X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/593a1d5fd87cdf5b46dd5fcb84467b432cea0f91..7ee9d059c4eecf68ae4f8b0fb99ae2471eda79af:/bsd/netinet/raw_ip.c diff --git a/bsd/netinet/raw_ip.c b/bsd/netinet/raw_ip.c index 533184f4e..0b63a3c0d 100644 --- a/bsd/netinet/raw_ip.c +++ b/bsd/netinet/raw_ip.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2000-2007 Apple Inc. All rights reserved. + * Copyright (c) 2000-2010 Apple Inc. All rights reserved. * * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ * @@ -94,6 +94,10 @@ #include #include +#if INET6 +#include +#endif /* INET6 */ + #include #if IPSEC @@ -120,7 +124,6 @@ int rip_shutdown(struct socket *); extern int ipsec_bypass; #endif -extern u_long route_generation; struct inpcbhead ripcb; struct inpcbinfo ripcbinfo; @@ -197,7 +200,7 @@ rip_input(m, iphlen) register struct inpcb *inp; struct inpcb *last = 0; struct mbuf *opts = 0; - int skipit; + int skipit = 0, ret = 0; ripsrc.sin_addr = ip->ip_src; lck_rw_lock_shared(ripcbinfo.mtx); @@ -217,9 +220,9 @@ rip_input(m, iphlen) if (last) { struct mbuf *n = m_copy(m, 0, (int)M_COPYALL); + skipit = 0; #if IPSEC /* check AH/ESP integrity. */ - skipit = 0; if (ipsec_bypass == 0 && n) { if (ipsec4_in_reject_so(n, last->inp_socket)) { m_freem(n); @@ -232,27 +235,36 @@ rip_input(m, iphlen) #if CONFIG_MACF_NET if (n && skipit == 0) { if (mac_inpcb_check_deliver(last, n, AF_INET, - SOCK_RAW) != 0) + SOCK_RAW) != 0) { + m_freem(n); skipit = 1; + } } #endif if (n && skipit == 0) { int error = 0; - if (last->inp_flags & INP_CONTROLOPTS || - last->inp_socket->so_options & SO_TIMESTAMP) - ip_savecontrol(last, &opts, ip, n); + if ((last->inp_flags & INP_CONTROLOPTS) != 0 || + (last->inp_socket->so_options & SO_TIMESTAMP) != 0 || + (last->inp_socket->so_options & SO_TIMESTAMP_MONOTONIC) != 0) { + ret = ip_savecontrol(last, &opts, ip, n); + if (ret != 0) { + m_freem(n); + m_freem(opts); + last = inp; + continue; + } + } if (last->inp_flags & INP_STRIPHDR) { n->m_len -= iphlen; n->m_pkthdr.len -= iphlen; n->m_data += iphlen; } -// ###LOCK need to lock that socket? + so_recv_data_stat(last->inp_socket, m, 0); if (sbappendaddr(&last->inp_socket->so_rcv, (struct sockaddr *)&ripsrc, n, opts, &error) != 0) { sorwakeup(last->inp_socket); - } - else { + } else { if (error) { /* should notify about lost packet */ kprintf("rip_input can't append to socket\n"); @@ -263,15 +275,15 @@ rip_input(m, iphlen) } last = inp; } - lck_rw_done(ripcbinfo.mtx); + + skipit = 0; #if IPSEC /* check AH/ESP integrity. */ - skipit = 0; if (ipsec_bypass == 0 && last) { if (ipsec4_in_reject_so(m, last->inp_socket)) { m_freem(m); IPSEC_STAT_INCREMENT(ipsecstat.in_polvio); - OSAddAtomic(1, (SInt32*)&ipstat.ips_delivered); + OSAddAtomic(1, &ipstat.ips_delivered); /* do not inject data to pcb */ skipit = 1; } @@ -279,20 +291,30 @@ rip_input(m, iphlen) #endif /*IPSEC*/ #if CONFIG_MACF_NET if (last && skipit == 0) { - if (mac_inpcb_check_deliver(last, m, AF_INET, SOCK_RAW) != 0) + if (mac_inpcb_check_deliver(last, m, AF_INET, SOCK_RAW) != 0) { skipit = 1; + m_freem(m); + } } #endif if (skipit == 0) { if (last) { - if (last->inp_flags & INP_CONTROLOPTS || - last->inp_socket->so_options & SO_TIMESTAMP) - ip_savecontrol(last, &opts, ip, m); + if ((last->inp_flags & INP_CONTROLOPTS) != 0 || + (last->inp_socket->so_options & SO_TIMESTAMP) != 0 || + (last->inp_socket->so_options & SO_TIMESTAMP_MONOTONIC) != 0) { + ret = ip_savecontrol(last, &opts, ip, m); + if (ret != 0) { + m_freem(m); + m_freem(opts); + goto unlock; + } + } if (last->inp_flags & INP_STRIPHDR) { m->m_len -= iphlen; m->m_pkthdr.len -= iphlen; m->m_data += iphlen; } + so_recv_data_stat(last->inp_socket, m, 0); if (sbappendaddr(&last->inp_socket->so_rcv, (struct sockaddr *)&ripsrc, m, opts, NULL) != 0) { sorwakeup(last->inp_socket); @@ -301,10 +323,16 @@ rip_input(m, iphlen) } } else { m_freem(m); - OSAddAtomic(1, (SInt32*)&ipstat.ips_noproto); - OSAddAtomic(-1, (SInt32*)&ipstat.ips_delivered); + OSAddAtomic(1, &ipstat.ips_noproto); + OSAddAtomic(-1, &ipstat.ips_delivered); } } +unlock: + /* + * Keep the list locked because socket filter may force the socket lock + * to be released when calling sbappendaddr() -- see rdar://7627704 + */ + lck_rw_done(ripcbinfo.mtx); } /* @@ -312,14 +340,30 @@ rip_input(m, iphlen) * Tack on options user may have setup with control call. */ int -rip_output(m, so, dst) - register struct mbuf *m; - struct socket *so; - u_long dst; +rip_output( + struct mbuf *m, + struct socket *so, + u_int32_t dst, + struct mbuf *control) { register struct ip *ip; register struct inpcb *inp = sotoinpcb(so); int flags = (so->so_options & SO_DONTROUTE) | IP_ALLOWBROADCAST; + struct ip_out_args ipoa; + struct ip_moptions *imo; + int error = 0; + mbuf_traffic_class_t mtc = MBUF_TC_UNSPEC; + + if (control != NULL) { + mtc = mbuf_traffic_class_from_control(control); + + m_freem(control); + } + /* If socket was bound to an ifindex, tell ip_output about it */ + ipoa.ipoa_boundif = (inp->inp_flags & INP_BOUND_IF) ? + inp->inp_boundif : IFSCOPE_NONE; + ipoa.ipoa_nocell = (inp->inp_flags & INP_NO_IFT_CELLULAR) ? 1 : 0; + flags |= IP_OUTARGS; /* * If the user handed us a complete IP packet, use it. @@ -331,6 +375,8 @@ rip_output(m, so, dst) return(EMSGSIZE); } M_PREPEND(m, sizeof(struct ip), M_WAIT); + if (m == NULL) + return ENOBUFS; ip = mtod(m, struct ip *); ip->ip_tos = inp->inp_ip_tos; ip->ip_off = 0; @@ -362,7 +408,7 @@ rip_output(m, so, dst) #endif /* XXX prevent ip_output from overwriting header fields */ flags |= IP_RAWOUTPUT; - OSAddAtomic(1, (SInt32*)&ipstat.ips_rawout); + OSAddAtomic(1, &ipstat.ips_rawout); } #if IPSEC @@ -372,26 +418,60 @@ rip_output(m, so, dst) } #endif /*IPSEC*/ - if (inp->inp_route.ro_rt && inp->inp_route.ro_rt->generation_id != route_generation) { + if (inp->inp_route.ro_rt != NULL && + inp->inp_route.ro_rt->generation_id != route_generation) { rtfree(inp->inp_route.ro_rt); - inp->inp_route.ro_rt = (struct rtentry *)0; + inp->inp_route.ro_rt = NULL; } + set_packet_tclass(m, so, mtc, 0); + #if CONFIG_MACF_NET mac_mbuf_label_associate_inpcb(inp, m); #endif -#if CONFIG_IP_EDGEHOLE - ip_edgehole_mbuf_tag(inp, m); -#endif + imo = inp->inp_moptions; + if (imo != NULL) + IMO_ADDREF(imo); + /* + * The domain lock is held across ip_output, so it is okay + * to pass the PCB cached route pointer directly to IP and + * the modules beneath it. + */ + error = ip_output(m, inp->inp_options, &inp->inp_route, flags, + imo, &ipoa); -#if CONFIG_FORCE_OUT_IFP - return (ip_output_list(m, 0, inp->inp_options, &inp->inp_route, flags, - inp->inp_moptions, inp->pdp_ifp)); -#else - return (ip_output_list(m, 0, inp->inp_options, &inp->inp_route, flags, - inp->inp_moptions, NULL)); -#endif + if (imo != NULL) + IMO_REMREF(imo); + + if (inp->inp_route.ro_rt != NULL) { + struct rtentry *rt = inp->inp_route.ro_rt; + unsigned int outif; + + if ((rt->rt_flags & (RTF_MULTICAST|RTF_BROADCAST)) || + inp->inp_socket == NULL || + !(inp->inp_socket->so_state & SS_ISCONNECTED)) { + rt = NULL; /* unusable */ + } + /* + * Always discard the cached route for unconnected + * socket or if it is a multicast route. + */ + if (rt == NULL) { + rtfree(inp->inp_route.ro_rt); + inp->inp_route.ro_rt = NULL; + } + /* + * If this is a connected socket and the destination + * route is unicast, update outif with that of the route + * interface index used by IP. + */ + if (rt != NULL && + (outif = rt->rt_ifp->if_index) != inp->inp_last_outif) + inp->inp_last_outif = outif; + } + + return (error); } #if IPFIREWALL @@ -453,7 +533,7 @@ rip_ctloutput(so, sopt) else error = ENOPROTOOPT; break; -#endif IPFIREWALL +#endif /* IPFIREWALL */ #if DUMMYNET case IP_DUMMYNET_GET: @@ -595,15 +675,21 @@ rip_ctlinput( struct in_ifaddr *ia; struct ifnet *ifp; int err; - int flags; + int flags, done = 0; switch (cmd) { case PRC_IFDOWN: - lck_mtx_lock(rt_mtx); + lck_rw_lock_shared(in_ifaddr_rwlock); for (ia = in_ifaddrhead.tqh_first; ia; ia = ia->ia_link.tqe_next) { - if (ia->ia_ifa.ifa_addr == sa - && (ia->ia_flags & IFA_ROUTE)) { + IFA_LOCK(&ia->ia_ifa); + if (ia->ia_ifa.ifa_addr == sa && + (ia->ia_flags & IFA_ROUTE)) { + done = 1; + IFA_ADDREF_LOCKED(&ia->ia_ifa); + IFA_UNLOCK(&ia->ia_ifa); + lck_rw_done(in_ifaddr_rwlock); + lck_mtx_lock(rnh_lock); /* * in_ifscrub kills the interface route. */ @@ -615,23 +701,38 @@ rip_ctlinput( * a routing process they will come back. */ in_ifadown(&ia->ia_ifa, 1); + lck_mtx_unlock(rnh_lock); + IFA_REMREF(&ia->ia_ifa); break; } + IFA_UNLOCK(&ia->ia_ifa); } - lck_mtx_unlock(rt_mtx); + if (!done) + lck_rw_done(in_ifaddr_rwlock); break; case PRC_IFUP: - lck_mtx_lock(rt_mtx); + lck_rw_lock_shared(in_ifaddr_rwlock); for (ia = in_ifaddrhead.tqh_first; ia; ia = ia->ia_link.tqe_next) { - if (ia->ia_ifa.ifa_addr == sa) + IFA_LOCK(&ia->ia_ifa); + if (ia->ia_ifa.ifa_addr == sa) { + /* keep it locked */ break; + } + IFA_UNLOCK(&ia->ia_ifa); } - if (ia == 0 || (ia->ia_flags & IFA_ROUTE)) { - lck_mtx_unlock(rt_mtx); + if (ia == NULL || (ia->ia_flags & IFA_ROUTE) || + (ia->ia_ifa.ifa_debug & IFD_NOTREADY)) { + if (ia != NULL) + IFA_UNLOCK(&ia->ia_ifa); + lck_rw_done(in_ifaddr_rwlock); return; } + IFA_ADDREF_LOCKED(&ia->ia_ifa); + IFA_UNLOCK(&ia->ia_ifa); + lck_rw_done(in_ifaddr_rwlock); + flags = RTF_UP; ifp = ia->ia_ifa.ifa_ifp; @@ -639,20 +740,23 @@ rip_ctlinput( || (ifp->if_flags & IFF_POINTOPOINT)) flags |= RTF_HOST; - err = rtinit_locked(&ia->ia_ifa, RTM_ADD, flags); - lck_mtx_unlock(rt_mtx); - if (err == 0) + err = rtinit(&ia->ia_ifa, RTM_ADD, flags); + if (err == 0) { + IFA_LOCK_SPIN(&ia->ia_ifa); ia->ia_flags |= IFA_ROUTE; + IFA_UNLOCK(&ia->ia_ifa); + } + IFA_REMREF(&ia->ia_ifa); break; } } -u_long rip_sendspace = RIPSNDQ; -u_long rip_recvspace = RIPRCVQ; +u_int32_t rip_sendspace = RIPSNDQ; +u_int32_t rip_recvspace = RIPRCVQ; -SYSCTL_INT(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW, +SYSCTL_INT(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW | CTLFLAG_LOCKED, &rip_sendspace, 0, "Maximum outgoing raw IP datagram size"); -SYSCTL_INT(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW, +SYSCTL_INT(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW | CTLFLAG_LOCKED, &rip_recvspace, 0, "Maximum incoming raw IP datagram size"); static int @@ -720,6 +824,7 @@ rip_bind(struct socket *so, struct sockaddr *nam, __unused struct proc *p) struct inpcb *inp = sotoinpcb(so); struct sockaddr_in *addr = (struct sockaddr_in *)nam; struct ifaddr *ifa = NULL; + unsigned int outif = 0; if (nam->sa_len != sizeof(*addr)) return EINVAL; @@ -731,10 +836,13 @@ rip_bind(struct socket *so, struct sockaddr *nam, __unused struct proc *p) return EADDRNOTAVAIL; } else if (ifa) { - ifafree(ifa); - ifa = NULL; + IFA_LOCK(ifa); + outif = ifa->ifa_ifp->if_index; + IFA_UNLOCK(ifa); + IFA_REMREF(ifa); } inp->inp_laddr = addr->sin_addr; + inp->inp_last_outif = outif; return 0; } @@ -765,10 +873,10 @@ rip_shutdown(struct socket *so) __private_extern__ int rip_send(struct socket *so, __unused int flags, struct mbuf *m, struct sockaddr *nam, - __unused struct mbuf *control, __unused struct proc *p) + struct mbuf *control, __unused struct proc *p) { struct inpcb *inp = sotoinpcb(so); - register u_long dst; + register u_int32_t dst; if (so->so_state & SS_ISCONNECTED) { if (nam) { @@ -783,36 +891,48 @@ rip_send(struct socket *so, __unused int flags, struct mbuf *m, struct sockaddr } dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr; } - return rip_output(m, so, dst); + return rip_output(m, so, dst, control); } /* note: rip_unlock is called from different protos instead of the generic socket_unlock, * it will handle the socket dealloc on last reference * */ int -rip_unlock(struct socket *so, int refcount, int debug) +rip_unlock(struct socket *so, int refcount, void *debug) { - int lr_saved; + void *lr_saved; struct inpcb *inp = sotoinpcb(so); - if (debug == 0) - lr_saved = (unsigned int) __builtin_return_address(0); - else lr_saved = debug; + if (debug == NULL) + lr_saved = __builtin_return_address(0); + else + lr_saved = debug; if (refcount) { - if (so->so_usecount <= 0) - panic("rip_unlock: bad refoucnt so=%p val=%x\n", so, so->so_usecount); + if (so->so_usecount <= 0) { + panic("rip_unlock: bad refoucnt so=%p val=%x lrh= %s\n", + so, so->so_usecount, solockhistory_nr(so)); + /* NOTREACHED */ + } so->so_usecount--; if (so->so_usecount == 0 && (inp->inp_wantcnt == WNT_STOPUSING)) { /* cleanup after last reference */ lck_mtx_unlock(so->so_proto->pr_domain->dom_mtx); lck_rw_lock_exclusive(ripcbinfo.mtx); + if (inp->inp_state != INPCB_STATE_DEAD) { +#if INET6 + if (INP_CHECK_SOCKAF(so, AF_INET6)) + in6_pcbdetach(inp); + else +#endif /* INET6 */ + in_pcbdetach(inp); + } in_pcbdispose(inp); lck_rw_done(ripcbinfo.mtx); return(0); } } - so->unlock_lr[so->next_unlock_lr] = (u_int32_t)lr_saved; + so->unlock_lr[so->next_unlock_lr] = lr_saved; so->next_unlock_lr = (so->next_unlock_lr+1) % SO_LCKDBG_MAX; lck_mtx_unlock(so->so_proto->pr_domain->dom_mtx); return(0); @@ -917,9 +1037,129 @@ rip_pcblist SYSCTL_HANDLER_ARGS return error; } -SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, CTLFLAG_RD, 0, 0, +SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0, rip_pcblist, "S,xinpcb", "List of active raw IP sockets"); +#if !CONFIG_EMBEDDED + +static int +rip_pcblist64 SYSCTL_HANDLER_ARGS +{ +#pragma unused(oidp, arg1, arg2) + int error, i, n; + struct inpcb *inp, **inp_list; + inp_gen_t gencnt; + struct xinpgen xig; + + /* + * The process of preparing the TCB list is too time-consuming and + * resource-intensive to repeat twice on every request. + */ + lck_rw_lock_exclusive(ripcbinfo.mtx); + if (req->oldptr == USER_ADDR_NULL) { + n = ripcbinfo.ipi_count; + req->oldidx = 2 * (sizeof xig) + + (n + n/8) * sizeof(struct xinpcb64); + lck_rw_done(ripcbinfo.mtx); + return 0; + } + + if (req->newptr != USER_ADDR_NULL) { + lck_rw_done(ripcbinfo.mtx); + return EPERM; + } + + /* + * OK, now we're committed to doing something. + */ + gencnt = ripcbinfo.ipi_gencnt; + n = ripcbinfo.ipi_count; + + bzero(&xig, sizeof(xig)); + xig.xig_len = sizeof xig; + xig.xig_count = n; + xig.xig_gen = gencnt; + xig.xig_sogen = so_gencnt; + error = SYSCTL_OUT(req, &xig, sizeof xig); + if (error) { + lck_rw_done(ripcbinfo.mtx); + return error; + } + /* + * We are done if there is no pcb + */ + if (n == 0) { + lck_rw_done(ripcbinfo.mtx); + return 0; + } + + inp_list = _MALLOC(n * sizeof *inp_list, M_TEMP, M_WAITOK); + if (inp_list == 0) { + lck_rw_done(ripcbinfo.mtx); + return ENOMEM; + } + + for (inp = ripcbinfo.listhead->lh_first, i = 0; inp && i < n; + inp = inp->inp_list.le_next) { + if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) + inp_list[i++] = inp; + } + n = i; + + error = 0; + for (i = 0; i < n; i++) { + inp = inp_list[i]; + if (inp->inp_gencnt <= gencnt && inp->inp_state != INPCB_STATE_DEAD) { + struct xinpcb64 xi; + + bzero(&xi, sizeof(xi)); + xi.xi_len = sizeof xi; + inpcb_to_xinpcb64(inp, &xi); + if (inp->inp_socket) + sotoxsocket64(inp->inp_socket, &xi.xi_socket); + error = SYSCTL_OUT(req, &xi, sizeof xi); + } + } + if (!error) { + /* + * Give the user an updated idea of our state. + * If the generation differs from what we told + * her before, she knows that something happened + * while we were processing this request, and it + * might be necessary to retry. + */ + bzero(&xig, sizeof(xig)); + xig.xig_len = sizeof xig; + xig.xig_gen = ripcbinfo.ipi_gencnt; + xig.xig_sogen = so_gencnt; + xig.xig_count = ripcbinfo.ipi_count; + error = SYSCTL_OUT(req, &xig, sizeof xig); + } + FREE(inp_list, M_TEMP); + lck_rw_done(ripcbinfo.mtx); + return error; +} + +SYSCTL_PROC(_net_inet_raw, OID_AUTO, pcblist64, CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0, + rip_pcblist64, "S,xinpcb64", "List of active raw IP sockets"); + +#endif /* !CONFIG_EMBEDDED */ + + +static int +rip_pcblist_n SYSCTL_HANDLER_ARGS +{ +#pragma unused(oidp, arg1, arg2) + int error = 0; + + error = get_pcblist_n(IPPROTO_IP, req, &ripcbinfo); + + return error; +} + +SYSCTL_PROC(_net_inet_raw, OID_AUTO, pcblist_n, CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0, + rip_pcblist_n, "S,xinpcb_n", "List of active raw IP sockets"); + struct pr_usrreqs rip_usrreqs = { rip_abort, pru_accept_notsupp, rip_attach, rip_bind, rip_connect, pru_connect2_notsupp, in_control, rip_detach, rip_disconnect,