X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/b0d623f7f2ae71ed96e60569f61f9a9a27016e80..99c3a10404e5d1ef94397ab4df5a8b74711fc4d3:/bsd/netinet/udp_usrreq.c diff --git a/bsd/netinet/udp_usrreq.c b/bsd/netinet/udp_usrreq.c index b661c10ea..e5462393d 100644 --- a/bsd/netinet/udp_usrreq.c +++ b/bsd/netinet/udp_usrreq.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2000-2008 Apple Inc. All rights reserved. + * Copyright (c) 2000-2012 Apple Inc. All rights reserved. * * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ * @@ -72,6 +72,8 @@ #include #include #include +#include +#include #include @@ -121,35 +123,35 @@ static int udpcksum = 1; #else static int udpcksum = 0; /* XXX */ #endif -SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_RW, +SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_RW | CTLFLAG_LOCKED, &udpcksum, 0, ""); static u_int32_t udps_in_sw_cksum; -SYSCTL_UINT(_net_inet_udp, OID_AUTO, in_sw_cksum, CTLFLAG_RD, +SYSCTL_UINT(_net_inet_udp, OID_AUTO, in_sw_cksum, CTLFLAG_RD | CTLFLAG_LOCKED, &udps_in_sw_cksum, 0, "Number of received packets checksummed in software"); static u_int64_t udps_in_sw_cksum_bytes; -SYSCTL_QUAD(_net_inet_udp, OID_AUTO, in_sw_cksum_bytes, CTLFLAG_RD, +SYSCTL_QUAD(_net_inet_udp, OID_AUTO, in_sw_cksum_bytes, CTLFLAG_RD | CTLFLAG_LOCKED, &udps_in_sw_cksum_bytes, "Amount of received data checksummed in software"); static u_int32_t udps_out_sw_cksum; -SYSCTL_UINT(_net_inet_udp, OID_AUTO, out_sw_cksum, CTLFLAG_RD, +SYSCTL_UINT(_net_inet_udp, OID_AUTO, out_sw_cksum, CTLFLAG_RD | CTLFLAG_LOCKED, &udps_out_sw_cksum, 0, "Number of transmitted packets checksummed in software"); static u_int64_t udps_out_sw_cksum_bytes; -SYSCTL_QUAD(_net_inet_udp, OID_AUTO, out_sw_cksum_bytes, CTLFLAG_RD, +SYSCTL_QUAD(_net_inet_udp, OID_AUTO, out_sw_cksum_bytes, CTLFLAG_RD | CTLFLAG_LOCKED, &udps_out_sw_cksum_bytes, "Amount of transmitted data checksummed in software"); int log_in_vain = 0; -SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW, +SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW | CTLFLAG_LOCKED, &log_in_vain, 0, "Log all incoming UDP packets"); static int blackhole = 0; -SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW, +SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW | CTLFLAG_LOCKED, &blackhole, 0, "Do not send port unreachables for refused connects"); struct inpcbhead udb; /* from udp_var.h */ @@ -179,13 +181,13 @@ static int udp_gc_done = FALSE; /* Garbage collection performed last slowtimo */ #endif struct udpstat udpstat; /* from udp_var.h */ -SYSCTL_STRUCT(_net_inet_udp, UDPCTL_STATS, stats, CTLFLAG_RD, +SYSCTL_STRUCT(_net_inet_udp, UDPCTL_STATS, stats, CTLFLAG_RD | CTLFLAG_LOCKED, &udpstat, udpstat, "UDP statistics (struct udpstat, netinet/udp_var.h)"); -SYSCTL_INT(_net_inet_udp, OID_AUTO, pcbcount, CTLFLAG_RD, +SYSCTL_INT(_net_inet_udp, OID_AUTO, pcbcount, CTLFLAG_RD | CTLFLAG_LOCKED, &udbinfo.ipi_count, 0, "Number of active PCBs"); __private_extern__ int udp_use_randomport = 1; -SYSCTL_INT(_net_inet_udp, OID_AUTO, randomize_ports, CTLFLAG_RW, +SYSCTL_INT(_net_inet_udp, OID_AUTO, randomize_ports, CTLFLAG_RW | CTLFLAG_LOCKED, &udp_use_randomport, 0, "Randomize UDP port numbers"); #if INET6 @@ -254,13 +256,15 @@ udp_input(m, iphlen) register struct udphdr *uh; register struct inpcb *inp; struct mbuf *opts = 0; - int len; + int len, isbroadcast; struct ip save_ip; struct sockaddr *append_sa; struct inpcbinfo *pcbinfo = &udbinfo; struct sockaddr_in udp_in = { sizeof (udp_in), AF_INET, 0, { 0 }, { 0, 0, 0, 0, 0, 0, 0, 0 } }; + struct ip_moptions *imo = NULL; + int foundmembership = 0, ret = 0; #if INET6 struct udp_in6 udp_in6 = { { sizeof (udp_in6.uin6_sin), AF_INET6, 0, 0, @@ -269,6 +273,7 @@ udp_input(m, iphlen) }; struct udp_ip6 udp_ip6; #endif /* INET6 */ + struct ifnet *ifp = (m->m_pkthdr.rcvif != NULL) ? m->m_pkthdr.rcvif: NULL; udpstat.udps_ipackets++; @@ -276,6 +281,9 @@ udp_input(m, iphlen) if (m->m_pkthdr.csum_flags & CSUM_TCP_SUM16) m->m_pkthdr.csum_flags = 0; /* invalidate hwcksum for UDP */ + /* Expect 32-bit aligned data pointer on strict-align platforms */ + MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m); + /* * Strip IP options, if any; should skip this, * make available to user, and use on returned packets, @@ -299,11 +307,16 @@ udp_input(m, iphlen) } ip = mtod(m, struct ip *); } - uh = (struct udphdr *)((caddr_t)ip + iphlen); + uh = (struct udphdr *)(void *)((caddr_t)ip + iphlen); /* destination port of 0 is illegal, based on RFC768. */ - if (uh->uh_dport == 0) + if (uh->uh_dport == 0) { + + if (ifp->if_udp_stat != NULL) + atomic_add_64(&ifp->if_udp_stat->port0, 1); + goto bad; + } KERNEL_DEBUG(DBG_LAYER_IN_BEG, uh->uh_dport, uh->uh_sport, ip->ip_src.s_addr, ip->ip_dst.s_addr, uh->uh_ulen); @@ -316,6 +329,10 @@ udp_input(m, iphlen) if (ip->ip_len != len) { if (len > ip->ip_len || len < sizeof(struct udphdr)) { udpstat.udps_badlen++; + + if (ifp->if_udp_stat != NULL) + atomic_add_64(&ifp->if_udp_stat->badlength, 1); + goto bad; } m_adj(m, len - ip->ip_len); @@ -340,21 +357,23 @@ udp_input(m, iphlen) } else { char b[9]; doudpcksum: - *(uint32_t*)&b[0] = *(uint32_t*)&((struct ipovly *)ip)->ih_x1[0]; - *(uint32_t*)&b[4] = *(uint32_t*)&((struct ipovly *)ip)->ih_x1[4]; - *(uint8_t*)&b[8] = *(uint8_t*)&((struct ipovly *)ip)->ih_x1[8]; - - bzero(((struct ipovly *)ip)->ih_x1, 9); + bcopy(((struct ipovly *)ip)->ih_x1, b, + sizeof (((struct ipovly *)ip)->ih_x1)); + bzero(((struct ipovly *)ip)->ih_x1, + sizeof (((struct ipovly *)ip)->ih_x1)); ((struct ipovly *)ip)->ih_len = uh->uh_ulen; uh->uh_sum = in_cksum(m, len + sizeof (struct ip)); - - *(uint32_t*)&((struct ipovly *)ip)->ih_x1[0] = *(uint32_t*)&b[0]; - *(uint32_t*)&((struct ipovly *)ip)->ih_x1[4] = *(uint32_t*)&b[4]; - *(uint8_t*)&((struct ipovly *)ip)->ih_x1[8] = *(uint8_t*)&b[8]; + bcopy(b, ((struct ipovly *)ip)->ih_x1, + sizeof (((struct ipovly *)ip)->ih_x1)); + udp_in_cksum_stats(len); } if (uh->uh_sum) { udpstat.udps_badsum++; + + if (ifp->if_udp_stat != NULL) + atomic_add_64(&ifp->if_udp_stat->badchksum, 1); + m_freem(m); KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0,0,0,0,0); return; @@ -365,8 +384,9 @@ doudpcksum: udpstat.udps_nosum++; #endif - if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || - in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) { + isbroadcast = in_broadcast(ip->ip_dst, ifp); + + if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || isbroadcast) { int reuse_sock = 0, mcast_delivered = 0; @@ -409,6 +429,16 @@ doudpcksum: if ((inp->inp_vflag & INP_IPV4) == 0) continue; #endif + if (ip_restrictrecvif && ifp != NULL && + (ifp->if_eflags & IFEF_RESTRICTED_RECV) && + !(inp->inp_flags & INP_RECV_ANYIF)) + continue; + + if ((inp->inp_moptions == NULL) && + (ntohl(ip->ip_dst.s_addr) != INADDR_ALLHOSTS_GROUP) && + (isbroadcast == 0) ) + continue; + if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) == WNT_STOPUSING) { continue; @@ -441,6 +471,35 @@ doudpcksum: } } + if (isbroadcast == 0 && (ntohl(ip->ip_dst.s_addr) != INADDR_ALLHOSTS_GROUP)) { + if((imo = inp->inp_moptions) == NULL) { + udp_unlock(inp->inp_socket, 1, 0); + continue; + } else { + struct sockaddr_in group; + int blocked; + + IMO_LOCK(imo); + + bzero(&group, sizeof(struct sockaddr_in)); + group.sin_len = sizeof(struct sockaddr_in); + group.sin_family = AF_INET; + group.sin_addr = ip->ip_dst; + + blocked = imo_multi_filter(imo, ifp, + (struct sockaddr *)&group, + (struct sockaddr *)&udp_in); + if (blocked == MCAST_PASS) + foundmembership = 1; + + IMO_UNLOCK(imo); + if (!foundmembership) { + udp_unlock(inp->inp_socket, 1, 0); + continue; + } + foundmembership = 0; + } + } reuse_sock = inp->inp_socket->so_options& (SO_REUSEPORT|SO_REUSEADDR); { #if IPSEC @@ -485,11 +544,17 @@ doudpcksum: */ if (reuse_sock == 0 || m == NULL) break; + + /* + * Expect 32-bit aligned data pointer on strict-align + * platforms. + */ + MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m); /* * Recompute IP and UDP header pointers for new mbuf */ ip = mtod(m, struct ip *); - uh = (struct udphdr *)((caddr_t)ip + iphlen); + uh = (struct udphdr *)(void *)((caddr_t)ip + iphlen); } lck_rw_done(pcbinfo->mtx); @@ -500,6 +565,10 @@ doudpcksum: * for a broadcast or multicast datgram.) */ udpstat.udps_noportbcast++; + + if (ifp->if_udp_stat != NULL) + atomic_add_64(&ifp->if_udp_stat->port_unreach, 1); + goto bad; } @@ -526,8 +595,14 @@ doudpcksum: KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0,0,0,0,0); return; } + /* + * Expect 32-bit aligned data pointer on strict-align + * platforms. + */ + MBUF_STRICT_DATA_ALIGNMENT_CHECK_32(m); + ip = mtod(m, struct ip *); - uh = (struct udphdr *)((caddr_t)ip + iphlen); + uh = (struct udphdr *)(void *)((caddr_t)ip + iphlen); } /* Check for NAT keepalive packet */ if (payload_len == 1 && *(u_int8_t*)((caddr_t)uh + sizeof(struct udphdr)) == 0xFF) { @@ -535,23 +610,11 @@ doudpcksum: KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0,0,0,0,0); return; } - else if (payload_len == 4 && *(u_int32_t*)((caddr_t)uh + sizeof(struct udphdr)) != 0) { + else if (payload_len == 4 && *(u_int32_t*)(void *)((caddr_t)uh + sizeof(struct udphdr)) != 0) { /* UDP encapsulated IPSec packet to pass through NAT */ - size_t stripsiz; - - stripsiz = sizeof(struct udphdr); - - ip = mtod(m, struct ip *); - ovbcopy((caddr_t)ip, (caddr_t)(((u_char *)ip) + stripsiz), iphlen); - m->m_data += stripsiz; - m->m_len -= stripsiz; - m->m_pkthdr.len -= stripsiz; - ip = mtod(m, struct ip *); - ip->ip_len = ip->ip_len - stripsiz; - ip->ip_p = IPPROTO_ESP; - KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0,0,0,0,0); - esp4_input(m, iphlen); + /* preserve the udp header */ + esp4_input(m, iphlen + sizeof(struct udphdr)); return; } } @@ -561,8 +624,12 @@ doudpcksum: * Locate pcb for datagram. */ inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport, - ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif); + ip->ip_dst, uh->uh_dport, 1, ifp); if (inp == NULL) { + + if (ifp->if_udp_stat != NULL) + atomic_add_64(&ifp->if_udp_stat->port_unreach, 1); + if (log_in_vain) { char buf[MAX_IPv4_STR_LEN]; char buf2[MAX_IPv4_STR_LEN]; @@ -594,7 +661,7 @@ doudpcksum: goto bad; #endif if (blackhole) - if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type != IFT_LOOP) + if (ifp && ifp->if_type != IFT_LOOP) goto bad; *ip = save_ip; ip->ip_len += iphlen; @@ -606,6 +673,10 @@ doudpcksum: if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) { udp_unlock(inp->inp_socket, 1, 0); + + if (ifp->if_udp_stat != NULL) + atomic_add_64(&ifp->if_udp_stat->cleanup, 1); + goto bad; } #if IPSEC @@ -613,8 +684,12 @@ doudpcksum: if (ipsec4_in_reject_so(m, inp->inp_socket)) { IPSEC_STAT_INCREMENT(ipsecstat.in_polvio); udp_unlock(inp->inp_socket, 1, 0); + + if (ifp->if_udp_stat != NULL) + atomic_add_64(&ifp->if_udp_stat->badipsec, 1); + goto bad; - } + } } #endif /*IPSEC*/ @@ -624,8 +699,9 @@ doudpcksum: */ udp_in.sin_port = uh->uh_sport; udp_in.sin_addr = ip->ip_src; - if (inp->inp_flags & INP_CONTROLOPTS - || inp->inp_socket->so_options & SO_TIMESTAMP) { + if ((inp->inp_flags & INP_CONTROLOPTS) != 0 + || (inp->inp_socket->so_options & SO_TIMESTAMP) != 0 + || (inp->inp_socket->so_options & SO_TIMESTAMP_MONOTONIC) != 0) { #if INET6 if (inp->inp_vflag & INP_IPV6) { int savedflags; @@ -633,11 +709,17 @@ doudpcksum: ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip); savedflags = inp->inp_flags; inp->inp_flags &= ~INP_UNMAPPABLEOPTS; - ip6_savecontrol(inp, &opts, &udp_ip6.uip6_ip6, m); + ret = ip6_savecontrol(inp, m, &opts); inp->inp_flags = savedflags; } else #endif - ip_savecontrol(inp, &opts, ip, m); + { + ret = ip_savecontrol(inp, &opts, ip, m); + } + if (ret != 0) { + udp_unlock(inp->inp_socket, 1, 0); + goto bad; + } } m_adj(m, iphlen + sizeof(struct udphdr)); @@ -651,10 +733,14 @@ doudpcksum: } else #endif append_sa = (struct sockaddr *)&udp_in; + if (nstat_collect) { + locked_add_64(&inp->inp_stat->rxpackets, 1); + locked_add_64(&inp->inp_stat->rxbytes, m->m_pkthdr.len); + } + so_recv_data_stat(inp->inp_socket, m, 0); if (sbappendaddr(&inp->inp_socket->so_rcv, append_sa, m, opts, NULL) == 0) { udpstat.udps_fullsock++; - } - else { + } else { sorwakeup(inp->inp_socket); } udp_unlock(inp->inp_socket, 1, 0); @@ -680,10 +766,14 @@ ip_2_ip6_hdr(ip6, ip) ip6->ip6_plen = ip->ip_len; ip6->ip6_nxt = ip->ip_p; ip6->ip6_hlim = ip->ip_ttl; - ip6->ip6_src.s6_addr32[2] = ip6->ip6_dst.s6_addr32[2] = - IPV6_ADDR_INT32_SMP; - ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr; - ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr; + if (ip->ip_src.s_addr) { + ip6->ip6_src.s6_addr32[2] = IPV6_ADDR_INT32_SMP; + ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr; + } + if (ip->ip_dst.s_addr) { + ip6->ip6_dst.s6_addr32[2] = IPV6_ADDR_INT32_SMP; + ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr; + } } #endif @@ -702,6 +792,7 @@ udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n, int off, { struct sockaddr *append_sa; struct mbuf *opts = 0; + int ret = 0; #if CONFIG_MACF_NET if (mac_inpcb_check_deliver(last, n, AF_INET, SOCK_DGRAM) != 0) { @@ -709,8 +800,9 @@ udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n, int off, return; } #endif - if (last->inp_flags & INP_CONTROLOPTS || - last->inp_socket->so_options & SO_TIMESTAMP) { + if ((last->inp_flags & INP_CONTROLOPTS) != 0 || + (last->inp_socket->so_options & SO_TIMESTAMP) != 0 || + (last->inp_socket->so_options & SO_TIMESTAMP_MONOTONIC) != 0) { #if INET6 if (last->inp_vflag & INP_IPV6) { int savedflags; @@ -721,11 +813,20 @@ udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n, int off, } savedflags = last->inp_flags; last->inp_flags &= ~INP_UNMAPPABLEOPTS; - ip6_savecontrol(last, &opts, &pudp_ip6->uip6_ip6, n); + ret = ip6_savecontrol(last, n, &opts); + if (ret != 0) { + last->inp_flags = savedflags; + goto error; + } last->inp_flags = savedflags; } else #endif - ip_savecontrol(last, &opts, ip, n); + { + ret = ip_savecontrol(last, &opts, ip, n); + if (ret != 0) { + goto error; + } + } } #if INET6 if (last->inp_vflag & INP_IPV6) { @@ -737,11 +838,22 @@ udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n, int off, } else #endif append_sa = (struct sockaddr *)pudp_in; + if (nstat_collect) { + locked_add_64(&last->inp_stat->rxpackets, 1); + locked_add_64(&last->inp_stat->rxbytes, n->m_pkthdr.len); + } + so_recv_data_stat(last->inp_socket, n, 0); m_adj(n, off); if (sbappendaddr(&last->inp_socket->so_rcv, append_sa, n, opts, NULL) == 0) { udpstat.udps_fullsock++; - } else + } else { sorwakeup(last->inp_socket); + } + return; +error: + m_freem(n); + m_freem(opts); + return; } /* @@ -765,14 +877,13 @@ udp_ctlinput(cmd, sa, vip) void *vip; { struct ip *ip = vip; - struct udphdr *uh; void (*notify)(struct inpcb *, int) = udp_notify; struct in_addr faddr; struct inpcb *inp; - faddr = ((struct sockaddr_in *)sa)->sin_addr; + faddr = ((struct sockaddr_in *)(void *)sa)->sin_addr; if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY) - return; + return; if (PRC_IS_REDIRECT(cmd)) { ip = 0; @@ -782,12 +893,15 @@ udp_ctlinput(cmd, sa, vip) else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) return; if (ip) { - uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2)); - inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport, - ip->ip_src, uh->uh_sport, 0, NULL); + struct udphdr uh; + + bcopy(((caddr_t)ip + (ip->ip_hl << 2)), &uh, sizeof (uh)); + inp = in_pcblookup_hash(&udbinfo, faddr, uh.uh_dport, + ip->ip_src, uh.uh_sport, 0, NULL); if (inp != NULL && inp->inp_socket != NULL) { udp_lock(inp->inp_socket, 1, 0); - if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) { + if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == + WNT_STOPUSING) { udp_unlock(inp->inp_socket, 1, 0); return; } @@ -804,7 +918,9 @@ udp_ctloutput(struct socket *so, struct sockopt *sopt) int error, optval; struct inpcb *inp; - if (sopt->sopt_level != IPPROTO_UDP) + /* Allow at this level */ + if (sopt->sopt_level != IPPROTO_UDP && + !(sopt->sopt_level == SOL_SOCKET && sopt->sopt_name == SO_FLUSH)) return (ip_ctloutput(so, sopt)); error = 0; @@ -830,6 +946,14 @@ udp_ctloutput(struct socket *so, struct sockopt *sopt) inp->inp_flags &= ~INP_UDP_NOCKSUM; break; + case SO_FLUSH: + if ((error = sooptcopyin(sopt, &optval, sizeof (optval), + sizeof (optval))) != 0) + break; + + error = inp_flush(inp, optval); + break; + default: error = ENOPROTOOPT; break; @@ -952,7 +1076,7 @@ udp_pcblist SYSCTL_HANDLER_ARGS return error; } -SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, 0, 0, +SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0, udp_pcblist, "S,xinpcb", "List of active UDP sockets"); #if !CONFIG_EMBEDDED @@ -1055,11 +1179,38 @@ udp_pcblist64 SYSCTL_HANDLER_ARGS return error; } -SYSCTL_PROC(_net_inet_udp, OID_AUTO, pcblist64, CTLFLAG_RD, 0, 0, +SYSCTL_PROC(_net_inet_udp, OID_AUTO, pcblist64, CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0, udp_pcblist64, "S,xinpcb64", "List of active UDP sockets"); #endif /* !CONFIG_EMBEDDED */ +static int +udp_pcblist_n SYSCTL_HANDLER_ARGS +{ +#pragma unused(oidp, arg1, arg2) + int error = 0; + + error = get_pcblist_n(IPPROTO_UDP, req, &udbinfo); + + return error; +} + +SYSCTL_PROC(_net_inet_udp, OID_AUTO, pcblist_n, CTLFLAG_RD | CTLFLAG_LOCKED, 0, 0, + udp_pcblist_n, "S,xinpcb_n", "List of active UDP sockets"); + + +__private_extern__ void +udp_get_ports_used(unsigned int ifindex, uint8_t *bitfield) +{ + inpcb_get_ports_used(ifindex, bitfield, &udbinfo); +} + +__private_extern__ uint32_t +udp_count_opportunistic(unsigned int ifindex, u_int32_t flags) +{ + return inpcb_count_opportunistic(ifindex, &udbinfo, flags); +} + static __inline__ u_int16_t get_socket_id(struct socket * s) { @@ -1075,6 +1226,71 @@ get_socket_id(struct socket * s) return (val); } +static int +udp_check_pktinfo(struct mbuf *control, struct ifnet **outif, struct in_addr *laddr) +{ + struct cmsghdr *cm = 0; + struct in_pktinfo *pktinfo; + struct ifnet *ifp; + + /* + * XXX: Currently, we assume all the optional information is stored + * in a single mbuf. + */ + if (control->m_next) + return (EINVAL); + + if (control->m_len < CMSG_LEN(0)) + return (EINVAL); + + for (cm = M_FIRST_CMSGHDR(control); cm; cm = M_NXT_CMSGHDR(control, cm)) { + if (cm->cmsg_len < sizeof(struct cmsghdr) || cm->cmsg_len > control->m_len) + return (EINVAL); + + if (cm->cmsg_level != IPPROTO_IP || cm->cmsg_type != IP_PKTINFO) + continue; + + if (cm->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo))) + return (EINVAL); + + pktinfo = (struct in_pktinfo *)(void *)CMSG_DATA(cm); + + /* Check for a valid ifindex in pktinfo */ + ifnet_head_lock_shared(); + + if (pktinfo->ipi_ifindex > if_index) { + ifnet_head_done(); + return (ENXIO); + } + + /* If ipi_ifindex is specified it takes precedence over ipi_spec_dst */ + + if (pktinfo->ipi_ifindex) { + ifp = ifindex2ifnet[pktinfo->ipi_ifindex]; + if (ifp == NULL) { + ifnet_head_done(); + return (ENXIO); + } + + ifnet_head_done(); + + if (outif != NULL) + *outif = ifp; + laddr->s_addr = INADDR_ANY; + break; + } + + ifnet_head_done(); + + /* Use the provided ipi_spec_dst address for temp source address */ + if (outif != NULL) + *outif = NULL; + *laddr = pktinfo->ipi_spec_dst; + break; + } + return (0); +} + static int udp_output(inp, m, addr, control, p) register struct inpcb *inp; @@ -1086,21 +1302,42 @@ udp_output(inp, m, addr, control, p) register struct udpiphdr *ui; register int len = m->m_pkthdr.len; struct sockaddr_in *sin; - struct in_addr origladdr, laddr, faddr; + struct in_addr origladdr, laddr, faddr, pi_laddr; u_short lport, fport; - struct sockaddr_in *ifaddr; - int error = 0, udp_dodisconnect = 0; + struct sockaddr_in ifaddr; + int error = 0, udp_dodisconnect = 0, pktinfo = 0; struct socket *so = inp->inp_socket; int soopts = 0; struct mbuf *inpopts; struct ip_moptions *mopts; struct route ro; - struct ip_out_args ipoa; + struct ip_out_args ipoa = { IFSCOPE_NONE, { 0 }, IPOAF_SELECT_SRCIF }; + struct ifnet *outif = NULL; + struct flowadv *adv = &ipoa.ipoa_flowadv; + mbuf_svc_class_t msc = MBUF_SC_UNSPEC; + struct ifnet *origoutifp; + int flowadv = 0; + + /* Enable flow advisory only when connected */ + flowadv = (so->so_state & SS_ISCONNECTED) ? 1 : 0; + + pi_laddr.s_addr = INADDR_ANY; KERNEL_DEBUG(DBG_FNC_UDP_OUTPUT | DBG_FUNC_START, 0,0,0,0,0); - if (control) - m_freem(control); /* XXX */ + lck_mtx_assert(&inp->inpcb_mtx, LCK_MTX_ASSERT_OWNED); + if (control != NULL) { + msc = mbuf_service_class_from_control(control); + + error = udp_check_pktinfo(control, &outif, &pi_laddr); + + m_freem(control); + if (error) + goto release; + pktinfo++; + if (outif != NULL) + ipoa.ipoa_boundif = outif->if_index; + } KERNEL_DEBUG(DBG_LAYER_OUT_BEG, inp->inp_fport, inp->inp_lport, inp->inp_laddr.s_addr, inp->inp_faddr.s_addr, @@ -1111,11 +1348,26 @@ udp_output(inp, m, addr, control, p) goto release; } - lck_mtx_assert(inp->inpcb_mtx, LCK_MTX_ASSERT_OWNED); - - /* If socket was bound to an ifindex, tell ip_output about it */ - ipoa.ipoa_ifscope = (inp->inp_flags & INP_BOUND_IF) ? - inp->inp_boundif : IFSCOPE_NONE; + if (flowadv && INP_WAIT_FOR_IF_FEEDBACK(inp)) { + /* + * The socket is flow-controlled, drop the packets + * until the inp is not flow controlled + */ + error = ENOBUFS; + goto release; + } + /* + * If socket was bound to an ifindex, tell ip_output about it. + * If the ancillary IP_PKTINFO option contains an interface index, + * it takes precedence over the one specified by IP_BOUND_IF. + */ + if (ipoa.ipoa_boundif == IFSCOPE_NONE && + (inp->inp_flags & INP_BOUND_IF)) { + outif = inp->inp_boundifp; + ipoa.ipoa_boundif = outif->if_index; + } + if (inp->inp_flags & INP_NO_IFT_CELLULAR) + ipoa.ipoa_flags |= IPOAF_NO_CELLULAR; soopts |= IP_OUTARGS; /* If there was a routing change, discard cached route and check @@ -1128,28 +1380,55 @@ udp_output(inp, m, addr, control, p) /* src address is gone? */ if ((ia = ifa_foraddr(inp->inp_laddr.s_addr)) == NULL) { - if (inp->inp_flags & INP_INADDR_ANY) { - /* new src will be set later */ - inp->inp_laddr.s_addr = INADDR_ANY; - } else { + if (((inp->inp_flags & INP_INADDR_ANY) == 0) || + (so->so_state & SS_ISCONNECTED)) { + /* Rdar://5448998 + * If the source address is gone, return an error if: + * - the source was specified + * - the socket was already connected + */ + soevent(so, + (SO_FILT_HINT_LOCKED | + SO_FILT_HINT_NOSRCADDR)); error = EADDRNOTAVAIL; goto release; + } else { + /* new src will be set later */ + inp->inp_laddr.s_addr = INADDR_ANY; + inp->inp_last_outifp = NULL; } } if (ia != NULL) - ifafree(&ia->ia_ifa); + IFA_REMREF(&ia->ia_ifa); if (inp->inp_route.ro_rt != NULL) rtfree(inp->inp_route.ro_rt); inp->inp_route.ro_rt = NULL; } - origladdr= laddr = inp->inp_laddr; + origoutifp = inp->inp_last_outifp; + + /* IP_PKTINFO option check. + * If a temporary scope or src address is provided, use it for this packet only + * and make sure we forget it after sending this datagram. + */ + + if (pi_laddr.s_addr != INADDR_ANY || + (ipoa.ipoa_boundif != IFSCOPE_NONE && pktinfo)) { + laddr = pi_laddr; /* temp src address for this datagram only */ + origladdr.s_addr = INADDR_ANY; + udp_dodisconnect = 1; /* we don't want to keep the laddr or route */ + inp->inp_flags |= INP_INADDR_ANY; /* remember we don't care about src addr.*/ + } else { + origladdr = laddr = inp->inp_laddr; + } + + origoutifp = inp->inp_last_outifp; faddr = inp->inp_faddr; lport = inp->inp_lport; fport = inp->inp_fport; if (addr) { - sin = (struct sockaddr_in *)addr; + sin = (struct sockaddr_in *)(void *)addr; if (faddr.s_addr != INADDR_ANY) { error = EISCONN; goto release; @@ -1159,8 +1438,11 @@ udp_output(inp, m, addr, control, p) * In case we don't have a local port set, go through the full connect. * We don't have a local port yet (ie, we can't be looked up), * so it's not an issue if the input runs at the same time we do this. - */ - error = in_pcbconnect(inp, addr, p); + */ + + if (pi_laddr.s_addr != INADDR_ANY) /* if we have a source address specified, use that */ + inp->inp_laddr = pi_laddr; + error = in_pcbconnect(inp, addr, p, &outif); /* if a scope is specified, use it */ if (error) { goto release; } @@ -1169,20 +1451,26 @@ udp_output(inp, m, addr, control, p) faddr = inp->inp_faddr; fport = inp->inp_fport; udp_dodisconnect = 1; + ipoa.ipoa_boundif = (outif != NULL) ? + outif->if_index : IFSCOPE_NONE; } - else { + else { /* Fast path case * we have a full address and a local port. * use those info to build the packet without changing the pcb * and interfering with the input path. See 3851370 + * Note: if we may have a scope from IP_PKTINFO but the + * priority is always given to the scope provided by INP_BOUND_IF. */ if (laddr.s_addr == INADDR_ANY) { - if ((error = in_pcbladdr(inp, addr, &ifaddr)) != 0) - goto release; - laddr = ifaddr->sin_addr; - inp->inp_flags |= INP_INADDR_ANY; /* from pcbconnect: remember we don't care about src addr.*/ + if ((error = in_pcbladdr(inp, addr, &ifaddr, &outif)) != 0) + goto release; + laddr = ifaddr.sin_addr; + inp->inp_flags |= INP_INADDR_ANY; /* from pcbconnect: remember we don't care about src addr.*/ + ipoa.ipoa_boundif = (outif != NULL) ? + outif->if_index : IFSCOPE_NONE; } - + faddr = sin->sin_addr; fport = sin->sin_port; } @@ -1196,6 +1484,8 @@ udp_output(inp, m, addr, control, p) #if CONFIG_MACF_NET mac_mbuf_label_associate_inpcb(inp, m); #endif + if (inp->inp_flowhash == 0) + inp->inp_flowhash = inp_calc_flowhash(inp); /* * Calculate data length and get a mbuf @@ -1245,45 +1535,107 @@ udp_output(inp, m, addr, control, p) goto abort; } #endif /*IPSEC*/ - m->m_pkthdr.socket_id = get_socket_id(inp->inp_socket); inpopts = inp->inp_options; soopts |= (inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST)); mopts = inp->inp_moptions; + if (mopts != NULL) { + IMO_LOCK(mopts); + IMO_ADDREF_LOCKED(mopts); + if (IN_MULTICAST(ntohl(ui->ui_dst.s_addr)) && + mopts->imo_multicast_ifp != NULL) { + inp->inp_last_outifp = mopts->imo_multicast_ifp; + } + IMO_UNLOCK(mopts); + } /* Copy the cached route and take an extra reference */ inp_route_copyout(inp, &ro); + set_packet_service_class(m, so, msc, 0); + m->m_pkthdr.socket_id = get_socket_id(inp->inp_socket); + m->m_pkthdr.m_flowhash = inp->inp_flowhash; + m->m_pkthdr.m_fhflags |= PF_TAG_FLOWHASH; + if (flowadv) + m->m_pkthdr.m_fhflags |= PF_TAG_FLOWADV; + + if (ipoa.ipoa_boundif != IFSCOPE_NONE) + ipoa.ipoa_flags |= IPOAF_BOUND_IF; + + if (laddr.s_addr != INADDR_ANY) + ipoa.ipoa_flags |= IPOAF_BOUND_SRCADDR; + + inp->inp_sndinprog_cnt++; + socket_unlock(so, 0); - /* XXX jgraessley please look at XXX */ error = ip_output_list(m, 0, inpopts, &ro, soopts, mopts, &ipoa); + m = NULL; socket_lock(so, 0); + if (mopts != NULL) + IMO_REMREF(mopts); + + if (error == 0 && nstat_collect) { + locked_add_64(&inp->inp_stat->txpackets, 1); + locked_add_64(&inp->inp_stat->txbytes, len); + } + + if (flowadv && (adv->code == FADV_FLOW_CONTROLLED || + adv->code == FADV_SUSPENDED)) { + /* return a hint to the application that + * the packet has been dropped + */ + error = ENOBUFS; + inp_set_fc_state(inp, adv->code); + } + + VERIFY(inp->inp_sndinprog_cnt > 0); + if ( --inp->inp_sndinprog_cnt == 0) + inp->inp_flags &= ~(INP_FC_FEEDBACK); /* Synchronize PCB cached route */ inp_route_copyin(inp, &ro); +abort: if (udp_dodisconnect) { + /* Always discard the cached route for unconnected socket */ + if (inp->inp_route.ro_rt != NULL) { + rtfree(inp->inp_route.ro_rt); + inp->inp_route.ro_rt = NULL; + } in_pcbdisconnect(inp); inp->inp_laddr = origladdr; /* XXX rehash? */ - } - KERNEL_DEBUG(DBG_FNC_UDP_OUTPUT | DBG_FUNC_END, error, 0,0,0,0); - return (error); + inp->inp_last_outifp = origoutifp; + } else if (inp->inp_route.ro_rt != NULL) { + struct rtentry *rt = inp->inp_route.ro_rt; + struct ifnet *outifp; -abort: - if (udp_dodisconnect) { - in_pcbdisconnect(inp); - inp->inp_laddr = origladdr; /* XXX rehash? */ - } + if (rt->rt_flags & (RTF_MULTICAST|RTF_BROADCAST)) + rt = NULL; /* unusable */ + /* + * Always discard if it is a multicast or broadcast route. + */ + if (rt == NULL) { + rtfree(inp->inp_route.ro_rt); + inp->inp_route.ro_rt = NULL; + } + /* + * If the destination route is unicast, update outifp with + * that of the route interface used by IP. + */ + if (rt != NULL && (outifp = rt->rt_ifp) != inp->inp_last_outifp) + inp->inp_last_outifp = outifp; + } release: - m_freem(m); + if (m != NULL) + m_freem(m); KERNEL_DEBUG(DBG_FNC_UDP_OUTPUT | DBG_FUNC_END, error, 0,0,0,0); return (error); } u_int32_t udp_sendspace = 9216; /* really max datagram size */ -/* 40 1K datagrams */ -u_int32_t udp_recvspace = 40 * (1024 + +/* 187 1K datagrams (approx 192 KB) */ +u_int32_t udp_recvspace = 187 * (1024 + #if INET6 sizeof(struct sockaddr_in6) #else @@ -1321,10 +1673,10 @@ sysctl_udp_sospace(struct sysctl_oid *oidp, __unused void *arg1, return error; } -SYSCTL_PROC(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLTYPE_INT | CTLFLAG_RW, +SYSCTL_PROC(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, &udp_recvspace, 0, &sysctl_udp_sospace, "IU", "Maximum incoming UDP datagram size"); -SYSCTL_PROC(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLTYPE_INT | CTLFLAG_RW, +SYSCTL_PROC(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, &udp_sendspace, 0, &sysctl_udp_sospace, "IU", "Maximum outgoing UDP datagram size"); static int @@ -1359,6 +1711,8 @@ udp_attach(struct socket *so, __unused int proto, struct proc *p) inp = (struct inpcb *)so->so_pcb; inp->inp_vflag |= INP_IPV4; inp->inp_ip_ttl = ip_defttl; + if (nstat_collect) + nstat_udp_new_pcb(inp); return 0; } @@ -1390,9 +1744,12 @@ udp_connect(struct socket *so, struct sockaddr *nam, struct proc *p) return EINVAL; if (inp->inp_faddr.s_addr != INADDR_ANY) return EISCONN; - error = in_pcbconnect(inp, nam, p); - if (error == 0) + error = in_pcbconnect(inp, nam, p, NULL); + if (error == 0) { soisconnected(so); + if (inp->inp_flowhash == 0) + inp->inp_flowhash = inp_calc_flowhash(inp); + } return error; } @@ -1421,8 +1778,13 @@ udp_disconnect(struct socket *so) return ENOTCONN; in_pcbdisconnect(inp); + + /* reset flow controlled state, just in case */ + inp_reset_fc_state(inp); + inp->inp_laddr.s_addr = INADDR_ANY; so->so_state &= ~SS_ISCONNECTED; /* XXX */ + inp->inp_last_outifp = NULL; return 0; } @@ -1473,9 +1835,9 @@ udp_lock(struct socket *so, int refcount, void *debug) lr_saved = debug; if (so->so_pcb) { - lck_mtx_assert(((struct inpcb *)so->so_pcb)->inpcb_mtx, + lck_mtx_assert(&((struct inpcb *)so->so_pcb)->inpcb_mtx, LCK_MTX_ASSERT_NOTOWNED); - lck_mtx_lock(((struct inpcb *)so->so_pcb)->inpcb_mtx); + lck_mtx_lock(&((struct inpcb *)so->so_pcb)->inpcb_mtx); } else { panic("udp_lock: so=%p NO PCB! lr=%p lrh= %s\n", so, lr_saved, solockhistory_nr(so)); @@ -1507,11 +1869,11 @@ udp_unlock(struct socket *so, int refcount, void *debug) so, lr_saved, solockhistory_nr(so)); /* NOTREACHED */ } else { - lck_mtx_assert(((struct inpcb *)so->so_pcb)->inpcb_mtx, + lck_mtx_assert(&((struct inpcb *)so->so_pcb)->inpcb_mtx, LCK_MTX_ASSERT_OWNED); so->unlock_lr[so->next_unlock_lr] = lr_saved; so->next_unlock_lr = (so->next_unlock_lr+1) % SO_LCKDBG_MAX; - lck_mtx_unlock(((struct inpcb *)so->so_pcb)->inpcb_mtx); + lck_mtx_unlock(&((struct inpcb *)so->so_pcb)->inpcb_mtx); } @@ -1525,7 +1887,7 @@ udp_getlock(struct socket *so, __unused int locktype) if (so->so_pcb) - return(inp->inpcb_mtx); + return(&inp->inpcb_mtx); else { panic("udp_getlock: so=%p NULL so_pcb lrh= %s\n", so, solockhistory_nr(so)); @@ -1557,7 +1919,7 @@ udp_slowtimo() continue; so = inp->inp_socket; - if (!lck_mtx_try_lock(inp->inpcb_mtx)) /* skip if busy, no hurry for cleanup... */ + if (!lck_mtx_try_lock(&inp->inpcb_mtx)) /* skip if busy, no hurry for cleanup... */ continue; if (so->so_usecount == 0) { @@ -1571,7 +1933,7 @@ udp_slowtimo() } in_pcbdispose(inp); } else { - lck_mtx_unlock(inp->inpcb_mtx); + lck_mtx_unlock(&inp->inpcb_mtx); } } lck_rw_done(pcbinfo->mtx);