2 * Copyright (C) 1997-2003
3 * Sony Computer Science Laboratories Inc. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
14 * THIS SOFTWARE IS PROVIDED BY SONY CSL AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED. IN NO EVENT SHALL SONY CSL OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 * $KAME: altq_subr.c,v 1.21 2003/11/06 06:32:53 kjc Exp $
32 #include "opt_inet6.h"
34 #include <sys/param.h>
35 #include <sys/malloc.h>
37 #include <sys/systm.h>
39 #include <sys/socket.h>
40 #include <sys/socketvar.h>
41 #include <sys/kernel.h>
42 #include <sys/errno.h>
43 #include <sys/syslog.h>
44 #include <sys/sysctl.h>
45 #include <sys/queue.h>
48 #include <net/if_var.h>
49 #include <net/if_dl.h>
50 #include <net/if_types.h>
53 #include <netinet/in.h>
54 #include <netinet/in_systm.h>
55 #include <netinet/ip.h>
57 #include <netinet/ip6.h>
59 #include <netinet/tcp.h>
60 #include <netinet/udp.h>
62 #include <netpfil/pf/pf.h>
63 #include <netpfil/pf/pf_altq.h>
64 #include <net/altq/altq.h>
66 #include <net/altq/altq_conf.h>
69 /* machine dependent clock related includes */
72 #include <sys/eventhandler.h>
73 #include <machine/clock.h>
74 #if defined(__amd64__) || defined(__i386__)
75 #include <machine/cpufunc.h> /* for pentium tsc */
76 #include <machine/specialreg.h> /* for CPUID_TSC */
77 #include <machine/md_var.h> /* for cpu_feature */
78 #endif /* __amd64 || __i386__ */
81 * internal function prototypes
83 static void tbr_timeout(void *);
84 int (*altq_input)(struct mbuf *, int) = NULL;
85 static struct mbuf *tbr_dequeue(struct ifaltq *, int);
86 static int tbr_timer = 0; /* token bucket regulator timer */
87 #if !defined(__FreeBSD__) || (__FreeBSD_version < 600000)
88 static struct callout tbr_callout = CALLOUT_INITIALIZER;
90 static struct callout tbr_callout;
93 #ifdef ALTQ3_CLFIER_COMPAT
94 static int extract_ports4(struct mbuf *, struct ip *, struct flowinfo_in *);
96 static int extract_ports6(struct mbuf *, struct ip6_hdr *,
97 struct flowinfo_in6 *);
99 static int apply_filter4(u_int32_t, struct flow_filter *,
100 struct flowinfo_in *);
101 static int apply_ppfilter4(u_int32_t, struct flow_filter *,
102 struct flowinfo_in *);
104 static int apply_filter6(u_int32_t, struct flow_filter6 *,
105 struct flowinfo_in6 *);
107 static int apply_tosfilter4(u_int32_t, struct flow_filter *,
108 struct flowinfo_in *);
109 static u_long get_filt_handle(struct acc_classifier *, int);
110 static struct acc_filter *filth_to_filtp(struct acc_classifier *, u_long);
111 static u_int32_t filt2fibmask(struct flow_filter *);
113 static void ip4f_cache(struct ip *, struct flowinfo_in *);
114 static int ip4f_lookup(struct ip *, struct flowinfo_in *);
115 static int ip4f_init(void);
116 static struct ip4_frag *ip4f_alloc(void);
117 static void ip4f_free(struct ip4_frag *);
118 #endif /* ALTQ3_CLFIER_COMPAT */
121 * alternate queueing support routines
124 /* look up the queue state by the interface name and the queueing type. */
126 altq_lookup(name, type)
132 if ((ifp = ifunit(name)) != NULL) {
133 /* read if_snd unlocked */
134 if (type != ALTQT_NONE && ifp->if_snd.altq_type == type)
135 return (ifp->if_snd.altq_disc);
142 altq_attach(ifq, type, discipline, enqueue, dequeue, request, clfier, classify)
146 int (*enqueue)(struct ifaltq *, struct mbuf *, struct altq_pktattr *);
147 struct mbuf *(*dequeue)(struct ifaltq *, int);
148 int (*request)(struct ifaltq *, int, void *);
150 void *(*classify)(void *, struct mbuf *, int);
153 if (!ALTQ_IS_READY(ifq)) {
160 * pfaltq can override the existing discipline, but altq3 cannot.
161 * check these if clfier is not NULL (which implies altq3).
163 if (clfier != NULL) {
164 if (ALTQ_IS_ENABLED(ifq)) {
168 if (ALTQ_IS_ATTACHED(ifq)) {
174 ifq->altq_type = type;
175 ifq->altq_disc = discipline;
176 ifq->altq_enqueue = enqueue;
177 ifq->altq_dequeue = dequeue;
178 ifq->altq_request = request;
179 ifq->altq_clfier = clfier;
180 ifq->altq_classify = classify;
181 ifq->altq_flags &= (ALTQF_CANTCHANGE|ALTQF_ENABLED);
184 altq_module_incref(type);
197 if (!ALTQ_IS_READY(ifq)) {
201 if (ALTQ_IS_ENABLED(ifq)) {
205 if (!ALTQ_IS_ATTACHED(ifq)) {
211 altq_module_declref(ifq->altq_type);
215 ifq->altq_type = ALTQT_NONE;
216 ifq->altq_disc = NULL;
217 ifq->altq_enqueue = NULL;
218 ifq->altq_dequeue = NULL;
219 ifq->altq_request = NULL;
220 ifq->altq_clfier = NULL;
221 ifq->altq_classify = NULL;
222 ifq->altq_flags &= ALTQF_CANTCHANGE;
236 if (!ALTQ_IS_READY(ifq)) {
240 if (ALTQ_IS_ENABLED(ifq)) {
246 IFQ_PURGE_NOLOCK(ifq);
247 ASSERT(ifq->ifq_len == 0);
248 ifq->ifq_drv_maxlen = 0; /* disable bulk dequeue */
249 ifq->altq_flags |= ALTQF_ENABLED;
250 if (ifq->altq_clfier != NULL)
251 ifq->altq_flags |= ALTQF_CLASSIFY;
265 if (!ALTQ_IS_ENABLED(ifq)) {
271 IFQ_PURGE_NOLOCK(ifq);
272 ASSERT(ifq->ifq_len == 0);
273 ifq->altq_flags &= ~(ALTQF_ENABLED|ALTQF_CLASSIFY);
282 altq_assert(file, line, failedexpr)
283 const char *file, *failedexpr;
286 (void)printf("altq assertion \"%s\" failed: file \"%s\", line %d\n",
287 failedexpr, file, line);
288 panic("altq assertion");
294 * internal representation of token bucket parameters
295 * rate: byte_per_unittime << 32
296 * (((bits_per_sec) / 8) << 32) / machclk_freq
301 #define TBR_SCALE(x) ((int64_t)(x) << TBR_SHIFT)
302 #define TBR_UNSCALE(x) ((x) >> TBR_SHIFT)
309 struct tb_regulator *tbr;
314 IFQ_LOCK_ASSERT(ifq);
316 if (op == ALTDQ_REMOVE && tbr->tbr_lastop == ALTDQ_POLL) {
317 /* if this is a remove after poll, bypass tbr check */
319 /* update token only when it is negative */
320 if (tbr->tbr_token <= 0) {
321 now = read_machclk();
322 interval = now - tbr->tbr_last;
323 if (interval >= tbr->tbr_filluptime)
324 tbr->tbr_token = tbr->tbr_depth;
326 tbr->tbr_token += interval * tbr->tbr_rate;
327 if (tbr->tbr_token > tbr->tbr_depth)
328 tbr->tbr_token = tbr->tbr_depth;
332 /* if token is still negative, don't allow dequeue */
333 if (tbr->tbr_token <= 0)
337 if (ALTQ_IS_ENABLED(ifq))
338 m = (*ifq->altq_dequeue)(ifq, op);
340 if (op == ALTDQ_POLL)
346 if (m != NULL && op == ALTDQ_REMOVE)
347 tbr->tbr_token -= TBR_SCALE(m_pktlen(m));
348 tbr->tbr_lastop = op;
353 * set a token bucket regulator.
354 * if the specified rate is zero, the token bucket regulator is deleted.
357 tbr_set(ifq, profile)
359 struct tb_profile *profile;
361 struct tb_regulator *tbr, *otbr;
363 if (tbr_dequeue_ptr == NULL)
364 tbr_dequeue_ptr = tbr_dequeue;
366 if (machclk_freq == 0)
368 if (machclk_freq == 0) {
369 printf("tbr_set: no cpu clock available!\n");
374 if (profile->rate == 0) {
375 /* delete this tbr */
376 if ((tbr = ifq->altq_tbr) == NULL) {
380 ifq->altq_tbr = NULL;
386 tbr = malloc(sizeof(struct tb_regulator), M_DEVBUF, M_NOWAIT | M_ZERO);
392 tbr->tbr_rate = TBR_SCALE(profile->rate / 8) / machclk_freq;
393 tbr->tbr_depth = TBR_SCALE(profile->depth);
394 if (tbr->tbr_rate > 0)
395 tbr->tbr_filluptime = tbr->tbr_depth / tbr->tbr_rate;
397 tbr->tbr_filluptime = 0xffffffffffffffffLL;
398 tbr->tbr_token = tbr->tbr_depth;
399 tbr->tbr_last = read_machclk();
400 tbr->tbr_lastop = ALTDQ_REMOVE;
402 otbr = ifq->altq_tbr;
403 ifq->altq_tbr = tbr; /* set the new tbr */
406 free(otbr, M_DEVBUF);
408 if (tbr_timer == 0) {
409 CALLOUT_RESET(&tbr_callout, 1, tbr_timeout, (void *)0);
418 * tbr_timeout goes through the interface list, and kicks the drivers
427 VNET_ITERATOR_DECL(vnet_iter);
433 IFNET_RLOCK_NOSLEEP();
434 VNET_LIST_RLOCK_NOSLEEP();
435 VNET_FOREACH(vnet_iter) {
436 CURVNET_SET(vnet_iter);
437 for (ifp = TAILQ_FIRST(&V_ifnet); ifp;
438 ifp = TAILQ_NEXT(ifp, if_list)) {
439 /* read from if_snd unlocked */
440 if (!TBR_IS_ENABLED(&ifp->if_snd))
443 if (!IFQ_IS_EMPTY(&ifp->if_snd) &&
444 ifp->if_start != NULL)
445 (*ifp->if_start)(ifp);
449 VNET_LIST_RUNLOCK_NOSLEEP();
450 IFNET_RUNLOCK_NOSLEEP();
453 CALLOUT_RESET(&tbr_callout, 1, tbr_timeout, (void *)0);
455 tbr_timer = 0; /* don't need tbr_timer anymore */
459 * get token bucket regulator profile
462 tbr_get(ifq, profile)
464 struct tb_profile *profile;
466 struct tb_regulator *tbr;
469 if ((tbr = ifq->altq_tbr) == NULL) {
474 (u_int)TBR_UNSCALE(tbr->tbr_rate * 8 * machclk_freq);
475 profile->depth = (u_int)TBR_UNSCALE(tbr->tbr_depth);
482 * attach a discipline to the interface. if one already exists, it is
484 * Locking is done in the discipline specific attach functions. Basically
485 * they call back to altq_attach which takes care of the attach and locking.
488 altq_pfattach(struct pf_altq *a)
492 switch (a->scheduler) {
497 error = cbq_pfattach(a);
502 error = priq_pfattach(a);
507 error = hfsc_pfattach(a);
518 * detach a discipline from the interface.
519 * it is possible that the discipline was already overridden by another
523 altq_pfdetach(struct pf_altq *a)
528 if ((ifp = ifunit(a->ifname)) == NULL)
531 /* if this discipline is no longer referenced, just return */
532 /* read unlocked from if_snd */
533 if (a->altq_disc == NULL || a->altq_disc != ifp->if_snd.altq_disc)
537 /* read unlocked from if_snd, _disable and _detach take care */
538 if (ALTQ_IS_ENABLED(&ifp->if_snd))
539 error = altq_disable(&ifp->if_snd);
541 error = altq_detach(&ifp->if_snd);
548 * add a discipline or a queue
549 * Locking is done in the discipline specific functions with regards to
550 * malloc with WAITOK, also it is not yet clear which lock to use.
553 altq_add(struct pf_altq *a)
557 if (a->qname[0] != 0)
558 return (altq_add_queue(a));
560 if (machclk_freq == 0)
562 if (machclk_freq == 0)
563 panic("altq_add: no cpu clock");
565 switch (a->scheduler) {
568 error = cbq_add_altq(a);
573 error = priq_add_altq(a);
578 error = hfsc_add_altq(a);
589 * remove a discipline or a queue
590 * It is yet unclear what lock to use to protect this operation, the
591 * discipline specific functions will determine and grab it
594 altq_remove(struct pf_altq *a)
598 if (a->qname[0] != 0)
599 return (altq_remove_queue(a));
601 switch (a->scheduler) {
604 error = cbq_remove_altq(a);
609 error = priq_remove_altq(a);
614 error = hfsc_remove_altq(a);
625 * add a queue to the discipline
626 * It is yet unclear what lock to use to protect this operation, the
627 * discipline specific functions will determine and grab it
630 altq_add_queue(struct pf_altq *a)
634 switch (a->scheduler) {
637 error = cbq_add_queue(a);
642 error = priq_add_queue(a);
647 error = hfsc_add_queue(a);
658 * remove a queue from the discipline
659 * It is yet unclear what lock to use to protect this operation, the
660 * discipline specific functions will determine and grab it
663 altq_remove_queue(struct pf_altq *a)
667 switch (a->scheduler) {
670 error = cbq_remove_queue(a);
675 error = priq_remove_queue(a);
680 error = hfsc_remove_queue(a);
691 * get queue statistics
692 * Locking is done in the discipline specific functions with regards to
693 * copyout operations, also it is not yet clear which lock to use.
696 altq_getqstats(struct pf_altq *a, void *ubuf, int *nbytes)
700 switch (a->scheduler) {
703 error = cbq_getqstats(a, ubuf, nbytes);
708 error = priq_getqstats(a, ubuf, nbytes);
713 error = hfsc_getqstats(a, ubuf, nbytes);
724 * read and write diffserv field in IPv4 or IPv6 header
727 read_dsfield(m, pktattr)
729 struct altq_pktattr *pktattr;
732 u_int8_t ds_field = 0;
734 if (pktattr == NULL ||
735 (pktattr->pattr_af != AF_INET && pktattr->pattr_af != AF_INET6))
736 return ((u_int8_t)0);
738 /* verify that pattr_hdr is within the mbuf data */
739 for (m0 = m; m0 != NULL; m0 = m0->m_next)
740 if ((pktattr->pattr_hdr >= m0->m_data) &&
741 (pktattr->pattr_hdr < m0->m_data + m0->m_len))
744 /* ick, pattr_hdr is stale */
745 pktattr->pattr_af = AF_UNSPEC;
747 printf("read_dsfield: can't locate header!\n");
749 return ((u_int8_t)0);
752 if (pktattr->pattr_af == AF_INET) {
753 struct ip *ip = (struct ip *)pktattr->pattr_hdr;
756 return ((u_int8_t)0); /* version mismatch! */
757 ds_field = ip->ip_tos;
760 else if (pktattr->pattr_af == AF_INET6) {
761 struct ip6_hdr *ip6 = (struct ip6_hdr *)pktattr->pattr_hdr;
764 flowlabel = ntohl(ip6->ip6_flow);
765 if ((flowlabel >> 28) != 6)
766 return ((u_int8_t)0); /* version mismatch! */
767 ds_field = (flowlabel >> 20) & 0xff;
774 write_dsfield(struct mbuf *m, struct altq_pktattr *pktattr, u_int8_t dsfield)
778 if (pktattr == NULL ||
779 (pktattr->pattr_af != AF_INET && pktattr->pattr_af != AF_INET6))
782 /* verify that pattr_hdr is within the mbuf data */
783 for (m0 = m; m0 != NULL; m0 = m0->m_next)
784 if ((pktattr->pattr_hdr >= m0->m_data) &&
785 (pktattr->pattr_hdr < m0->m_data + m0->m_len))
788 /* ick, pattr_hdr is stale */
789 pktattr->pattr_af = AF_UNSPEC;
791 printf("write_dsfield: can't locate header!\n");
796 if (pktattr->pattr_af == AF_INET) {
797 struct ip *ip = (struct ip *)pktattr->pattr_hdr;
802 return; /* version mismatch! */
804 dsfield |= old & 3; /* leave CU bits */
807 ip->ip_tos = dsfield;
809 * update checksum (from RFC1624)
810 * HC' = ~(~HC + ~m + m')
812 sum = ~ntohs(ip->ip_sum) & 0xffff;
813 sum += 0xff00 + (~old & 0xff) + dsfield;
814 sum = (sum >> 16) + (sum & 0xffff);
815 sum += (sum >> 16); /* add carry */
817 ip->ip_sum = htons(~sum & 0xffff);
820 else if (pktattr->pattr_af == AF_INET6) {
821 struct ip6_hdr *ip6 = (struct ip6_hdr *)pktattr->pattr_hdr;
824 flowlabel = ntohl(ip6->ip6_flow);
825 if ((flowlabel >> 28) != 6)
826 return; /* version mismatch! */
827 flowlabel = (flowlabel & 0xf03fffff) | (dsfield << 20);
828 ip6->ip6_flow = htonl(flowlabel);
836 * high resolution clock support taking advantage of a machine dependent
837 * high resolution time counter (e.g., timestamp counter of intel pentium).
839 * - 64-bit-long monotonically-increasing counter
840 * - frequency range is 100M-4GHz (CPU speed)
842 /* if pcc is not available or disabled, emulate 256MHz using microtime() */
843 #define MACHCLK_SHIFT 8
846 u_int32_t machclk_freq;
847 u_int32_t machclk_per_tick;
849 #if defined(__i386__) && defined(__NetBSD__)
850 extern u_int64_t cpu_tsc_freq;
853 #if (__FreeBSD_version >= 700035)
854 /* Update TSC freq with the value indicated by the caller. */
856 tsc_freq_changed(void *arg, const struct cf_level *level, int status)
858 /* If there was an error during the transition, don't do anything. */
862 #if (__FreeBSD_version >= 701102) && (defined(__amd64__) || defined(__i386__))
863 /* If TSC is P-state invariant, don't do anything. */
864 if (tsc_is_invariant)
868 /* Total setting for this level gives the new frequency in MHz. */
871 EVENTHANDLER_DEFINE(cpufreq_post_change, tsc_freq_changed, NULL,
872 EVENTHANDLER_PRI_LAST);
873 #endif /* __FreeBSD_version >= 700035 */
876 init_machclk_setup(void)
878 #if (__FreeBSD_version >= 600000)
879 callout_init(&tbr_callout, 0);
884 #if (!defined(__amd64__) && !defined(__i386__)) || defined(ALTQ_NOPCC)
887 #if defined(__FreeBSD__) && defined(SMP)
890 #if defined(__NetBSD__) && defined(MULTIPROCESSOR)
893 #if defined(__amd64__) || defined(__i386__)
894 /* check if TSC is available */
895 if ((cpu_feature & CPUID_TSC) == 0 ||
896 atomic_load_acq_64(&tsc_freq) == 0)
906 /* Call one-time initialization function. */
908 init_machclk_setup();
912 if (machclk_usepcc == 0) {
913 /* emulate 256MHz using microtime() */
914 machclk_freq = 1000000 << MACHCLK_SHIFT;
915 machclk_per_tick = machclk_freq / hz;
917 printf("altq: emulate %uHz cpu clock\n", machclk_freq);
923 * if the clock frequency (of Pentium TSC or Alpha PCC) is
924 * accessible, just use it.
926 #if defined(__amd64__) || defined(__i386__)
927 machclk_freq = atomic_load_acq_64(&tsc_freq);
931 * if we don't know the clock frequency, measure it.
933 if (machclk_freq == 0) {
935 struct timeval tv_start, tv_end;
936 u_int64_t start, end, diff;
939 microtime(&tv_start);
940 start = read_machclk();
941 timo = hz; /* 1 sec */
942 (void)tsleep(&wait, PWAIT | PCATCH, "init_machclk", timo);
944 end = read_machclk();
945 diff = (u_int64_t)(tv_end.tv_sec - tv_start.tv_sec) * 1000000
946 + tv_end.tv_usec - tv_start.tv_usec;
948 machclk_freq = (u_int)((end - start) * 1000000 / diff);
951 machclk_per_tick = machclk_freq / hz;
954 printf("altq: CPU clock: %uHz\n", machclk_freq);
958 #if defined(__OpenBSD__) && defined(__i386__)
959 static __inline u_int64_t
963 __asm __volatile(".byte 0x0f, 0x31" : "=A" (rv));
966 #endif /* __OpenBSD__ && __i386__ */
973 if (machclk_usepcc) {
974 #if defined(__amd64__) || defined(__i386__)
977 panic("read_machclk");
983 val = (((u_int64_t)(tv.tv_sec - boottime.tv_sec) * 1000000
984 + tv.tv_usec) << MACHCLK_SHIFT);
989 #ifdef ALTQ3_CLFIER_COMPAT
992 #define IPPROTO_ESP 50 /* encapsulating security payload */
995 #define IPPROTO_AH 51 /* authentication header */
999 * extract flow information from a given packet.
1000 * filt_mask shows flowinfo fields required.
1001 * we assume the ip header is in one mbuf, and addresses and ports are
1002 * in network byte order.
1005 altq_extractflow(m, af, flow, filt_bmask)
1008 struct flowinfo *flow;
1009 u_int32_t filt_bmask;
1014 struct flowinfo_in *fin;
1017 ip = mtod(m, struct ip *);
1022 fin = (struct flowinfo_in *)flow;
1023 fin->fi_len = sizeof(struct flowinfo_in);
1024 fin->fi_family = AF_INET;
1026 fin->fi_proto = ip->ip_p;
1027 fin->fi_tos = ip->ip_tos;
1029 fin->fi_src.s_addr = ip->ip_src.s_addr;
1030 fin->fi_dst.s_addr = ip->ip_dst.s_addr;
1032 if (filt_bmask & FIMB4_PORTS)
1033 /* if port info is required, extract port numbers */
1034 extract_ports4(m, ip, fin);
1045 struct flowinfo_in6 *fin6;
1046 struct ip6_hdr *ip6;
1048 ip6 = mtod(m, struct ip6_hdr *);
1049 /* should we check the ip version? */
1051 fin6 = (struct flowinfo_in6 *)flow;
1052 fin6->fi6_len = sizeof(struct flowinfo_in6);
1053 fin6->fi6_family = AF_INET6;
1055 fin6->fi6_proto = ip6->ip6_nxt;
1056 fin6->fi6_tclass = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
1058 fin6->fi6_flowlabel = ip6->ip6_flow & htonl(0x000fffff);
1059 fin6->fi6_src = ip6->ip6_src;
1060 fin6->fi6_dst = ip6->ip6_dst;
1062 if ((filt_bmask & FIMB6_PORTS) ||
1063 ((filt_bmask & FIMB6_PROTO)
1064 && ip6->ip6_nxt > IPPROTO_IPV6))
1066 * if port info is required, or proto is required
1067 * but there are option headers, extract port
1068 * and protocol numbers.
1070 extract_ports6(m, ip6, fin6);
1072 fin6->fi6_sport = 0;
1073 fin6->fi6_dport = 0;
1085 flow->fi_len = sizeof(struct flowinfo);
1086 flow->fi_family = AF_UNSPEC;
1091 * helper routine to extract port numbers
1093 /* structure for ipsec and ipv6 option header template */
1095 u_int8_t opt6_nxt; /* next header */
1096 u_int8_t opt6_hlen; /* header extension length */
1098 u_int32_t ah_spi; /* security parameter index
1099 for authentication header */
1103 * extract port numbers from a ipv4 packet.
1106 extract_ports4(m, ip, fin)
1109 struct flowinfo_in *fin;
1120 ip_off = ntohs(ip->ip_off);
1121 /* if it is a fragment, try cached fragment info */
1122 if (ip_off & IP_OFFMASK) {
1123 ip4f_lookup(ip, fin);
1127 /* locate the mbuf containing the protocol header */
1128 for (m0 = m; m0 != NULL; m0 = m0->m_next)
1129 if (((caddr_t)ip >= m0->m_data) &&
1130 ((caddr_t)ip < m0->m_data + m0->m_len))
1134 printf("extract_ports4: can't locate header! ip=%p\n", ip);
1138 off = ((caddr_t)ip - m0->m_data) + (ip->ip_hl << 2);
1144 while (off >= m0->m_len) {
1148 return (0); /* bogus ip_hl! */
1150 if (m0->m_len < off + 4)
1158 udp = (struct udphdr *)(mtod(m0, caddr_t) + off);
1159 fin->fi_sport = udp->uh_sport;
1160 fin->fi_dport = udp->uh_dport;
1161 fin->fi_proto = proto;
1167 if (fin->fi_gpi == 0){
1170 gpi = (u_int32_t *)(mtod(m0, caddr_t) + off);
1173 fin->fi_proto = proto;
1177 /* get next header and header length */
1180 opt6 = (struct _opt6 *)(mtod(m0, caddr_t) + off);
1181 proto = opt6->opt6_nxt;
1182 off += 8 + (opt6->opt6_hlen * 4);
1183 if (fin->fi_gpi == 0 && m0->m_len >= off + 8)
1184 fin->fi_gpi = opt6->ah_spi;
1186 /* goto the next header */
1188 #endif /* ALTQ_IPSEC */
1191 fin->fi_proto = proto;
1195 /* if this is a first fragment, cache it. */
1197 ip4f_cache(ip, fin);
1204 extract_ports6(m, ip6, fin6)
1206 struct ip6_hdr *ip6;
1207 struct flowinfo_in6 *fin6;
1214 fin6->fi6_sport = 0;
1215 fin6->fi6_dport = 0;
1217 /* locate the mbuf containing the protocol header */
1218 for (m0 = m; m0 != NULL; m0 = m0->m_next)
1219 if (((caddr_t)ip6 >= m0->m_data) &&
1220 ((caddr_t)ip6 < m0->m_data + m0->m_len))
1224 printf("extract_ports6: can't locate header! ip6=%p\n", ip6);
1228 off = ((caddr_t)ip6 - m0->m_data) + sizeof(struct ip6_hdr);
1230 proto = ip6->ip6_nxt;
1232 while (off >= m0->m_len) {
1238 if (m0->m_len < off + 4)
1246 udp = (struct udphdr *)(mtod(m0, caddr_t) + off);
1247 fin6->fi6_sport = udp->uh_sport;
1248 fin6->fi6_dport = udp->uh_dport;
1249 fin6->fi6_proto = proto;
1254 if (fin6->fi6_gpi == 0) {
1257 gpi = (u_int32_t *)(mtod(m0, caddr_t) + off);
1258 fin6->fi6_gpi = *gpi;
1260 fin6->fi6_proto = proto;
1264 /* get next header and header length */
1267 opt6 = (struct _opt6 *)(mtod(m0, caddr_t) + off);
1268 if (fin6->fi6_gpi == 0 && m0->m_len >= off + 8)
1269 fin6->fi6_gpi = opt6->ah_spi;
1270 proto = opt6->opt6_nxt;
1271 off += 8 + (opt6->opt6_hlen * 4);
1272 /* goto the next header */
1276 case IPPROTO_HOPOPTS:
1277 case IPPROTO_ROUTING:
1278 case IPPROTO_DSTOPTS: {
1279 /* get next header and header length */
1282 opt6 = (struct _opt6 *)(mtod(m0, caddr_t) + off);
1283 proto = opt6->opt6_nxt;
1284 off += (opt6->opt6_hlen + 1) * 8;
1285 /* goto the next header */
1289 case IPPROTO_FRAGMENT:
1290 /* ipv6 fragmentations are not supported yet */
1292 fin6->fi6_proto = proto;
1301 * altq common classifier
1304 acc_add_filter(classifier, filter, class, phandle)
1305 struct acc_classifier *classifier;
1306 struct flow_filter *filter;
1310 struct acc_filter *afp, *prev, *tmp;
1314 if (filter->ff_flow.fi_family != AF_INET &&
1315 filter->ff_flow.fi_family != AF_INET6)
1318 if (filter->ff_flow.fi_family != AF_INET)
1322 afp = malloc(sizeof(struct acc_filter),
1323 M_DEVBUF, M_WAITOK);
1326 bzero(afp, sizeof(struct acc_filter));
1328 afp->f_filter = *filter;
1329 afp->f_class = class;
1331 i = ACC_WILDCARD_INDEX;
1332 if (filter->ff_flow.fi_family == AF_INET) {
1333 struct flow_filter *filter4 = &afp->f_filter;
1336 * if address is 0, it's a wildcard. if address mask
1337 * isn't set, use full mask.
1339 if (filter4->ff_flow.fi_dst.s_addr == 0)
1340 filter4->ff_mask.mask_dst.s_addr = 0;
1341 else if (filter4->ff_mask.mask_dst.s_addr == 0)
1342 filter4->ff_mask.mask_dst.s_addr = 0xffffffff;
1343 if (filter4->ff_flow.fi_src.s_addr == 0)
1344 filter4->ff_mask.mask_src.s_addr = 0;
1345 else if (filter4->ff_mask.mask_src.s_addr == 0)
1346 filter4->ff_mask.mask_src.s_addr = 0xffffffff;
1348 /* clear extra bits in addresses */
1349 filter4->ff_flow.fi_dst.s_addr &=
1350 filter4->ff_mask.mask_dst.s_addr;
1351 filter4->ff_flow.fi_src.s_addr &=
1352 filter4->ff_mask.mask_src.s_addr;
1355 * if dst address is a wildcard, use hash-entry
1356 * ACC_WILDCARD_INDEX.
1358 if (filter4->ff_mask.mask_dst.s_addr != 0xffffffff)
1359 i = ACC_WILDCARD_INDEX;
1361 i = ACC_GET_HASH_INDEX(filter4->ff_flow.fi_dst.s_addr);
1364 else if (filter->ff_flow.fi_family == AF_INET6) {
1365 struct flow_filter6 *filter6 =
1366 (struct flow_filter6 *)&afp->f_filter;
1367 #ifndef IN6MASK0 /* taken from kame ipv6 */
1368 #define IN6MASK0 {{{ 0, 0, 0, 0 }}}
1369 #define IN6MASK128 {{{ 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }}}
1370 const struct in6_addr in6mask0 = IN6MASK0;
1371 const struct in6_addr in6mask128 = IN6MASK128;
1374 if (IN6_IS_ADDR_UNSPECIFIED(&filter6->ff_flow6.fi6_dst))
1375 filter6->ff_mask6.mask6_dst = in6mask0;
1376 else if (IN6_IS_ADDR_UNSPECIFIED(&filter6->ff_mask6.mask6_dst))
1377 filter6->ff_mask6.mask6_dst = in6mask128;
1378 if (IN6_IS_ADDR_UNSPECIFIED(&filter6->ff_flow6.fi6_src))
1379 filter6->ff_mask6.mask6_src = in6mask0;
1380 else if (IN6_IS_ADDR_UNSPECIFIED(&filter6->ff_mask6.mask6_src))
1381 filter6->ff_mask6.mask6_src = in6mask128;
1383 /* clear extra bits in addresses */
1384 for (i = 0; i < 16; i++)
1385 filter6->ff_flow6.fi6_dst.s6_addr[i] &=
1386 filter6->ff_mask6.mask6_dst.s6_addr[i];
1387 for (i = 0; i < 16; i++)
1388 filter6->ff_flow6.fi6_src.s6_addr[i] &=
1389 filter6->ff_mask6.mask6_src.s6_addr[i];
1391 if (filter6->ff_flow6.fi6_flowlabel == 0)
1392 i = ACC_WILDCARD_INDEX;
1394 i = ACC_GET_HASH_INDEX(filter6->ff_flow6.fi6_flowlabel);
1398 afp->f_handle = get_filt_handle(classifier, i);
1400 /* update filter bitmask */
1401 afp->f_fbmask = filt2fibmask(filter);
1402 classifier->acc_fbmask |= afp->f_fbmask;
1405 * add this filter to the filter list.
1406 * filters are ordered from the highest rule number.
1410 LIST_FOREACH(tmp, &classifier->acc_filters[i], f_chain) {
1411 if (tmp->f_filter.ff_ruleno > afp->f_filter.ff_ruleno)
1417 LIST_INSERT_HEAD(&classifier->acc_filters[i], afp, f_chain);
1419 LIST_INSERT_AFTER(prev, afp, f_chain);
1422 *phandle = afp->f_handle;
1427 acc_delete_filter(classifier, handle)
1428 struct acc_classifier *classifier;
1431 struct acc_filter *afp;
1434 if ((afp = filth_to_filtp(classifier, handle)) == NULL)
1438 LIST_REMOVE(afp, f_chain);
1441 free(afp, M_DEVBUF);
1443 /* todo: update filt_bmask */
1449 * delete filters referencing to the specified class.
1450 * if the all flag is not 0, delete all the filters.
1453 acc_discard_filters(classifier, class, all)
1454 struct acc_classifier *classifier;
1458 struct acc_filter *afp;
1462 for (i = 0; i < ACC_FILTER_TABLESIZE; i++) {
1464 LIST_FOREACH(afp, &classifier->acc_filters[i], f_chain)
1465 if (all || afp->f_class == class) {
1466 LIST_REMOVE(afp, f_chain);
1467 free(afp, M_DEVBUF);
1468 /* start again from the head */
1471 } while (afp != NULL);
1476 classifier->acc_fbmask = 0;
1482 acc_classify(clfier, m, af)
1487 struct acc_classifier *classifier;
1488 struct flowinfo flow;
1489 struct acc_filter *afp;
1492 classifier = (struct acc_classifier *)clfier;
1493 altq_extractflow(m, af, &flow, classifier->acc_fbmask);
1495 if (flow.fi_family == AF_INET) {
1496 struct flowinfo_in *fp = (struct flowinfo_in *)&flow;
1498 if ((classifier->acc_fbmask & FIMB4_ALL) == FIMB4_TOS) {
1499 /* only tos is used */
1501 &classifier->acc_filters[ACC_WILDCARD_INDEX],
1503 if (apply_tosfilter4(afp->f_fbmask,
1504 &afp->f_filter, fp))
1505 /* filter matched */
1506 return (afp->f_class);
1507 } else if ((classifier->acc_fbmask &
1508 (~(FIMB4_PROTO|FIMB4_SPORT|FIMB4_DPORT) & FIMB4_ALL))
1510 /* only proto and ports are used */
1512 &classifier->acc_filters[ACC_WILDCARD_INDEX],
1514 if (apply_ppfilter4(afp->f_fbmask,
1515 &afp->f_filter, fp))
1516 /* filter matched */
1517 return (afp->f_class);
1519 /* get the filter hash entry from its dest address */
1520 i = ACC_GET_HASH_INDEX(fp->fi_dst.s_addr);
1523 * go through this loop twice. first for dst
1524 * hash, second for wildcards.
1526 LIST_FOREACH(afp, &classifier->acc_filters[i],
1528 if (apply_filter4(afp->f_fbmask,
1529 &afp->f_filter, fp))
1530 /* filter matched */
1531 return (afp->f_class);
1534 * check again for filters with a dst addr
1536 * (daddr == 0 || dmask != 0xffffffff).
1538 if (i != ACC_WILDCARD_INDEX)
1539 i = ACC_WILDCARD_INDEX;
1546 else if (flow.fi_family == AF_INET6) {
1547 struct flowinfo_in6 *fp6 = (struct flowinfo_in6 *)&flow;
1549 /* get the filter hash entry from its flow ID */
1550 if (fp6->fi6_flowlabel != 0)
1551 i = ACC_GET_HASH_INDEX(fp6->fi6_flowlabel);
1553 /* flowlable can be zero */
1554 i = ACC_WILDCARD_INDEX;
1556 /* go through this loop twice. first for flow hash, second
1559 LIST_FOREACH(afp, &classifier->acc_filters[i], f_chain)
1560 if (apply_filter6(afp->f_fbmask,
1561 (struct flow_filter6 *)&afp->f_filter,
1563 /* filter matched */
1564 return (afp->f_class);
1567 * check again for filters with a wildcard.
1569 if (i != ACC_WILDCARD_INDEX)
1570 i = ACC_WILDCARD_INDEX;
1577 /* no filter matched */
1582 apply_filter4(fbmask, filt, pkt)
1584 struct flow_filter *filt;
1585 struct flowinfo_in *pkt;
1587 if (filt->ff_flow.fi_family != AF_INET)
1589 if ((fbmask & FIMB4_SPORT) && filt->ff_flow.fi_sport != pkt->fi_sport)
1591 if ((fbmask & FIMB4_DPORT) && filt->ff_flow.fi_dport != pkt->fi_dport)
1593 if ((fbmask & FIMB4_DADDR) &&
1594 filt->ff_flow.fi_dst.s_addr !=
1595 (pkt->fi_dst.s_addr & filt->ff_mask.mask_dst.s_addr))
1597 if ((fbmask & FIMB4_SADDR) &&
1598 filt->ff_flow.fi_src.s_addr !=
1599 (pkt->fi_src.s_addr & filt->ff_mask.mask_src.s_addr))
1601 if ((fbmask & FIMB4_PROTO) && filt->ff_flow.fi_proto != pkt->fi_proto)
1603 if ((fbmask & FIMB4_TOS) && filt->ff_flow.fi_tos !=
1604 (pkt->fi_tos & filt->ff_mask.mask_tos))
1606 if ((fbmask & FIMB4_GPI) && filt->ff_flow.fi_gpi != (pkt->fi_gpi))
1613 * filter matching function optimized for a common case that checks
1614 * only protocol and port numbers
1617 apply_ppfilter4(fbmask, filt, pkt)
1619 struct flow_filter *filt;
1620 struct flowinfo_in *pkt;
1622 if (filt->ff_flow.fi_family != AF_INET)
1624 if ((fbmask & FIMB4_SPORT) && filt->ff_flow.fi_sport != pkt->fi_sport)
1626 if ((fbmask & FIMB4_DPORT) && filt->ff_flow.fi_dport != pkt->fi_dport)
1628 if ((fbmask & FIMB4_PROTO) && filt->ff_flow.fi_proto != pkt->fi_proto)
1635 * filter matching function only for tos field.
1638 apply_tosfilter4(fbmask, filt, pkt)
1640 struct flow_filter *filt;
1641 struct flowinfo_in *pkt;
1643 if (filt->ff_flow.fi_family != AF_INET)
1645 if ((fbmask & FIMB4_TOS) && filt->ff_flow.fi_tos !=
1646 (pkt->fi_tos & filt->ff_mask.mask_tos))
1654 apply_filter6(fbmask, filt, pkt)
1656 struct flow_filter6 *filt;
1657 struct flowinfo_in6 *pkt;
1661 if (filt->ff_flow6.fi6_family != AF_INET6)
1663 if ((fbmask & FIMB6_FLABEL) &&
1664 filt->ff_flow6.fi6_flowlabel != pkt->fi6_flowlabel)
1666 if ((fbmask & FIMB6_PROTO) &&
1667 filt->ff_flow6.fi6_proto != pkt->fi6_proto)
1669 if ((fbmask & FIMB6_SPORT) &&
1670 filt->ff_flow6.fi6_sport != pkt->fi6_sport)
1672 if ((fbmask & FIMB6_DPORT) &&
1673 filt->ff_flow6.fi6_dport != pkt->fi6_dport)
1675 if (fbmask & FIMB6_SADDR) {
1676 for (i = 0; i < 4; i++)
1677 if (filt->ff_flow6.fi6_src.s6_addr32[i] !=
1678 (pkt->fi6_src.s6_addr32[i] &
1679 filt->ff_mask6.mask6_src.s6_addr32[i]))
1682 if (fbmask & FIMB6_DADDR) {
1683 for (i = 0; i < 4; i++)
1684 if (filt->ff_flow6.fi6_dst.s6_addr32[i] !=
1685 (pkt->fi6_dst.s6_addr32[i] &
1686 filt->ff_mask6.mask6_dst.s6_addr32[i]))
1689 if ((fbmask & FIMB6_TCLASS) &&
1690 filt->ff_flow6.fi6_tclass !=
1691 (pkt->fi6_tclass & filt->ff_mask6.mask6_tclass))
1693 if ((fbmask & FIMB6_GPI) &&
1694 filt->ff_flow6.fi6_gpi != pkt->fi6_gpi)
1703 * bit 20-28: index to the filter hash table
1704 * bit 0-19: unique id in the hash bucket.
1707 get_filt_handle(classifier, i)
1708 struct acc_classifier *classifier;
1711 static u_long handle_number = 1;
1713 struct acc_filter *afp;
1716 handle = handle_number++ & 0x000fffff;
1718 if (LIST_EMPTY(&classifier->acc_filters[i]))
1721 LIST_FOREACH(afp, &classifier->acc_filters[i], f_chain)
1722 if ((afp->f_handle & 0x000fffff) == handle)
1726 /* this handle is already used, try again */
1729 return ((i << 20) | handle);
1732 /* convert filter handle to filter pointer */
1733 static struct acc_filter *
1734 filth_to_filtp(classifier, handle)
1735 struct acc_classifier *classifier;
1738 struct acc_filter *afp;
1741 i = ACC_GET_HINDEX(handle);
1743 LIST_FOREACH(afp, &classifier->acc_filters[i], f_chain)
1744 if (afp->f_handle == handle)
1750 /* create flowinfo bitmask */
1753 struct flow_filter *filt;
1757 struct flow_filter6 *filt6;
1760 switch (filt->ff_flow.fi_family) {
1762 if (filt->ff_flow.fi_proto != 0)
1763 mask |= FIMB4_PROTO;
1764 if (filt->ff_flow.fi_tos != 0)
1766 if (filt->ff_flow.fi_dst.s_addr != 0)
1767 mask |= FIMB4_DADDR;
1768 if (filt->ff_flow.fi_src.s_addr != 0)
1769 mask |= FIMB4_SADDR;
1770 if (filt->ff_flow.fi_sport != 0)
1771 mask |= FIMB4_SPORT;
1772 if (filt->ff_flow.fi_dport != 0)
1773 mask |= FIMB4_DPORT;
1774 if (filt->ff_flow.fi_gpi != 0)
1779 filt6 = (struct flow_filter6 *)filt;
1781 if (filt6->ff_flow6.fi6_proto != 0)
1782 mask |= FIMB6_PROTO;
1783 if (filt6->ff_flow6.fi6_tclass != 0)
1784 mask |= FIMB6_TCLASS;
1785 if (!IN6_IS_ADDR_UNSPECIFIED(&filt6->ff_flow6.fi6_dst))
1786 mask |= FIMB6_DADDR;
1787 if (!IN6_IS_ADDR_UNSPECIFIED(&filt6->ff_flow6.fi6_src))
1788 mask |= FIMB6_SADDR;
1789 if (filt6->ff_flow6.fi6_sport != 0)
1790 mask |= FIMB6_SPORT;
1791 if (filt6->ff_flow6.fi6_dport != 0)
1792 mask |= FIMB6_DPORT;
1793 if (filt6->ff_flow6.fi6_gpi != 0)
1795 if (filt6->ff_flow6.fi6_flowlabel != 0)
1796 mask |= FIMB6_FLABEL;
1805 * helper functions to handle IPv4 fragments.
1806 * currently only in-sequence fragments are handled.
1807 * - fragment info is cached in a LRU list.
1808 * - when a first fragment is found, cache its flow info.
1809 * - when a non-first fragment is found, lookup the cache.
1813 TAILQ_ENTRY(ip4_frag) ip4f_chain;
1816 struct flowinfo_in ip4f_info;
1819 static TAILQ_HEAD(ip4f_list, ip4_frag) ip4f_list; /* IPv4 fragment cache */
1821 #define IP4F_TABSIZE 16 /* IPv4 fragment cache size */
1827 struct flowinfo_in *fin;
1829 struct ip4_frag *fp;
1831 if (TAILQ_EMPTY(&ip4f_list)) {
1832 /* first time call, allocate fragment cache entries. */
1833 if (ip4f_init() < 0)
1834 /* allocation failed! */
1839 fp->ip4f_id = ip->ip_id;
1840 fp->ip4f_info.fi_proto = ip->ip_p;
1841 fp->ip4f_info.fi_src.s_addr = ip->ip_src.s_addr;
1842 fp->ip4f_info.fi_dst.s_addr = ip->ip_dst.s_addr;
1844 /* save port numbers */
1845 fp->ip4f_info.fi_sport = fin->fi_sport;
1846 fp->ip4f_info.fi_dport = fin->fi_dport;
1847 fp->ip4f_info.fi_gpi = fin->fi_gpi;
1851 ip4f_lookup(ip, fin)
1853 struct flowinfo_in *fin;
1855 struct ip4_frag *fp;
1857 for (fp = TAILQ_FIRST(&ip4f_list); fp != NULL && fp->ip4f_valid;
1858 fp = TAILQ_NEXT(fp, ip4f_chain))
1859 if (ip->ip_id == fp->ip4f_id &&
1860 ip->ip_src.s_addr == fp->ip4f_info.fi_src.s_addr &&
1861 ip->ip_dst.s_addr == fp->ip4f_info.fi_dst.s_addr &&
1862 ip->ip_p == fp->ip4f_info.fi_proto) {
1864 /* found the matching entry */
1865 fin->fi_sport = fp->ip4f_info.fi_sport;
1866 fin->fi_dport = fp->ip4f_info.fi_dport;
1867 fin->fi_gpi = fp->ip4f_info.fi_gpi;
1869 if ((ntohs(ip->ip_off) & IP_MF) == 0)
1870 /* this is the last fragment,
1871 release the entry. */
1877 /* no matching entry found */
1884 struct ip4_frag *fp;
1887 TAILQ_INIT(&ip4f_list);
1888 for (i=0; i<IP4F_TABSIZE; i++) {
1889 fp = malloc(sizeof(struct ip4_frag),
1890 M_DEVBUF, M_NOWAIT);
1892 printf("ip4f_init: can't alloc %dth entry!\n", i);
1898 TAILQ_INSERT_TAIL(&ip4f_list, fp, ip4f_chain);
1903 static struct ip4_frag *
1906 struct ip4_frag *fp;
1908 /* reclaim an entry at the tail, put it at the head */
1909 fp = TAILQ_LAST(&ip4f_list, ip4f_list);
1910 TAILQ_REMOVE(&ip4f_list, fp, ip4f_chain);
1912 TAILQ_INSERT_HEAD(&ip4f_list, fp, ip4f_chain);
1918 struct ip4_frag *fp;
1920 TAILQ_REMOVE(&ip4f_list, fp, ip4f_chain);
1922 TAILQ_INSERT_TAIL(&ip4f_list, fp, ip4f_chain);
1925 #endif /* ALTQ3_CLFIER_COMPAT */