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 #include <netinet/cc/cc.h> 123 #ifdef TCPDEBUG 124 #include <netinet/tcp_debug.h> 125 #endif /* TCPDEBUG */ 126 127 #include <machine/in_cksum.h> 128 129 VNET_DECLARE(struct uma_zone *, sack_hole_zone); 130 #define V_sack_hole_zone VNET(sack_hole_zone) 131 132 SYSCTL_NODE(_net_inet_tcp, OID_AUTO, sack, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 133 "TCP SACK"); 134 135 VNET_DEFINE(int, tcp_do_sack) = 1; 136 SYSCTL_INT(_net_inet_tcp_sack, OID_AUTO, enable, CTLFLAG_VNET | CTLFLAG_RW, 137 &VNET_NAME(tcp_do_sack), 0, 138 "Enable/Disable TCP SACK support"); 139 140 VNET_DEFINE(int, tcp_do_newsack) = 1; 141 SYSCTL_INT(_net_inet_tcp_sack, OID_AUTO, revised, CTLFLAG_VNET | CTLFLAG_RW, 142 &VNET_NAME(tcp_do_newsack), 0, 143 "Use revised SACK loss recovery per RFC 6675"); 144 145 VNET_DEFINE(int, tcp_sack_maxholes) = 128; 146 SYSCTL_INT(_net_inet_tcp_sack, OID_AUTO, maxholes, CTLFLAG_VNET | CTLFLAG_RW, 147 &VNET_NAME(tcp_sack_maxholes), 0, 148 "Maximum number of TCP SACK holes allowed per connection"); 149 150 VNET_DEFINE(int, tcp_sack_globalmaxholes) = 65536; 151 SYSCTL_INT(_net_inet_tcp_sack, OID_AUTO, globalmaxholes, CTLFLAG_VNET | CTLFLAG_RW, 152 &VNET_NAME(tcp_sack_globalmaxholes), 0, 153 "Global maximum number of TCP SACK holes"); 154 155 VNET_DEFINE(int, tcp_sack_globalholes) = 0; 156 SYSCTL_INT(_net_inet_tcp_sack, OID_AUTO, globalholes, CTLFLAG_VNET | CTLFLAG_RD, 157 &VNET_NAME(tcp_sack_globalholes), 0, 158 "Global number of TCP SACK holes currently allocated"); 159 160 int 161 tcp_dsack_block_exists(struct tcpcb *tp) 162 { 163 /* Return true if a DSACK block exists */ 164 if (tp->rcv_numsacks == 0) 165 return (0); 166 if (SEQ_LEQ(tp->sackblks[0].end, tp->rcv_nxt)) 167 return(1); 168 return (0); 169 } 170 171 /* 172 * This function will find overlaps with the currently stored sackblocks 173 * and add any overlap as a dsack block upfront 174 */ 175 void 176 tcp_update_dsack_list(struct tcpcb *tp, tcp_seq rcv_start, tcp_seq rcv_end) 177 { 178 struct sackblk head_blk,mid_blk,saved_blks[MAX_SACK_BLKS]; 179 int i, j, n, identical; 180 tcp_seq start, end; 181 182 INP_WLOCK_ASSERT(tp->t_inpcb); 183 184 KASSERT(SEQ_LT(rcv_start, rcv_end), ("rcv_start < rcv_end")); 185 186 if (SEQ_LT(rcv_end, tp->rcv_nxt) || 187 ((rcv_end == tp->rcv_nxt) && 188 (tp->rcv_numsacks > 0 ) && 189 (tp->sackblks[0].end == tp->rcv_nxt))) { 190 saved_blks[0].start = rcv_start; 191 saved_blks[0].end = rcv_end; 192 } else { 193 saved_blks[0].start = saved_blks[0].end = 0; 194 } 195 196 head_blk.start = head_blk.end = 0; 197 mid_blk.start = rcv_start; 198 mid_blk.end = rcv_end; 199 identical = 0; 200 201 for (i = 0; i < tp->rcv_numsacks; i++) { 202 start = tp->sackblks[i].start; 203 end = tp->sackblks[i].end; 204 if (SEQ_LT(rcv_end, start)) { 205 /* pkt left to sack blk */ 206 continue; 207 } 208 if (SEQ_GT(rcv_start, end)) { 209 /* pkt right to sack blk */ 210 continue; 211 } 212 if (SEQ_GT(tp->rcv_nxt, end)) { 213 if ((SEQ_MAX(rcv_start, start) != SEQ_MIN(rcv_end, end)) && 214 (SEQ_GT(head_blk.start, SEQ_MAX(rcv_start, start)) || 215 (head_blk.start == head_blk.end))) { 216 head_blk.start = SEQ_MAX(rcv_start, start); 217 head_blk.end = SEQ_MIN(rcv_end, end); 218 } 219 continue; 220 } 221 if (((head_blk.start == head_blk.end) || 222 SEQ_LT(start, head_blk.start)) && 223 (SEQ_GT(end, rcv_start) && 224 SEQ_LEQ(start, rcv_end))) { 225 head_blk.start = start; 226 head_blk.end = end; 227 } 228 mid_blk.start = SEQ_MIN(mid_blk.start, start); 229 mid_blk.end = SEQ_MAX(mid_blk.end, end); 230 if ((mid_blk.start == start) && 231 (mid_blk.end == end)) 232 identical = 1; 233 } 234 if (SEQ_LT(head_blk.start, head_blk.end)) { 235 /* store overlapping range */ 236 saved_blks[0].start = SEQ_MAX(rcv_start, head_blk.start); 237 saved_blks[0].end = SEQ_MIN(rcv_end, head_blk.end); 238 } 239 n = 1; 240 /* 241 * Second, if not ACKed, store the SACK block that 242 * overlaps with the DSACK block unless it is identical 243 */ 244 if ((SEQ_LT(tp->rcv_nxt, mid_blk.end) && 245 !((mid_blk.start == saved_blks[0].start) && 246 (mid_blk.end == saved_blks[0].end))) || 247 identical == 1) { 248 saved_blks[n].start = mid_blk.start; 249 saved_blks[n++].end = mid_blk.end; 250 } 251 for (j = 0; (j < tp->rcv_numsacks) && (n < MAX_SACK_BLKS); j++) { 252 if (((SEQ_LT(tp->sackblks[j].end, mid_blk.start) || 253 SEQ_GT(tp->sackblks[j].start, mid_blk.end)) && 254 (SEQ_GT(tp->sackblks[j].start, tp->rcv_nxt)))) 255 saved_blks[n++] = tp->sackblks[j]; 256 } 257 j = 0; 258 for (i = 0; i < n; i++) { 259 /* we can end up with a stale initial entry */ 260 if (SEQ_LT(saved_blks[i].start, saved_blks[i].end)) { 261 tp->sackblks[j++] = saved_blks[i]; 262 } 263 } 264 tp->rcv_numsacks = j; 265 } 266 267 /* 268 * This function is called upon receipt of new valid data (while not in 269 * header prediction mode), and it updates the ordered list of sacks. 270 */ 271 void 272 tcp_update_sack_list(struct tcpcb *tp, tcp_seq rcv_start, tcp_seq rcv_end) 273 { 274 /* 275 * First reported block MUST be the most recent one. Subsequent 276 * blocks SHOULD be in the order in which they arrived at the 277 * receiver. These two conditions make the implementation fully 278 * compliant with RFC 2018. 279 */ 280 struct sackblk head_blk, saved_blks[MAX_SACK_BLKS]; 281 int num_head, num_saved, i; 282 283 INP_WLOCK_ASSERT(tp->t_inpcb); 284 285 /* Check arguments. */ 286 KASSERT(SEQ_LEQ(rcv_start, rcv_end), ("rcv_start <= rcv_end")); 287 288 if ((rcv_start == rcv_end) && 289 (tp->rcv_numsacks >= 1) && 290 (rcv_end == tp->sackblks[0].end)) { 291 /* retaining DSACK block below rcv_nxt (todrop) */ 292 head_blk = tp->sackblks[0]; 293 } else { 294 /* SACK block for the received segment. */ 295 head_blk.start = rcv_start; 296 head_blk.end = rcv_end; 297 } 298 299 /* 300 * Merge updated SACK blocks into head_blk, and save unchanged SACK 301 * blocks into saved_blks[]. num_saved will have the number of the 302 * saved SACK blocks. 303 */ 304 num_saved = 0; 305 for (i = 0; i < tp->rcv_numsacks; i++) { 306 tcp_seq start = tp->sackblks[i].start; 307 tcp_seq end = tp->sackblks[i].end; 308 if (SEQ_GEQ(start, end) || SEQ_LEQ(start, tp->rcv_nxt)) { 309 /* 310 * Discard this SACK block. 311 */ 312 } else if (SEQ_LEQ(head_blk.start, end) && 313 SEQ_GEQ(head_blk.end, start)) { 314 /* 315 * Merge this SACK block into head_blk. This SACK 316 * block itself will be discarded. 317 */ 318 /* 319 * |-| 320 * |---| merge 321 * 322 * |-| 323 * |---| merge 324 * 325 * |-----| 326 * |-| DSACK smaller 327 * 328 * |-| 329 * |-----| DSACK smaller 330 */ 331 if (head_blk.start == end) 332 head_blk.start = start; 333 else if (head_blk.end == start) 334 head_blk.end = end; 335 else { 336 if (SEQ_LT(head_blk.start, start)) { 337 tcp_seq temp = start; 338 start = head_blk.start; 339 head_blk.start = temp; 340 } 341 if (SEQ_GT(head_blk.end, end)) { 342 tcp_seq temp = end; 343 end = head_blk.end; 344 head_blk.end = temp; 345 } 346 if ((head_blk.start != start) || 347 (head_blk.end != end)) { 348 if ((num_saved >= 1) && 349 SEQ_GEQ(saved_blks[num_saved-1].start, start) && 350 SEQ_LEQ(saved_blks[num_saved-1].end, end)) 351 num_saved--; 352 saved_blks[num_saved].start = start; 353 saved_blks[num_saved].end = end; 354 num_saved++; 355 } 356 } 357 } else { 358 /* 359 * This block supercedes the prior block 360 */ 361 if ((num_saved >= 1) && 362 SEQ_GEQ(saved_blks[num_saved-1].start, start) && 363 SEQ_LEQ(saved_blks[num_saved-1].end, end)) 364 num_saved--; 365 /* 366 * Save this SACK block. 367 */ 368 saved_blks[num_saved].start = start; 369 saved_blks[num_saved].end = end; 370 num_saved++; 371 } 372 } 373 374 /* 375 * Update SACK list in tp->sackblks[]. 376 */ 377 num_head = 0; 378 if (SEQ_LT(rcv_start, rcv_end)) { 379 /* 380 * The received data segment is an out-of-order segment. Put 381 * head_blk at the top of SACK list. 382 */ 383 tp->sackblks[0] = head_blk; 384 num_head = 1; 385 /* 386 * If the number of saved SACK blocks exceeds its limit, 387 * discard the last SACK block. 388 */ 389 if (num_saved >= MAX_SACK_BLKS) 390 num_saved--; 391 } 392 if ((rcv_start == rcv_end) && 393 (rcv_start == tp->sackblks[0].end)) { 394 num_head = 1; 395 } 396 if (num_saved > 0) { 397 /* 398 * Copy the saved SACK blocks back. 399 */ 400 bcopy(saved_blks, &tp->sackblks[num_head], 401 sizeof(struct sackblk) * num_saved); 402 } 403 404 /* Save the number of SACK blocks. */ 405 tp->rcv_numsacks = num_head + num_saved; 406 } 407 408 void 409 tcp_clean_dsack_blocks(struct tcpcb *tp) 410 { 411 struct sackblk saved_blks[MAX_SACK_BLKS]; 412 int num_saved, i; 413 414 INP_WLOCK_ASSERT(tp->t_inpcb); 415 /* 416 * Clean up any DSACK blocks that 417 * are in our queue of sack blocks. 418 * 419 */ 420 num_saved = 0; 421 for (i = 0; i < tp->rcv_numsacks; i++) { 422 tcp_seq start = tp->sackblks[i].start; 423 tcp_seq end = tp->sackblks[i].end; 424 if (SEQ_GEQ(start, end) || SEQ_LEQ(start, tp->rcv_nxt)) { 425 /* 426 * Discard this D-SACK block. 427 */ 428 continue; 429 } 430 /* 431 * Save this SACK block. 432 */ 433 saved_blks[num_saved].start = start; 434 saved_blks[num_saved].end = end; 435 num_saved++; 436 } 437 if (num_saved > 0) { 438 /* 439 * Copy the saved SACK blocks back. 440 */ 441 bcopy(saved_blks, &tp->sackblks[0], 442 sizeof(struct sackblk) * num_saved); 443 } 444 tp->rcv_numsacks = num_saved; 445 } 446 447 /* 448 * Delete all receiver-side SACK information. 449 */ 450 void 451 tcp_clean_sackreport(struct tcpcb *tp) 452 { 453 int i; 454 455 INP_WLOCK_ASSERT(tp->t_inpcb); 456 tp->rcv_numsacks = 0; 457 for (i = 0; i < MAX_SACK_BLKS; i++) 458 tp->sackblks[i].start = tp->sackblks[i].end=0; 459 } 460 461 /* 462 * Allocate struct sackhole. 463 */ 464 static struct sackhole * 465 tcp_sackhole_alloc(struct tcpcb *tp, tcp_seq start, tcp_seq end) 466 { 467 struct sackhole *hole; 468 469 if (tp->snd_numholes >= V_tcp_sack_maxholes || 470 V_tcp_sack_globalholes >= V_tcp_sack_globalmaxholes) { 471 TCPSTAT_INC(tcps_sack_sboverflow); 472 return NULL; 473 } 474 475 hole = (struct sackhole *)uma_zalloc(V_sack_hole_zone, M_NOWAIT); 476 if (hole == NULL) 477 return NULL; 478 479 hole->start = start; 480 hole->end = end; 481 hole->rxmit = start; 482 483 tp->snd_numholes++; 484 atomic_add_int(&V_tcp_sack_globalholes, 1); 485 486 return hole; 487 } 488 489 /* 490 * Free struct sackhole. 491 */ 492 static void 493 tcp_sackhole_free(struct tcpcb *tp, struct sackhole *hole) 494 { 495 496 uma_zfree(V_sack_hole_zone, hole); 497 498 tp->snd_numholes--; 499 atomic_subtract_int(&V_tcp_sack_globalholes, 1); 500 501 KASSERT(tp->snd_numholes >= 0, ("tp->snd_numholes >= 0")); 502 KASSERT(V_tcp_sack_globalholes >= 0, ("tcp_sack_globalholes >= 0")); 503 } 504 505 /* 506 * Insert new SACK hole into scoreboard. 507 */ 508 static struct sackhole * 509 tcp_sackhole_insert(struct tcpcb *tp, tcp_seq start, tcp_seq end, 510 struct sackhole *after) 511 { 512 struct sackhole *hole; 513 514 /* Allocate a new SACK hole. */ 515 hole = tcp_sackhole_alloc(tp, start, end); 516 if (hole == NULL) 517 return NULL; 518 519 /* Insert the new SACK hole into scoreboard. */ 520 if (after != NULL) 521 TAILQ_INSERT_AFTER(&tp->snd_holes, after, hole, scblink); 522 else 523 TAILQ_INSERT_TAIL(&tp->snd_holes, hole, scblink); 524 525 /* Update SACK hint. */ 526 if (tp->sackhint.nexthole == NULL) 527 tp->sackhint.nexthole = hole; 528 529 return hole; 530 } 531 532 /* 533 * Remove SACK hole from scoreboard. 534 */ 535 static void 536 tcp_sackhole_remove(struct tcpcb *tp, struct sackhole *hole) 537 { 538 539 /* Update SACK hint. */ 540 if (tp->sackhint.nexthole == hole) 541 tp->sackhint.nexthole = TAILQ_NEXT(hole, scblink); 542 543 /* Remove this SACK hole. */ 544 TAILQ_REMOVE(&tp->snd_holes, hole, scblink); 545 546 /* Free this SACK hole. */ 547 tcp_sackhole_free(tp, hole); 548 } 549 550 /* 551 * Process cumulative ACK and the TCP SACK option to update the scoreboard. 552 * tp->snd_holes is an ordered list of holes (oldest to newest, in terms of 553 * the sequence space). 554 * Returns 1 if incoming ACK has previously unknown SACK information, 555 * 0 otherwise. 556 */ 557 int 558 tcp_sack_doack(struct tcpcb *tp, struct tcpopt *to, tcp_seq th_ack) 559 { 560 struct sackhole *cur, *temp; 561 struct sackblk sack, sack_blocks[TCP_MAX_SACK + 1], *sblkp; 562 int i, j, num_sack_blks, sack_changed; 563 int delivered_data, left_edge_delta; 564 565 INP_WLOCK_ASSERT(tp->t_inpcb); 566 567 num_sack_blks = 0; 568 sack_changed = 0; 569 delivered_data = 0; 570 left_edge_delta = 0; 571 /* 572 * If SND.UNA will be advanced by SEG.ACK, and if SACK holes exist, 573 * treat [SND.UNA, SEG.ACK) as if it is a SACK block. 574 * Account changes to SND.UNA always in delivered data. 575 */ 576 if (SEQ_LT(tp->snd_una, th_ack) && !TAILQ_EMPTY(&tp->snd_holes)) { 577 left_edge_delta = th_ack - tp->snd_una; 578 sack_blocks[num_sack_blks].start = tp->snd_una; 579 sack_blocks[num_sack_blks++].end = th_ack; 580 /* 581 * Pulling snd_fack forward if we got here 582 * due to DSACK blocks 583 */ 584 if (SEQ_LT(tp->snd_fack, th_ack)) { 585 delivered_data += th_ack - tp->snd_una; 586 tp->snd_fack = th_ack; 587 sack_changed = 1; 588 } 589 } 590 /* 591 * Append received valid SACK blocks to sack_blocks[], but only if we 592 * received new blocks from the other side. 593 */ 594 if (to->to_flags & TOF_SACK) { 595 for (i = 0; i < to->to_nsacks; i++) { 596 bcopy((to->to_sacks + i * TCPOLEN_SACK), 597 &sack, sizeof(sack)); 598 sack.start = ntohl(sack.start); 599 sack.end = ntohl(sack.end); 600 if (SEQ_GT(sack.end, sack.start) && 601 SEQ_GT(sack.start, tp->snd_una) && 602 SEQ_GT(sack.start, th_ack) && 603 SEQ_LT(sack.start, tp->snd_max) && 604 SEQ_GT(sack.end, tp->snd_una) && 605 SEQ_LEQ(sack.end, tp->snd_max)) { 606 sack_blocks[num_sack_blks++] = sack; 607 } else if (SEQ_LEQ(sack.start, th_ack) && 608 SEQ_LEQ(sack.end, th_ack)) { 609 /* 610 * Its a D-SACK block. 611 */ 612 tcp_record_dsack(tp, sack.start, sack.end, 0); 613 } 614 } 615 } 616 /* 617 * Return if SND.UNA is not advanced and no valid SACK block is 618 * received. 619 */ 620 if (num_sack_blks == 0) 621 return (sack_changed); 622 623 /* 624 * Sort the SACK blocks so we can update the scoreboard with just one 625 * pass. The overhead of sorting up to 4+1 elements is less than 626 * making up to 4+1 passes over the scoreboard. 627 */ 628 for (i = 0; i < num_sack_blks; i++) { 629 for (j = i + 1; j < num_sack_blks; j++) { 630 if (SEQ_GT(sack_blocks[i].end, sack_blocks[j].end)) { 631 sack = sack_blocks[i]; 632 sack_blocks[i] = sack_blocks[j]; 633 sack_blocks[j] = sack; 634 } 635 } 636 } 637 if (TAILQ_EMPTY(&tp->snd_holes)) { 638 /* 639 * Empty scoreboard. Need to initialize snd_fack (it may be 640 * uninitialized or have a bogus value). Scoreboard holes 641 * (from the sack blocks received) are created later below 642 * (in the logic that adds holes to the tail of the 643 * scoreboard). 644 */ 645 tp->snd_fack = SEQ_MAX(tp->snd_una, th_ack); 646 tp->sackhint.sacked_bytes = 0; /* reset */ 647 } 648 /* 649 * In the while-loop below, incoming SACK blocks (sack_blocks[]) and 650 * SACK holes (snd_holes) are traversed from their tails with just 651 * one pass in order to reduce the number of compares especially when 652 * the bandwidth-delay product is large. 653 * 654 * Note: Typically, in the first RTT of SACK recovery, the highest 655 * three or four SACK blocks with the same ack number are received. 656 * In the second RTT, if retransmitted data segments are not lost, 657 * the highest three or four SACK blocks with ack number advancing 658 * are received. 659 */ 660 sblkp = &sack_blocks[num_sack_blks - 1]; /* Last SACK block */ 661 tp->sackhint.last_sack_ack = sblkp->end; 662 if (SEQ_LT(tp->snd_fack, sblkp->start)) { 663 /* 664 * The highest SACK block is beyond fack. First, 665 * check if there was a successful Rescue Retransmission, 666 * and move this hole left. With normal holes, snd_fack 667 * is always to the right of the end. 668 */ 669 if (((temp = TAILQ_LAST(&tp->snd_holes, sackhole_head)) != NULL) && 670 SEQ_LEQ(tp->snd_fack,temp->end)) { 671 temp->start = SEQ_MAX(tp->snd_fack, SEQ_MAX(tp->snd_una, th_ack)); 672 temp->end = sblkp->start; 673 temp->rxmit = temp->start; 674 delivered_data += sblkp->end - sblkp->start; 675 tp->snd_fack = sblkp->end; 676 sblkp--; 677 sack_changed = 1; 678 } else { 679 /* 680 * Append a new SACK hole at the tail. If the 681 * second or later highest SACK blocks are also 682 * beyond the current fack, they will be inserted 683 * by way of hole splitting in the while-loop below. 684 */ 685 temp = tcp_sackhole_insert(tp, tp->snd_fack,sblkp->start,NULL); 686 if (temp != NULL) { 687 delivered_data += sblkp->end - sblkp->start; 688 tp->snd_fack = sblkp->end; 689 /* Go to the previous sack block. */ 690 sblkp--; 691 sack_changed = 1; 692 } else { 693 /* 694 * We failed to add a new hole based on the current 695 * sack block. Skip over all the sack blocks that 696 * fall completely to the right of snd_fack and 697 * proceed to trim the scoreboard based on the 698 * remaining sack blocks. This also trims the 699 * scoreboard for th_ack (which is sack_blocks[0]). 700 */ 701 while (sblkp >= sack_blocks && 702 SEQ_LT(tp->snd_fack, sblkp->start)) 703 sblkp--; 704 if (sblkp >= sack_blocks && 705 SEQ_LT(tp->snd_fack, sblkp->end)) { 706 delivered_data += sblkp->end - tp->snd_fack; 707 tp->snd_fack = sblkp->end; 708 sack_changed = 1; 709 } 710 } 711 } 712 } else if (SEQ_LT(tp->snd_fack, sblkp->end)) { 713 /* fack is advanced. */ 714 delivered_data += sblkp->end - tp->snd_fack; 715 tp->snd_fack = sblkp->end; 716 sack_changed = 1; 717 } 718 cur = TAILQ_LAST(&tp->snd_holes, sackhole_head); /* Last SACK hole. */ 719 /* 720 * Since the incoming sack blocks are sorted, we can process them 721 * making one sweep of the scoreboard. 722 */ 723 while (sblkp >= sack_blocks && cur != NULL) { 724 if (SEQ_GEQ(sblkp->start, cur->end)) { 725 /* 726 * SACKs data beyond the current hole. Go to the 727 * previous sack block. 728 */ 729 sblkp--; 730 continue; 731 } 732 if (SEQ_LEQ(sblkp->end, cur->start)) { 733 /* 734 * SACKs data before the current hole. Go to the 735 * previous hole. 736 */ 737 cur = TAILQ_PREV(cur, sackhole_head, scblink); 738 continue; 739 } 740 tp->sackhint.sack_bytes_rexmit -= 741 (SEQ_MIN(cur->rxmit, cur->end) - cur->start); 742 KASSERT(tp->sackhint.sack_bytes_rexmit >= 0, 743 ("sackhint bytes rtx >= 0")); 744 sack_changed = 1; 745 if (SEQ_LEQ(sblkp->start, cur->start)) { 746 /* Data acks at least the beginning of hole. */ 747 if (SEQ_GEQ(sblkp->end, cur->end)) { 748 /* Acks entire hole, so delete hole. */ 749 delivered_data += (cur->end - cur->start); 750 temp = cur; 751 cur = TAILQ_PREV(cur, sackhole_head, scblink); 752 tcp_sackhole_remove(tp, temp); 753 /* 754 * The sack block may ack all or part of the 755 * next hole too, so continue onto the next 756 * hole. 757 */ 758 continue; 759 } else { 760 /* Move start of hole forward. */ 761 delivered_data += (sblkp->end - cur->start); 762 cur->start = sblkp->end; 763 cur->rxmit = SEQ_MAX(cur->rxmit, cur->start); 764 } 765 } else { 766 /* Data acks at least the end of hole. */ 767 if (SEQ_GEQ(sblkp->end, cur->end)) { 768 /* Move end of hole backward. */ 769 delivered_data += (cur->end - sblkp->start); 770 cur->end = sblkp->start; 771 cur->rxmit = SEQ_MIN(cur->rxmit, cur->end); 772 if ((tp->t_flags & TF_LRD) && SEQ_GEQ(cur->rxmit, cur->end)) 773 cur->rxmit = tp->snd_recover; 774 } else { 775 /* 776 * ACKs some data in middle of a hole; need 777 * to split current hole 778 */ 779 temp = tcp_sackhole_insert(tp, sblkp->end, 780 cur->end, cur); 781 if (temp != NULL) { 782 if (SEQ_GT(cur->rxmit, temp->rxmit)) { 783 temp->rxmit = cur->rxmit; 784 tp->sackhint.sack_bytes_rexmit += 785 (SEQ_MIN(temp->rxmit, 786 temp->end) - temp->start); 787 } 788 cur->end = sblkp->start; 789 cur->rxmit = SEQ_MIN(cur->rxmit, 790 cur->end); 791 if ((tp->t_flags & TF_LRD) && SEQ_GEQ(cur->rxmit, cur->end)) 792 cur->rxmit = tp->snd_recover; 793 delivered_data += (sblkp->end - sblkp->start); 794 } 795 } 796 } 797 tp->sackhint.sack_bytes_rexmit += 798 (SEQ_MIN(cur->rxmit, cur->end) - cur->start); 799 /* 800 * Testing sblkp->start against cur->start tells us whether 801 * we're done with the sack block or the sack hole. 802 * Accordingly, we advance one or the other. 803 */ 804 if (SEQ_LEQ(sblkp->start, cur->start)) 805 cur = TAILQ_PREV(cur, sackhole_head, scblink); 806 else 807 sblkp--; 808 } 809 if (!(to->to_flags & TOF_SACK)) 810 /* 811 * If this ACK did not contain any 812 * SACK blocks, any only moved the 813 * left edge right, it is a pure 814 * cumulative ACK. Do not count 815 * DupAck for this. Also required 816 * for RFC6675 rescue retransmission. 817 */ 818 sack_changed = 0; 819 tp->sackhint.delivered_data = delivered_data; 820 tp->sackhint.sacked_bytes += delivered_data - left_edge_delta; 821 KASSERT((delivered_data >= 0), ("delivered_data < 0")); 822 KASSERT((tp->sackhint.sacked_bytes >= 0), ("sacked_bytes < 0")); 823 return (sack_changed); 824 } 825 826 /* 827 * Free all SACK holes to clear the scoreboard. 828 */ 829 void 830 tcp_free_sackholes(struct tcpcb *tp) 831 { 832 struct sackhole *q; 833 834 INP_WLOCK_ASSERT(tp->t_inpcb); 835 while ((q = TAILQ_FIRST(&tp->snd_holes)) != NULL) 836 tcp_sackhole_remove(tp, q); 837 tp->sackhint.sack_bytes_rexmit = 0; 838 839 KASSERT(tp->snd_numholes == 0, ("tp->snd_numholes == 0")); 840 KASSERT(tp->sackhint.nexthole == NULL, 841 ("tp->sackhint.nexthole == NULL")); 842 } 843 844 /* 845 * Partial ack handling within a sack recovery episode. Keeping this very 846 * simple for now. When a partial ack is received, force snd_cwnd to a value 847 * that will allow the sender to transmit no more than 2 segments. If 848 * necessary, a better scheme can be adopted at a later point, but for now, 849 * the goal is to prevent the sender from bursting a large amount of data in 850 * the midst of sack recovery. 851 */ 852 void 853 tcp_sack_partialack(struct tcpcb *tp, struct tcphdr *th) 854 { 855 int num_segs = 1; 856 u_int maxseg = tcp_maxseg(tp); 857 858 INP_WLOCK_ASSERT(tp->t_inpcb); 859 tcp_timer_activate(tp, TT_REXMT, 0); 860 tp->t_rtttime = 0; 861 /* Send one or 2 segments based on how much new data was acked. */ 862 if ((BYTES_THIS_ACK(tp, th) / maxseg) >= 2) 863 num_segs = 2; 864 tp->snd_cwnd = (tp->sackhint.sack_bytes_rexmit + 865 (tp->snd_nxt - tp->snd_recover) + num_segs * maxseg); 866 if (tp->snd_cwnd > tp->snd_ssthresh) 867 tp->snd_cwnd = tp->snd_ssthresh; 868 tp->t_flags |= TF_ACKNOW; 869 /* 870 * RFC6675 rescue retransmission 871 * Add a hole between th_ack (snd_una is not yet set) and snd_max, 872 * if this was a pure cumulative ACK and no data was send beyond 873 * recovery point. Since the data in the socket has not been freed 874 * at this point, we check if the scoreboard is empty, and the ACK 875 * delivered some new data, indicating a full ACK. Also, if the 876 * recovery point is still at snd_max, we are probably application 877 * limited. However, this inference might not always be true. The 878 * rescue retransmission may rarely be slightly premature 879 * compared to RFC6675. 880 * The corresponding ACK+SACK will cause any further outstanding 881 * segments to be retransmitted. This addresses a corner case, when 882 * the trailing packets of a window are lost and no further data 883 * is available for sending. 884 */ 885 if ((V_tcp_do_newsack) && 886 SEQ_LT(th->th_ack, tp->snd_recover) && 887 (tp->snd_recover == tp->snd_max) && 888 TAILQ_EMPTY(&tp->snd_holes) && 889 (tp->sackhint.delivered_data > 0)) { 890 /* 891 * Exclude FIN sequence space in 892 * the hole for the rescue retransmission, 893 * and also don't create a hole, if only 894 * the ACK for a FIN is outstanding. 895 */ 896 tcp_seq highdata = tp->snd_max; 897 if (tp->t_flags & TF_SENTFIN) 898 highdata--; 899 if (th->th_ack != highdata) { 900 tp->snd_fack = th->th_ack; 901 (void)tcp_sackhole_insert(tp, SEQ_MAX(th->th_ack, 902 highdata - maxseg), highdata, NULL); 903 } 904 } 905 (void) tcp_output(tp); 906 } 907 908 #if 0 909 /* 910 * Debug version of tcp_sack_output() that walks the scoreboard. Used for 911 * now to sanity check the hint. 912 */ 913 static struct sackhole * 914 tcp_sack_output_debug(struct tcpcb *tp, int *sack_bytes_rexmt) 915 { 916 struct sackhole *p; 917 918 INP_WLOCK_ASSERT(tp->t_inpcb); 919 *sack_bytes_rexmt = 0; 920 TAILQ_FOREACH(p, &tp->snd_holes, scblink) { 921 if (SEQ_LT(p->rxmit, p->end)) { 922 if (SEQ_LT(p->rxmit, tp->snd_una)) {/* old SACK hole */ 923 continue; 924 } 925 *sack_bytes_rexmt += (p->rxmit - p->start); 926 break; 927 } 928 *sack_bytes_rexmt += (SEQ_MIN(p->rxmit, p->end) - p->start); 929 } 930 return (p); 931 } 932 #endif 933 934 /* 935 * Returns the next hole to retransmit and the number of retransmitted bytes 936 * from the scoreboard. We store both the next hole and the number of 937 * retransmitted bytes as hints (and recompute these on the fly upon SACK/ACK 938 * reception). This avoids scoreboard traversals completely. 939 * 940 * The loop here will traverse *at most* one link. Here's the argument. For 941 * the loop to traverse more than 1 link before finding the next hole to 942 * retransmit, we would need to have at least 1 node following the current 943 * hint with (rxmit == end). But, for all holes following the current hint, 944 * (start == rxmit), since we have not yet retransmitted from them. 945 * Therefore, in order to traverse more 1 link in the loop below, we need to 946 * have at least one node following the current hint with (start == rxmit == 947 * end). But that can't happen, (start == end) means that all the data in 948 * that hole has been sacked, in which case, the hole would have been removed 949 * from the scoreboard. 950 */ 951 struct sackhole * 952 tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt) 953 { 954 struct sackhole *hole = NULL; 955 956 INP_WLOCK_ASSERT(tp->t_inpcb); 957 *sack_bytes_rexmt = tp->sackhint.sack_bytes_rexmit; 958 hole = tp->sackhint.nexthole; 959 if (hole == NULL || SEQ_LT(hole->rxmit, hole->end)) 960 goto out; 961 while ((hole = TAILQ_NEXT(hole, scblink)) != NULL) { 962 if (SEQ_LT(hole->rxmit, hole->end)) { 963 tp->sackhint.nexthole = hole; 964 break; 965 } 966 } 967 out: 968 return (hole); 969 } 970 971 /* 972 * After a timeout, the SACK list may be rebuilt. This SACK information 973 * should be used to avoid retransmitting SACKed data. This function 974 * traverses the SACK list to see if snd_nxt should be moved forward. 975 */ 976 void 977 tcp_sack_adjust(struct tcpcb *tp) 978 { 979 struct sackhole *p, *cur = TAILQ_FIRST(&tp->snd_holes); 980 981 INP_WLOCK_ASSERT(tp->t_inpcb); 982 if (cur == NULL) 983 return; /* No holes */ 984 if (SEQ_GEQ(tp->snd_nxt, tp->snd_fack)) 985 return; /* We're already beyond any SACKed blocks */ 986 /*- 987 * Two cases for which we want to advance snd_nxt: 988 * i) snd_nxt lies between end of one hole and beginning of another 989 * ii) snd_nxt lies between end of last hole and snd_fack 990 */ 991 while ((p = TAILQ_NEXT(cur, scblink)) != NULL) { 992 if (SEQ_LT(tp->snd_nxt, cur->end)) 993 return; 994 if (SEQ_GEQ(tp->snd_nxt, p->start)) 995 cur = p; 996 else { 997 tp->snd_nxt = p->start; 998 return; 999 } 1000 } 1001 if (SEQ_LT(tp->snd_nxt, cur->end)) 1002 return; 1003 tp->snd_nxt = tp->snd_fack; 1004 } 1005 1006 /* 1007 * Lost Retransmission Detection 1008 * Check is FACK is beyond the rexmit of the leftmost hole. 1009 * If yes, we restart sending from still existing holes, 1010 * and adjust cwnd via the congestion control module. 1011 */ 1012 void 1013 tcp_sack_lost_retransmission(struct tcpcb *tp, struct tcphdr *th) 1014 { 1015 struct sackhole *temp; 1016 1017 if (IN_RECOVERY(tp->t_flags) && 1018 SEQ_GT(tp->snd_fack, tp->snd_recover) && 1019 ((temp = TAILQ_FIRST(&tp->snd_holes)) != NULL) && 1020 SEQ_GEQ(temp->rxmit, temp->end) && 1021 SEQ_GEQ(tp->snd_fack, temp->rxmit)) { 1022 TCPSTAT_INC(tcps_sack_lostrexmt); 1023 /* 1024 * Start retransmissions from the first hole, and 1025 * subsequently all other remaining holes, including 1026 * those, which had been sent completely before. 1027 */ 1028 tp->sackhint.nexthole = temp; 1029 TAILQ_FOREACH(temp, &tp->snd_holes, scblink) { 1030 if (SEQ_GEQ(tp->snd_fack, temp->rxmit) && 1031 SEQ_GEQ(temp->rxmit, temp->end)) 1032 temp->rxmit = temp->start; 1033 } 1034 /* 1035 * Remember the old ssthresh, to deduct the beta factor used 1036 * by the CC module. Finally, set cwnd to ssthresh just 1037 * prior to invoking another cwnd reduction by the CC 1038 * module, to not shrink it excessively. 1039 */ 1040 tp->snd_cwnd = tp->snd_ssthresh; 1041 /* 1042 * Formally exit recovery, and let the CC module adjust 1043 * ssthresh as intended. 1044 */ 1045 EXIT_RECOVERY(tp->t_flags); 1046 cc_cong_signal(tp, th, CC_NDUPACK); 1047 /* 1048 * For PRR, adjust recover_fs as if this new reduction 1049 * initialized this variable. 1050 * cwnd will be adjusted by SACK or PRR processing 1051 * subsequently, only set it to a safe value here. 1052 */ 1053 tp->snd_cwnd = tcp_maxseg(tp); 1054 tp->sackhint.recover_fs = (tp->snd_max - tp->snd_una) - 1055 tp->sackhint.recover_fs; 1056 } 1057 } 1058