altq_subr.c revision 1.33.46.1
1 /* $NetBSD: altq_subr.c,v 1.33.46.1 2023/11/11 13:16:30 thorpej Exp $ */ 2 /* $KAME: altq_subr.c,v 1.24 2005/04/13 03:44:25 suz Exp $ */ 3 4 /* 5 * Copyright (C) 1997-2003 6 * Sony Computer Science Laboratories Inc. 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 * 17 * THIS SOFTWARE IS PROVIDED BY SONY CSL 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 SONY CSL 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 30 #include <sys/cdefs.h> 31 __KERNEL_RCSID(0, "$NetBSD: altq_subr.c,v 1.33.46.1 2023/11/11 13:16:30 thorpej Exp $"); 32 33 #ifdef _KERNEL_OPT 34 #include "opt_altq.h" 35 #include "opt_inet.h" 36 #include "pf.h" 37 #endif 38 39 #include <sys/param.h> 40 #include <sys/malloc.h> 41 #include <sys/mbuf.h> 42 #include <sys/systm.h> 43 #include <sys/proc.h> 44 #include <sys/socket.h> 45 #include <sys/socketvar.h> 46 #include <sys/kernel.h> 47 #include <sys/kmem.h> 48 #include <sys/errno.h> 49 #include <sys/syslog.h> 50 #include <sys/sysctl.h> 51 #include <sys/queue.h> 52 53 #include <net/if.h> 54 #include <net/if_dl.h> 55 #include <net/if_types.h> 56 57 #include <netinet/in.h> 58 #include <netinet/in_systm.h> 59 #include <netinet/ip.h> 60 #ifdef INET6 61 #include <netinet/ip6.h> 62 #endif 63 #include <netinet/tcp.h> 64 #include <netinet/udp.h> 65 66 #if NPF > 0 67 #include <net/pfvar.h> 68 #endif 69 #include <altq/altq.h> 70 #ifdef ALTQ3_COMPAT 71 #include <altq/altq_conf.h> 72 #endif 73 74 /* 75 * internal function prototypes 76 */ 77 static void tbr_timeout(void *); 78 int (*altq_input)(struct mbuf *, int) = NULL; 79 static int tbr_timer = 0; /* token bucket regulator timer */ 80 static struct callout tbr_callout; 81 82 #ifdef ALTQ3_CLFIER_COMPAT 83 static int extract_ports4(struct mbuf *, struct ip *, struct flowinfo_in *); 84 #ifdef INET6 85 static int extract_ports6(struct mbuf *, struct ip6_hdr *, 86 struct flowinfo_in6 *); 87 #endif 88 static int apply_filter4(u_int32_t, struct flow_filter *, 89 struct flowinfo_in *); 90 static int apply_ppfilter4(u_int32_t, struct flow_filter *, 91 struct flowinfo_in *); 92 #ifdef INET6 93 static int apply_filter6(u_int32_t, struct flow_filter6 *, 94 struct flowinfo_in6 *); 95 #endif 96 static int apply_tosfilter4(u_int32_t, struct flow_filter *, 97 struct flowinfo_in *); 98 static u_long get_filt_handle(struct acc_classifier *, int); 99 static struct acc_filter *filth_to_filtp(struct acc_classifier *, u_long); 100 static u_int32_t filt2fibmask(struct flow_filter *); 101 102 static void ip4f_cache(struct ip *, struct flowinfo_in *); 103 static int ip4f_lookup(struct ip *, struct flowinfo_in *); 104 static int ip4f_init(void); 105 static struct ip4_frag *ip4f_alloc(void); 106 static void ip4f_free(struct ip4_frag *); 107 #endif /* ALTQ3_CLFIER_COMPAT */ 108 109 /* 110 * alternate queueing support routines 111 */ 112 113 /* 114 * Allocate an ifaltq and associate it with the specified ifqueue and 115 * ifnet. 116 */ 117 void 118 altq_alloc(struct ifqueue *ifq, struct ifnet *ifp) 119 { 120 if (ifq->ifq_altq == NULL) { 121 ifq->ifq_altq = kmem_zalloc(sizeof(*ifq->ifq_altq), KM_SLEEP); 122 ifq->ifq_altq->altq_ifq = ifq; 123 124 /* 125 * This allows the ALTQ_*() mcaros to work with either 126 * an ifqueue or an ifaltq argument. Yes, it's a hack, 127 * but it results in less code churn. 128 */ 129 ifq->ifq_altq->ifq_altq = ifq->ifq_altq; 130 } 131 if (ifp != NULL) { 132 ifq->ifq_altq->altq_ifp = ifp; 133 } 134 } 135 136 /* 137 * Free the ifaltq structure associated with an ifqueue. 138 */ 139 void 140 altq_free(struct ifqueue *ifq) 141 { 142 if (ifq->ifq_altq != NULL) { 143 ifq->ifq_altq->altq_ifp = NULL; 144 kmem_free(ifq->ifq_altq, sizeof(*ifq->ifq_altq)); 145 ifq->ifq_altq = NULL; 146 } 147 } 148 149 /* 150 * Mark's a device's send queue as being ready for (as in 151 * "knowledgeable about") ALTQ processing. 152 */ 153 void 154 altq_set_ready(struct ifqueue *ifq) 155 { 156 altq_alloc(ifq, NULL); 157 ifq->ifq_altq->altq_flags = ALTQF_READY; 158 } 159 160 /* look up the queue state by the interface name and the queueing type. */ 161 void * 162 altq_lookup(char *name, int type) 163 { 164 struct ifnet *ifp; 165 166 if ((ifp = ifunit(name)) != NULL) { 167 struct ifaltq *altq = ifp->if_snd.ifq_altq; 168 if (type != ALTQT_NONE && altq->altq_type == type) 169 return (altq->altq_disc); 170 } 171 172 return NULL; 173 } 174 175 int 176 altq_attach(struct ifaltq *ifq, int type, void *discipline, 177 int (*enqueue)(struct ifaltq *, struct mbuf *), 178 struct mbuf *(*dequeue)(struct ifaltq *, int), 179 int (*request)(struct ifaltq *, int, void *), 180 void *clfier, void *(*classify)(void *, struct mbuf *, int)) 181 { 182 if (!ALTQ_IS_READY(ifq)) 183 return ENXIO; 184 185 #ifdef ALTQ3_COMPAT 186 /* 187 * pfaltq can override the existing discipline, but altq3 cannot. 188 * check these if clfier is not NULL (which implies altq3). 189 */ 190 if (clfier != NULL) { 191 if (ALTQ_IS_ENABLED(ifq)) 192 return EBUSY; 193 if (ALTQ_IS_ATTACHED(ifq)) 194 return EEXIST; 195 } 196 #endif 197 ifq->altq_type = type; 198 ifq->altq_disc = discipline; 199 ifq->altq_enqueue = enqueue; 200 ifq->altq_dequeue = dequeue; 201 ifq->altq_request = request; 202 ifq->altq_clfier = clfier; 203 ifq->altq_classify = classify; 204 ifq->altq_flags &= (ALTQF_CANTCHANGE|ALTQF_ENABLED); 205 #ifdef ALTQ3_COMPAT 206 #ifdef ALTQ_KLD 207 altq_module_incref(type); 208 #endif 209 #endif 210 return 0; 211 } 212 213 int 214 altq_detach(struct ifaltq *ifq) 215 { 216 if (!ALTQ_IS_READY(ifq)) 217 return ENXIO; 218 if (ALTQ_IS_ENABLED(ifq)) 219 return EBUSY; 220 if (!ALTQ_IS_ATTACHED(ifq)) 221 return (0); 222 #ifdef ALTQ3_COMPAT 223 #ifdef ALTQ_KLD 224 altq_module_declref(ifq->altq_type); 225 #endif 226 #endif 227 228 ifq->altq_type = ALTQT_NONE; 229 ifq->altq_disc = NULL; 230 ifq->altq_enqueue = NULL; 231 ifq->altq_dequeue = NULL; 232 ifq->altq_request = NULL; 233 ifq->altq_clfier = NULL; 234 ifq->altq_classify = NULL; 235 ifq->altq_flags &= ALTQF_CANTCHANGE; 236 return 0; 237 } 238 239 int 240 altq_enable(struct ifaltq *ifq) 241 { 242 int s; 243 244 if (!ALTQ_IS_READY(ifq)) 245 return ENXIO; 246 if (ALTQ_IS_ENABLED(ifq)) 247 return 0; 248 249 s = splnet(); 250 IFQ_PURGE(ifq->altq_ifq); 251 ASSERT(ALTQ_GET_LEN(ifq) == 0); 252 ifq->altq_flags |= ALTQF_ENABLED; 253 if (ifq->altq_clfier != NULL) 254 ifq->altq_flags |= ALTQF_CLASSIFY; 255 splx(s); 256 257 return 0; 258 } 259 260 int 261 altq_disable(struct ifaltq *ifq) 262 { 263 int s; 264 265 if (!ALTQ_IS_ENABLED(ifq)) 266 return 0; 267 268 s = splnet(); 269 IFQ_PURGE(ifq->altq_ifq); 270 ASSERT(ALTQ_GET_LEN(ifq) == 0); 271 ifq->altq_flags &= ~(ALTQF_ENABLED|ALTQF_CLASSIFY); 272 splx(s); 273 return 0; 274 } 275 276 #ifdef ALTQ_DEBUG 277 void 278 altq_assert(const char *file, int line, const char *failedexpr) 279 { 280 (void)printf("altq assertion \"%s\" failed: file \"%s\", line %d\n", 281 failedexpr, file, line); 282 panic("altq assertion"); 283 /* NOTREACHED */ 284 } 285 #endif 286 287 /* 288 * internal representation of token bucket parameters 289 * rate: byte_per_unittime << 32 290 * (((bits_per_sec) / 8) << 32) / machclk_freq 291 * depth: byte << 32 292 * 293 */ 294 #define TBR_SHIFT 32 295 #define TBR_SCALE(x) ((int64_t)(x) << TBR_SHIFT) 296 #define TBR_UNSCALE(x) ((x) >> TBR_SHIFT) 297 298 struct mbuf * 299 tbr_dequeue(struct ifaltq *ifq, int op) 300 { 301 struct tb_regulator *tbr; 302 struct mbuf *m; 303 int64_t interval; 304 u_int64_t now; 305 306 tbr = ifq->altq_tbr; 307 if (op == ALTDQ_REMOVE && tbr->tbr_lastop == ALTDQ_POLL) { 308 /* if this is a remove after poll, bypass tbr check */ 309 } else { 310 /* update token only when it is negative */ 311 if (tbr->tbr_token <= 0) { 312 now = read_machclk(); 313 interval = now - tbr->tbr_last; 314 if (interval >= tbr->tbr_filluptime) 315 tbr->tbr_token = tbr->tbr_depth; 316 else { 317 tbr->tbr_token += interval * tbr->tbr_rate; 318 if (tbr->tbr_token > tbr->tbr_depth) 319 tbr->tbr_token = tbr->tbr_depth; 320 } 321 tbr->tbr_last = now; 322 } 323 /* if token is still negative, don't allow dequeue */ 324 if (tbr->tbr_token <= 0) 325 return (NULL); 326 } 327 328 if (ALTQ_IS_ENABLED(ifq)) 329 m = (*ifq->altq_dequeue)(ifq, op); 330 else { 331 if (op == ALTDQ_POLL) 332 IF_POLL(ifq->altq_ifq, m); 333 else 334 IF_DEQUEUE(ifq->altq_ifq, m); 335 } 336 337 if (m != NULL && op == ALTDQ_REMOVE) 338 tbr->tbr_token -= TBR_SCALE(m_pktlen(m)); 339 tbr->tbr_lastop = op; 340 return (m); 341 } 342 343 /* 344 * set a token bucket regulator. 345 * if the specified rate is zero, the token bucket regulator is deleted. 346 */ 347 int 348 tbr_set(struct ifaltq *ifq, struct tb_profile *profile) 349 { 350 struct tb_regulator *tbr, *otbr; 351 352 if (machclk_freq == 0) 353 init_machclk(); 354 if (machclk_freq == 0) { 355 printf("tbr_set: no CPU clock available!\n"); 356 return (ENXIO); 357 } 358 359 if (profile->rate == 0) { 360 /* delete this tbr */ 361 if ((tbr = ifq->altq_tbr) == NULL) 362 return (ENOENT); 363 ifq->altq_tbr = NULL; 364 free(tbr, M_DEVBUF); 365 return (0); 366 } 367 368 tbr = malloc(sizeof(struct tb_regulator), M_DEVBUF, M_WAITOK|M_ZERO); 369 if (tbr == NULL) 370 return (ENOMEM); 371 372 tbr->tbr_rate = TBR_SCALE(profile->rate / 8) / machclk_freq; 373 tbr->tbr_depth = TBR_SCALE(profile->depth); 374 if (tbr->tbr_rate > 0) 375 tbr->tbr_filluptime = tbr->tbr_depth / tbr->tbr_rate; 376 else 377 tbr->tbr_filluptime = 0xffffffffffffffffLL; 378 tbr->tbr_token = tbr->tbr_depth; 379 tbr->tbr_last = read_machclk(); 380 tbr->tbr_lastop = ALTDQ_REMOVE; 381 382 otbr = ifq->altq_tbr; 383 ifq->altq_tbr = tbr; /* set the new tbr */ 384 385 if (otbr != NULL) { 386 free(otbr, M_DEVBUF); 387 } else { 388 if (tbr_timer == 0) { 389 CALLOUT_RESET(&tbr_callout, 1, tbr_timeout, (void *)0); 390 tbr_timer = 1; 391 } 392 } 393 return (0); 394 } 395 396 /* 397 * tbr_timeout goes through the interface list, and kicks the drivers 398 * if necessary. 399 */ 400 static void 401 tbr_timeout(void *arg) 402 { 403 struct ifnet *ifp; 404 int active, s; 405 406 active = 0; 407 s = pserialize_read_enter(); 408 IFNET_READER_FOREACH(ifp) { 409 struct psref psref; 410 if (!TBR_IS_ENABLED(&ifp->if_snd)) 411 continue; 412 if_acquire(ifp, &psref); 413 pserialize_read_exit(s); 414 415 active++; 416 if (!IFQ_IS_EMPTY(&ifp->if_snd) && ifp->if_start != NULL) { 417 int _s = splnet(); 418 if_start_lock(ifp); 419 splx(_s); 420 } 421 422 s = pserialize_read_enter(); 423 if_release(ifp, &psref); 424 } 425 pserialize_read_exit(s); 426 427 if (active > 0) 428 CALLOUT_RESET(&tbr_callout, 1, tbr_timeout, (void *)0); 429 else 430 tbr_timer = 0; /* don't need tbr_timer anymore */ 431 } 432 433 /* 434 * get token bucket regulator profile 435 */ 436 int 437 tbr_get(struct ifaltq *ifq, struct tb_profile *profile) 438 { 439 struct tb_regulator *tbr; 440 441 if ((tbr = ifq->altq_tbr) == NULL) { 442 profile->rate = 0; 443 profile->depth = 0; 444 } else { 445 profile->rate = 446 (u_int)TBR_UNSCALE(tbr->tbr_rate * 8 * machclk_freq); 447 profile->depth = (u_int)TBR_UNSCALE(tbr->tbr_depth); 448 } 449 return (0); 450 } 451 452 #if NPF > 0 453 /* 454 * attach a discipline to the interface. if one already exists, it is 455 * overridden. 456 */ 457 int 458 altq_pfattach(struct pf_altq *a) 459 { 460 int error = 0; 461 462 switch (a->scheduler) { 463 case ALTQT_NONE: 464 break; 465 #ifdef ALTQ_CBQ 466 case ALTQT_CBQ: 467 error = cbq_pfattach(a); 468 break; 469 #endif 470 #ifdef ALTQ_PRIQ 471 case ALTQT_PRIQ: 472 error = priq_pfattach(a); 473 break; 474 #endif 475 #ifdef ALTQ_HFSC 476 case ALTQT_HFSC: 477 error = hfsc_pfattach(a); 478 break; 479 #endif 480 default: 481 error = ENXIO; 482 } 483 484 return (error); 485 } 486 487 /* 488 * detach a discipline from the interface. 489 * it is possible that the discipline was already overridden by another 490 * discipline. 491 */ 492 int 493 altq_pfdetach(struct pf_altq *a) 494 { 495 struct ifnet *ifp; 496 int s, error = 0; 497 498 if ((ifp = ifunit(a->ifname)) == NULL) 499 return (EINVAL); 500 501 /* if this discipline is no longer referenced, just return */ 502 if (a->altq_disc == NULL || a->altq_disc != ifp->if_snd.altq_disc) 503 return (0); 504 505 s = splnet(); 506 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 507 error = altq_disable(&ifp->if_snd); 508 if (error == 0) 509 error = altq_detach(&ifp->if_snd); 510 splx(s); 511 512 return (error); 513 } 514 515 /* 516 * add a discipline or a queue 517 */ 518 int 519 altq_add(struct pf_altq *a) 520 { 521 int error = 0; 522 523 if (a->qname[0] != 0) 524 return (altq_add_queue(a)); 525 526 if (machclk_freq == 0) 527 init_machclk(); 528 if (machclk_freq == 0) 529 panic("altq_add: no CPU clock"); 530 531 switch (a->scheduler) { 532 #ifdef ALTQ_CBQ 533 case ALTQT_CBQ: 534 error = cbq_add_altq(a); 535 break; 536 #endif 537 #ifdef ALTQ_PRIQ 538 case ALTQT_PRIQ: 539 error = priq_add_altq(a); 540 break; 541 #endif 542 #ifdef ALTQ_HFSC 543 case ALTQT_HFSC: 544 error = hfsc_add_altq(a); 545 break; 546 #endif 547 default: 548 error = ENXIO; 549 } 550 551 return (error); 552 } 553 554 /* 555 * remove a discipline or a queue 556 */ 557 int 558 altq_remove(struct pf_altq *a) 559 { 560 int error = 0; 561 562 if (a->qname[0] != 0) 563 return (altq_remove_queue(a)); 564 565 switch (a->scheduler) { 566 #ifdef ALTQ_CBQ 567 case ALTQT_CBQ: 568 error = cbq_remove_altq(a); 569 break; 570 #endif 571 #ifdef ALTQ_PRIQ 572 case ALTQT_PRIQ: 573 error = priq_remove_altq(a); 574 break; 575 #endif 576 #ifdef ALTQ_HFSC 577 case ALTQT_HFSC: 578 error = hfsc_remove_altq(a); 579 break; 580 #endif 581 default: 582 error = ENXIO; 583 } 584 585 return (error); 586 } 587 588 /* 589 * add a queue to the discipline 590 */ 591 int 592 altq_add_queue(struct pf_altq *a) 593 { 594 int error = 0; 595 596 switch (a->scheduler) { 597 #ifdef ALTQ_CBQ 598 case ALTQT_CBQ: 599 error = cbq_add_queue(a); 600 break; 601 #endif 602 #ifdef ALTQ_PRIQ 603 case ALTQT_PRIQ: 604 error = priq_add_queue(a); 605 break; 606 #endif 607 #ifdef ALTQ_HFSC 608 case ALTQT_HFSC: 609 error = hfsc_add_queue(a); 610 break; 611 #endif 612 default: 613 error = ENXIO; 614 } 615 616 return (error); 617 } 618 619 /* 620 * remove a queue from the discipline 621 */ 622 int 623 altq_remove_queue(struct pf_altq *a) 624 { 625 int error = 0; 626 627 switch (a->scheduler) { 628 #ifdef ALTQ_CBQ 629 case ALTQT_CBQ: 630 error = cbq_remove_queue(a); 631 break; 632 #endif 633 #ifdef ALTQ_PRIQ 634 case ALTQT_PRIQ: 635 error = priq_remove_queue(a); 636 break; 637 #endif 638 #ifdef ALTQ_HFSC 639 case ALTQT_HFSC: 640 error = hfsc_remove_queue(a); 641 break; 642 #endif 643 default: 644 error = ENXIO; 645 } 646 647 return (error); 648 } 649 650 /* 651 * get queue statistics 652 */ 653 int 654 altq_getqstats(struct pf_altq *a, void *ubuf, int *nbytes) 655 { 656 int error = 0; 657 658 switch (a->scheduler) { 659 #ifdef ALTQ_CBQ 660 case ALTQT_CBQ: 661 error = cbq_getqstats(a, ubuf, nbytes); 662 break; 663 #endif 664 #ifdef ALTQ_PRIQ 665 case ALTQT_PRIQ: 666 error = priq_getqstats(a, ubuf, nbytes); 667 break; 668 #endif 669 #ifdef ALTQ_HFSC 670 case ALTQT_HFSC: 671 error = hfsc_getqstats(a, ubuf, nbytes); 672 break; 673 #endif 674 default: 675 error = ENXIO; 676 } 677 678 return (error); 679 } 680 #endif /* NPF > 0 */ 681 682 /* 683 * read and write diffserv field in IPv4 or IPv6 header 684 */ 685 u_int8_t 686 read_dsfield(struct mbuf *m, struct altq_pktattr *pktattr) 687 { 688 struct mbuf *m0; 689 u_int8_t ds_field = 0; 690 691 if (pktattr == NULL || 692 (pktattr->pattr_af != AF_INET && pktattr->pattr_af != AF_INET6)) 693 return ((u_int8_t)0); 694 695 /* verify that pattr_hdr is within the mbuf data */ 696 for (m0 = m; m0 != NULL; m0 = m0->m_next) 697 if (((char *)pktattr->pattr_hdr >= m0->m_data) && 698 ((char *)pktattr->pattr_hdr < m0->m_data + m0->m_len)) 699 break; 700 if (m0 == NULL) { 701 /* ick, pattr_hdr is stale */ 702 pktattr->pattr_af = AF_UNSPEC; 703 #ifdef ALTQ_DEBUG 704 printf("read_dsfield: can't locate header!\n"); 705 #endif 706 return ((u_int8_t)0); 707 } 708 709 if (pktattr->pattr_af == AF_INET) { 710 struct ip *ip = (struct ip *)pktattr->pattr_hdr; 711 712 if (ip->ip_v != 4) 713 return ((u_int8_t)0); /* version mismatch! */ 714 ds_field = ip->ip_tos; 715 } 716 #ifdef INET6 717 else if (pktattr->pattr_af == AF_INET6) { 718 struct ip6_hdr *ip6 = (struct ip6_hdr *)pktattr->pattr_hdr; 719 u_int32_t flowlabel; 720 721 flowlabel = ntohl(ip6->ip6_flow); 722 if ((flowlabel >> 28) != 6) 723 return ((u_int8_t)0); /* version mismatch! */ 724 ds_field = (flowlabel >> 20) & 0xff; 725 } 726 #endif 727 return (ds_field); 728 } 729 730 void 731 write_dsfield(struct mbuf *m, struct altq_pktattr *pktattr, u_int8_t dsfield) 732 { 733 struct mbuf *m0; 734 735 if (pktattr == NULL || 736 (pktattr->pattr_af != AF_INET && pktattr->pattr_af != AF_INET6)) 737 return; 738 739 /* verify that pattr_hdr is within the mbuf data */ 740 for (m0 = m; m0 != NULL; m0 = m0->m_next) 741 if (((char *)pktattr->pattr_hdr >= m0->m_data) && 742 ((char *)pktattr->pattr_hdr < m0->m_data + m0->m_len)) 743 break; 744 if (m0 == NULL) { 745 /* ick, pattr_hdr is stale */ 746 pktattr->pattr_af = AF_UNSPEC; 747 #ifdef ALTQ_DEBUG 748 printf("write_dsfield: can't locate header!\n"); 749 #endif 750 return; 751 } 752 753 if (pktattr->pattr_af == AF_INET) { 754 struct ip *ip = (struct ip *)pktattr->pattr_hdr; 755 u_int8_t old; 756 int32_t sum; 757 758 if (ip->ip_v != 4) 759 return; /* version mismatch! */ 760 old = ip->ip_tos; 761 dsfield |= old & 3; /* leave CU bits */ 762 if (old == dsfield) 763 return; 764 ip->ip_tos = dsfield; 765 /* 766 * update checksum (from RFC1624) 767 * HC' = ~(~HC + ~m + m') 768 */ 769 sum = ~ntohs(ip->ip_sum) & 0xffff; 770 sum += 0xff00 + (~old & 0xff) + dsfield; 771 sum = (sum >> 16) + (sum & 0xffff); 772 sum += (sum >> 16); /* add carry */ 773 774 ip->ip_sum = htons(~sum & 0xffff); 775 } 776 #ifdef INET6 777 else if (pktattr->pattr_af == AF_INET6) { 778 struct ip6_hdr *ip6 = (struct ip6_hdr *)pktattr->pattr_hdr; 779 u_int32_t flowlabel; 780 781 flowlabel = ntohl(ip6->ip6_flow); 782 if ((flowlabel >> 28) != 6) 783 return; /* version mismatch! */ 784 flowlabel = (flowlabel & 0xf03fffff) | (dsfield << 20); 785 ip6->ip6_flow = htonl(flowlabel); 786 } 787 #endif 788 return; 789 } 790 791 #define BINTIME_SHIFT 2 792 793 u_int32_t machclk_freq = 0; 794 u_int32_t machclk_per_tick = 0; 795 796 void 797 init_machclk(void) 798 { 799 800 callout_init(&tbr_callout, 0); 801 802 /* 803 * Always emulate 1GiHz counter using bintime(9) 804 * since it has enough resolution via timecounter(9). 805 * Using machine dependent cpu_counter() is not MP safe 806 * and it won't work even on UP with Speedstep etc. 807 */ 808 machclk_freq = 1024 * 1024 * 1024; /* 2^30 to emulate ~1GHz */ 809 machclk_per_tick = machclk_freq / hz; 810 #ifdef ALTQ_DEBUG 811 printf("altq: emulate %uHz CPU clock\n", machclk_freq); 812 #endif 813 } 814 815 u_int64_t 816 read_machclk(void) 817 { 818 struct bintime bt; 819 u_int64_t val; 820 821 binuptime(&bt); 822 val = (((u_int64_t)bt.sec << 32) + (bt.frac >> 32)) >> BINTIME_SHIFT; 823 return (val); 824 } 825 826 #ifdef ALTQ3_CLFIER_COMPAT 827 828 #ifndef IPPROTO_ESP 829 #define IPPROTO_ESP 50 /* encapsulating security payload */ 830 #endif 831 #ifndef IPPROTO_AH 832 #define IPPROTO_AH 51 /* authentication header */ 833 #endif 834 835 /* 836 * extract flow information from a given packet. 837 * filt_mask shows flowinfo fields required. 838 * we assume the ip header is in one mbuf, and addresses and ports are 839 * in network byte order. 840 */ 841 int 842 altq_extractflow(struct mbuf *m, int af, struct flowinfo *flow, 843 u_int32_t filt_bmask) 844 { 845 846 switch (af) { 847 case PF_INET: { 848 struct flowinfo_in *fin; 849 struct ip *ip; 850 851 ip = mtod(m, struct ip *); 852 853 if (ip->ip_v != 4) 854 break; 855 856 fin = (struct flowinfo_in *)flow; 857 fin->fi_len = sizeof(struct flowinfo_in); 858 fin->fi_family = AF_INET; 859 860 fin->fi_proto = ip->ip_p; 861 fin->fi_tos = ip->ip_tos; 862 863 fin->fi_src.s_addr = ip->ip_src.s_addr; 864 fin->fi_dst.s_addr = ip->ip_dst.s_addr; 865 866 if (filt_bmask & FIMB4_PORTS) 867 /* if port info is required, extract port numbers */ 868 extract_ports4(m, ip, fin); 869 else { 870 fin->fi_sport = 0; 871 fin->fi_dport = 0; 872 fin->fi_gpi = 0; 873 } 874 return (1); 875 } 876 877 #ifdef INET6 878 case PF_INET6: { 879 struct flowinfo_in6 *fin6; 880 struct ip6_hdr *ip6; 881 882 ip6 = mtod(m, struct ip6_hdr *); 883 /* should we check the ip version? */ 884 885 fin6 = (struct flowinfo_in6 *)flow; 886 fin6->fi6_len = sizeof(struct flowinfo_in6); 887 fin6->fi6_family = AF_INET6; 888 889 fin6->fi6_proto = ip6->ip6_nxt; 890 fin6->fi6_tclass = (ntohl(ip6->ip6_flow) >> 20) & 0xff; 891 892 fin6->fi6_flowlabel = ip6->ip6_flow & htonl(0x000fffff); 893 fin6->fi6_src = ip6->ip6_src; 894 fin6->fi6_dst = ip6->ip6_dst; 895 896 if ((filt_bmask & FIMB6_PORTS) || 897 ((filt_bmask & FIMB6_PROTO) 898 && ip6->ip6_nxt > IPPROTO_IPV6)) 899 /* 900 * if port info is required, or proto is required 901 * but there are option headers, extract port 902 * and protocol numbers. 903 */ 904 extract_ports6(m, ip6, fin6); 905 else { 906 fin6->fi6_sport = 0; 907 fin6->fi6_dport = 0; 908 fin6->fi6_gpi = 0; 909 } 910 return (1); 911 } 912 #endif /* INET6 */ 913 914 default: 915 break; 916 } 917 918 /* failed */ 919 flow->fi_len = sizeof(struct flowinfo); 920 flow->fi_family = AF_UNSPEC; 921 return (0); 922 } 923 924 /* 925 * helper routine to extract port numbers 926 */ 927 /* structure for ipsec and ipv6 option header template */ 928 struct _opt6 { 929 u_int8_t opt6_nxt; /* next header */ 930 u_int8_t opt6_hlen; /* header extension length */ 931 u_int16_t _pad; 932 u_int32_t ah_spi; /* security parameter index 933 for authentication header */ 934 }; 935 936 /* 937 * extract port numbers from a ipv4 packet. 938 */ 939 static int 940 extract_ports4(struct mbuf *m, struct ip *ip, struct flowinfo_in *fin) 941 { 942 struct mbuf *m0; 943 u_short ip_off; 944 u_int8_t proto; 945 int off; 946 947 fin->fi_sport = 0; 948 fin->fi_dport = 0; 949 fin->fi_gpi = 0; 950 951 ip_off = ntohs(ip->ip_off); 952 /* if it is a fragment, try cached fragment info */ 953 if (ip_off & IP_OFFMASK) { 954 ip4f_lookup(ip, fin); 955 return (1); 956 } 957 958 /* locate the mbuf containing the protocol header */ 959 for (m0 = m; m0 != NULL; m0 = m0->m_next) 960 if (((char *)ip >= m0->m_data) && 961 ((char *)ip < m0->m_data + m0->m_len)) 962 break; 963 if (m0 == NULL) { 964 #ifdef ALTQ_DEBUG 965 printf("extract_ports4: can't locate header! ip=%p\n", ip); 966 #endif 967 return (0); 968 } 969 off = ((char *)ip - m0->m_data) + (ip->ip_hl << 2); 970 proto = ip->ip_p; 971 972 #ifdef ALTQ_IPSEC 973 again: 974 #endif 975 while (off >= m0->m_len) { 976 off -= m0->m_len; 977 m0 = m0->m_next; 978 if (m0 == NULL) 979 return (0); /* bogus ip_hl! */ 980 } 981 if (m0->m_len < off + 4) 982 return (0); 983 984 switch (proto) { 985 case IPPROTO_TCP: 986 case IPPROTO_UDP: { 987 struct udphdr *udp; 988 989 udp = (struct udphdr *)(mtod(m0, char *) + off); 990 fin->fi_sport = udp->uh_sport; 991 fin->fi_dport = udp->uh_dport; 992 fin->fi_proto = proto; 993 } 994 break; 995 996 #ifdef ALTQ_IPSEC 997 case IPPROTO_ESP: 998 if (fin->fi_gpi == 0){ 999 u_int32_t *gpi; 1000 1001 gpi = (u_int32_t *)(mtod(m0, char *) + off); 1002 fin->fi_gpi = *gpi; 1003 } 1004 fin->fi_proto = proto; 1005 break; 1006 1007 case IPPROTO_AH: { 1008 /* get next header and header length */ 1009 struct _opt6 *opt6; 1010 1011 opt6 = (struct _opt6 *)(mtod(m0, char *) + off); 1012 proto = opt6->opt6_nxt; 1013 off += 8 + (opt6->opt6_hlen * 4); 1014 if (fin->fi_gpi == 0 && m0->m_len >= off + 8) 1015 fin->fi_gpi = opt6->ah_spi; 1016 } 1017 /* goto the next header */ 1018 goto again; 1019 #endif /* ALTQ_IPSEC */ 1020 1021 default: 1022 fin->fi_proto = proto; 1023 return (0); 1024 } 1025 1026 /* if this is a first fragment, cache it. */ 1027 if (ip_off & IP_MF) 1028 ip4f_cache(ip, fin); 1029 1030 return (1); 1031 } 1032 1033 #ifdef INET6 1034 static int 1035 extract_ports6(struct mbuf *m, struct ip6_hdr *ip6, struct flowinfo_in6 *fin6) 1036 { 1037 struct mbuf *m0; 1038 int off; 1039 u_int8_t proto; 1040 1041 fin6->fi6_gpi = 0; 1042 fin6->fi6_sport = 0; 1043 fin6->fi6_dport = 0; 1044 1045 /* locate the mbuf containing the protocol header */ 1046 for (m0 = m; m0 != NULL; m0 = m0->m_next) 1047 if (((char *)ip6 >= m0->m_data) && 1048 ((char *)ip6 < m0->m_data + m0->m_len)) 1049 break; 1050 if (m0 == NULL) { 1051 #ifdef ALTQ_DEBUG 1052 printf("extract_ports6: can't locate header! ip6=%p\n", ip6); 1053 #endif 1054 return (0); 1055 } 1056 off = ((char *)ip6 - m0->m_data) + sizeof(struct ip6_hdr); 1057 1058 proto = ip6->ip6_nxt; 1059 do { 1060 while (off >= m0->m_len) { 1061 off -= m0->m_len; 1062 m0 = m0->m_next; 1063 if (m0 == NULL) 1064 return (0); 1065 } 1066 if (m0->m_len < off + 4) 1067 return (0); 1068 1069 switch (proto) { 1070 case IPPROTO_TCP: 1071 case IPPROTO_UDP: { 1072 struct udphdr *udp; 1073 1074 udp = (struct udphdr *)(mtod(m0, char *) + off); 1075 fin6->fi6_sport = udp->uh_sport; 1076 fin6->fi6_dport = udp->uh_dport; 1077 fin6->fi6_proto = proto; 1078 } 1079 return (1); 1080 1081 case IPPROTO_ESP: 1082 if (fin6->fi6_gpi == 0) { 1083 u_int32_t *gpi; 1084 1085 gpi = (u_int32_t *)(mtod(m0, char *) + off); 1086 fin6->fi6_gpi = *gpi; 1087 } 1088 fin6->fi6_proto = proto; 1089 return (1); 1090 1091 case IPPROTO_AH: { 1092 /* get next header and header length */ 1093 struct _opt6 *opt6; 1094 1095 opt6 = (struct _opt6 *)(mtod(m0, char *) + off); 1096 if (fin6->fi6_gpi == 0 && m0->m_len >= off + 8) 1097 fin6->fi6_gpi = opt6->ah_spi; 1098 proto = opt6->opt6_nxt; 1099 off += 8 + (opt6->opt6_hlen * 4); 1100 /* goto the next header */ 1101 break; 1102 } 1103 1104 case IPPROTO_HOPOPTS: 1105 case IPPROTO_ROUTING: 1106 case IPPROTO_DSTOPTS: { 1107 /* get next header and header length */ 1108 struct _opt6 *opt6; 1109 1110 opt6 = (struct _opt6 *)(mtod(m0, char *) + off); 1111 proto = opt6->opt6_nxt; 1112 off += (opt6->opt6_hlen + 1) * 8; 1113 /* goto the next header */ 1114 break; 1115 } 1116 1117 case IPPROTO_FRAGMENT: 1118 /* ipv6 fragmentations are not supported yet */ 1119 default: 1120 fin6->fi6_proto = proto; 1121 return (0); 1122 } 1123 } while (1); 1124 /*NOTREACHED*/ 1125 } 1126 #endif /* INET6 */ 1127 1128 /* 1129 * altq common classifier 1130 */ 1131 int 1132 acc_add_filter(struct acc_classifier *classifier, struct flow_filter *filter, 1133 void *class, u_long *phandle) 1134 { 1135 struct acc_filter *afp, *prev, *tmp; 1136 int i, s; 1137 1138 #ifdef INET6 1139 if (filter->ff_flow.fi_family != AF_INET && 1140 filter->ff_flow.fi_family != AF_INET6) 1141 return (EINVAL); 1142 #else 1143 if (filter->ff_flow.fi_family != AF_INET) 1144 return (EINVAL); 1145 #endif 1146 1147 afp = malloc(sizeof(struct acc_filter), M_DEVBUF, M_WAITOK|M_ZERO); 1148 if (afp == NULL) 1149 return (ENOMEM); 1150 1151 afp->f_filter = *filter; 1152 afp->f_class = class; 1153 1154 i = ACC_WILDCARD_INDEX; 1155 if (filter->ff_flow.fi_family == AF_INET) { 1156 struct flow_filter *filter4 = &afp->f_filter; 1157 1158 /* 1159 * if address is 0, it's a wildcard. if address mask 1160 * isn't set, use full mask. 1161 */ 1162 if (filter4->ff_flow.fi_dst.s_addr == 0) 1163 filter4->ff_mask.mask_dst.s_addr = 0; 1164 else if (filter4->ff_mask.mask_dst.s_addr == 0) 1165 filter4->ff_mask.mask_dst.s_addr = 0xffffffff; 1166 if (filter4->ff_flow.fi_src.s_addr == 0) 1167 filter4->ff_mask.mask_src.s_addr = 0; 1168 else if (filter4->ff_mask.mask_src.s_addr == 0) 1169 filter4->ff_mask.mask_src.s_addr = 0xffffffff; 1170 1171 /* clear extra bits in addresses */ 1172 filter4->ff_flow.fi_dst.s_addr &= 1173 filter4->ff_mask.mask_dst.s_addr; 1174 filter4->ff_flow.fi_src.s_addr &= 1175 filter4->ff_mask.mask_src.s_addr; 1176 1177 /* 1178 * if dst address is a wildcard, use hash-entry 1179 * ACC_WILDCARD_INDEX. 1180 */ 1181 if (filter4->ff_mask.mask_dst.s_addr != 0xffffffff) 1182 i = ACC_WILDCARD_INDEX; 1183 else 1184 i = ACC_GET_HASH_INDEX(filter4->ff_flow.fi_dst.s_addr); 1185 } 1186 #ifdef INET6 1187 else if (filter->ff_flow.fi_family == AF_INET6) { 1188 struct flow_filter6 *filter6 = 1189 (struct flow_filter6 *)&afp->f_filter; 1190 #ifndef IN6MASK0 /* taken from kame ipv6 */ 1191 #define IN6MASK0 {{{ 0, 0, 0, 0 }}} 1192 #define IN6MASK128 {{{ 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }}} 1193 const struct in6_addr in6mask0 = IN6MASK0; 1194 const struct in6_addr in6mask128 = IN6MASK128; 1195 #endif 1196 1197 if (IN6_IS_ADDR_UNSPECIFIED(&filter6->ff_flow6.fi6_dst)) 1198 filter6->ff_mask6.mask6_dst = in6mask0; 1199 else if (IN6_IS_ADDR_UNSPECIFIED(&filter6->ff_mask6.mask6_dst)) 1200 filter6->ff_mask6.mask6_dst = in6mask128; 1201 if (IN6_IS_ADDR_UNSPECIFIED(&filter6->ff_flow6.fi6_src)) 1202 filter6->ff_mask6.mask6_src = in6mask0; 1203 else if (IN6_IS_ADDR_UNSPECIFIED(&filter6->ff_mask6.mask6_src)) 1204 filter6->ff_mask6.mask6_src = in6mask128; 1205 1206 /* clear extra bits in addresses */ 1207 for (i = 0; i < 16; i++) 1208 filter6->ff_flow6.fi6_dst.s6_addr[i] &= 1209 filter6->ff_mask6.mask6_dst.s6_addr[i]; 1210 for (i = 0; i < 16; i++) 1211 filter6->ff_flow6.fi6_src.s6_addr[i] &= 1212 filter6->ff_mask6.mask6_src.s6_addr[i]; 1213 1214 if (filter6->ff_flow6.fi6_flowlabel == 0) 1215 i = ACC_WILDCARD_INDEX; 1216 else 1217 i = ACC_GET_HASH_INDEX(filter6->ff_flow6.fi6_flowlabel); 1218 } 1219 #endif /* INET6 */ 1220 1221 afp->f_handle = get_filt_handle(classifier, i); 1222 1223 /* update filter bitmask */ 1224 afp->f_fbmask = filt2fibmask(filter); 1225 classifier->acc_fbmask |= afp->f_fbmask; 1226 1227 /* 1228 * add this filter to the filter list. 1229 * filters are ordered from the highest rule number. 1230 */ 1231 s = splnet(); 1232 prev = NULL; 1233 LIST_FOREACH(tmp, &classifier->acc_filters[i], f_chain) { 1234 if (tmp->f_filter.ff_ruleno > afp->f_filter.ff_ruleno) 1235 prev = tmp; 1236 else 1237 break; 1238 } 1239 if (prev == NULL) 1240 LIST_INSERT_HEAD(&classifier->acc_filters[i], afp, f_chain); 1241 else 1242 LIST_INSERT_AFTER(prev, afp, f_chain); 1243 splx(s); 1244 1245 *phandle = afp->f_handle; 1246 return (0); 1247 } 1248 1249 int 1250 acc_delete_filter(struct acc_classifier *classifier, u_long handle) 1251 { 1252 struct acc_filter *afp; 1253 int s; 1254 1255 if ((afp = filth_to_filtp(classifier, handle)) == NULL) 1256 return (EINVAL); 1257 1258 s = splnet(); 1259 LIST_REMOVE(afp, f_chain); 1260 splx(s); 1261 1262 free(afp, M_DEVBUF); 1263 1264 /* todo: update filt_bmask */ 1265 1266 return (0); 1267 } 1268 1269 /* 1270 * delete filters referencing to the specified class. 1271 * if the all flag is not 0, delete all the filters. 1272 */ 1273 int 1274 acc_discard_filters(struct acc_classifier *classifier, void *class, int all) 1275 { 1276 struct acc_filter *afp; 1277 int i, s; 1278 1279 s = splnet(); 1280 for (i = 0; i < ACC_FILTER_TABLESIZE; i++) { 1281 do { 1282 LIST_FOREACH(afp, &classifier->acc_filters[i], f_chain) 1283 if (all || afp->f_class == class) { 1284 LIST_REMOVE(afp, f_chain); 1285 free(afp, M_DEVBUF); 1286 /* start again from the head */ 1287 break; 1288 } 1289 } while (afp != NULL); 1290 } 1291 splx(s); 1292 1293 if (all) 1294 classifier->acc_fbmask = 0; 1295 1296 return (0); 1297 } 1298 1299 void * 1300 acc_classify(void *clfier, struct mbuf *m, int af) 1301 { 1302 struct acc_classifier *classifier; 1303 struct flowinfo flow; 1304 struct acc_filter *afp; 1305 int i; 1306 1307 classifier = (struct acc_classifier *)clfier; 1308 altq_extractflow(m, af, &flow, classifier->acc_fbmask); 1309 1310 if (flow.fi_family == AF_INET) { 1311 struct flowinfo_in *fp = (struct flowinfo_in *)&flow; 1312 1313 if ((classifier->acc_fbmask & FIMB4_ALL) == FIMB4_TOS) { 1314 /* only tos is used */ 1315 LIST_FOREACH(afp, 1316 &classifier->acc_filters[ACC_WILDCARD_INDEX], 1317 f_chain) 1318 if (apply_tosfilter4(afp->f_fbmask, 1319 &afp->f_filter, fp)) 1320 /* filter matched */ 1321 return (afp->f_class); 1322 } else if ((classifier->acc_fbmask & 1323 (~(FIMB4_PROTO|FIMB4_SPORT|FIMB4_DPORT) & FIMB4_ALL)) 1324 == 0) { 1325 /* only proto and ports are used */ 1326 LIST_FOREACH(afp, 1327 &classifier->acc_filters[ACC_WILDCARD_INDEX], 1328 f_chain) 1329 if (apply_ppfilter4(afp->f_fbmask, 1330 &afp->f_filter, fp)) 1331 /* filter matched */ 1332 return (afp->f_class); 1333 } else { 1334 /* get the filter hash entry from its dest address */ 1335 i = ACC_GET_HASH_INDEX(fp->fi_dst.s_addr); 1336 do { 1337 /* 1338 * go through this loop twice. first for dst 1339 * hash, second for wildcards. 1340 */ 1341 LIST_FOREACH(afp, &classifier->acc_filters[i], 1342 f_chain) 1343 if (apply_filter4(afp->f_fbmask, 1344 &afp->f_filter, fp)) 1345 /* filter matched */ 1346 return (afp->f_class); 1347 1348 /* 1349 * check again for filters with a dst addr 1350 * wildcard. 1351 * (daddr == 0 || dmask != 0xffffffff). 1352 */ 1353 if (i != ACC_WILDCARD_INDEX) 1354 i = ACC_WILDCARD_INDEX; 1355 else 1356 break; 1357 } while (1); 1358 } 1359 } 1360 #ifdef INET6 1361 else if (flow.fi_family == AF_INET6) { 1362 struct flowinfo_in6 *fp6 = (struct flowinfo_in6 *)&flow; 1363 1364 /* get the filter hash entry from its flow ID */ 1365 if (fp6->fi6_flowlabel != 0) 1366 i = ACC_GET_HASH_INDEX(fp6->fi6_flowlabel); 1367 else 1368 /* flowlable can be zero */ 1369 i = ACC_WILDCARD_INDEX; 1370 1371 /* go through this loop twice. first for flow hash, second 1372 for wildcards. */ 1373 do { 1374 LIST_FOREACH(afp, &classifier->acc_filters[i], f_chain) 1375 if (apply_filter6(afp->f_fbmask, 1376 (struct flow_filter6 *)&afp->f_filter, 1377 fp6)) 1378 /* filter matched */ 1379 return (afp->f_class); 1380 1381 /* 1382 * check again for filters with a wildcard. 1383 */ 1384 if (i != ACC_WILDCARD_INDEX) 1385 i = ACC_WILDCARD_INDEX; 1386 else 1387 break; 1388 } while (1); 1389 } 1390 #endif /* INET6 */ 1391 1392 /* no filter matched */ 1393 return (NULL); 1394 } 1395 1396 static int 1397 apply_filter4(u_int32_t fbmask, struct flow_filter *filt, 1398 struct flowinfo_in *pkt) 1399 { 1400 if (filt->ff_flow.fi_family != AF_INET) 1401 return (0); 1402 if ((fbmask & FIMB4_SPORT) && filt->ff_flow.fi_sport != pkt->fi_sport) 1403 return (0); 1404 if ((fbmask & FIMB4_DPORT) && filt->ff_flow.fi_dport != pkt->fi_dport) 1405 return (0); 1406 if ((fbmask & FIMB4_DADDR) && 1407 filt->ff_flow.fi_dst.s_addr != 1408 (pkt->fi_dst.s_addr & filt->ff_mask.mask_dst.s_addr)) 1409 return (0); 1410 if ((fbmask & FIMB4_SADDR) && 1411 filt->ff_flow.fi_src.s_addr != 1412 (pkt->fi_src.s_addr & filt->ff_mask.mask_src.s_addr)) 1413 return (0); 1414 if ((fbmask & FIMB4_PROTO) && filt->ff_flow.fi_proto != pkt->fi_proto) 1415 return (0); 1416 if ((fbmask & FIMB4_TOS) && filt->ff_flow.fi_tos != 1417 (pkt->fi_tos & filt->ff_mask.mask_tos)) 1418 return (0); 1419 if ((fbmask & FIMB4_GPI) && filt->ff_flow.fi_gpi != (pkt->fi_gpi)) 1420 return (0); 1421 /* match */ 1422 return (1); 1423 } 1424 1425 /* 1426 * filter matching function optimized for a common case that checks 1427 * only protocol and port numbers 1428 */ 1429 static int 1430 apply_ppfilter4(u_int32_t fbmask, struct flow_filter *filt, 1431 struct flowinfo_in *pkt) 1432 { 1433 if (filt->ff_flow.fi_family != AF_INET) 1434 return (0); 1435 if ((fbmask & FIMB4_SPORT) && filt->ff_flow.fi_sport != pkt->fi_sport) 1436 return (0); 1437 if ((fbmask & FIMB4_DPORT) && filt->ff_flow.fi_dport != pkt->fi_dport) 1438 return (0); 1439 if ((fbmask & FIMB4_PROTO) && filt->ff_flow.fi_proto != pkt->fi_proto) 1440 return (0); 1441 /* match */ 1442 return (1); 1443 } 1444 1445 /* 1446 * filter matching function only for tos field. 1447 */ 1448 static int 1449 apply_tosfilter4(u_int32_t fbmask, struct flow_filter *filt, 1450 struct flowinfo_in *pkt) 1451 { 1452 if (filt->ff_flow.fi_family != AF_INET) 1453 return (0); 1454 if ((fbmask & FIMB4_TOS) && filt->ff_flow.fi_tos != 1455 (pkt->fi_tos & filt->ff_mask.mask_tos)) 1456 return (0); 1457 /* match */ 1458 return (1); 1459 } 1460 1461 #ifdef INET6 1462 static int 1463 apply_filter6(u_int32_t fbmask, struct flow_filter6 *filt, 1464 struct flowinfo_in6 *pkt) 1465 { 1466 int i; 1467 1468 if (filt->ff_flow6.fi6_family != AF_INET6) 1469 return (0); 1470 if ((fbmask & FIMB6_FLABEL) && 1471 filt->ff_flow6.fi6_flowlabel != pkt->fi6_flowlabel) 1472 return (0); 1473 if ((fbmask & FIMB6_PROTO) && 1474 filt->ff_flow6.fi6_proto != pkt->fi6_proto) 1475 return (0); 1476 if ((fbmask & FIMB6_SPORT) && 1477 filt->ff_flow6.fi6_sport != pkt->fi6_sport) 1478 return (0); 1479 if ((fbmask & FIMB6_DPORT) && 1480 filt->ff_flow6.fi6_dport != pkt->fi6_dport) 1481 return (0); 1482 if (fbmask & FIMB6_SADDR) { 1483 for (i = 0; i < 4; i++) 1484 if (filt->ff_flow6.fi6_src.s6_addr32[i] != 1485 (pkt->fi6_src.s6_addr32[i] & 1486 filt->ff_mask6.mask6_src.s6_addr32[i])) 1487 return (0); 1488 } 1489 if (fbmask & FIMB6_DADDR) { 1490 for (i = 0; i < 4; i++) 1491 if (filt->ff_flow6.fi6_dst.s6_addr32[i] != 1492 (pkt->fi6_dst.s6_addr32[i] & 1493 filt->ff_mask6.mask6_dst.s6_addr32[i])) 1494 return (0); 1495 } 1496 if ((fbmask & FIMB6_TCLASS) && 1497 filt->ff_flow6.fi6_tclass != 1498 (pkt->fi6_tclass & filt->ff_mask6.mask6_tclass)) 1499 return (0); 1500 if ((fbmask & FIMB6_GPI) && 1501 filt->ff_flow6.fi6_gpi != pkt->fi6_gpi) 1502 return (0); 1503 /* match */ 1504 return (1); 1505 } 1506 #endif /* INET6 */ 1507 1508 /* 1509 * filter handle: 1510 * bit 20-28: index to the filter hash table 1511 * bit 0-19: unique id in the hash bucket. 1512 */ 1513 static u_long 1514 get_filt_handle(struct acc_classifier *classifier, int i) 1515 { 1516 static u_long handle_number = 1; 1517 u_long handle; 1518 struct acc_filter *afp; 1519 1520 while (1) { 1521 handle = handle_number++ & 0x000fffff; 1522 1523 if (LIST_EMPTY(&classifier->acc_filters[i])) 1524 break; 1525 1526 LIST_FOREACH(afp, &classifier->acc_filters[i], f_chain) 1527 if ((afp->f_handle & 0x000fffff) == handle) 1528 break; 1529 if (afp == NULL) 1530 break; 1531 /* this handle is already used, try again */ 1532 } 1533 1534 return ((i << 20) | handle); 1535 } 1536 1537 /* convert filter handle to filter pointer */ 1538 static struct acc_filter * 1539 filth_to_filtp(struct acc_classifier *classifier, u_long handle) 1540 { 1541 struct acc_filter *afp; 1542 int i; 1543 1544 i = ACC_GET_HINDEX(handle); 1545 1546 LIST_FOREACH(afp, &classifier->acc_filters[i], f_chain) 1547 if (afp->f_handle == handle) 1548 return (afp); 1549 1550 return (NULL); 1551 } 1552 1553 /* create flowinfo bitmask */ 1554 static u_int32_t 1555 filt2fibmask(struct flow_filter *filt) 1556 { 1557 u_int32_t mask = 0; 1558 #ifdef INET6 1559 struct flow_filter6 *filt6; 1560 #endif 1561 1562 switch (filt->ff_flow.fi_family) { 1563 case AF_INET: 1564 if (filt->ff_flow.fi_proto != 0) 1565 mask |= FIMB4_PROTO; 1566 if (filt->ff_flow.fi_tos != 0) 1567 mask |= FIMB4_TOS; 1568 if (filt->ff_flow.fi_dst.s_addr != 0) 1569 mask |= FIMB4_DADDR; 1570 if (filt->ff_flow.fi_src.s_addr != 0) 1571 mask |= FIMB4_SADDR; 1572 if (filt->ff_flow.fi_sport != 0) 1573 mask |= FIMB4_SPORT; 1574 if (filt->ff_flow.fi_dport != 0) 1575 mask |= FIMB4_DPORT; 1576 if (filt->ff_flow.fi_gpi != 0) 1577 mask |= FIMB4_GPI; 1578 break; 1579 #ifdef INET6 1580 case AF_INET6: 1581 filt6 = (struct flow_filter6 *)filt; 1582 1583 if (filt6->ff_flow6.fi6_proto != 0) 1584 mask |= FIMB6_PROTO; 1585 if (filt6->ff_flow6.fi6_tclass != 0) 1586 mask |= FIMB6_TCLASS; 1587 if (!IN6_IS_ADDR_UNSPECIFIED(&filt6->ff_flow6.fi6_dst)) 1588 mask |= FIMB6_DADDR; 1589 if (!IN6_IS_ADDR_UNSPECIFIED(&filt6->ff_flow6.fi6_src)) 1590 mask |= FIMB6_SADDR; 1591 if (filt6->ff_flow6.fi6_sport != 0) 1592 mask |= FIMB6_SPORT; 1593 if (filt6->ff_flow6.fi6_dport != 0) 1594 mask |= FIMB6_DPORT; 1595 if (filt6->ff_flow6.fi6_gpi != 0) 1596 mask |= FIMB6_GPI; 1597 if (filt6->ff_flow6.fi6_flowlabel != 0) 1598 mask |= FIMB6_FLABEL; 1599 break; 1600 #endif /* INET6 */ 1601 } 1602 return (mask); 1603 } 1604 1605 1606 /* 1607 * helper functions to handle IPv4 fragments. 1608 * currently only in-sequence fragments are handled. 1609 * - fragment info is cached in a LRU list. 1610 * - when a first fragment is found, cache its flow info. 1611 * - when a non-first fragment is found, lookup the cache. 1612 */ 1613 1614 struct ip4_frag { 1615 TAILQ_ENTRY(ip4_frag) ip4f_chain; 1616 char ip4f_valid; 1617 u_short ip4f_id; 1618 struct flowinfo_in ip4f_info; 1619 }; 1620 1621 static TAILQ_HEAD(ip4f_list, ip4_frag) ip4f_list; /* IPv4 fragment cache */ 1622 1623 #define IP4F_TABSIZE 16 /* IPv4 fragment cache size */ 1624 1625 1626 static void 1627 ip4f_cache(struct ip *ip, struct flowinfo_in *fin) 1628 { 1629 struct ip4_frag *fp; 1630 1631 if (TAILQ_EMPTY(&ip4f_list)) { 1632 /* first time call, allocate fragment cache entries. */ 1633 if (ip4f_init() < 0) 1634 /* allocation failed! */ 1635 return; 1636 } 1637 1638 fp = ip4f_alloc(); 1639 fp->ip4f_id = ip->ip_id; 1640 fp->ip4f_info.fi_proto = ip->ip_p; 1641 fp->ip4f_info.fi_src.s_addr = ip->ip_src.s_addr; 1642 fp->ip4f_info.fi_dst.s_addr = ip->ip_dst.s_addr; 1643 1644 /* save port numbers */ 1645 fp->ip4f_info.fi_sport = fin->fi_sport; 1646 fp->ip4f_info.fi_dport = fin->fi_dport; 1647 fp->ip4f_info.fi_gpi = fin->fi_gpi; 1648 } 1649 1650 static int 1651 ip4f_lookup(struct ip *ip, struct flowinfo_in *fin) 1652 { 1653 struct ip4_frag *fp; 1654 1655 for (fp = TAILQ_FIRST(&ip4f_list); fp != NULL && fp->ip4f_valid; 1656 fp = TAILQ_NEXT(fp, ip4f_chain)) 1657 if (ip->ip_id == fp->ip4f_id && 1658 ip->ip_src.s_addr == fp->ip4f_info.fi_src.s_addr && 1659 ip->ip_dst.s_addr == fp->ip4f_info.fi_dst.s_addr && 1660 ip->ip_p == fp->ip4f_info.fi_proto) { 1661 1662 /* found the matching entry */ 1663 fin->fi_sport = fp->ip4f_info.fi_sport; 1664 fin->fi_dport = fp->ip4f_info.fi_dport; 1665 fin->fi_gpi = fp->ip4f_info.fi_gpi; 1666 1667 if ((ntohs(ip->ip_off) & IP_MF) == 0) 1668 /* this is the last fragment, 1669 release the entry. */ 1670 ip4f_free(fp); 1671 1672 return (1); 1673 } 1674 1675 /* no matching entry found */ 1676 return (0); 1677 } 1678 1679 static int 1680 ip4f_init(void) 1681 { 1682 struct ip4_frag *fp; 1683 int i; 1684 1685 TAILQ_INIT(&ip4f_list); 1686 for (i=0; i<IP4F_TABSIZE; i++) { 1687 fp = malloc(sizeof(struct ip4_frag), M_DEVBUF, M_NOWAIT); 1688 if (fp == NULL) { 1689 printf("ip4f_init: can't alloc %dth entry!\n", i); 1690 if (i == 0) 1691 return (-1); 1692 return (0); 1693 } 1694 fp->ip4f_valid = 0; 1695 TAILQ_INSERT_TAIL(&ip4f_list, fp, ip4f_chain); 1696 } 1697 return (0); 1698 } 1699 1700 static struct ip4_frag * 1701 ip4f_alloc(void) 1702 { 1703 struct ip4_frag *fp; 1704 1705 /* reclaim an entry at the tail, put it at the head */ 1706 fp = TAILQ_LAST(&ip4f_list, ip4f_list); 1707 TAILQ_REMOVE(&ip4f_list, fp, ip4f_chain); 1708 fp->ip4f_valid = 1; 1709 TAILQ_INSERT_HEAD(&ip4f_list, fp, ip4f_chain); 1710 return (fp); 1711 } 1712 1713 static void 1714 ip4f_free(struct ip4_frag *fp) 1715 { 1716 TAILQ_REMOVE(&ip4f_list, fp, ip4f_chain); 1717 fp->ip4f_valid = 0; 1718 TAILQ_INSERT_TAIL(&ip4f_list, fp, ip4f_chain); 1719 } 1720 1721 #endif /* ALTQ3_CLFIER_COMPAT */ 1722
Indexes created Fri Oct 17 17:09:57 GMT 2025