altq_subr.c revision 263086
1/* $FreeBSD: stable/10/sys/contrib/altq/altq/altq_subr.c 263086 2014-03-12 10:45:58Z glebius $ */ 2/* $KAME: altq_subr.c,v 1.21 2003/11/06 06:32:53 kjc 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#if defined(__FreeBSD__) || defined(__NetBSD__) 31#include "opt_altq.h" 32#include "opt_inet.h" 33#ifdef __FreeBSD__ 34#include "opt_inet6.h" 35#endif 36#endif /* __FreeBSD__ || __NetBSD__ */ 37 38#include <sys/param.h> 39#include <sys/malloc.h> 40#include <sys/mbuf.h> 41#include <sys/systm.h> 42#include <sys/proc.h> 43#include <sys/socket.h> 44#include <sys/socketvar.h> 45#include <sys/kernel.h> 46#include <sys/errno.h> 47#include <sys/syslog.h> 48#include <sys/sysctl.h> 49#include <sys/queue.h> 50 51#include <net/if.h> 52#include <net/if_var.h> 53#include <net/if_dl.h> 54#include <net/if_types.h> 55#ifdef __FreeBSD__ 56#include <net/vnet.h> 57#endif 58 59#include <netinet/in.h> 60#include <netinet/in_systm.h> 61#include <netinet/ip.h> 62#ifdef INET6 63#include <netinet/ip6.h> 64#endif 65#include <netinet/tcp.h> 66#include <netinet/udp.h> 67 68#include <netpfil/pf/pf.h> 69#include <netpfil/pf/pf_altq.h> 70#include <altq/altq.h> 71#ifdef ALTQ3_COMPAT 72#include <altq/altq_conf.h> 73#endif 74 75/* machine dependent clock related includes */ 76#ifdef __FreeBSD__ 77#include <sys/bus.h> 78#include <sys/cpu.h> 79#include <sys/eventhandler.h> 80#include <machine/clock.h> 81#endif 82#if defined(__amd64__) || defined(__i386__) 83#include <machine/cpufunc.h> /* for pentium tsc */ 84#include <machine/specialreg.h> /* for CPUID_TSC */ 85#ifdef __FreeBSD__ 86#include <machine/md_var.h> /* for cpu_feature */ 87#elif defined(__NetBSD__) || defined(__OpenBSD__) 88#include <machine/cpu.h> /* for cpu_feature */ 89#endif 90#endif /* __amd64 || __i386__ */ 91 92/* 93 * internal function prototypes 94 */ 95static void tbr_timeout(void *); 96int (*altq_input)(struct mbuf *, int) = NULL; 97static struct mbuf *tbr_dequeue(struct ifaltq *, int); 98static int tbr_timer = 0; /* token bucket regulator timer */ 99#if !defined(__FreeBSD__) || (__FreeBSD_version < 600000) 100static struct callout tbr_callout = CALLOUT_INITIALIZER; 101#else 102static struct callout tbr_callout; 103#endif 104 105#ifdef ALTQ3_CLFIER_COMPAT 106static int extract_ports4(struct mbuf *, struct ip *, struct flowinfo_in *); 107#ifdef INET6 108static int extract_ports6(struct mbuf *, struct ip6_hdr *, 109 struct flowinfo_in6 *); 110#endif 111static int apply_filter4(u_int32_t, struct flow_filter *, 112 struct flowinfo_in *); 113static int apply_ppfilter4(u_int32_t, struct flow_filter *, 114 struct flowinfo_in *); 115#ifdef INET6 116static int apply_filter6(u_int32_t, struct flow_filter6 *, 117 struct flowinfo_in6 *); 118#endif 119static int apply_tosfilter4(u_int32_t, struct flow_filter *, 120 struct flowinfo_in *); 121static u_long get_filt_handle(struct acc_classifier *, int); 122static struct acc_filter *filth_to_filtp(struct acc_classifier *, u_long); 123static u_int32_t filt2fibmask(struct flow_filter *); 124 125static void ip4f_cache(struct ip *, struct flowinfo_in *); 126static int ip4f_lookup(struct ip *, struct flowinfo_in *); 127static int ip4f_init(void); 128static struct ip4_frag *ip4f_alloc(void); 129static void ip4f_free(struct ip4_frag *); 130#endif /* ALTQ3_CLFIER_COMPAT */ 131 132/* 133 * alternate queueing support routines 134 */ 135 136/* look up the queue state by the interface name and the queueing type. */ 137void * 138altq_lookup(name, type) 139 char *name; 140 int type; 141{ 142 struct ifnet *ifp; 143 144 if ((ifp = ifunit(name)) != NULL) { 145 /* read if_snd unlocked */ 146 if (type != ALTQT_NONE && ifp->if_snd.altq_type == type) 147 return (ifp->if_snd.altq_disc); 148 } 149 150 return NULL; 151} 152 153int 154altq_attach(ifq, type, discipline, enqueue, dequeue, request, clfier, classify) 155 struct ifaltq *ifq; 156 int type; 157 void *discipline; 158 int (*enqueue)(struct ifaltq *, struct mbuf *, struct altq_pktattr *); 159 struct mbuf *(*dequeue)(struct ifaltq *, int); 160 int (*request)(struct ifaltq *, int, void *); 161 void *clfier; 162 void *(*classify)(void *, struct mbuf *, int); 163{ 164 IFQ_LOCK(ifq); 165 if (!ALTQ_IS_READY(ifq)) { 166 IFQ_UNLOCK(ifq); 167 return ENXIO; 168 } 169 170#ifdef ALTQ3_COMPAT 171 /* 172 * pfaltq can override the existing discipline, but altq3 cannot. 173 * check these if clfier is not NULL (which implies altq3). 174 */ 175 if (clfier != NULL) { 176 if (ALTQ_IS_ENABLED(ifq)) { 177 IFQ_UNLOCK(ifq); 178 return EBUSY; 179 } 180 if (ALTQ_IS_ATTACHED(ifq)) { 181 IFQ_UNLOCK(ifq); 182 return EEXIST; 183 } 184 } 185#endif 186 ifq->altq_type = type; 187 ifq->altq_disc = discipline; 188 ifq->altq_enqueue = enqueue; 189 ifq->altq_dequeue = dequeue; 190 ifq->altq_request = request; 191 ifq->altq_clfier = clfier; 192 ifq->altq_classify = classify; 193 ifq->altq_flags &= (ALTQF_CANTCHANGE|ALTQF_ENABLED); 194#ifdef ALTQ3_COMPAT 195#ifdef ALTQ_KLD 196 altq_module_incref(type); 197#endif 198#endif 199 IFQ_UNLOCK(ifq); 200 return 0; 201} 202 203int 204altq_detach(ifq) 205 struct ifaltq *ifq; 206{ 207 IFQ_LOCK(ifq); 208 209 if (!ALTQ_IS_READY(ifq)) { 210 IFQ_UNLOCK(ifq); 211 return ENXIO; 212 } 213 if (ALTQ_IS_ENABLED(ifq)) { 214 IFQ_UNLOCK(ifq); 215 return EBUSY; 216 } 217 if (!ALTQ_IS_ATTACHED(ifq)) { 218 IFQ_UNLOCK(ifq); 219 return (0); 220 } 221#ifdef ALTQ3_COMPAT 222#ifdef ALTQ_KLD 223 altq_module_declref(ifq->altq_type); 224#endif 225#endif 226 227 ifq->altq_type = ALTQT_NONE; 228 ifq->altq_disc = NULL; 229 ifq->altq_enqueue = NULL; 230 ifq->altq_dequeue = NULL; 231 ifq->altq_request = NULL; 232 ifq->altq_clfier = NULL; 233 ifq->altq_classify = NULL; 234 ifq->altq_flags &= ALTQF_CANTCHANGE; 235 236 IFQ_UNLOCK(ifq); 237 return 0; 238} 239 240int 241altq_enable(ifq) 242 struct ifaltq *ifq; 243{ 244 int s; 245 246 IFQ_LOCK(ifq); 247 248 if (!ALTQ_IS_READY(ifq)) { 249 IFQ_UNLOCK(ifq); 250 return ENXIO; 251 } 252 if (ALTQ_IS_ENABLED(ifq)) { 253 IFQ_UNLOCK(ifq); 254 return 0; 255 } 256 257#ifdef __NetBSD__ 258 s = splnet(); 259#else 260 s = splimp(); 261#endif 262 IFQ_PURGE_NOLOCK(ifq); 263 ASSERT(ifq->ifq_len == 0); 264 ifq->ifq_drv_maxlen = 0; /* disable bulk dequeue */ 265 ifq->altq_flags |= ALTQF_ENABLED; 266 if (ifq->altq_clfier != NULL) 267 ifq->altq_flags |= ALTQF_CLASSIFY; 268 splx(s); 269 270 IFQ_UNLOCK(ifq); 271 return 0; 272} 273 274int 275altq_disable(ifq) 276 struct ifaltq *ifq; 277{ 278 int s; 279 280 IFQ_LOCK(ifq); 281 if (!ALTQ_IS_ENABLED(ifq)) { 282 IFQ_UNLOCK(ifq); 283 return 0; 284 } 285 286#ifdef __NetBSD__ 287 s = splnet(); 288#else 289 s = splimp(); 290#endif 291 IFQ_PURGE_NOLOCK(ifq); 292 ASSERT(ifq->ifq_len == 0); 293 ifq->altq_flags &= ~(ALTQF_ENABLED|ALTQF_CLASSIFY); 294 splx(s); 295 296 IFQ_UNLOCK(ifq); 297 return 0; 298} 299 300#ifdef ALTQ_DEBUG 301void 302altq_assert(file, line, failedexpr) 303 const char *file, *failedexpr; 304 int line; 305{ 306 (void)printf("altq assertion \"%s\" failed: file \"%s\", line %d\n", 307 failedexpr, file, line); 308 panic("altq assertion"); 309 /* NOTREACHED */ 310} 311#endif 312 313/* 314 * internal representation of token bucket parameters 315 * rate: byte_per_unittime << 32 316 * (((bits_per_sec) / 8) << 32) / machclk_freq 317 * depth: byte << 32 318 * 319 */ 320#define TBR_SHIFT 32 321#define TBR_SCALE(x) ((int64_t)(x) << TBR_SHIFT) 322#define TBR_UNSCALE(x) ((x) >> TBR_SHIFT) 323 324static struct mbuf * 325tbr_dequeue(ifq, op) 326 struct ifaltq *ifq; 327 int op; 328{ 329 struct tb_regulator *tbr; 330 struct mbuf *m; 331 int64_t interval; 332 u_int64_t now; 333 334 IFQ_LOCK_ASSERT(ifq); 335 tbr = ifq->altq_tbr; 336 if (op == ALTDQ_REMOVE && tbr->tbr_lastop == ALTDQ_POLL) { 337 /* if this is a remove after poll, bypass tbr check */ 338 } else { 339 /* update token only when it is negative */ 340 if (tbr->tbr_token <= 0) { 341 now = read_machclk(); 342 interval = now - tbr->tbr_last; 343 if (interval >= tbr->tbr_filluptime) 344 tbr->tbr_token = tbr->tbr_depth; 345 else { 346 tbr->tbr_token += interval * tbr->tbr_rate; 347 if (tbr->tbr_token > tbr->tbr_depth) 348 tbr->tbr_token = tbr->tbr_depth; 349 } 350 tbr->tbr_last = now; 351 } 352 /* if token is still negative, don't allow dequeue */ 353 if (tbr->tbr_token <= 0) 354 return (NULL); 355 } 356 357 if (ALTQ_IS_ENABLED(ifq)) 358 m = (*ifq->altq_dequeue)(ifq, op); 359 else { 360 if (op == ALTDQ_POLL) 361 _IF_POLL(ifq, m); 362 else 363 _IF_DEQUEUE(ifq, m); 364 } 365 366 if (m != NULL && op == ALTDQ_REMOVE) 367 tbr->tbr_token -= TBR_SCALE(m_pktlen(m)); 368 tbr->tbr_lastop = op; 369 return (m); 370} 371 372/* 373 * set a token bucket regulator. 374 * if the specified rate is zero, the token bucket regulator is deleted. 375 */ 376int 377tbr_set(ifq, profile) 378 struct ifaltq *ifq; 379 struct tb_profile *profile; 380{ 381 struct tb_regulator *tbr, *otbr; 382 383 if (tbr_dequeue_ptr == NULL) 384 tbr_dequeue_ptr = tbr_dequeue; 385 386 if (machclk_freq == 0) 387 init_machclk(); 388 if (machclk_freq == 0) { 389 printf("tbr_set: no cpu clock available!\n"); 390 return (ENXIO); 391 } 392 393 IFQ_LOCK(ifq); 394 if (profile->rate == 0) { 395 /* delete this tbr */ 396 if ((tbr = ifq->altq_tbr) == NULL) { 397 IFQ_UNLOCK(ifq); 398 return (ENOENT); 399 } 400 ifq->altq_tbr = NULL; 401 free(tbr, M_DEVBUF); 402 IFQ_UNLOCK(ifq); 403 return (0); 404 } 405 406 tbr = malloc(sizeof(struct tb_regulator), M_DEVBUF, M_NOWAIT | M_ZERO); 407 if (tbr == NULL) { 408 IFQ_UNLOCK(ifq); 409 return (ENOMEM); 410 } 411 412 tbr->tbr_rate = TBR_SCALE(profile->rate / 8) / machclk_freq; 413 tbr->tbr_depth = TBR_SCALE(profile->depth); 414 if (tbr->tbr_rate > 0) 415 tbr->tbr_filluptime = tbr->tbr_depth / tbr->tbr_rate; 416 else 417 tbr->tbr_filluptime = 0xffffffffffffffffLL; 418 tbr->tbr_token = tbr->tbr_depth; 419 tbr->tbr_last = read_machclk(); 420 tbr->tbr_lastop = ALTDQ_REMOVE; 421 422 otbr = ifq->altq_tbr; 423 ifq->altq_tbr = tbr; /* set the new tbr */ 424 425 if (otbr != NULL) 426 free(otbr, M_DEVBUF); 427 else { 428 if (tbr_timer == 0) { 429 CALLOUT_RESET(&tbr_callout, 1, tbr_timeout, (void *)0); 430 tbr_timer = 1; 431 } 432 } 433 IFQ_UNLOCK(ifq); 434 return (0); 435} 436 437/* 438 * tbr_timeout goes through the interface list, and kicks the drivers 439 * if necessary. 440 * 441 * MPSAFE 442 */ 443static void 444tbr_timeout(arg) 445 void *arg; 446{ 447#ifdef __FreeBSD__ 448 VNET_ITERATOR_DECL(vnet_iter); 449#endif 450 struct ifnet *ifp; 451 int active, s; 452 453 active = 0; 454#ifdef __NetBSD__ 455 s = splnet(); 456#else 457 s = splimp(); 458#endif 459#ifdef __FreeBSD__ 460 IFNET_RLOCK_NOSLEEP(); 461 VNET_LIST_RLOCK_NOSLEEP(); 462 VNET_FOREACH(vnet_iter) { 463 CURVNET_SET(vnet_iter); 464#endif 465 for (ifp = TAILQ_FIRST(&V_ifnet); ifp; 466 ifp = TAILQ_NEXT(ifp, if_list)) { 467 /* read from if_snd unlocked */ 468 if (!TBR_IS_ENABLED(&ifp->if_snd)) 469 continue; 470 active++; 471 if (!IFQ_IS_EMPTY(&ifp->if_snd) && 472 ifp->if_start != NULL) 473 (*ifp->if_start)(ifp); 474 } 475#ifdef __FreeBSD__ 476 CURVNET_RESTORE(); 477 } 478 VNET_LIST_RUNLOCK_NOSLEEP(); 479 IFNET_RUNLOCK_NOSLEEP(); 480#endif 481 splx(s); 482 if (active > 0) 483 CALLOUT_RESET(&tbr_callout, 1, tbr_timeout, (void *)0); 484 else 485 tbr_timer = 0; /* don't need tbr_timer anymore */ 486} 487 488/* 489 * get token bucket regulator profile 490 */ 491int 492tbr_get(ifq, profile) 493 struct ifaltq *ifq; 494 struct tb_profile *profile; 495{ 496 struct tb_regulator *tbr; 497 498 IFQ_LOCK(ifq); 499 if ((tbr = ifq->altq_tbr) == NULL) { 500 profile->rate = 0; 501 profile->depth = 0; 502 } else { 503 profile->rate = 504 (u_int)TBR_UNSCALE(tbr->tbr_rate * 8 * machclk_freq); 505 profile->depth = (u_int)TBR_UNSCALE(tbr->tbr_depth); 506 } 507 IFQ_UNLOCK(ifq); 508 return (0); 509} 510 511/* 512 * attach a discipline to the interface. if one already exists, it is 513 * overridden. 514 * Locking is done in the discipline specific attach functions. Basically 515 * they call back to altq_attach which takes care of the attach and locking. 516 */ 517int 518altq_pfattach(struct pf_altq *a) 519{ 520 int error = 0; 521 522 switch (a->scheduler) { 523 case ALTQT_NONE: 524 break; 525#ifdef ALTQ_CBQ 526 case ALTQT_CBQ: 527 error = cbq_pfattach(a); 528 break; 529#endif 530#ifdef ALTQ_PRIQ 531 case ALTQT_PRIQ: 532 error = priq_pfattach(a); 533 break; 534#endif 535#ifdef ALTQ_HFSC 536 case ALTQT_HFSC: 537 error = hfsc_pfattach(a); 538 break; 539#endif 540 default: 541 error = ENXIO; 542 } 543 544 return (error); 545} 546 547/* 548 * detach a discipline from the interface. 549 * it is possible that the discipline was already overridden by another 550 * discipline. 551 */ 552int 553altq_pfdetach(struct pf_altq *a) 554{ 555 struct ifnet *ifp; 556 int s, error = 0; 557 558 if ((ifp = ifunit(a->ifname)) == NULL) 559 return (EINVAL); 560 561 /* if this discipline is no longer referenced, just return */ 562 /* read unlocked from if_snd */ 563 if (a->altq_disc == NULL || a->altq_disc != ifp->if_snd.altq_disc) 564 return (0); 565 566#ifdef __NetBSD__ 567 s = splnet(); 568#else 569 s = splimp(); 570#endif 571 /* read unlocked from if_snd, _disable and _detach take care */ 572 if (ALTQ_IS_ENABLED(&ifp->if_snd)) 573 error = altq_disable(&ifp->if_snd); 574 if (error == 0) 575 error = altq_detach(&ifp->if_snd); 576 splx(s); 577 578 return (error); 579} 580 581/* 582 * add a discipline or a queue 583 * Locking is done in the discipline specific functions with regards to 584 * malloc with WAITOK, also it is not yet clear which lock to use. 585 */ 586int 587altq_add(struct pf_altq *a) 588{ 589 int error = 0; 590 591 if (a->qname[0] != 0) 592 return (altq_add_queue(a)); 593 594 if (machclk_freq == 0) 595 init_machclk(); 596 if (machclk_freq == 0) 597 panic("altq_add: no cpu clock"); 598 599 switch (a->scheduler) { 600#ifdef ALTQ_CBQ 601 case ALTQT_CBQ: 602 error = cbq_add_altq(a); 603 break; 604#endif 605#ifdef ALTQ_PRIQ 606 case ALTQT_PRIQ: 607 error = priq_add_altq(a); 608 break; 609#endif 610#ifdef ALTQ_HFSC 611 case ALTQT_HFSC: 612 error = hfsc_add_altq(a); 613 break; 614#endif 615 default: 616 error = ENXIO; 617 } 618 619 return (error); 620} 621 622/* 623 * remove a discipline or a queue 624 * It is yet unclear what lock to use to protect this operation, the 625 * discipline specific functions will determine and grab it 626 */ 627int 628altq_remove(struct pf_altq *a) 629{ 630 int error = 0; 631 632 if (a->qname[0] != 0) 633 return (altq_remove_queue(a)); 634 635 switch (a->scheduler) { 636#ifdef ALTQ_CBQ 637 case ALTQT_CBQ: 638 error = cbq_remove_altq(a); 639 break; 640#endif 641#ifdef ALTQ_PRIQ 642 case ALTQT_PRIQ: 643 error = priq_remove_altq(a); 644 break; 645#endif 646#ifdef ALTQ_HFSC 647 case ALTQT_HFSC: 648 error = hfsc_remove_altq(a); 649 break; 650#endif 651 default: 652 error = ENXIO; 653 } 654 655 return (error); 656} 657 658/* 659 * add a queue to the discipline 660 * It is yet unclear what lock to use to protect this operation, the 661 * discipline specific functions will determine and grab it 662 */ 663int 664altq_add_queue(struct pf_altq *a) 665{ 666 int error = 0; 667 668 switch (a->scheduler) { 669#ifdef ALTQ_CBQ 670 case ALTQT_CBQ: 671 error = cbq_add_queue(a); 672 break; 673#endif 674#ifdef ALTQ_PRIQ 675 case ALTQT_PRIQ: 676 error = priq_add_queue(a); 677 break; 678#endif 679#ifdef ALTQ_HFSC 680 case ALTQT_HFSC: 681 error = hfsc_add_queue(a); 682 break; 683#endif 684 default: 685 error = ENXIO; 686 } 687 688 return (error); 689} 690 691/* 692 * remove a queue from the discipline 693 * It is yet unclear what lock to use to protect this operation, the 694 * discipline specific functions will determine and grab it 695 */ 696int 697altq_remove_queue(struct pf_altq *a) 698{ 699 int error = 0; 700 701 switch (a->scheduler) { 702#ifdef ALTQ_CBQ 703 case ALTQT_CBQ: 704 error = cbq_remove_queue(a); 705 break; 706#endif 707#ifdef ALTQ_PRIQ 708 case ALTQT_PRIQ: 709 error = priq_remove_queue(a); 710 break; 711#endif 712#ifdef ALTQ_HFSC 713 case ALTQT_HFSC: 714 error = hfsc_remove_queue(a); 715 break; 716#endif 717 default: 718 error = ENXIO; 719 } 720 721 return (error); 722} 723 724/* 725 * get queue statistics 726 * Locking is done in the discipline specific functions with regards to 727 * copyout operations, also it is not yet clear which lock to use. 728 */ 729int 730altq_getqstats(struct pf_altq *a, void *ubuf, int *nbytes) 731{ 732 int error = 0; 733 734 switch (a->scheduler) { 735#ifdef ALTQ_CBQ 736 case ALTQT_CBQ: 737 error = cbq_getqstats(a, ubuf, nbytes); 738 break; 739#endif 740#ifdef ALTQ_PRIQ 741 case ALTQT_PRIQ: 742 error = priq_getqstats(a, ubuf, nbytes); 743 break; 744#endif 745#ifdef ALTQ_HFSC 746 case ALTQT_HFSC: 747 error = hfsc_getqstats(a, ubuf, nbytes); 748 break; 749#endif 750 default: 751 error = ENXIO; 752 } 753 754 return (error); 755} 756 757/* 758 * read and write diffserv field in IPv4 or IPv6 header 759 */ 760u_int8_t 761read_dsfield(m, pktattr) 762 struct mbuf *m; 763 struct altq_pktattr *pktattr; 764{ 765 struct mbuf *m0; 766 u_int8_t ds_field = 0; 767 768 if (pktattr == NULL || 769 (pktattr->pattr_af != AF_INET && pktattr->pattr_af != AF_INET6)) 770 return ((u_int8_t)0); 771 772 /* verify that pattr_hdr is within the mbuf data */ 773 for (m0 = m; m0 != NULL; m0 = m0->m_next) 774 if ((pktattr->pattr_hdr >= m0->m_data) && 775 (pktattr->pattr_hdr < m0->m_data + m0->m_len)) 776 break; 777 if (m0 == NULL) { 778 /* ick, pattr_hdr is stale */ 779 pktattr->pattr_af = AF_UNSPEC; 780#ifdef ALTQ_DEBUG 781 printf("read_dsfield: can't locate header!\n"); 782#endif 783 return ((u_int8_t)0); 784 } 785 786 if (pktattr->pattr_af == AF_INET) { 787 struct ip *ip = (struct ip *)pktattr->pattr_hdr; 788 789 if (ip->ip_v != 4) 790 return ((u_int8_t)0); /* version mismatch! */ 791 ds_field = ip->ip_tos; 792 } 793#ifdef INET6 794 else if (pktattr->pattr_af == AF_INET6) { 795 struct ip6_hdr *ip6 = (struct ip6_hdr *)pktattr->pattr_hdr; 796 u_int32_t flowlabel; 797 798 flowlabel = ntohl(ip6->ip6_flow); 799 if ((flowlabel >> 28) != 6) 800 return ((u_int8_t)0); /* version mismatch! */ 801 ds_field = (flowlabel >> 20) & 0xff; 802 } 803#endif 804 return (ds_field); 805} 806 807void 808write_dsfield(struct mbuf *m, struct altq_pktattr *pktattr, u_int8_t dsfield) 809{ 810 struct mbuf *m0; 811 812 if (pktattr == NULL || 813 (pktattr->pattr_af != AF_INET && pktattr->pattr_af != AF_INET6)) 814 return; 815 816 /* verify that pattr_hdr is within the mbuf data */ 817 for (m0 = m; m0 != NULL; m0 = m0->m_next) 818 if ((pktattr->pattr_hdr >= m0->m_data) && 819 (pktattr->pattr_hdr < m0->m_data + m0->m_len)) 820 break; 821 if (m0 == NULL) { 822 /* ick, pattr_hdr is stale */ 823 pktattr->pattr_af = AF_UNSPEC; 824#ifdef ALTQ_DEBUG 825 printf("write_dsfield: can't locate header!\n"); 826#endif 827 return; 828 } 829 830 if (pktattr->pattr_af == AF_INET) { 831 struct ip *ip = (struct ip *)pktattr->pattr_hdr; 832 u_int8_t old; 833 int32_t sum; 834 835 if (ip->ip_v != 4) 836 return; /* version mismatch! */ 837 old = ip->ip_tos; 838 dsfield |= old & 3; /* leave CU bits */ 839 if (old == dsfield) 840 return; 841 ip->ip_tos = dsfield; 842 /* 843 * update checksum (from RFC1624) 844 * HC' = ~(~HC + ~m + m') 845 */ 846 sum = ~ntohs(ip->ip_sum) & 0xffff; 847 sum += 0xff00 + (~old & 0xff) + dsfield; 848 sum = (sum >> 16) + (sum & 0xffff); 849 sum += (sum >> 16); /* add carry */ 850 851 ip->ip_sum = htons(~sum & 0xffff); 852 } 853#ifdef INET6 854 else if (pktattr->pattr_af == AF_INET6) { 855 struct ip6_hdr *ip6 = (struct ip6_hdr *)pktattr->pattr_hdr; 856 u_int32_t flowlabel; 857 858 flowlabel = ntohl(ip6->ip6_flow); 859 if ((flowlabel >> 28) != 6) 860 return; /* version mismatch! */ 861 flowlabel = (flowlabel & 0xf03fffff) | (dsfield << 20); 862 ip6->ip6_flow = htonl(flowlabel); 863 } 864#endif 865 return; 866} 867 868 869/* 870 * high resolution clock support taking advantage of a machine dependent 871 * high resolution time counter (e.g., timestamp counter of intel pentium). 872 * we assume 873 * - 64-bit-long monotonically-increasing counter 874 * - frequency range is 100M-4GHz (CPU speed) 875 */ 876/* if pcc is not available or disabled, emulate 256MHz using microtime() */ 877#define MACHCLK_SHIFT 8 878 879int machclk_usepcc; 880u_int32_t machclk_freq; 881u_int32_t machclk_per_tick; 882 883#if defined(__i386__) && defined(__NetBSD__) 884extern u_int64_t cpu_tsc_freq; 885#endif 886 887#if (__FreeBSD_version >= 700035) 888/* Update TSC freq with the value indicated by the caller. */ 889static void 890tsc_freq_changed(void *arg, const struct cf_level *level, int status) 891{ 892 /* If there was an error during the transition, don't do anything. */ 893 if (status != 0) 894 return; 895 896#if (__FreeBSD_version >= 701102) && (defined(__amd64__) || defined(__i386__)) 897 /* If TSC is P-state invariant, don't do anything. */ 898 if (tsc_is_invariant) 899 return; 900#endif 901 902 /* Total setting for this level gives the new frequency in MHz. */ 903 init_machclk(); 904} 905EVENTHANDLER_DEFINE(cpufreq_post_change, tsc_freq_changed, NULL, 906 EVENTHANDLER_PRI_LAST); 907#endif /* __FreeBSD_version >= 700035 */ 908 909static void 910init_machclk_setup(void) 911{ 912#if (__FreeBSD_version >= 600000) 913 callout_init(&tbr_callout, 0); 914#endif 915 916 machclk_usepcc = 1; 917 918#if (!defined(__amd64__) && !defined(__i386__)) || defined(ALTQ_NOPCC) 919 machclk_usepcc = 0; 920#endif 921#if defined(__FreeBSD__) && defined(SMP) 922 machclk_usepcc = 0; 923#endif 924#if defined(__NetBSD__) && defined(MULTIPROCESSOR) 925 machclk_usepcc = 0; 926#endif 927#if defined(__amd64__) || defined(__i386__) 928 /* check if TSC is available */ 929#ifdef __FreeBSD__ 930 if ((cpu_feature & CPUID_TSC) == 0 || 931 atomic_load_acq_64(&tsc_freq) == 0) 932#else 933 if ((cpu_feature & CPUID_TSC) == 0) 934#endif 935 machclk_usepcc = 0; 936#endif 937} 938 939void 940init_machclk(void) 941{ 942 static int called; 943 944 /* Call one-time initialization function. */ 945 if (!called) { 946 init_machclk_setup(); 947 called = 1; 948 } 949 950 if (machclk_usepcc == 0) { 951 /* emulate 256MHz using microtime() */ 952 machclk_freq = 1000000 << MACHCLK_SHIFT; 953 machclk_per_tick = machclk_freq / hz; 954#ifdef ALTQ_DEBUG 955 printf("altq: emulate %uHz cpu clock\n", machclk_freq); 956#endif 957 return; 958 } 959 960 /* 961 * if the clock frequency (of Pentium TSC or Alpha PCC) is 962 * accessible, just use it. 963 */ 964#if defined(__amd64__) || defined(__i386__) 965#ifdef __FreeBSD__ 966 machclk_freq = atomic_load_acq_64(&tsc_freq); 967#elif defined(__NetBSD__) 968 machclk_freq = (u_int32_t)cpu_tsc_freq; 969#elif defined(__OpenBSD__) && (defined(I586_CPU) || defined(I686_CPU)) 970 machclk_freq = pentium_mhz * 1000000; 971#endif 972#endif 973 974 /* 975 * if we don't know the clock frequency, measure it. 976 */ 977 if (machclk_freq == 0) { 978 static int wait; 979 struct timeval tv_start, tv_end; 980 u_int64_t start, end, diff; 981 int timo; 982 983 microtime(&tv_start); 984 start = read_machclk(); 985 timo = hz; /* 1 sec */ 986 (void)tsleep(&wait, PWAIT | PCATCH, "init_machclk", timo); 987 microtime(&tv_end); 988 end = read_machclk(); 989 diff = (u_int64_t)(tv_end.tv_sec - tv_start.tv_sec) * 1000000 990 + tv_end.tv_usec - tv_start.tv_usec; 991 if (diff != 0) 992 machclk_freq = (u_int)((end - start) * 1000000 / diff); 993 } 994 995 machclk_per_tick = machclk_freq / hz; 996 997#ifdef ALTQ_DEBUG 998 printf("altq: CPU clock: %uHz\n", machclk_freq); 999#endif 1000} 1001 1002#if defined(__OpenBSD__) && defined(__i386__) 1003static __inline u_int64_t 1004rdtsc(void) 1005{ 1006 u_int64_t rv; 1007 __asm __volatile(".byte 0x0f, 0x31" : "=A" (rv)); 1008 return (rv); 1009} 1010#endif /* __OpenBSD__ && __i386__ */ 1011 1012u_int64_t 1013read_machclk(void) 1014{ 1015 u_int64_t val; 1016 1017 if (machclk_usepcc) { 1018#if defined(__amd64__) || defined(__i386__) 1019 val = rdtsc(); 1020#else 1021 panic("read_machclk"); 1022#endif 1023 } else { 1024 struct timeval tv; 1025 1026 microtime(&tv); 1027 val = (((u_int64_t)(tv.tv_sec - boottime.tv_sec) * 1000000 1028 + tv.tv_usec) << MACHCLK_SHIFT); 1029 } 1030 return (val); 1031} 1032 1033#ifdef ALTQ3_CLFIER_COMPAT 1034 1035#ifndef IPPROTO_ESP 1036#define IPPROTO_ESP 50 /* encapsulating security payload */ 1037#endif 1038#ifndef IPPROTO_AH 1039#define IPPROTO_AH 51 /* authentication header */ 1040#endif 1041 1042/* 1043 * extract flow information from a given packet. 1044 * filt_mask shows flowinfo fields required. 1045 * we assume the ip header is in one mbuf, and addresses and ports are 1046 * in network byte order. 1047 */ 1048int 1049altq_extractflow(m, af, flow, filt_bmask) 1050 struct mbuf *m; 1051 int af; 1052 struct flowinfo *flow; 1053 u_int32_t filt_bmask; 1054{ 1055 1056 switch (af) { 1057 case PF_INET: { 1058 struct flowinfo_in *fin; 1059 struct ip *ip; 1060 1061 ip = mtod(m, struct ip *); 1062 1063 if (ip->ip_v != 4) 1064 break; 1065 1066 fin = (struct flowinfo_in *)flow; 1067 fin->fi_len = sizeof(struct flowinfo_in); 1068 fin->fi_family = AF_INET; 1069 1070 fin->fi_proto = ip->ip_p; 1071 fin->fi_tos = ip->ip_tos; 1072 1073 fin->fi_src.s_addr = ip->ip_src.s_addr; 1074 fin->fi_dst.s_addr = ip->ip_dst.s_addr; 1075 1076 if (filt_bmask & FIMB4_PORTS) 1077 /* if port info is required, extract port numbers */ 1078 extract_ports4(m, ip, fin); 1079 else { 1080 fin->fi_sport = 0; 1081 fin->fi_dport = 0; 1082 fin->fi_gpi = 0; 1083 } 1084 return (1); 1085 } 1086 1087#ifdef INET6 1088 case PF_INET6: { 1089 struct flowinfo_in6 *fin6; 1090 struct ip6_hdr *ip6; 1091 1092 ip6 = mtod(m, struct ip6_hdr *); 1093 /* should we check the ip version? */ 1094 1095 fin6 = (struct flowinfo_in6 *)flow; 1096 fin6->fi6_len = sizeof(struct flowinfo_in6); 1097 fin6->fi6_family = AF_INET6; 1098 1099 fin6->fi6_proto = ip6->ip6_nxt; 1100 fin6->fi6_tclass = (ntohl(ip6->ip6_flow) >> 20) & 0xff; 1101 1102 fin6->fi6_flowlabel = ip6->ip6_flow & htonl(0x000fffff); 1103 fin6->fi6_src = ip6->ip6_src; 1104 fin6->fi6_dst = ip6->ip6_dst; 1105 1106 if ((filt_bmask & FIMB6_PORTS) || 1107 ((filt_bmask & FIMB6_PROTO) 1108 && ip6->ip6_nxt > IPPROTO_IPV6)) 1109 /* 1110 * if port info is required, or proto is required 1111 * but there are option headers, extract port 1112 * and protocol numbers. 1113 */ 1114 extract_ports6(m, ip6, fin6); 1115 else { 1116 fin6->fi6_sport = 0; 1117 fin6->fi6_dport = 0; 1118 fin6->fi6_gpi = 0; 1119 } 1120 return (1); 1121 } 1122#endif /* INET6 */ 1123 1124 default: 1125 break; 1126 } 1127 1128 /* failed */ 1129 flow->fi_len = sizeof(struct flowinfo); 1130 flow->fi_family = AF_UNSPEC; 1131 return (0); 1132} 1133 1134/* 1135 * helper routine to extract port numbers 1136 */ 1137/* structure for ipsec and ipv6 option header template */ 1138struct _opt6 { 1139 u_int8_t opt6_nxt; /* next header */ 1140 u_int8_t opt6_hlen; /* header extension length */ 1141 u_int16_t _pad; 1142 u_int32_t ah_spi; /* security parameter index 1143 for authentication header */ 1144}; 1145 1146/* 1147 * extract port numbers from a ipv4 packet. 1148 */ 1149static int 1150extract_ports4(m, ip, fin) 1151 struct mbuf *m; 1152 struct ip *ip; 1153 struct flowinfo_in *fin; 1154{ 1155 struct mbuf *m0; 1156 u_short ip_off; 1157 u_int8_t proto; 1158 int off; 1159 1160 fin->fi_sport = 0; 1161 fin->fi_dport = 0; 1162 fin->fi_gpi = 0; 1163 1164 ip_off = ntohs(ip->ip_off); 1165 /* if it is a fragment, try cached fragment info */ 1166 if (ip_off & IP_OFFMASK) { 1167 ip4f_lookup(ip, fin); 1168 return (1); 1169 } 1170 1171 /* locate the mbuf containing the protocol header */ 1172 for (m0 = m; m0 != NULL; m0 = m0->m_next) 1173 if (((caddr_t)ip >= m0->m_data) && 1174 ((caddr_t)ip < m0->m_data + m0->m_len)) 1175 break; 1176 if (m0 == NULL) { 1177#ifdef ALTQ_DEBUG 1178 printf("extract_ports4: can't locate header! ip=%p\n", ip); 1179#endif 1180 return (0); 1181 } 1182 off = ((caddr_t)ip - m0->m_data) + (ip->ip_hl << 2); 1183 proto = ip->ip_p; 1184 1185#ifdef ALTQ_IPSEC 1186 again: 1187#endif 1188 while (off >= m0->m_len) { 1189 off -= m0->m_len; 1190 m0 = m0->m_next; 1191 if (m0 == NULL) 1192 return (0); /* bogus ip_hl! */ 1193 } 1194 if (m0->m_len < off + 4) 1195 return (0); 1196 1197 switch (proto) { 1198 case IPPROTO_TCP: 1199 case IPPROTO_UDP: { 1200 struct udphdr *udp; 1201 1202 udp = (struct udphdr *)(mtod(m0, caddr_t) + off); 1203 fin->fi_sport = udp->uh_sport; 1204 fin->fi_dport = udp->uh_dport; 1205 fin->fi_proto = proto; 1206 } 1207 break; 1208 1209#ifdef ALTQ_IPSEC 1210 case IPPROTO_ESP: 1211 if (fin->fi_gpi == 0){ 1212 u_int32_t *gpi; 1213 1214 gpi = (u_int32_t *)(mtod(m0, caddr_t) + off); 1215 fin->fi_gpi = *gpi; 1216 } 1217 fin->fi_proto = proto; 1218 break; 1219 1220 case IPPROTO_AH: { 1221 /* get next header and header length */ 1222 struct _opt6 *opt6; 1223 1224 opt6 = (struct _opt6 *)(mtod(m0, caddr_t) + off); 1225 proto = opt6->opt6_nxt; 1226 off += 8 + (opt6->opt6_hlen * 4); 1227 if (fin->fi_gpi == 0 && m0->m_len >= off + 8) 1228 fin->fi_gpi = opt6->ah_spi; 1229 } 1230 /* goto the next header */ 1231 goto again; 1232#endif /* ALTQ_IPSEC */ 1233 1234 default: 1235 fin->fi_proto = proto; 1236 return (0); 1237 } 1238 1239 /* if this is a first fragment, cache it. */ 1240 if (ip_off & IP_MF) 1241 ip4f_cache(ip, fin); 1242 1243 return (1); 1244} 1245 1246#ifdef INET6 1247static int 1248extract_ports6(m, ip6, fin6) 1249 struct mbuf *m; 1250 struct ip6_hdr *ip6; 1251 struct flowinfo_in6 *fin6; 1252{ 1253 struct mbuf *m0; 1254 int off; 1255 u_int8_t proto; 1256 1257 fin6->fi6_gpi = 0; 1258 fin6->fi6_sport = 0; 1259 fin6->fi6_dport = 0; 1260 1261 /* locate the mbuf containing the protocol header */ 1262 for (m0 = m; m0 != NULL; m0 = m0->m_next) 1263 if (((caddr_t)ip6 >= m0->m_data) && 1264 ((caddr_t)ip6 < m0->m_data + m0->m_len)) 1265 break; 1266 if (m0 == NULL) { 1267#ifdef ALTQ_DEBUG 1268 printf("extract_ports6: can't locate header! ip6=%p\n", ip6); 1269#endif 1270 return (0); 1271 } 1272 off = ((caddr_t)ip6 - m0->m_data) + sizeof(struct ip6_hdr); 1273 1274 proto = ip6->ip6_nxt; 1275 do { 1276 while (off >= m0->m_len) { 1277 off -= m0->m_len; 1278 m0 = m0->m_next; 1279 if (m0 == NULL) 1280 return (0); 1281 } 1282 if (m0->m_len < off + 4) 1283 return (0); 1284 1285 switch (proto) { 1286 case IPPROTO_TCP: 1287 case IPPROTO_UDP: { 1288 struct udphdr *udp; 1289 1290 udp = (struct udphdr *)(mtod(m0, caddr_t) + off); 1291 fin6->fi6_sport = udp->uh_sport; 1292 fin6->fi6_dport = udp->uh_dport; 1293 fin6->fi6_proto = proto; 1294 } 1295 return (1); 1296 1297 case IPPROTO_ESP: 1298 if (fin6->fi6_gpi == 0) { 1299 u_int32_t *gpi; 1300 1301 gpi = (u_int32_t *)(mtod(m0, caddr_t) + off); 1302 fin6->fi6_gpi = *gpi; 1303 } 1304 fin6->fi6_proto = proto; 1305 return (1); 1306 1307 case IPPROTO_AH: { 1308 /* get next header and header length */ 1309 struct _opt6 *opt6; 1310 1311 opt6 = (struct _opt6 *)(mtod(m0, caddr_t) + off); 1312 if (fin6->fi6_gpi == 0 && m0->m_len >= off + 8) 1313 fin6->fi6_gpi = opt6->ah_spi; 1314 proto = opt6->opt6_nxt; 1315 off += 8 + (opt6->opt6_hlen * 4); 1316 /* goto the next header */ 1317 break; 1318 } 1319 1320 case IPPROTO_HOPOPTS: 1321 case IPPROTO_ROUTING: 1322 case IPPROTO_DSTOPTS: { 1323 /* get next header and header length */ 1324 struct _opt6 *opt6; 1325 1326 opt6 = (struct _opt6 *)(mtod(m0, caddr_t) + off); 1327 proto = opt6->opt6_nxt; 1328 off += (opt6->opt6_hlen + 1) * 8; 1329 /* goto the next header */ 1330 break; 1331 } 1332 1333 case IPPROTO_FRAGMENT: 1334 /* ipv6 fragmentations are not supported yet */ 1335 default: 1336 fin6->fi6_proto = proto; 1337 return (0); 1338 } 1339 } while (1); 1340 /*NOTREACHED*/ 1341} 1342#endif /* INET6 */ 1343 1344/* 1345 * altq common classifier 1346 */ 1347int 1348acc_add_filter(classifier, filter, class, phandle) 1349 struct acc_classifier *classifier; 1350 struct flow_filter *filter; 1351 void *class; 1352 u_long *phandle; 1353{ 1354 struct acc_filter *afp, *prev, *tmp; 1355 int i, s; 1356 1357#ifdef INET6 1358 if (filter->ff_flow.fi_family != AF_INET && 1359 filter->ff_flow.fi_family != AF_INET6) 1360 return (EINVAL); 1361#else 1362 if (filter->ff_flow.fi_family != AF_INET) 1363 return (EINVAL); 1364#endif 1365 1366 afp = malloc(sizeof(struct acc_filter), 1367 M_DEVBUF, M_WAITOK); 1368 if (afp == NULL) 1369 return (ENOMEM); 1370 bzero(afp, sizeof(struct acc_filter)); 1371 1372 afp->f_filter = *filter; 1373 afp->f_class = class; 1374 1375 i = ACC_WILDCARD_INDEX; 1376 if (filter->ff_flow.fi_family == AF_INET) { 1377 struct flow_filter *filter4 = &afp->f_filter; 1378 1379 /* 1380 * if address is 0, it's a wildcard. if address mask 1381 * isn't set, use full mask. 1382 */ 1383 if (filter4->ff_flow.fi_dst.s_addr == 0) 1384 filter4->ff_mask.mask_dst.s_addr = 0; 1385 else if (filter4->ff_mask.mask_dst.s_addr == 0) 1386 filter4->ff_mask.mask_dst.s_addr = 0xffffffff; 1387 if (filter4->ff_flow.fi_src.s_addr == 0) 1388 filter4->ff_mask.mask_src.s_addr = 0; 1389 else if (filter4->ff_mask.mask_src.s_addr == 0) 1390 filter4->ff_mask.mask_src.s_addr = 0xffffffff; 1391 1392 /* clear extra bits in addresses */ 1393 filter4->ff_flow.fi_dst.s_addr &= 1394 filter4->ff_mask.mask_dst.s_addr; 1395 filter4->ff_flow.fi_src.s_addr &= 1396 filter4->ff_mask.mask_src.s_addr; 1397 1398 /* 1399 * if dst address is a wildcard, use hash-entry 1400 * ACC_WILDCARD_INDEX. 1401 */ 1402 if (filter4->ff_mask.mask_dst.s_addr != 0xffffffff) 1403 i = ACC_WILDCARD_INDEX; 1404 else 1405 i = ACC_GET_HASH_INDEX(filter4->ff_flow.fi_dst.s_addr); 1406 } 1407#ifdef INET6 1408 else if (filter->ff_flow.fi_family == AF_INET6) { 1409 struct flow_filter6 *filter6 = 1410 (struct flow_filter6 *)&afp->f_filter; 1411#ifndef IN6MASK0 /* taken from kame ipv6 */ 1412#define IN6MASK0 {{{ 0, 0, 0, 0 }}} 1413#define IN6MASK128 {{{ 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }}} 1414 const struct in6_addr in6mask0 = IN6MASK0; 1415 const struct in6_addr in6mask128 = IN6MASK128; 1416#endif 1417 1418 if (IN6_IS_ADDR_UNSPECIFIED(&filter6->ff_flow6.fi6_dst)) 1419 filter6->ff_mask6.mask6_dst = in6mask0; 1420 else if (IN6_IS_ADDR_UNSPECIFIED(&filter6->ff_mask6.mask6_dst)) 1421 filter6->ff_mask6.mask6_dst = in6mask128; 1422 if (IN6_IS_ADDR_UNSPECIFIED(&filter6->ff_flow6.fi6_src)) 1423 filter6->ff_mask6.mask6_src = in6mask0; 1424 else if (IN6_IS_ADDR_UNSPECIFIED(&filter6->ff_mask6.mask6_src)) 1425 filter6->ff_mask6.mask6_src = in6mask128; 1426 1427 /* clear extra bits in addresses */ 1428 for (i = 0; i < 16; i++) 1429 filter6->ff_flow6.fi6_dst.s6_addr[i] &= 1430 filter6->ff_mask6.mask6_dst.s6_addr[i]; 1431 for (i = 0; i < 16; i++) 1432 filter6->ff_flow6.fi6_src.s6_addr[i] &= 1433 filter6->ff_mask6.mask6_src.s6_addr[i]; 1434 1435 if (filter6->ff_flow6.fi6_flowlabel == 0) 1436 i = ACC_WILDCARD_INDEX; 1437 else 1438 i = ACC_GET_HASH_INDEX(filter6->ff_flow6.fi6_flowlabel); 1439 } 1440#endif /* INET6 */ 1441 1442 afp->f_handle = get_filt_handle(classifier, i); 1443 1444 /* update filter bitmask */ 1445 afp->f_fbmask = filt2fibmask(filter); 1446 classifier->acc_fbmask |= afp->f_fbmask; 1447 1448 /* 1449 * add this filter to the filter list. 1450 * filters are ordered from the highest rule number. 1451 */ 1452#ifdef __NetBSD__ 1453 s = splnet(); 1454#else 1455 s = splimp(); 1456#endif 1457 prev = NULL; 1458 LIST_FOREACH(tmp, &classifier->acc_filters[i], f_chain) { 1459 if (tmp->f_filter.ff_ruleno > afp->f_filter.ff_ruleno) 1460 prev = tmp; 1461 else 1462 break; 1463 } 1464 if (prev == NULL) 1465 LIST_INSERT_HEAD(&classifier->acc_filters[i], afp, f_chain); 1466 else 1467 LIST_INSERT_AFTER(prev, afp, f_chain); 1468 splx(s); 1469 1470 *phandle = afp->f_handle; 1471 return (0); 1472} 1473 1474int 1475acc_delete_filter(classifier, handle) 1476 struct acc_classifier *classifier; 1477 u_long handle; 1478{ 1479 struct acc_filter *afp; 1480 int s; 1481 1482 if ((afp = filth_to_filtp(classifier, handle)) == NULL) 1483 return (EINVAL); 1484 1485#ifdef __NetBSD__ 1486 s = splnet(); 1487#else 1488 s = splimp(); 1489#endif 1490 LIST_REMOVE(afp, f_chain); 1491 splx(s); 1492 1493 free(afp, M_DEVBUF); 1494 1495 /* todo: update filt_bmask */ 1496 1497 return (0); 1498} 1499 1500/* 1501 * delete filters referencing to the specified class. 1502 * if the all flag is not 0, delete all the filters. 1503 */ 1504int 1505acc_discard_filters(classifier, class, all) 1506 struct acc_classifier *classifier; 1507 void *class; 1508 int all; 1509{ 1510 struct acc_filter *afp; 1511 int i, s; 1512 1513#ifdef __NetBSD__ 1514 s = splnet(); 1515#else 1516 s = splimp(); 1517#endif 1518 for (i = 0; i < ACC_FILTER_TABLESIZE; i++) { 1519 do { 1520 LIST_FOREACH(afp, &classifier->acc_filters[i], f_chain) 1521 if (all || afp->f_class == class) { 1522 LIST_REMOVE(afp, f_chain); 1523 free(afp, M_DEVBUF); 1524 /* start again from the head */ 1525 break; 1526 } 1527 } while (afp != NULL); 1528 } 1529 splx(s); 1530 1531 if (all) 1532 classifier->acc_fbmask = 0; 1533 1534 return (0); 1535} 1536 1537void * 1538acc_classify(clfier, m, af) 1539 void *clfier; 1540 struct mbuf *m; 1541 int af; 1542{ 1543 struct acc_classifier *classifier; 1544 struct flowinfo flow; 1545 struct acc_filter *afp; 1546 int i; 1547 1548 classifier = (struct acc_classifier *)clfier; 1549 altq_extractflow(m, af, &flow, classifier->acc_fbmask); 1550 1551 if (flow.fi_family == AF_INET) { 1552 struct flowinfo_in *fp = (struct flowinfo_in *)&flow; 1553 1554 if ((classifier->acc_fbmask & FIMB4_ALL) == FIMB4_TOS) { 1555 /* only tos is used */ 1556 LIST_FOREACH(afp, 1557 &classifier->acc_filters[ACC_WILDCARD_INDEX], 1558 f_chain) 1559 if (apply_tosfilter4(afp->f_fbmask, 1560 &afp->f_filter, fp)) 1561 /* filter matched */ 1562 return (afp->f_class); 1563 } else if ((classifier->acc_fbmask & 1564 (~(FIMB4_PROTO|FIMB4_SPORT|FIMB4_DPORT) & FIMB4_ALL)) 1565 == 0) { 1566 /* only proto and ports are used */ 1567 LIST_FOREACH(afp, 1568 &classifier->acc_filters[ACC_WILDCARD_INDEX], 1569 f_chain) 1570 if (apply_ppfilter4(afp->f_fbmask, 1571 &afp->f_filter, fp)) 1572 /* filter matched */ 1573 return (afp->f_class); 1574 } else { 1575 /* get the filter hash entry from its dest address */ 1576 i = ACC_GET_HASH_INDEX(fp->fi_dst.s_addr); 1577 do { 1578 /* 1579 * go through this loop twice. first for dst 1580 * hash, second for wildcards. 1581 */ 1582 LIST_FOREACH(afp, &classifier->acc_filters[i], 1583 f_chain) 1584 if (apply_filter4(afp->f_fbmask, 1585 &afp->f_filter, fp)) 1586 /* filter matched */ 1587 return (afp->f_class); 1588 1589 /* 1590 * check again for filters with a dst addr 1591 * wildcard. 1592 * (daddr == 0 || dmask != 0xffffffff). 1593 */ 1594 if (i != ACC_WILDCARD_INDEX) 1595 i = ACC_WILDCARD_INDEX; 1596 else 1597 break; 1598 } while (1); 1599 } 1600 } 1601#ifdef INET6 1602 else if (flow.fi_family == AF_INET6) { 1603 struct flowinfo_in6 *fp6 = (struct flowinfo_in6 *)&flow; 1604 1605 /* get the filter hash entry from its flow ID */ 1606 if (fp6->fi6_flowlabel != 0) 1607 i = ACC_GET_HASH_INDEX(fp6->fi6_flowlabel); 1608 else 1609 /* flowlable can be zero */ 1610 i = ACC_WILDCARD_INDEX; 1611 1612 /* go through this loop twice. first for flow hash, second 1613 for wildcards. */ 1614 do { 1615 LIST_FOREACH(afp, &classifier->acc_filters[i], f_chain) 1616 if (apply_filter6(afp->f_fbmask, 1617 (struct flow_filter6 *)&afp->f_filter, 1618 fp6)) 1619 /* filter matched */ 1620 return (afp->f_class); 1621 1622 /* 1623 * check again for filters with a wildcard. 1624 */ 1625 if (i != ACC_WILDCARD_INDEX) 1626 i = ACC_WILDCARD_INDEX; 1627 else 1628 break; 1629 } while (1); 1630 } 1631#endif /* INET6 */ 1632 1633 /* no filter matched */ 1634 return (NULL); 1635} 1636 1637static int 1638apply_filter4(fbmask, filt, pkt) 1639 u_int32_t fbmask; 1640 struct flow_filter *filt; 1641 struct flowinfo_in *pkt; 1642{ 1643 if (filt->ff_flow.fi_family != AF_INET) 1644 return (0); 1645 if ((fbmask & FIMB4_SPORT) && filt->ff_flow.fi_sport != pkt->fi_sport) 1646 return (0); 1647 if ((fbmask & FIMB4_DPORT) && filt->ff_flow.fi_dport != pkt->fi_dport) 1648 return (0); 1649 if ((fbmask & FIMB4_DADDR) && 1650 filt->ff_flow.fi_dst.s_addr != 1651 (pkt->fi_dst.s_addr & filt->ff_mask.mask_dst.s_addr)) 1652 return (0); 1653 if ((fbmask & FIMB4_SADDR) && 1654 filt->ff_flow.fi_src.s_addr != 1655 (pkt->fi_src.s_addr & filt->ff_mask.mask_src.s_addr)) 1656 return (0); 1657 if ((fbmask & FIMB4_PROTO) && filt->ff_flow.fi_proto != pkt->fi_proto) 1658 return (0); 1659 if ((fbmask & FIMB4_TOS) && filt->ff_flow.fi_tos != 1660 (pkt->fi_tos & filt->ff_mask.mask_tos)) 1661 return (0); 1662 if ((fbmask & FIMB4_GPI) && filt->ff_flow.fi_gpi != (pkt->fi_gpi)) 1663 return (0); 1664 /* match */ 1665 return (1); 1666} 1667 1668/* 1669 * filter matching function optimized for a common case that checks 1670 * only protocol and port numbers 1671 */ 1672static int 1673apply_ppfilter4(fbmask, filt, pkt) 1674 u_int32_t fbmask; 1675 struct flow_filter *filt; 1676 struct flowinfo_in *pkt; 1677{ 1678 if (filt->ff_flow.fi_family != AF_INET) 1679 return (0); 1680 if ((fbmask & FIMB4_SPORT) && filt->ff_flow.fi_sport != pkt->fi_sport) 1681 return (0); 1682 if ((fbmask & FIMB4_DPORT) && filt->ff_flow.fi_dport != pkt->fi_dport) 1683 return (0); 1684 if ((fbmask & FIMB4_PROTO) && filt->ff_flow.fi_proto != pkt->fi_proto) 1685 return (0); 1686 /* match */ 1687 return (1); 1688} 1689 1690/* 1691 * filter matching function only for tos field. 1692 */ 1693static int 1694apply_tosfilter4(fbmask, filt, pkt) 1695 u_int32_t fbmask; 1696 struct flow_filter *filt; 1697 struct flowinfo_in *pkt; 1698{ 1699 if (filt->ff_flow.fi_family != AF_INET) 1700 return (0); 1701 if ((fbmask & FIMB4_TOS) && filt->ff_flow.fi_tos != 1702 (pkt->fi_tos & filt->ff_mask.mask_tos)) 1703 return (0); 1704 /* match */ 1705 return (1); 1706} 1707 1708#ifdef INET6 1709static int 1710apply_filter6(fbmask, filt, pkt) 1711 u_int32_t fbmask; 1712 struct flow_filter6 *filt; 1713 struct flowinfo_in6 *pkt; 1714{ 1715 int i; 1716 1717 if (filt->ff_flow6.fi6_family != AF_INET6) 1718 return (0); 1719 if ((fbmask & FIMB6_FLABEL) && 1720 filt->ff_flow6.fi6_flowlabel != pkt->fi6_flowlabel) 1721 return (0); 1722 if ((fbmask & FIMB6_PROTO) && 1723 filt->ff_flow6.fi6_proto != pkt->fi6_proto) 1724 return (0); 1725 if ((fbmask & FIMB6_SPORT) && 1726 filt->ff_flow6.fi6_sport != pkt->fi6_sport) 1727 return (0); 1728 if ((fbmask & FIMB6_DPORT) && 1729 filt->ff_flow6.fi6_dport != pkt->fi6_dport) 1730 return (0); 1731 if (fbmask & FIMB6_SADDR) { 1732 for (i = 0; i < 4; i++) 1733 if (filt->ff_flow6.fi6_src.s6_addr32[i] != 1734 (pkt->fi6_src.s6_addr32[i] & 1735 filt->ff_mask6.mask6_src.s6_addr32[i])) 1736 return (0); 1737 } 1738 if (fbmask & FIMB6_DADDR) { 1739 for (i = 0; i < 4; i++) 1740 if (filt->ff_flow6.fi6_dst.s6_addr32[i] != 1741 (pkt->fi6_dst.s6_addr32[i] & 1742 filt->ff_mask6.mask6_dst.s6_addr32[i])) 1743 return (0); 1744 } 1745 if ((fbmask & FIMB6_TCLASS) && 1746 filt->ff_flow6.fi6_tclass != 1747 (pkt->fi6_tclass & filt->ff_mask6.mask6_tclass)) 1748 return (0); 1749 if ((fbmask & FIMB6_GPI) && 1750 filt->ff_flow6.fi6_gpi != pkt->fi6_gpi) 1751 return (0); 1752 /* match */ 1753 return (1); 1754} 1755#endif /* INET6 */ 1756 1757/* 1758 * filter handle: 1759 * bit 20-28: index to the filter hash table 1760 * bit 0-19: unique id in the hash bucket. 1761 */ 1762static u_long 1763get_filt_handle(classifier, i) 1764 struct acc_classifier *classifier; 1765 int i; 1766{ 1767 static u_long handle_number = 1; 1768 u_long handle; 1769 struct acc_filter *afp; 1770 1771 while (1) { 1772 handle = handle_number++ & 0x000fffff; 1773 1774 if (LIST_EMPTY(&classifier->acc_filters[i])) 1775 break; 1776 1777 LIST_FOREACH(afp, &classifier->acc_filters[i], f_chain) 1778 if ((afp->f_handle & 0x000fffff) == handle) 1779 break; 1780 if (afp == NULL) 1781 break; 1782 /* this handle is already used, try again */ 1783 } 1784 1785 return ((i << 20) | handle); 1786} 1787 1788/* convert filter handle to filter pointer */ 1789static struct acc_filter * 1790filth_to_filtp(classifier, handle) 1791 struct acc_classifier *classifier; 1792 u_long handle; 1793{ 1794 struct acc_filter *afp; 1795 int i; 1796 1797 i = ACC_GET_HINDEX(handle); 1798 1799 LIST_FOREACH(afp, &classifier->acc_filters[i], f_chain) 1800 if (afp->f_handle == handle) 1801 return (afp); 1802 1803 return (NULL); 1804} 1805 1806/* create flowinfo bitmask */ 1807static u_int32_t 1808filt2fibmask(filt) 1809 struct flow_filter *filt; 1810{ 1811 u_int32_t mask = 0; 1812#ifdef INET6 1813 struct flow_filter6 *filt6; 1814#endif 1815 1816 switch (filt->ff_flow.fi_family) { 1817 case AF_INET: 1818 if (filt->ff_flow.fi_proto != 0) 1819 mask |= FIMB4_PROTO; 1820 if (filt->ff_flow.fi_tos != 0) 1821 mask |= FIMB4_TOS; 1822 if (filt->ff_flow.fi_dst.s_addr != 0) 1823 mask |= FIMB4_DADDR; 1824 if (filt->ff_flow.fi_src.s_addr != 0) 1825 mask |= FIMB4_SADDR; 1826 if (filt->ff_flow.fi_sport != 0) 1827 mask |= FIMB4_SPORT; 1828 if (filt->ff_flow.fi_dport != 0) 1829 mask |= FIMB4_DPORT; 1830 if (filt->ff_flow.fi_gpi != 0) 1831 mask |= FIMB4_GPI; 1832 break; 1833#ifdef INET6 1834 case AF_INET6: 1835 filt6 = (struct flow_filter6 *)filt; 1836 1837 if (filt6->ff_flow6.fi6_proto != 0) 1838 mask |= FIMB6_PROTO; 1839 if (filt6->ff_flow6.fi6_tclass != 0) 1840 mask |= FIMB6_TCLASS; 1841 if (!IN6_IS_ADDR_UNSPECIFIED(&filt6->ff_flow6.fi6_dst)) 1842 mask |= FIMB6_DADDR; 1843 if (!IN6_IS_ADDR_UNSPECIFIED(&filt6->ff_flow6.fi6_src)) 1844 mask |= FIMB6_SADDR; 1845 if (filt6->ff_flow6.fi6_sport != 0) 1846 mask |= FIMB6_SPORT; 1847 if (filt6->ff_flow6.fi6_dport != 0) 1848 mask |= FIMB6_DPORT; 1849 if (filt6->ff_flow6.fi6_gpi != 0) 1850 mask |= FIMB6_GPI; 1851 if (filt6->ff_flow6.fi6_flowlabel != 0) 1852 mask |= FIMB6_FLABEL; 1853 break; 1854#endif /* INET6 */ 1855 } 1856 return (mask); 1857} 1858 1859 1860/* 1861 * helper functions to handle IPv4 fragments. 1862 * currently only in-sequence fragments are handled. 1863 * - fragment info is cached in a LRU list. 1864 * - when a first fragment is found, cache its flow info. 1865 * - when a non-first fragment is found, lookup the cache. 1866 */ 1867 1868struct ip4_frag { 1869 TAILQ_ENTRY(ip4_frag) ip4f_chain; 1870 char ip4f_valid; 1871 u_short ip4f_id; 1872 struct flowinfo_in ip4f_info; 1873}; 1874 1875static TAILQ_HEAD(ip4f_list, ip4_frag) ip4f_list; /* IPv4 fragment cache */ 1876 1877#define IP4F_TABSIZE 16 /* IPv4 fragment cache size */ 1878 1879 1880static void 1881ip4f_cache(ip, fin) 1882 struct ip *ip; 1883 struct flowinfo_in *fin; 1884{ 1885 struct ip4_frag *fp; 1886 1887 if (TAILQ_EMPTY(&ip4f_list)) { 1888 /* first time call, allocate fragment cache entries. */ 1889 if (ip4f_init() < 0) 1890 /* allocation failed! */ 1891 return; 1892 } 1893 1894 fp = ip4f_alloc(); 1895 fp->ip4f_id = ip->ip_id; 1896 fp->ip4f_info.fi_proto = ip->ip_p; 1897 fp->ip4f_info.fi_src.s_addr = ip->ip_src.s_addr; 1898 fp->ip4f_info.fi_dst.s_addr = ip->ip_dst.s_addr; 1899 1900 /* save port numbers */ 1901 fp->ip4f_info.fi_sport = fin->fi_sport; 1902 fp->ip4f_info.fi_dport = fin->fi_dport; 1903 fp->ip4f_info.fi_gpi = fin->fi_gpi; 1904} 1905 1906static int 1907ip4f_lookup(ip, fin) 1908 struct ip *ip; 1909 struct flowinfo_in *fin; 1910{ 1911 struct ip4_frag *fp; 1912 1913 for (fp = TAILQ_FIRST(&ip4f_list); fp != NULL && fp->ip4f_valid; 1914 fp = TAILQ_NEXT(fp, ip4f_chain)) 1915 if (ip->ip_id == fp->ip4f_id && 1916 ip->ip_src.s_addr == fp->ip4f_info.fi_src.s_addr && 1917 ip->ip_dst.s_addr == fp->ip4f_info.fi_dst.s_addr && 1918 ip->ip_p == fp->ip4f_info.fi_proto) { 1919 1920 /* found the matching entry */ 1921 fin->fi_sport = fp->ip4f_info.fi_sport; 1922 fin->fi_dport = fp->ip4f_info.fi_dport; 1923 fin->fi_gpi = fp->ip4f_info.fi_gpi; 1924 1925 if ((ntohs(ip->ip_off) & IP_MF) == 0) 1926 /* this is the last fragment, 1927 release the entry. */ 1928 ip4f_free(fp); 1929 1930 return (1); 1931 } 1932 1933 /* no matching entry found */ 1934 return (0); 1935} 1936 1937static int 1938ip4f_init(void) 1939{ 1940 struct ip4_frag *fp; 1941 int i; 1942 1943 TAILQ_INIT(&ip4f_list); 1944 for (i=0; i<IP4F_TABSIZE; i++) { 1945 fp = malloc(sizeof(struct ip4_frag), 1946 M_DEVBUF, M_NOWAIT); 1947 if (fp == NULL) { 1948 printf("ip4f_init: can't alloc %dth entry!\n", i); 1949 if (i == 0) 1950 return (-1); 1951 return (0); 1952 } 1953 fp->ip4f_valid = 0; 1954 TAILQ_INSERT_TAIL(&ip4f_list, fp, ip4f_chain); 1955 } 1956 return (0); 1957} 1958 1959static struct ip4_frag * 1960ip4f_alloc(void) 1961{ 1962 struct ip4_frag *fp; 1963 1964 /* reclaim an entry at the tail, put it at the head */ 1965 fp = TAILQ_LAST(&ip4f_list, ip4f_list); 1966 TAILQ_REMOVE(&ip4f_list, fp, ip4f_chain); 1967 fp->ip4f_valid = 1; 1968 TAILQ_INSERT_HEAD(&ip4f_list, fp, ip4f_chain); 1969 return (fp); 1970} 1971 1972static void 1973ip4f_free(fp) 1974 struct ip4_frag *fp; 1975{ 1976 TAILQ_REMOVE(&ip4f_list, fp, ip4f_chain); 1977 fp->ip4f_valid = 0; 1978 TAILQ_INSERT_TAIL(&ip4f_list, fp, ip4f_chain); 1979} 1980 1981#endif /* ALTQ3_CLFIER_COMPAT */ 1982