/* $OpenBSD: bridgectl.c,v 1.20 2019/07/09 15:15:49 mpi Exp $ */ /* * Copyright (c) 1999, 2000 Jason L. Wright (jason@thought.net) * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE * DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE * POSSIBILITY OF SUCH DAMAGE. * * Effort sponsored in part by the Defense Advanced Research Projects * Agency (DARPA) and Air Force Research Laboratory, Air Force * Materiel Command, USAF, under agreement number F30602-01-2-0537. * */ #include "pf.h" #include #include #include #include #include #include #include #include #include #include #include #include int bridge_rtfind(struct bridge_softc *, struct ifbaconf *); int bridge_rtdaddr(struct bridge_softc *, struct ether_addr *); u_int32_t bridge_hash(struct bridge_softc *, struct ether_addr *); int bridge_brlconf(struct bridge_iflist *, struct ifbrlconf *); int bridge_addrule(struct bridge_iflist *, struct ifbrlreq *, int out); int bridgectl_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data) { struct bridge_softc *sc = (struct bridge_softc *)ifp->if_softc; struct ifbreq *req = (struct ifbreq *)data; struct ifbrlreq *brlreq = (struct ifbrlreq *)data; struct ifbrlconf *bc = (struct ifbrlconf *)data; struct ifbareq *bareq = (struct ifbareq *)data; struct ifbrparam *bparam = (struct ifbrparam *)data; struct bridge_iflist *bif; struct ifnet *ifs; int error = 0; switch (cmd) { case SIOCBRDGRTS: error = bridge_rtfind(sc, (struct ifbaconf *)data); break; case SIOCBRDGFLUSH: bridge_rtflush(sc, req->ifbr_ifsflags); break; case SIOCBRDGSADDR: ifs = ifunit(bareq->ifba_ifsname); if (ifs == NULL) { /* no such interface */ error = ENOENT; break; } if (ifs->if_bridgeidx != ifp->if_index) { error = ESRCH; break; } if (bridge_rtupdate(sc, &bareq->ifba_dst, ifs, 1, bareq->ifba_flags, NULL)) error = ENOMEM; break; case SIOCBRDGDADDR: error = bridge_rtdaddr(sc, &bareq->ifba_dst); break; case SIOCBRDGGCACHE: bparam->ifbrp_csize = sc->sc_brtmax; break; case SIOCBRDGSCACHE: mtx_enter(&sc->sc_mtx); sc->sc_brtmax = bparam->ifbrp_csize; mtx_leave(&sc->sc_mtx); break; case SIOCBRDGSTO: if (bparam->ifbrp_ctime < 0 || bparam->ifbrp_ctime > INT_MAX / hz) { error = EINVAL; break; } sc->sc_brttimeout = bparam->ifbrp_ctime; if (bparam->ifbrp_ctime != 0) timeout_add_sec(&sc->sc_brtimeout, sc->sc_brttimeout); else timeout_del(&sc->sc_brtimeout); break; case SIOCBRDGGTO: bparam->ifbrp_ctime = sc->sc_brttimeout; break; case SIOCBRDGARL: ifs = ifunit(brlreq->ifbr_ifsname); if (ifs == NULL) { error = ENOENT; break; } if (ifs->if_bridgeidx != ifp->if_index) { error = ESRCH; break; } if ((brlreq->ifbr_action != BRL_ACTION_BLOCK && brlreq->ifbr_action != BRL_ACTION_PASS) || (brlreq->ifbr_flags & (BRL_FLAG_IN|BRL_FLAG_OUT)) == 0) { error = EINVAL; break; } bif = bridge_getbif(ifs); if (brlreq->ifbr_flags & BRL_FLAG_IN) { error = bridge_addrule(bif, brlreq, 0); if (error) break; } if (brlreq->ifbr_flags & BRL_FLAG_OUT) { error = bridge_addrule(bif, brlreq, 1); if (error) break; } break; case SIOCBRDGFRL: ifs = ifunit(brlreq->ifbr_ifsname); if (ifs == NULL) { error = ENOENT; break; } if (ifs->if_bridgeidx != ifp->if_index) { error = ESRCH; break; } bif = bridge_getbif(ifs); bridge_flushrule(bif); break; case SIOCBRDGGRL: ifs = ifunit(bc->ifbrl_ifsname); if (ifs == NULL) { error = ENOENT; break; } if (ifs->if_bridgeidx != ifp->if_index) { error = ESRCH; break; } bif = bridge_getbif(ifs); error = bridge_brlconf(bif, bc); break; default: break; } return (error); } int bridge_rtupdate(struct bridge_softc *sc, struct ether_addr *ea, struct ifnet *ifp, int setflags, u_int8_t flags, struct mbuf *m) { struct bridge_rtnode *p, *q; struct bridge_tunneltag *brtag = NULL; u_int32_t h; int dir, error = 0; if (m != NULL) { /* Check if the mbuf was tagged with a tunnel endpoint addr */ brtag = bridge_tunnel(m); } h = bridge_hash(sc, ea); mtx_enter(&sc->sc_mtx); p = LIST_FIRST(&sc->sc_rts[h]); if (p == NULL) { if (sc->sc_brtcnt >= sc->sc_brtmax) goto done; p = malloc(sizeof(*p), M_DEVBUF, M_NOWAIT); if (p == NULL) goto done; bcopy(ea, &p->brt_addr, sizeof(p->brt_addr)); p->brt_ifidx = ifp->if_index; p->brt_age = 1; bridge_copytag(brtag, &p->brt_tunnel); if (setflags) p->brt_flags = flags; else p->brt_flags = IFBAF_DYNAMIC; LIST_INSERT_HEAD(&sc->sc_rts[h], p, brt_next); sc->sc_brtcnt++; goto want; } do { q = p; p = LIST_NEXT(p, brt_next); dir = memcmp(ea, &q->brt_addr, sizeof(q->brt_addr)); if (dir == 0) { if (setflags) { q->brt_ifidx = ifp->if_index; q->brt_flags = flags; } else if (!(q->brt_flags & IFBAF_STATIC)) q->brt_ifidx = ifp->if_index; if (q->brt_ifidx == ifp->if_index) q->brt_age = 1; bridge_copytag(brtag, &q->brt_tunnel); goto want; } if (dir > 0) { if (sc->sc_brtcnt >= sc->sc_brtmax) goto done; p = malloc(sizeof(*p), M_DEVBUF, M_NOWAIT); if (p == NULL) goto done; bcopy(ea, &p->brt_addr, sizeof(p->brt_addr)); p->brt_ifidx = ifp->if_index; p->brt_age = 1; bridge_copytag(brtag, &p->brt_tunnel); if (setflags) p->brt_flags = flags; else p->brt_flags = IFBAF_DYNAMIC; LIST_INSERT_BEFORE(q, p, brt_next); sc->sc_brtcnt++; goto want; } if (p == NULL) { if (sc->sc_brtcnt >= sc->sc_brtmax) goto done; p = malloc(sizeof(*p), M_DEVBUF, M_NOWAIT); if (p == NULL) goto done; bcopy(ea, &p->brt_addr, sizeof(p->brt_addr)); p->brt_ifidx = ifp->if_index; p->brt_age = 1; bridge_copytag(brtag, &p->brt_tunnel); if (setflags) p->brt_flags = flags; else p->brt_flags = IFBAF_DYNAMIC; LIST_INSERT_AFTER(q, p, brt_next); sc->sc_brtcnt++; goto want; } } while (p != NULL); done: error = 1; want: mtx_leave(&sc->sc_mtx); return (error); } unsigned int bridge_rtlookup(struct ifnet *brifp, struct ether_addr *ea, struct mbuf *m) { struct bridge_softc *sc = brifp->if_softc; struct bridge_rtnode *p = NULL; unsigned int ifidx = 0; u_int32_t h; int dir; h = bridge_hash(sc, ea); mtx_enter(&sc->sc_mtx); LIST_FOREACH(p, &sc->sc_rts[h], brt_next) { dir = memcmp(ea, &p->brt_addr, sizeof(p->brt_addr)); if (dir == 0) break; if (dir > 0) { p = NULL; break; } } if (p != NULL) { ifidx = p->brt_ifidx; if (p->brt_family != AF_UNSPEC && m != NULL) { struct bridge_tunneltag *brtag; brtag = bridge_tunneltag(m); if (brtag != NULL) bridge_copytag(&p->brt_tunnel, brtag); } } mtx_leave(&sc->sc_mtx); return (ifidx); } u_int32_t bridge_hash(struct bridge_softc *sc, struct ether_addr *addr) { return SipHash24((SIPHASH_KEY *)sc->sc_hashkey, addr, ETHER_ADDR_LEN) & BRIDGE_RTABLE_MASK; } /* * Perform an aging cycle */ void bridge_rtage(void *vsc) { struct bridge_softc *sc = vsc; struct ifnet *ifp = &sc->sc_if; struct bridge_rtnode *n, *p; int i; if (!ISSET(ifp->if_flags, IFF_RUNNING)) return; mtx_enter(&sc->sc_mtx); for (i = 0; i < BRIDGE_RTABLE_SIZE; i++) { n = LIST_FIRST(&sc->sc_rts[i]); while (n != NULL) { if ((n->brt_flags & IFBAF_TYPEMASK) == IFBAF_STATIC) { n->brt_age = !n->brt_age; if (n->brt_age) n->brt_age = 0; n = LIST_NEXT(n, brt_next); } else if (n->brt_age) { n->brt_age = 0; n = LIST_NEXT(n, brt_next); } else { p = LIST_NEXT(n, brt_next); LIST_REMOVE(n, brt_next); sc->sc_brtcnt--; free(n, M_DEVBUF, sizeof *n); n = p; } } } mtx_leave(&sc->sc_mtx); if (sc->sc_brttimeout != 0) timeout_add_sec(&sc->sc_brtimeout, sc->sc_brttimeout); } void bridge_rtagenode(struct ifnet *ifp, int age) { struct bridge_softc *sc; struct bridge_rtnode *n; struct ifnet *bifp; int i; bifp = if_get(ifp->if_bridgeidx); if (bifp == NULL) return; sc = bifp->if_softc; /* * If the age is zero then flush, otherwise set all the expiry times to * age for the interface */ if (age == 0) bridge_rtdelete(sc, ifp, 1); else { mtx_enter(&sc->sc_mtx); for (i = 0; i < BRIDGE_RTABLE_SIZE; i++) { LIST_FOREACH(n, &sc->sc_rts[i], brt_next) { /* Cap the expiry time to 'age' */ if (n->brt_ifidx == ifp->if_index && n->brt_age > time_uptime + age && (n->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) n->brt_age = time_uptime + age; } } mtx_leave(&sc->sc_mtx); } if_put(bifp); } /* * Remove all dynamic addresses from the cache */ void bridge_rtflush(struct bridge_softc *sc, int full) { int i; struct bridge_rtnode *p, *n; mtx_enter(&sc->sc_mtx); for (i = 0; i < BRIDGE_RTABLE_SIZE; i++) { n = LIST_FIRST(&sc->sc_rts[i]); while (n != NULL) { if (full || (n->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) { p = LIST_NEXT(n, brt_next); LIST_REMOVE(n, brt_next); sc->sc_brtcnt--; free(n, M_DEVBUF, sizeof *n); n = p; } else n = LIST_NEXT(n, brt_next); } } mtx_leave(&sc->sc_mtx); } /* * Remove an address from the cache */ int bridge_rtdaddr(struct bridge_softc *sc, struct ether_addr *ea) { int h; struct bridge_rtnode *p; h = bridge_hash(sc, ea); mtx_enter(&sc->sc_mtx); LIST_FOREACH(p, &sc->sc_rts[h], brt_next) { if (memcmp(ea, &p->brt_addr, sizeof(p->brt_addr)) == 0) { LIST_REMOVE(p, brt_next); sc->sc_brtcnt--; mtx_leave(&sc->sc_mtx); free(p, M_DEVBUF, sizeof *p); return (0); } } mtx_leave(&sc->sc_mtx); return (ENOENT); } /* * Delete routes to a specific interface member. */ void bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int dynonly) { int i; struct bridge_rtnode *n, *p; /* * Loop through all of the hash buckets and traverse each * chain looking for routes to this interface. */ mtx_enter(&sc->sc_mtx); for (i = 0; i < BRIDGE_RTABLE_SIZE; i++) { n = LIST_FIRST(&sc->sc_rts[i]); while (n != NULL) { if (n->brt_ifidx != ifp->if_index) { /* Not ours */ n = LIST_NEXT(n, brt_next); continue; } if (dynonly && (n->brt_flags & IFBAF_TYPEMASK) != IFBAF_DYNAMIC) { /* only deleting dynamics */ n = LIST_NEXT(n, brt_next); continue; } p = LIST_NEXT(n, brt_next); LIST_REMOVE(n, brt_next); sc->sc_brtcnt--; free(n, M_DEVBUF, sizeof *n); n = p; } } mtx_leave(&sc->sc_mtx); } /* * Gather all of the routes for this interface. */ int bridge_rtfind(struct bridge_softc *sc, struct ifbaconf *baconf) { struct ifbareq *bareq, *bareqs = NULL; struct bridge_rtnode *n; u_int32_t i = 0, total = 0; int k, error = 0; mtx_enter(&sc->sc_mtx); for (k = 0; k < BRIDGE_RTABLE_SIZE; k++) { LIST_FOREACH(n, &sc->sc_rts[k], brt_next) total++; } mtx_leave(&sc->sc_mtx); if (baconf->ifbac_len == 0) { i = total; goto done; } total = MIN(total, baconf->ifbac_len / sizeof(*bareqs)); bareqs = mallocarray(total, sizeof(*bareqs), M_TEMP, M_NOWAIT|M_ZERO); if (bareqs == NULL) goto done; mtx_enter(&sc->sc_mtx); for (k = 0; k < BRIDGE_RTABLE_SIZE; k++) { LIST_FOREACH(n, &sc->sc_rts[k], brt_next) { struct ifnet *ifp; if (i >= total) { mtx_leave(&sc->sc_mtx); goto done; } bareq = &bareqs[i]; ifp = if_get(n->brt_ifidx); if (ifp == NULL) continue; bcopy(ifp->if_xname, bareq->ifba_ifsname, sizeof(bareq->ifba_ifsname)); if_put(ifp); bcopy(sc->sc_if.if_xname, bareq->ifba_name, sizeof(bareq->ifba_name)); bcopy(&n->brt_addr, &bareq->ifba_dst, sizeof(bareq->ifba_dst)); bridge_copyaddr(&n->brt_tunnel.brtag_peer.sa, sstosa(&bareq->ifba_dstsa)); bareq->ifba_age = n->brt_age; bareq->ifba_flags = n->brt_flags; i++; } } mtx_leave(&sc->sc_mtx); error = copyout(bareqs, baconf->ifbac_req, i * sizeof(*bareqs)); done: free(bareqs, M_TEMP, total * sizeof(*bareqs)); baconf->ifbac_len = i * sizeof(*bareqs); return (error); } void bridge_update(struct ifnet *ifp, struct ether_addr *ea, int delete) { struct bridge_softc *sc; struct bridge_iflist *bif; u_int8_t *addr; addr = (u_int8_t *)ea; bif = bridge_getbif(ifp); if (bif == NULL) return; sc = bif->bridge_sc; if (sc == NULL) return; /* * Update the bridge interface if it is in * the learning state. */ if ((bif->bif_flags & IFBIF_LEARNING) && (ETHER_IS_MULTICAST(addr) == 0) && !(addr[0] == 0 && addr[1] == 0 && addr[2] == 0 && addr[3] == 0 && addr[4] == 0 && addr[5] == 0)) { /* Care must be taken with spanning tree */ if ((bif->bif_flags & IFBIF_STP) && (bif->bif_state == BSTP_IFSTATE_DISCARDING)) return; /* Delete the address from the bridge */ bridge_rtdaddr(sc, ea); if (!delete) { /* Update the bridge table */ bridge_rtupdate(sc, ea, ifp, 0, IFBAF_DYNAMIC, NULL); } } } /* * bridge filter/matching rules */ int bridge_brlconf(struct bridge_iflist *bif, struct ifbrlconf *bc) { struct bridge_softc *sc = bif->bridge_sc; struct brl_node *n; struct ifbrlreq *req, *reqs = NULL; int error = 0; u_int32_t i = 0, total = 0; SIMPLEQ_FOREACH(n, &bif->bif_brlin, brl_next) { total++; } SIMPLEQ_FOREACH(n, &bif->bif_brlout, brl_next) { total++; } if (bc->ifbrl_len == 0) { i = total; goto done; } reqs = mallocarray(total, sizeof(*reqs), M_TEMP, M_NOWAIT|M_ZERO); if (reqs == NULL) goto done; SIMPLEQ_FOREACH(n, &bif->bif_brlin, brl_next) { if (bc->ifbrl_len < (i + 1) * sizeof(*reqs)) goto done; req = &reqs[i]; strlcpy(req->ifbr_name, sc->sc_if.if_xname, IFNAMSIZ); strlcpy(req->ifbr_ifsname, bif->ifp->if_xname, IFNAMSIZ); req->ifbr_action = n->brl_action; req->ifbr_flags = n->brl_flags; req->ifbr_src = n->brl_src; req->ifbr_dst = n->brl_dst; req->ifbr_arpf = n->brl_arpf; #if NPF > 0 req->ifbr_tagname[0] = '\0'; if (n->brl_tag) pf_tag2tagname(n->brl_tag, req->ifbr_tagname); #endif i++; } SIMPLEQ_FOREACH(n, &bif->bif_brlout, brl_next) { if (bc->ifbrl_len < (i + 1) * sizeof(*reqs)) goto done; req = &reqs[i]; strlcpy(req->ifbr_name, sc->sc_if.if_xname, IFNAMSIZ); strlcpy(req->ifbr_ifsname, bif->ifp->if_xname, IFNAMSIZ); req->ifbr_action = n->brl_action; req->ifbr_flags = n->brl_flags; req->ifbr_src = n->brl_src; req->ifbr_dst = n->brl_dst; req->ifbr_arpf = n->brl_arpf; #if NPF > 0 req->ifbr_tagname[0] = '\0'; if (n->brl_tag) pf_tag2tagname(n->brl_tag, req->ifbr_tagname); #endif i++; } error = copyout(reqs, bc->ifbrl_buf, i * sizeof(*reqs)); done: free(reqs, M_TEMP, total * sizeof(*reqs)); bc->ifbrl_len = i * sizeof(*reqs); return (error); } u_int8_t bridge_arpfilter(struct brl_node *n, struct ether_header *eh, struct mbuf *m) { struct ether_arp ea; if (!(n->brl_arpf.brla_flags & (BRLA_ARP|BRLA_RARP))) return (1); if (ntohs(eh->ether_type) != ETHERTYPE_ARP) return (0); if (m->m_pkthdr.len <= ETHER_HDR_LEN + sizeof(ea)) return (0); /* log error? */ m_copydata(m, ETHER_HDR_LEN, sizeof(ea), (caddr_t)&ea); if (ntohs(ea.arp_hrd) != ARPHRD_ETHER || ntohs(ea.arp_pro) != ETHERTYPE_IP || ea.arp_hln != ETHER_ADDR_LEN || ea.arp_pln != sizeof(struct in_addr)) return (0); if ((n->brl_arpf.brla_flags & BRLA_ARP) && ntohs(ea.arp_op) != ARPOP_REQUEST && ntohs(ea.arp_op) != ARPOP_REPLY) return (0); if ((n->brl_arpf.brla_flags & BRLA_RARP) && ntohs(ea.arp_op) != ARPOP_REVREQUEST && ntohs(ea.arp_op) != ARPOP_REVREPLY) return (0); if (n->brl_arpf.brla_op && ntohs(ea.arp_op) != n->brl_arpf.brla_op) return (0); if (n->brl_arpf.brla_flags & BRLA_SHA && memcmp(ea.arp_sha, &n->brl_arpf.brla_sha, ETHER_ADDR_LEN)) return (0); if (n->brl_arpf.brla_flags & BRLA_THA && memcmp(ea.arp_tha, &n->brl_arpf.brla_tha, ETHER_ADDR_LEN)) return (0); if (n->brl_arpf.brla_flags & BRLA_SPA && memcmp(ea.arp_spa, &n->brl_arpf.brla_spa, sizeof(struct in_addr))) return (0); if (n->brl_arpf.brla_flags & BRLA_TPA && memcmp(ea.arp_tpa, &n->brl_arpf.brla_tpa, sizeof(struct in_addr))) return (0); return (1); } u_int8_t bridge_filterrule(struct brl_head *h, struct ether_header *eh, struct mbuf *m) { struct brl_node *n; u_int8_t action, flags; if (SIMPLEQ_EMPTY(h)) return (BRL_ACTION_PASS); KERNEL_LOCK(); SIMPLEQ_FOREACH(n, h, brl_next) { if (!bridge_arpfilter(n, eh, m)) continue; flags = n->brl_flags & (BRL_FLAG_SRCVALID|BRL_FLAG_DSTVALID); if (flags == 0) goto return_action; if (flags == (BRL_FLAG_SRCVALID|BRL_FLAG_DSTVALID)) { if (memcmp(eh->ether_shost, &n->brl_src, ETHER_ADDR_LEN)) continue; if (memcmp(eh->ether_dhost, &n->brl_dst, ETHER_ADDR_LEN)) continue; goto return_action; } if (flags == BRL_FLAG_SRCVALID) { if (memcmp(eh->ether_shost, &n->brl_src, ETHER_ADDR_LEN)) continue; goto return_action; } if (flags == BRL_FLAG_DSTVALID) { if (memcmp(eh->ether_dhost, &n->brl_dst, ETHER_ADDR_LEN)) continue; goto return_action; } } KERNEL_UNLOCK(); return (BRL_ACTION_PASS); return_action: #if NPF > 0 pf_tag_packet(m, n->brl_tag, -1); #endif action = n->brl_action; KERNEL_UNLOCK(); return (action); } int bridge_addrule(struct bridge_iflist *bif, struct ifbrlreq *req, int out) { struct brl_node *n; n = malloc(sizeof(*n), M_DEVBUF, M_NOWAIT); if (n == NULL) return (ENOMEM); bcopy(&req->ifbr_src, &n->brl_src, sizeof(struct ether_addr)); bcopy(&req->ifbr_dst, &n->brl_dst, sizeof(struct ether_addr)); n->brl_action = req->ifbr_action; n->brl_flags = req->ifbr_flags; n->brl_arpf = req->ifbr_arpf; #if NPF > 0 if (req->ifbr_tagname[0]) n->brl_tag = pf_tagname2tag(req->ifbr_tagname, 1); else n->brl_tag = 0; #endif KERNEL_ASSERT_LOCKED(); if (out) { n->brl_flags &= ~BRL_FLAG_IN; n->brl_flags |= BRL_FLAG_OUT; SIMPLEQ_INSERT_TAIL(&bif->bif_brlout, n, brl_next); } else { n->brl_flags &= ~BRL_FLAG_OUT; n->brl_flags |= BRL_FLAG_IN; SIMPLEQ_INSERT_TAIL(&bif->bif_brlin, n, brl_next); } return (0); } void bridge_flushrule(struct bridge_iflist *bif) { struct brl_node *p; KERNEL_ASSERT_LOCKED(); while (!SIMPLEQ_EMPTY(&bif->bif_brlin)) { p = SIMPLEQ_FIRST(&bif->bif_brlin); SIMPLEQ_REMOVE_HEAD(&bif->bif_brlin, brl_next); #if NPF > 0 pf_tag_unref(p->brl_tag); #endif free(p, M_DEVBUF, sizeof *p); } while (!SIMPLEQ_EMPTY(&bif->bif_brlout)) { p = SIMPLEQ_FIRST(&bif->bif_brlout); SIMPLEQ_REMOVE_HEAD(&bif->bif_brlout, brl_next); #if NPF > 0 pf_tag_unref(p->brl_tag); #endif free(p, M_DEVBUF, sizeof *p); } } struct bridge_tunneltag * bridge_tunnel(struct mbuf *m) { struct m_tag *mtag; if ((mtag = m_tag_find(m, PACKET_TAG_TUNNEL, NULL)) == NULL) return (NULL); return ((struct bridge_tunneltag *)(mtag + 1)); } struct bridge_tunneltag * bridge_tunneltag(struct mbuf *m) { struct m_tag *mtag; if ((mtag = m_tag_find(m, PACKET_TAG_TUNNEL, NULL)) == NULL) { mtag = m_tag_get(PACKET_TAG_TUNNEL, sizeof(struct bridge_tunneltag), M_NOWAIT); if (mtag == NULL) return (NULL); bzero(mtag + 1, sizeof(struct bridge_tunneltag)); m_tag_prepend(m, mtag); } return ((struct bridge_tunneltag *)(mtag + 1)); } void bridge_tunneluntag(struct mbuf *m) { struct m_tag *mtag; if ((mtag = m_tag_find(m, PACKET_TAG_TUNNEL, NULL)) != NULL) m_tag_delete(m, mtag); } void bridge_copyaddr(struct sockaddr *src, struct sockaddr *dst) { if (src != NULL && src->sa_family != AF_UNSPEC) memcpy(dst, src, src->sa_len); else { dst->sa_family = AF_UNSPEC; dst->sa_len = 0; } } void bridge_copytag(struct bridge_tunneltag *src, struct bridge_tunneltag *dst) { if (src == NULL) { memset(dst, 0, sizeof(*dst)); } else { bridge_copyaddr(&src->brtag_peer.sa, &dst->brtag_peer.sa); bridge_copyaddr(&src->brtag_local.sa, &dst->brtag_local.sa); dst->brtag_id = src->brtag_id; } }