Please note that diffs are not public domain; they are subject to the copyright notices on the relevant files. =================================================================== RCS file: /ftp/cvs/cvsroot/src/sys/netinet6/mld6.c,v rcsdiff: /ftp/cvs/cvsroot/src/sys/netinet6/mld6.c,v: warning: Unknown phrases like `commitid ...;' are present. retrieving revision 1.36.4.1 retrieving revision 1.85 diff -u -p -r1.36.4.1 -r1.85 --- src/sys/netinet6/mld6.c 2007/03/12 05:59:59 1.36.4.1 +++ src/sys/netinet6/mld6.c 2017/03/01 09:09:37 1.85 @@ -1,4 +1,4 @@ -/* $NetBSD: mld6.c,v 1.36.4.1 2007/03/12 05:59:59 rmind Exp $ */ +/* $NetBSD: mld6.c,v 1.85 2017/03/01 09:09:37 ozaki-r Exp $ */ /* $KAME: mld6.c,v 1.25 2001/01/16 14:14:18 itojun Exp $ */ /* @@ -102,19 +102,24 @@ */ #include -__KERNEL_RCSID(0, "$NetBSD: mld6.c,v 1.36.4.1 2007/03/12 05:59:59 rmind Exp $"); +__KERNEL_RCSID(0, "$NetBSD: mld6.c,v 1.85 2017/03/01 09:09:37 ozaki-r Exp $"); +#ifdef _KERNEL_OPT #include "opt_inet.h" +#include "opt_net_mpsafe.h" +#endif #include #include #include #include -#include +#include #include #include #include #include +#include +#include #include @@ -125,23 +130,13 @@ __KERNEL_RCSID(0, "$NetBSD: mld6.c,v 1.3 #include #include #include +#include #include #include -/* - * This structure is used to keep track of in6_multi chains which belong to - * deleted interface addresses. - */ -static LIST_HEAD(, multi6_kludge) in6_mk; /* XXX BSS initialization */ - -struct multi6_kludge { - LIST_ENTRY(multi6_kludge) mk_entry; - struct ifnet *mk_ifp; - struct in6_multihead mk_head; -}; - +static krwlock_t in6_multilock __cacheline_aligned; /* * Protocol constants @@ -163,11 +158,14 @@ static struct mld_hdr * mld_allocbuf(str static void mld_sendpkt(struct in6_multi *, int, const struct in6_addr *); static void mld_starttimer(struct in6_multi *); static void mld_stoptimer(struct in6_multi *); -static void mld_timeo(struct in6_multi *); static u_long mld_timerresid(struct in6_multi *); +static void in6m_ref(struct in6_multi *); +static void in6m_unref(struct in6_multi *); +static void in6m_destroy(struct in6_multi *); + void -mld_init() +mld_init(void) { static u_int8_t hbh_buf[8]; struct ip6_hbh *hbh = (struct ip6_hbh *)hbh_buf; @@ -181,19 +179,24 @@ mld_init() hbh_buf[3] = 0; hbh_buf[4] = IP6OPT_RTALERT; hbh_buf[5] = IP6OPT_RTALERT_LEN - 2; - bcopy((void *)&rtalert_code, &hbh_buf[6], sizeof(u_int16_t)); + memcpy(&hbh_buf[6], (void *)&rtalert_code, sizeof(u_int16_t)); ip6_opts.ip6po_hbh = hbh; /* We will specify the hoplimit by a multicast option. */ ip6_opts.ip6po_hlim = -1; + ip6_opts.ip6po_prefer_tempaddr = IP6PO_TEMPADDR_NOTPREFER; + + rw_init(&in6_multilock); } static void -mld_starttimer(in6m) - struct in6_multi *in6m; +mld_starttimer(struct in6_multi *in6m) { struct timeval now; + KASSERT(rw_write_held(&in6_multilock)); + KASSERT(in6m->in6m_timer != IN6M_TIMER_UNDEF); + microtime(&now); in6m->in6m_timer_expire.tv_sec = now.tv_sec + in6m->in6m_timer / hz; in6m->in6m_timer_expire.tv_usec = now.tv_usec + @@ -204,31 +207,50 @@ mld_starttimer(in6m) } /* start or restart the timer */ - callout_reset(in6m->in6m_timer_ch, in6m->in6m_timer, - (void (*) __P((void *)))mld_timeo, in6m); + callout_schedule(&in6m->in6m_timer_ch, in6m->in6m_timer); } +/* + * mld_stoptimer releases in6_multilock when calling callout_halt. + * The caller must ensure in6m won't be freed while releasing the lock. + */ static void -mld_stoptimer(in6m) - struct in6_multi *in6m; +mld_stoptimer(struct in6_multi *in6m) { + + KASSERT(rw_write_held(&in6_multilock)); + if (in6m->in6m_timer == IN6M_TIMER_UNDEF) return; - callout_stop(in6m->in6m_timer_ch); + rw_exit(&in6_multilock); + + if (mutex_owned(softnet_lock)) + callout_halt(&in6m->in6m_timer_ch, softnet_lock); + else + callout_halt(&in6m->in6m_timer_ch, NULL); + + rw_enter(&in6_multilock, RW_WRITER); in6m->in6m_timer = IN6M_TIMER_UNDEF; } static void -mld_timeo(in6m) - struct in6_multi *in6m; +mld_timeo(void *arg) { - int s = splsoftnet(); + struct in6_multi *in6m = arg; - in6m->in6m_timer = IN6M_TIMER_UNDEF; + KASSERT(in6m->in6m_refcount > 0); - callout_stop(in6m->in6m_timer_ch); +#ifndef NET_MPSAFE + mutex_enter(softnet_lock); + KERNEL_LOCK(1, NULL); +#endif + rw_enter(&in6_multilock, RW_WRITER); + if (in6m->in6m_timer == IN6M_TIMER_UNDEF) + goto out; + + in6m->in6m_timer = IN6M_TIMER_UNDEF; switch (in6m->in6m_state) { case MLD_REPORTPENDING: @@ -239,12 +261,18 @@ mld_timeo(in6m) break; } - splx(s); +out: + rw_exit(&in6_multilock); +#ifndef NET_MPSAFE + KERNEL_UNLOCK_ONE(NULL); + mutex_exit(softnet_lock); +#else + return; +#endif } static u_long -mld_timerresid(in6m) - struct in6_multi *in6m; +mld_timerresid(struct in6_multi *in6m) { struct timeval now, diff; @@ -264,15 +292,16 @@ mld_timerresid(in6m) } /* return the remaining time in milliseconds */ - return (((u_long)(diff.tv_sec * 1000000 + diff.tv_usec)) / 1000); + return diff.tv_sec * 1000 + diff.tv_usec / 1000; } static void -mld_start_listening(in6m) - struct in6_multi *in6m; +mld_start_listening(struct in6_multi *in6m) { struct in6_addr all_in6; + KASSERT(rw_write_held(&in6_multilock)); + /* * RFC2710 page 10: * The node never sends a Report or Done for the link-scope all-nodes @@ -292,7 +321,7 @@ mld_start_listening(in6m) in6m->in6m_state = MLD_OTHERLISTENER; } else { mld_sendpkt(in6m, MLD_LISTENER_REPORT, NULL); - in6m->in6m_timer = arc4random() % + in6m->in6m_timer = cprng_fast32() % (MLD_UNSOLICITED_REPORT_INTERVAL * hz); in6m->in6m_state = MLD_IREPORTEDLAST; @@ -301,11 +330,12 @@ mld_start_listening(in6m) } static void -mld_stop_listening(in6m) - struct in6_multi *in6m; +mld_stop_listening(struct in6_multi *in6m) { struct in6_addr allnode, allrouter; + KASSERT(rw_lock_held(&in6_multilock)); + allnode = in6addr_linklocal_allnodes; if (in6_setscope(&allnode, in6m->in6m_ifp, NULL)) { /* XXX: this should not happen! */ @@ -326,22 +356,23 @@ mld_stop_listening(in6m) } void -mld_input(m, off) - struct mbuf *m; - int off; +mld_input(struct mbuf *m, int off) { - struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); + struct ip6_hdr *ip6; struct mld_hdr *mldh; - struct ifnet *ifp = m->m_pkthdr.rcvif; + struct ifnet *ifp; struct in6_multi *in6m = NULL; struct in6_addr mld_addr, all_in6; - struct in6_ifaddr *ia; - int timer = 0; /* timer value in the MLD query header */ + u_long timer = 0; /* timer value in the MLD query header */ + struct psref psref; + ifp = m_get_rcvif_psref(m, &psref); + if (__predict_false(ifp == NULL)) + goto out; IP6_EXTHDR_GET(mldh, struct mld_hdr *, m, off, sizeof(*mldh)); if (mldh == NULL) { - icmp6stat.icp6s_tooshort++; - return; + ICMP6_STATINC(ICMP6_STAT_TOOSHORT); + goto out_nodrop; } /* source address validation */ @@ -367,12 +398,14 @@ mld_input(m, off) * though RFC3590 says "SHOULD log" if the source of a query * is the unspecified address. */ + char ip6bufs[INET6_ADDRSTRLEN]; + char ip6bufm[INET6_ADDRSTRLEN]; log(LOG_INFO, "mld_input: src %s is not link-local (grp=%s)\n", - ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&mldh->mld_addr)); + IN6_PRINT(ip6bufs,&ip6->ip6_src), + IN6_PRINT(ip6bufm, &mldh->mld_addr)); #endif - m_freem(m); - return; + goto out; } /* @@ -381,8 +414,7 @@ mld_input(m, off) mld_addr = mldh->mld_addr; if (in6_setscope(&mld_addr, ifp, NULL)) { /* XXX: this should not happen! */ - m_free(m); - return; + goto out; } /* @@ -398,7 +430,9 @@ mld_input(m, off) * if we sent the last report. */ switch (mldh->mld_type) { - case MLD_LISTENER_QUERY: + case MLD_LISTENER_QUERY: { + struct in6_multi *next; + if (ifp->if_flags & IFF_LOOPBACK) break; @@ -424,11 +458,17 @@ mld_input(m, off) */ timer = ntohs(mldh->mld_maxdelay); - IFP_TO_IA6(ifp, ia); - if (ia == NULL) - break; - - LIST_FOREACH(in6m, &ia->ia6_multiaddrs, in6m_entry) { + rw_enter(&in6_multilock, RW_WRITER); + /* + * mld_stoptimer and mld_sendpkt release in6_multilock + * temporarily, so we have to prevent in6m from being freed + * while releasing the lock by having an extra reference to it. + * + * Also in6_purge_multi might remove items from the list of the + * ifp while releasing the lock. Fortunately in6_purge_multi is + * never executed as long as we have a psref of the ifp. + */ + LIST_FOREACH_SAFE(in6m, &ifp->if_multiaddrs, in6m_entry, next) { if (IN6_ARE_ADDR_EQUAL(&in6m->in6m_addr, &all_in6) || IPV6_ADDR_MC_SCOPE(&in6m->in6m_addr) < IPV6_ADDR_SCOPE_LINKLOCAL) @@ -442,18 +482,24 @@ mld_input(m, off) continue; if (timer == 0) { + in6m_ref(in6m); + /* send a report immediately */ mld_stoptimer(in6m); mld_sendpkt(in6m, MLD_LISTENER_REPORT, NULL); in6m->in6m_state = MLD_IREPORTEDLAST; + + in6m_unref(in6m); /* May free in6m */ } else if (in6m->in6m_timer == IN6M_TIMER_UNDEF || - mld_timerresid(in6m) > (u_long)timer) { - in6m->in6m_timer = arc4random() % - (int)(((long)timer * hz) / 1000); + mld_timerresid(in6m) > timer) { + in6m->in6m_timer = + 1 + (cprng_fast32() % timer) * hz / 1000; mld_starttimer(in6m); } } + rw_exit(&in6_multilock); break; + } case MLD_LISTENER_REPORT: /* @@ -475,11 +521,16 @@ mld_input(m, off) * If we belong to the group being reported, stop * our timer for that group. */ - IN6_LOOKUP_MULTI(mld_addr, ifp, in6m); + rw_enter(&in6_multilock, RW_WRITER); + in6m = in6_lookup_multi(&mld_addr, ifp); if (in6m) { + in6m_ref(in6m); mld_stoptimer(in6m); /* transit to idle state */ in6m->in6m_state = MLD_OTHERLISTENER; /* clear flag */ + in6m_unref(in6m); + in6m = NULL; /* in6m might be freed */ } + rw_exit(&in6_multilock); break; default: /* this is impossible */ #if 0 @@ -493,14 +544,20 @@ mld_input(m, off) break; } +out: m_freem(m); +out_nodrop: + m_put_rcvif_psref(ifp, &psref); } +/* + * XXX mld_sendpkt must be called with in6_multilock held and + * will release in6_multilock before calling ip6_output and + * returning to avoid locking against myself in ip6_output. + */ static void -mld_sendpkt(in6m, type, dst) - struct in6_multi *in6m; - int type; - const struct in6_addr *dst; +mld_sendpkt(struct in6_multi *in6m, int type, + const struct in6_addr *dst) { struct mbuf *mh; struct mld_hdr *mldh; @@ -509,6 +566,10 @@ mld_sendpkt(in6m, type, dst) struct in6_ifaddr *ia = NULL; struct ifnet *ifp = in6m->in6m_ifp; int ignflags; + struct psref psref; + int bound; + + KASSERT(rw_write_held(&in6_multilock)); /* * At first, find a link local address on the outgoing interface @@ -517,20 +578,31 @@ mld_sendpkt(in6m, type, dst) * the case where we first join a link-local address. */ ignflags = (IN6_IFF_NOTREADY|IN6_IFF_ANYCAST) & ~IN6_IFF_TENTATIVE; - if ((ia = in6ifa_ifpforlinklocal(ifp, ignflags)) == NULL) + bound = curlwp_bind(); + ia = in6ifa_ifpforlinklocal_psref(ifp, ignflags, &psref); + if (ia == NULL) { + curlwp_bindx(bound); return; - if ((ia->ia6_flags & IN6_IFF_TENTATIVE)) + } + if ((ia->ia6_flags & IN6_IFF_TENTATIVE)) { + ia6_release(ia, &psref); ia = NULL; + } /* Allocate two mbufs to store IPv6 header and MLD header */ mldh = mld_allocbuf(&mh, sizeof(struct mld_hdr), in6m, type); - if (mldh == NULL) + if (mldh == NULL) { + ia6_release(ia, &psref); + curlwp_bindx(bound); return; + } /* fill src/dst here */ ip6 = mtod(mh, struct ip6_hdr *); ip6->ip6_src = ia ? ia->ia_addr.sin6_addr : in6addr_any; ip6->ip6_dst = dst ? *dst : in6m->in6m_addr; + ia6_release(ia, &psref); + curlwp_bindx(bound); mldh->mld_addr = in6m->in6m_addr; in6_clearscope(&mldh->mld_addr); /* XXX */ @@ -539,7 +611,7 @@ mld_sendpkt(in6m, type, dst) /* construct multicast option */ memset(&im6o, 0, sizeof(im6o)); - im6o.im6o_multicast_ifp = ifp; + im6o.im6o_multicast_if_index = if_get_index(ifp); im6o.im6o_multicast_hlim = 1; /* @@ -549,7 +621,7 @@ mld_sendpkt(in6m, type, dst) im6o.im6o_multicast_loop = (ip6_mrouter != NULL); /* increment output statictics */ - icmp6stat.icp6s_outhist[type]++; + ICMP6_STATINC(ICMP6_STAT_OUTHIST + type); icmp6_ifstat_inc(ifp, ifs6_out_msg); switch (type) { case MLD_LISTENER_QUERY: @@ -563,8 +635,13 @@ mld_sendpkt(in6m, type, dst) break; } + /* XXX we cannot call ip6_output with holding in6_multilock */ + rw_exit(&in6_multilock); + ip6_output(mh, &ip6_opts, NULL, ia ? 0 : IPV6_UNSPECSRC, - &im6o, (struct socket *)NULL, NULL); + &im6o, NULL, NULL); + + rw_enter(&in6_multilock, RW_WRITER); } static struct mld_hdr * @@ -592,7 +669,7 @@ mld_allocbuf(struct mbuf **mh, int len, (*mh)->m_next = md; md->m_next = NULL; - (*mh)->m_pkthdr.rcvif = NULL; + m_reset_rcvif((*mh)); (*mh)->m_pkthdr.len = sizeof(struct ip6_hdr) + len; (*mh)->m_len = sizeof(struct ip6_hdr); MH_ALIGN(*mh, sizeof(struct ip6_hdr)); @@ -616,26 +693,40 @@ mld_allocbuf(struct mbuf **mh, int len, return mldh; } +static void +in6m_ref(struct in6_multi *in6m) +{ + + KASSERT(rw_write_held(&in6_multilock)); + in6m->in6m_refcount++; +} + +static void +in6m_unref(struct in6_multi *in6m) +{ + + KASSERT(rw_write_held(&in6_multilock)); + if (--in6m->in6m_refcount == 0) + in6m_destroy(in6m); +} + /* * Add an address to the list of IP6 multicast addresses for a given interface. */ struct in6_multi * -in6_addmulti(maddr6, ifp, errorp, timer) - struct in6_addr *maddr6; - struct ifnet *ifp; - int *errorp, timer; +in6_addmulti(struct in6_addr *maddr6, struct ifnet *ifp, + int *errorp, int timer) { - struct in6_ifaddr *ia; - struct in6_ifreq ifr; + struct sockaddr_in6 sin6; struct in6_multi *in6m; - int s = splsoftnet(); *errorp = 0; + rw_enter(&in6_multilock, RW_WRITER); /* * See if address already in list. */ - IN6_LOOKUP_MULTI(*maddr6, ifp, in6m); + in6m = in6_lookup_multi(maddr6, ifp); if (in6m != NULL) { /* * Found it; just increment the refrence count. @@ -647,67 +738,40 @@ in6_addmulti(maddr6, ifp, errorp, timer) * and link it into the interface's multicast list. */ in6m = (struct in6_multi *) - malloc(sizeof(*in6m), M_IPMADDR, M_NOWAIT); + malloc(sizeof(*in6m), M_IPMADDR, M_NOWAIT|M_ZERO); if (in6m == NULL) { - splx(s); *errorp = ENOBUFS; - return (NULL); + goto out; } - memset(in6m, 0, sizeof(*in6m)); in6m->in6m_addr = *maddr6; in6m->in6m_ifp = ifp; in6m->in6m_refcount = 1; in6m->in6m_timer = IN6M_TIMER_UNDEF; - in6m->in6m_timer_ch = - malloc(sizeof(*in6m->in6m_timer_ch), M_IPMADDR, M_NOWAIT); - if (in6m->in6m_timer_ch == NULL) { - free(in6m, M_IPMADDR); - splx(s); - return (NULL); - } - IFP_TO_IA6(ifp, ia); - if (ia == NULL) { - /* leaks in6m_timer_ch */ - free(in6m, M_IPMADDR); - splx(s); - *errorp = EADDRNOTAVAIL; /* appropriate? */ - return (NULL); - } - in6m->in6m_ia = ia; - IFAREF(&ia->ia_ifa); /* gain a reference */ - LIST_INSERT_HEAD(&ia->ia6_multiaddrs, in6m, in6m_entry); + callout_init(&in6m->in6m_timer_ch, CALLOUT_MPSAFE); + callout_setfunc(&in6m->in6m_timer_ch, mld_timeo, in6m); + + LIST_INSERT_HEAD(&ifp->if_multiaddrs, in6m, in6m_entry); /* * Ask the network driver to update its multicast reception * filter appropriately for the new address. */ - memset(&ifr.ifr_addr, 0, sizeof(struct sockaddr_in6)); - ifr.ifr_addr.sin6_family = AF_INET6; - ifr.ifr_addr.sin6_len = sizeof(struct sockaddr_in6); - ifr.ifr_addr.sin6_addr = *maddr6; - if (ifp->if_ioctl == NULL) - *errorp = ENXIO; /* XXX: appropriate? */ - else - *errorp = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, - (void *)&ifr); + sockaddr_in6_init(&sin6, maddr6, 0, 0, 0); + *errorp = if_mcast_op(ifp, SIOCADDMULTI, sin6tosa(&sin6)); if (*errorp) { + callout_destroy(&in6m->in6m_timer_ch); LIST_REMOVE(in6m, in6m_entry); - /* leaks in6m_timer_ch */ free(in6m, M_IPMADDR); - IFAFREE(&ia->ia_ifa); - splx(s); - return (NULL); + in6m = NULL; + goto out; } - callout_init(in6m->in6m_timer_ch); in6m->in6m_timer = timer; if (in6m->in6m_timer > 0) { in6m->in6m_state = MLD_REPORTPENDING; mld_starttimer(in6m); - - splx(s); - return (in6m); + goto out; } /* @@ -716,83 +780,149 @@ in6_addmulti(maddr6, ifp, errorp, timer) */ mld_start_listening(in6m); } - splx(s); - return (in6m); +out: + rw_exit(&in6_multilock); + return in6m; +} + +static void +in6m_destroy(struct in6_multi *in6m) +{ + struct in6_ifaddr *ia; + struct sockaddr_in6 sin6; + int s; + + KASSERT(rw_write_held(&in6_multilock)); + KASSERT(in6m->in6m_refcount == 0); + + /* + * No remaining claims to this record; let MLD6 know + * that we are leaving the multicast group. + */ + mld_stop_listening(in6m); + + /* + * Unlink from list. + */ + LIST_REMOVE(in6m, in6m_entry); + + /* + * Delete all references of this multicasting group from + * the membership arrays + */ + s = pserialize_read_enter(); + IN6_ADDRLIST_READER_FOREACH(ia) { + struct in6_multi_mship *imm; + LIST_FOREACH(imm, &ia->ia6_memberships, i6mm_chain) { + if (imm->i6mm_maddr == in6m) + imm->i6mm_maddr = NULL; + } + } + pserialize_read_exit(s); + + /* + * Notify the network driver to update its multicast + * reception filter. + */ + sockaddr_in6_init(&sin6, &in6m->in6m_addr, 0, 0, 0); + if_mcast_op(in6m->in6m_ifp, SIOCDELMULTI, sin6tosa(&sin6)); + + /* Tell mld_timeo we're halting the timer */ + in6m->in6m_timer = IN6M_TIMER_UNDEF; + if (mutex_owned(softnet_lock)) + callout_halt(&in6m->in6m_timer_ch, softnet_lock); + else + callout_halt(&in6m->in6m_timer_ch, NULL); + callout_destroy(&in6m->in6m_timer_ch); + + free(in6m, M_IPMADDR); } /* * Delete a multicast address record. */ void -in6_delmulti(in6m) - struct in6_multi *in6m; +in6_delmulti(struct in6_multi *in6m) { - struct in6_ifreq ifr; - struct in6_ifaddr *ia; - int s = splsoftnet(); + KASSERT(in6m->in6m_refcount > 0); + + rw_enter(&in6_multilock, RW_WRITER); + /* + * The caller should have a reference to in6m. So we don't need to care + * of releasing the lock in mld_stoptimer. + */ mld_stoptimer(in6m); + if (--in6m->in6m_refcount == 0) + in6m_destroy(in6m); + rw_exit(&in6_multilock); +} - if (--in6m->in6m_refcount == 0) { - /* - * No remaining claims to this record; let MLD6 know - * that we are leaving the multicast group. - */ - mld_stop_listening(in6m); +/* + * Look up the in6_multi record for a given IP6 multicast address + * on a given interface. If no matching record is found, "in6m" + * returns NULL. + */ +struct in6_multi * +in6_lookup_multi(const struct in6_addr *addr, const struct ifnet *ifp) +{ + struct in6_multi *in6m; - /* - * Unlink from list. - */ - LIST_REMOVE(in6m, in6m_entry); - if (in6m->in6m_ia != NULL) { - IFAFREE(&in6m->in6m_ia->ia_ifa); /* release reference */ - in6m->in6m_ia = NULL; - } + KASSERT(rw_lock_held(&in6_multilock)); - /* - * Delete all references of this multicasting group from - * the membership arrays - */ - for (ia = in6_ifaddr; ia; ia = ia->ia_next) { - struct in6_multi_mship *imm; - LIST_FOREACH(imm, &ia->ia6_memberships, i6mm_chain) { - if (imm->i6mm_maddr == in6m) - imm->i6mm_maddr = NULL; - } - } + LIST_FOREACH(in6m, &ifp->if_multiaddrs, in6m_entry) { + if (IN6_ARE_ADDR_EQUAL(&in6m->in6m_addr, addr)) + break; + } + return in6m; +} + +bool +in6_multi_group(const struct in6_addr *addr, const struct ifnet *ifp) +{ + bool ingroup; + rw_enter(&in6_multilock, RW_READER); + ingroup = in6_lookup_multi(addr, ifp) != NULL; + rw_exit(&in6_multilock); + + return ingroup; +} + +/* + * Purge in6_multi records associated to the interface. + */ +void +in6_purge_multi(struct ifnet *ifp) +{ + struct in6_multi *in6m, *next; + + rw_enter(&in6_multilock, RW_WRITER); + LIST_FOREACH_SAFE(in6m, &ifp->if_multiaddrs, in6m_entry, next) { /* - * Notify the network driver to update its multicast - * reception filter. + * Normally multicast addresses are already purged at this + * point. Remaining references aren't accessible via ifp, + * so what we can do here is to prevent ifp from being + * accessed via in6m by removing it from the list of ifp. */ - memset(&ifr.ifr_addr, 0, sizeof(struct sockaddr_in6)); - ifr.ifr_addr.sin6_family = AF_INET6; - ifr.ifr_addr.sin6_len = sizeof(struct sockaddr_in6); - ifr.ifr_addr.sin6_addr = in6m->in6m_addr; - (*in6m->in6m_ifp->if_ioctl)(in6m->in6m_ifp, - SIOCDELMULTI, (void *)&ifr); - free(in6m->in6m_timer_ch, M_IPMADDR); - free(in6m, M_IPMADDR); + mld_stoptimer(in6m); + LIST_REMOVE(in6m, in6m_entry); } - splx(s); + rw_exit(&in6_multilock); } - struct in6_multi_mship * -in6_joingroup(ifp, addr, errorp, timer) - struct ifnet *ifp; - struct in6_addr *addr; - int *errorp, timer; +in6_joingroup(struct ifnet *ifp, struct in6_addr *addr, + int *errorp, int timer) { struct in6_multi_mship *imm; - imm = malloc(sizeof(*imm), M_IPMADDR, M_NOWAIT); - if (!imm) { + imm = malloc(sizeof(*imm), M_IPMADDR, M_NOWAIT|M_ZERO); + if (imm == NULL) { *errorp = ENOBUFS; return NULL; } - memset(imm, 0, sizeof(*imm)); imm->i6mm_maddr = in6_addmulti(addr, ifp, errorp, timer); if (!imm->i6mm_maddr) { /* *errorp is already set */ @@ -803,8 +933,7 @@ in6_joingroup(ifp, addr, errorp, timer) } int -in6_leavegroup(imm) - struct in6_multi_mship *imm; +in6_leavegroup(struct in6_multi_mship *imm) { if (imm->i6mm_maddr) { @@ -814,126 +943,138 @@ in6_leavegroup(imm) return 0; } - /* - * Multicast address kludge: - * If there were any multicast addresses attached to this interface address, - * either move them to another address on this interface, or save them until - * such time as this interface is reconfigured for IPv6. + * DEPRECATED: keep it just to avoid breaking old sysctl users. */ -void -in6_savemkludge(oia) - struct in6_ifaddr *oia; +static int +in6_mkludge_sysctl(SYSCTLFN_ARGS) { - struct in6_ifaddr *ia; - struct in6_multi *in6m; - IFP_TO_IA6(oia->ia_ifp, ia); - if (ia) { /* there is another address */ - KASSERT(ia != oia); - while ((in6m = LIST_FIRST(&oia->ia6_multiaddrs)) != NULL) { - LIST_REMOVE(in6m, in6m_entry); - IFAREF(&ia->ia_ifa); - IFAFREE(&in6m->in6m_ia->ia_ifa); - in6m->in6m_ia = ia; - LIST_INSERT_HEAD(&ia->ia6_multiaddrs, in6m, in6m_entry); - } - } else { /* last address on this if deleted, save */ - struct multi6_kludge *mk; - - LIST_FOREACH(mk, &in6_mk, mk_entry) { - if (mk->mk_ifp == oia->ia_ifp) - break; - } - if (mk == NULL) /* this should not happen! */ - panic("in6_savemkludge: no kludge space"); - - while ((in6m = LIST_FIRST(&oia->ia6_multiaddrs)) != NULL) { - LIST_REMOVE(in6m, in6m_entry); - IFAFREE(&in6m->in6m_ia->ia_ifa); /* release reference */ - in6m->in6m_ia = NULL; - LIST_INSERT_HEAD(&mk->mk_head, in6m, in6m_entry); - } - } + if (namelen != 1) + return EINVAL; + *oldlenp = 0; + return 0; } -/* - * Continuation of multicast address hack: - * If there was a multicast group list previously saved for this interface, - * then we re-attach it to the first address configured on the i/f. - */ -void -in6_restoremkludge(ia, ifp) - struct in6_ifaddr *ia; - struct ifnet *ifp; +static int +in6_multicast_sysctl(SYSCTLFN_ARGS) { - struct multi6_kludge *mk; + struct ifnet *ifp; + struct ifaddr *ifa; + struct in6_ifaddr *ia6; struct in6_multi *in6m; + uint32_t tmp; + int error; + size_t written; + struct psref psref, psref_ia; + int bound, s; + + if (namelen != 1) + return EINVAL; + + rw_enter(&in6_multilock, RW_READER); + + bound = curlwp_bind(); + ifp = if_get_byindex(name[0], &psref); + if (ifp == NULL) { + curlwp_bindx(bound); + rw_exit(&in6_multilock); + return ENODEV; + } + + if (oldp == NULL) { + *oldlenp = 0; + s = pserialize_read_enter(); + IFADDR_READER_FOREACH(ifa, ifp) { + LIST_FOREACH(in6m, &ifp->if_multiaddrs, in6m_entry) { + *oldlenp += 2 * sizeof(struct in6_addr) + + sizeof(uint32_t); + } + } + pserialize_read_exit(s); + if_put(ifp, &psref); + curlwp_bindx(bound); + rw_exit(&in6_multilock); + return 0; + } + + error = 0; + written = 0; + s = pserialize_read_enter(); + IFADDR_READER_FOREACH(ifa, ifp) { + if (ifa->ifa_addr->sa_family != AF_INET6) + continue; + + ifa_acquire(ifa, &psref_ia); + pserialize_read_exit(s); + + ia6 = ifatoia6(ifa); + LIST_FOREACH(in6m, &ifp->if_multiaddrs, in6m_entry) { + if (written + 2 * sizeof(struct in6_addr) + + sizeof(uint32_t) > *oldlenp) + goto done; + /* + * XXX return the first IPv6 address to keep backward + * compatibility, however now multicast addresses + * don't belong to any IPv6 addresses so it should be + * unnecessary. + */ + error = sysctl_copyout(l, &ia6->ia_addr.sin6_addr, + oldp, sizeof(struct in6_addr)); + if (error) + goto done; + oldp = (char *)oldp + sizeof(struct in6_addr); + written += sizeof(struct in6_addr); + error = sysctl_copyout(l, &in6m->in6m_addr, + oldp, sizeof(struct in6_addr)); + if (error) + goto done; + oldp = (char *)oldp + sizeof(struct in6_addr); + written += sizeof(struct in6_addr); + tmp = in6m->in6m_refcount; + error = sysctl_copyout(l, &tmp, oldp, sizeof(tmp)); + if (error) + goto done; + oldp = (char *)oldp + sizeof(tmp); + written += sizeof(tmp); + } - LIST_FOREACH(mk, &in6_mk, mk_entry) { - if (mk->mk_ifp == ifp) - break; - } - if (mk == NULL) - return; - while ((in6m = LIST_FIRST(&mk->mk_head)) != NULL) { - LIST_REMOVE(in6m, in6m_entry); - in6m->in6m_ia = ia; - IFAREF(&ia->ia_ifa); - LIST_INSERT_HEAD(&ia->ia6_multiaddrs, in6m, in6m_entry); - } -} - -/* - * Allocate space for the kludge at interface initialization time. - * Formerly, we dynamically allocated the space in in6_savemkludge() with - * malloc(M_WAITOK). However, it was wrong since the function could be called - * under an interrupt context (software timer on address lifetime expiration). - * Also, we cannot just give up allocating the strucutre, since the group - * membership structure is very complex and we need to keep it anyway. - * Of course, this function MUST NOT be called under an interrupt context. - * Specifically, it is expected to be called only from in6_ifattach(), though - * it is a global function. - */ -void -in6_createmkludge(ifp) - struct ifnet *ifp; -{ - struct multi6_kludge *mk; + s = pserialize_read_enter(); + ifa_release(ifa, &psref_ia); - LIST_FOREACH(mk, &in6_mk, mk_entry) { - /* If we've already had one, do not allocate. */ - if (mk->mk_ifp == ifp) - return; + break; } - - mk = malloc(sizeof(*mk), M_IPMADDR, M_WAITOK); - - memset(mk, 0, sizeof(*mk)); - LIST_INIT(&mk->mk_head); - mk->mk_ifp = ifp; - LIST_INSERT_HEAD(&in6_mk, mk, mk_entry); + pserialize_read_exit(s); +done: + ifa_release(ifa, &psref_ia); + if_put(ifp, &psref); + curlwp_bindx(bound); + rw_exit(&in6_multilock); + *oldlenp = written; + return error; } void -in6_purgemkludge(ifp) - struct ifnet *ifp; +in6_sysctl_multicast_setup(struct sysctllog **clog) { - struct multi6_kludge *mk; - struct in6_multi *in6m, *next; - LIST_FOREACH(mk, &in6_mk, mk_entry) { - if (mk->mk_ifp == ifp) - break; - } - if (mk == NULL) - return; - - /* leave from all multicast groups joined */ - for (in6m = LIST_FIRST(&mk->mk_head); in6m != NULL; in6m = next) { - next = LIST_NEXT(in6m, in6m_entry); - in6_delmulti(in6m); - } - LIST_REMOVE(mk, mk_entry); - free(mk, M_IPMADDR); + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT, + CTLTYPE_NODE, "inet6", NULL, + NULL, 0, NULL, 0, + CTL_NET, PF_INET6, CTL_EOL); + + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT, + CTLTYPE_NODE, "multicast", + SYSCTL_DESCR("Multicast information"), + in6_multicast_sysctl, 0, NULL, 0, + CTL_NET, PF_INET6, CTL_CREATE, CTL_EOL); + + sysctl_createv(clog, 0, NULL, NULL, + CTLFLAG_PERMANENT, + CTLTYPE_NODE, "multicast_kludge", + SYSCTL_DESCR("multicast kludge information"), + in6_mkludge_sysctl, 0, NULL, 0, + CTL_NET, PF_INET6, CTL_CREATE, CTL_EOL); }