altq_subr.c revision 1.4.2.3
1 /* $NetBSD: altq_subr.c,v 1.4.2.3 2001/04/21 17:46:12 bouyer Exp $ */ 2 /* $KAME: altq_subr.c,v 1.8 2000/12/14 08:12:46 thorpej Exp $ */ 3 4 /* 5 * Copyright (C) 1997-2000 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 #if defined(__FreeBSD__) || defined(__NetBSD__) 31 #include "opt_altq.h" 32 #if (__FreeBSD__ != 2) 33 #include "opt_inet.h" 34 #ifdef __FreeBSD__ 35 #include "opt_inet6.h" 36 #endif 37 #endif 38 #endif /* __FreeBSD__ || __NetBSD__ */ 39 40 #include <sys/param.h> 41 #include <sys/malloc.h> 42 #include <sys/mbuf.h> 43 #include <sys/systm.h> 44 #include <sys/proc.h> 45 #include <sys/socket.h> 46 #include <sys/socketvar.h> 47 #include <sys/kernel.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 #include <altq/altq.h> 67 #include <altq/altq_conf.h> 68 69 #ifdef __FreeBSD__ 70 #include "opt_cpu.h" /* for FreeBSD-2.2.8 to get i586_ctr_freq */ 71 #include <machine/clock.h> 72 #endif 73 74 /* 75 * internal function prototypes 76 */ 77 static void tbr_timeout __P((void *)); 78 static int extract_ports4 __P((struct mbuf *, struct ip *, 79 struct flowinfo_in *)); 80 #ifdef INET6 81 static int extract_ports6 __P((struct mbuf *, struct ip6_hdr *, 82 struct flowinfo_in6 *)); 83 #endif 84 static int apply_filter4 __P((u_int32_t, struct flow_filter *, 85 struct flowinfo_in *)); 86 static int apply_ppfilter4 __P((u_int32_t, struct flow_filter *, 87 struct flowinfo_in *)); 88 #ifdef INET6 89 static int apply_filter6 __P((u_int32_t, struct flow_filter6 *, 90 struct flowinfo_in6 *)); 91 #endif 92 static int apply_tosfilter4 __P((u_int32_t, struct flow_filter *, 93 struct flowinfo_in *)); 94 static u_long get_filt_handle __P((struct acc_classifier *, int)); 95 static struct acc_filter *filth_to_filtp __P((struct acc_classifier *, 96 u_long)); 97 static u_int32_t filt2fibmask __P((struct flow_filter *)); 98 99 static void ip4f_cache __P((struct ip *, struct flowinfo_in *)); 100 static int ip4f_lookup __P((struct ip *, struct flowinfo_in *)); 101 static int ip4f_init __P((void)); 102 static struct ip4_frag *ip4f_alloc __P((void)); 103 static void ip4f_free __P((struct ip4_frag *)); 104 105 int (*altq_input) __P((struct mbuf *, int)) = NULL; 106 static int tbr_timer = 0; /* token bucket regulator timer */ 107 static struct callout tbr_callout = CALLOUT_INITIALIZER; 108 109 /* 110 * alternate queueing support routines 111 */ 112 113 /* look up the queue state by the interface name and the queuing type. */ 114 void * 115 altq_lookup(name, type) 116 char *name; 117 int type; 118 { 119 struct ifnet *ifp; 120 121 if ((ifp = ifunit(name)) != NULL) { 122 if (type != ALTQT_NONE && ifp->if_snd.altq_type == type) 123 return (ifp->if_snd.altq_disc); 124 } 125 126 return NULL; 127 } 128 129 int 130 altq_attach(ifq, type, discipline, enqueue, dequeue, request, clfier, classify) 131 struct ifaltq *ifq; 132 int type; 133 void *discipline; 134 int (*enqueue)(struct ifaltq *, struct mbuf *, struct altq_pktattr *); 135 struct mbuf *(*dequeue)(struct ifaltq *, int); 136 int (*request)(struct ifaltq *, int, void *); 137 void *clfier; 138 void *(*classify)(void *, struct mbuf *, int); 139 { 140 if (!ALTQ_IS_READY(ifq)) 141 return ENXIO; 142 if (ALTQ_IS_ENABLED(ifq)) 143 return EBUSY; 144 if (ALTQ_IS_ATTACHED(ifq)) 145 return EEXIST; 146 ifq->altq_type = type; 147 ifq->altq_disc = discipline; 148 ifq->altq_enqueue = enqueue; 149 ifq->altq_dequeue = dequeue; 150 ifq->altq_request = request; 151 ifq->altq_clfier = clfier; 152 ifq->altq_classify = classify; 153 ifq->altq_flags &= ALTQF_CANTCHANGE; 154 #ifdef ALTQ_KLD 155 altq_module_incref(type); 156 #endif 157 return 0; 158 } 159 160 int 161 altq_detach(ifq) 162 struct ifaltq *ifq; 163 { 164 if (!ALTQ_IS_READY(ifq)) 165 return ENXIO; 166 if (ALTQ_IS_ENABLED(ifq)) 167 return EBUSY; 168 if (!ALTQ_IS_ATTACHED(ifq)) 169 return (0); 170 171 #ifdef ALTQ_KLD 172 altq_module_declref(ifq->altq_type); 173 #endif 174 ifq->altq_type = ALTQT_NONE; 175 ifq->altq_disc = NULL; 176 ifq->altq_enqueue = NULL; 177 ifq->altq_dequeue = NULL; 178 ifq->altq_request = NULL; 179 ifq->altq_clfier = NULL; 180 ifq->altq_classify = NULL; 181 ifq->altq_flags &= ALTQF_CANTCHANGE; 182 return 0; 183 } 184 185 int 186 altq_enable(ifq) 187 struct ifaltq *ifq; 188 { 189 int s; 190 191 if (!ALTQ_IS_READY(ifq)) 192 return ENXIO; 193 if (ALTQ_IS_ENABLED(ifq)) 194 return 0; 195 196 s = splnet(); 197 IFQ_PURGE(ifq); 198 ASSERT(ifq->ifq_len == 0); 199 ifq->altq_flags |= ALTQF_ENABLED; 200 if (ifq->altq_clfier != NULL) 201 ifq->altq_flags |= ALTQF_CLASSIFY; 202 splx(s); 203 204 return 0; 205 } 206 207 int 208 altq_disable(ifq) 209 struct ifaltq *ifq; 210 { 211 int s; 212 213 if (!ALTQ_IS_ENABLED(ifq)) 214 return 0; 215 216 s = splnet(); 217 IFQ_PURGE(ifq); 218 ASSERT(ifq->ifq_len == 0); 219 ifq->altq_flags &= ~(ALTQF_ENABLED|ALTQF_CLASSIFY); 220 splx(s); 221 return 0; 222 } 223 224 void 225 altq_assert(file, line, failedexpr) 226 const char *file, *failedexpr; 227 int line; 228 { 229 (void)printf("altq assertion \"%s\" failed: file \"%s\", line %d\n", 230 failedexpr, file, line); 231 panic("altq assertion"); 232 /* NOTREACHED */ 233 } 234 235 /* 236 * internal representation of token bucket parameters 237 * rate: byte_per_unittime << 32 238 * (((bits_per_sec) / 8) << 32) / machclk_freq 239 * depth: byte << 32 240 * 241 */ 242 #define TBR_SHIFT 32 243 #define TBR_SCALE(x) ((int64_t)(x) << TBR_SHIFT) 244 #define TBR_UNSCALE(x) ((x) >> TBR_SHIFT) 245 246 struct mbuf * 247 tbr_dequeue(ifq, op) 248 struct ifaltq *ifq; 249 int op; 250 { 251 struct tb_regulator *tbr; 252 struct mbuf *m; 253 int64_t interval; 254 u_int64_t now; 255 256 tbr = ifq->altq_tbr; 257 if (op == ALTDQ_REMOVE && tbr->tbr_lastop == ALTDQ_POLL) { 258 /* if this is a remove after poll, bypass tbr check */ 259 } else { 260 /* update token only when it is negative */ 261 if (tbr->tbr_token <= 0) { 262 now = read_machclk(); 263 interval = now - tbr->tbr_last; 264 if (interval >= tbr->tbr_filluptime) 265 tbr->tbr_token = tbr->tbr_depth; 266 else { 267 tbr->tbr_token += interval * tbr->tbr_rate; 268 if (tbr->tbr_token > tbr->tbr_depth) 269 tbr->tbr_token = tbr->tbr_depth; 270 } 271 tbr->tbr_last = now; 272 } 273 /* if token is still negative, don't allow dequeue */ 274 if (tbr->tbr_token <= 0) 275 return (NULL); 276 } 277 278 if (ALTQ_IS_ENABLED(ifq)) 279 m = (*ifq->altq_dequeue)(ifq, op); 280 else { 281 if (op == ALTDQ_POLL) 282 IF_POLL(ifq, m); 283 else 284 IF_DEQUEUE(ifq, m); 285 } 286 287 if (m != NULL && op == ALTDQ_REMOVE) 288 tbr->tbr_token -= TBR_SCALE(m_pktlen(m)); 289 tbr->tbr_lastop = op; 290 return (m); 291 } 292 293 /* 294 * set a token bucket regulator. 295 * if the specified rate is zero, the token bucket regulator is deleted. 296 */ 297 int 298 tbr_set(ifq, profile) 299 struct ifaltq *ifq; 300 struct tb_profile *profile; 301 { 302 struct tb_regulator *tbr, *otbr; 303 304 if (machclk_freq == 0) 305 init_machclk(); 306 if (machclk_freq == 0) { 307 printf("tbr_set: no cpu clock available!\n"); 308 return (ENXIO); 309 } 310 311 if (profile->rate == 0) { 312 /* delete this tbr */ 313 if ((tbr = ifq->altq_tbr) == NULL) 314 return (ENOENT); 315 ifq->altq_tbr = NULL; 316 FREE(tbr, M_DEVBUF); 317 return (0); 318 } 319 320 MALLOC(tbr, struct tb_regulator *, sizeof(struct tb_regulator), 321 M_DEVBUF, M_WAITOK); 322 if (tbr == NULL) 323 return (ENOMEM); 324 bzero(tbr, sizeof(struct tb_regulator)); 325 326 tbr->tbr_rate = TBR_SCALE(profile->rate / 8) / machclk_freq; 327 tbr->tbr_depth = TBR_SCALE(profile->depth); 328 if (tbr->tbr_rate > 0) 329 tbr->tbr_filluptime = tbr->tbr_depth / tbr->tbr_rate; 330 else 331 tbr->tbr_filluptime = 0xffffffffffffffffLL; 332 tbr->tbr_token = tbr->tbr_depth; 333 tbr->tbr_last = read_machclk(); 334 tbr->tbr_lastop = ALTDQ_REMOVE; 335 336 otbr = ifq->altq_tbr; 337 ifq->altq_tbr = tbr; /* set the new tbr */ 338 339 if (otbr != NULL) 340 FREE(otbr, M_DEVBUF); 341 else { 342 if (tbr_timer == 0) { 343 CALLOUT_RESET(&tbr_callout, 1, tbr_timeout, (void *)0); 344 tbr_timer = 1; 345 } 346 } 347 return (0); 348 } 349 350 /* 351 * tbr_timeout goes through the interface list, and kicks the drivers 352 * if necessary. 353 */ 354 static void 355 tbr_timeout(arg) 356 void *arg; 357 { 358 struct ifnet *ifp; 359 int active, s; 360 361 active = 0; 362 s = splnet(); 363 #ifdef __FreeBSD__ 364 #if (__FreeBSD_version < 300000) 365 for (ifp = ifnet; ifp; ifp = ifp->if_next) 366 #else 367 for (ifp = ifnet.tqh_first; ifp != NULL; ifp = ifp->if_link.tqe_next) 368 #endif 369 #else /* !FreeBSD */ 370 for (ifp = ifnet.tqh_first; ifp != NULL; ifp = ifp->if_list.tqe_next) 371 #endif 372 { 373 if (!TBR_IS_ENABLED(&ifp->if_snd)) 374 continue; 375 active++; 376 if (!IFQ_IS_EMPTY(&ifp->if_snd) && ifp->if_start != NULL) 377 (*ifp->if_start)(ifp); 378 } 379 splx(s); 380 if (active > 0) 381 CALLOUT_RESET(&tbr_callout, 1, tbr_timeout, (void *)0); 382 else 383 tbr_timer = 0; /* don't need tbr_timer anymore */ 384 #if defined(__alpha__) && !defined(ALTQ_NOPCC) 385 { 386 /* 387 * XXX read out the machine dependent clock once a second 388 * to detect counter wrap-around. 389 */ 390 static u_int cnt; 391 392 if (++cnt >= hz) { 393 (void)read_machclk(); 394 cnt = 0; 395 } 396 } 397 #endif /* __alpha__ && !ALTQ_NOPCC */ 398 } 399 400 /* 401 * get token bucket regulator profile 402 */ 403 int 404 tbr_get(ifq, profile) 405 struct ifaltq *ifq; 406 struct tb_profile *profile; 407 { 408 struct tb_regulator *tbr; 409 410 if ((tbr = ifq->altq_tbr) == NULL) { 411 profile->rate = 0; 412 profile->depth = 0; 413 } else { 414 profile->rate = 415 (u_int)TBR_UNSCALE(tbr->tbr_rate * 8 * machclk_freq); 416 profile->depth = (u_int)TBR_UNSCALE(tbr->tbr_depth); 417 } 418 return (0); 419 } 420 421 422 #ifndef IPPROTO_ESP 423 #define IPPROTO_ESP 50 /* encapsulating security payload */ 424 #endif 425 #ifndef IPPROTO_AH 426 #define IPPROTO_AH 51 /* authentication header */ 427 #endif 428 429 /* 430 * extract flow information from a given packet. 431 * filt_mask shows flowinfo fields required. 432 * we assume the ip header is in one mbuf, and addresses and ports are 433 * in network byte order. 434 */ 435 int 436 altq_extractflow(m, af, flow, filt_bmask) 437 struct mbuf *m; 438 int af; 439 struct flowinfo *flow; 440 u_int32_t filt_bmask; 441 { 442 443 switch (af) { 444 case PF_INET: { 445 struct flowinfo_in *fin; 446 struct ip *ip; 447 448 ip = mtod(m, struct ip *); 449 450 if (ip->ip_v != 4) 451 break; 452 453 fin = (struct flowinfo_in *)flow; 454 fin->fi_len = sizeof(struct flowinfo_in); 455 fin->fi_family = AF_INET; 456 457 fin->fi_proto = ip->ip_p; 458 fin->fi_tos = ip->ip_tos; 459 460 fin->fi_src.s_addr = ip->ip_src.s_addr; 461 fin->fi_dst.s_addr = ip->ip_dst.s_addr; 462 463 if (filt_bmask & FIMB4_PORTS) 464 /* if port info is required, extract port numbers */ 465 extract_ports4(m, ip, fin); 466 else { 467 fin->fi_sport = 0; 468 fin->fi_dport = 0; 469 fin->fi_gpi = 0; 470 } 471 return (1); 472 } 473 474 #ifdef INET6 475 case PF_INET6: { 476 struct flowinfo_in6 *fin6; 477 struct ip6_hdr *ip6; 478 479 ip6 = mtod(m, struct ip6_hdr *); 480 /* should we check the ip version? */ 481 482 fin6 = (struct flowinfo_in6 *)flow; 483 fin6->fi6_len = sizeof(struct flowinfo_in6); 484 fin6->fi6_family = AF_INET6; 485 486 fin6->fi6_proto = ip6->ip6_nxt; 487 fin6->fi6_tclass = (ntohl(ip6->ip6_flow) >> 20) & 0xff; 488 489 fin6->fi6_flowlabel = ip6->ip6_flow & htonl(0x000fffff); 490 fin6->fi6_src = ip6->ip6_src; 491 fin6->fi6_dst = ip6->ip6_dst; 492 493 if ((filt_bmask & FIMB6_PORTS) || 494 ((filt_bmask & FIMB6_PROTO) 495 && ip6->ip6_nxt > IPPROTO_IPV6)) 496 /* 497 * if port info is required, or proto is required 498 * but there are option headers, extract port 499 * and protocol numbers. 500 */ 501 extract_ports6(m, ip6, fin6); 502 else { 503 fin6->fi6_sport = 0; 504 fin6->fi6_dport = 0; 505 fin6->fi6_gpi = 0; 506 } 507 return (1); 508 } 509 #endif /* INET6 */ 510 511 default: 512 break; 513 } 514 515 /* failed */ 516 flow->fi_len = sizeof(struct flowinfo); 517 flow->fi_family = AF_UNSPEC; 518 return (0); 519 } 520 521 /* 522 * helper routine to extract port numbers 523 */ 524 /* structure for ipsec and ipv6 option header template */ 525 struct _opt6 { 526 u_int8_t opt6_nxt; /* next header */ 527 u_int8_t opt6_hlen; /* header extension length */ 528 u_int16_t _pad; 529 u_int32_t ah_spi; /* security parameter index 530 for authentication header */ 531 }; 532 533 /* 534 * extract port numbers from a ipv4 packet. 535 */ 536 static int 537 extract_ports4(m, ip, fin) 538 struct mbuf *m; 539 struct ip *ip; 540 struct flowinfo_in *fin; 541 { 542 struct mbuf *m0; 543 u_short ip_off; 544 u_int8_t proto; 545 int off; 546 547 fin->fi_sport = 0; 548 fin->fi_dport = 0; 549 fin->fi_gpi = 0; 550 551 ip_off = ntohs(ip->ip_off); 552 /* if it is a fragment, try cached fragment info */ 553 if (ip_off & IP_OFFMASK) { 554 ip4f_lookup(ip, fin); 555 return (1); 556 } 557 558 /* locate the mbuf containing the protocol header */ 559 for (m0 = m; m0 != NULL; m0 = m0->m_next) 560 if (((caddr_t)ip >= m0->m_data) && 561 ((caddr_t)ip < m0->m_data + m0->m_len)) 562 break; 563 if (m0 == NULL) { 564 #ifdef ALTQ_DEBUG 565 printf("extract_ports4: can't locate header! ip=%p\n", ip); 566 #endif 567 return (0); 568 } 569 off = ((caddr_t)ip - m0->m_data) + (ip->ip_hl << 2); 570 proto = ip->ip_p; 571 572 #ifdef ALTQ_IPSEC 573 again: 574 #endif 575 while (off >= m0->m_len) { 576 off -= m0->m_len; 577 m0 = m0->m_next; 578 } 579 ASSERT(m0->m_len >= off + 4); 580 581 switch (proto) { 582 case IPPROTO_TCP: 583 case IPPROTO_UDP: { 584 struct udphdr *udp; 585 586 udp = (struct udphdr *)(mtod(m0, caddr_t) + off); 587 fin->fi_sport = udp->uh_sport; 588 fin->fi_dport = udp->uh_dport; 589 fin->fi_proto = proto; 590 } 591 break; 592 593 #ifdef ALTQ_IPSEC 594 case IPPROTO_ESP: 595 if (fin->fi_gpi == 0){ 596 u_int32_t *gpi; 597 598 gpi = (u_int32_t *)(mtod(m0, caddr_t) + off); 599 fin->fi_gpi = *gpi; 600 } 601 fin->fi_proto = proto; 602 break; 603 604 case IPPROTO_AH: { 605 /* get next header and header length */ 606 struct _opt6 *opt6; 607 608 opt6 = (struct _opt6 *)(mtod(m0, caddr_t) + off); 609 proto = opt6->opt6_nxt; 610 off += 8 + (opt6->opt6_hlen * 4); 611 if (fin->fi_gpi == 0) 612 fin->fi_gpi = opt6->ah_spi; 613 } 614 /* goto the next header */ 615 goto again; 616 #endif /* ALTQ_IPSEC */ 617 618 default: 619 fin->fi_proto = proto; 620 return (0); 621 } 622 623 /* if this is a first fragment, cache it. */ 624 if (ip_off & IP_MF) 625 ip4f_cache(ip, fin); 626 627 return (1); 628 } 629 630 #ifdef INET6 631 static int 632 extract_ports6(m, ip6, fin6) 633 struct mbuf *m; 634 struct ip6_hdr *ip6; 635 struct flowinfo_in6 *fin6; 636 { 637 struct mbuf *m0; 638 int off; 639 u_int8_t proto; 640 641 fin6->fi6_gpi = 0; 642 fin6->fi6_sport = 0; 643 fin6->fi6_dport = 0; 644 645 /* locate the mbuf containing the protocol header */ 646 for (m0 = m; m0 != NULL; m0 = m0->m_next) 647 if (((caddr_t)ip6 >= m0->m_data) && 648 ((caddr_t)ip6 < m0->m_data + m0->m_len)) 649 break; 650 if (m0 == NULL) { 651 #ifdef ALTQ_DEBUG 652 printf("extract_ports6: can't locate header! ip6=%p\n", ip6); 653 #endif 654 return (0); 655 } 656 off = ((caddr_t)ip6 - m0->m_data) + sizeof(struct ip6_hdr); 657 658 proto = ip6->ip6_nxt; 659 do { 660 while (off >= m0->m_len) { 661 off -= m0->m_len; 662 m0 = m0->m_next; 663 } 664 ASSERT(m0->m_len >= off + 4); 665 666 switch (proto) { 667 case IPPROTO_TCP: 668 case IPPROTO_UDP: { 669 struct udphdr *udp; 670 671 udp = (struct udphdr *)(mtod(m0, caddr_t) + off); 672 fin6->fi6_sport = udp->uh_sport; 673 fin6->fi6_dport = udp->uh_dport; 674 fin6->fi6_proto = proto; 675 } 676 return (1); 677 678 case IPPROTO_ESP: 679 if (fin6->fi6_gpi == 0) { 680 u_int32_t *gpi; 681 682 gpi = (u_int32_t *)(mtod(m0, caddr_t) + off); 683 fin6->fi6_gpi = *gpi; 684 } 685 fin6->fi6_proto = proto; 686 return (1); 687 688 case IPPROTO_AH: { 689 /* get next header and header length */ 690 struct _opt6 *opt6; 691 692 opt6 = (struct _opt6 *)(mtod(m0, caddr_t) + off); 693 if (fin6->fi6_gpi == 0) 694 fin6->fi6_gpi = opt6->ah_spi; 695 proto = opt6->opt6_nxt; 696 off += 8 + (opt6->opt6_hlen * 4); 697 /* goto the next header */ 698 break; 699 } 700 701 case IPPROTO_HOPOPTS: 702 case IPPROTO_ROUTING: 703 case IPPROTO_DSTOPTS: { 704 /* get next header and header length */ 705 struct _opt6 *opt6; 706 707 opt6 = (struct _opt6 *)(mtod(m0, caddr_t) + off); 708 proto = opt6->opt6_nxt; 709 off += (opt6->opt6_hlen + 1) * 8; 710 /* goto the next header */ 711 break; 712 } 713 714 case IPPROTO_FRAGMENT: 715 /* ipv6 fragmentations are not supported yet */ 716 default: 717 fin6->fi6_proto = proto; 718 return (0); 719 } 720 } while (1); 721 /*NOTREACHED*/ 722 } 723 #endif /* INET6 */ 724 725 /* 726 * altq common classifier 727 */ 728 int 729 acc_add_filter(classifier, filter, class, phandle) 730 struct acc_classifier *classifier; 731 struct flow_filter *filter; 732 void *class; 733 u_long *phandle; 734 { 735 struct acc_filter *afp, *prev, *tmp; 736 int i, s; 737 738 #ifdef INET6 739 if (filter->ff_flow.fi_family != AF_INET && 740 filter->ff_flow.fi_family != AF_INET6) 741 return (EINVAL); 742 #else 743 if (filter->ff_flow.fi_family != AF_INET) 744 return (EINVAL); 745 #endif 746 747 MALLOC(afp, struct acc_filter *, sizeof(struct acc_filter), 748 M_DEVBUF, M_WAITOK); 749 if (afp == NULL) 750 return (ENOMEM); 751 bzero(afp, sizeof(struct acc_filter)); 752 753 afp->f_filter = *filter; 754 afp->f_class = class; 755 756 i = ACC_WILDCARD_INDEX; 757 if (filter->ff_flow.fi_family == AF_INET) { 758 struct flow_filter *filter4 = &afp->f_filter; 759 760 /* 761 * if address is 0, it's a wildcard. if address mask 762 * isn't set, use full mask. 763 */ 764 if (filter4->ff_flow.fi_dst.s_addr == 0) 765 filter4->ff_mask.mask_dst.s_addr = 0; 766 else if (filter4->ff_mask.mask_dst.s_addr == 0) 767 filter4->ff_mask.mask_dst.s_addr = 0xffffffff; 768 if (filter4->ff_flow.fi_src.s_addr == 0) 769 filter4->ff_mask.mask_src.s_addr = 0; 770 else if (filter4->ff_mask.mask_src.s_addr == 0) 771 filter4->ff_mask.mask_src.s_addr = 0xffffffff; 772 773 /* clear extra bits in addresses */ 774 filter4->ff_flow.fi_dst.s_addr &= 775 filter4->ff_mask.mask_dst.s_addr; 776 filter4->ff_flow.fi_src.s_addr &= 777 filter4->ff_mask.mask_src.s_addr; 778 779 /* 780 * if dst address is a wildcard, use hash-entry 781 * ACC_WILDCARD_INDEX. 782 */ 783 if (filter4->ff_mask.mask_dst.s_addr != 0xffffffff) 784 i = ACC_WILDCARD_INDEX; 785 else 786 i = ACC_GET_HASH_INDEX(filter4->ff_flow.fi_dst.s_addr); 787 } 788 #ifdef INET6 789 else if (filter->ff_flow.fi_family == AF_INET6) { 790 struct flow_filter6 *filter6 = 791 (struct flow_filter6 *)&afp->f_filter; 792 #ifndef IN6MASK0 /* taken from kame ipv6 */ 793 #define IN6MASK0 {{{ 0, 0, 0, 0 }}} 794 #define IN6MASK128 {{{ 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }}} 795 const struct in6_addr in6mask0 = IN6MASK0; 796 const struct in6_addr in6mask128 = IN6MASK128; 797 #endif 798 799 if (IN6_IS_ADDR_UNSPECIFIED(&filter6->ff_flow6.fi6_dst)) 800 filter6->ff_mask6.mask6_dst = in6mask0; 801 else if (IN6_IS_ADDR_UNSPECIFIED(&filter6->ff_mask6.mask6_dst)) 802 filter6->ff_mask6.mask6_dst = in6mask128; 803 if (IN6_IS_ADDR_UNSPECIFIED(&filter6->ff_flow6.fi6_src)) 804 filter6->ff_mask6.mask6_src = in6mask0; 805 else if (IN6_IS_ADDR_UNSPECIFIED(&filter6->ff_mask6.mask6_src)) 806 filter6->ff_mask6.mask6_src = in6mask128; 807 808 /* clear extra bits in addresses */ 809 for (i = 0; i < 16; i++) 810 filter6->ff_flow6.fi6_dst.s6_addr[i] &= 811 filter6->ff_mask6.mask6_dst.s6_addr[i]; 812 for (i = 0; i < 16; i++) 813 filter6->ff_flow6.fi6_src.s6_addr[i] &= 814 filter6->ff_mask6.mask6_src.s6_addr[i]; 815 816 if (filter6->ff_flow6.fi6_flowlabel == 0) 817 i = ACC_WILDCARD_INDEX; 818 else 819 i = ACC_GET_HASH_INDEX(filter6->ff_flow6.fi6_flowlabel); 820 } 821 #endif /* INET6 */ 822 823 afp->f_handle = get_filt_handle(classifier, i); 824 825 /* update filter bitmask */ 826 afp->f_fbmask = filt2fibmask(filter); 827 classifier->acc_fbmask |= afp->f_fbmask; 828 829 /* 830 * add this filter to the filter list. 831 * filters are ordered from the highest rule number. 832 */ 833 s = splnet(); 834 prev = NULL; 835 LIST_FOREACH(tmp, &classifier->acc_filters[i], f_chain) { 836 if (tmp->f_filter.ff_ruleno > afp->f_filter.ff_ruleno) 837 prev = tmp; 838 else 839 break; 840 } 841 if (prev == NULL) 842 LIST_INSERT_HEAD(&classifier->acc_filters[i], afp, f_chain); 843 else 844 LIST_INSERT_AFTER(prev, afp, f_chain); 845 splx(s); 846 847 *phandle = afp->f_handle; 848 return (0); 849 } 850 851 int 852 acc_delete_filter(classifier, handle) 853 struct acc_classifier *classifier; 854 u_long handle; 855 { 856 struct acc_filter *afp; 857 int s; 858 859 if ((afp = filth_to_filtp(classifier, handle)) == NULL) 860 return (EINVAL); 861 862 s = splnet(); 863 LIST_REMOVE(afp, f_chain); 864 splx(s); 865 866 FREE(afp, M_DEVBUF); 867 868 /* todo: update filt_bmask */ 869 870 return (0); 871 } 872 873 /* 874 * delete filters referencing to the specified class. 875 * if the all flag is not 0, delete all the filters. 876 */ 877 int 878 acc_discard_filters(classifier, class, all) 879 struct acc_classifier *classifier; 880 void *class; 881 int all; 882 { 883 struct acc_filter *afp; 884 int i, s; 885 886 s = splnet(); 887 for (i = 0; i < ACC_FILTER_TABLESIZE; i++) { 888 do { 889 LIST_FOREACH(afp, &classifier->acc_filters[i], f_chain) 890 if (all || afp->f_class == class) { 891 LIST_REMOVE(afp, f_chain); 892 FREE(afp, M_DEVBUF); 893 /* start again from the head */ 894 break; 895 } 896 } while (afp != NULL); 897 } 898 splx(s); 899 900 if (all) 901 classifier->acc_fbmask = 0; 902 903 return (0); 904 } 905 906 void * 907 acc_classify(clfier, m, af) 908 void *clfier; 909 struct mbuf *m; 910 int af; 911 { 912 struct acc_classifier *classifier; 913 struct flowinfo flow; 914 struct acc_filter *afp; 915 int i; 916 917 classifier = (struct acc_classifier *)clfier; 918 altq_extractflow(m, af, &flow, classifier->acc_fbmask); 919 920 if (flow.fi_family == AF_INET) { 921 struct flowinfo_in *fp = (struct flowinfo_in *)&flow; 922 923 if ((classifier->acc_fbmask & FIMB4_ALL) == FIMB4_TOS) { 924 /* only tos is used */ 925 LIST_FOREACH(afp, 926 &classifier->acc_filters[ACC_WILDCARD_INDEX], 927 f_chain) 928 if (apply_tosfilter4(afp->f_fbmask, 929 &afp->f_filter, fp)) 930 /* filter matched */ 931 return (afp->f_class); 932 } else if ((classifier->acc_fbmask & 933 (~(FIMB4_PROTO|FIMB4_SPORT|FIMB4_DPORT) & FIMB4_ALL)) 934 == 0) { 935 /* only proto and ports are used */ 936 LIST_FOREACH(afp, 937 &classifier->acc_filters[ACC_WILDCARD_INDEX], 938 f_chain) 939 if (apply_ppfilter4(afp->f_fbmask, 940 &afp->f_filter, fp)) 941 /* filter matched */ 942 return (afp->f_class); 943 } else { 944 /* get the filter hash entry from its dest address */ 945 i = ACC_GET_HASH_INDEX(fp->fi_dst.s_addr); 946 do { 947 /* 948 * go through this loop twice. first for dst 949 * hash, second for wildcards. 950 */ 951 LIST_FOREACH(afp, &classifier->acc_filters[i], 952 f_chain) 953 if (apply_filter4(afp->f_fbmask, 954 &afp->f_filter, fp)) 955 /* filter matched */ 956 return (afp->f_class); 957 958 /* 959 * check again for filters with a dst addr 960 * wildcard. 961 * (daddr == 0 || dmask != 0xffffffff). 962 */ 963 if (i != ACC_WILDCARD_INDEX) 964 i = ACC_WILDCARD_INDEX; 965 else 966 break; 967 } while (1); 968 } 969 } 970 #ifdef INET6 971 else if (flow.fi_family == AF_INET6) { 972 struct flowinfo_in6 *fp6 = (struct flowinfo_in6 *)&flow; 973 974 /* get the filter hash entry from its flow ID */ 975 if (fp6->fi6_flowlabel != 0) 976 i = ACC_GET_HASH_INDEX(fp6->fi6_flowlabel); 977 else 978 /* flowlable can be zero */ 979 i = ACC_WILDCARD_INDEX; 980 981 /* go through this loop twice. first for flow hash, second 982 for wildcards. */ 983 do { 984 LIST_FOREACH(afp, &classifier->acc_filters[i], f_chain) 985 if (apply_filter6(afp->f_fbmask, 986 (struct flow_filter6 *)&afp->f_filter, 987 fp6)) 988 /* filter matched */ 989 return (afp->f_class); 990 991 /* 992 * check again for filters with a wildcard. 993 */ 994 if (i != ACC_WILDCARD_INDEX) 995 i = ACC_WILDCARD_INDEX; 996 else 997 break; 998 } while (1); 999 } 1000 #endif /* INET6 */ 1001 1002 /* no filter matched */ 1003 return (NULL); 1004 } 1005 1006 static int 1007 apply_filter4(fbmask, filt, pkt) 1008 u_int32_t fbmask; 1009 struct flow_filter *filt; 1010 struct flowinfo_in *pkt; 1011 { 1012 if (filt->ff_flow.fi_family != AF_INET) 1013 return (0); 1014 if ((fbmask & FIMB4_SPORT) && filt->ff_flow.fi_sport != pkt->fi_sport) 1015 return (0); 1016 if ((fbmask & FIMB4_DPORT) && filt->ff_flow.fi_dport != pkt->fi_dport) 1017 return (0); 1018 if ((fbmask & FIMB4_DADDR) && 1019 filt->ff_flow.fi_dst.s_addr != 1020 (pkt->fi_dst.s_addr & filt->ff_mask.mask_dst.s_addr)) 1021 return (0); 1022 if ((fbmask & FIMB4_SADDR) && 1023 filt->ff_flow.fi_src.s_addr != 1024 (pkt->fi_src.s_addr & filt->ff_mask.mask_src.s_addr)) 1025 return (0); 1026 if ((fbmask & FIMB4_PROTO) && filt->ff_flow.fi_proto != pkt->fi_proto) 1027 return (0); 1028 if ((fbmask & FIMB4_TOS) && filt->ff_flow.fi_tos != 1029 (pkt->fi_tos & filt->ff_mask.mask_tos)) 1030 return (0); 1031 if ((fbmask & FIMB4_GPI) && filt->ff_flow.fi_gpi != (pkt->fi_gpi)) 1032 return (0); 1033 /* match */ 1034 return (1); 1035 } 1036 1037 /* 1038 * filter matching function optimized for a common case that checks 1039 * only protocol and port numbers 1040 */ 1041 static int 1042 apply_ppfilter4(fbmask, filt, pkt) 1043 u_int32_t fbmask; 1044 struct flow_filter *filt; 1045 struct flowinfo_in *pkt; 1046 { 1047 if (filt->ff_flow.fi_family != AF_INET) 1048 return (0); 1049 if ((fbmask & FIMB4_SPORT) && filt->ff_flow.fi_sport != pkt->fi_sport) 1050 return (0); 1051 if ((fbmask & FIMB4_DPORT) && filt->ff_flow.fi_dport != pkt->fi_dport) 1052 return (0); 1053 if ((fbmask & FIMB4_PROTO) && filt->ff_flow.fi_proto != pkt->fi_proto) 1054 return (0); 1055 /* match */ 1056 return (1); 1057 } 1058 1059 /* 1060 * filter matching function only for tos field. 1061 */ 1062 static int 1063 apply_tosfilter4(fbmask, filt, pkt) 1064 u_int32_t fbmask; 1065 struct flow_filter *filt; 1066 struct flowinfo_in *pkt; 1067 { 1068 if (filt->ff_flow.fi_family != AF_INET) 1069 return (0); 1070 if ((fbmask & FIMB4_TOS) && filt->ff_flow.fi_tos != 1071 (pkt->fi_tos & filt->ff_mask.mask_tos)) 1072 return (0); 1073 /* match */ 1074 return (1); 1075 } 1076 1077 #ifdef INET6 1078 static int 1079 apply_filter6(fbmask, filt, pkt) 1080 u_int32_t fbmask; 1081 struct flow_filter6 *filt; 1082 struct flowinfo_in6 *pkt; 1083 { 1084 int i; 1085 1086 if (filt->ff_flow6.fi6_family != AF_INET6) 1087 return (0); 1088 if ((fbmask & FIMB6_FLABEL) && 1089 filt->ff_flow6.fi6_flowlabel != pkt->fi6_flowlabel) 1090 return (0); 1091 if ((fbmask & FIMB6_PROTO) && 1092 filt->ff_flow6.fi6_proto != pkt->fi6_proto) 1093 return (0); 1094 if ((fbmask & FIMB6_SPORT) && 1095 filt->ff_flow6.fi6_sport != pkt->fi6_sport) 1096 return (0); 1097 if ((fbmask & FIMB6_DPORT) && 1098 filt->ff_flow6.fi6_dport != pkt->fi6_dport) 1099 return (0); 1100 if (fbmask & FIMB6_SADDR) { 1101 for (i = 0; i < 4; i++) 1102 if (filt->ff_flow6.fi6_src.s6_addr32[i] != 1103 (pkt->fi6_src.s6_addr32[i] & 1104 filt->ff_mask6.mask6_src.s6_addr32[i])) 1105 return (0); 1106 } 1107 if (fbmask & FIMB6_DADDR) { 1108 for (i = 0; i < 4; i++) 1109 if (filt->ff_flow6.fi6_dst.s6_addr32[i] != 1110 (pkt->fi6_dst.s6_addr32[i] & 1111 filt->ff_mask6.mask6_dst.s6_addr32[i])) 1112 return (0); 1113 } 1114 if ((fbmask & FIMB6_TCLASS) && 1115 filt->ff_flow6.fi6_tclass != 1116 (pkt->fi6_tclass & filt->ff_mask6.mask6_tclass)) 1117 return (0); 1118 if ((fbmask & FIMB6_GPI) && 1119 filt->ff_flow6.fi6_gpi != pkt->fi6_gpi) 1120 return (0); 1121 /* match */ 1122 return (1); 1123 } 1124 #endif /* INET6 */ 1125 1126 /* 1127 * filter handle: 1128 * bit 20-28: index to the filter hash table 1129 * bit 0-19: unique id in the hash bucket. 1130 */ 1131 static u_long 1132 get_filt_handle(classifier, i) 1133 struct acc_classifier *classifier; 1134 int i; 1135 { 1136 static u_long handle_number = 1; 1137 u_long handle; 1138 struct acc_filter *afp; 1139 1140 while (1) { 1141 handle = handle_number++ & 0x000fffff; 1142 1143 if (LIST_EMPTY(&classifier->acc_filters[i])) 1144 break; 1145 1146 LIST_FOREACH(afp, &classifier->acc_filters[i], f_chain) 1147 if ((afp->f_handle & 0x000fffff) == handle) 1148 break; 1149 if (afp == NULL) 1150 break; 1151 /* this handle is already used, try again */ 1152 } 1153 1154 return ((i << 20) | handle); 1155 } 1156 1157 /* convert filter handle to filter pointer */ 1158 static struct acc_filter * 1159 filth_to_filtp(classifier, handle) 1160 struct acc_classifier *classifier; 1161 u_long handle; 1162 { 1163 struct acc_filter *afp; 1164 int i; 1165 1166 i = ACC_GET_HINDEX(handle); 1167 1168 LIST_FOREACH(afp, &classifier->acc_filters[i], f_chain) 1169 if (afp->f_handle == handle) 1170 return (afp); 1171 1172 return (NULL); 1173 } 1174 1175 /* create flowinfo bitmask */ 1176 static u_int32_t 1177 filt2fibmask(filt) 1178 struct flow_filter *filt; 1179 { 1180 u_int32_t mask = 0; 1181 #ifdef INET6 1182 struct flow_filter6 *filt6; 1183 #endif 1184 1185 switch (filt->ff_flow.fi_family) { 1186 case AF_INET: 1187 if (filt->ff_flow.fi_proto != 0) 1188 mask |= FIMB4_PROTO; 1189 if (filt->ff_flow.fi_tos != 0) 1190 mask |= FIMB4_TOS; 1191 if (filt->ff_flow.fi_dst.s_addr != 0) 1192 mask |= FIMB4_DADDR; 1193 if (filt->ff_flow.fi_src.s_addr != 0) 1194 mask |= FIMB4_SADDR; 1195 if (filt->ff_flow.fi_sport != 0) 1196 mask |= FIMB4_SPORT; 1197 if (filt->ff_flow.fi_dport != 0) 1198 mask |= FIMB4_DPORT; 1199 if (filt->ff_flow.fi_gpi != 0) 1200 mask |= FIMB4_GPI; 1201 break; 1202 #ifdef INET6 1203 case AF_INET6: 1204 filt6 = (struct flow_filter6 *)filt; 1205 1206 if (filt6->ff_flow6.fi6_proto != 0) 1207 mask |= FIMB6_PROTO; 1208 if (filt6->ff_flow6.fi6_tclass != 0) 1209 mask |= FIMB6_TCLASS; 1210 if (!IN6_IS_ADDR_UNSPECIFIED(&filt6->ff_flow6.fi6_dst)) 1211 mask |= FIMB6_DADDR; 1212 if (!IN6_IS_ADDR_UNSPECIFIED(&filt6->ff_flow6.fi6_src)) 1213 mask |= FIMB6_SADDR; 1214 if (filt6->ff_flow6.fi6_sport != 0) 1215 mask |= FIMB6_SPORT; 1216 if (filt6->ff_flow6.fi6_dport != 0) 1217 mask |= FIMB6_DPORT; 1218 if (filt6->ff_flow6.fi6_gpi != 0) 1219 mask |= FIMB6_GPI; 1220 if (filt6->ff_flow6.fi6_flowlabel != 0) 1221 mask |= FIMB6_FLABEL; 1222 break; 1223 #endif /* INET6 */ 1224 } 1225 return (mask); 1226 } 1227 1228 1229 /* 1230 * helper functions to handle IPv4 fragments. 1231 * currently only in-sequence fragments are handled. 1232 * - fragment info is cached in a LRU list. 1233 * - when a first fragment is found, cache its flow info. 1234 * - when a non-first fragment is found, lookup the cache. 1235 */ 1236 1237 struct ip4_frag { 1238 TAILQ_ENTRY(ip4_frag) ip4f_chain; 1239 char ip4f_valid; 1240 u_short ip4f_id; 1241 struct flowinfo_in ip4f_info; 1242 }; 1243 1244 static TAILQ_HEAD(ip4f_list, ip4_frag) ip4f_list; /* IPv4 fragment cache */ 1245 1246 #define IP4F_TABSIZE 16 /* IPv4 fragment cache size */ 1247 1248 1249 static void 1250 ip4f_cache(ip, fin) 1251 struct ip *ip; 1252 struct flowinfo_in *fin; 1253 { 1254 struct ip4_frag *fp; 1255 1256 if (TAILQ_EMPTY(&ip4f_list)) { 1257 /* first time call, allocate fragment cache entries. */ 1258 if (ip4f_init() < 0) 1259 /* allocation failed! */ 1260 return; 1261 } 1262 1263 fp = ip4f_alloc(); 1264 fp->ip4f_id = ip->ip_id; 1265 1266 /* save port numbers */ 1267 fp->ip4f_info.fi_sport = fin->fi_sport; 1268 fp->ip4f_info.fi_dport = fin->fi_dport; 1269 fp->ip4f_info.fi_gpi = fin->fi_gpi; 1270 } 1271 1272 static int 1273 ip4f_lookup(ip, fin) 1274 struct ip *ip; 1275 struct flowinfo_in *fin; 1276 { 1277 struct ip4_frag *fp; 1278 1279 for (fp = TAILQ_FIRST(&ip4f_list); fp != NULL && fp->ip4f_valid; 1280 fp = TAILQ_NEXT(fp, ip4f_chain)) 1281 if (ip->ip_id == fp->ip4f_id && 1282 ip->ip_src.s_addr == fp->ip4f_info.fi_src.s_addr && 1283 ip->ip_dst.s_addr == fp->ip4f_info.fi_dst.s_addr && 1284 ip->ip_p == fp->ip4f_info.fi_proto) { 1285 1286 /* found the matching entry */ 1287 fin->fi_sport = fp->ip4f_info.fi_sport; 1288 fin->fi_dport = fp->ip4f_info.fi_dport; 1289 fin->fi_gpi = fp->ip4f_info.fi_gpi; 1290 1291 if ((ntohs(ip->ip_off) & IP_MF) == 0) 1292 /* this is the last fragment, 1293 release the entry. */ 1294 ip4f_free(fp); 1295 1296 return (1); 1297 } 1298 1299 /* no matching entry found */ 1300 return (0); 1301 } 1302 1303 static int 1304 ip4f_init(void) 1305 { 1306 struct ip4_frag *fp; 1307 int i; 1308 1309 TAILQ_INIT(&ip4f_list); 1310 for (i=0; i<IP4F_TABSIZE; i++) { 1311 MALLOC(fp, struct ip4_frag *, sizeof(struct ip4_frag), 1312 M_DEVBUF, M_NOWAIT); 1313 if (fp == NULL) { 1314 printf("ip4f_init: can't alloc %dth entry!\n", i); 1315 if (i == 0) 1316 return (-1); 1317 return (0); 1318 } 1319 fp->ip4f_valid = 0; 1320 TAILQ_INSERT_TAIL(&ip4f_list, fp, ip4f_chain); 1321 } 1322 return (0); 1323 } 1324 1325 static struct ip4_frag * 1326 ip4f_alloc(void) 1327 { 1328 struct ip4_frag *fp; 1329 1330 /* reclaim an entry at the tail, put it at the head */ 1331 fp = TAILQ_LAST(&ip4f_list, ip4f_list); 1332 TAILQ_REMOVE(&ip4f_list, fp, ip4f_chain); 1333 fp->ip4f_valid = 1; 1334 TAILQ_INSERT_HEAD(&ip4f_list, fp, ip4f_chain); 1335 return (fp); 1336 } 1337 1338 static void 1339 ip4f_free(fp) 1340 struct ip4_frag *fp; 1341 { 1342 TAILQ_REMOVE(&ip4f_list, fp, ip4f_chain); 1343 fp->ip4f_valid = 0; 1344 TAILQ_INSERT_TAIL(&ip4f_list, fp, ip4f_chain); 1345 } 1346 1347 /* 1348 * read and write diffserv field in IPv4 or IPv6 header 1349 */ 1350 u_int8_t 1351 read_dsfield(m, pktattr) 1352 struct mbuf *m; 1353 struct altq_pktattr *pktattr; 1354 { 1355 struct mbuf *m0; 1356 u_int8_t ds_field = 0; 1357 1358 if (pktattr == NULL || 1359 (pktattr->pattr_af != AF_INET && pktattr->pattr_af != AF_INET6)) 1360 return ((u_int8_t)0); 1361 1362 /* verify that pattr_hdr is within the mbuf data */ 1363 for (m0 = m; m0 != NULL; m0 = m0->m_next) 1364 if ((pktattr->pattr_hdr >= m0->m_data) && 1365 (pktattr->pattr_hdr < m0->m_data + m0->m_len)) 1366 break; 1367 if (m0 == NULL) { 1368 /* ick, pattr_hdr is stale */ 1369 pktattr->pattr_af = AF_UNSPEC; 1370 #ifdef ALTQ_DEBUG 1371 printf("read_dsfield: can't locate header!\n"); 1372 #endif 1373 return ((u_int8_t)0); 1374 } 1375 1376 if (pktattr->pattr_af == AF_INET) { 1377 struct ip *ip = (struct ip *)pktattr->pattr_hdr; 1378 1379 if (ip->ip_v != 4) 1380 return ((u_int8_t)0); /* version mismatch! */ 1381 ds_field = ip->ip_tos; 1382 } 1383 #ifdef INET6 1384 else if (pktattr->pattr_af == AF_INET6) { 1385 struct ip6_hdr *ip6 = (struct ip6_hdr *)pktattr->pattr_hdr; 1386 u_int32_t flowlabel; 1387 1388 flowlabel = ntohl(ip6->ip6_flow); 1389 if ((flowlabel >> 28) != 6) 1390 return ((u_int8_t)0); /* version mismatch! */ 1391 ds_field = (flowlabel >> 20) & 0xff; 1392 } 1393 #endif 1394 return (ds_field); 1395 } 1396 1397 void 1398 write_dsfield(m, pktattr, dsfield) 1399 struct mbuf *m; 1400 struct altq_pktattr *pktattr; 1401 u_int8_t dsfield; 1402 { 1403 struct mbuf *m0; 1404 1405 if (pktattr == NULL || 1406 (pktattr->pattr_af != AF_INET && pktattr->pattr_af != AF_INET6)) 1407 return; 1408 1409 /* verify that pattr_hdr is within the mbuf data */ 1410 for (m0 = m; m0 != NULL; m0 = m0->m_next) 1411 if ((pktattr->pattr_hdr >= m0->m_data) && 1412 (pktattr->pattr_hdr < m0->m_data + m0->m_len)) 1413 break; 1414 if (m0 == NULL) { 1415 /* ick, pattr_hdr is stale */ 1416 pktattr->pattr_af = AF_UNSPEC; 1417 #ifdef ALTQ_DEBUG 1418 printf("write_dsfield: can't locate header!\n"); 1419 #endif 1420 return; 1421 } 1422 1423 if (pktattr->pattr_af == AF_INET) { 1424 struct ip *ip = (struct ip *)pktattr->pattr_hdr; 1425 u_int8_t old; 1426 int32_t sum; 1427 1428 if (ip->ip_v != 4) 1429 return; /* version mismatch! */ 1430 old = ip->ip_tos; 1431 dsfield |= old & 3; /* leave CU bits */ 1432 if (old == dsfield) 1433 return; 1434 ip->ip_tos = dsfield; 1435 /* 1436 * update checksum (from RFC1624) 1437 * HC' = ~(~HC + ~m + m') 1438 */ 1439 sum = ~ntohs(ip->ip_sum) & 0xffff; 1440 sum += 0xff00 + (~old & 0xff) + dsfield; 1441 sum = (sum >> 16) + (sum & 0xffff); 1442 sum += (sum >> 16); /* add carry */ 1443 1444 ip->ip_sum = htons(~sum & 0xffff); 1445 } 1446 #ifdef INET6 1447 else if (pktattr->pattr_af == AF_INET6) { 1448 struct ip6_hdr *ip6 = (struct ip6_hdr *)pktattr->pattr_hdr; 1449 u_int32_t flowlabel; 1450 1451 flowlabel = ntohl(ip6->ip6_flow); 1452 if ((flowlabel >> 28) != 6) 1453 return; /* version mismatch! */ 1454 flowlabel = (flowlabel & 0xf03fffff) | (dsfield << 20); 1455 ip6->ip6_flow = htonl(flowlabel); 1456 } 1457 #endif 1458 return; 1459 } 1460 1461 1462 /* 1463 * high resolution clock support taking advantage of a machine dependent 1464 * high resolution time counter (e.g., timestamp counter of intel pentium). 1465 * we assume 1466 * - 64-bit-long monotonically-increasing counter 1467 * - frequency range is 100M-4GHz (CPU speed) 1468 */ 1469 u_int32_t machclk_freq = 0; 1470 u_int32_t machclk_per_tick = 0; 1471 1472 #if (defined(__i386__) || defined(__alpha__)) && !defined(ALTQ_NOPCC) 1473 #ifdef __FreeBSD__ 1474 /* freebsd makes clock frequency accessible */ 1475 #ifdef __alpha__ 1476 extern u_int32_t cycles_per_sec; /* alpha cpu clock frequency */ 1477 #endif 1478 void 1479 init_machclk(void) 1480 { 1481 #if defined(__i386__) 1482 #if (__FreeBSD_version > 300000) 1483 machclk_freq = tsc_freq; 1484 #else 1485 machclk_freq = i586_ctr_freq; 1486 #endif 1487 #elif defined(__alpha__) 1488 machclk_freq = cycles_per_sec; 1489 #endif /* __alpha__ */ 1490 machclk_per_tick = machclk_freq / hz; 1491 } 1492 #else /* !__FreeBSD__ */ 1493 /* 1494 * measure Pentium TSC or Alpha PCC clock frequency 1495 */ 1496 void 1497 init_machclk(void) 1498 { 1499 static int wait; 1500 struct timeval tv_start, tv_end; 1501 u_int64_t start, end, diff; 1502 int timo; 1503 1504 microtime(&tv_start); 1505 start = read_machclk(); 1506 timo = hz; /* 1 sec */ 1507 (void)tsleep(&wait, PWAIT | PCATCH, "init_machclk", timo); 1508 microtime(&tv_end); 1509 end = read_machclk(); 1510 diff = (u_int64_t)(tv_end.tv_sec - tv_start.tv_sec) * 1000000 1511 + tv_end.tv_usec - tv_start.tv_usec; 1512 if (diff != 0) 1513 machclk_freq = (u_int)((end - start) * 1000000 / diff); 1514 machclk_per_tick = machclk_freq / hz; 1515 1516 printf("altq: CPU clock: %uHz\n", machclk_freq); 1517 } 1518 #endif /* !__FreeBSD__ */ 1519 #ifdef __alpha__ 1520 /* 1521 * make a 64bit counter value out of the 32bit alpha processor cycle counter. 1522 * read_machclk must be called within a half of its wrap-around cycle 1523 * (about 5 sec for 400MHz cpu) to properly detect a counter wrap-around. 1524 * tbr_timeout calls read_machclk once a second. 1525 */ 1526 u_int64_t 1527 read_machclk(void) 1528 { 1529 static u_int32_t last_pcc, upper; 1530 u_int32_t pcc; 1531 1532 pcc = (u_int32_t)alpha_rpcc(); 1533 if (pcc <= last_pcc) 1534 upper++; 1535 last_pcc = pcc; 1536 return (((u_int64_t)upper << 32) + pcc); 1537 } 1538 #endif /* __alpha__ */ 1539 #else /* !i386 && !alpha */ 1540 /* use microtime() for now */ 1541 void 1542 init_machclk(void) 1543 { 1544 machclk_freq = 1000000 << MACHCLK_SHIFT; 1545 machclk_per_tick = machclk_freq / hz; 1546 printf("altq: emulate %uHz cpu clock\n", machclk_freq); 1547 } 1548 #endif /* !i386 && !alpha */ 1549
Indexes created Wed Oct 15 16:09:53 GMT 2025