1c398230bSWarner Losh /*- 22469dd60SGarrett Wollman * Copyright (c) 1982, 1986, 1991, 1993, 1995 3497057eeSRobert Watson * The Regents of the University of California. 4111d57a6SRobert Watson * Copyright (c) 2007-2009 Robert N. M. Watson 579bdc6e5SRobert Watson * Copyright (c) 2010-2011 Juniper Networks, Inc. 6497057eeSRobert Watson * All rights reserved. 7df8bae1dSRodney W. Grimes * 879bdc6e5SRobert Watson * Portions of this software were developed by Robert N. M. Watson under 979bdc6e5SRobert Watson * contract to Juniper Networks, Inc. 1079bdc6e5SRobert Watson * 11df8bae1dSRodney W. Grimes * Redistribution and use in source and binary forms, with or without 12df8bae1dSRodney W. Grimes * modification, are permitted provided that the following conditions 13df8bae1dSRodney W. Grimes * are met: 14df8bae1dSRodney W. Grimes * 1. Redistributions of source code must retain the above copyright 15df8bae1dSRodney W. Grimes * notice, this list of conditions and the following disclaimer. 16df8bae1dSRodney W. Grimes * 2. Redistributions in binary form must reproduce the above copyright 17df8bae1dSRodney W. Grimes * notice, this list of conditions and the following disclaimer in the 18df8bae1dSRodney W. Grimes * documentation and/or other materials provided with the distribution. 19df8bae1dSRodney W. Grimes * 4. Neither the name of the University nor the names of its contributors 20df8bae1dSRodney W. Grimes * may be used to endorse or promote products derived from this software 21df8bae1dSRodney W. Grimes * without specific prior written permission. 22df8bae1dSRodney W. Grimes * 23df8bae1dSRodney W. Grimes * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 24df8bae1dSRodney W. Grimes * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 25df8bae1dSRodney W. Grimes * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 26df8bae1dSRodney W. Grimes * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 27df8bae1dSRodney W. Grimes * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 28df8bae1dSRodney W. Grimes * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 29df8bae1dSRodney W. Grimes * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 30df8bae1dSRodney W. Grimes * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 31df8bae1dSRodney W. Grimes * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 32df8bae1dSRodney W. Grimes * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 33df8bae1dSRodney W. Grimes * SUCH DAMAGE. 34df8bae1dSRodney W. Grimes * 352469dd60SGarrett Wollman * @(#)in_pcb.c 8.4 (Berkeley) 5/24/95 36df8bae1dSRodney W. Grimes */ 37df8bae1dSRodney W. Grimes 384b421e2dSMike Silbersack #include <sys/cdefs.h> 394b421e2dSMike Silbersack __FBSDID("$FreeBSD$"); 404b421e2dSMike Silbersack 41497057eeSRobert Watson #include "opt_ddb.h" 426a800098SYoshinobu Inoue #include "opt_ipsec.h" 43efc76f72SBjoern A. Zeeb #include "opt_inet.h" 44cfa1ca9dSYoshinobu Inoue #include "opt_inet6.h" 4552cd27cbSRobert Watson #include "opt_pcbgroup.h" 467527624eSRobert Watson #include "opt_rss.h" 47cfa1ca9dSYoshinobu Inoue 48df8bae1dSRodney W. Grimes #include <sys/param.h> 49df8bae1dSRodney W. Grimes #include <sys/systm.h> 50cc0a3c8cSAndrey V. Elsukov #include <sys/lock.h> 51df8bae1dSRodney W. Grimes #include <sys/malloc.h> 52df8bae1dSRodney W. Grimes #include <sys/mbuf.h> 53aae49dd3SBjoern A. Zeeb #include <sys/callout.h> 54ea8d1492SAlexander V. Chernikov #include <sys/eventhandler.h> 55cfa1ca9dSYoshinobu Inoue #include <sys/domain.h> 56df8bae1dSRodney W. Grimes #include <sys/protosw.h> 57cc0a3c8cSAndrey V. Elsukov #include <sys/rmlock.h> 58df8bae1dSRodney W. Grimes #include <sys/socket.h> 59df8bae1dSRodney W. Grimes #include <sys/socketvar.h> 60acd3428bSRobert Watson #include <sys/priv.h> 61df8bae1dSRodney W. Grimes #include <sys/proc.h> 6279bdc6e5SRobert Watson #include <sys/refcount.h> 6375c13541SPoul-Henning Kamp #include <sys/jail.h> 64101f9fc8SPeter Wemm #include <sys/kernel.h> 65101f9fc8SPeter Wemm #include <sys/sysctl.h> 668781d8e9SBruce Evans 67497057eeSRobert Watson #ifdef DDB 68497057eeSRobert Watson #include <ddb/ddb.h> 69497057eeSRobert Watson #endif 70497057eeSRobert Watson 7169c2d429SJeff Roberson #include <vm/uma.h> 72df8bae1dSRodney W. Grimes 73df8bae1dSRodney W. Grimes #include <net/if.h> 7476039bc8SGleb Smirnoff #include <net/if_var.h> 75cfa1ca9dSYoshinobu Inoue #include <net/if_types.h> 76df8bae1dSRodney W. Grimes #include <net/route.h> 77b2bdc62aSAdrian Chadd #include <net/rss_config.h> 78530c0060SRobert Watson #include <net/vnet.h> 79df8bae1dSRodney W. Grimes 8067107f45SBjoern A. Zeeb #if defined(INET) || defined(INET6) 81df8bae1dSRodney W. Grimes #include <netinet/in.h> 82df8bae1dSRodney W. Grimes #include <netinet/in_pcb.h> 83df8bae1dSRodney W. Grimes #include <netinet/ip_var.h> 84340c35deSJonathan Lemon #include <netinet/tcp_var.h> 855f311da2SMike Silbersack #include <netinet/udp.h> 865f311da2SMike Silbersack #include <netinet/udp_var.h> 8767107f45SBjoern A. Zeeb #endif 8867107f45SBjoern A. Zeeb #ifdef INET 8967107f45SBjoern A. Zeeb #include <netinet/in_var.h> 9067107f45SBjoern A. Zeeb #endif 91cfa1ca9dSYoshinobu Inoue #ifdef INET6 92cfa1ca9dSYoshinobu Inoue #include <netinet/ip6.h> 93efc76f72SBjoern A. Zeeb #include <netinet6/in6_pcb.h> 9467107f45SBjoern A. Zeeb #include <netinet6/in6_var.h> 9567107f45SBjoern A. Zeeb #include <netinet6/ip6_var.h> 96cfa1ca9dSYoshinobu Inoue #endif /* INET6 */ 97cfa1ca9dSYoshinobu Inoue 98df8bae1dSRodney W. Grimes 99b2630c29SGeorge V. Neville-Neil #ifdef IPSEC 100b9234fafSSam Leffler #include <netipsec/ipsec.h> 101b9234fafSSam Leffler #include <netipsec/key.h> 102b2630c29SGeorge V. Neville-Neil #endif /* IPSEC */ 103b9234fafSSam Leffler 104aed55708SRobert Watson #include <security/mac/mac_framework.h> 105aed55708SRobert Watson 106aae49dd3SBjoern A. Zeeb static struct callout ipport_tick_callout; 107aae49dd3SBjoern A. Zeeb 108101f9fc8SPeter Wemm /* 109101f9fc8SPeter Wemm * These configure the range of local port addresses assigned to 110101f9fc8SPeter Wemm * "unspecified" outgoing connections/packets/whatever. 111101f9fc8SPeter Wemm */ 112eddfbb76SRobert Watson VNET_DEFINE(int, ipport_lowfirstauto) = IPPORT_RESERVED - 1; /* 1023 */ 113eddfbb76SRobert Watson VNET_DEFINE(int, ipport_lowlastauto) = IPPORT_RESERVEDSTART; /* 600 */ 114eddfbb76SRobert Watson VNET_DEFINE(int, ipport_firstauto) = IPPORT_EPHEMERALFIRST; /* 10000 */ 115eddfbb76SRobert Watson VNET_DEFINE(int, ipport_lastauto) = IPPORT_EPHEMERALLAST; /* 65535 */ 116eddfbb76SRobert Watson VNET_DEFINE(int, ipport_hifirstauto) = IPPORT_HIFIRSTAUTO; /* 49152 */ 117eddfbb76SRobert Watson VNET_DEFINE(int, ipport_hilastauto) = IPPORT_HILASTAUTO; /* 65535 */ 118101f9fc8SPeter Wemm 119b0d22693SCrist J. Clark /* 120b0d22693SCrist J. Clark * Reserved ports accessible only to root. There are significant 121b0d22693SCrist J. Clark * security considerations that must be accounted for when changing these, 122b0d22693SCrist J. Clark * but the security benefits can be great. Please be careful. 123b0d22693SCrist J. Clark */ 124eddfbb76SRobert Watson VNET_DEFINE(int, ipport_reservedhigh) = IPPORT_RESERVED - 1; /* 1023 */ 125eddfbb76SRobert Watson VNET_DEFINE(int, ipport_reservedlow); 126b0d22693SCrist J. Clark 1275f311da2SMike Silbersack /* Variables dealing with random ephemeral port allocation. */ 128eddfbb76SRobert Watson VNET_DEFINE(int, ipport_randomized) = 1; /* user controlled via sysctl */ 129eddfbb76SRobert Watson VNET_DEFINE(int, ipport_randomcps) = 10; /* user controlled via sysctl */ 130eddfbb76SRobert Watson VNET_DEFINE(int, ipport_randomtime) = 45; /* user controlled via sysctl */ 131eddfbb76SRobert Watson VNET_DEFINE(int, ipport_stoprandom); /* toggled by ipport_tick */ 132eddfbb76SRobert Watson VNET_DEFINE(int, ipport_tcpallocs); 1333e288e62SDimitry Andric static VNET_DEFINE(int, ipport_tcplastcount); 134eddfbb76SRobert Watson 1351e77c105SRobert Watson #define V_ipport_tcplastcount VNET(ipport_tcplastcount) 1366ac48b74SMike Silbersack 137d2025bd0SBjoern A. Zeeb static void in_pcbremlists(struct inpcb *inp); 138d2025bd0SBjoern A. Zeeb #ifdef INET 139fa046d87SRobert Watson static struct inpcb *in_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, 140fa046d87SRobert Watson struct in_addr faddr, u_int fport_arg, 141fa046d87SRobert Watson struct in_addr laddr, u_int lport_arg, 142fa046d87SRobert Watson int lookupflags, struct ifnet *ifp); 14367107f45SBjoern A. Zeeb 144bbd42ad0SPeter Wemm #define RANGECHK(var, min, max) \ 145bbd42ad0SPeter Wemm if ((var) < (min)) { (var) = (min); } \ 146bbd42ad0SPeter Wemm else if ((var) > (max)) { (var) = (max); } 147bbd42ad0SPeter Wemm 148bbd42ad0SPeter Wemm static int 14982d9ae4eSPoul-Henning Kamp sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS) 150bbd42ad0SPeter Wemm { 15130a4ab08SBruce Evans int error; 15230a4ab08SBruce Evans 153f6dfe47aSMarko Zec error = sysctl_handle_int(oidp, arg1, arg2, req); 15430a4ab08SBruce Evans if (error == 0) { 155603724d3SBjoern A. Zeeb RANGECHK(V_ipport_lowfirstauto, 1, IPPORT_RESERVED - 1); 156603724d3SBjoern A. Zeeb RANGECHK(V_ipport_lowlastauto, 1, IPPORT_RESERVED - 1); 157603724d3SBjoern A. Zeeb RANGECHK(V_ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX); 158603724d3SBjoern A. Zeeb RANGECHK(V_ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX); 159603724d3SBjoern A. Zeeb RANGECHK(V_ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX); 160603724d3SBjoern A. Zeeb RANGECHK(V_ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX); 161bbd42ad0SPeter Wemm } 16230a4ab08SBruce Evans return (error); 163bbd42ad0SPeter Wemm } 164bbd42ad0SPeter Wemm 165bbd42ad0SPeter Wemm #undef RANGECHK 166bbd42ad0SPeter Wemm 1676472ac3dSEd Schouten static SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, 1686472ac3dSEd Schouten "IP Ports"); 16933b3ac06SPeter Wemm 1706df8a710SGleb Smirnoff SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, 1716df8a710SGleb Smirnoff CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, 1726df8a710SGleb Smirnoff &VNET_NAME(ipport_lowfirstauto), 0, &sysctl_net_ipport_check, "I", ""); 1736df8a710SGleb Smirnoff SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, 1746df8a710SGleb Smirnoff CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, 1756df8a710SGleb Smirnoff &VNET_NAME(ipport_lowlastauto), 0, &sysctl_net_ipport_check, "I", ""); 1766df8a710SGleb Smirnoff SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, 1776df8a710SGleb Smirnoff CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, 1786df8a710SGleb Smirnoff &VNET_NAME(ipport_firstauto), 0, &sysctl_net_ipport_check, "I", ""); 1796df8a710SGleb Smirnoff SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, 1806df8a710SGleb Smirnoff CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, 1816df8a710SGleb Smirnoff &VNET_NAME(ipport_lastauto), 0, &sysctl_net_ipport_check, "I", ""); 1826df8a710SGleb Smirnoff SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, 1836df8a710SGleb Smirnoff CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, 1846df8a710SGleb Smirnoff &VNET_NAME(ipport_hifirstauto), 0, &sysctl_net_ipport_check, "I", ""); 1856df8a710SGleb Smirnoff SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, 1866df8a710SGleb Smirnoff CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, 1876df8a710SGleb Smirnoff &VNET_NAME(ipport_hilastauto), 0, &sysctl_net_ipport_check, "I", ""); 1886df8a710SGleb Smirnoff SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh, 1896df8a710SGleb Smirnoff CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE, 1906df8a710SGleb Smirnoff &VNET_NAME(ipport_reservedhigh), 0, ""); 1916df8a710SGleb Smirnoff SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow, 192eddfbb76SRobert Watson CTLFLAG_RW|CTLFLAG_SECURE, &VNET_NAME(ipport_reservedlow), 0, ""); 1936df8a710SGleb Smirnoff SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomized, 1946df8a710SGleb Smirnoff CTLFLAG_VNET | CTLFLAG_RW, 195eddfbb76SRobert Watson &VNET_NAME(ipport_randomized), 0, "Enable random port allocation"); 1966df8a710SGleb Smirnoff SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomcps, 1976df8a710SGleb Smirnoff CTLFLAG_VNET | CTLFLAG_RW, 198eddfbb76SRobert Watson &VNET_NAME(ipport_randomcps), 0, "Maximum number of random port " 1996ee79c59SMaxim Konovalov "allocations before switching to a sequental one"); 2006df8a710SGleb Smirnoff SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomtime, 2016df8a710SGleb Smirnoff CTLFLAG_VNET | CTLFLAG_RW, 202eddfbb76SRobert Watson &VNET_NAME(ipport_randomtime), 0, 2038b615593SMarko Zec "Minimum time to keep sequental port " 2046ee79c59SMaxim Konovalov "allocation before switching to a random one"); 20583e521ecSBjoern A. Zeeb #endif /* INET */ 2060312fbe9SPoul-Henning Kamp 207c3229e05SDavid Greenman /* 208c3229e05SDavid Greenman * in_pcb.c: manage the Protocol Control Blocks. 209c3229e05SDavid Greenman * 210de35559fSRobert Watson * NOTE: It is assumed that most of these functions will be called with 211de35559fSRobert Watson * the pcbinfo lock held, and often, the inpcb lock held, as these utility 212de35559fSRobert Watson * functions often modify hash chains or addresses in pcbs. 213c3229e05SDavid Greenman */ 214c3229e05SDavid Greenman 215c3229e05SDavid Greenman /* 2169bcd427bSRobert Watson * Initialize an inpcbinfo -- we should be able to reduce the number of 2179bcd427bSRobert Watson * arguments in time. 2189bcd427bSRobert Watson */ 2199bcd427bSRobert Watson void 2209bcd427bSRobert Watson in_pcbinfo_init(struct inpcbinfo *pcbinfo, const char *name, 2219bcd427bSRobert Watson struct inpcbhead *listhead, int hash_nelements, int porthash_nelements, 2229bcd427bSRobert Watson char *inpcbzone_name, uma_init inpcbzone_init, uma_fini inpcbzone_fini, 22352cd27cbSRobert Watson uint32_t inpcbzone_flags, u_int hashfields) 2249bcd427bSRobert Watson { 2259bcd427bSRobert Watson 2269bcd427bSRobert Watson INP_INFO_LOCK_INIT(pcbinfo, name); 227fa046d87SRobert Watson INP_HASH_LOCK_INIT(pcbinfo, "pcbinfohash"); /* XXXRW: argument? */ 228ff9b006dSJulien Charbon INP_LIST_LOCK_INIT(pcbinfo, "pcbinfolist"); 2299bcd427bSRobert Watson #ifdef VIMAGE 2309bcd427bSRobert Watson pcbinfo->ipi_vnet = curvnet; 2319bcd427bSRobert Watson #endif 2329bcd427bSRobert Watson pcbinfo->ipi_listhead = listhead; 2339bcd427bSRobert Watson LIST_INIT(pcbinfo->ipi_listhead); 234fa046d87SRobert Watson pcbinfo->ipi_count = 0; 2359bcd427bSRobert Watson pcbinfo->ipi_hashbase = hashinit(hash_nelements, M_PCB, 2369bcd427bSRobert Watson &pcbinfo->ipi_hashmask); 2379bcd427bSRobert Watson pcbinfo->ipi_porthashbase = hashinit(porthash_nelements, M_PCB, 2389bcd427bSRobert Watson &pcbinfo->ipi_porthashmask); 23952cd27cbSRobert Watson #ifdef PCBGROUP 24052cd27cbSRobert Watson in_pcbgroup_init(pcbinfo, hashfields, hash_nelements); 24152cd27cbSRobert Watson #endif 2429bcd427bSRobert Watson pcbinfo->ipi_zone = uma_zcreate(inpcbzone_name, sizeof(struct inpcb), 2439bcd427bSRobert Watson NULL, NULL, inpcbzone_init, inpcbzone_fini, UMA_ALIGN_PTR, 2449bcd427bSRobert Watson inpcbzone_flags); 2459bcd427bSRobert Watson uma_zone_set_max(pcbinfo->ipi_zone, maxsockets); 2466acd596eSPawel Jakub Dawidek uma_zone_set_warning(pcbinfo->ipi_zone, 2476acd596eSPawel Jakub Dawidek "kern.ipc.maxsockets limit reached"); 2489bcd427bSRobert Watson } 2499bcd427bSRobert Watson 2509bcd427bSRobert Watson /* 2519bcd427bSRobert Watson * Destroy an inpcbinfo. 2529bcd427bSRobert Watson */ 2539bcd427bSRobert Watson void 2549bcd427bSRobert Watson in_pcbinfo_destroy(struct inpcbinfo *pcbinfo) 2559bcd427bSRobert Watson { 2569bcd427bSRobert Watson 257fa046d87SRobert Watson KASSERT(pcbinfo->ipi_count == 0, 258fa046d87SRobert Watson ("%s: ipi_count = %u", __func__, pcbinfo->ipi_count)); 259fa046d87SRobert Watson 2609bcd427bSRobert Watson hashdestroy(pcbinfo->ipi_hashbase, M_PCB, pcbinfo->ipi_hashmask); 2619bcd427bSRobert Watson hashdestroy(pcbinfo->ipi_porthashbase, M_PCB, 2629bcd427bSRobert Watson pcbinfo->ipi_porthashmask); 26352cd27cbSRobert Watson #ifdef PCBGROUP 26452cd27cbSRobert Watson in_pcbgroup_destroy(pcbinfo); 26552cd27cbSRobert Watson #endif 2669bcd427bSRobert Watson uma_zdestroy(pcbinfo->ipi_zone); 267ff9b006dSJulien Charbon INP_LIST_LOCK_DESTROY(pcbinfo); 268fa046d87SRobert Watson INP_HASH_LOCK_DESTROY(pcbinfo); 2699bcd427bSRobert Watson INP_INFO_LOCK_DESTROY(pcbinfo); 2709bcd427bSRobert Watson } 2719bcd427bSRobert Watson 2729bcd427bSRobert Watson /* 273c3229e05SDavid Greenman * Allocate a PCB and associate it with the socket. 274d915b280SStephan Uphoff * On success return with the PCB locked. 275c3229e05SDavid Greenman */ 276df8bae1dSRodney W. Grimes int 277d915b280SStephan Uphoff in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo) 278df8bae1dSRodney W. Grimes { 279136d4f1cSRobert Watson struct inpcb *inp; 28013cf67f3SHajimu UMEMOTO int error; 281a557af22SRobert Watson 282ff9b006dSJulien Charbon #ifdef INVARIANTS 283ff9b006dSJulien Charbon if (pcbinfo == &V_tcbinfo) { 284ff9b006dSJulien Charbon INP_INFO_RLOCK_ASSERT(pcbinfo); 285ff9b006dSJulien Charbon } else { 28659daba27SSam Leffler INP_INFO_WLOCK_ASSERT(pcbinfo); 287ff9b006dSJulien Charbon } 288ff9b006dSJulien Charbon #endif 289ff9b006dSJulien Charbon 290a557af22SRobert Watson error = 0; 291d915b280SStephan Uphoff inp = uma_zalloc(pcbinfo->ipi_zone, M_NOWAIT); 292df8bae1dSRodney W. Grimes if (inp == NULL) 293df8bae1dSRodney W. Grimes return (ENOBUFS); 294d915b280SStephan Uphoff bzero(inp, inp_zero_size); 29515bd2b43SDavid Greenman inp->inp_pcbinfo = pcbinfo; 296df8bae1dSRodney W. Grimes inp->inp_socket = so; 29786d02c5cSBjoern A. Zeeb inp->inp_cred = crhold(so->so_cred); 2988b07e49aSJulian Elischer inp->inp_inc.inc_fibnum = so->so_fibnum; 299a557af22SRobert Watson #ifdef MAC 30030d239bcSRobert Watson error = mac_inpcb_init(inp, M_NOWAIT); 301a557af22SRobert Watson if (error != 0) 302a557af22SRobert Watson goto out; 30330d239bcSRobert Watson mac_inpcb_create(so, inp); 304a557af22SRobert Watson #endif 305b2630c29SGeorge V. Neville-Neil #ifdef IPSEC 30613cf67f3SHajimu UMEMOTO error = ipsec_init_policy(so, &inp->inp_sp); 3070bffde27SRobert Watson if (error != 0) { 3080bffde27SRobert Watson #ifdef MAC 3090bffde27SRobert Watson mac_inpcb_destroy(inp); 3100bffde27SRobert Watson #endif 311a557af22SRobert Watson goto out; 3120bffde27SRobert Watson } 313b2630c29SGeorge V. Neville-Neil #endif /*IPSEC*/ 314e3fd5ffdSRobert Watson #ifdef INET6 315340c35deSJonathan Lemon if (INP_SOCKAF(so) == AF_INET6) { 316340c35deSJonathan Lemon inp->inp_vflag |= INP_IPV6PROTO; 317603724d3SBjoern A. Zeeb if (V_ip6_v6only) 31833841545SHajimu UMEMOTO inp->inp_flags |= IN6P_IPV6_V6ONLY; 319340c35deSJonathan Lemon } 32075daea93SPaul Saab #endif 321ff9b006dSJulien Charbon INP_WLOCK(inp); 322ff9b006dSJulien Charbon INP_LIST_WLOCK(pcbinfo); 323712fc218SRobert Watson LIST_INSERT_HEAD(pcbinfo->ipi_listhead, inp, inp_list); 3243d4d47f3SGarrett Wollman pcbinfo->ipi_count++; 325df8bae1dSRodney W. Grimes so->so_pcb = (caddr_t)inp; 32633841545SHajimu UMEMOTO #ifdef INET6 327603724d3SBjoern A. Zeeb if (V_ip6_auto_flowlabel) 32833841545SHajimu UMEMOTO inp->inp_flags |= IN6P_AUTOFLOWLABEL; 32933841545SHajimu UMEMOTO #endif 330d915b280SStephan Uphoff inp->inp_gencnt = ++pcbinfo->ipi_gencnt; 33179bdc6e5SRobert Watson refcount_init(&inp->inp_refcount, 1); /* Reference from inpcbinfo */ 332ff9b006dSJulien Charbon INP_LIST_WUNLOCK(pcbinfo); 333b2630c29SGeorge V. Neville-Neil #if defined(IPSEC) || defined(MAC) 334a557af22SRobert Watson out: 33586d02c5cSBjoern A. Zeeb if (error != 0) { 33686d02c5cSBjoern A. Zeeb crfree(inp->inp_cred); 337a557af22SRobert Watson uma_zfree(pcbinfo->ipi_zone, inp); 33886d02c5cSBjoern A. Zeeb } 339a557af22SRobert Watson #endif 340a557af22SRobert Watson return (error); 341df8bae1dSRodney W. Grimes } 342df8bae1dSRodney W. Grimes 34367107f45SBjoern A. Zeeb #ifdef INET 344df8bae1dSRodney W. Grimes int 345136d4f1cSRobert Watson in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred) 346df8bae1dSRodney W. Grimes { 3474b932371SIan Dowse int anonport, error; 3484b932371SIan Dowse 3498501a69cSRobert Watson INP_WLOCK_ASSERT(inp); 350fa046d87SRobert Watson INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); 35159daba27SSam Leffler 3524b932371SIan Dowse if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY) 3534b932371SIan Dowse return (EINVAL); 3545cd3dcaaSNavdeep Parhar anonport = nam == NULL || ((struct sockaddr_in *)nam)->sin_port == 0; 3554b932371SIan Dowse error = in_pcbbind_setup(inp, nam, &inp->inp_laddr.s_addr, 356b0330ed9SPawel Jakub Dawidek &inp->inp_lport, cred); 3574b932371SIan Dowse if (error) 3584b932371SIan Dowse return (error); 3594b932371SIan Dowse if (in_pcbinshash(inp) != 0) { 3604b932371SIan Dowse inp->inp_laddr.s_addr = INADDR_ANY; 3614b932371SIan Dowse inp->inp_lport = 0; 3624b932371SIan Dowse return (EAGAIN); 3634b932371SIan Dowse } 3644b932371SIan Dowse if (anonport) 3654b932371SIan Dowse inp->inp_flags |= INP_ANONPORT; 3664b932371SIan Dowse return (0); 3674b932371SIan Dowse } 36867107f45SBjoern A. Zeeb #endif 3694b932371SIan Dowse 37039c8c62eSHiren Panchasara /* 37139c8c62eSHiren Panchasara * Select a local port (number) to use. 37239c8c62eSHiren Panchasara */ 373efc76f72SBjoern A. Zeeb #if defined(INET) || defined(INET6) 374efc76f72SBjoern A. Zeeb int 375efc76f72SBjoern A. Zeeb in_pcb_lport(struct inpcb *inp, struct in_addr *laddrp, u_short *lportp, 37668e0d7e0SRobert Watson struct ucred *cred, int lookupflags) 377efc76f72SBjoern A. Zeeb { 378efc76f72SBjoern A. Zeeb struct inpcbinfo *pcbinfo; 379efc76f72SBjoern A. Zeeb struct inpcb *tmpinp; 380efc76f72SBjoern A. Zeeb unsigned short *lastport; 381efc76f72SBjoern A. Zeeb int count, dorandom, error; 382efc76f72SBjoern A. Zeeb u_short aux, first, last, lport; 383efc76f72SBjoern A. Zeeb #ifdef INET 384efc76f72SBjoern A. Zeeb struct in_addr laddr; 385efc76f72SBjoern A. Zeeb #endif 386efc76f72SBjoern A. Zeeb 387efc76f72SBjoern A. Zeeb pcbinfo = inp->inp_pcbinfo; 388efc76f72SBjoern A. Zeeb 389efc76f72SBjoern A. Zeeb /* 390efc76f72SBjoern A. Zeeb * Because no actual state changes occur here, a global write lock on 391efc76f72SBjoern A. Zeeb * the pcbinfo isn't required. 392efc76f72SBjoern A. Zeeb */ 393efc76f72SBjoern A. Zeeb INP_LOCK_ASSERT(inp); 394fa046d87SRobert Watson INP_HASH_LOCK_ASSERT(pcbinfo); 395efc76f72SBjoern A. Zeeb 396efc76f72SBjoern A. Zeeb if (inp->inp_flags & INP_HIGHPORT) { 397efc76f72SBjoern A. Zeeb first = V_ipport_hifirstauto; /* sysctl */ 398efc76f72SBjoern A. Zeeb last = V_ipport_hilastauto; 399efc76f72SBjoern A. Zeeb lastport = &pcbinfo->ipi_lasthi; 400efc76f72SBjoern A. Zeeb } else if (inp->inp_flags & INP_LOWPORT) { 401efc76f72SBjoern A. Zeeb error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0); 402efc76f72SBjoern A. Zeeb if (error) 403efc76f72SBjoern A. Zeeb return (error); 404efc76f72SBjoern A. Zeeb first = V_ipport_lowfirstauto; /* 1023 */ 405efc76f72SBjoern A. Zeeb last = V_ipport_lowlastauto; /* 600 */ 406efc76f72SBjoern A. Zeeb lastport = &pcbinfo->ipi_lastlow; 407efc76f72SBjoern A. Zeeb } else { 408efc76f72SBjoern A. Zeeb first = V_ipport_firstauto; /* sysctl */ 409efc76f72SBjoern A. Zeeb last = V_ipport_lastauto; 410efc76f72SBjoern A. Zeeb lastport = &pcbinfo->ipi_lastport; 411efc76f72SBjoern A. Zeeb } 412efc76f72SBjoern A. Zeeb /* 413e06e816fSKevin Lo * For UDP(-Lite), use random port allocation as long as the user 414efc76f72SBjoern A. Zeeb * allows it. For TCP (and as of yet unknown) connections, 415efc76f72SBjoern A. Zeeb * use random port allocation only if the user allows it AND 416efc76f72SBjoern A. Zeeb * ipport_tick() allows it. 417efc76f72SBjoern A. Zeeb */ 418efc76f72SBjoern A. Zeeb if (V_ipport_randomized && 419e06e816fSKevin Lo (!V_ipport_stoprandom || pcbinfo == &V_udbinfo || 420e06e816fSKevin Lo pcbinfo == &V_ulitecbinfo)) 421efc76f72SBjoern A. Zeeb dorandom = 1; 422efc76f72SBjoern A. Zeeb else 423efc76f72SBjoern A. Zeeb dorandom = 0; 424efc76f72SBjoern A. Zeeb /* 425efc76f72SBjoern A. Zeeb * It makes no sense to do random port allocation if 426efc76f72SBjoern A. Zeeb * we have the only port available. 427efc76f72SBjoern A. Zeeb */ 428efc76f72SBjoern A. Zeeb if (first == last) 429efc76f72SBjoern A. Zeeb dorandom = 0; 430e06e816fSKevin Lo /* Make sure to not include UDP(-Lite) packets in the count. */ 431e06e816fSKevin Lo if (pcbinfo != &V_udbinfo || pcbinfo != &V_ulitecbinfo) 432efc76f72SBjoern A. Zeeb V_ipport_tcpallocs++; 433efc76f72SBjoern A. Zeeb /* 434efc76f72SBjoern A. Zeeb * Instead of having two loops further down counting up or down 435efc76f72SBjoern A. Zeeb * make sure that first is always <= last and go with only one 436efc76f72SBjoern A. Zeeb * code path implementing all logic. 437efc76f72SBjoern A. Zeeb */ 438efc76f72SBjoern A. Zeeb if (first > last) { 439efc76f72SBjoern A. Zeeb aux = first; 440efc76f72SBjoern A. Zeeb first = last; 441efc76f72SBjoern A. Zeeb last = aux; 442efc76f72SBjoern A. Zeeb } 443efc76f72SBjoern A. Zeeb 444efc76f72SBjoern A. Zeeb #ifdef INET 445efc76f72SBjoern A. Zeeb /* Make the compiler happy. */ 446efc76f72SBjoern A. Zeeb laddr.s_addr = 0; 4474d457387SBjoern A. Zeeb if ((inp->inp_vflag & (INP_IPV4|INP_IPV6)) == INP_IPV4) { 448efc76f72SBjoern A. Zeeb KASSERT(laddrp != NULL, ("%s: laddrp NULL for v4 inp %p", 449efc76f72SBjoern A. Zeeb __func__, inp)); 450efc76f72SBjoern A. Zeeb laddr = *laddrp; 451efc76f72SBjoern A. Zeeb } 452efc76f72SBjoern A. Zeeb #endif 45367107f45SBjoern A. Zeeb tmpinp = NULL; /* Make compiler happy. */ 454efc76f72SBjoern A. Zeeb lport = *lportp; 455efc76f72SBjoern A. Zeeb 456efc76f72SBjoern A. Zeeb if (dorandom) 457efc76f72SBjoern A. Zeeb *lastport = first + (arc4random() % (last - first)); 458efc76f72SBjoern A. Zeeb 459efc76f72SBjoern A. Zeeb count = last - first; 460efc76f72SBjoern A. Zeeb 461efc76f72SBjoern A. Zeeb do { 462efc76f72SBjoern A. Zeeb if (count-- < 0) /* completely used? */ 463efc76f72SBjoern A. Zeeb return (EADDRNOTAVAIL); 464efc76f72SBjoern A. Zeeb ++*lastport; 465efc76f72SBjoern A. Zeeb if (*lastport < first || *lastport > last) 466efc76f72SBjoern A. Zeeb *lastport = first; 467efc76f72SBjoern A. Zeeb lport = htons(*lastport); 468efc76f72SBjoern A. Zeeb 469efc76f72SBjoern A. Zeeb #ifdef INET6 470efc76f72SBjoern A. Zeeb if ((inp->inp_vflag & INP_IPV6) != 0) 471efc76f72SBjoern A. Zeeb tmpinp = in6_pcblookup_local(pcbinfo, 47268e0d7e0SRobert Watson &inp->in6p_laddr, lport, lookupflags, cred); 473efc76f72SBjoern A. Zeeb #endif 474efc76f72SBjoern A. Zeeb #if defined(INET) && defined(INET6) 475efc76f72SBjoern A. Zeeb else 476efc76f72SBjoern A. Zeeb #endif 477efc76f72SBjoern A. Zeeb #ifdef INET 478efc76f72SBjoern A. Zeeb tmpinp = in_pcblookup_local(pcbinfo, laddr, 47968e0d7e0SRobert Watson lport, lookupflags, cred); 480efc76f72SBjoern A. Zeeb #endif 481efc76f72SBjoern A. Zeeb } while (tmpinp != NULL); 482efc76f72SBjoern A. Zeeb 483efc76f72SBjoern A. Zeeb #ifdef INET 4844d457387SBjoern A. Zeeb if ((inp->inp_vflag & (INP_IPV4|INP_IPV6)) == INP_IPV4) 485efc76f72SBjoern A. Zeeb laddrp->s_addr = laddr.s_addr; 486efc76f72SBjoern A. Zeeb #endif 487efc76f72SBjoern A. Zeeb *lportp = lport; 488efc76f72SBjoern A. Zeeb 489efc76f72SBjoern A. Zeeb return (0); 490efc76f72SBjoern A. Zeeb } 491efdf104bSMikolaj Golub 492efdf104bSMikolaj Golub /* 493efdf104bSMikolaj Golub * Return cached socket options. 494efdf104bSMikolaj Golub */ 495efdf104bSMikolaj Golub short 496efdf104bSMikolaj Golub inp_so_options(const struct inpcb *inp) 497efdf104bSMikolaj Golub { 498efdf104bSMikolaj Golub short so_options; 499efdf104bSMikolaj Golub 500efdf104bSMikolaj Golub so_options = 0; 501efdf104bSMikolaj Golub 502efdf104bSMikolaj Golub if ((inp->inp_flags2 & INP_REUSEPORT) != 0) 503efdf104bSMikolaj Golub so_options |= SO_REUSEPORT; 504efdf104bSMikolaj Golub if ((inp->inp_flags2 & INP_REUSEADDR) != 0) 505efdf104bSMikolaj Golub so_options |= SO_REUSEADDR; 506efdf104bSMikolaj Golub return (so_options); 507efdf104bSMikolaj Golub } 508efc76f72SBjoern A. Zeeb #endif /* INET || INET6 */ 509efc76f72SBjoern A. Zeeb 5104b932371SIan Dowse /* 5110a100a6fSAdrian Chadd * Check if a new BINDMULTI socket is allowed to be created. 5120a100a6fSAdrian Chadd * 5130a100a6fSAdrian Chadd * ni points to the new inp. 5140a100a6fSAdrian Chadd * oi points to the exisitng inp. 5150a100a6fSAdrian Chadd * 5160a100a6fSAdrian Chadd * This checks whether the existing inp also has BINDMULTI and 5170a100a6fSAdrian Chadd * whether the credentials match. 5180a100a6fSAdrian Chadd */ 519d5bb8bd3SAdrian Chadd int 5200a100a6fSAdrian Chadd in_pcbbind_check_bindmulti(const struct inpcb *ni, const struct inpcb *oi) 5210a100a6fSAdrian Chadd { 5220a100a6fSAdrian Chadd /* Check permissions match */ 5230a100a6fSAdrian Chadd if ((ni->inp_flags2 & INP_BINDMULTI) && 5240a100a6fSAdrian Chadd (ni->inp_cred->cr_uid != 5250a100a6fSAdrian Chadd oi->inp_cred->cr_uid)) 5260a100a6fSAdrian Chadd return (0); 5270a100a6fSAdrian Chadd 5280a100a6fSAdrian Chadd /* Check the existing inp has BINDMULTI set */ 5290a100a6fSAdrian Chadd if ((ni->inp_flags2 & INP_BINDMULTI) && 5300a100a6fSAdrian Chadd ((oi->inp_flags2 & INP_BINDMULTI) == 0)) 5310a100a6fSAdrian Chadd return (0); 5320a100a6fSAdrian Chadd 5330a100a6fSAdrian Chadd /* 5340a100a6fSAdrian Chadd * We're okay - either INP_BINDMULTI isn't set on ni, or 5350a100a6fSAdrian Chadd * it is and it matches the checks. 5360a100a6fSAdrian Chadd */ 5370a100a6fSAdrian Chadd return (1); 5380a100a6fSAdrian Chadd } 5390a100a6fSAdrian Chadd 540d5bb8bd3SAdrian Chadd #ifdef INET 5410a100a6fSAdrian Chadd /* 5424b932371SIan Dowse * Set up a bind operation on a PCB, performing port allocation 5434b932371SIan Dowse * as required, but do not actually modify the PCB. Callers can 5444b932371SIan Dowse * either complete the bind by setting inp_laddr/inp_lport and 5454b932371SIan Dowse * calling in_pcbinshash(), or they can just use the resulting 5464b932371SIan Dowse * port and address to authorise the sending of a once-off packet. 5474b932371SIan Dowse * 5484b932371SIan Dowse * On error, the values of *laddrp and *lportp are not changed. 5494b932371SIan Dowse */ 5504b932371SIan Dowse int 551136d4f1cSRobert Watson in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp, 552136d4f1cSRobert Watson u_short *lportp, struct ucred *cred) 5534b932371SIan Dowse { 5544b932371SIan Dowse struct socket *so = inp->inp_socket; 55515bd2b43SDavid Greenman struct sockaddr_in *sin; 556c3229e05SDavid Greenman struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 5574b932371SIan Dowse struct in_addr laddr; 558df8bae1dSRodney W. Grimes u_short lport = 0; 55968e0d7e0SRobert Watson int lookupflags = 0, reuseport = (so->so_options & SO_REUSEPORT); 560413628a7SBjoern A. Zeeb int error; 561df8bae1dSRodney W. Grimes 5628501a69cSRobert Watson /* 563fa046d87SRobert Watson * No state changes, so read locks are sufficient here. 5648501a69cSRobert Watson */ 56559daba27SSam Leffler INP_LOCK_ASSERT(inp); 566fa046d87SRobert Watson INP_HASH_LOCK_ASSERT(pcbinfo); 56759daba27SSam Leffler 568603724d3SBjoern A. Zeeb if (TAILQ_EMPTY(&V_in_ifaddrhead)) /* XXX broken! */ 569df8bae1dSRodney W. Grimes return (EADDRNOTAVAIL); 5704b932371SIan Dowse laddr.s_addr = *laddrp; 5714b932371SIan Dowse if (nam != NULL && laddr.s_addr != INADDR_ANY) 572df8bae1dSRodney W. Grimes return (EINVAL); 573c3229e05SDavid Greenman if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0) 57468e0d7e0SRobert Watson lookupflags = INPLOOKUP_WILDCARD; 5757c2f3cb9SJamie Gritton if (nam == NULL) { 5767c2f3cb9SJamie Gritton if ((error = prison_local_ip4(cred, &laddr)) != 0) 5777c2f3cb9SJamie Gritton return (error); 5787c2f3cb9SJamie Gritton } else { 57957bf258eSGarrett Wollman sin = (struct sockaddr_in *)nam; 58057bf258eSGarrett Wollman if (nam->sa_len != sizeof (*sin)) 581df8bae1dSRodney W. Grimes return (EINVAL); 582df8bae1dSRodney W. Grimes #ifdef notdef 583df8bae1dSRodney W. Grimes /* 584df8bae1dSRodney W. Grimes * We should check the family, but old programs 585df8bae1dSRodney W. Grimes * incorrectly fail to initialize it. 586df8bae1dSRodney W. Grimes */ 587df8bae1dSRodney W. Grimes if (sin->sin_family != AF_INET) 588df8bae1dSRodney W. Grimes return (EAFNOSUPPORT); 589df8bae1dSRodney W. Grimes #endif 590b89e82ddSJamie Gritton error = prison_local_ip4(cred, &sin->sin_addr); 591b89e82ddSJamie Gritton if (error) 592b89e82ddSJamie Gritton return (error); 5934b932371SIan Dowse if (sin->sin_port != *lportp) { 5944b932371SIan Dowse /* Don't allow the port to change. */ 5954b932371SIan Dowse if (*lportp != 0) 5964b932371SIan Dowse return (EINVAL); 597df8bae1dSRodney W. Grimes lport = sin->sin_port; 5984b932371SIan Dowse } 5994b932371SIan Dowse /* NB: lport is left as 0 if the port isn't being changed. */ 600df8bae1dSRodney W. Grimes if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) { 601df8bae1dSRodney W. Grimes /* 602df8bae1dSRodney W. Grimes * Treat SO_REUSEADDR as SO_REUSEPORT for multicast; 603df8bae1dSRodney W. Grimes * allow complete duplication of binding if 604df8bae1dSRodney W. Grimes * SO_REUSEPORT is set, or if SO_REUSEADDR is set 605df8bae1dSRodney W. Grimes * and a multicast address is bound on both 606df8bae1dSRodney W. Grimes * new and duplicated sockets. 607df8bae1dSRodney W. Grimes */ 608f122b319SMikolaj Golub if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) != 0) 609df8bae1dSRodney W. Grimes reuseport = SO_REUSEADDR|SO_REUSEPORT; 610df8bae1dSRodney W. Grimes } else if (sin->sin_addr.s_addr != INADDR_ANY) { 611df8bae1dSRodney W. Grimes sin->sin_port = 0; /* yech... */ 61283103a73SAndrew R. Reiter bzero(&sin->sin_zero, sizeof(sin->sin_zero)); 6134209e01aSAdrian Chadd /* 6144209e01aSAdrian Chadd * Is the address a local IP address? 615f44270e7SPawel Jakub Dawidek * If INP_BINDANY is set, then the socket may be bound 6168696873dSAdrian Chadd * to any endpoint address, local or not. 6174209e01aSAdrian Chadd */ 618f44270e7SPawel Jakub Dawidek if ((inp->inp_flags & INP_BINDANY) == 0 && 6198896f83aSRobert Watson ifa_ifwithaddr_check((struct sockaddr *)sin) == 0) 620df8bae1dSRodney W. Grimes return (EADDRNOTAVAIL); 621df8bae1dSRodney W. Grimes } 6224b932371SIan Dowse laddr = sin->sin_addr; 623df8bae1dSRodney W. Grimes if (lport) { 624df8bae1dSRodney W. Grimes struct inpcb *t; 625ae0e7143SRobert Watson struct tcptw *tw; 626ae0e7143SRobert Watson 627df8bae1dSRodney W. Grimes /* GROSS */ 628603724d3SBjoern A. Zeeb if (ntohs(lport) <= V_ipport_reservedhigh && 629603724d3SBjoern A. Zeeb ntohs(lport) >= V_ipport_reservedlow && 630acd3428bSRobert Watson priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 63132f9753cSRobert Watson 0)) 6322469dd60SGarrett Wollman return (EACCES); 633835d4b89SPawel Jakub Dawidek if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) && 63486d02c5cSBjoern A. Zeeb priv_check_cred(inp->inp_cred, 63532f9753cSRobert Watson PRIV_NETINET_REUSEPORT, 0) != 0) { 636078b7042SBjoern A. Zeeb t = in_pcblookup_local(pcbinfo, sin->sin_addr, 637413628a7SBjoern A. Zeeb lport, INPLOOKUP_WILDCARD, cred); 638340c35deSJonathan Lemon /* 639340c35deSJonathan Lemon * XXX 640340c35deSJonathan Lemon * This entire block sorely needs a rewrite. 641340c35deSJonathan Lemon */ 6424cc20ab1SSeigo Tanimura if (t && 6430a100a6fSAdrian Chadd ((inp->inp_flags2 & INP_BINDMULTI) == 0) && 644ad71fe3cSRobert Watson ((t->inp_flags & INP_TIMEWAIT) == 0) && 6454658dc83SYaroslav Tykhiy (so->so_type != SOCK_STREAM || 6464658dc83SYaroslav Tykhiy ntohl(t->inp_faddr.s_addr) == INADDR_ANY) && 6474cc20ab1SSeigo Tanimura (ntohl(sin->sin_addr.s_addr) != INADDR_ANY || 64852b65dbeSBill Fenner ntohl(t->inp_laddr.s_addr) != INADDR_ANY || 649fc06cd42SMikolaj Golub (t->inp_flags2 & INP_REUSEPORT) == 0) && 65086d02c5cSBjoern A. Zeeb (inp->inp_cred->cr_uid != 65186d02c5cSBjoern A. Zeeb t->inp_cred->cr_uid)) 6524049a042SGuido van Rooij return (EADDRINUSE); 6530a100a6fSAdrian Chadd 6540a100a6fSAdrian Chadd /* 6550a100a6fSAdrian Chadd * If the socket is a BINDMULTI socket, then 6560a100a6fSAdrian Chadd * the credentials need to match and the 6570a100a6fSAdrian Chadd * original socket also has to have been bound 6580a100a6fSAdrian Chadd * with BINDMULTI. 6590a100a6fSAdrian Chadd */ 6600a100a6fSAdrian Chadd if (t && (! in_pcbbind_check_bindmulti(inp, t))) 6610a100a6fSAdrian Chadd return (EADDRINUSE); 6624049a042SGuido van Rooij } 663c3229e05SDavid Greenman t = in_pcblookup_local(pcbinfo, sin->sin_addr, 66468e0d7e0SRobert Watson lport, lookupflags, cred); 665ad71fe3cSRobert Watson if (t && (t->inp_flags & INP_TIMEWAIT)) { 666ae0e7143SRobert Watson /* 667ae0e7143SRobert Watson * XXXRW: If an incpb has had its timewait 668ae0e7143SRobert Watson * state recycled, we treat the address as 669ae0e7143SRobert Watson * being in use (for now). This is better 670ae0e7143SRobert Watson * than a panic, but not desirable. 671ae0e7143SRobert Watson */ 672ec95b709SMikolaj Golub tw = intotw(t); 673ae0e7143SRobert Watson if (tw == NULL || 674ae0e7143SRobert Watson (reuseport & tw->tw_so_options) == 0) 675340c35deSJonathan Lemon return (EADDRINUSE); 6760a100a6fSAdrian Chadd } else if (t && 6770a100a6fSAdrian Chadd ((inp->inp_flags2 & INP_BINDMULTI) == 0) && 6780a100a6fSAdrian Chadd (reuseport & inp_so_options(t)) == 0) { 679e3fd5ffdSRobert Watson #ifdef INET6 68033841545SHajimu UMEMOTO if (ntohl(sin->sin_addr.s_addr) != 681cfa1ca9dSYoshinobu Inoue INADDR_ANY || 682cfa1ca9dSYoshinobu Inoue ntohl(t->inp_laddr.s_addr) != 683cfa1ca9dSYoshinobu Inoue INADDR_ANY || 684fc06cd42SMikolaj Golub (inp->inp_vflag & INP_IPV6PROTO) == 0 || 685fc06cd42SMikolaj Golub (t->inp_vflag & INP_IPV6PROTO) == 0) 686e3fd5ffdSRobert Watson #endif 687df8bae1dSRodney W. Grimes return (EADDRINUSE); 6880a100a6fSAdrian Chadd if (t && (! in_pcbbind_check_bindmulti(inp, t))) 6890a100a6fSAdrian Chadd return (EADDRINUSE); 690df8bae1dSRodney W. Grimes } 691cfa1ca9dSYoshinobu Inoue } 692df8bae1dSRodney W. Grimes } 6934b932371SIan Dowse if (*lportp != 0) 6944b932371SIan Dowse lport = *lportp; 69533b3ac06SPeter Wemm if (lport == 0) { 69668e0d7e0SRobert Watson error = in_pcb_lport(inp, &laddr, &lport, cred, lookupflags); 697efc76f72SBjoern A. Zeeb if (error != 0) 698efc76f72SBjoern A. Zeeb return (error); 69933b3ac06SPeter Wemm 70033b3ac06SPeter Wemm } 7014b932371SIan Dowse *laddrp = laddr.s_addr; 7024b932371SIan Dowse *lportp = lport; 703df8bae1dSRodney W. Grimes return (0); 704df8bae1dSRodney W. Grimes } 705df8bae1dSRodney W. Grimes 706999f1343SGarrett Wollman /* 7075200e00eSIan Dowse * Connect from a socket to a specified address. 7085200e00eSIan Dowse * Both address and port must be specified in argument sin. 7095200e00eSIan Dowse * If don't have a local address for this socket yet, 7105200e00eSIan Dowse * then pick one. 711999f1343SGarrett Wollman */ 712999f1343SGarrett Wollman int 713d3c1f003SRobert Watson in_pcbconnect_mbuf(struct inpcb *inp, struct sockaddr *nam, 714d3c1f003SRobert Watson struct ucred *cred, struct mbuf *m) 715999f1343SGarrett Wollman { 7165200e00eSIan Dowse u_short lport, fport; 7175200e00eSIan Dowse in_addr_t laddr, faddr; 7185200e00eSIan Dowse int anonport, error; 719df8bae1dSRodney W. Grimes 7208501a69cSRobert Watson INP_WLOCK_ASSERT(inp); 721fa046d87SRobert Watson INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); 72227f74fd0SRobert Watson 7235200e00eSIan Dowse lport = inp->inp_lport; 7245200e00eSIan Dowse laddr = inp->inp_laddr.s_addr; 7255200e00eSIan Dowse anonport = (lport == 0); 7265200e00eSIan Dowse error = in_pcbconnect_setup(inp, nam, &laddr, &lport, &faddr, &fport, 727b0330ed9SPawel Jakub Dawidek NULL, cred); 7285200e00eSIan Dowse if (error) 7295200e00eSIan Dowse return (error); 7305200e00eSIan Dowse 7315200e00eSIan Dowse /* Do the initial binding of the local address if required. */ 7325200e00eSIan Dowse if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) { 7335200e00eSIan Dowse inp->inp_lport = lport; 7345200e00eSIan Dowse inp->inp_laddr.s_addr = laddr; 7355200e00eSIan Dowse if (in_pcbinshash(inp) != 0) { 7365200e00eSIan Dowse inp->inp_laddr.s_addr = INADDR_ANY; 7375200e00eSIan Dowse inp->inp_lport = 0; 7385200e00eSIan Dowse return (EAGAIN); 7395200e00eSIan Dowse } 7405200e00eSIan Dowse } 7415200e00eSIan Dowse 7425200e00eSIan Dowse /* Commit the remaining changes. */ 7435200e00eSIan Dowse inp->inp_lport = lport; 7445200e00eSIan Dowse inp->inp_laddr.s_addr = laddr; 7455200e00eSIan Dowse inp->inp_faddr.s_addr = faddr; 7465200e00eSIan Dowse inp->inp_fport = fport; 747d3c1f003SRobert Watson in_pcbrehash_mbuf(inp, m); 7482cb64cb2SGeorge V. Neville-Neil 7495200e00eSIan Dowse if (anonport) 7505200e00eSIan Dowse inp->inp_flags |= INP_ANONPORT; 7515200e00eSIan Dowse return (0); 7525200e00eSIan Dowse } 7535200e00eSIan Dowse 754d3c1f003SRobert Watson int 755d3c1f003SRobert Watson in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred) 756d3c1f003SRobert Watson { 757d3c1f003SRobert Watson 758d3c1f003SRobert Watson return (in_pcbconnect_mbuf(inp, nam, cred, NULL)); 759d3c1f003SRobert Watson } 760d3c1f003SRobert Watson 7615200e00eSIan Dowse /* 7620895aec3SBjoern A. Zeeb * Do proper source address selection on an unbound socket in case 7630895aec3SBjoern A. Zeeb * of connect. Take jails into account as well. 7640895aec3SBjoern A. Zeeb */ 765ae190832SSteven Hartland int 7660895aec3SBjoern A. Zeeb in_pcbladdr(struct inpcb *inp, struct in_addr *faddr, struct in_addr *laddr, 7670895aec3SBjoern A. Zeeb struct ucred *cred) 7680895aec3SBjoern A. Zeeb { 7690895aec3SBjoern A. Zeeb struct ifaddr *ifa; 7700895aec3SBjoern A. Zeeb struct sockaddr *sa; 7710895aec3SBjoern A. Zeeb struct sockaddr_in *sin; 7720895aec3SBjoern A. Zeeb struct route sro; 7730895aec3SBjoern A. Zeeb int error; 7740895aec3SBjoern A. Zeeb 775413628a7SBjoern A. Zeeb KASSERT(laddr != NULL, ("%s: laddr NULL", __func__)); 7760895aec3SBjoern A. Zeeb 777592bcae8SBjoern A. Zeeb /* 778592bcae8SBjoern A. Zeeb * Bypass source address selection and use the primary jail IP 779592bcae8SBjoern A. Zeeb * if requested. 780592bcae8SBjoern A. Zeeb */ 781592bcae8SBjoern A. Zeeb if (cred != NULL && !prison_saddrsel_ip4(cred, laddr)) 782592bcae8SBjoern A. Zeeb return (0); 783592bcae8SBjoern A. Zeeb 7840895aec3SBjoern A. Zeeb error = 0; 7850895aec3SBjoern A. Zeeb bzero(&sro, sizeof(sro)); 7860895aec3SBjoern A. Zeeb 7870895aec3SBjoern A. Zeeb sin = (struct sockaddr_in *)&sro.ro_dst; 7880895aec3SBjoern A. Zeeb sin->sin_family = AF_INET; 7890895aec3SBjoern A. Zeeb sin->sin_len = sizeof(struct sockaddr_in); 7900895aec3SBjoern A. Zeeb sin->sin_addr.s_addr = faddr->s_addr; 7910895aec3SBjoern A. Zeeb 7920895aec3SBjoern A. Zeeb /* 7930895aec3SBjoern A. Zeeb * If route is known our src addr is taken from the i/f, 7940895aec3SBjoern A. Zeeb * else punt. 7950895aec3SBjoern A. Zeeb * 7960895aec3SBjoern A. Zeeb * Find out route to destination. 7970895aec3SBjoern A. Zeeb */ 7980895aec3SBjoern A. Zeeb if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0) 7996e6b3f7cSQing Li in_rtalloc_ign(&sro, 0, inp->inp_inc.inc_fibnum); 8000895aec3SBjoern A. Zeeb 8010895aec3SBjoern A. Zeeb /* 8020895aec3SBjoern A. Zeeb * If we found a route, use the address corresponding to 8030895aec3SBjoern A. Zeeb * the outgoing interface. 8040895aec3SBjoern A. Zeeb * 8050895aec3SBjoern A. Zeeb * Otherwise assume faddr is reachable on a directly connected 8060895aec3SBjoern A. Zeeb * network and try to find a corresponding interface to take 8070895aec3SBjoern A. Zeeb * the source address from. 8080895aec3SBjoern A. Zeeb */ 8090895aec3SBjoern A. Zeeb if (sro.ro_rt == NULL || sro.ro_rt->rt_ifp == NULL) { 8108c0fec80SRobert Watson struct in_ifaddr *ia; 8110895aec3SBjoern A. Zeeb struct ifnet *ifp; 8120895aec3SBjoern A. Zeeb 8134f8585e0SAlan Somers ia = ifatoia(ifa_ifwithdstaddr((struct sockaddr *)sin, 81458a39d8cSAlan Somers inp->inp_socket->so_fibnum)); 8150895aec3SBjoern A. Zeeb if (ia == NULL) 8164f8585e0SAlan Somers ia = ifatoia(ifa_ifwithnet((struct sockaddr *)sin, 0, 81758a39d8cSAlan Somers inp->inp_socket->so_fibnum)); 8180895aec3SBjoern A. Zeeb if (ia == NULL) { 8190895aec3SBjoern A. Zeeb error = ENETUNREACH; 8200895aec3SBjoern A. Zeeb goto done; 8210895aec3SBjoern A. Zeeb } 8220895aec3SBjoern A. Zeeb 8230304c731SJamie Gritton if (cred == NULL || !prison_flag(cred, PR_IP4)) { 8240895aec3SBjoern A. Zeeb laddr->s_addr = ia->ia_addr.sin_addr.s_addr; 8258c0fec80SRobert Watson ifa_free(&ia->ia_ifa); 8260895aec3SBjoern A. Zeeb goto done; 8270895aec3SBjoern A. Zeeb } 8280895aec3SBjoern A. Zeeb 8290895aec3SBjoern A. Zeeb ifp = ia->ia_ifp; 8308c0fec80SRobert Watson ifa_free(&ia->ia_ifa); 8310895aec3SBjoern A. Zeeb ia = NULL; 832137f91e8SJohn Baldwin IF_ADDR_RLOCK(ifp); 8330895aec3SBjoern A. Zeeb TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 8340895aec3SBjoern A. Zeeb 8350895aec3SBjoern A. Zeeb sa = ifa->ifa_addr; 8360895aec3SBjoern A. Zeeb if (sa->sa_family != AF_INET) 8370895aec3SBjoern A. Zeeb continue; 8380895aec3SBjoern A. Zeeb sin = (struct sockaddr_in *)sa; 839b89e82ddSJamie Gritton if (prison_check_ip4(cred, &sin->sin_addr) == 0) { 8400895aec3SBjoern A. Zeeb ia = (struct in_ifaddr *)ifa; 8410895aec3SBjoern A. Zeeb break; 8420895aec3SBjoern A. Zeeb } 8430895aec3SBjoern A. Zeeb } 8440895aec3SBjoern A. Zeeb if (ia != NULL) { 8450895aec3SBjoern A. Zeeb laddr->s_addr = ia->ia_addr.sin_addr.s_addr; 846137f91e8SJohn Baldwin IF_ADDR_RUNLOCK(ifp); 8470895aec3SBjoern A. Zeeb goto done; 8480895aec3SBjoern A. Zeeb } 849137f91e8SJohn Baldwin IF_ADDR_RUNLOCK(ifp); 8500895aec3SBjoern A. Zeeb 8510895aec3SBjoern A. Zeeb /* 3. As a last resort return the 'default' jail address. */ 852b89e82ddSJamie Gritton error = prison_get_ip4(cred, laddr); 8530895aec3SBjoern A. Zeeb goto done; 8540895aec3SBjoern A. Zeeb } 8550895aec3SBjoern A. Zeeb 8560895aec3SBjoern A. Zeeb /* 8570895aec3SBjoern A. Zeeb * If the outgoing interface on the route found is not 8580895aec3SBjoern A. Zeeb * a loopback interface, use the address from that interface. 8590895aec3SBjoern A. Zeeb * In case of jails do those three steps: 8600895aec3SBjoern A. Zeeb * 1. check if the interface address belongs to the jail. If so use it. 8610895aec3SBjoern A. Zeeb * 2. check if we have any address on the outgoing interface 8620895aec3SBjoern A. Zeeb * belonging to this jail. If so use it. 8630895aec3SBjoern A. Zeeb * 3. as a last resort return the 'default' jail address. 8640895aec3SBjoern A. Zeeb */ 8650895aec3SBjoern A. Zeeb if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0) { 8668c0fec80SRobert Watson struct in_ifaddr *ia; 8679317b04eSRobert Watson struct ifnet *ifp; 8680895aec3SBjoern A. Zeeb 8690895aec3SBjoern A. Zeeb /* If not jailed, use the default returned. */ 8700304c731SJamie Gritton if (cred == NULL || !prison_flag(cred, PR_IP4)) { 8710895aec3SBjoern A. Zeeb ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa; 8720895aec3SBjoern A. Zeeb laddr->s_addr = ia->ia_addr.sin_addr.s_addr; 8730895aec3SBjoern A. Zeeb goto done; 8740895aec3SBjoern A. Zeeb } 8750895aec3SBjoern A. Zeeb 8760895aec3SBjoern A. Zeeb /* Jailed. */ 8770895aec3SBjoern A. Zeeb /* 1. Check if the iface address belongs to the jail. */ 8780895aec3SBjoern A. Zeeb sin = (struct sockaddr_in *)sro.ro_rt->rt_ifa->ifa_addr; 879b89e82ddSJamie Gritton if (prison_check_ip4(cred, &sin->sin_addr) == 0) { 8800895aec3SBjoern A. Zeeb ia = (struct in_ifaddr *)sro.ro_rt->rt_ifa; 8810895aec3SBjoern A. Zeeb laddr->s_addr = ia->ia_addr.sin_addr.s_addr; 8820895aec3SBjoern A. Zeeb goto done; 8830895aec3SBjoern A. Zeeb } 8840895aec3SBjoern A. Zeeb 8850895aec3SBjoern A. Zeeb /* 8860895aec3SBjoern A. Zeeb * 2. Check if we have any address on the outgoing interface 8870895aec3SBjoern A. Zeeb * belonging to this jail. 8880895aec3SBjoern A. Zeeb */ 8898c0fec80SRobert Watson ia = NULL; 8909317b04eSRobert Watson ifp = sro.ro_rt->rt_ifp; 891137f91e8SJohn Baldwin IF_ADDR_RLOCK(ifp); 8929317b04eSRobert Watson TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 8930895aec3SBjoern A. Zeeb sa = ifa->ifa_addr; 8940895aec3SBjoern A. Zeeb if (sa->sa_family != AF_INET) 8950895aec3SBjoern A. Zeeb continue; 8960895aec3SBjoern A. Zeeb sin = (struct sockaddr_in *)sa; 897b89e82ddSJamie Gritton if (prison_check_ip4(cred, &sin->sin_addr) == 0) { 8980895aec3SBjoern A. Zeeb ia = (struct in_ifaddr *)ifa; 8990895aec3SBjoern A. Zeeb break; 9000895aec3SBjoern A. Zeeb } 9010895aec3SBjoern A. Zeeb } 9020895aec3SBjoern A. Zeeb if (ia != NULL) { 9030895aec3SBjoern A. Zeeb laddr->s_addr = ia->ia_addr.sin_addr.s_addr; 904137f91e8SJohn Baldwin IF_ADDR_RUNLOCK(ifp); 9050895aec3SBjoern A. Zeeb goto done; 9060895aec3SBjoern A. Zeeb } 907137f91e8SJohn Baldwin IF_ADDR_RUNLOCK(ifp); 9080895aec3SBjoern A. Zeeb 9090895aec3SBjoern A. Zeeb /* 3. As a last resort return the 'default' jail address. */ 910b89e82ddSJamie Gritton error = prison_get_ip4(cred, laddr); 9110895aec3SBjoern A. Zeeb goto done; 9120895aec3SBjoern A. Zeeb } 9130895aec3SBjoern A. Zeeb 9140895aec3SBjoern A. Zeeb /* 9150895aec3SBjoern A. Zeeb * The outgoing interface is marked with 'loopback net', so a route 9160895aec3SBjoern A. Zeeb * to ourselves is here. 9170895aec3SBjoern A. Zeeb * Try to find the interface of the destination address and then 9180895aec3SBjoern A. Zeeb * take the address from there. That interface is not necessarily 9190895aec3SBjoern A. Zeeb * a loopback interface. 9200895aec3SBjoern A. Zeeb * In case of jails, check that it is an address of the jail 9210895aec3SBjoern A. Zeeb * and if we cannot find, fall back to the 'default' jail address. 9220895aec3SBjoern A. Zeeb */ 9230895aec3SBjoern A. Zeeb if ((sro.ro_rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) { 9240895aec3SBjoern A. Zeeb struct sockaddr_in sain; 9258c0fec80SRobert Watson struct in_ifaddr *ia; 9260895aec3SBjoern A. Zeeb 9270895aec3SBjoern A. Zeeb bzero(&sain, sizeof(struct sockaddr_in)); 9280895aec3SBjoern A. Zeeb sain.sin_family = AF_INET; 9290895aec3SBjoern A. Zeeb sain.sin_len = sizeof(struct sockaddr_in); 9300895aec3SBjoern A. Zeeb sain.sin_addr.s_addr = faddr->s_addr; 9310895aec3SBjoern A. Zeeb 93258a39d8cSAlan Somers ia = ifatoia(ifa_ifwithdstaddr(sintosa(&sain), 93358a39d8cSAlan Somers inp->inp_socket->so_fibnum)); 9340895aec3SBjoern A. Zeeb if (ia == NULL) 9354f8585e0SAlan Somers ia = ifatoia(ifa_ifwithnet(sintosa(&sain), 0, 93658a39d8cSAlan Somers inp->inp_socket->so_fibnum)); 937f0bb05fcSQing Li if (ia == NULL) 938f0bb05fcSQing Li ia = ifatoia(ifa_ifwithaddr(sintosa(&sain))); 9390895aec3SBjoern A. Zeeb 9400304c731SJamie Gritton if (cred == NULL || !prison_flag(cred, PR_IP4)) { 9410895aec3SBjoern A. Zeeb if (ia == NULL) { 9420895aec3SBjoern A. Zeeb error = ENETUNREACH; 9430895aec3SBjoern A. Zeeb goto done; 9440895aec3SBjoern A. Zeeb } 9450895aec3SBjoern A. Zeeb laddr->s_addr = ia->ia_addr.sin_addr.s_addr; 9468c0fec80SRobert Watson ifa_free(&ia->ia_ifa); 9470895aec3SBjoern A. Zeeb goto done; 9480895aec3SBjoern A. Zeeb } 9490895aec3SBjoern A. Zeeb 9500895aec3SBjoern A. Zeeb /* Jailed. */ 9510895aec3SBjoern A. Zeeb if (ia != NULL) { 9520895aec3SBjoern A. Zeeb struct ifnet *ifp; 9530895aec3SBjoern A. Zeeb 9540895aec3SBjoern A. Zeeb ifp = ia->ia_ifp; 9558c0fec80SRobert Watson ifa_free(&ia->ia_ifa); 9560895aec3SBjoern A. Zeeb ia = NULL; 957137f91e8SJohn Baldwin IF_ADDR_RLOCK(ifp); 9580895aec3SBjoern A. Zeeb TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 9590895aec3SBjoern A. Zeeb 9600895aec3SBjoern A. Zeeb sa = ifa->ifa_addr; 9610895aec3SBjoern A. Zeeb if (sa->sa_family != AF_INET) 9620895aec3SBjoern A. Zeeb continue; 9630895aec3SBjoern A. Zeeb sin = (struct sockaddr_in *)sa; 964b89e82ddSJamie Gritton if (prison_check_ip4(cred, 965b89e82ddSJamie Gritton &sin->sin_addr) == 0) { 9660895aec3SBjoern A. Zeeb ia = (struct in_ifaddr *)ifa; 9670895aec3SBjoern A. Zeeb break; 9680895aec3SBjoern A. Zeeb } 9690895aec3SBjoern A. Zeeb } 9700895aec3SBjoern A. Zeeb if (ia != NULL) { 9710895aec3SBjoern A. Zeeb laddr->s_addr = ia->ia_addr.sin_addr.s_addr; 972137f91e8SJohn Baldwin IF_ADDR_RUNLOCK(ifp); 9730895aec3SBjoern A. Zeeb goto done; 9740895aec3SBjoern A. Zeeb } 975137f91e8SJohn Baldwin IF_ADDR_RUNLOCK(ifp); 9760895aec3SBjoern A. Zeeb } 9770895aec3SBjoern A. Zeeb 9780895aec3SBjoern A. Zeeb /* 3. As a last resort return the 'default' jail address. */ 979b89e82ddSJamie Gritton error = prison_get_ip4(cred, laddr); 9800895aec3SBjoern A. Zeeb goto done; 9810895aec3SBjoern A. Zeeb } 9820895aec3SBjoern A. Zeeb 9830895aec3SBjoern A. Zeeb done: 9840895aec3SBjoern A. Zeeb if (sro.ro_rt != NULL) 9850895aec3SBjoern A. Zeeb RTFREE(sro.ro_rt); 9860895aec3SBjoern A. Zeeb return (error); 9870895aec3SBjoern A. Zeeb } 9880895aec3SBjoern A. Zeeb 9890895aec3SBjoern A. Zeeb /* 9905200e00eSIan Dowse * Set up for a connect from a socket to the specified address. 9915200e00eSIan Dowse * On entry, *laddrp and *lportp should contain the current local 9925200e00eSIan Dowse * address and port for the PCB; these are updated to the values 9935200e00eSIan Dowse * that should be placed in inp_laddr and inp_lport to complete 9945200e00eSIan Dowse * the connect. 9955200e00eSIan Dowse * 9965200e00eSIan Dowse * On success, *faddrp and *fportp will be set to the remote address 9975200e00eSIan Dowse * and port. These are not updated in the error case. 9985200e00eSIan Dowse * 9995200e00eSIan Dowse * If the operation fails because the connection already exists, 10005200e00eSIan Dowse * *oinpp will be set to the PCB of that connection so that the 10015200e00eSIan Dowse * caller can decide to override it. In all other cases, *oinpp 10025200e00eSIan Dowse * is set to NULL. 10035200e00eSIan Dowse */ 10045200e00eSIan Dowse int 1005136d4f1cSRobert Watson in_pcbconnect_setup(struct inpcb *inp, struct sockaddr *nam, 1006136d4f1cSRobert Watson in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp, 1007136d4f1cSRobert Watson struct inpcb **oinpp, struct ucred *cred) 10085200e00eSIan Dowse { 1009cc0a3c8cSAndrey V. Elsukov struct rm_priotracker in_ifa_tracker; 10105200e00eSIan Dowse struct sockaddr_in *sin = (struct sockaddr_in *)nam; 10115200e00eSIan Dowse struct in_ifaddr *ia; 10125200e00eSIan Dowse struct inpcb *oinp; 1013b89e82ddSJamie Gritton struct in_addr laddr, faddr; 10145200e00eSIan Dowse u_short lport, fport; 10155200e00eSIan Dowse int error; 10165200e00eSIan Dowse 10178501a69cSRobert Watson /* 10188501a69cSRobert Watson * Because a global state change doesn't actually occur here, a read 10198501a69cSRobert Watson * lock is sufficient. 10208501a69cSRobert Watson */ 102127f74fd0SRobert Watson INP_LOCK_ASSERT(inp); 1022fa046d87SRobert Watson INP_HASH_LOCK_ASSERT(inp->inp_pcbinfo); 102327f74fd0SRobert Watson 10245200e00eSIan Dowse if (oinpp != NULL) 10255200e00eSIan Dowse *oinpp = NULL; 102657bf258eSGarrett Wollman if (nam->sa_len != sizeof (*sin)) 1027df8bae1dSRodney W. Grimes return (EINVAL); 1028df8bae1dSRodney W. Grimes if (sin->sin_family != AF_INET) 1029df8bae1dSRodney W. Grimes return (EAFNOSUPPORT); 1030df8bae1dSRodney W. Grimes if (sin->sin_port == 0) 1031df8bae1dSRodney W. Grimes return (EADDRNOTAVAIL); 10325200e00eSIan Dowse laddr.s_addr = *laddrp; 10335200e00eSIan Dowse lport = *lportp; 10345200e00eSIan Dowse faddr = sin->sin_addr; 10355200e00eSIan Dowse fport = sin->sin_port; 10360895aec3SBjoern A. Zeeb 1037603724d3SBjoern A. Zeeb if (!TAILQ_EMPTY(&V_in_ifaddrhead)) { 1038df8bae1dSRodney W. Grimes /* 1039df8bae1dSRodney W. Grimes * If the destination address is INADDR_ANY, 1040df8bae1dSRodney W. Grimes * use the primary local address. 1041df8bae1dSRodney W. Grimes * If the supplied address is INADDR_BROADCAST, 1042df8bae1dSRodney W. Grimes * and the primary interface supports broadcast, 1043df8bae1dSRodney W. Grimes * choose the broadcast address for that interface. 1044df8bae1dSRodney W. Grimes */ 1045413628a7SBjoern A. Zeeb if (faddr.s_addr == INADDR_ANY) { 1046cc0a3c8cSAndrey V. Elsukov IN_IFADDR_RLOCK(&in_ifa_tracker); 1047413628a7SBjoern A. Zeeb faddr = 1048b89e82ddSJamie Gritton IA_SIN(TAILQ_FIRST(&V_in_ifaddrhead))->sin_addr; 1049cc0a3c8cSAndrey V. Elsukov IN_IFADDR_RUNLOCK(&in_ifa_tracker); 1050b89e82ddSJamie Gritton if (cred != NULL && 1051b89e82ddSJamie Gritton (error = prison_get_ip4(cred, &faddr)) != 0) 1052b89e82ddSJamie Gritton return (error); 10532d9cfabaSRobert Watson } else if (faddr.s_addr == (u_long)INADDR_BROADCAST) { 1054cc0a3c8cSAndrey V. Elsukov IN_IFADDR_RLOCK(&in_ifa_tracker); 10552d9cfabaSRobert Watson if (TAILQ_FIRST(&V_in_ifaddrhead)->ia_ifp->if_flags & 10562d9cfabaSRobert Watson IFF_BROADCAST) 10575200e00eSIan Dowse faddr = satosin(&TAILQ_FIRST( 1058603724d3SBjoern A. Zeeb &V_in_ifaddrhead)->ia_broadaddr)->sin_addr; 1059cc0a3c8cSAndrey V. Elsukov IN_IFADDR_RUNLOCK(&in_ifa_tracker); 10602d9cfabaSRobert Watson } 1061df8bae1dSRodney W. Grimes } 10625200e00eSIan Dowse if (laddr.s_addr == INADDR_ANY) { 1063d79fdd98SDaniel Eischen error = in_pcbladdr(inp, &faddr, &laddr, cred); 1064df8bae1dSRodney W. Grimes /* 1065df8bae1dSRodney W. Grimes * If the destination address is multicast and an outgoing 1066d79fdd98SDaniel Eischen * interface has been set as a multicast option, prefer the 1067df8bae1dSRodney W. Grimes * address of that interface as our source address. 1068df8bae1dSRodney W. Grimes */ 10695200e00eSIan Dowse if (IN_MULTICAST(ntohl(faddr.s_addr)) && 1070df8bae1dSRodney W. Grimes inp->inp_moptions != NULL) { 1071df8bae1dSRodney W. Grimes struct ip_moptions *imo; 1072df8bae1dSRodney W. Grimes struct ifnet *ifp; 1073df8bae1dSRodney W. Grimes 1074df8bae1dSRodney W. Grimes imo = inp->inp_moptions; 1075df8bae1dSRodney W. Grimes if (imo->imo_multicast_ifp != NULL) { 1076df8bae1dSRodney W. Grimes ifp = imo->imo_multicast_ifp; 1077cc0a3c8cSAndrey V. Elsukov IN_IFADDR_RLOCK(&in_ifa_tracker); 1078e691be70SDaniel Eischen TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 1079e691be70SDaniel Eischen if ((ia->ia_ifp == ifp) && 1080e691be70SDaniel Eischen (cred == NULL || 1081e691be70SDaniel Eischen prison_check_ip4(cred, 1082e691be70SDaniel Eischen &ia->ia_addr.sin_addr) == 0)) 1083df8bae1dSRodney W. Grimes break; 1084e691be70SDaniel Eischen } 1085e691be70SDaniel Eischen if (ia == NULL) 1086d79fdd98SDaniel Eischen error = EADDRNOTAVAIL; 1087e691be70SDaniel Eischen else { 10885200e00eSIan Dowse laddr = ia->ia_addr.sin_addr; 1089d79fdd98SDaniel Eischen error = 0; 1090999f1343SGarrett Wollman } 1091cc0a3c8cSAndrey V. Elsukov IN_IFADDR_RUNLOCK(&in_ifa_tracker); 1092d79fdd98SDaniel Eischen } 1093d79fdd98SDaniel Eischen } 109404215ed2SRandall Stewart if (error) 109504215ed2SRandall Stewart return (error); 10960895aec3SBjoern A. Zeeb } 1097fa046d87SRobert Watson oinp = in_pcblookup_hash_locked(inp->inp_pcbinfo, faddr, fport, 1098fa046d87SRobert Watson laddr, lport, 0, NULL); 10995200e00eSIan Dowse if (oinp != NULL) { 11005200e00eSIan Dowse if (oinpp != NULL) 11015200e00eSIan Dowse *oinpp = oinp; 1102df8bae1dSRodney W. Grimes return (EADDRINUSE); 1103c3229e05SDavid Greenman } 11045200e00eSIan Dowse if (lport == 0) { 1105b0330ed9SPawel Jakub Dawidek error = in_pcbbind_setup(inp, NULL, &laddr.s_addr, &lport, 1106b0330ed9SPawel Jakub Dawidek cred); 11075a903f8dSPierre Beyssac if (error) 11085a903f8dSPierre Beyssac return (error); 11095a903f8dSPierre Beyssac } 11105200e00eSIan Dowse *laddrp = laddr.s_addr; 11115200e00eSIan Dowse *lportp = lport; 11125200e00eSIan Dowse *faddrp = faddr.s_addr; 11135200e00eSIan Dowse *fportp = fport; 1114df8bae1dSRodney W. Grimes return (0); 1115df8bae1dSRodney W. Grimes } 1116df8bae1dSRodney W. Grimes 111726f9a767SRodney W. Grimes void 1118136d4f1cSRobert Watson in_pcbdisconnect(struct inpcb *inp) 1119df8bae1dSRodney W. Grimes { 11206b348152SRobert Watson 11218501a69cSRobert Watson INP_WLOCK_ASSERT(inp); 1122fa046d87SRobert Watson INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); 1123df8bae1dSRodney W. Grimes 1124df8bae1dSRodney W. Grimes inp->inp_faddr.s_addr = INADDR_ANY; 1125df8bae1dSRodney W. Grimes inp->inp_fport = 0; 112615bd2b43SDavid Greenman in_pcbrehash(inp); 1127df8bae1dSRodney W. Grimes } 112883e521ecSBjoern A. Zeeb #endif /* INET */ 1129df8bae1dSRodney W. Grimes 11304c7c478dSRobert Watson /* 113128696211SRobert Watson * in_pcbdetach() is responsibe for disassociating a socket from an inpcb. 1132c0a211c5SRobert Watson * For most protocols, this will be invoked immediately prior to calling 113328696211SRobert Watson * in_pcbfree(). However, with TCP the inpcb may significantly outlive the 113428696211SRobert Watson * socket, in which case in_pcbfree() is deferred. 11354c7c478dSRobert Watson */ 113626f9a767SRodney W. Grimes void 1137136d4f1cSRobert Watson in_pcbdetach(struct inpcb *inp) 1138df8bae1dSRodney W. Grimes { 11394c7c478dSRobert Watson 1140a7df09e8SBjoern A. Zeeb KASSERT(inp->inp_socket != NULL, ("%s: inp_socket == NULL", __func__)); 1141c0a211c5SRobert Watson 11424c7c478dSRobert Watson inp->inp_socket->so_pcb = NULL; 11434c7c478dSRobert Watson inp->inp_socket = NULL; 11444c7c478dSRobert Watson } 11454c7c478dSRobert Watson 1146c0a211c5SRobert Watson /* 114779bdc6e5SRobert Watson * in_pcbref() bumps the reference count on an inpcb in order to maintain 114879bdc6e5SRobert Watson * stability of an inpcb pointer despite the inpcb lock being released. This 114979bdc6e5SRobert Watson * is used in TCP when the inpcbinfo lock needs to be acquired or upgraded, 115052cd27cbSRobert Watson * but where the inpcb lock may already held, or when acquiring a reference 115152cd27cbSRobert Watson * via a pcbgroup. 115279bdc6e5SRobert Watson * 115379bdc6e5SRobert Watson * in_pcbref() should be used only to provide brief memory stability, and 115479bdc6e5SRobert Watson * must always be followed by a call to INP_WLOCK() and in_pcbrele() to 115579bdc6e5SRobert Watson * garbage collect the inpcb if it has been in_pcbfree()'d from another 115679bdc6e5SRobert Watson * context. Until in_pcbrele() has returned that the inpcb is still valid, 115779bdc6e5SRobert Watson * lock and rele are the *only* safe operations that may be performed on the 115879bdc6e5SRobert Watson * inpcb. 115979bdc6e5SRobert Watson * 116079bdc6e5SRobert Watson * While the inpcb will not be freed, releasing the inpcb lock means that the 116179bdc6e5SRobert Watson * connection's state may change, so the caller should be careful to 116279bdc6e5SRobert Watson * revalidate any cached state on reacquiring the lock. Drop the reference 116379bdc6e5SRobert Watson * using in_pcbrele(). 1164c0a211c5SRobert Watson */ 116579bdc6e5SRobert Watson void 116679bdc6e5SRobert Watson in_pcbref(struct inpcb *inp) 11674c7c478dSRobert Watson { 1168df8bae1dSRodney W. Grimes 116979bdc6e5SRobert Watson KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__)); 117079bdc6e5SRobert Watson 117179bdc6e5SRobert Watson refcount_acquire(&inp->inp_refcount); 117279bdc6e5SRobert Watson } 117379bdc6e5SRobert Watson 117479bdc6e5SRobert Watson /* 117579bdc6e5SRobert Watson * Drop a refcount on an inpcb elevated using in_pcbref(); because a call to 117679bdc6e5SRobert Watson * in_pcbfree() may have been made between in_pcbref() and in_pcbrele(), we 117779bdc6e5SRobert Watson * return a flag indicating whether or not the inpcb remains valid. If it is 117879bdc6e5SRobert Watson * valid, we return with the inpcb lock held. 117979bdc6e5SRobert Watson * 118079bdc6e5SRobert Watson * Notice that, unlike in_pcbref(), the inpcb lock must be held to drop a 118179bdc6e5SRobert Watson * reference on an inpcb. Historically more work was done here (actually, in 118279bdc6e5SRobert Watson * in_pcbfree_internal()) but has been moved to in_pcbfree() to avoid the 118379bdc6e5SRobert Watson * need for the pcbinfo lock in in_pcbrele(). Deferring the free is entirely 118479bdc6e5SRobert Watson * about memory stability (and continued use of the write lock). 118579bdc6e5SRobert Watson */ 118679bdc6e5SRobert Watson int 118779bdc6e5SRobert Watson in_pcbrele_rlocked(struct inpcb *inp) 118879bdc6e5SRobert Watson { 118979bdc6e5SRobert Watson struct inpcbinfo *pcbinfo; 119079bdc6e5SRobert Watson 119179bdc6e5SRobert Watson KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__)); 119279bdc6e5SRobert Watson 119379bdc6e5SRobert Watson INP_RLOCK_ASSERT(inp); 119479bdc6e5SRobert Watson 1195df4e91d3SGleb Smirnoff if (refcount_release(&inp->inp_refcount) == 0) { 1196df4e91d3SGleb Smirnoff /* 1197df4e91d3SGleb Smirnoff * If the inpcb has been freed, let the caller know, even if 1198df4e91d3SGleb Smirnoff * this isn't the last reference. 1199df4e91d3SGleb Smirnoff */ 1200df4e91d3SGleb Smirnoff if (inp->inp_flags2 & INP_FREED) { 1201df4e91d3SGleb Smirnoff INP_RUNLOCK(inp); 1202df4e91d3SGleb Smirnoff return (1); 1203df4e91d3SGleb Smirnoff } 120479bdc6e5SRobert Watson return (0); 1205df4e91d3SGleb Smirnoff } 120679bdc6e5SRobert Watson 120779bdc6e5SRobert Watson KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__)); 120879bdc6e5SRobert Watson 120979bdc6e5SRobert Watson INP_RUNLOCK(inp); 121079bdc6e5SRobert Watson pcbinfo = inp->inp_pcbinfo; 121179bdc6e5SRobert Watson uma_zfree(pcbinfo->ipi_zone, inp); 121279bdc6e5SRobert Watson return (1); 121379bdc6e5SRobert Watson } 121479bdc6e5SRobert Watson 121579bdc6e5SRobert Watson int 121679bdc6e5SRobert Watson in_pcbrele_wlocked(struct inpcb *inp) 121779bdc6e5SRobert Watson { 121879bdc6e5SRobert Watson struct inpcbinfo *pcbinfo; 121979bdc6e5SRobert Watson 122079bdc6e5SRobert Watson KASSERT(inp->inp_refcount > 0, ("%s: refcount 0", __func__)); 122179bdc6e5SRobert Watson 122279bdc6e5SRobert Watson INP_WLOCK_ASSERT(inp); 122379bdc6e5SRobert Watson 1224edd0e0b0SFabien Thomas if (refcount_release(&inp->inp_refcount) == 0) { 1225edd0e0b0SFabien Thomas /* 1226edd0e0b0SFabien Thomas * If the inpcb has been freed, let the caller know, even if 1227edd0e0b0SFabien Thomas * this isn't the last reference. 1228edd0e0b0SFabien Thomas */ 1229edd0e0b0SFabien Thomas if (inp->inp_flags2 & INP_FREED) { 1230edd0e0b0SFabien Thomas INP_WUNLOCK(inp); 1231edd0e0b0SFabien Thomas return (1); 1232edd0e0b0SFabien Thomas } 123379bdc6e5SRobert Watson return (0); 1234edd0e0b0SFabien Thomas } 123579bdc6e5SRobert Watson 123679bdc6e5SRobert Watson KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__)); 123779bdc6e5SRobert Watson 123879bdc6e5SRobert Watson INP_WUNLOCK(inp); 123979bdc6e5SRobert Watson pcbinfo = inp->inp_pcbinfo; 124079bdc6e5SRobert Watson uma_zfree(pcbinfo->ipi_zone, inp); 124179bdc6e5SRobert Watson return (1); 124279bdc6e5SRobert Watson } 124379bdc6e5SRobert Watson 124479bdc6e5SRobert Watson /* 124579bdc6e5SRobert Watson * Temporary wrapper. 124679bdc6e5SRobert Watson */ 124779bdc6e5SRobert Watson int 124879bdc6e5SRobert Watson in_pcbrele(struct inpcb *inp) 124979bdc6e5SRobert Watson { 125079bdc6e5SRobert Watson 125179bdc6e5SRobert Watson return (in_pcbrele_wlocked(inp)); 125279bdc6e5SRobert Watson } 125379bdc6e5SRobert Watson 125479bdc6e5SRobert Watson /* 125579bdc6e5SRobert Watson * Unconditionally schedule an inpcb to be freed by decrementing its 125679bdc6e5SRobert Watson * reference count, which should occur only after the inpcb has been detached 125779bdc6e5SRobert Watson * from its socket. If another thread holds a temporary reference (acquired 125879bdc6e5SRobert Watson * using in_pcbref()) then the free is deferred until that reference is 125979bdc6e5SRobert Watson * released using in_pcbrele(), but the inpcb is still unlocked. Almost all 126079bdc6e5SRobert Watson * work, including removal from global lists, is done in this context, where 126179bdc6e5SRobert Watson * the pcbinfo lock is held. 126279bdc6e5SRobert Watson */ 126379bdc6e5SRobert Watson void 126479bdc6e5SRobert Watson in_pcbfree(struct inpcb *inp) 126579bdc6e5SRobert Watson { 126679bdc6e5SRobert Watson struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 126779bdc6e5SRobert Watson 126879bdc6e5SRobert Watson KASSERT(inp->inp_socket == NULL, ("%s: inp_socket != NULL", __func__)); 126979bdc6e5SRobert Watson 1270ff9b006dSJulien Charbon #ifdef INVARIANTS 1271ff9b006dSJulien Charbon if (pcbinfo == &V_tcbinfo) { 1272079672cbSJulien Charbon INP_INFO_LOCK_ASSERT(pcbinfo); 1273ff9b006dSJulien Charbon } else { 127479bdc6e5SRobert Watson INP_INFO_WLOCK_ASSERT(pcbinfo); 1275ff9b006dSJulien Charbon } 1276ff9b006dSJulien Charbon #endif 127779bdc6e5SRobert Watson INP_WLOCK_ASSERT(inp); 127879bdc6e5SRobert Watson 127979bdc6e5SRobert Watson /* XXXRW: Do as much as possible here. */ 1280b2630c29SGeorge V. Neville-Neil #ifdef IPSEC 12816aee2fc5SBjoern A. Zeeb if (inp->inp_sp != NULL) 12826974bd9eSBjoern A. Zeeb ipsec_delete_pcbpolicy(inp); 128383e521ecSBjoern A. Zeeb #endif 1284ff9b006dSJulien Charbon INP_LIST_WLOCK(pcbinfo); 128579bdc6e5SRobert Watson inp->inp_gencnt = ++pcbinfo->ipi_gencnt; 1286c3229e05SDavid Greenman in_pcbremlists(inp); 1287ff9b006dSJulien Charbon INP_LIST_WUNLOCK(pcbinfo); 12886aee2fc5SBjoern A. Zeeb #ifdef INET6 12896aee2fc5SBjoern A. Zeeb if (inp->inp_vflag & INP_IPV6PROTO) { 12906aee2fc5SBjoern A. Zeeb ip6_freepcbopts(inp->in6p_outputopts); 12911096332aSBruce M Simpson if (inp->in6p_moptions != NULL) 12926aee2fc5SBjoern A. Zeeb ip6_freemoptions(inp->in6p_moptions); 12936aee2fc5SBjoern A. Zeeb } 12946aee2fc5SBjoern A. Zeeb #endif 1295df8bae1dSRodney W. Grimes if (inp->inp_options) 1296df8bae1dSRodney W. Grimes (void)m_free(inp->inp_options); 129767107f45SBjoern A. Zeeb #ifdef INET 129871498f30SBruce M Simpson if (inp->inp_moptions != NULL) 129971498f30SBruce M Simpson inp_freemoptions(inp->inp_moptions); 130067107f45SBjoern A. Zeeb #endif 1301*84cc0778SGeorge V. Neville-Neil if (inp->inp_route.ro_rt) { 1302*84cc0778SGeorge V. Neville-Neil RTFREE(inp->inp_route.ro_rt); 1303*84cc0778SGeorge V. Neville-Neil inp->inp_route.ro_rt = (struct rtentry *)NULL; 1304*84cc0778SGeorge V. Neville-Neil } 1305*84cc0778SGeorge V. Neville-Neil 1306cfa1ca9dSYoshinobu Inoue inp->inp_vflag = 0; 1307df4e91d3SGleb Smirnoff inp->inp_flags2 |= INP_FREED; 130886d02c5cSBjoern A. Zeeb crfree(inp->inp_cred); 1309a557af22SRobert Watson #ifdef MAC 131030d239bcSRobert Watson mac_inpcb_destroy(inp); 1311a557af22SRobert Watson #endif 131279bdc6e5SRobert Watson if (!in_pcbrele_wlocked(inp)) 131328696211SRobert Watson INP_WUNLOCK(inp); 131428696211SRobert Watson } 131528696211SRobert Watson 131628696211SRobert Watson /* 1317c0a211c5SRobert Watson * in_pcbdrop() removes an inpcb from hashed lists, releasing its address and 1318c0a211c5SRobert Watson * port reservation, and preventing it from being returned by inpcb lookups. 1319c0a211c5SRobert Watson * 1320c0a211c5SRobert Watson * It is used by TCP to mark an inpcb as unused and avoid future packet 1321c0a211c5SRobert Watson * delivery or event notification when a socket remains open but TCP has 1322c0a211c5SRobert Watson * closed. This might occur as a result of a shutdown()-initiated TCP close 1323c0a211c5SRobert Watson * or a RST on the wire, and allows the port binding to be reused while still 1324c0a211c5SRobert Watson * maintaining the invariant that so_pcb always points to a valid inpcb until 1325c0a211c5SRobert Watson * in_pcbdetach(). 1326c0a211c5SRobert Watson * 1327c0a211c5SRobert Watson * XXXRW: Possibly in_pcbdrop() should also prevent future notifications by 1328c0a211c5SRobert Watson * in_pcbnotifyall() and in_pcbpurgeif0()? 132910702a28SRobert Watson */ 133010702a28SRobert Watson void 133110702a28SRobert Watson in_pcbdrop(struct inpcb *inp) 133210702a28SRobert Watson { 133310702a28SRobert Watson 13348501a69cSRobert Watson INP_WLOCK_ASSERT(inp); 133510702a28SRobert Watson 1336fa046d87SRobert Watson /* 1337fa046d87SRobert Watson * XXXRW: Possibly we should protect the setting of INP_DROPPED with 1338fa046d87SRobert Watson * the hash lock...? 1339fa046d87SRobert Watson */ 1340ad71fe3cSRobert Watson inp->inp_flags |= INP_DROPPED; 1341111d57a6SRobert Watson if (inp->inp_flags & INP_INHASHLIST) { 134210702a28SRobert Watson struct inpcbport *phd = inp->inp_phd; 134310702a28SRobert Watson 1344fa046d87SRobert Watson INP_HASH_WLOCK(inp->inp_pcbinfo); 134510702a28SRobert Watson LIST_REMOVE(inp, inp_hash); 134610702a28SRobert Watson LIST_REMOVE(inp, inp_portlist); 134710702a28SRobert Watson if (LIST_FIRST(&phd->phd_pcblist) == NULL) { 134810702a28SRobert Watson LIST_REMOVE(phd, phd_hash); 134910702a28SRobert Watson free(phd, M_PCB); 135010702a28SRobert Watson } 1351fa046d87SRobert Watson INP_HASH_WUNLOCK(inp->inp_pcbinfo); 1352111d57a6SRobert Watson inp->inp_flags &= ~INP_INHASHLIST; 135352cd27cbSRobert Watson #ifdef PCBGROUP 135452cd27cbSRobert Watson in_pcbgroup_remove(inp); 135552cd27cbSRobert Watson #endif 135610702a28SRobert Watson } 135710702a28SRobert Watson } 135810702a28SRobert Watson 135967107f45SBjoern A. Zeeb #ifdef INET 136054d642bbSRobert Watson /* 136154d642bbSRobert Watson * Common routines to return the socket addresses associated with inpcbs. 136254d642bbSRobert Watson */ 136326ef6ac4SDon Lewis struct sockaddr * 1364136d4f1cSRobert Watson in_sockaddr(in_port_t port, struct in_addr *addr_p) 136526ef6ac4SDon Lewis { 136626ef6ac4SDon Lewis struct sockaddr_in *sin; 136726ef6ac4SDon Lewis 13681ede983cSDag-Erling Smørgrav sin = malloc(sizeof *sin, M_SONAME, 1369a163d034SWarner Losh M_WAITOK | M_ZERO); 137026ef6ac4SDon Lewis sin->sin_family = AF_INET; 137126ef6ac4SDon Lewis sin->sin_len = sizeof(*sin); 137226ef6ac4SDon Lewis sin->sin_addr = *addr_p; 137326ef6ac4SDon Lewis sin->sin_port = port; 137426ef6ac4SDon Lewis 137526ef6ac4SDon Lewis return (struct sockaddr *)sin; 137626ef6ac4SDon Lewis } 137726ef6ac4SDon Lewis 1378117bcae7SGarrett Wollman int 137954d642bbSRobert Watson in_getsockaddr(struct socket *so, struct sockaddr **nam) 1380df8bae1dSRodney W. Grimes { 1381136d4f1cSRobert Watson struct inpcb *inp; 138226ef6ac4SDon Lewis struct in_addr addr; 138326ef6ac4SDon Lewis in_port_t port; 138442fa505bSDavid Greenman 1385fdc984f7STor Egge inp = sotoinpcb(so); 138654d642bbSRobert Watson KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL")); 13876466b28aSRobert Watson 1388a69042a5SRobert Watson INP_RLOCK(inp); 138926ef6ac4SDon Lewis port = inp->inp_lport; 139026ef6ac4SDon Lewis addr = inp->inp_laddr; 1391a69042a5SRobert Watson INP_RUNLOCK(inp); 139242fa505bSDavid Greenman 139326ef6ac4SDon Lewis *nam = in_sockaddr(port, &addr); 1394117bcae7SGarrett Wollman return 0; 1395df8bae1dSRodney W. Grimes } 1396df8bae1dSRodney W. Grimes 1397117bcae7SGarrett Wollman int 139854d642bbSRobert Watson in_getpeeraddr(struct socket *so, struct sockaddr **nam) 1399df8bae1dSRodney W. Grimes { 1400136d4f1cSRobert Watson struct inpcb *inp; 140126ef6ac4SDon Lewis struct in_addr addr; 140226ef6ac4SDon Lewis in_port_t port; 140342fa505bSDavid Greenman 1404fdc984f7STor Egge inp = sotoinpcb(so); 140554d642bbSRobert Watson KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL")); 14066466b28aSRobert Watson 1407a69042a5SRobert Watson INP_RLOCK(inp); 140826ef6ac4SDon Lewis port = inp->inp_fport; 140926ef6ac4SDon Lewis addr = inp->inp_faddr; 1410a69042a5SRobert Watson INP_RUNLOCK(inp); 141142fa505bSDavid Greenman 141226ef6ac4SDon Lewis *nam = in_sockaddr(port, &addr); 1413117bcae7SGarrett Wollman return 0; 1414df8bae1dSRodney W. Grimes } 1415df8bae1dSRodney W. Grimes 141626f9a767SRodney W. Grimes void 1417136d4f1cSRobert Watson in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno, 1418136d4f1cSRobert Watson struct inpcb *(*notify)(struct inpcb *, int)) 1419d1c54148SJesper Skriver { 1420f457d580SRobert Watson struct inpcb *inp, *inp_temp; 1421d1c54148SJesper Skriver 14223dc7ebf9SJeffrey Hsu INP_INFO_WLOCK(pcbinfo); 1423f457d580SRobert Watson LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) { 14248501a69cSRobert Watson INP_WLOCK(inp); 1425d1c54148SJesper Skriver #ifdef INET6 1426f76fcf6dSJeffrey Hsu if ((inp->inp_vflag & INP_IPV4) == 0) { 14278501a69cSRobert Watson INP_WUNLOCK(inp); 1428d1c54148SJesper Skriver continue; 1429f76fcf6dSJeffrey Hsu } 1430d1c54148SJesper Skriver #endif 1431d1c54148SJesper Skriver if (inp->inp_faddr.s_addr != faddr.s_addr || 1432f76fcf6dSJeffrey Hsu inp->inp_socket == NULL) { 14338501a69cSRobert Watson INP_WUNLOCK(inp); 1434d1c54148SJesper Skriver continue; 1435d1c54148SJesper Skriver } 14363dc7ebf9SJeffrey Hsu if ((*notify)(inp, errno)) 14378501a69cSRobert Watson INP_WUNLOCK(inp); 1438f76fcf6dSJeffrey Hsu } 14393dc7ebf9SJeffrey Hsu INP_INFO_WUNLOCK(pcbinfo); 1440d1c54148SJesper Skriver } 1441d1c54148SJesper Skriver 1442e43cc4aeSHajimu UMEMOTO void 1443136d4f1cSRobert Watson in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp) 1444e43cc4aeSHajimu UMEMOTO { 1445e43cc4aeSHajimu UMEMOTO struct inpcb *inp; 1446e43cc4aeSHajimu UMEMOTO struct ip_moptions *imo; 1447e43cc4aeSHajimu UMEMOTO int i, gap; 1448e43cc4aeSHajimu UMEMOTO 1449ff9b006dSJulien Charbon INP_INFO_WLOCK(pcbinfo); 1450712fc218SRobert Watson LIST_FOREACH(inp, pcbinfo->ipi_listhead, inp_list) { 14518501a69cSRobert Watson INP_WLOCK(inp); 1452e43cc4aeSHajimu UMEMOTO imo = inp->inp_moptions; 1453e43cc4aeSHajimu UMEMOTO if ((inp->inp_vflag & INP_IPV4) && 1454e43cc4aeSHajimu UMEMOTO imo != NULL) { 1455e43cc4aeSHajimu UMEMOTO /* 1456e43cc4aeSHajimu UMEMOTO * Unselect the outgoing interface if it is being 1457e43cc4aeSHajimu UMEMOTO * detached. 1458e43cc4aeSHajimu UMEMOTO */ 1459e43cc4aeSHajimu UMEMOTO if (imo->imo_multicast_ifp == ifp) 1460e43cc4aeSHajimu UMEMOTO imo->imo_multicast_ifp = NULL; 1461e43cc4aeSHajimu UMEMOTO 1462e43cc4aeSHajimu UMEMOTO /* 1463e43cc4aeSHajimu UMEMOTO * Drop multicast group membership if we joined 1464e43cc4aeSHajimu UMEMOTO * through the interface being detached. 1465e43cc4aeSHajimu UMEMOTO */ 1466e43cc4aeSHajimu UMEMOTO for (i = 0, gap = 0; i < imo->imo_num_memberships; 1467e43cc4aeSHajimu UMEMOTO i++) { 1468e43cc4aeSHajimu UMEMOTO if (imo->imo_membership[i]->inm_ifp == ifp) { 1469e43cc4aeSHajimu UMEMOTO in_delmulti(imo->imo_membership[i]); 1470e43cc4aeSHajimu UMEMOTO gap++; 1471e43cc4aeSHajimu UMEMOTO } else if (gap != 0) 1472e43cc4aeSHajimu UMEMOTO imo->imo_membership[i - gap] = 1473e43cc4aeSHajimu UMEMOTO imo->imo_membership[i]; 1474e43cc4aeSHajimu UMEMOTO } 1475e43cc4aeSHajimu UMEMOTO imo->imo_num_memberships -= gap; 1476e43cc4aeSHajimu UMEMOTO } 14778501a69cSRobert Watson INP_WUNLOCK(inp); 1478e43cc4aeSHajimu UMEMOTO } 1479ff9b006dSJulien Charbon INP_INFO_WUNLOCK(pcbinfo); 1480e43cc4aeSHajimu UMEMOTO } 1481e43cc4aeSHajimu UMEMOTO 1482df8bae1dSRodney W. Grimes /* 1483fa046d87SRobert Watson * Lookup a PCB based on the local address and port. Caller must hold the 1484fa046d87SRobert Watson * hash lock. No inpcb locks or references are acquired. 1485c3229e05SDavid Greenman */ 1486d5e8a67eSHajimu UMEMOTO #define INP_LOOKUP_MAPPED_PCB_COST 3 1487df8bae1dSRodney W. Grimes struct inpcb * 1488136d4f1cSRobert Watson in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr, 148968e0d7e0SRobert Watson u_short lport, int lookupflags, struct ucred *cred) 1490df8bae1dSRodney W. Grimes { 1491136d4f1cSRobert Watson struct inpcb *inp; 1492d5e8a67eSHajimu UMEMOTO #ifdef INET6 1493d5e8a67eSHajimu UMEMOTO int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST; 1494d5e8a67eSHajimu UMEMOTO #else 1495d5e8a67eSHajimu UMEMOTO int matchwild = 3; 1496d5e8a67eSHajimu UMEMOTO #endif 1497d5e8a67eSHajimu UMEMOTO int wildcard; 14987bc4aca7SDavid Greenman 149968e0d7e0SRobert Watson KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0, 150068e0d7e0SRobert Watson ("%s: invalid lookup flags %d", __func__, lookupflags)); 150168e0d7e0SRobert Watson 1502fa046d87SRobert Watson INP_HASH_LOCK_ASSERT(pcbinfo); 15031b73ca0bSSam Leffler 150468e0d7e0SRobert Watson if ((lookupflags & INPLOOKUP_WILDCARD) == 0) { 1505c3229e05SDavid Greenman struct inpcbhead *head; 1506c3229e05SDavid Greenman /* 1507c3229e05SDavid Greenman * Look for an unconnected (wildcard foreign addr) PCB that 1508c3229e05SDavid Greenman * matches the local address and port we're looking for. 1509c3229e05SDavid Greenman */ 1510712fc218SRobert Watson head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, 1511712fc218SRobert Watson 0, pcbinfo->ipi_hashmask)]; 1512fc2ffbe6SPoul-Henning Kamp LIST_FOREACH(inp, head, inp_hash) { 1513cfa1ca9dSYoshinobu Inoue #ifdef INET6 1514413628a7SBjoern A. Zeeb /* XXX inp locking */ 1515369dc8ceSEivind Eklund if ((inp->inp_vflag & INP_IPV4) == 0) 1516cfa1ca9dSYoshinobu Inoue continue; 1517cfa1ca9dSYoshinobu Inoue #endif 1518c3229e05SDavid Greenman if (inp->inp_faddr.s_addr == INADDR_ANY && 1519c3229e05SDavid Greenman inp->inp_laddr.s_addr == laddr.s_addr && 1520c3229e05SDavid Greenman inp->inp_lport == lport) { 1521c3229e05SDavid Greenman /* 1522413628a7SBjoern A. Zeeb * Found? 1523c3229e05SDavid Greenman */ 1524413628a7SBjoern A. Zeeb if (cred == NULL || 15250304c731SJamie Gritton prison_equal_ip4(cred->cr_prison, 15260304c731SJamie Gritton inp->inp_cred->cr_prison)) 1527c3229e05SDavid Greenman return (inp); 1528df8bae1dSRodney W. Grimes } 1529c3229e05SDavid Greenman } 1530c3229e05SDavid Greenman /* 1531c3229e05SDavid Greenman * Not found. 1532c3229e05SDavid Greenman */ 1533c3229e05SDavid Greenman return (NULL); 1534c3229e05SDavid Greenman } else { 1535c3229e05SDavid Greenman struct inpcbporthead *porthash; 1536c3229e05SDavid Greenman struct inpcbport *phd; 1537c3229e05SDavid Greenman struct inpcb *match = NULL; 1538c3229e05SDavid Greenman /* 1539c3229e05SDavid Greenman * Best fit PCB lookup. 1540c3229e05SDavid Greenman * 1541c3229e05SDavid Greenman * First see if this local port is in use by looking on the 1542c3229e05SDavid Greenman * port hash list. 1543c3229e05SDavid Greenman */ 1544712fc218SRobert Watson porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport, 1545712fc218SRobert Watson pcbinfo->ipi_porthashmask)]; 1546fc2ffbe6SPoul-Henning Kamp LIST_FOREACH(phd, porthash, phd_hash) { 1547c3229e05SDavid Greenman if (phd->phd_port == lport) 1548c3229e05SDavid Greenman break; 1549c3229e05SDavid Greenman } 1550c3229e05SDavid Greenman if (phd != NULL) { 1551c3229e05SDavid Greenman /* 1552c3229e05SDavid Greenman * Port is in use by one or more PCBs. Look for best 1553c3229e05SDavid Greenman * fit. 1554c3229e05SDavid Greenman */ 155537d40066SPoul-Henning Kamp LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) { 1556c3229e05SDavid Greenman wildcard = 0; 1557413628a7SBjoern A. Zeeb if (cred != NULL && 15580304c731SJamie Gritton !prison_equal_ip4(inp->inp_cred->cr_prison, 15590304c731SJamie Gritton cred->cr_prison)) 1560413628a7SBjoern A. Zeeb continue; 1561cfa1ca9dSYoshinobu Inoue #ifdef INET6 1562413628a7SBjoern A. Zeeb /* XXX inp locking */ 1563369dc8ceSEivind Eklund if ((inp->inp_vflag & INP_IPV4) == 0) 1564cfa1ca9dSYoshinobu Inoue continue; 1565d5e8a67eSHajimu UMEMOTO /* 1566d5e8a67eSHajimu UMEMOTO * We never select the PCB that has 1567d5e8a67eSHajimu UMEMOTO * INP_IPV6 flag and is bound to :: if 1568d5e8a67eSHajimu UMEMOTO * we have another PCB which is bound 1569d5e8a67eSHajimu UMEMOTO * to 0.0.0.0. If a PCB has the 1570d5e8a67eSHajimu UMEMOTO * INP_IPV6 flag, then we set its cost 1571d5e8a67eSHajimu UMEMOTO * higher than IPv4 only PCBs. 1572d5e8a67eSHajimu UMEMOTO * 1573d5e8a67eSHajimu UMEMOTO * Note that the case only happens 1574d5e8a67eSHajimu UMEMOTO * when a socket is bound to ::, under 1575d5e8a67eSHajimu UMEMOTO * the condition that the use of the 1576d5e8a67eSHajimu UMEMOTO * mapped address is allowed. 1577d5e8a67eSHajimu UMEMOTO */ 1578d5e8a67eSHajimu UMEMOTO if ((inp->inp_vflag & INP_IPV6) != 0) 1579d5e8a67eSHajimu UMEMOTO wildcard += INP_LOOKUP_MAPPED_PCB_COST; 1580cfa1ca9dSYoshinobu Inoue #endif 1581c3229e05SDavid Greenman if (inp->inp_faddr.s_addr != INADDR_ANY) 1582c3229e05SDavid Greenman wildcard++; 158315bd2b43SDavid Greenman if (inp->inp_laddr.s_addr != INADDR_ANY) { 158415bd2b43SDavid Greenman if (laddr.s_addr == INADDR_ANY) 158515bd2b43SDavid Greenman wildcard++; 158615bd2b43SDavid Greenman else if (inp->inp_laddr.s_addr != laddr.s_addr) 158715bd2b43SDavid Greenman continue; 158815bd2b43SDavid Greenman } else { 158915bd2b43SDavid Greenman if (laddr.s_addr != INADDR_ANY) 159015bd2b43SDavid Greenman wildcard++; 159115bd2b43SDavid Greenman } 1592df8bae1dSRodney W. Grimes if (wildcard < matchwild) { 1593df8bae1dSRodney W. Grimes match = inp; 1594df8bae1dSRodney W. Grimes matchwild = wildcard; 1595413628a7SBjoern A. Zeeb if (matchwild == 0) 1596df8bae1dSRodney W. Grimes break; 1597df8bae1dSRodney W. Grimes } 1598df8bae1dSRodney W. Grimes } 15993dbdc25cSDavid Greenman } 1600df8bae1dSRodney W. Grimes return (match); 1601df8bae1dSRodney W. Grimes } 1602c3229e05SDavid Greenman } 1603d5e8a67eSHajimu UMEMOTO #undef INP_LOOKUP_MAPPED_PCB_COST 160415bd2b43SDavid Greenman 160552cd27cbSRobert Watson #ifdef PCBGROUP 160652cd27cbSRobert Watson /* 160752cd27cbSRobert Watson * Lookup PCB in hash list, using pcbgroup tables. 160852cd27cbSRobert Watson */ 160952cd27cbSRobert Watson static struct inpcb * 161052cd27cbSRobert Watson in_pcblookup_group(struct inpcbinfo *pcbinfo, struct inpcbgroup *pcbgroup, 161152cd27cbSRobert Watson struct in_addr faddr, u_int fport_arg, struct in_addr laddr, 161252cd27cbSRobert Watson u_int lport_arg, int lookupflags, struct ifnet *ifp) 161352cd27cbSRobert Watson { 161452cd27cbSRobert Watson struct inpcbhead *head; 161552cd27cbSRobert Watson struct inpcb *inp, *tmpinp; 161652cd27cbSRobert Watson u_short fport = fport_arg, lport = lport_arg; 161752cd27cbSRobert Watson 161852cd27cbSRobert Watson /* 161952cd27cbSRobert Watson * First look for an exact match. 162052cd27cbSRobert Watson */ 162152cd27cbSRobert Watson tmpinp = NULL; 162252cd27cbSRobert Watson INP_GROUP_LOCK(pcbgroup); 162352cd27cbSRobert Watson head = &pcbgroup->ipg_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport, 162452cd27cbSRobert Watson pcbgroup->ipg_hashmask)]; 162552cd27cbSRobert Watson LIST_FOREACH(inp, head, inp_pcbgrouphash) { 162652cd27cbSRobert Watson #ifdef INET6 162752cd27cbSRobert Watson /* XXX inp locking */ 162852cd27cbSRobert Watson if ((inp->inp_vflag & INP_IPV4) == 0) 162952cd27cbSRobert Watson continue; 163052cd27cbSRobert Watson #endif 163152cd27cbSRobert Watson if (inp->inp_faddr.s_addr == faddr.s_addr && 163252cd27cbSRobert Watson inp->inp_laddr.s_addr == laddr.s_addr && 163352cd27cbSRobert Watson inp->inp_fport == fport && 163452cd27cbSRobert Watson inp->inp_lport == lport) { 163552cd27cbSRobert Watson /* 163652cd27cbSRobert Watson * XXX We should be able to directly return 163752cd27cbSRobert Watson * the inp here, without any checks. 163852cd27cbSRobert Watson * Well unless both bound with SO_REUSEPORT? 163952cd27cbSRobert Watson */ 164052cd27cbSRobert Watson if (prison_flag(inp->inp_cred, PR_IP4)) 164152cd27cbSRobert Watson goto found; 164252cd27cbSRobert Watson if (tmpinp == NULL) 164352cd27cbSRobert Watson tmpinp = inp; 164452cd27cbSRobert Watson } 164552cd27cbSRobert Watson } 164652cd27cbSRobert Watson if (tmpinp != NULL) { 164752cd27cbSRobert Watson inp = tmpinp; 164852cd27cbSRobert Watson goto found; 164952cd27cbSRobert Watson } 165052cd27cbSRobert Watson 16510a100a6fSAdrian Chadd #ifdef RSS 16520a100a6fSAdrian Chadd /* 16530a100a6fSAdrian Chadd * For incoming connections, we may wish to do a wildcard 16540a100a6fSAdrian Chadd * match for an RSS-local socket. 16550a100a6fSAdrian Chadd */ 16560a100a6fSAdrian Chadd if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { 16570a100a6fSAdrian Chadd struct inpcb *local_wild = NULL, *local_exact = NULL; 16580a100a6fSAdrian Chadd #ifdef INET6 16590a100a6fSAdrian Chadd struct inpcb *local_wild_mapped = NULL; 16600a100a6fSAdrian Chadd #endif 16610a100a6fSAdrian Chadd struct inpcb *jail_wild = NULL; 16620a100a6fSAdrian Chadd struct inpcbhead *head; 16630a100a6fSAdrian Chadd int injail; 16640a100a6fSAdrian Chadd 16650a100a6fSAdrian Chadd /* 16660a100a6fSAdrian Chadd * Order of socket selection - we always prefer jails. 16670a100a6fSAdrian Chadd * 1. jailed, non-wild. 16680a100a6fSAdrian Chadd * 2. jailed, wild. 16690a100a6fSAdrian Chadd * 3. non-jailed, non-wild. 16700a100a6fSAdrian Chadd * 4. non-jailed, wild. 16710a100a6fSAdrian Chadd */ 16720a100a6fSAdrian Chadd 16730a100a6fSAdrian Chadd head = &pcbgroup->ipg_hashbase[INP_PCBHASH(INADDR_ANY, 16740a100a6fSAdrian Chadd lport, 0, pcbgroup->ipg_hashmask)]; 16750a100a6fSAdrian Chadd LIST_FOREACH(inp, head, inp_pcbgrouphash) { 16760a100a6fSAdrian Chadd #ifdef INET6 16770a100a6fSAdrian Chadd /* XXX inp locking */ 16780a100a6fSAdrian Chadd if ((inp->inp_vflag & INP_IPV4) == 0) 16790a100a6fSAdrian Chadd continue; 16800a100a6fSAdrian Chadd #endif 16810a100a6fSAdrian Chadd if (inp->inp_faddr.s_addr != INADDR_ANY || 16820a100a6fSAdrian Chadd inp->inp_lport != lport) 16830a100a6fSAdrian Chadd continue; 16840a100a6fSAdrian Chadd 16850a100a6fSAdrian Chadd injail = prison_flag(inp->inp_cred, PR_IP4); 16860a100a6fSAdrian Chadd if (injail) { 16870a100a6fSAdrian Chadd if (prison_check_ip4(inp->inp_cred, 16880a100a6fSAdrian Chadd &laddr) != 0) 16890a100a6fSAdrian Chadd continue; 16900a100a6fSAdrian Chadd } else { 16910a100a6fSAdrian Chadd if (local_exact != NULL) 16920a100a6fSAdrian Chadd continue; 16930a100a6fSAdrian Chadd } 16940a100a6fSAdrian Chadd 16950a100a6fSAdrian Chadd if (inp->inp_laddr.s_addr == laddr.s_addr) { 16960a100a6fSAdrian Chadd if (injail) 16970a100a6fSAdrian Chadd goto found; 16980a100a6fSAdrian Chadd else 16990a100a6fSAdrian Chadd local_exact = inp; 17000a100a6fSAdrian Chadd } else if (inp->inp_laddr.s_addr == INADDR_ANY) { 17010a100a6fSAdrian Chadd #ifdef INET6 17020a100a6fSAdrian Chadd /* XXX inp locking, NULL check */ 17030a100a6fSAdrian Chadd if (inp->inp_vflag & INP_IPV6PROTO) 17040a100a6fSAdrian Chadd local_wild_mapped = inp; 17050a100a6fSAdrian Chadd else 17060a100a6fSAdrian Chadd #endif 17070a100a6fSAdrian Chadd if (injail) 17080a100a6fSAdrian Chadd jail_wild = inp; 17090a100a6fSAdrian Chadd else 17100a100a6fSAdrian Chadd local_wild = inp; 17110a100a6fSAdrian Chadd } 17120a100a6fSAdrian Chadd } /* LIST_FOREACH */ 17130a100a6fSAdrian Chadd 17140a100a6fSAdrian Chadd inp = jail_wild; 17150a100a6fSAdrian Chadd if (inp == NULL) 17160a100a6fSAdrian Chadd inp = local_exact; 17170a100a6fSAdrian Chadd if (inp == NULL) 17180a100a6fSAdrian Chadd inp = local_wild; 17190a100a6fSAdrian Chadd #ifdef INET6 17200a100a6fSAdrian Chadd if (inp == NULL) 17210a100a6fSAdrian Chadd inp = local_wild_mapped; 17220a100a6fSAdrian Chadd #endif 17230a100a6fSAdrian Chadd if (inp != NULL) 17240a100a6fSAdrian Chadd goto found; 17250a100a6fSAdrian Chadd } 17260a100a6fSAdrian Chadd #endif 17270a100a6fSAdrian Chadd 172852cd27cbSRobert Watson /* 172952cd27cbSRobert Watson * Then look for a wildcard match, if requested. 173052cd27cbSRobert Watson */ 173152cd27cbSRobert Watson if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { 173252cd27cbSRobert Watson struct inpcb *local_wild = NULL, *local_exact = NULL; 173352cd27cbSRobert Watson #ifdef INET6 173452cd27cbSRobert Watson struct inpcb *local_wild_mapped = NULL; 173552cd27cbSRobert Watson #endif 173652cd27cbSRobert Watson struct inpcb *jail_wild = NULL; 173752cd27cbSRobert Watson struct inpcbhead *head; 173852cd27cbSRobert Watson int injail; 173952cd27cbSRobert Watson 174052cd27cbSRobert Watson /* 174152cd27cbSRobert Watson * Order of socket selection - we always prefer jails. 174252cd27cbSRobert Watson * 1. jailed, non-wild. 174352cd27cbSRobert Watson * 2. jailed, wild. 174452cd27cbSRobert Watson * 3. non-jailed, non-wild. 174552cd27cbSRobert Watson * 4. non-jailed, wild. 174652cd27cbSRobert Watson */ 174752cd27cbSRobert Watson head = &pcbinfo->ipi_wildbase[INP_PCBHASH(INADDR_ANY, lport, 174852cd27cbSRobert Watson 0, pcbinfo->ipi_wildmask)]; 174952cd27cbSRobert Watson LIST_FOREACH(inp, head, inp_pcbgroup_wild) { 175052cd27cbSRobert Watson #ifdef INET6 175152cd27cbSRobert Watson /* XXX inp locking */ 175252cd27cbSRobert Watson if ((inp->inp_vflag & INP_IPV4) == 0) 175352cd27cbSRobert Watson continue; 175452cd27cbSRobert Watson #endif 175552cd27cbSRobert Watson if (inp->inp_faddr.s_addr != INADDR_ANY || 175652cd27cbSRobert Watson inp->inp_lport != lport) 175752cd27cbSRobert Watson continue; 175852cd27cbSRobert Watson 175952cd27cbSRobert Watson injail = prison_flag(inp->inp_cred, PR_IP4); 176052cd27cbSRobert Watson if (injail) { 176152cd27cbSRobert Watson if (prison_check_ip4(inp->inp_cred, 176252cd27cbSRobert Watson &laddr) != 0) 176352cd27cbSRobert Watson continue; 176452cd27cbSRobert Watson } else { 176552cd27cbSRobert Watson if (local_exact != NULL) 176652cd27cbSRobert Watson continue; 176752cd27cbSRobert Watson } 176852cd27cbSRobert Watson 176952cd27cbSRobert Watson if (inp->inp_laddr.s_addr == laddr.s_addr) { 177052cd27cbSRobert Watson if (injail) 177152cd27cbSRobert Watson goto found; 177252cd27cbSRobert Watson else 177352cd27cbSRobert Watson local_exact = inp; 177452cd27cbSRobert Watson } else if (inp->inp_laddr.s_addr == INADDR_ANY) { 177552cd27cbSRobert Watson #ifdef INET6 177652cd27cbSRobert Watson /* XXX inp locking, NULL check */ 177752cd27cbSRobert Watson if (inp->inp_vflag & INP_IPV6PROTO) 177852cd27cbSRobert Watson local_wild_mapped = inp; 177952cd27cbSRobert Watson else 178083e521ecSBjoern A. Zeeb #endif 178152cd27cbSRobert Watson if (injail) 178252cd27cbSRobert Watson jail_wild = inp; 178352cd27cbSRobert Watson else 178452cd27cbSRobert Watson local_wild = inp; 178552cd27cbSRobert Watson } 178652cd27cbSRobert Watson } /* LIST_FOREACH */ 178752cd27cbSRobert Watson inp = jail_wild; 178852cd27cbSRobert Watson if (inp == NULL) 178952cd27cbSRobert Watson inp = local_exact; 179052cd27cbSRobert Watson if (inp == NULL) 179152cd27cbSRobert Watson inp = local_wild; 179252cd27cbSRobert Watson #ifdef INET6 179352cd27cbSRobert Watson if (inp == NULL) 179452cd27cbSRobert Watson inp = local_wild_mapped; 179583e521ecSBjoern A. Zeeb #endif 179652cd27cbSRobert Watson if (inp != NULL) 179752cd27cbSRobert Watson goto found; 179852cd27cbSRobert Watson } /* if (lookupflags & INPLOOKUP_WILDCARD) */ 179952cd27cbSRobert Watson INP_GROUP_UNLOCK(pcbgroup); 180052cd27cbSRobert Watson return (NULL); 180152cd27cbSRobert Watson 180252cd27cbSRobert Watson found: 180352cd27cbSRobert Watson in_pcbref(inp); 180452cd27cbSRobert Watson INP_GROUP_UNLOCK(pcbgroup); 180552cd27cbSRobert Watson if (lookupflags & INPLOOKUP_WLOCKPCB) { 180652cd27cbSRobert Watson INP_WLOCK(inp); 180752cd27cbSRobert Watson if (in_pcbrele_wlocked(inp)) 180852cd27cbSRobert Watson return (NULL); 180952cd27cbSRobert Watson } else if (lookupflags & INPLOOKUP_RLOCKPCB) { 181052cd27cbSRobert Watson INP_RLOCK(inp); 181152cd27cbSRobert Watson if (in_pcbrele_rlocked(inp)) 181252cd27cbSRobert Watson return (NULL); 181352cd27cbSRobert Watson } else 181452cd27cbSRobert Watson panic("%s: locking bug", __func__); 181552cd27cbSRobert Watson return (inp); 181652cd27cbSRobert Watson } 181752cd27cbSRobert Watson #endif /* PCBGROUP */ 181852cd27cbSRobert Watson 181915bd2b43SDavid Greenman /* 1820fa046d87SRobert Watson * Lookup PCB in hash list, using pcbinfo tables. This variation assumes 1821fa046d87SRobert Watson * that the caller has locked the hash list, and will not perform any further 1822fa046d87SRobert Watson * locking or reference operations on either the hash list or the connection. 182315bd2b43SDavid Greenman */ 1824fa046d87SRobert Watson static struct inpcb * 1825fa046d87SRobert Watson in_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in_addr faddr, 182668e0d7e0SRobert Watson u_int fport_arg, struct in_addr laddr, u_int lport_arg, int lookupflags, 1827136d4f1cSRobert Watson struct ifnet *ifp) 182815bd2b43SDavid Greenman { 182915bd2b43SDavid Greenman struct inpcbhead *head; 1830413628a7SBjoern A. Zeeb struct inpcb *inp, *tmpinp; 183115bd2b43SDavid Greenman u_short fport = fport_arg, lport = lport_arg; 183215bd2b43SDavid Greenman 183368e0d7e0SRobert Watson KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0, 183468e0d7e0SRobert Watson ("%s: invalid lookup flags %d", __func__, lookupflags)); 183568e0d7e0SRobert Watson 1836fa046d87SRobert Watson INP_HASH_LOCK_ASSERT(pcbinfo); 1837602cc7f1SRobert Watson 183815bd2b43SDavid Greenman /* 183915bd2b43SDavid Greenman * First look for an exact match. 184015bd2b43SDavid Greenman */ 1841413628a7SBjoern A. Zeeb tmpinp = NULL; 1842712fc218SRobert Watson head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr.s_addr, lport, fport, 1843712fc218SRobert Watson pcbinfo->ipi_hashmask)]; 1844fc2ffbe6SPoul-Henning Kamp LIST_FOREACH(inp, head, inp_hash) { 1845cfa1ca9dSYoshinobu Inoue #ifdef INET6 1846413628a7SBjoern A. Zeeb /* XXX inp locking */ 1847369dc8ceSEivind Eklund if ((inp->inp_vflag & INP_IPV4) == 0) 1848cfa1ca9dSYoshinobu Inoue continue; 1849cfa1ca9dSYoshinobu Inoue #endif 18506d6a026bSDavid Greenman if (inp->inp_faddr.s_addr == faddr.s_addr && 1851ca98b82cSDavid Greenman inp->inp_laddr.s_addr == laddr.s_addr && 1852ca98b82cSDavid Greenman inp->inp_fport == fport && 1853413628a7SBjoern A. Zeeb inp->inp_lport == lport) { 1854413628a7SBjoern A. Zeeb /* 1855413628a7SBjoern A. Zeeb * XXX We should be able to directly return 1856413628a7SBjoern A. Zeeb * the inp here, without any checks. 1857413628a7SBjoern A. Zeeb * Well unless both bound with SO_REUSEPORT? 1858413628a7SBjoern A. Zeeb */ 18590304c731SJamie Gritton if (prison_flag(inp->inp_cred, PR_IP4)) 1860c3229e05SDavid Greenman return (inp); 1861413628a7SBjoern A. Zeeb if (tmpinp == NULL) 1862413628a7SBjoern A. Zeeb tmpinp = inp; 1863c3229e05SDavid Greenman } 1864413628a7SBjoern A. Zeeb } 1865413628a7SBjoern A. Zeeb if (tmpinp != NULL) 1866413628a7SBjoern A. Zeeb return (tmpinp); 1867e3fd5ffdSRobert Watson 1868e3fd5ffdSRobert Watson /* 1869e3fd5ffdSRobert Watson * Then look for a wildcard match, if requested. 1870e3fd5ffdSRobert Watson */ 187168e0d7e0SRobert Watson if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { 1872413628a7SBjoern A. Zeeb struct inpcb *local_wild = NULL, *local_exact = NULL; 1873e3fd5ffdSRobert Watson #ifdef INET6 1874cfa1ca9dSYoshinobu Inoue struct inpcb *local_wild_mapped = NULL; 1875e3fd5ffdSRobert Watson #endif 1876413628a7SBjoern A. Zeeb struct inpcb *jail_wild = NULL; 1877413628a7SBjoern A. Zeeb int injail; 1878413628a7SBjoern A. Zeeb 1879413628a7SBjoern A. Zeeb /* 1880413628a7SBjoern A. Zeeb * Order of socket selection - we always prefer jails. 1881413628a7SBjoern A. Zeeb * 1. jailed, non-wild. 1882413628a7SBjoern A. Zeeb * 2. jailed, wild. 1883413628a7SBjoern A. Zeeb * 3. non-jailed, non-wild. 1884413628a7SBjoern A. Zeeb * 4. non-jailed, wild. 1885413628a7SBjoern A. Zeeb */ 18866d6a026bSDavid Greenman 1887712fc218SRobert Watson head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, 1888712fc218SRobert Watson 0, pcbinfo->ipi_hashmask)]; 1889fc2ffbe6SPoul-Henning Kamp LIST_FOREACH(inp, head, inp_hash) { 1890cfa1ca9dSYoshinobu Inoue #ifdef INET6 1891413628a7SBjoern A. Zeeb /* XXX inp locking */ 1892369dc8ceSEivind Eklund if ((inp->inp_vflag & INP_IPV4) == 0) 1893cfa1ca9dSYoshinobu Inoue continue; 1894cfa1ca9dSYoshinobu Inoue #endif 1895413628a7SBjoern A. Zeeb if (inp->inp_faddr.s_addr != INADDR_ANY || 1896413628a7SBjoern A. Zeeb inp->inp_lport != lport) 1897413628a7SBjoern A. Zeeb continue; 1898413628a7SBjoern A. Zeeb 18990304c731SJamie Gritton injail = prison_flag(inp->inp_cred, PR_IP4); 1900413628a7SBjoern A. Zeeb if (injail) { 1901b89e82ddSJamie Gritton if (prison_check_ip4(inp->inp_cred, 1902b89e82ddSJamie Gritton &laddr) != 0) 1903413628a7SBjoern A. Zeeb continue; 1904413628a7SBjoern A. Zeeb } else { 1905413628a7SBjoern A. Zeeb if (local_exact != NULL) 1906413628a7SBjoern A. Zeeb continue; 1907413628a7SBjoern A. Zeeb } 1908413628a7SBjoern A. Zeeb 1909413628a7SBjoern A. Zeeb if (inp->inp_laddr.s_addr == laddr.s_addr) { 1910413628a7SBjoern A. Zeeb if (injail) 1911c3229e05SDavid Greenman return (inp); 1912413628a7SBjoern A. Zeeb else 1913413628a7SBjoern A. Zeeb local_exact = inp; 1914413628a7SBjoern A. Zeeb } else if (inp->inp_laddr.s_addr == INADDR_ANY) { 1915e3fd5ffdSRobert Watson #ifdef INET6 1916413628a7SBjoern A. Zeeb /* XXX inp locking, NULL check */ 19175cd54324SBjoern A. Zeeb if (inp->inp_vflag & INP_IPV6PROTO) 1918cfa1ca9dSYoshinobu Inoue local_wild_mapped = inp; 1919cfa1ca9dSYoshinobu Inoue else 192083e521ecSBjoern A. Zeeb #endif 1921413628a7SBjoern A. Zeeb if (injail) 1922413628a7SBjoern A. Zeeb jail_wild = inp; 1923413628a7SBjoern A. Zeeb else 19246d6a026bSDavid Greenman local_wild = inp; 19256d6a026bSDavid Greenman } 1926413628a7SBjoern A. Zeeb } /* LIST_FOREACH */ 1927413628a7SBjoern A. Zeeb if (jail_wild != NULL) 1928413628a7SBjoern A. Zeeb return (jail_wild); 1929413628a7SBjoern A. Zeeb if (local_exact != NULL) 1930413628a7SBjoern A. Zeeb return (local_exact); 1931413628a7SBjoern A. Zeeb if (local_wild != NULL) 1932c3229e05SDavid Greenman return (local_wild); 1933413628a7SBjoern A. Zeeb #ifdef INET6 1934413628a7SBjoern A. Zeeb if (local_wild_mapped != NULL) 1935413628a7SBjoern A. Zeeb return (local_wild_mapped); 193683e521ecSBjoern A. Zeeb #endif 193768e0d7e0SRobert Watson } /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */ 1938413628a7SBjoern A. Zeeb 19396d6a026bSDavid Greenman return (NULL); 194015bd2b43SDavid Greenman } 1941fa046d87SRobert Watson 1942fa046d87SRobert Watson /* 1943fa046d87SRobert Watson * Lookup PCB in hash list, using pcbinfo tables. This variation locks the 1944fa046d87SRobert Watson * hash list lock, and will return the inpcb locked (i.e., requires 1945fa046d87SRobert Watson * INPLOOKUP_LOCKPCB). 1946fa046d87SRobert Watson */ 1947fa046d87SRobert Watson static struct inpcb * 1948fa046d87SRobert Watson in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr, 1949fa046d87SRobert Watson u_int fport, struct in_addr laddr, u_int lport, int lookupflags, 1950fa046d87SRobert Watson struct ifnet *ifp) 1951fa046d87SRobert Watson { 1952fa046d87SRobert Watson struct inpcb *inp; 1953fa046d87SRobert Watson 1954fa046d87SRobert Watson INP_HASH_RLOCK(pcbinfo); 1955fa046d87SRobert Watson inp = in_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport, 1956fa046d87SRobert Watson (lookupflags & ~(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)), ifp); 1957fa046d87SRobert Watson if (inp != NULL) { 1958fa046d87SRobert Watson in_pcbref(inp); 1959fa046d87SRobert Watson INP_HASH_RUNLOCK(pcbinfo); 1960fa046d87SRobert Watson if (lookupflags & INPLOOKUP_WLOCKPCB) { 1961fa046d87SRobert Watson INP_WLOCK(inp); 1962fa046d87SRobert Watson if (in_pcbrele_wlocked(inp)) 1963fa046d87SRobert Watson return (NULL); 1964fa046d87SRobert Watson } else if (lookupflags & INPLOOKUP_RLOCKPCB) { 1965fa046d87SRobert Watson INP_RLOCK(inp); 1966fa046d87SRobert Watson if (in_pcbrele_rlocked(inp)) 1967fa046d87SRobert Watson return (NULL); 1968fa046d87SRobert Watson } else 1969fa046d87SRobert Watson panic("%s: locking bug", __func__); 1970fa046d87SRobert Watson } else 1971fa046d87SRobert Watson INP_HASH_RUNLOCK(pcbinfo); 1972fa046d87SRobert Watson return (inp); 1973fa046d87SRobert Watson } 1974fa046d87SRobert Watson 1975fa046d87SRobert Watson /* 1976d3c1f003SRobert Watson * Public inpcb lookup routines, accepting a 4-tuple, and optionally, an mbuf 1977d3c1f003SRobert Watson * from which a pre-calculated hash value may be extracted. 197852cd27cbSRobert Watson * 197952cd27cbSRobert Watson * Possibly more of this logic should be in in_pcbgroup.c. 1980fa046d87SRobert Watson */ 1981fa046d87SRobert Watson struct inpcb * 1982fa046d87SRobert Watson in_pcblookup(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport, 1983fa046d87SRobert Watson struct in_addr laddr, u_int lport, int lookupflags, struct ifnet *ifp) 1984fa046d87SRobert Watson { 19857527624eSRobert Watson #if defined(PCBGROUP) && !defined(RSS) 198652cd27cbSRobert Watson struct inpcbgroup *pcbgroup; 198752cd27cbSRobert Watson #endif 1988fa046d87SRobert Watson 1989fa046d87SRobert Watson KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0, 1990fa046d87SRobert Watson ("%s: invalid lookup flags %d", __func__, lookupflags)); 1991fa046d87SRobert Watson KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0, 1992fa046d87SRobert Watson ("%s: LOCKPCB not set", __func__)); 1993fa046d87SRobert Watson 19947527624eSRobert Watson /* 19957527624eSRobert Watson * When not using RSS, use connection groups in preference to the 19967527624eSRobert Watson * reservation table when looking up 4-tuples. When using RSS, just 19977527624eSRobert Watson * use the reservation table, due to the cost of the Toeplitz hash 19987527624eSRobert Watson * in software. 19997527624eSRobert Watson * 20007527624eSRobert Watson * XXXRW: This policy belongs in the pcbgroup code, as in principle 20017527624eSRobert Watson * we could be doing RSS with a non-Toeplitz hash that is affordable 20027527624eSRobert Watson * in software. 20037527624eSRobert Watson */ 20047527624eSRobert Watson #if defined(PCBGROUP) && !defined(RSS) 200552cd27cbSRobert Watson if (in_pcbgroup_enabled(pcbinfo)) { 200652cd27cbSRobert Watson pcbgroup = in_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr, 200752cd27cbSRobert Watson fport); 200852cd27cbSRobert Watson return (in_pcblookup_group(pcbinfo, pcbgroup, faddr, fport, 200952cd27cbSRobert Watson laddr, lport, lookupflags, ifp)); 201052cd27cbSRobert Watson } 201152cd27cbSRobert Watson #endif 2012fa046d87SRobert Watson return (in_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport, 2013fa046d87SRobert Watson lookupflags, ifp)); 2014fa046d87SRobert Watson } 2015d3c1f003SRobert Watson 2016d3c1f003SRobert Watson struct inpcb * 2017d3c1f003SRobert Watson in_pcblookup_mbuf(struct inpcbinfo *pcbinfo, struct in_addr faddr, 2018d3c1f003SRobert Watson u_int fport, struct in_addr laddr, u_int lport, int lookupflags, 2019d3c1f003SRobert Watson struct ifnet *ifp, struct mbuf *m) 2020d3c1f003SRobert Watson { 202152cd27cbSRobert Watson #ifdef PCBGROUP 202252cd27cbSRobert Watson struct inpcbgroup *pcbgroup; 202352cd27cbSRobert Watson #endif 2024d3c1f003SRobert Watson 2025d3c1f003SRobert Watson KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0, 2026d3c1f003SRobert Watson ("%s: invalid lookup flags %d", __func__, lookupflags)); 2027d3c1f003SRobert Watson KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0, 2028d3c1f003SRobert Watson ("%s: LOCKPCB not set", __func__)); 2029d3c1f003SRobert Watson 203052cd27cbSRobert Watson #ifdef PCBGROUP 20317527624eSRobert Watson /* 20327527624eSRobert Watson * If we can use a hardware-generated hash to look up the connection 20337527624eSRobert Watson * group, use that connection group to find the inpcb. Otherwise 20347527624eSRobert Watson * fall back on a software hash -- or the reservation table if we're 20357527624eSRobert Watson * using RSS. 20367527624eSRobert Watson * 20377527624eSRobert Watson * XXXRW: As above, that policy belongs in the pcbgroup code. 20387527624eSRobert Watson */ 20397527624eSRobert Watson if (in_pcbgroup_enabled(pcbinfo) && 20407527624eSRobert Watson !(M_HASHTYPE_TEST(m, M_HASHTYPE_NONE))) { 204152cd27cbSRobert Watson pcbgroup = in_pcbgroup_byhash(pcbinfo, M_HASHTYPE_GET(m), 204252cd27cbSRobert Watson m->m_pkthdr.flowid); 204352cd27cbSRobert Watson if (pcbgroup != NULL) 204452cd27cbSRobert Watson return (in_pcblookup_group(pcbinfo, pcbgroup, faddr, 204552cd27cbSRobert Watson fport, laddr, lport, lookupflags, ifp)); 20467527624eSRobert Watson #ifndef RSS 204752cd27cbSRobert Watson pcbgroup = in_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr, 204852cd27cbSRobert Watson fport); 204952cd27cbSRobert Watson return (in_pcblookup_group(pcbinfo, pcbgroup, faddr, fport, 205052cd27cbSRobert Watson laddr, lport, lookupflags, ifp)); 20517527624eSRobert Watson #endif 205252cd27cbSRobert Watson } 205352cd27cbSRobert Watson #endif 2054d3c1f003SRobert Watson return (in_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport, 2055d3c1f003SRobert Watson lookupflags, ifp)); 2056d3c1f003SRobert Watson } 205767107f45SBjoern A. Zeeb #endif /* INET */ 205815bd2b43SDavid Greenman 20597bc4aca7SDavid Greenman /* 2060c3229e05SDavid Greenman * Insert PCB onto various hash lists. 20617bc4aca7SDavid Greenman */ 206252cd27cbSRobert Watson static int 206352cd27cbSRobert Watson in_pcbinshash_internal(struct inpcb *inp, int do_pcbgroup_update) 206415bd2b43SDavid Greenman { 2065c3229e05SDavid Greenman struct inpcbhead *pcbhash; 2066c3229e05SDavid Greenman struct inpcbporthead *pcbporthash; 2067c3229e05SDavid Greenman struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 2068c3229e05SDavid Greenman struct inpcbport *phd; 2069cfa1ca9dSYoshinobu Inoue u_int32_t hashkey_faddr; 207015bd2b43SDavid Greenman 20718501a69cSRobert Watson INP_WLOCK_ASSERT(inp); 2072fa046d87SRobert Watson INP_HASH_WLOCK_ASSERT(pcbinfo); 2073fa046d87SRobert Watson 2074111d57a6SRobert Watson KASSERT((inp->inp_flags & INP_INHASHLIST) == 0, 2075111d57a6SRobert Watson ("in_pcbinshash: INP_INHASHLIST")); 2076602cc7f1SRobert Watson 2077cfa1ca9dSYoshinobu Inoue #ifdef INET6 2078cfa1ca9dSYoshinobu Inoue if (inp->inp_vflag & INP_IPV6) 20791b44e5ffSAndrey V. Elsukov hashkey_faddr = INP6_PCBHASHKEY(&inp->in6p_faddr); 2080cfa1ca9dSYoshinobu Inoue else 208183e521ecSBjoern A. Zeeb #endif 2082cfa1ca9dSYoshinobu Inoue hashkey_faddr = inp->inp_faddr.s_addr; 2083cfa1ca9dSYoshinobu Inoue 2084712fc218SRobert Watson pcbhash = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr, 2085712fc218SRobert Watson inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)]; 208615bd2b43SDavid Greenman 2087712fc218SRobert Watson pcbporthash = &pcbinfo->ipi_porthashbase[ 2088712fc218SRobert Watson INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)]; 2089c3229e05SDavid Greenman 2090c3229e05SDavid Greenman /* 2091c3229e05SDavid Greenman * Go through port list and look for a head for this lport. 2092c3229e05SDavid Greenman */ 2093fc2ffbe6SPoul-Henning Kamp LIST_FOREACH(phd, pcbporthash, phd_hash) { 2094c3229e05SDavid Greenman if (phd->phd_port == inp->inp_lport) 2095c3229e05SDavid Greenman break; 2096c3229e05SDavid Greenman } 2097c3229e05SDavid Greenman /* 2098c3229e05SDavid Greenman * If none exists, malloc one and tack it on. 2099c3229e05SDavid Greenman */ 2100c3229e05SDavid Greenman if (phd == NULL) { 21011ede983cSDag-Erling Smørgrav phd = malloc(sizeof(struct inpcbport), M_PCB, M_NOWAIT); 2102c3229e05SDavid Greenman if (phd == NULL) { 2103c3229e05SDavid Greenman return (ENOBUFS); /* XXX */ 2104c3229e05SDavid Greenman } 2105c3229e05SDavid Greenman phd->phd_port = inp->inp_lport; 2106c3229e05SDavid Greenman LIST_INIT(&phd->phd_pcblist); 2107c3229e05SDavid Greenman LIST_INSERT_HEAD(pcbporthash, phd, phd_hash); 2108c3229e05SDavid Greenman } 2109c3229e05SDavid Greenman inp->inp_phd = phd; 2110c3229e05SDavid Greenman LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist); 2111c3229e05SDavid Greenman LIST_INSERT_HEAD(pcbhash, inp, inp_hash); 2112111d57a6SRobert Watson inp->inp_flags |= INP_INHASHLIST; 211352cd27cbSRobert Watson #ifdef PCBGROUP 211452cd27cbSRobert Watson if (do_pcbgroup_update) 211552cd27cbSRobert Watson in_pcbgroup_update(inp); 211652cd27cbSRobert Watson #endif 2117c3229e05SDavid Greenman return (0); 211815bd2b43SDavid Greenman } 211915bd2b43SDavid Greenman 2120c3229e05SDavid Greenman /* 212152cd27cbSRobert Watson * For now, there are two public interfaces to insert an inpcb into the hash 212252cd27cbSRobert Watson * lists -- one that does update pcbgroups, and one that doesn't. The latter 212352cd27cbSRobert Watson * is used only in the TCP syncache, where in_pcbinshash is called before the 212452cd27cbSRobert Watson * full 4-tuple is set for the inpcb, and we don't want to install in the 212552cd27cbSRobert Watson * pcbgroup until later. 212652cd27cbSRobert Watson * 212752cd27cbSRobert Watson * XXXRW: This seems like a misfeature. in_pcbinshash should always update 212852cd27cbSRobert Watson * connection groups, and partially initialised inpcbs should not be exposed 212952cd27cbSRobert Watson * to either reservation hash tables or pcbgroups. 213052cd27cbSRobert Watson */ 213152cd27cbSRobert Watson int 213252cd27cbSRobert Watson in_pcbinshash(struct inpcb *inp) 213352cd27cbSRobert Watson { 213452cd27cbSRobert Watson 213552cd27cbSRobert Watson return (in_pcbinshash_internal(inp, 1)); 213652cd27cbSRobert Watson } 213752cd27cbSRobert Watson 213852cd27cbSRobert Watson int 213952cd27cbSRobert Watson in_pcbinshash_nopcbgroup(struct inpcb *inp) 214052cd27cbSRobert Watson { 214152cd27cbSRobert Watson 214252cd27cbSRobert Watson return (in_pcbinshash_internal(inp, 0)); 214352cd27cbSRobert Watson } 214452cd27cbSRobert Watson 214552cd27cbSRobert Watson /* 2146c3229e05SDavid Greenman * Move PCB to the proper hash bucket when { faddr, fport } have been 2147c3229e05SDavid Greenman * changed. NOTE: This does not handle the case of the lport changing (the 2148c3229e05SDavid Greenman * hashed port list would have to be updated as well), so the lport must 2149c3229e05SDavid Greenman * not change after in_pcbinshash() has been called. 2150c3229e05SDavid Greenman */ 215115bd2b43SDavid Greenman void 2152d3c1f003SRobert Watson in_pcbrehash_mbuf(struct inpcb *inp, struct mbuf *m) 215315bd2b43SDavid Greenman { 215459daba27SSam Leffler struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 215515bd2b43SDavid Greenman struct inpcbhead *head; 2156cfa1ca9dSYoshinobu Inoue u_int32_t hashkey_faddr; 215715bd2b43SDavid Greenman 21588501a69cSRobert Watson INP_WLOCK_ASSERT(inp); 2159fa046d87SRobert Watson INP_HASH_WLOCK_ASSERT(pcbinfo); 2160fa046d87SRobert Watson 2161111d57a6SRobert Watson KASSERT(inp->inp_flags & INP_INHASHLIST, 2162111d57a6SRobert Watson ("in_pcbrehash: !INP_INHASHLIST")); 2163602cc7f1SRobert Watson 2164cfa1ca9dSYoshinobu Inoue #ifdef INET6 2165cfa1ca9dSYoshinobu Inoue if (inp->inp_vflag & INP_IPV6) 21661b44e5ffSAndrey V. Elsukov hashkey_faddr = INP6_PCBHASHKEY(&inp->in6p_faddr); 2167cfa1ca9dSYoshinobu Inoue else 216883e521ecSBjoern A. Zeeb #endif 2169cfa1ca9dSYoshinobu Inoue hashkey_faddr = inp->inp_faddr.s_addr; 2170cfa1ca9dSYoshinobu Inoue 2171712fc218SRobert Watson head = &pcbinfo->ipi_hashbase[INP_PCBHASH(hashkey_faddr, 2172712fc218SRobert Watson inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask)]; 217315bd2b43SDavid Greenman 2174c3229e05SDavid Greenman LIST_REMOVE(inp, inp_hash); 217515bd2b43SDavid Greenman LIST_INSERT_HEAD(head, inp, inp_hash); 217652cd27cbSRobert Watson 217752cd27cbSRobert Watson #ifdef PCBGROUP 217852cd27cbSRobert Watson if (m != NULL) 217952cd27cbSRobert Watson in_pcbgroup_update_mbuf(inp, m); 218052cd27cbSRobert Watson else 218152cd27cbSRobert Watson in_pcbgroup_update(inp); 218252cd27cbSRobert Watson #endif 2183c3229e05SDavid Greenman } 2184c3229e05SDavid Greenman 2185d3c1f003SRobert Watson void 2186d3c1f003SRobert Watson in_pcbrehash(struct inpcb *inp) 2187d3c1f003SRobert Watson { 2188d3c1f003SRobert Watson 2189d3c1f003SRobert Watson in_pcbrehash_mbuf(inp, NULL); 2190d3c1f003SRobert Watson } 2191d3c1f003SRobert Watson 2192c3229e05SDavid Greenman /* 2193c3229e05SDavid Greenman * Remove PCB from various lists. 2194c3229e05SDavid Greenman */ 21956d888973SRobert Watson static void 2196136d4f1cSRobert Watson in_pcbremlists(struct inpcb *inp) 2197c3229e05SDavid Greenman { 219859daba27SSam Leffler struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 219959daba27SSam Leffler 2200ff9b006dSJulien Charbon #ifdef INVARIANTS 2201ff9b006dSJulien Charbon if (pcbinfo == &V_tcbinfo) { 2202ff9b006dSJulien Charbon INP_INFO_RLOCK_ASSERT(pcbinfo); 2203ff9b006dSJulien Charbon } else { 220459daba27SSam Leffler INP_INFO_WLOCK_ASSERT(pcbinfo); 2205ff9b006dSJulien Charbon } 2206ff9b006dSJulien Charbon #endif 2207ff9b006dSJulien Charbon 22088501a69cSRobert Watson INP_WLOCK_ASSERT(inp); 2209ff9b006dSJulien Charbon INP_LIST_WLOCK_ASSERT(pcbinfo); 221059daba27SSam Leffler 221159daba27SSam Leffler inp->inp_gencnt = ++pcbinfo->ipi_gencnt; 2212111d57a6SRobert Watson if (inp->inp_flags & INP_INHASHLIST) { 2213c3229e05SDavid Greenman struct inpcbport *phd = inp->inp_phd; 2214c3229e05SDavid Greenman 2215fa046d87SRobert Watson INP_HASH_WLOCK(pcbinfo); 2216c3229e05SDavid Greenman LIST_REMOVE(inp, inp_hash); 2217c3229e05SDavid Greenman LIST_REMOVE(inp, inp_portlist); 2218fc2ffbe6SPoul-Henning Kamp if (LIST_FIRST(&phd->phd_pcblist) == NULL) { 2219c3229e05SDavid Greenman LIST_REMOVE(phd, phd_hash); 2220c3229e05SDavid Greenman free(phd, M_PCB); 2221c3229e05SDavid Greenman } 2222fa046d87SRobert Watson INP_HASH_WUNLOCK(pcbinfo); 2223111d57a6SRobert Watson inp->inp_flags &= ~INP_INHASHLIST; 2224c3229e05SDavid Greenman } 2225c3229e05SDavid Greenman LIST_REMOVE(inp, inp_list); 222659daba27SSam Leffler pcbinfo->ipi_count--; 222752cd27cbSRobert Watson #ifdef PCBGROUP 222852cd27cbSRobert Watson in_pcbgroup_remove(inp); 222952cd27cbSRobert Watson #endif 223015bd2b43SDavid Greenman } 223175c13541SPoul-Henning Kamp 2232a557af22SRobert Watson /* 2233*84cc0778SGeorge V. Neville-Neil * Check for alternatives when higher level complains 2234*84cc0778SGeorge V. Neville-Neil * about service problems. For now, invalidate cached 2235*84cc0778SGeorge V. Neville-Neil * routing information. If the route was created dynamically 2236*84cc0778SGeorge V. Neville-Neil * (by a redirect), time to try a default gateway again. 2237*84cc0778SGeorge V. Neville-Neil */ 2238*84cc0778SGeorge V. Neville-Neil void 2239*84cc0778SGeorge V. Neville-Neil in_losing(struct inpcb *inp) 2240*84cc0778SGeorge V. Neville-Neil { 2241*84cc0778SGeorge V. Neville-Neil 2242*84cc0778SGeorge V. Neville-Neil if (inp->inp_route.ro_rt) { 2243*84cc0778SGeorge V. Neville-Neil RTFREE(inp->inp_route.ro_rt); 2244*84cc0778SGeorge V. Neville-Neil inp->inp_route.ro_rt = (struct rtentry *)NULL; 2245*84cc0778SGeorge V. Neville-Neil } 2246*84cc0778SGeorge V. Neville-Neil return; 2247*84cc0778SGeorge V. Neville-Neil } 2248*84cc0778SGeorge V. Neville-Neil 2249*84cc0778SGeorge V. Neville-Neil /* 2250a557af22SRobert Watson * A set label operation has occurred at the socket layer, propagate the 2251a557af22SRobert Watson * label change into the in_pcb for the socket. 2252a557af22SRobert Watson */ 2253a557af22SRobert Watson void 2254136d4f1cSRobert Watson in_pcbsosetlabel(struct socket *so) 2255a557af22SRobert Watson { 2256a557af22SRobert Watson #ifdef MAC 2257a557af22SRobert Watson struct inpcb *inp; 2258a557af22SRobert Watson 22594c7c478dSRobert Watson inp = sotoinpcb(so); 22604c7c478dSRobert Watson KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL")); 2261602cc7f1SRobert Watson 22628501a69cSRobert Watson INP_WLOCK(inp); 2263310e7cebSRobert Watson SOCK_LOCK(so); 2264a557af22SRobert Watson mac_inpcb_sosetlabel(so, inp); 2265310e7cebSRobert Watson SOCK_UNLOCK(so); 22668501a69cSRobert Watson INP_WUNLOCK(inp); 2267a557af22SRobert Watson #endif 2268a557af22SRobert Watson } 22695f311da2SMike Silbersack 22705f311da2SMike Silbersack /* 2271ad3a630fSRobert Watson * ipport_tick runs once per second, determining if random port allocation 2272ad3a630fSRobert Watson * should be continued. If more than ipport_randomcps ports have been 2273ad3a630fSRobert Watson * allocated in the last second, then we return to sequential port 2274ad3a630fSRobert Watson * allocation. We return to random allocation only once we drop below 2275ad3a630fSRobert Watson * ipport_randomcps for at least ipport_randomtime seconds. 22765f311da2SMike Silbersack */ 2277aae49dd3SBjoern A. Zeeb static void 2278136d4f1cSRobert Watson ipport_tick(void *xtp) 22795f311da2SMike Silbersack { 22808b615593SMarko Zec VNET_ITERATOR_DECL(vnet_iter); 2281ad3a630fSRobert Watson 22825ee847d3SRobert Watson VNET_LIST_RLOCK_NOSLEEP(); 22838b615593SMarko Zec VNET_FOREACH(vnet_iter) { 22848b615593SMarko Zec CURVNET_SET(vnet_iter); /* XXX appease INVARIANTS here */ 22858b615593SMarko Zec if (V_ipport_tcpallocs <= 22868b615593SMarko Zec V_ipport_tcplastcount + V_ipport_randomcps) { 2287603724d3SBjoern A. Zeeb if (V_ipport_stoprandom > 0) 2288603724d3SBjoern A. Zeeb V_ipport_stoprandom--; 2289ad3a630fSRobert Watson } else 2290603724d3SBjoern A. Zeeb V_ipport_stoprandom = V_ipport_randomtime; 2291603724d3SBjoern A. Zeeb V_ipport_tcplastcount = V_ipport_tcpallocs; 22928b615593SMarko Zec CURVNET_RESTORE(); 22938b615593SMarko Zec } 22945ee847d3SRobert Watson VNET_LIST_RUNLOCK_NOSLEEP(); 22955f311da2SMike Silbersack callout_reset(&ipport_tick_callout, hz, ipport_tick, NULL); 22965f311da2SMike Silbersack } 2297497057eeSRobert Watson 2298aae49dd3SBjoern A. Zeeb static void 2299aae49dd3SBjoern A. Zeeb ip_fini(void *xtp) 2300aae49dd3SBjoern A. Zeeb { 2301aae49dd3SBjoern A. Zeeb 2302aae49dd3SBjoern A. Zeeb callout_stop(&ipport_tick_callout); 2303aae49dd3SBjoern A. Zeeb } 2304aae49dd3SBjoern A. Zeeb 2305aae49dd3SBjoern A. Zeeb /* 2306aae49dd3SBjoern A. Zeeb * The ipport_callout should start running at about the time we attach the 2307aae49dd3SBjoern A. Zeeb * inet or inet6 domains. 2308aae49dd3SBjoern A. Zeeb */ 2309aae49dd3SBjoern A. Zeeb static void 2310aae49dd3SBjoern A. Zeeb ipport_tick_init(const void *unused __unused) 2311aae49dd3SBjoern A. Zeeb { 2312aae49dd3SBjoern A. Zeeb 2313aae49dd3SBjoern A. Zeeb /* Start ipport_tick. */ 2314fd90e2edSJung-uk Kim callout_init(&ipport_tick_callout, 1); 2315aae49dd3SBjoern A. Zeeb callout_reset(&ipport_tick_callout, 1, ipport_tick, NULL); 2316aae49dd3SBjoern A. Zeeb EVENTHANDLER_REGISTER(shutdown_pre_sync, ip_fini, NULL, 2317aae49dd3SBjoern A. Zeeb SHUTDOWN_PRI_DEFAULT); 2318aae49dd3SBjoern A. Zeeb } 2319aae49dd3SBjoern A. Zeeb SYSINIT(ipport_tick_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, 2320aae49dd3SBjoern A. Zeeb ipport_tick_init, NULL); 2321aae49dd3SBjoern A. Zeeb 23223d585327SKip Macy void 23233d585327SKip Macy inp_wlock(struct inpcb *inp) 23243d585327SKip Macy { 23253d585327SKip Macy 23268501a69cSRobert Watson INP_WLOCK(inp); 23273d585327SKip Macy } 23283d585327SKip Macy 23293d585327SKip Macy void 23303d585327SKip Macy inp_wunlock(struct inpcb *inp) 23313d585327SKip Macy { 23323d585327SKip Macy 23338501a69cSRobert Watson INP_WUNLOCK(inp); 23343d585327SKip Macy } 23353d585327SKip Macy 23363d585327SKip Macy void 23373d585327SKip Macy inp_rlock(struct inpcb *inp) 23383d585327SKip Macy { 23393d585327SKip Macy 2340a69042a5SRobert Watson INP_RLOCK(inp); 23413d585327SKip Macy } 23423d585327SKip Macy 23433d585327SKip Macy void 23443d585327SKip Macy inp_runlock(struct inpcb *inp) 23453d585327SKip Macy { 23463d585327SKip Macy 2347a69042a5SRobert Watson INP_RUNLOCK(inp); 23483d585327SKip Macy } 23493d585327SKip Macy 23503d585327SKip Macy #ifdef INVARIANTS 23513d585327SKip Macy void 2352e79dd20dSKip Macy inp_lock_assert(struct inpcb *inp) 23533d585327SKip Macy { 23543d585327SKip Macy 23558501a69cSRobert Watson INP_WLOCK_ASSERT(inp); 23563d585327SKip Macy } 23573d585327SKip Macy 23583d585327SKip Macy void 2359e79dd20dSKip Macy inp_unlock_assert(struct inpcb *inp) 23603d585327SKip Macy { 23613d585327SKip Macy 23623d585327SKip Macy INP_UNLOCK_ASSERT(inp); 23633d585327SKip Macy } 23643d585327SKip Macy #endif 23653d585327SKip Macy 23669378e437SKip Macy void 23679378e437SKip Macy inp_apply_all(void (*func)(struct inpcb *, void *), void *arg) 23689378e437SKip Macy { 23699378e437SKip Macy struct inpcb *inp; 23709378e437SKip Macy 2371ff9b006dSJulien Charbon INP_INFO_WLOCK(&V_tcbinfo); 237297021c24SMarko Zec LIST_FOREACH(inp, V_tcbinfo.ipi_listhead, inp_list) { 23739378e437SKip Macy INP_WLOCK(inp); 23749378e437SKip Macy func(inp, arg); 23759378e437SKip Macy INP_WUNLOCK(inp); 23769378e437SKip Macy } 2377ff9b006dSJulien Charbon INP_INFO_WUNLOCK(&V_tcbinfo); 23789378e437SKip Macy } 23799378e437SKip Macy 23809378e437SKip Macy struct socket * 23819378e437SKip Macy inp_inpcbtosocket(struct inpcb *inp) 23829378e437SKip Macy { 23839378e437SKip Macy 23849378e437SKip Macy INP_WLOCK_ASSERT(inp); 23859378e437SKip Macy return (inp->inp_socket); 23869378e437SKip Macy } 23879378e437SKip Macy 23889378e437SKip Macy struct tcpcb * 23899378e437SKip Macy inp_inpcbtotcpcb(struct inpcb *inp) 23909378e437SKip Macy { 23919378e437SKip Macy 23929378e437SKip Macy INP_WLOCK_ASSERT(inp); 23939378e437SKip Macy return ((struct tcpcb *)inp->inp_ppcb); 23949378e437SKip Macy } 23959378e437SKip Macy 23969378e437SKip Macy int 23979378e437SKip Macy inp_ip_tos_get(const struct inpcb *inp) 23989378e437SKip Macy { 23999378e437SKip Macy 24009378e437SKip Macy return (inp->inp_ip_tos); 24019378e437SKip Macy } 24029378e437SKip Macy 24039378e437SKip Macy void 24049378e437SKip Macy inp_ip_tos_set(struct inpcb *inp, int val) 24059378e437SKip Macy { 24069378e437SKip Macy 24079378e437SKip Macy inp->inp_ip_tos = val; 24089378e437SKip Macy } 24099378e437SKip Macy 24109378e437SKip Macy void 2411df9cf830STai-hwa Liang inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp, 24129d29c635SKip Macy uint32_t *faddr, uint16_t *fp) 24139378e437SKip Macy { 24149378e437SKip Macy 24159d29c635SKip Macy INP_LOCK_ASSERT(inp); 2416df9cf830STai-hwa Liang *laddr = inp->inp_laddr.s_addr; 2417df9cf830STai-hwa Liang *faddr = inp->inp_faddr.s_addr; 24189378e437SKip Macy *lp = inp->inp_lport; 24199378e437SKip Macy *fp = inp->inp_fport; 24209378e437SKip Macy } 24219378e437SKip Macy 2422dd0e6c38SKip Macy struct inpcb * 2423dd0e6c38SKip Macy so_sotoinpcb(struct socket *so) 2424dd0e6c38SKip Macy { 2425dd0e6c38SKip Macy 2426dd0e6c38SKip Macy return (sotoinpcb(so)); 2427dd0e6c38SKip Macy } 2428dd0e6c38SKip Macy 2429dd0e6c38SKip Macy struct tcpcb * 2430dd0e6c38SKip Macy so_sototcpcb(struct socket *so) 2431dd0e6c38SKip Macy { 2432dd0e6c38SKip Macy 2433dd0e6c38SKip Macy return (sototcpcb(so)); 2434dd0e6c38SKip Macy } 2435dd0e6c38SKip Macy 2436497057eeSRobert Watson #ifdef DDB 2437497057eeSRobert Watson static void 2438497057eeSRobert Watson db_print_indent(int indent) 2439497057eeSRobert Watson { 2440497057eeSRobert Watson int i; 2441497057eeSRobert Watson 2442497057eeSRobert Watson for (i = 0; i < indent; i++) 2443497057eeSRobert Watson db_printf(" "); 2444497057eeSRobert Watson } 2445497057eeSRobert Watson 2446497057eeSRobert Watson static void 2447497057eeSRobert Watson db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent) 2448497057eeSRobert Watson { 2449497057eeSRobert Watson char faddr_str[48], laddr_str[48]; 2450497057eeSRobert Watson 2451497057eeSRobert Watson db_print_indent(indent); 2452497057eeSRobert Watson db_printf("%s at %p\n", name, inc); 2453497057eeSRobert Watson 2454497057eeSRobert Watson indent += 2; 2455497057eeSRobert Watson 245603dc38a4SRobert Watson #ifdef INET6 2457dcdb4371SBjoern A. Zeeb if (inc->inc_flags & INC_ISIPV6) { 2458497057eeSRobert Watson /* IPv6. */ 2459497057eeSRobert Watson ip6_sprintf(laddr_str, &inc->inc6_laddr); 2460497057eeSRobert Watson ip6_sprintf(faddr_str, &inc->inc6_faddr); 246183e521ecSBjoern A. Zeeb } else 246203dc38a4SRobert Watson #endif 246383e521ecSBjoern A. Zeeb { 2464497057eeSRobert Watson /* IPv4. */ 2465497057eeSRobert Watson inet_ntoa_r(inc->inc_laddr, laddr_str); 2466497057eeSRobert Watson inet_ntoa_r(inc->inc_faddr, faddr_str); 2467497057eeSRobert Watson } 2468497057eeSRobert Watson db_print_indent(indent); 2469497057eeSRobert Watson db_printf("inc_laddr %s inc_lport %u\n", laddr_str, 2470497057eeSRobert Watson ntohs(inc->inc_lport)); 2471497057eeSRobert Watson db_print_indent(indent); 2472497057eeSRobert Watson db_printf("inc_faddr %s inc_fport %u\n", faddr_str, 2473497057eeSRobert Watson ntohs(inc->inc_fport)); 2474497057eeSRobert Watson } 2475497057eeSRobert Watson 2476497057eeSRobert Watson static void 2477497057eeSRobert Watson db_print_inpflags(int inp_flags) 2478497057eeSRobert Watson { 2479497057eeSRobert Watson int comma; 2480497057eeSRobert Watson 2481497057eeSRobert Watson comma = 0; 2482497057eeSRobert Watson if (inp_flags & INP_RECVOPTS) { 2483497057eeSRobert Watson db_printf("%sINP_RECVOPTS", comma ? ", " : ""); 2484497057eeSRobert Watson comma = 1; 2485497057eeSRobert Watson } 2486497057eeSRobert Watson if (inp_flags & INP_RECVRETOPTS) { 2487497057eeSRobert Watson db_printf("%sINP_RECVRETOPTS", comma ? ", " : ""); 2488497057eeSRobert Watson comma = 1; 2489497057eeSRobert Watson } 2490497057eeSRobert Watson if (inp_flags & INP_RECVDSTADDR) { 2491497057eeSRobert Watson db_printf("%sINP_RECVDSTADDR", comma ? ", " : ""); 2492497057eeSRobert Watson comma = 1; 2493497057eeSRobert Watson } 2494497057eeSRobert Watson if (inp_flags & INP_HDRINCL) { 2495497057eeSRobert Watson db_printf("%sINP_HDRINCL", comma ? ", " : ""); 2496497057eeSRobert Watson comma = 1; 2497497057eeSRobert Watson } 2498497057eeSRobert Watson if (inp_flags & INP_HIGHPORT) { 2499497057eeSRobert Watson db_printf("%sINP_HIGHPORT", comma ? ", " : ""); 2500497057eeSRobert Watson comma = 1; 2501497057eeSRobert Watson } 2502497057eeSRobert Watson if (inp_flags & INP_LOWPORT) { 2503497057eeSRobert Watson db_printf("%sINP_LOWPORT", comma ? ", " : ""); 2504497057eeSRobert Watson comma = 1; 2505497057eeSRobert Watson } 2506497057eeSRobert Watson if (inp_flags & INP_ANONPORT) { 2507497057eeSRobert Watson db_printf("%sINP_ANONPORT", comma ? ", " : ""); 2508497057eeSRobert Watson comma = 1; 2509497057eeSRobert Watson } 2510497057eeSRobert Watson if (inp_flags & INP_RECVIF) { 2511497057eeSRobert Watson db_printf("%sINP_RECVIF", comma ? ", " : ""); 2512497057eeSRobert Watson comma = 1; 2513497057eeSRobert Watson } 2514497057eeSRobert Watson if (inp_flags & INP_MTUDISC) { 2515497057eeSRobert Watson db_printf("%sINP_MTUDISC", comma ? ", " : ""); 2516497057eeSRobert Watson comma = 1; 2517497057eeSRobert Watson } 2518497057eeSRobert Watson if (inp_flags & INP_RECVTTL) { 2519497057eeSRobert Watson db_printf("%sINP_RECVTTL", comma ? ", " : ""); 2520497057eeSRobert Watson comma = 1; 2521497057eeSRobert Watson } 2522497057eeSRobert Watson if (inp_flags & INP_DONTFRAG) { 2523497057eeSRobert Watson db_printf("%sINP_DONTFRAG", comma ? ", " : ""); 2524497057eeSRobert Watson comma = 1; 2525497057eeSRobert Watson } 25263cca425bSMichael Tuexen if (inp_flags & INP_RECVTOS) { 25273cca425bSMichael Tuexen db_printf("%sINP_RECVTOS", comma ? ", " : ""); 25283cca425bSMichael Tuexen comma = 1; 25293cca425bSMichael Tuexen } 2530497057eeSRobert Watson if (inp_flags & IN6P_IPV6_V6ONLY) { 2531497057eeSRobert Watson db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : ""); 2532497057eeSRobert Watson comma = 1; 2533497057eeSRobert Watson } 2534497057eeSRobert Watson if (inp_flags & IN6P_PKTINFO) { 2535497057eeSRobert Watson db_printf("%sIN6P_PKTINFO", comma ? ", " : ""); 2536497057eeSRobert Watson comma = 1; 2537497057eeSRobert Watson } 2538497057eeSRobert Watson if (inp_flags & IN6P_HOPLIMIT) { 2539497057eeSRobert Watson db_printf("%sIN6P_HOPLIMIT", comma ? ", " : ""); 2540497057eeSRobert Watson comma = 1; 2541497057eeSRobert Watson } 2542497057eeSRobert Watson if (inp_flags & IN6P_HOPOPTS) { 2543497057eeSRobert Watson db_printf("%sIN6P_HOPOPTS", comma ? ", " : ""); 2544497057eeSRobert Watson comma = 1; 2545497057eeSRobert Watson } 2546497057eeSRobert Watson if (inp_flags & IN6P_DSTOPTS) { 2547497057eeSRobert Watson db_printf("%sIN6P_DSTOPTS", comma ? ", " : ""); 2548497057eeSRobert Watson comma = 1; 2549497057eeSRobert Watson } 2550497057eeSRobert Watson if (inp_flags & IN6P_RTHDR) { 2551497057eeSRobert Watson db_printf("%sIN6P_RTHDR", comma ? ", " : ""); 2552497057eeSRobert Watson comma = 1; 2553497057eeSRobert Watson } 2554497057eeSRobert Watson if (inp_flags & IN6P_RTHDRDSTOPTS) { 2555497057eeSRobert Watson db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : ""); 2556497057eeSRobert Watson comma = 1; 2557497057eeSRobert Watson } 2558497057eeSRobert Watson if (inp_flags & IN6P_TCLASS) { 2559497057eeSRobert Watson db_printf("%sIN6P_TCLASS", comma ? ", " : ""); 2560497057eeSRobert Watson comma = 1; 2561497057eeSRobert Watson } 2562497057eeSRobert Watson if (inp_flags & IN6P_AUTOFLOWLABEL) { 2563497057eeSRobert Watson db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : ""); 2564497057eeSRobert Watson comma = 1; 2565497057eeSRobert Watson } 2566ad71fe3cSRobert Watson if (inp_flags & INP_TIMEWAIT) { 2567ad71fe3cSRobert Watson db_printf("%sINP_TIMEWAIT", comma ? ", " : ""); 2568ad71fe3cSRobert Watson comma = 1; 2569ad71fe3cSRobert Watson } 2570ad71fe3cSRobert Watson if (inp_flags & INP_ONESBCAST) { 2571ad71fe3cSRobert Watson db_printf("%sINP_ONESBCAST", comma ? ", " : ""); 2572ad71fe3cSRobert Watson comma = 1; 2573ad71fe3cSRobert Watson } 2574ad71fe3cSRobert Watson if (inp_flags & INP_DROPPED) { 2575ad71fe3cSRobert Watson db_printf("%sINP_DROPPED", comma ? ", " : ""); 2576ad71fe3cSRobert Watson comma = 1; 2577ad71fe3cSRobert Watson } 2578ad71fe3cSRobert Watson if (inp_flags & INP_SOCKREF) { 2579ad71fe3cSRobert Watson db_printf("%sINP_SOCKREF", comma ? ", " : ""); 2580ad71fe3cSRobert Watson comma = 1; 2581ad71fe3cSRobert Watson } 2582497057eeSRobert Watson if (inp_flags & IN6P_RFC2292) { 2583497057eeSRobert Watson db_printf("%sIN6P_RFC2292", comma ? ", " : ""); 2584497057eeSRobert Watson comma = 1; 2585497057eeSRobert Watson } 2586497057eeSRobert Watson if (inp_flags & IN6P_MTU) { 2587497057eeSRobert Watson db_printf("IN6P_MTU%s", comma ? ", " : ""); 2588497057eeSRobert Watson comma = 1; 2589497057eeSRobert Watson } 2590497057eeSRobert Watson } 2591497057eeSRobert Watson 2592497057eeSRobert Watson static void 2593497057eeSRobert Watson db_print_inpvflag(u_char inp_vflag) 2594497057eeSRobert Watson { 2595497057eeSRobert Watson int comma; 2596497057eeSRobert Watson 2597497057eeSRobert Watson comma = 0; 2598497057eeSRobert Watson if (inp_vflag & INP_IPV4) { 2599497057eeSRobert Watson db_printf("%sINP_IPV4", comma ? ", " : ""); 2600497057eeSRobert Watson comma = 1; 2601497057eeSRobert Watson } 2602497057eeSRobert Watson if (inp_vflag & INP_IPV6) { 2603497057eeSRobert Watson db_printf("%sINP_IPV6", comma ? ", " : ""); 2604497057eeSRobert Watson comma = 1; 2605497057eeSRobert Watson } 2606497057eeSRobert Watson if (inp_vflag & INP_IPV6PROTO) { 2607497057eeSRobert Watson db_printf("%sINP_IPV6PROTO", comma ? ", " : ""); 2608497057eeSRobert Watson comma = 1; 2609497057eeSRobert Watson } 2610497057eeSRobert Watson } 2611497057eeSRobert Watson 26126d888973SRobert Watson static void 2613497057eeSRobert Watson db_print_inpcb(struct inpcb *inp, const char *name, int indent) 2614497057eeSRobert Watson { 2615497057eeSRobert Watson 2616497057eeSRobert Watson db_print_indent(indent); 2617497057eeSRobert Watson db_printf("%s at %p\n", name, inp); 2618497057eeSRobert Watson 2619497057eeSRobert Watson indent += 2; 2620497057eeSRobert Watson 2621497057eeSRobert Watson db_print_indent(indent); 2622497057eeSRobert Watson db_printf("inp_flow: 0x%x\n", inp->inp_flow); 2623497057eeSRobert Watson 2624497057eeSRobert Watson db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent); 2625497057eeSRobert Watson 2626497057eeSRobert Watson db_print_indent(indent); 2627497057eeSRobert Watson db_printf("inp_ppcb: %p inp_pcbinfo: %p inp_socket: %p\n", 2628497057eeSRobert Watson inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket); 2629497057eeSRobert Watson 2630497057eeSRobert Watson db_print_indent(indent); 2631497057eeSRobert Watson db_printf("inp_label: %p inp_flags: 0x%x (", 2632497057eeSRobert Watson inp->inp_label, inp->inp_flags); 2633497057eeSRobert Watson db_print_inpflags(inp->inp_flags); 2634497057eeSRobert Watson db_printf(")\n"); 2635497057eeSRobert Watson 2636497057eeSRobert Watson db_print_indent(indent); 2637497057eeSRobert Watson db_printf("inp_sp: %p inp_vflag: 0x%x (", inp->inp_sp, 2638497057eeSRobert Watson inp->inp_vflag); 2639497057eeSRobert Watson db_print_inpvflag(inp->inp_vflag); 2640497057eeSRobert Watson db_printf(")\n"); 2641497057eeSRobert Watson 2642497057eeSRobert Watson db_print_indent(indent); 2643497057eeSRobert Watson db_printf("inp_ip_ttl: %d inp_ip_p: %d inp_ip_minttl: %d\n", 2644497057eeSRobert Watson inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl); 2645497057eeSRobert Watson 2646497057eeSRobert Watson db_print_indent(indent); 2647497057eeSRobert Watson #ifdef INET6 2648497057eeSRobert Watson if (inp->inp_vflag & INP_IPV6) { 2649497057eeSRobert Watson db_printf("in6p_options: %p in6p_outputopts: %p " 2650497057eeSRobert Watson "in6p_moptions: %p\n", inp->in6p_options, 2651497057eeSRobert Watson inp->in6p_outputopts, inp->in6p_moptions); 2652497057eeSRobert Watson db_printf("in6p_icmp6filt: %p in6p_cksum %d " 2653497057eeSRobert Watson "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum, 2654497057eeSRobert Watson inp->in6p_hops); 2655497057eeSRobert Watson } else 2656497057eeSRobert Watson #endif 2657497057eeSRobert Watson { 2658497057eeSRobert Watson db_printf("inp_ip_tos: %d inp_ip_options: %p " 2659497057eeSRobert Watson "inp_ip_moptions: %p\n", inp->inp_ip_tos, 2660497057eeSRobert Watson inp->inp_options, inp->inp_moptions); 2661497057eeSRobert Watson } 2662497057eeSRobert Watson 2663497057eeSRobert Watson db_print_indent(indent); 2664497057eeSRobert Watson db_printf("inp_phd: %p inp_gencnt: %ju\n", inp->inp_phd, 2665497057eeSRobert Watson (uintmax_t)inp->inp_gencnt); 2666497057eeSRobert Watson } 2667497057eeSRobert Watson 2668497057eeSRobert Watson DB_SHOW_COMMAND(inpcb, db_show_inpcb) 2669497057eeSRobert Watson { 2670497057eeSRobert Watson struct inpcb *inp; 2671497057eeSRobert Watson 2672497057eeSRobert Watson if (!have_addr) { 2673497057eeSRobert Watson db_printf("usage: show inpcb <addr>\n"); 2674497057eeSRobert Watson return; 2675497057eeSRobert Watson } 2676497057eeSRobert Watson inp = (struct inpcb *)addr; 2677497057eeSRobert Watson 2678497057eeSRobert Watson db_print_inpcb(inp, "inpcb", 0); 2679497057eeSRobert Watson } 268083e521ecSBjoern A. Zeeb #endif /* DDB */ 2681