without modification, of the software and documentation 40 * created at NRL provided that the following conditions are met: 41 * 42 * 1. Redistributions of source code must retain the above copyright 43 * notice, this list of conditions and the following disclaimer. 44 * 2. Redistributions in binary form must reproduce the above copyright 45 * notice, this list of conditions and the following disclaimer in the 46 * documentation and/or other materials provided with the distribution. 47 * 3. All advertising materials mentioning features or use of this software 48 * must display the following acknowledgements: 49 * This product includes software developed by the University of 50 * California, Berkeley and its contributors. 51 * This product includes software developed at the Information 52 * Technology Division, US Naval Research Laboratory. 53 * 4. Neither the name of the NRL nor the names of its contributors 54 * may be used to endorse or promote products derived from this software 55 * without specific prior written permission. 56 * 57 * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS 58 * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 59 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A 60 * PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL NRL OR 61 * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, 62 * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 63 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR 64 * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF 65 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING 66 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 67 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 68 * 69 * The views and conclusions contained in the software and documentation 70 * are those of the authors and should not be interpreted as representing 71 * official policies, either expressed or implied, of the US Naval 72 * Research Laboratory (NRL). 73 */ 74 75 #include <sys/cdefs.h> 76 __FBSDID("$FreeBSD$"); 77 78 #include "opt_inet.h" 79 #include "opt_inet6.h" 80 #include "opt_tcpdebug.h" 81 82 #include <sys/param.h> 83 #include <sys/systm.h> 84 #include <sys/kernel.h> 85 #include <sys/sysctl.h> 86 #include <sys/malloc.h> 87 #include <sys/mbuf.h> 88 #include <sys/proc.h> /* for proc0 declaration */ 89 #include <sys/protosw.h> 90 #include <sys/socket.h> 91 #include <sys/socketvar.h> 92 #include <sys/syslog.h> 93 #include <sys/systm.h> 94 95 #include <machine/cpu.h> /* before tcp_seq.h, for tcp_random18() */ 96 97 #include <vm/uma.h> 98 99 #include <net/if.h> 100 #include <net/if_var.h> 101 #include <net/route.h> 102 #include <net/vnet.h> 103 104 #include <netinet/in.h> 105 #include <netinet/in_systm.h> 106 #include <netinet/ip.h> 107 #include <netinet/in_var.h> 108 #include <netinet/in_pcb.h> 109 #include <netinet/ip_var.h> 110 #include <netinet/ip6.h> 111 #include <netinet/icmp6.h> 112 #include <netinet6/nd6.h> 113 #include <netinet6/ip6_var.h> 114 #include <netinet6/in6_pcb.h> 115 #include <netinet/tcp.h> 116 #include <netinet/tcp_fsm.h> 117 #include <netinet/tcp_seq.h> 118 #include <netinet/tcp_timer.h> 119 #include <netinet/tcp_var.h> 120 #include <netinet6/tcp6_var.h> 121 #include <netinet/tcpip.h> 122 #ifdef TCPDEBUG 123 #include <netinet/tcp_debug.h> 124 #endif /* TCPDEBUG */ 125 126 #include <machine/in_cksum.h> 127 128 VNET_DECLARE(struct uma_zone *, sack_hole_zone); 129 #define V_sack_hole_zone VNET(sack_hole_zone) 130 131 SYSCTL_NODE(_net_inet_tcp, OID_AUTO, sack, CTLFLAG_RW, 0, "TCP SACK"); 132 VNET_DEFINE(int, tcp_do_sack) = 1; 133 #define V_tcp_do_sack VNET(tcp_do_sack) 134 SYSCTL_INT(_net_inet_tcp_sack, OID_AUTO, enable, CTLFLAG_VNET | CTLFLAG_RW, 135 &VNET_NAME(tcp_do_sack), 0, "Enable/Disable TCP SACK support"); 136 137 VNET_DEFINE(int, tcp_sack_maxholes) = 128; 138 SYSCTL_INT(_net_inet_tcp_sack, OID_AUTO, maxholes, CTLFLAG_VNET | CTLFLAG_RW, 139 &VNET_NAME(tcp_sack_maxholes), 0, 140 "Maximum number of TCP SACK holes allowed per connection"); 141 142 VNET_DEFINE(int, tcp_sack_globalmaxholes) = 65536; 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 SYSCTL_INT(_net_inet_tcp_sack, OID_AUTO, globalholes, CTLFLAG_VNET | CTLFLAG_RD, 149 &VNET_NAME(tcp_sack_globalholes), 0, 150 "Global number of TCP SACK holes currently allocated"); 151 152 153 /* 154 * This function will find overlaps with the currently stored sackblocks 155 * and add any overlap as a dsack block upfront 156 */ 157 void 158 tcp_update_dsack_list(struct tcpcb *tp, tcp_seq rcv_start, tcp_seq rcv_end) 159 { 160 struct sackblk head_blk,mid_blk,saved_blks[MAX_SACK_BLKS]; 161 int i, j, n, identical; 162 tcp_seq start, end; 163 164 INP_WLOCK_ASSERT(tp->t_inpcb); 165 166 KASSERT(SEQ_LT(rcv_start, rcv_end), ("rcv_start < rcv_end")); 167 168 if (SEQ_LT(rcv_end, tp->rcv_nxt) || 169 ((rcv_end == tp->rcv_nxt) && 170 (tp->rcv_numsacks > 0 ) && 171 (tp->sackblks[0].end == tp->rcv_nxt))) { 172 saved_blks[0].start = rcv_start; 173 saved_blks[0].end = rcv_end; 174 } else { 175 saved_blks[0].start = saved_blks[0].end = 0; 176 } 177 178 head_blk.start = head_blk.end = 0; 179 mid_blk.start = rcv_start; 180 mid_blk.end = rcv_end; 181 identical = 0; 182 183 for (i = 0; i < tp->rcv_numsacks; i++) { 184 start = tp->sackblks[i].start; 185 end = tp->sackblks[i].end; 186 if (SEQ_LT(rcv_end, start)) { 187 /* pkt left to sack blk */ 188 continue; 189 } 190 if (SEQ_GT(rcv_start, end)) { 191 /* pkt right to sack blk */ 192 continue; 193 } 194 if (SEQ_GT(tp->rcv_nxt, end)) { 195 if ((SEQ_MAX(rcv_start, start) != SEQ_MIN(rcv_end, end)) && 196 (SEQ_GT(head_blk.start, SEQ_MAX(rcv_start, start)) || 197 (head_blk.start == head_blk.end))) { 198 head_blk.start = SEQ_MAX(rcv_start, start); 199 head_blk.end = SEQ_MIN(rcv_end, end); 200 } 201 continue; 202 } 203 if (((head_blk.start == head_blk.end) || 204 SEQ_LT(start, head_blk.start)) && 205 (SEQ_GT(end, rcv_start) && 206 SEQ_LEQ(start, rcv_end))) { 207 head_blk.start = start; 208 head_blk.end = end; 209 } 210 mid_blk.start = SEQ_MIN(mid_blk.start, start); 211 mid_blk.end = SEQ_MAX(mid_blk.end, end); 212 if ((mid_blk.start == start) && 213 (mid_blk.end == end)) 214 identical = 1; 215 } 216 if (SEQ_LT(head_blk.start, head_blk.end)) { 217 /* store overlapping range */ 218 saved_blks[0].start = SEQ_MAX(rcv_start, head_blk.start); 219 saved_blks[0].end = SEQ_MIN(rcv_end, head_blk.end); 220 } 221 n = 1; 222 /* 223 * Second, if not ACKed, store the SACK block that 224 * overlaps with the DSACK block unless it is identical 225 */ 226 if ((SEQ_LT(tp->rcv_nxt, mid_blk.end) && 227 !((mid_blk.start == saved_blks[0].start) && 228 (mid_blk.end == saved_blks[0].end))) || 229 identical == 1) { 230 saved_blks[n].start = mid_blk.start; 231 saved_blks[n++].end = mid_blk.end; 232 } 233 for (j = 0; (j < tp->rcv_numsacks) && (n < MAX_SACK_BLKS); j++) { 234 if (((SEQ_LT(tp->sackblks[j].end, mid_blk.start) || 235 SEQ_GT(tp->sackblks[j].start, mid_blk.end)) && 236 (SEQ_GT(tp->sackblks[j].start, tp->rcv_nxt)))) 237 saved_blks[n++] = tp->sackblks[j]; 238 } 239 j = 0; 240 for (i = 0; i < n; i++) { 241 /* we can end up with a stale initial entry */ 242 if (SEQ_LT(saved_blks[i].start, saved_blks[i].end)) { 243 tp->sackblks[j++] = saved_blks[i]; 244 } 245 } 246 tp->rcv_numsacks = j; 247 } 248 249 /* 250 * This function is called upon receipt of new valid data (while not in 251 * header prediction mode), and it updates the ordered list of sacks. 252 */ 253 void 254 tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_start, tcp_seq rcv_end) 255 { 256 /* 257 * First reported block MUST be the most recent one. Subsequent 258 * blocks SHOULD be in the order in which they arrived at the 259 * receiver. These two conditions make the implementation fully 260 * compliant with RFC 2018. 261 */ 262 struct sackblk head_blk, saved_blks[MAX_SACK_BLKS]; 263 int num_head, num_saved, i; 264 265 INP_WLOCK_ASSERT(tp->t_inpcb); 266 267 /* Check arguments. */ 268 KASSERT(SEQ_LEQ(rcv_start, rcv_end), ("rcv_start <= rcv_end")); 269 270 if ((rcv_start == rcv_end) && 271 (tp->rcv_numsacks >= 1) && 272 (rcv_end == tp->sackblks[0].end)) { 273 /* retaining DSACK block below rcv_nxt (todrop) */ 274 head_blk = tp->sackblks[0]; 275 } else { 276 /* SACK block for the received segment. */ 277 head_blk.start = rcv_start; 278 head_blk.end = rcv_end; 279 } 280 281 /* 282 * Merge updated SACK blocks into head_blk, and save unchanged SACK 283 * blocks into saved_blks[]. num_saved will have the number of the 284 * saved SACK blocks. 285 */ 286 num_saved = 0; 287 for (i = 0; i < tp->rcv_numsacks; i++) { 288 tcp_seq start = tp->sackblks[i].start; 289 tcp_seq end = tp->sackblks[i].end; 290 if (SEQ_GEQ(start, end) || SEQ_LEQ(start, tp->rcv_nxt)) { 291 /* 292 * Discard this SACK block. 293 */ 294 } else if (SEQ_LEQ(head_blk.start, end) && 295 SEQ_GEQ(head_blk.end, start)) { 296 /* 297 * Merge this SACK block into head_blk. This SACK 298 * block itself will be discarded. 299 */ 300 /* 301 * |-| 302 * |---| merge 303 * 304 * |-| 305 * |---| merge 306 * 307 * |-----| 308 * |-| DSACK smaller 309 * 310 * |-| 311 * |-----| DSACK smaller 312 */ 313 if (head_blk.start == end) 314 head_blk.start = start; 315 else if (head_blk.end == start) 316 head_blk.end = end; 317 else { 318 if (SEQ_LT(head_blk.start, start)) { 319 tcp_seq temp = start; 320 start = head_blk.start; 321 head_blk.start = temp; 322 } 323 if (SEQ_GT(head_blk.end, end)) { 324 tcp_seq temp = end; 325 end = head_blk.end; 326 head_blk.end = temp; 327 } 328 if ((head_blk.start != start) || 329 (head_blk.end != end)) { 330 if ((num_saved >= 1) && 331 SEQ_GEQ(saved_blks[num_saved-1].start, start) && 332 SEQ_LEQ(saved_blks[num_saved-1].end, end)) 333 num_saved--; 334 saved_blks[num_saved].start = start; 335 saved_blks[num_saved].end = end; 336 num_saved++; 337 } 338 } 339 } else { 340 /* 341 * This block supercedes the prior block 342 */ 343 if ((num_saved >= 1) && 344 SEQ_GEQ(saved_blks[num_saved-1].start, start) && 345 SEQ_LEQ(saved_blks[num_saved-1].end, end)) 346 num_saved--; 347 /* 348 * Save this SACK block. 349 */ 350 saved_blks[num_saved].start = start; 351 saved_blks[num_saved].end = end; 352 num_saved++; 353 } 354 } 355 356 /* 357 * Update SACK list in tp->sackblks[]. 358 */ 359 num_head = 0; 360 if (SEQ_LT(rcv_start, rcv_end)) { 361 /* 362 * The received data segment is an out-of-order segment. Put 363 * head_blk at the top of SACK list. 364 */ 365 tp->sackblks[0] = head_blk; 366 num_head = 1; 367 /* 368 * If the number of saved SACK blocks exceeds its limit, 369 * discard the last SACK block. 370 */ 371 if (num_saved >= MAX_SACK_BLKS) 372 num_saved--; 373 } 374 if ((rcv_start == rcv_end) && 375 (rcv_start == tp->sackblks[0].end)) { 376 num_head = 1; 377 } 378 if (num_saved > 0) { 379 /* 380 * Copy the saved SACK blocks back. 381 */ 382 bcopy(saved_blks, &tp->sackblks[num_head], 383 sizeof(struct sackblk) * num_saved); 384 } 385 386 /* Save the number of SACK blocks. */ 387 tp->rcv_numsacks = num_head + num_saved; 388 } 389 390 void 391 tcp_clean_dsack_blocks(struct tcpcb *tp) 392 { 393 struct sackblk saved_blks[MAX_SACK_BLKS]; 394 int num_saved, i; 395 396 INP_WLOCK_ASSERT(tp->t_inpcb); 397 /* 398 * Clean up any DSACK blocks that 399 * are in our queue of sack blocks. 400 * 401 */ 402 num_saved = 0; 403 for (i = 0; i < tp->rcv_numsacks; i++) { 404 tcp_seq start = tp->sackblks[i].start; 405 tcp_seq end = tp->sackblks[i].end; 406 if (SEQ_GEQ(start, end) || SEQ_LEQ(start, tp->rcv_nxt)) { 407 /* 408 * Discard this D-SACK block. 409 */ 410 continue; 411 } 412 /* 413 * Save this SACK block. 414 */ 415 saved_blks[num_saved].start = start; 416 saved_blks[num_saved].end = end; 417 num_saved++; 418 } 419 if (num_saved > 0) { 420 /* 421 * Copy the saved SACK blocks back. 422 */ 423 bcopy(saved_blks, &tp->sackblks[0], 424 sizeof(struct sackblk) * num_saved); 425 } 426 tp->rcv_numsacks = num_saved; 427 } 428 429 /* 430 * Delete all receiver-side SACK information. 431 */ 432 void 433 tcp_clean_sackreport(struct tcpcb *tp) 434 { 435 int i; 436 437 INP_WLOCK_ASSERT(tp->t_inpcb); 438 tp->rcv_numsacks = 0; 439 for (i = 0; i < MAX_SACK_BLKS; i++) 440 tp->sackblks[i].start = tp->sackblks[i].end=0; 441 } 442 443 /* 444 * Allocate struct sackhole. 445 */ 446 static struct sackhole * 447 tcp_sackhole_alloc(struct tcpcb *tp, tcp_seq start, tcp_seq end) 448 { 449 struct sackhole *hole; 450 451 if (tp->snd_numholes >= V_tcp_sack_maxholes || 452 V_tcp_sack_globalholes >= V_tcp_sack_globalmaxholes) { 453 TCPSTAT_INC(tcps_sack_sboverflow); 454 return NULL; 455 } 456 457 hole = (struct sackhole *)uma_zalloc(V_sack_hole_zone, M_NOWAIT); 458 if (hole == NULL) 459 return NULL; 460 461 hole->start = start; 462 hole->end = end; 463 hole->rxmit = start; 464 465 tp->snd_numholes++; 466 atomic_add_int(&V_tcp_sack_globalholes, 1); 467 468 return hole; 469 } 470 471 /* 472 * Free struct sackhole. 473 */ 474 static void 475 tcp_sackhole_free(struct tcpcb *tp, struct sackhole *hole) 476 { 477 478 uma_zfree(V_sack_hole_zone, hole); 479 480 tp->snd_numholes--; 481 atomic_subtract_int(&V_tcp_sack_globalholes, 1); 482 483 KASSERT(tp->snd_numholes >= 0, ("tp->snd_numholes >= 0")); 484 KASSERT(V_tcp_sack_globalholes >= 0, ("tcp_sack_globalholes >= 0")); 485 } 486 487 /* 488 * Insert new SACK hole into scoreboard. 489 */ 490 static struct sackhole * 491 tcp_sackhole_insert(struct tcpcb *tp, tcp_seq start, tcp_seq end, 492 struct sackhole *after) 493 { 494 struct sackhole *hole; 495 496 /* Allocate a new SACK hole. */ 497 hole = tcp_sackhole_alloc(tp, start, end); 498 if (hole == NULL) 499 return NULL; 500 501 /* Insert the new SACK hole into scoreboard. */ 502 if (after != NULL) 503 TAILQ_INSERT_AFTER(&tp->snd_holes, after, hole, scblink); 504 else 505 TAILQ_INSERT_TAIL(&tp->snd_holes, hole, scblink); 506 507 /* Update SACK hint. */ 508 if (tp->sackhint.nexthole == NULL) 509 tp->sackhint.nexthole = hole; 510 511 return hole; 512 } 513 514 /* 515 * Remove SACK hole from scoreboard. 516 */ 517 static void 518 tcp_sackhole_remove(struct tcpcb *tp, struct sackhole *hole) 519 { 520 521 /* Update SACK hint. */ 522 if (tp->sackhint.nexthole == hole) 523 tp->sackhint.nexthole = TAILQ_NEXT(hole, scblink); 524 525 /* Remove this SACK hole. */ 526 TAILQ_REMOVE(&tp->snd_holes, hole, scblink); 527 528 /* Free this SACK hole. */ 529 tcp_sackhole_free(tp, hole); 530 } 531 532 /* 533 * Process cumulative ACK and the TCP SACK option to update the scoreboard. 534 * tp->snd_holes is an ordered list of holes (oldest to newest, in terms of 535 * the sequence space). 536 * Returns 1 if incoming ACK has previously unknown SACK information, 537 * 0 otherwise. Note: We treat (snd_una, th_ack) as a sack block so any changes 538 * to that (i.e. left edge moving) would also be considered a change in SACK 539 * information which is slightly different than rfc6675. 540 */ 541 int 542 tcp_sack_doack(struct tcpcb *tp, struct tcpopt *to, tcp_seq th_ack) 543 { 544 struct sackhole *cur, *temp; 545 struct sackblk sack, sack_blocks[TCP_MAX_SACK + 1], *sblkp; 546 int i, j, num_sack_blks, sack_changed; 547 548 INP_WLOCK_ASSERT(tp->t_inpcb); 549 550 num_sack_blks = 0; 551 sack_changed = 0; 552 /* 553 * If SND.UNA will be advanced by SEG.ACK, and if SACK holes exist, 554 * treat [SND.UNA, SEG.ACK) as if it is a SACK block. 555 */ 556 if (SEQ_LT(tp->snd_una, th_ack) && !TAILQ_EMPTY(&tp->snd_holes)) { 557 sack_blocks[num_sack_blks].start = tp->snd_una; 558 sack_blocks[num_sack_blks++].end = th_ack; 559 } 560 /* 561 * Append received valid SACK blocks to sack_blocks[], but only if we 562 * received new blocks from the other side. 563 */ 564 if (to->to_flags & TOF_SACK) { 565 tp->sackhint.sacked_bytes = 0; /* reset */ 566 for (i = 0; i < to->to_nsacks; i++) { 567 bcopy((to->to_sacks + i * TCPOLEN_SACK), 568 &sack, sizeof(sack)); 569 sack.start = ntohl(sack.start); 570 sack.end = ntohl(sack.end); 571 if (SEQ_GT(sack.end, sack.start) && 572 SEQ_GT(sack.start, tp->snd_una) && 573 SEQ_GT(sack.start, th_ack) && 574 SEQ_LT(sack.start, tp->snd_max) && 575 SEQ_GT(sack.end, tp->snd_una) && 576 SEQ_LEQ(sack.end, tp->snd_max)) { 577 sack_blocks[num_sack_blks++] = sack; 578 tp->sackhint.sacked_bytes += 579 (sack.end-sack.start); 580 } 581 } 582 } 583 /* 584 * Return if SND.UNA is not advanced and no valid SACK block is 585 * received. 586 */ 587 if (num_sack_blks == 0) 588 return (sack_changed); 589 590 /* 591 * Sort the SACK blocks so we can update the scoreboard with just one 592 * pass. The overhead of sorting up to 4+1 elements is less than 593 * making up to 4+1 passes over the scoreboard. 594 */ 595 for (i = 0; i < num_sack_blks; i++) { 596 for (j = i + 1; j < num_sack_blks; j++) { 597 if (SEQ_GT(sack_blocks[i].end, sack_blocks[j].end)) { 598 sack = sack_blocks[i]; 599 sack_blocks[i] = sack_blocks[j]; 600 sack_blocks[j] = sack; 601 } 602 } 603 } 604 if (TAILQ_EMPTY(&tp->snd_holes)) 605 /* 606 * Empty scoreboard. Need to initialize snd_fack (it may be 607 * uninitialized or have a bogus value). Scoreboard holes 608 * (from the sack blocks received) are created later below 609 * (in the logic that adds holes to the tail of the 610 * scoreboard). 611 */ 612 tp->snd_fack = SEQ_MAX(tp->snd_una, th_ack); 613 /* 614 * In the while-loop below, incoming SACK blocks (sack_blocks[]) and 615 * SACK holes (snd_holes) are traversed from their tails with just 616 * one pass in order to reduce the number of compares especially when 617 * the bandwidth-delay product is large. 618 * 619 * Note: Typically, in the first RTT of SACK recovery, the highest 620 * three or four SACK blocks with the same ack number are received. 621 * In the second RTT, if retransmitted data segments are not lost, 622 * the highest three or four SACK blocks with ack number advancing 623 * are received. 624 */ 625 sblkp = &sack_blocks[num_sack_blks - 1]; /* Last SACK block */ 626 tp->sackhint.last_sack_ack = sblkp->end; 627 if (SEQ_LT(tp->snd_fack, sblkp->start)) { 628 /* 629 * The highest SACK block is beyond fack. Append new SACK 630 * hole at the tail. If the second or later highest SACK 631 * blocks are also beyond the current fack, they will be 632 * inserted by way of hole splitting in the while-loop below. 633 */ 634 temp = tcp_sackhole_insert(tp, tp->snd_fack,sblkp->start,NULL); 635 if (temp != NULL) { 636 tp->snd_fack = sblkp->end; 637 /* Go to the previous sack block. */ 638 sblkp--; 639 sack_changed = 1; 640 } else { 641 /* 642 * We failed to add a new hole based on the current 643 * sack block. Skip over all the sack blocks that 644 * fall completely to the right of snd_fack and 645 * proceed to trim the scoreboard based on the 646 * remaining sack blocks. This also trims the 647 * scoreboard for th_ack (which is sack_blocks[0]). 648 */ 649 while (sblkp >= sack_blocks && 650 SEQ_LT(tp->snd_fack, sblkp->start)) 651 sblkp--; 652 if (sblkp >= sack_blocks && 653 SEQ_LT(tp->snd_fack, sblkp->end)) 654 tp->snd_fack = sblkp->end; 655 } 656 } else if (SEQ_LT(tp->snd_fack, sblkp->end)) { 657 /* fack is advanced. */ 658 tp->snd_fack = sblkp->end; 659 sack_changed = 1; 660 } 661 cur = TAILQ_LAST(&tp->snd_holes, sackhole_head); /* Last SACK hole. */ 662 /* 663 * Since the incoming sack blocks are sorted, we can process them 664 * making one sweep of the scoreboard. 665 */ 666 while (sblkp >= sack_blocks && cur != NULL) { 667 if (SEQ_GEQ(sblkp->start, cur->end)) { 668 /* 669 * SACKs data beyond the current hole. Go to the 670 * previous sack block. 671 */ 672 sblkp--; 673 continue; 674 } 675 if (SEQ_LEQ(sblkp->end, cur->start)) { 676 /* 677 * SACKs data before the current hole. Go to the 678 * previous hole. 679 */ 680 cur = TAILQ_PREV(cur, sackhole_head, scblink); 681 continue; 682 } 683 tp->sackhint.sack_bytes_rexmit -= (cur->rxmit - cur->start); 684 KASSERT(tp->sackhint.sack_bytes_rexmit >= 0, 685 ("sackhint bytes rtx >= 0")); 686 sack_changed = 1; 687 if (SEQ_LEQ(sblkp->start, cur->start)) { 688 /* Data acks at least the beginning of hole. */ 689 if (SEQ_GEQ(sblkp->end, cur->end)) { 690 /* Acks entire hole, so delete hole. */ 691 temp = cur; 692 cur = TAILQ_PREV(cur, sackhole_head, scblink); 693 tcp_sackhole_remove(tp, temp); 694 /* 695 * The sack block may ack all or part of the 696 * next hole too, so continue onto the next 697 * hole. 698 */ 699 continue; 700 } else { 701 /* Move start of hole forward. */ 702 cur->start = sblkp->end; 703 cur->rxmit = SEQ_MAX(cur->rxmit, cur->start); 704 } 705 } else { 706 /* Data acks at least the end of hole. */ 707 if (SEQ_GEQ(sblkp->end, cur->end)) { 708 /* Move end of hole backward. */ 709 cur->end = sblkp->start; 710 cur->rxmit = SEQ_MIN(cur->rxmit, cur->end); 711 } else { 712 /* 713 * ACKs some data in middle of a hole; need 714 * to split current hole 715 */ 716 temp = tcp_sackhole_insert(tp, sblkp->end, 717 cur->end, cur); 718 if (temp != NULL) { 719 if (SEQ_GT(cur->rxmit, temp->rxmit)) { 720 temp->rxmit = cur->rxmit; 721 tp->sackhint.sack_bytes_rexmit 722 += (temp->rxmit 723 - temp->start); 724 } 725 cur->end = sblkp->start; 726 cur->rxmit = SEQ_MIN(cur->rxmit, 727 cur->end); 728 } 729 } 730 } 731 tp->sackhint.sack_bytes_rexmit += (cur->rxmit - cur->start); 732 /* 733 * Testing sblkp->start against cur->start tells us whether 734 * we're done with the sack block or the sack hole. 735 * Accordingly, we advance one or the other. 736 */ 737 if (SEQ_LEQ(sblkp->start, cur->start)) 738 cur = TAILQ_PREV(cur, sackhole_head, scblink); 739 else 740 sblkp--; 741 } 742 return (sack_changed); 743 } 744 745 /* 746 * Free all SACK holes to clear the scoreboard. 747 */ 748 void 749 tcp_free_sackholes(struct tcpcb *tp) 750 { 751 struct sackhole *q; 752 753 INP_WLOCK_ASSERT(tp->t_inpcb); 754 while ((q = TAILQ_FIRST(&tp->snd_holes)) != NULL) 755 tcp_sackhole_remove(tp, q); 756 tp->sackhint.sack_bytes_rexmit = 0; 757 758 KASSERT(tp->snd_numholes == 0, ("tp->snd_numholes == 0")); 759 KASSERT(tp->sackhint.nexthole == NULL, 760 ("tp->sackhint.nexthole == NULL")); 761 } 762 763 /* 764 * Partial ack handling within a sack recovery episode. Keeping this very 765 * simple for now. When a partial ack is received, force snd_cwnd to a value 766 * that will allow the sender to transmit no more than 2 segments. If 767 * necessary, a better scheme can be adopted at a later point, but for now, 768 * the goal is to prevent the sender from bursting a large amount of data in 769 * the midst of sack recovery. 770 */ 771 void 772 tcp_sack_partialack(struct tcpcb *tp, struct tcphdr *th) 773 { 774 int num_segs = 1; 775 776 INP_WLOCK_ASSERT(tp->t_inpcb); 777 tcp_timer_activate(tp, TT_REXMT, 0); 778 tp->t_rtttime = 0; 779 /* Send one or 2 segments based on how much new data was acked. */ 780 if ((BYTES_THIS_ACK(tp, th) / tp->t_maxseg) >= 2) 781 num_segs = 2; 782 tp->snd_cwnd = (tp->sackhint.sack_bytes_rexmit + 783 (tp->snd_nxt - tp->snd_recover) + num_segs * tp->t_maxseg); 784 if (tp->snd_cwnd > tp->snd_ssthresh) 785 tp->snd_cwnd = tp->snd_ssthresh; 786 tp->t_flags |= TF_ACKNOW; 787 (void) tp->t_fb->tfb_tcp_output(tp); 788 } 789 790 #if 0 791 /* 792 * Debug version of tcp_sack_output() that walks the scoreboard. Used for 793 * now to sanity check the hint. 794 */ 795 static struct sackhole * 796 tcp_sack_output_debug(struct tcpcb *tp, int *sack_bytes_rexmt) 797 { 798 struct sackhole *p; 799 800 INP_WLOCK_ASSERT(tp->t_inpcb); 801 *sack_bytes_rexmt = 0; 802 TAILQ_FOREACH(p, &tp->snd_holes, scblink) { 803 if (SEQ_LT(p->rxmit, p->end)) { 804 if (SEQ_LT(p->rxmit, tp->snd_una)) {/* old SACK hole */ 805 continue; 806 } 807 *sack_bytes_rexmt += (p->rxmit - p->start); 808 break; 809 } 810 *sack_bytes_rexmt += (p->rxmit - p->start); 811 } 812 return (p); 813 } 814 #endif 815 816 /* 817 * Returns the next hole to retransmit and the number of retransmitted bytes 818 * from the scoreboard. We store both the next hole and the number of 819 * retransmitted bytes as hints (and recompute these on the fly upon SACK/ACK 820 * reception). This avoids scoreboard traversals completely. 821 * 822 * The loop here will traverse *at most* one link. Here's the argument. For 823 * the loop to traverse more than 1 link before finding the next hole to 824 * retransmit, we would need to have at least 1 node following the current 825 * hint with (rxmit == end). But, for all holes following the current hint, 826 * (start == rxmit), since we have not yet retransmitted from them. 827 * Therefore, in order to traverse more 1 link in the loop below, we need to 828 * have at least one node following the current hint with (start == rxmit == 829 * end). But that can't happen, (start == end) means that all the data in 830 * that hole has been sacked, in which case, the hole would have been removed 831 * from the scoreboard. 832 */ 833 struct sackhole * 834 tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt) 835 { 836 struct sackhole *hole = NULL; 837 838 INP_WLOCK_ASSERT(tp->t_inpcb); 839 *sack_bytes_rexmt = tp->sackhint.sack_bytes_rexmit; 840 hole = tp->sackhint.nexthole; 841 if (hole == NULL || SEQ_LT(hole->rxmit, hole->end)) 842 goto out; 843 while ((hole = TAILQ_NEXT(hole, scblink)) != NULL) { 844 if (SEQ_LT(hole->rxmit, hole->end)) { 845 tp->sackhint.nexthole = hole; 846 break; 847 } 848 } 849 out: 850 return (hole); 851 } 852 853 /* 854 * After a timeout, the SACK list may be rebuilt. This SACK information 855 * should be used to avoid retransmitting SACKed data. This function 856 * traverses the SACK list to see if snd_nxt should be moved forward. 857 */ 858 void 859 tcp_sack_adjust(struct tcpcb *tp) 860 { 861 struct sackhole *p, *cur = TAILQ_FIRST(&tp->snd_holes); 862 863 INP_WLOCK_ASSERT(tp->t_inpcb); 864 if (cur == NULL) 865 return; /* No holes */ 866 if (SEQ_GEQ(tp->snd_nxt, tp->snd_fack)) 867 return; /* We're already beyond any SACKed blocks */ 868 /*- 869 * Two cases for which we want to advance snd_nxt: 870 * i) snd_nxt lies between end of one hole and beginning of another 871 * ii) snd_nxt lies between end of last hole and snd_fack 872 */ 873 while ((p = TAILQ_NEXT(cur, scblink)) != NULL) { 874 if (SEQ_LT(tp->snd_nxt, cur->end)) 875 return; 876 if (SEQ_GEQ(tp->snd_nxt, p->start)) 877 cur = p; 878 else { 879 tp->snd_nxt = p->start; 880 return; 881 } 882 } 883 if (SEQ_LT(tp->snd_nxt, cur->end)) 884 return; 885 tp->snd_nxt = tp->snd_fack; 886 } 887