[BACK]Return to if.c CVS log [TXT][DIR] Up to [cvs.NetBSD.org] / src / sys / net

Annotation of src/sys/net/if.c, Revision 1.423

1.423   ! ozaki-r     1: /*     $NetBSD: if.c,v 1.422 2018/05/14 02:55:03 ozaki-r Exp $ */
1.53      thorpej     2:
                      3: /*-
1.219     ad          4:  * Copyright (c) 1999, 2000, 2001, 2008 The NetBSD Foundation, Inc.
1.53      thorpej     5:  * All rights reserved.
                      6:  *
                      7:  * This code is derived from software contributed to The NetBSD Foundation
1.150     peter       8:  * by William Studenmund and Jason R. Thorpe.
1.53      thorpej     9:  *
                     10:  * Redistribution and use in source and binary forms, with or without
                     11:  * modification, are permitted provided that the following conditions
                     12:  * are met:
                     13:  * 1. Redistributions of source code must retain the above copyright
                     14:  *    notice, this list of conditions and the following disclaimer.
                     15:  * 2. Redistributions in binary form must reproduce the above copyright
                     16:  *    notice, this list of conditions and the following disclaimer in the
                     17:  *    documentation and/or other materials provided with the distribution.
                     18:  *
                     19:  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
                     20:  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
                     21:  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
                     22:  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
                     23:  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
                     24:  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
                     25:  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
                     26:  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
                     27:  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
                     28:  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
                     29:  * POSSIBILITY OF SUCH DAMAGE.
                     30:  */
1.49      itojun     31:
                     32: /*
                     33:  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
                     34:  * All rights reserved.
1.154     perry      35:  *
1.49      itojun     36:  * Redistribution and use in source and binary forms, with or without
                     37:  * modification, are permitted provided that the following conditions
                     38:  * are met:
                     39:  * 1. Redistributions of source code must retain the above copyright
                     40:  *    notice, this list of conditions and the following disclaimer.
                     41:  * 2. Redistributions in binary form must reproduce the above copyright
                     42:  *    notice, this list of conditions and the following disclaimer in the
                     43:  *    documentation and/or other materials provided with the distribution.
                     44:  * 3. Neither the name of the project nor the names of its contributors
                     45:  *    may be used to endorse or promote products derived from this software
                     46:  *    without specific prior written permission.
1.154     perry      47:  *
1.49      itojun     48:  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
                     49:  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
                     50:  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
                     51:  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
                     52:  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
                     53:  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
                     54:  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
                     55:  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
                     56:  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
                     57:  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
                     58:  * SUCH DAMAGE.
                     59:  */
1.16      cgd        60:
1.1       cgd        61: /*
1.15      mycroft    62:  * Copyright (c) 1980, 1986, 1993
                     63:  *     The Regents of the University of California.  All rights reserved.
1.1       cgd        64:  *
                     65:  * Redistribution and use in source and binary forms, with or without
                     66:  * modification, are permitted provided that the following conditions
                     67:  * are met:
                     68:  * 1. Redistributions of source code must retain the above copyright
                     69:  *    notice, this list of conditions and the following disclaimer.
                     70:  * 2. Redistributions in binary form must reproduce the above copyright
                     71:  *    notice, this list of conditions and the following disclaimer in the
                     72:  *    documentation and/or other materials provided with the distribution.
1.126     agc        73:  * 3. Neither the name of the University nor the names of its contributors
1.1       cgd        74:  *    may be used to endorse or promote products derived from this software
                     75:  *    without specific prior written permission.
                     76:  *
                     77:  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
                     78:  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
                     79:  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
                     80:  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
                     81:  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
                     82:  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
                     83:  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
                     84:  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
                     85:  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
                     86:  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
                     87:  * SUCH DAMAGE.
                     88:  *
1.44      fvdl       89:  *     @(#)if.c        8.5 (Berkeley) 1/9/95
1.1       cgd        90:  */
1.99      lukem      91:
                     92: #include <sys/cdefs.h>
1.423   ! ozaki-r    93: __KERNEL_RCSID(0, "$NetBSD: if.c,v 1.422 2018/05/14 02:55:03 ozaki-r Exp $");
1.50      thorpej    94:
1.308     ozaki-r    95: #if defined(_KERNEL_OPT)
1.50      thorpej    96: #include "opt_inet.h"
1.352     knakahar   97: #include "opt_ipsec.h"
1.51      bouyer     98: #include "opt_atalk.h"
1.120     martin     99: #include "opt_natm.h"
1.288     ozaki-r   100: #include "opt_wlan.h"
1.308     ozaki-r   101: #include "opt_net_mpsafe.h"
1.417     ozaki-r   102: #include "opt_mrouting.h"
1.308     ozaki-r   103: #endif
1.1       cgd       104:
1.8       mycroft   105: #include <sys/param.h>
                    106: #include <sys/mbuf.h>
                    107: #include <sys/systm.h>
1.59      thorpej   108: #include <sys/callout.h>
1.15      mycroft   109: #include <sys/proc.h>
1.8       mycroft   110: #include <sys/socket.h>
                    111: #include <sys/socketvar.h>
1.56      thorpej   112: #include <sys/domain.h>
1.8       mycroft   113: #include <sys/protosw.h>
                    114: #include <sys/kernel.h>
                    115: #include <sys/ioctl.h>
1.133     jonathan  116: #include <sys/sysctl.h>
1.159     dyoung    117: #include <sys/syslog.h>
1.165     elad      118: #include <sys/kauth.h>
1.254     dyoung    119: #include <sys/kmem.h>
1.276     rmind     120: #include <sys/xcall.h>
1.323     ozaki-r   121: #include <sys/cpu.h>
                    122: #include <sys/intr.h>
1.1       cgd       123:
1.8       mycroft   124: #include <net/if.h>
                    125: #include <net/if_dl.h>
1.66      onoe      126: #include <net/if_ether.h>
1.124     dyoung    127: #include <net/if_media.h>
1.132     dyoung    128: #include <net80211/ieee80211.h>
                    129: #include <net80211/ieee80211_ioctl.h>
1.8       mycroft   130: #include <net/if_types.h>
1.53      thorpej   131: #include <net/route.h>
1.95      itojun    132: #include <net/netisr.h>
1.262     christos  133: #include <sys/module.h>
1.51      bouyer    134: #ifdef NETATALK
                    135: #include <netatalk/at_extern.h>
                    136: #include <netatalk/at.h>
                    137: #endif
1.143     itojun    138: #include <net/pfil.h>
1.278     he        139: #include <netinet/in.h>
1.276     rmind     140: #include <netinet/in_var.h>
1.352     knakahar  141: #include <netinet/ip_encap.h>
1.368     ozaki-r   142: #include <net/bpf.h>
1.1       cgd       143:
1.49      itojun    144: #ifdef INET6
1.72      thorpej   145: #include <netinet6/in6_var.h>
1.108     itojun    146: #include <netinet6/nd6.h>
1.49      itojun    147: #endif
1.117     thorpej   148:
1.288     ozaki-r   149: #include "ether.h"
                    150: #include "fddi.h"
                    151: #include "token.h"
                    152:
1.166     liamjfoy  153: #include "carp.h"
                    154: #if NCARP > 0
                    155: #include <netinet/ip_carp.h>
                    156: #endif
                    157:
1.186     christos  158: #include <compat/sys/sockio.h>
1.161     christos  159: #include <compat/sys/socket.h>
                    160:
1.117     thorpej   161: MALLOC_DEFINE(M_IFADDR, "ifaddr", "interface address");
                    162: MALLOC_DEFINE(M_IFMADDR, "ether_multi", "link-level multicast address");
1.49      itojun    163:
1.274     rmind     164: /*
                    165:  * Global list of interfaces.
                    166:  */
1.334     ozaki-r   167: /* DEPRECATED. Remove it once kvm(3) users disappeared */
1.274     rmind     168: struct ifnet_head              ifnet_list;
1.334     ozaki-r   169:
                    170: struct pslist_head             ifnet_pslist;
1.274     rmind     171: static ifnet_t **              ifindex2ifnet = NULL;
                    172: static u_int                   if_index = 1;
                    173: static size_t                  if_indexlim = 0;
                    174: static uint64_t                        index_gen;
1.334     ozaki-r   175: /* Mutex to protect the above objects. */
                    176: kmutex_t                       ifnet_mtx __cacheline_aligned;
1.385     ozaki-r   177: static struct psref_class      *ifnet_psref_class __read_mostly;
1.334     ozaki-r   178: static pserialize_t            ifnet_psz;
                    179:
1.292     christos  180: static kmutex_t                        if_clone_mtx;
1.274     rmind     181:
                    182: struct ifnet *lo0ifp;
1.1       cgd       183: int    ifqmaxlen = IFQ_MAXLEN;
1.104     matt      184:
1.357     ozaki-r   185: struct psref_class             *ifa_psref_class __read_mostly;
                    186:
1.362     ozaki-r   187: static int     if_delroute_matcher(struct rtentry *, void *);
1.53      thorpej   188:
1.379     knakahar  189: static bool if_is_unit(const char *);
1.163     thorpej   190: static struct if_clone *if_clone_lookup(const char *, int *);
1.63      thorpej   191:
1.163     thorpej   192: static LIST_HEAD(, if_clone) if_cloners = LIST_HEAD_INITIALIZER(if_cloners);
                    193: static int if_cloners_count;
1.63      thorpej   194:
1.265     rmind     195: /* Packet filtering hook for interfaces. */
1.369     christos  196: pfil_head_t *                  if_pfil __read_mostly;
1.143     itojun    197:
1.239     elad      198: static kauth_listener_t if_listener;
                    199:
1.273     pooka     200: static int doifioctl(struct socket *, u_long, void *, struct lwp *);
1.163     thorpej   201: static void if_detach_queues(struct ifnet *, struct ifqueue *);
1.234     dyoung    202: static void sysctl_sndq_setup(struct sysctllog **, const char *,
                    203:     struct ifaltq *);
1.294     ozaki-r   204: static void if_slowtimo(void *);
1.300     ozaki-r   205: static void if_free_sadl(struct ifnet *);
1.302     ozaki-r   206: static void if_attachdomain1(struct ifnet *);
                    207: static int ifconf(u_long, void *);
1.345     knakahar  208: static int if_transmit(struct ifnet *, struct mbuf *);
1.302     ozaki-r   209: static int if_clone_create(const char *);
                    210: static int if_clone_destroy(const char *);
1.324     ozaki-r   211: static void if_link_state_change_si(void *);
1.403     ozaki-r   212: static void if_up_locked(struct ifnet *);
                    213: static void _if_down(struct ifnet *);
                    214: static void if_down_deactivated(struct ifnet *);
1.95      itojun    215:
1.323     ozaki-r   216: struct if_percpuq {
                    217:        struct ifnet    *ipq_ifp;
                    218:        void            *ipq_si;
                    219:        struct percpu   *ipq_ifqs;      /* struct ifqueue */
                    220: };
                    221:
                    222: static struct mbuf *if_percpuq_dequeue(struct if_percpuq *);
                    223:
1.327     knakahar  224: static void if_percpuq_drops(void *, void *, struct cpu_info *);
                    225: static int sysctl_percpuq_drops_handler(SYSCTLFN_PROTO);
                    226: static void sysctl_percpuq_setup(struct sysctllog **, const char *,
                    227:     struct if_percpuq *);
                    228:
1.364     ozaki-r   229: struct if_deferred_start {
                    230:        struct ifnet    *ids_ifp;
                    231:        void            (*ids_if_start)(struct ifnet *);
                    232:        void            *ids_si;
                    233: };
                    234:
                    235: static void if_deferred_start_softint(void *);
                    236: static void if_deferred_start_common(struct ifnet *);
                    237: static void if_deferred_start_destroy(struct ifnet *);
                    238:
1.240     cegger    239: #if defined(INET) || defined(INET6)
1.276     rmind     240: static void sysctl_net_pktq_setup(struct sysctllog **, int);
1.240     cegger    241: #endif
1.237     pooka     242:
1.370     ozaki-r   243: static void if_sysctl_setup(struct sysctllog **);
                    244:
1.420     christos  245: /* Compatibility vector functions */
                    246: u_long (*vec_compat_cvtcmd)(u_long) = NULL;
1.361     pgoyette  247: int (*vec_compat_ifioctl)(struct socket *, u_long, u_long, void *,
                    248:        struct lwp *) = NULL;
1.420     christos  249: int (*vec_compat_ifconf)(struct lwp *, u_long, void *) = (void *)enosys;
                    250: int (*vec_compat_ifdatareq)(struct lwp *, u_long, void *) = (void *)enosys;
1.361     pgoyette  251:
1.239     elad      252: static int
                    253: if_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
                    254:     void *arg0, void *arg1, void *arg2, void *arg3)
                    255: {
                    256:        int result;
                    257:        enum kauth_network_req req;
                    258:
                    259:        result = KAUTH_RESULT_DEFER;
                    260:        req = (enum kauth_network_req)arg1;
                    261:
                    262:        if (action != KAUTH_NETWORK_INTERFACE)
                    263:                return result;
                    264:
                    265:        if ((req == KAUTH_REQ_NETWORK_INTERFACE_GET) ||
                    266:            (req == KAUTH_REQ_NETWORK_INTERFACE_SET))
                    267:                result = KAUTH_RESULT_ALLOW;
                    268:
                    269:        return result;
                    270: }
                    271:
1.1       cgd       272: /*
                    273:  * Network interface utility routines.
                    274:  *
                    275:  * Routines with ifa_ifwith* names take sockaddr *'s as
                    276:  * parameters.
                    277:  */
1.4       andrew    278: void
1.163     thorpej   279: ifinit(void)
1.1       cgd       280: {
1.370     ozaki-r   281:
                    282:        if_sysctl_setup(NULL);
1.1       cgd       283:
1.392     ozaki-r   284: #if (defined(INET) || defined(INET6))
1.352     knakahar  285:        encapinit();
                    286: #endif
                    287:
1.239     elad      288:        if_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
                    289:            if_listener_cb, NULL);
1.273     pooka     290:
                    291:        /* interfaces are available, inform socket code */
                    292:        ifioctl = doifioctl;
1.227     yamt      293: }
                    294:
                    295: /*
                    296:  * XXX Initialization before configure().
                    297:  * XXX hack to get pfil_add_hook working in autoconf.
                    298:  */
                    299: void
                    300: ifinit1(void)
                    301: {
1.292     christos  302:        mutex_init(&if_clone_mtx, MUTEX_DEFAULT, IPL_NONE);
1.357     ozaki-r   303:
1.274     rmind     304:        TAILQ_INIT(&ifnet_list);
1.334     ozaki-r   305:        mutex_init(&ifnet_mtx, MUTEX_DEFAULT, IPL_NONE);
                    306:        ifnet_psz = pserialize_create();
                    307:        ifnet_psref_class = psref_class_create("ifnet", IPL_SOFTNET);
1.357     ozaki-r   308:        ifa_psref_class = psref_class_create("ifa", IPL_SOFTNET);
1.334     ozaki-r   309:        PSLIST_INIT(&ifnet_pslist);
1.357     ozaki-r   310:
1.274     rmind     311:        if_indexlim = 8;
                    312:
1.265     rmind     313:        if_pfil = pfil_head_create(PFIL_TYPE_IFNET, NULL);
                    314:        KASSERT(if_pfil != NULL);
1.286     ozaki-r   315:
1.288     ozaki-r   316: #if NETHER > 0 || NFDDI > 0 || defined(NETATALK) || NTOKEN > 0 || defined(WLAN)
1.286     ozaki-r   317:        etherinit();
1.288     ozaki-r   318: #endif
1.1       cgd       319: }
                    320:
1.274     rmind     321: ifnet_t *
1.226     christos  322: if_alloc(u_char type)
                    323: {
1.274     rmind     324:        return kmem_zalloc(sizeof(ifnet_t), KM_SLEEP);
1.226     christos  325: }
                    326:
                    327: void
1.274     rmind     328: if_free(ifnet_t *ifp)
1.251     dyoung    329: {
1.274     rmind     330:        kmem_free(ifp, sizeof(ifnet_t));
1.251     dyoung    331: }
                    332:
                    333: void
1.226     christos  334: if_initname(struct ifnet *ifp, const char *name, int unit)
                    335: {
                    336:        (void)snprintf(ifp->if_xname, sizeof(ifp->if_xname),
                    337:            "%s%d", name, unit);
                    338: }
                    339:
1.53      thorpej   340: /*
                    341:  * Null routines used while an interface is going away.  These routines
                    342:  * just return an error.
                    343:  */
                    344:
                    345: int
1.177     christos  346: if_nulloutput(struct ifnet *ifp, struct mbuf *m,
1.331     ozaki-r   347:     const struct sockaddr *so, const struct rtentry *rt)
1.53      thorpej   348: {
                    349:
1.185     dyoung    350:        return ENXIO;
1.53      thorpej   351: }
                    352:
                    353: void
1.177     christos  354: if_nullinput(struct ifnet *ifp, struct mbuf *m)
1.53      thorpej   355: {
                    356:
                    357:        /* Nothing. */
                    358: }
                    359:
                    360: void
1.177     christos  361: if_nullstart(struct ifnet *ifp)
1.53      thorpej   362: {
                    363:
                    364:        /* Nothing. */
                    365: }
                    366:
                    367: int
1.332     knakahar  368: if_nulltransmit(struct ifnet *ifp, struct mbuf *m)
                    369: {
                    370:
1.383     knakahar  371:        m_freem(m);
1.332     knakahar  372:        return ENXIO;
                    373: }
                    374:
                    375: int
1.183     christos  376: if_nullioctl(struct ifnet *ifp, u_long cmd, void *data)
1.53      thorpej   377: {
                    378:
1.185     dyoung    379:        return ENXIO;
1.53      thorpej   380: }
                    381:
                    382: int
1.177     christos  383: if_nullinit(struct ifnet *ifp)
1.53      thorpej   384: {
                    385:
1.185     dyoung    386:        return ENXIO;
1.53      thorpej   387: }
                    388:
                    389: void
1.177     christos  390: if_nullstop(struct ifnet *ifp, int disable)
1.75      thorpej   391: {
                    392:
                    393:        /* Nothing. */
                    394: }
                    395:
                    396: void
1.295     ozaki-r   397: if_nullslowtimo(struct ifnet *ifp)
1.53      thorpej   398: {
                    399:
                    400:        /* Nothing. */
                    401: }
                    402:
                    403: void
1.177     christos  404: if_nulldrain(struct ifnet *ifp)
1.53      thorpej   405: {
                    406:
                    407:        /* Nothing. */
                    408: }
                    409:
1.210     dyoung    410: void
1.231     dyoung    411: if_set_sadl(struct ifnet *ifp, const void *lla, u_char addrlen, bool factory)
1.210     dyoung    412: {
                    413:        struct ifaddr *ifa;
                    414:        struct sockaddr_dl *sdl;
                    415:
                    416:        ifp->if_addrlen = addrlen;
                    417:        if_alloc_sadl(ifp);
                    418:        ifa = ifp->if_dl;
                    419:        sdl = satosdl(ifa->ifa_addr);
                    420:
                    421:        (void)sockaddr_dl_setaddr(sdl, sdl->sdl_len, lla, ifp->if_addrlen);
1.231     dyoung    422:        if (factory) {
                    423:                ifp->if_hwdl = ifp->if_dl;
1.291     rmind     424:                ifaref(ifp->if_hwdl);
1.231     dyoung    425:        }
1.223     dyoung    426:        /* TBD routing socket */
1.210     dyoung    427: }
                    428:
1.211     dyoung    429: struct ifaddr *
                    430: if_dl_create(const struct ifnet *ifp, const struct sockaddr_dl **sdlp)
                    431: {
                    432:        unsigned socksize, ifasize;
                    433:        int addrlen, namelen;
                    434:        struct sockaddr_dl *mask, *sdl;
                    435:        struct ifaddr *ifa;
                    436:
                    437:        namelen = strlen(ifp->if_xname);
                    438:        addrlen = ifp->if_addrlen;
                    439:        socksize = roundup(sockaddr_dl_measure(namelen, addrlen), sizeof(long));
                    440:        ifasize = sizeof(*ifa) + 2 * socksize;
1.398     christos  441:        ifa = malloc(ifasize, M_IFADDR, M_WAITOK|M_ZERO);
1.211     dyoung    442:
                    443:        sdl = (struct sockaddr_dl *)(ifa + 1);
                    444:        mask = (struct sockaddr_dl *)(socksize + (char *)sdl);
                    445:
                    446:        sockaddr_dl_init(sdl, socksize, ifp->if_index, ifp->if_type,
                    447:            ifp->if_xname, namelen, NULL, addrlen);
1.398     christos  448:        mask->sdl_family = AF_LINK;
1.211     dyoung    449:        mask->sdl_len = sockaddr_dl_measure(namelen, 0);
                    450:        memset(&mask->sdl_data[0], 0xff, namelen);
                    451:        ifa->ifa_rtrequest = link_rtrequest;
                    452:        ifa->ifa_addr = (struct sockaddr *)sdl;
                    453:        ifa->ifa_netmask = (struct sockaddr *)mask;
1.357     ozaki-r   454:        ifa_psref_init(ifa);
1.211     dyoung    455:
                    456:        *sdlp = sdl;
                    457:
                    458:        return ifa;
                    459: }
                    460:
1.223     dyoung    461: static void
                    462: if_sadl_setrefs(struct ifnet *ifp, struct ifaddr *ifa)
                    463: {
                    464:        const struct sockaddr_dl *sdl;
1.316     ozaki-r   465:
1.223     dyoung    466:        ifp->if_dl = ifa;
1.291     rmind     467:        ifaref(ifa);
1.223     dyoung    468:        sdl = satosdl(ifa->ifa_addr);
                    469:        ifp->if_sadl = sdl;
                    470: }
                    471:
1.1       cgd       472: /*
1.81      thorpej   473:  * Allocate the link level name for the specified interface.  This
                    474:  * is an attachment helper.  It must be called after ifp->if_addrlen
                    475:  * is initialized, which may not be the case when if_attach() is
                    476:  * called.
                    477:  */
                    478: void
                    479: if_alloc_sadl(struct ifnet *ifp)
                    480: {
                    481:        struct ifaddr *ifa;
1.211     dyoung    482:        const struct sockaddr_dl *sdl;
1.84      thorpej   483:
                    484:        /*
                    485:         * If the interface already has a link name, release it
                    486:         * now.  This is useful for interfaces that can change
                    487:         * link types, and thus switch link names often.
                    488:         */
                    489:        if (ifp->if_sadl != NULL)
                    490:                if_free_sadl(ifp);
1.81      thorpej   491:
1.211     dyoung    492:        ifa = if_dl_create(ifp, &sdl);
1.195     dyoung    493:
1.207     dyoung    494:        ifa_insert(ifp, ifa);
1.223     dyoung    495:        if_sadl_setrefs(ifp, ifa);
                    496: }
                    497:
                    498: static void
                    499: if_deactivate_sadl(struct ifnet *ifp)
                    500: {
                    501:        struct ifaddr *ifa;
                    502:
                    503:        KASSERT(ifp->if_dl != NULL);
                    504:
                    505:        ifa = ifp->if_dl;
                    506:
                    507:        ifp->if_sadl = NULL;
                    508:
                    509:        ifp->if_dl = NULL;
1.291     rmind     510:        ifafree(ifa);
1.223     dyoung    511: }
                    512:
1.378     ozaki-r   513: static void
                    514: if_replace_sadl(struct ifnet *ifp, struct ifaddr *ifa)
                    515: {
                    516:        struct ifaddr *old;
                    517:
                    518:        KASSERT(ifp->if_dl != NULL);
                    519:
                    520:        old = ifp->if_dl;
                    521:
                    522:        ifaref(ifa);
                    523:        /* XXX Update if_dl and if_sadl atomically */
                    524:        ifp->if_dl = ifa;
                    525:        ifp->if_sadl = satosdl(ifa->ifa_addr);
                    526:
                    527:        ifafree(old);
                    528: }
                    529:
1.224     dyoung    530: void
1.357     ozaki-r   531: if_activate_sadl(struct ifnet *ifp, struct ifaddr *ifa0,
1.223     dyoung    532:     const struct sockaddr_dl *sdl)
                    533: {
1.357     ozaki-r   534:        int s, ss;
                    535:        struct ifaddr *ifa;
                    536:        int bound = curlwp_bind();
1.223     dyoung    537:
1.378     ozaki-r   538:        KASSERT(ifa_held(ifa0));
                    539:
1.373     ozaki-r   540:        s = splsoftnet();
1.223     dyoung    541:
1.378     ozaki-r   542:        if_replace_sadl(ifp, ifa0);
1.357     ozaki-r   543:
                    544:        ss = pserialize_read_enter();
                    545:        IFADDR_READER_FOREACH(ifa, ifp) {
                    546:                struct psref psref;
                    547:                ifa_acquire(ifa, &psref);
                    548:                pserialize_read_exit(ss);
                    549:
1.231     dyoung    550:                rtinit(ifa, RTM_LLINFO_UPD, 0);
1.354     ozaki-r   551:
1.357     ozaki-r   552:                ss = pserialize_read_enter();
                    553:                ifa_release(ifa, &psref);
                    554:        }
                    555:        pserialize_read_exit(ss);
                    556:
1.223     dyoung    557:        splx(s);
1.357     ozaki-r   558:        curlwp_bindx(bound);
1.81      thorpej   559: }
                    560:
                    561: /*
                    562:  * Free the link level name for the specified interface.  This is
1.300     ozaki-r   563:  * a detach helper.  This is called from if_detach().
1.81      thorpej   564:  */
1.300     ozaki-r   565: static void
1.81      thorpej   566: if_free_sadl(struct ifnet *ifp)
                    567: {
                    568:        struct ifaddr *ifa;
                    569:        int s;
                    570:
1.316     ozaki-r   571:        ifa = ifp->if_dl;
1.81      thorpej   572:        if (ifa == NULL) {
                    573:                KASSERT(ifp->if_sadl == NULL);
                    574:                return;
                    575:        }
                    576:
                    577:        KASSERT(ifp->if_sadl != NULL);
                    578:
1.373     ozaki-r   579:        s = splsoftnet();
1.81      thorpej   580:        rtinit(ifa, RTM_DELETE, 0);
1.207     dyoung    581:        ifa_remove(ifp, ifa);
1.223     dyoung    582:        if_deactivate_sadl(ifp);
1.231     dyoung    583:        if (ifp->if_hwdl == ifa) {
1.291     rmind     584:                ifafree(ifa);
1.231     dyoung    585:                ifp->if_hwdl = NULL;
                    586:        }
1.81      thorpej   587:        splx(s);
                    588: }
                    589:
1.274     rmind     590: static void
                    591: if_getindex(ifnet_t *ifp)
1.1       cgd       592: {
1.274     rmind     593:        bool hitlimit = false;
1.231     dyoung    594:
1.234     dyoung    595:        ifp->if_index_gen = index_gen++;
                    596:
1.102     atatat    597:        ifp->if_index = if_index;
1.274     rmind     598:        if (ifindex2ifnet == NULL) {
1.102     atatat    599:                if_index++;
1.274     rmind     600:                goto skip;
                    601:        }
                    602:        while (if_byindex(ifp->if_index)) {
                    603:                /*
                    604:                 * If we hit USHRT_MAX, we skip back to 0 since
                    605:                 * there are a number of places where the value
                    606:                 * of if_index or if_index itself is compared
                    607:                 * to or stored in an unsigned short.  By
                    608:                 * jumping back, we won't botch those assignments
                    609:                 * or comparisons.
                    610:                 */
                    611:                if (++if_index == 0) {
                    612:                        if_index = 1;
                    613:                } else if (if_index == USHRT_MAX) {
1.102     atatat    614:                        /*
1.274     rmind     615:                         * However, if we have to jump back to
                    616:                         * zero *twice* without finding an empty
                    617:                         * slot in ifindex2ifnet[], then there
                    618:                         * there are too many (>65535) interfaces.
1.102     atatat    619:                         */
1.274     rmind     620:                        if (hitlimit) {
                    621:                                panic("too many interfaces");
1.102     atatat    622:                        }
1.274     rmind     623:                        hitlimit = true;
                    624:                        if_index = 1;
1.102     atatat    625:                }
1.274     rmind     626:                ifp->if_index = if_index;
                    627:        }
                    628: skip:
1.49      itojun    629:        /*
1.316     ozaki-r   630:         * ifindex2ifnet is indexed by if_index. Since if_index will
                    631:         * grow dynamically, it should grow too.
1.49      itojun    632:         */
1.316     ozaki-r   633:        if (ifindex2ifnet == NULL || ifp->if_index >= if_indexlim) {
1.131     itojun    634:                size_t m, n, oldlim;
1.183     christos  635:                void *q;
1.154     perry     636:
1.131     itojun    637:                oldlim = if_indexlim;
1.53      thorpej   638:                while (ifp->if_index >= if_indexlim)
1.49      itojun    639:                        if_indexlim <<= 1;
                    640:
                    641:                /* grow ifindex2ifnet */
1.131     itojun    642:                m = oldlim * sizeof(struct ifnet *);
1.49      itojun    643:                n = if_indexlim * sizeof(struct ifnet *);
1.230     dyoung    644:                q = malloc(n, M_IFADDR, M_WAITOK|M_ZERO);
1.185     dyoung    645:                if (ifindex2ifnet != NULL) {
1.230     dyoung    646:                        memcpy(q, ifindex2ifnet, m);
                    647:                        free(ifindex2ifnet, M_IFADDR);
1.49      itojun    648:                }
                    649:                ifindex2ifnet = (struct ifnet **)q;
1.1       cgd       650:        }
1.274     rmind     651:        ifindex2ifnet[ifp->if_index] = ifp;
                    652: }
                    653:
                    654: /*
1.307     ozaki-r   655:  * Initialize an interface and assign an index for it.
                    656:  *
                    657:  * It must be called prior to a device specific attach routine
                    658:  * (e.g., ether_ifattach and ieee80211_ifattach) or if_alloc_sadl,
                    659:  * and be followed by if_register:
                    660:  *
                    661:  *     if_initialize(ifp);
                    662:  *     ether_ifattach(ifp, enaddr);
                    663:  *     if_register(ifp);
1.274     rmind     664:  */
1.396     msaitoh   665: int
1.307     ozaki-r   666: if_initialize(ifnet_t *ifp)
1.274     rmind     667: {
1.396     msaitoh   668:        int rv = 0;
                    669:
1.274     rmind     670:        KASSERT(if_indexlim > 0);
                    671:        TAILQ_INIT(&ifp->if_addrlist);
1.49      itojun    672:
1.1       cgd       673:        /*
1.81      thorpej   674:         * Link level name is allocated later by a separate call to
                    675:         * if_alloc_sadl().
1.1       cgd       676:         */
1.81      thorpej   677:
1.40      thorpej   678:        if (ifp->if_snd.ifq_maxlen == 0)
1.94      itojun    679:                ifp->if_snd.ifq_maxlen = ifqmaxlen;
1.234     dyoung    680:
1.42      is        681:        ifp->if_broadcastaddr = 0; /* reliably crash if used uninitialized */
1.57      thorpej   682:
                    683:        ifp->if_link_state = LINK_STATE_UNKNOWN;
1.325     roy       684:        ifp->if_link_queue = -1; /* all bits set, see link_state_change() */
1.57      thorpej   685:
1.89      thorpej   686:        ifp->if_capenable = 0;
1.97      thorpej   687:        ifp->if_csum_flags_tx = 0;
                    688:        ifp->if_csum_flags_rx = 0;
1.89      thorpej   689:
1.86      thorpej   690: #ifdef ALTQ
                    691:        ifp->if_snd.altq_type = 0;
                    692:        ifp->if_snd.altq_disc = NULL;
                    693:        ifp->if_snd.altq_flags &= ALTQF_CANTCHANGE;
                    694:        ifp->if_snd.altq_tbr  = NULL;
                    695:        ifp->if_snd.altq_ifp  = ifp;
                    696: #endif
                    697:
1.358     ozaki-r   698:        IFQ_LOCK_INIT(&ifp->if_snd);
1.285     ozaki-r   699:
1.265     rmind     700:        ifp->if_pfil = pfil_head_create(PFIL_TYPE_IFNET, ifp);
1.369     christos  701:        pfil_run_ifhooks(if_pfil, PFIL_IFNET_ATTACH, ifp);
1.87      thorpej   702:
1.318     ozaki-r   703:        IF_AFDATA_LOCK_INIT(ifp);
                    704:
1.347     knakahar  705:        if (if_is_link_state_changeable(ifp)) {
1.406     ozaki-r   706:                u_int flags = SOFTINT_NET;
1.421     ozaki-r   707:                flags |= if_is_mpsafe(ifp) ? SOFTINT_MPSAFE : 0;
1.406     ozaki-r   708:                ifp->if_link_si = softint_establish(flags,
1.346     knakahar  709:                    if_link_state_change_si, ifp);
1.396     msaitoh   710:                if (ifp->if_link_si == NULL) {
                    711:                        rv = ENOMEM;
                    712:                        goto fail;
                    713:                }
1.346     knakahar  714:        }
1.324     ozaki-r   715:
1.334     ozaki-r   716:        PSLIST_ENTRY_INIT(ifp, if_pslist_entry);
1.354     ozaki-r   717:        PSLIST_INIT(&ifp->if_addr_pslist);
1.334     ozaki-r   718:        psref_target_init(&ifp->if_psref, ifnet_psref_class);
1.336     ozaki-r   719:        ifp->if_ioctl_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
1.381     ozaki-r   720:        LIST_INIT(&ifp->if_multiaddrs);
1.334     ozaki-r   721:
1.412     ozaki-r   722:        IFNET_GLOBAL_LOCK();
1.307     ozaki-r   723:        if_getindex(ifp);
1.412     ozaki-r   724:        IFNET_GLOBAL_UNLOCK();
1.396     msaitoh   725:
                    726:        return 0;
                    727:
                    728: fail:
                    729:        IF_AFDATA_LOCK_DESTROY(ifp);
                    730:
                    731:        pfil_run_ifhooks(if_pfil, PFIL_IFNET_DETACH, ifp);
                    732:        (void)pfil_head_destroy(ifp->if_pfil);
                    733:
                    734:        IFQ_LOCK_DESTROY(&ifp->if_snd);
                    735:
                    736:        return rv;
1.307     ozaki-r   737: }
                    738:
                    739: /*
                    740:  * Register an interface to the list of "active" interfaces.
                    741:  */
                    742: void
                    743: if_register(ifnet_t *ifp)
                    744: {
1.336     ozaki-r   745:        /*
                    746:         * If the driver has not supplied its own if_ioctl, then
                    747:         * supply the default.
                    748:         */
                    749:        if (ifp->if_ioctl == NULL)
                    750:                ifp->if_ioctl = ifioctl_common;
1.307     ozaki-r   751:
                    752:        sysctl_sndq_setup(&ifp->if_sysctl_log, ifp->if_xname, &ifp->if_snd);
                    753:
1.152     matt      754:        if (!STAILQ_EMPTY(&domains))
1.147     tron      755:                if_attachdomain1(ifp);
                    756:
1.107     itojun    757:        /* Announce the interface. */
                    758:        rt_ifannouncemsg(ifp, IFAN_ARRIVAL);
1.296     ozaki-r   759:
                    760:        if (ifp->if_slowtimo != NULL) {
1.297     ozaki-r   761:                ifp->if_slowtimo_ch =
                    762:                    kmem_zalloc(sizeof(*ifp->if_slowtimo_ch), KM_SLEEP);
1.411     ozaki-r   763:                callout_init(ifp->if_slowtimo_ch, 0);
1.297     ozaki-r   764:                callout_setfunc(ifp->if_slowtimo_ch, if_slowtimo, ifp);
1.296     ozaki-r   765:                if_slowtimo(ifp);
                    766:        }
1.307     ozaki-r   767:
1.332     knakahar  768:        if (ifp->if_transmit == NULL || ifp->if_transmit == if_nulltransmit)
                    769:                ifp->if_transmit = if_transmit;
                    770:
1.412     ozaki-r   771:        IFNET_GLOBAL_LOCK();
1.307     ozaki-r   772:        TAILQ_INSERT_TAIL(&ifnet_list, ifp, if_list);
1.334     ozaki-r   773:        IFNET_WRITER_INSERT_TAIL(ifp);
1.412     ozaki-r   774:        IFNET_GLOBAL_UNLOCK();
1.307     ozaki-r   775: }
                    776:
                    777: /*
1.323     ozaki-r   778:  * The if_percpuq framework
                    779:  *
                    780:  * It allows network device drivers to execute the network stack
                    781:  * in softint (so called softint-based if_input). It utilizes
                    782:  * softint and percpu ifqueue. It doesn't distribute any packets
                    783:  * between CPUs, unlike pktqueue(9).
                    784:  *
                    785:  * Currently we support two options for device drivers to apply the framework:
                    786:  * - Use it implicitly with less changes
                    787:  *   - If you use if_attach in driver's _attach function and if_input in
                    788:  *     driver's Rx interrupt handler, a packet is queued and a softint handles
                    789:  *     the packet implicitly
                    790:  * - Use it explicitly in each driver (recommended)
                    791:  *   - You can use if_percpuq_* directly in your driver
                    792:  *   - In this case, you need to allocate struct if_percpuq in driver's softc
                    793:  *   - See wm(4) as a reference implementation
                    794:  */
                    795:
                    796: static void
                    797: if_percpuq_softint(void *arg)
                    798: {
                    799:        struct if_percpuq *ipq = arg;
                    800:        struct ifnet *ifp = ipq->ipq_ifp;
                    801:        struct mbuf *m;
                    802:
1.368     ozaki-r   803:        while ((m = if_percpuq_dequeue(ipq)) != NULL) {
                    804:                ifp->if_ipackets++;
                    805:                bpf_mtap(ifp, m);
                    806:
1.323     ozaki-r   807:                ifp->_if_input(ifp, m);
1.368     ozaki-r   808:        }
1.323     ozaki-r   809: }
                    810:
                    811: static void
                    812: if_percpuq_init_ifq(void *p, void *arg __unused, struct cpu_info *ci __unused)
                    813: {
                    814:        struct ifqueue *const ifq = p;
                    815:
                    816:        memset(ifq, 0, sizeof(*ifq));
                    817:        ifq->ifq_maxlen = IFQ_MAXLEN;
                    818: }
                    819:
                    820: struct if_percpuq *
                    821: if_percpuq_create(struct ifnet *ifp)
                    822: {
                    823:        struct if_percpuq *ipq;
1.422     ozaki-r   824:        u_int flags = SOFTINT_NET;
                    825:
                    826:        flags |= if_is_mpsafe(ifp) ? SOFTINT_MPSAFE : 0;
1.323     ozaki-r   827:
                    828:        ipq = kmem_zalloc(sizeof(*ipq), KM_SLEEP);
                    829:        ipq->ipq_ifp = ifp;
1.422     ozaki-r   830:        ipq->ipq_si = softint_establish(flags, if_percpuq_softint, ipq);
1.323     ozaki-r   831:        ipq->ipq_ifqs = percpu_alloc(sizeof(struct ifqueue));
                    832:        percpu_foreach(ipq->ipq_ifqs, &if_percpuq_init_ifq, NULL);
                    833:
1.327     knakahar  834:        sysctl_percpuq_setup(&ifp->if_sysctl_log, ifp->if_xname, ipq);
                    835:
1.323     ozaki-r   836:        return ipq;
                    837: }
                    838:
                    839: static struct mbuf *
                    840: if_percpuq_dequeue(struct if_percpuq *ipq)
                    841: {
                    842:        struct mbuf *m;
                    843:        struct ifqueue *ifq;
                    844:        int s;
                    845:
                    846:        s = splnet();
                    847:        ifq = percpu_getref(ipq->ipq_ifqs);
                    848:        IF_DEQUEUE(ifq, m);
                    849:        percpu_putref(ipq->ipq_ifqs);
                    850:        splx(s);
                    851:
                    852:        return m;
                    853: }
                    854:
                    855: static void
                    856: if_percpuq_purge_ifq(void *p, void *arg __unused, struct cpu_info *ci __unused)
                    857: {
                    858:        struct ifqueue *const ifq = p;
                    859:
                    860:        IF_PURGE(ifq);
                    861: }
                    862:
                    863: void
                    864: if_percpuq_destroy(struct if_percpuq *ipq)
                    865: {
                    866:
                    867:        /* if_detach may already destroy it */
                    868:        if (ipq == NULL)
                    869:                return;
                    870:
                    871:        softint_disestablish(ipq->ipq_si);
                    872:        percpu_foreach(ipq->ipq_ifqs, &if_percpuq_purge_ifq, NULL);
                    873:        percpu_free(ipq->ipq_ifqs, sizeof(struct ifqueue));
1.363     ozaki-r   874:        kmem_free(ipq, sizeof(*ipq));
1.323     ozaki-r   875: }
                    876:
                    877: void
                    878: if_percpuq_enqueue(struct if_percpuq *ipq, struct mbuf *m)
                    879: {
                    880:        struct ifqueue *ifq;
                    881:        int s;
                    882:
                    883:        KASSERT(ipq != NULL);
                    884:
                    885:        s = splnet();
                    886:        ifq = percpu_getref(ipq->ipq_ifqs);
                    887:        if (IF_QFULL(ifq)) {
                    888:                IF_DROP(ifq);
1.326     ozaki-r   889:                percpu_putref(ipq->ipq_ifqs);
1.323     ozaki-r   890:                m_freem(m);
                    891:                goto out;
                    892:        }
                    893:        IF_ENQUEUE(ifq, m);
                    894:        percpu_putref(ipq->ipq_ifqs);
                    895:
                    896:        softint_schedule(ipq->ipq_si);
                    897: out:
                    898:        splx(s);
                    899: }
                    900:
1.327     knakahar  901: static void
                    902: if_percpuq_drops(void *p, void *arg, struct cpu_info *ci __unused)
                    903: {
                    904:        struct ifqueue *const ifq = p;
                    905:        int *sum = arg;
                    906:
                    907:        *sum += ifq->ifq_drops;
                    908: }
                    909:
                    910: static int
                    911: sysctl_percpuq_drops_handler(SYSCTLFN_ARGS)
                    912: {
                    913:        struct sysctlnode node;
                    914:        struct if_percpuq *ipq;
                    915:        int sum = 0;
                    916:        int error;
                    917:
                    918:        node = *rnode;
                    919:        ipq = node.sysctl_data;
                    920:
                    921:        percpu_foreach(ipq->ipq_ifqs, if_percpuq_drops, &sum);
                    922:
                    923:        node.sysctl_data = &sum;
                    924:        error = sysctl_lookup(SYSCTLFN_CALL(&node));
                    925:        if (error != 0 || newp == NULL)
                    926:                return error;
                    927:
                    928:        return 0;
                    929: }
                    930:
                    931: static void
                    932: sysctl_percpuq_setup(struct sysctllog **clog, const char* ifname,
                    933:     struct if_percpuq *ipq)
                    934: {
                    935:        const struct sysctlnode *cnode, *rnode;
                    936:
                    937:        if (sysctl_createv(clog, 0, NULL, &rnode,
                    938:                       CTLFLAG_PERMANENT,
                    939:                       CTLTYPE_NODE, "interfaces",
                    940:                       SYSCTL_DESCR("Per-interface controls"),
                    941:                       NULL, 0, NULL, 0,
                    942:                       CTL_NET, CTL_CREATE, CTL_EOL) != 0)
                    943:                goto bad;
                    944:
                    945:        if (sysctl_createv(clog, 0, &rnode, &rnode,
                    946:                       CTLFLAG_PERMANENT,
                    947:                       CTLTYPE_NODE, ifname,
                    948:                       SYSCTL_DESCR("Interface controls"),
                    949:                       NULL, 0, NULL, 0,
                    950:                       CTL_CREATE, CTL_EOL) != 0)
                    951:                goto bad;
                    952:
                    953:        if (sysctl_createv(clog, 0, &rnode, &rnode,
                    954:                       CTLFLAG_PERMANENT,
                    955:                       CTLTYPE_NODE, "rcvq",
                    956:                       SYSCTL_DESCR("Interface input queue controls"),
                    957:                       NULL, 0, NULL, 0,
                    958:                       CTL_CREATE, CTL_EOL) != 0)
                    959:                goto bad;
                    960:
                    961: #ifdef NOTYET
                    962:        /* XXX Should show each per-CPU queue length? */
                    963:        if (sysctl_createv(clog, 0, &rnode, &rnode,
                    964:                       CTLFLAG_PERMANENT,
                    965:                       CTLTYPE_INT, "len",
                    966:                       SYSCTL_DESCR("Current input queue length"),
                    967:                       sysctl_percpuq_len, 0, NULL, 0,
                    968:                       CTL_CREATE, CTL_EOL) != 0)
                    969:                goto bad;
                    970:
                    971:        if (sysctl_createv(clog, 0, &rnode, &cnode,
                    972:                       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
                    973:                       CTLTYPE_INT, "maxlen",
                    974:                       SYSCTL_DESCR("Maximum allowed input queue length"),
                    975:                       sysctl_percpuq_maxlen_handler, 0, (void *)ipq, 0,
                    976:                       CTL_CREATE, CTL_EOL) != 0)
                    977:                goto bad;
                    978: #endif
                    979:
                    980:        if (sysctl_createv(clog, 0, &rnode, &cnode,
                    981:                       CTLFLAG_PERMANENT,
                    982:                       CTLTYPE_INT, "drops",
                    983:                       SYSCTL_DESCR("Total packets dropped due to full input queue"),
                    984:                       sysctl_percpuq_drops_handler, 0, (void *)ipq, 0,
                    985:                       CTL_CREATE, CTL_EOL) != 0)
                    986:                goto bad;
                    987:
                    988:        return;
                    989: bad:
                    990:        printf("%s: could not attach sysctl nodes\n", ifname);
                    991:        return;
                    992: }
                    993:
1.364     ozaki-r   994: /*
                    995:  * The deferred if_start framework
                    996:  *
                    997:  * The common APIs to defer if_start to softint when if_start is requested
                    998:  * from a device driver running in hardware interrupt context.
                    999:  */
                   1000: /*
                   1001:  * Call ifp->if_start (or equivalent) in a dedicated softint for
                   1002:  * deferred if_start.
                   1003:  */
                   1004: static void
                   1005: if_deferred_start_softint(void *arg)
                   1006: {
                   1007:        struct if_deferred_start *ids = arg;
                   1008:        struct ifnet *ifp = ids->ids_ifp;
                   1009:
                   1010:        ids->ids_if_start(ifp);
                   1011: }
                   1012:
                   1013: /*
                   1014:  * The default callback function for deferred if_start.
                   1015:  */
                   1016: static void
                   1017: if_deferred_start_common(struct ifnet *ifp)
                   1018: {
1.382     ozaki-r  1019:        int s;
1.364     ozaki-r  1020:
1.382     ozaki-r  1021:        s = splnet();
1.364     ozaki-r  1022:        if_start_lock(ifp);
1.382     ozaki-r  1023:        splx(s);
1.364     ozaki-r  1024: }
                   1025:
                   1026: static inline bool
                   1027: if_snd_is_used(struct ifnet *ifp)
                   1028: {
                   1029:
                   1030:        return ifp->if_transmit == NULL || ifp->if_transmit == if_nulltransmit ||
                   1031:            ALTQ_IS_ENABLED(&ifp->if_snd);
                   1032: }
                   1033:
                   1034: /*
                   1035:  * Schedule deferred if_start.
                   1036:  */
                   1037: void
                   1038: if_schedule_deferred_start(struct ifnet *ifp)
                   1039: {
                   1040:
                   1041:        KASSERT(ifp->if_deferred_start != NULL);
                   1042:
                   1043:        if (if_snd_is_used(ifp) && IFQ_IS_EMPTY(&ifp->if_snd))
                   1044:                return;
                   1045:
                   1046:        softint_schedule(ifp->if_deferred_start->ids_si);
                   1047: }
                   1048:
                   1049: /*
                   1050:  * Create an instance of deferred if_start. A driver should call the function
                   1051:  * only if the driver needs deferred if_start. Drivers can setup their own
                   1052:  * deferred if_start function via 2nd argument.
                   1053:  */
                   1054: void
                   1055: if_deferred_start_init(struct ifnet *ifp, void (*func)(struct ifnet *))
                   1056: {
                   1057:        struct if_deferred_start *ids;
1.423   ! ozaki-r  1058:        u_int flags = SOFTINT_NET;
        !          1059:
        !          1060:        flags |= if_is_mpsafe(ifp) ? SOFTINT_MPSAFE : 0;
1.364     ozaki-r  1061:
                   1062:        ids = kmem_zalloc(sizeof(*ids), KM_SLEEP);
                   1063:        ids->ids_ifp = ifp;
1.423   ! ozaki-r  1064:        ids->ids_si = softint_establish(flags, if_deferred_start_softint, ids);
1.364     ozaki-r  1065:        if (func != NULL)
                   1066:                ids->ids_if_start = func;
                   1067:        else
                   1068:                ids->ids_if_start = if_deferred_start_common;
                   1069:
                   1070:        ifp->if_deferred_start = ids;
                   1071: }
                   1072:
                   1073: static void
                   1074: if_deferred_start_destroy(struct ifnet *ifp)
                   1075: {
                   1076:
                   1077:        if (ifp->if_deferred_start == NULL)
                   1078:                return;
                   1079:
                   1080:        softint_disestablish(ifp->if_deferred_start->ids_si);
                   1081:        kmem_free(ifp->if_deferred_start, sizeof(*ifp->if_deferred_start));
                   1082:        ifp->if_deferred_start = NULL;
                   1083: }
1.327     knakahar 1084:
1.323     ozaki-r  1085: /*
                   1086:  * The common interface input routine that is called by device drivers,
                   1087:  * which should be used only when the driver's rx handler already runs
                   1088:  * in softint.
                   1089:  */
                   1090: void
                   1091: if_input(struct ifnet *ifp, struct mbuf *m)
                   1092: {
                   1093:
                   1094:        KASSERT(ifp->if_percpuq == NULL);
                   1095:        KASSERT(!cpu_intr_p());
                   1096:
1.368     ozaki-r  1097:        ifp->if_ipackets++;
                   1098:        bpf_mtap(ifp, m);
                   1099:
1.323     ozaki-r  1100:        ifp->_if_input(ifp, m);
                   1101: }
                   1102:
                   1103: /*
                   1104:  * DEPRECATED. Use if_initialize and if_register instead.
1.307     ozaki-r  1105:  * See the above comment of if_initialize.
1.323     ozaki-r  1106:  *
                   1107:  * Note that it implicitly enables if_percpuq to make drivers easy to
1.333     skrll    1108:  * migrate softint-based if_input without much changes. If you don't
1.323     ozaki-r  1109:  * want to enable it, use if_initialize instead.
1.307     ozaki-r  1110:  */
1.396     msaitoh  1111: int
1.307     ozaki-r  1112: if_attach(ifnet_t *ifp)
                   1113: {
1.396     msaitoh  1114:        int rv;
                   1115:
                   1116:        rv = if_initialize(ifp);
                   1117:        if (rv != 0)
                   1118:                return rv;
1.323     ozaki-r  1119:
                   1120:        ifp->if_percpuq = if_percpuq_create(ifp);
1.307     ozaki-r  1121:        if_register(ifp);
1.396     msaitoh  1122:
                   1123:        return 0;
1.107     itojun   1124: }
                   1125:
                   1126: void
1.163     thorpej  1127: if_attachdomain(void)
1.107     itojun   1128: {
                   1129:        struct ifnet *ifp;
1.110     itojun   1130:        int s;
1.339     ozaki-r  1131:        int bound = curlwp_bind();
1.107     itojun   1132:
1.334     ozaki-r  1133:        s = pserialize_read_enter();
                   1134:        IFNET_READER_FOREACH(ifp) {
                   1135:                struct psref psref;
                   1136:                psref_acquire(&psref, &ifp->if_psref, ifnet_psref_class);
                   1137:                pserialize_read_exit(s);
1.107     itojun   1138:                if_attachdomain1(ifp);
1.334     ozaki-r  1139:                s = pserialize_read_enter();
                   1140:                psref_release(&psref, &ifp->if_psref, ifnet_psref_class);
                   1141:        }
                   1142:        pserialize_read_exit(s);
1.339     ozaki-r  1143:        curlwp_bindx(bound);
1.107     itojun   1144: }
                   1145:
1.302     ozaki-r  1146: static void
1.163     thorpej  1147: if_attachdomain1(struct ifnet *ifp)
1.107     itojun   1148: {
                   1149:        struct domain *dp;
1.109     itojun   1150:        int s;
                   1151:
1.373     ozaki-r  1152:        s = splsoftnet();
1.107     itojun   1153:
1.106     itojun   1154:        /* address family dependent data region */
                   1155:        memset(ifp->if_afdata, 0, sizeof(ifp->if_afdata));
1.152     matt     1156:        DOMAIN_FOREACH(dp) {
1.185     dyoung   1157:                if (dp->dom_ifattach != NULL)
1.106     itojun   1158:                        ifp->if_afdata[dp->dom_family] =
                   1159:                            (*dp->dom_ifattach)(ifp);
                   1160:        }
1.109     itojun   1161:
                   1162:        splx(s);
1.1       cgd      1163: }
1.53      thorpej  1164:
                   1165: /*
                   1166:  * Deactivate an interface.  This points all of the procedure
                   1167:  * handles at error stubs.  May be called from interrupt context.
                   1168:  */
                   1169: void
1.163     thorpej  1170: if_deactivate(struct ifnet *ifp)
1.53      thorpej  1171: {
                   1172:        int s;
                   1173:
1.373     ozaki-r  1174:        s = splsoftnet();
1.53      thorpej  1175:
                   1176:        ifp->if_output   = if_nulloutput;
1.323     ozaki-r  1177:        ifp->_if_input   = if_nullinput;
1.53      thorpej  1178:        ifp->if_start    = if_nullstart;
1.332     knakahar 1179:        ifp->if_transmit = if_nulltransmit;
1.53      thorpej  1180:        ifp->if_ioctl    = if_nullioctl;
1.75      thorpej  1181:        ifp->if_init     = if_nullinit;
                   1182:        ifp->if_stop     = if_nullstop;
1.295     ozaki-r  1183:        ifp->if_slowtimo = if_nullslowtimo;
1.53      thorpej  1184:        ifp->if_drain    = if_nulldrain;
                   1185:
                   1186:        /* No more packets may be enqueued. */
                   1187:        ifp->if_snd.ifq_maxlen = 0;
                   1188:
                   1189:        splx(s);
                   1190: }
                   1191:
1.348     ozaki-r  1192: bool
1.367     ozaki-r  1193: if_is_deactivated(const struct ifnet *ifp)
1.348     ozaki-r  1194: {
                   1195:
                   1196:        return ifp->if_output == if_nulloutput;
                   1197: }
                   1198:
1.206     dyoung   1199: void
1.218     dyoung   1200: if_purgeaddrs(struct ifnet *ifp, int family, void (*purgeaddr)(struct ifaddr *))
1.206     dyoung   1201: {
1.289     ozaki-r  1202:        struct ifaddr *ifa, *nifa;
1.357     ozaki-r  1203:        int s;
1.206     dyoung   1204:
1.357     ozaki-r  1205:        s = pserialize_read_enter();
1.354     ozaki-r  1206:        for (ifa = IFADDR_READER_FIRST(ifp); ifa; ifa = nifa) {
                   1207:                nifa = IFADDR_READER_NEXT(ifa);
1.206     dyoung   1208:                if (ifa->ifa_addr->sa_family != family)
                   1209:                        continue;
1.357     ozaki-r  1210:                pserialize_read_exit(s);
                   1211:
1.206     dyoung   1212:                (*purgeaddr)(ifa);
1.357     ozaki-r  1213:
                   1214:                s = pserialize_read_enter();
1.206     dyoung   1215:        }
1.357     ozaki-r  1216:        pserialize_read_exit(s);
1.206     dyoung   1217: }
                   1218:
1.349     ozaki-r  1219: #ifdef IFAREF_DEBUG
                   1220: static struct ifaddr **ifa_list;
                   1221: static int ifa_list_size;
                   1222:
                   1223: /* Depends on only one if_attach runs at once */
                   1224: static void
                   1225: if_build_ifa_list(struct ifnet *ifp)
                   1226: {
                   1227:        struct ifaddr *ifa;
                   1228:        int i;
                   1229:
                   1230:        KASSERT(ifa_list == NULL);
                   1231:        KASSERT(ifa_list_size == 0);
                   1232:
1.354     ozaki-r  1233:        IFADDR_READER_FOREACH(ifa, ifp)
1.349     ozaki-r  1234:                ifa_list_size++;
                   1235:
                   1236:        ifa_list = kmem_alloc(sizeof(*ifa) * ifa_list_size, KM_SLEEP);
                   1237:        i = 0;
1.354     ozaki-r  1238:        IFADDR_READER_FOREACH(ifa, ifp) {
1.349     ozaki-r  1239:                ifa_list[i++] = ifa;
                   1240:                ifaref(ifa);
                   1241:        }
                   1242: }
                   1243:
                   1244: static void
                   1245: if_check_and_free_ifa_list(struct ifnet *ifp)
                   1246: {
                   1247:        int i;
                   1248:        struct ifaddr *ifa;
                   1249:
                   1250:        if (ifa_list == NULL)
                   1251:                return;
                   1252:
                   1253:        for (i = 0; i < ifa_list_size; i++) {
                   1254:                char buf[64];
                   1255:
                   1256:                ifa = ifa_list[i];
                   1257:                sockaddr_format(ifa->ifa_addr, buf, sizeof(buf));
                   1258:                if (ifa->ifa_refcnt > 1) {
                   1259:                        log(LOG_WARNING,
                   1260:                            "ifa(%s) still referenced (refcnt=%d)\n",
                   1261:                            buf, ifa->ifa_refcnt - 1);
                   1262:                } else
                   1263:                        log(LOG_DEBUG,
                   1264:                            "ifa(%s) not referenced (refcnt=%d)\n",
                   1265:                            buf, ifa->ifa_refcnt - 1);
                   1266:                ifafree(ifa);
                   1267:        }
                   1268:
                   1269:        kmem_free(ifa_list, sizeof(*ifa) * ifa_list_size);
                   1270:        ifa_list = NULL;
                   1271:        ifa_list_size = 0;
                   1272: }
                   1273: #endif
                   1274:
1.53      thorpej  1275: /*
                   1276:  * Detach an interface from the list of "active" interfaces,
                   1277:  * freeing any resources as we go along.
                   1278:  *
                   1279:  * NOTE: This routine must be called with a valid thread context,
                   1280:  * as it may block.
                   1281:  */
                   1282: void
1.163     thorpej  1283: if_detach(struct ifnet *ifp)
1.53      thorpej  1284: {
1.56      thorpej  1285:        struct socket so;
1.178     dyoung   1286:        struct ifaddr *ifa;
1.53      thorpej  1287: #ifdef IFAREF_DEBUG
                   1288:        struct ifaddr *last_ifa = NULL;
                   1289: #endif
1.56      thorpej  1290:        struct domain *dp;
1.141     matt     1291:        const struct protosw *pr;
1.322     riastrad 1292:        int s, i, family, purged;
1.276     rmind    1293:        uint64_t xc;
1.53      thorpej  1294:
1.349     ozaki-r  1295: #ifdef IFAREF_DEBUG
                   1296:        if_build_ifa_list(ifp);
                   1297: #endif
1.56      thorpej  1298:        /*
                   1299:         * XXX It's kind of lame that we have to have the
                   1300:         * XXX socket structure...
                   1301:         */
                   1302:        memset(&so, 0, sizeof(so));
1.53      thorpej  1303:
1.88      thorpej  1304:        s = splnet();
1.53      thorpej  1305:
1.334     ozaki-r  1306:        sysctl_teardown(&ifp->if_sysctl_log);
1.413     ozaki-r  1307:        IFNET_LOCK(ifp);
1.350     ozaki-r  1308:        if_deactivate(ifp);
1.413     ozaki-r  1309:        IFNET_UNLOCK(ifp);
1.334     ozaki-r  1310:
1.350     ozaki-r  1311:        if (ifp->if_slowtimo != NULL && ifp->if_slowtimo_ch != NULL) {
1.305     martin   1312:                ifp->if_slowtimo = NULL;
1.297     ozaki-r  1313:                callout_halt(ifp->if_slowtimo_ch, NULL);
                   1314:                callout_destroy(ifp->if_slowtimo_ch);
                   1315:                kmem_free(ifp->if_slowtimo_ch, sizeof(*ifp->if_slowtimo_ch));
1.296     ozaki-r  1316:        }
1.364     ozaki-r  1317:        if_deferred_start_destroy(ifp);
1.296     ozaki-r  1318:
1.53      thorpej  1319:        /*
                   1320:         * Do an if_down() to give protocols a chance to do something.
                   1321:         */
1.403     ozaki-r  1322:        if_down_deactivated(ifp);
1.86      thorpej  1323:
                   1324: #ifdef ALTQ
                   1325:        if (ALTQ_IS_ENABLED(&ifp->if_snd))
                   1326:                altq_disable(&ifp->if_snd);
                   1327:        if (ALTQ_IS_ATTACHED(&ifp->if_snd))
                   1328:                altq_detach(&ifp->if_snd);
1.87      thorpej  1329: #endif
                   1330:
1.166     liamjfoy 1331: #if NCARP > 0
                   1332:        /* Remove the interface from any carp group it is a part of.  */
1.185     dyoung   1333:        if (ifp->if_carp != NULL && ifp->if_type != IFT_CARP)
1.166     liamjfoy 1334:                carp_ifdetach(ifp);
                   1335: #endif
                   1336:
1.53      thorpej  1337:        /*
                   1338:         * Rip all the addresses off the interface.  This should make
                   1339:         * all of the routes go away.
1.178     dyoung   1340:         *
                   1341:         * pr_usrreq calls can remove an arbitrary number of ifaddrs
                   1342:         * from the list, including our "cursor", ifa.  For safety,
                   1343:         * and to honor the TAILQ abstraction, I just restart the
                   1344:         * loop after each removal.  Note that the loop will exit
                   1345:         * when all of the remaining ifaddrs belong to the AF_LINK
                   1346:         * family.  I am counting on the historical fact that at
                   1347:         * least one pr_usrreq in each address domain removes at
                   1348:         * least one ifaddr.
1.53      thorpej  1349:         */
1.178     dyoung   1350: again:
1.357     ozaki-r  1351:        /*
                   1352:         * At this point, no other one tries to remove ifa in the list,
1.407     ozaki-r  1353:         * so we don't need to take a lock or psref.  Avoid using
                   1354:         * IFADDR_READER_FOREACH to pass over an inspection of contract
                   1355:         * violations of pserialize.
1.357     ozaki-r  1356:         */
1.407     ozaki-r  1357:        IFADDR_WRITER_FOREACH(ifa, ifp) {
1.56      thorpej  1358:                family = ifa->ifa_addr->sa_family;
1.53      thorpej  1359: #ifdef IFAREF_DEBUG
                   1360:                printf("if_detach: ifaddr %p, family %d, refcnt %d\n",
1.56      thorpej  1361:                    ifa, family, ifa->ifa_refcnt);
1.53      thorpej  1362:                if (last_ifa != NULL && ifa == last_ifa)
1.56      thorpej  1363:                        panic("if_detach: loop detected");
1.53      thorpej  1364:                last_ifa = ifa;
                   1365: #endif
1.178     dyoung   1366:                if (family == AF_LINK)
1.118     itojun   1367:                        continue;
                   1368:                dp = pffinddomain(family);
1.384     ozaki-r  1369:                KASSERTMSG(dp != NULL, "no domain for AF %d", family);
1.160     gdt      1370:                /*
                   1371:                 * XXX These PURGEIF calls are redundant with the
                   1372:                 * purge-all-families calls below, but are left in for
                   1373:                 * now both to make a smaller change, and to avoid
                   1374:                 * unplanned interactions with clearing of
                   1375:                 * ifp->if_addrlist.
                   1376:                 */
1.118     itojun   1377:                purged = 0;
1.322     riastrad 1378:                for (pr = dp->dom_protosw;
                   1379:                     pr < dp->dom_protoswNPROTOSW; pr++) {
1.118     itojun   1380:                        so.so_proto = pr;
1.275     rmind    1381:                        if (pr->pr_usrreqs) {
1.290     rtr      1382:                                (void) (*pr->pr_usrreqs->pr_purgeif)(&so, ifp);
1.118     itojun   1383:                                purged = 1;
1.53      thorpej  1384:                        }
1.118     itojun   1385:                }
                   1386:                if (purged == 0) {
                   1387:                        /*
                   1388:                         * XXX What's really the best thing to do
1.135     keihan   1389:                         * XXX here?  --thorpej@NetBSD.org
1.118     itojun   1390:                         */
                   1391:                        printf("if_detach: WARNING: AF %d not purged\n",
                   1392:                            family);
1.207     dyoung   1393:                        ifa_remove(ifp, ifa);
1.53      thorpej  1394:                }
1.178     dyoung   1395:                goto again;
1.53      thorpej  1396:        }
1.118     itojun   1397:
                   1398:        if_free_sadl(ifp);
1.53      thorpej  1399:
1.362     ozaki-r  1400:        /* Delete stray routes from the routing table. */
                   1401:        for (i = 0; i <= AF_MAX; i++)
                   1402:                rt_delete_matched_entries(i, if_delroute_matcher, ifp);
1.106     itojun   1403:
1.152     matt     1404:        DOMAIN_FOREACH(dp) {
1.185     dyoung   1405:                if (dp->dom_ifdetach != NULL && ifp->if_afdata[dp->dom_family])
1.260     christos 1406:                {
                   1407:                        void *p = ifp->if_afdata[dp->dom_family];
                   1408:                        if (p) {
                   1409:                                ifp->if_afdata[dp->dom_family] = NULL;
                   1410:                                (*dp->dom_ifdetach)(ifp, p);
                   1411:                        }
                   1412:                }
1.160     gdt      1413:
                   1414:                /*
                   1415:                 * One would expect multicast memberships (INET and
                   1416:                 * INET6) on UDP sockets to be purged by the PURGEIF
                   1417:                 * calls above, but if all addresses were removed from
                   1418:                 * the interface prior to destruction, the calls will
                   1419:                 * not be made (e.g. ppp, for which pppd(8) generally
                   1420:                 * removes addresses before destroying the interface).
                   1421:                 * Because there is no invariant that multicast
                   1422:                 * memberships only exist for interfaces with IPv4
                   1423:                 * addresses, we must call PURGEIF regardless of
                   1424:                 * addresses.  (Protocols which might store ifnet
                   1425:                 * pointers are marked with PR_PURGEIF.)
                   1426:                 */
1.185     dyoung   1427:                for (pr = dp->dom_protosw; pr < dp->dom_protoswNPROTOSW; pr++) {
1.160     gdt      1428:                        so.so_proto = pr;
1.275     rmind    1429:                        if (pr->pr_usrreqs && pr->pr_flags & PR_PURGEIF)
1.290     rtr      1430:                                (void)(*pr->pr_usrreqs->pr_purgeif)(&so, ifp);
1.160     gdt      1431:                }
1.53      thorpej  1432:        }
1.57      thorpej  1433:
1.414     ozaki-r  1434:        /* Wait for all readers to drain before freeing.  */
                   1435:        IFNET_GLOBAL_LOCK();
                   1436:        ifindex2ifnet[ifp->if_index] = NULL;
                   1437:        TAILQ_REMOVE(&ifnet_list, ifp, if_list);
                   1438:        IFNET_WRITER_REMOVE(ifp);
                   1439:        pserialize_perform(ifnet_psz);
                   1440:        IFNET_GLOBAL_UNLOCK();
                   1441:
                   1442:        psref_target_destroy(&ifp->if_psref, ifnet_psref_class);
                   1443:        PSLIST_ENTRY_DESTROY(ifp, if_pslist_entry);
                   1444:
1.369     christos 1445:        pfil_run_ifhooks(if_pfil, PFIL_IFNET_DETACH, ifp);
1.265     rmind    1446:        (void)pfil_head_destroy(ifp->if_pfil);
1.184     dyoung   1447:
1.57      thorpej  1448:        /* Announce that the interface is gone. */
                   1449:        rt_ifannouncemsg(ifp, IFAN_DEPARTURE);
1.93      itojun   1450:
1.320     ozaki-r  1451:        IF_AFDATA_LOCK_DESTROY(ifp);
                   1452:
1.347     knakahar 1453:        if (if_is_link_state_changeable(ifp)) {
1.346     knakahar 1454:                softint_disestablish(ifp->if_link_si);
                   1455:                ifp->if_link_si = NULL;
                   1456:        }
1.324     ozaki-r  1457:
1.95      itojun   1458:        /*
1.168     matt     1459:         * remove packets that came from ifp, from software interrupt queues.
1.95      itojun   1460:         */
1.168     matt     1461:        DOMAIN_FOREACH(dp) {
                   1462:                for (i = 0; i < __arraycount(dp->dom_ifqueues); i++) {
1.260     christos 1463:                        struct ifqueue *iq = dp->dom_ifqueues[i];
                   1464:                        if (iq == NULL)
1.168     matt     1465:                                break;
1.260     christos 1466:                        dp->dom_ifqueues[i] = NULL;
                   1467:                        if_detach_queues(ifp, iq);
1.168     matt     1468:                }
                   1469:        }
1.95      itojun   1470:
1.276     rmind    1471:        /*
                   1472:         * IP queues have to be processed separately: net-queue barrier
                   1473:         * ensures that the packets are dequeued while a cross-call will
                   1474:         * ensure that the interrupts have completed. FIXME: not quite..
                   1475:         */
1.278     he       1476: #ifdef INET
1.276     rmind    1477:        pktq_barrier(ip_pktq);
1.278     he       1478: #endif
1.281     rmind    1479: #ifdef INET6
1.293     pooka    1480:        if (in6_present)
                   1481:                pktq_barrier(ip6_pktq);
1.281     rmind    1482: #endif
1.276     rmind    1483:        xc = xc_broadcast(0, (xcfunc_t)nullop, NULL, NULL);
                   1484:        xc_wait(xc);
                   1485:
1.323     ozaki-r  1486:        if (ifp->if_percpuq != NULL) {
                   1487:                if_percpuq_destroy(ifp->if_percpuq);
                   1488:                ifp->if_percpuq = NULL;
                   1489:        }
                   1490:
1.414     ozaki-r  1491:        mutex_obj_free(ifp->if_ioctl_lock);
                   1492:        ifp->if_ioctl_lock = NULL;
1.419     ozaki-r  1493:        mutex_obj_free(ifp->if_snd.ifq_lock);
1.414     ozaki-r  1494:
1.53      thorpej  1495:        splx(s);
1.349     ozaki-r  1496:
                   1497: #ifdef IFAREF_DEBUG
                   1498:        if_check_and_free_ifa_list(ifp);
                   1499: #endif
1.95      itojun   1500: }
                   1501:
                   1502: static void
1.163     thorpej  1503: if_detach_queues(struct ifnet *ifp, struct ifqueue *q)
1.95      itojun   1504: {
                   1505:        struct mbuf *m, *prev, *next;
                   1506:
                   1507:        prev = NULL;
1.185     dyoung   1508:        for (m = q->ifq_head; m != NULL; m = next) {
1.274     rmind    1509:                KASSERT((m->m_flags & M_PKTHDR) != 0);
                   1510:
1.95      itojun   1511:                next = m->m_nextpkt;
1.338     ozaki-r  1512:                if (m->m_pkthdr.rcvif_index != ifp->if_index) {
1.96      itojun   1513:                        prev = m;
1.95      itojun   1514:                        continue;
1.96      itojun   1515:                }
1.95      itojun   1516:
1.185     dyoung   1517:                if (prev != NULL)
1.95      itojun   1518:                        prev->m_nextpkt = m->m_nextpkt;
                   1519:                else
                   1520:                        q->ifq_head = m->m_nextpkt;
                   1521:                if (q->ifq_tail == m)
                   1522:                        q->ifq_tail = prev;
                   1523:                q->ifq_len--;
                   1524:
                   1525:                m->m_nextpkt = NULL;
                   1526:                m_freem(m);
                   1527:                IF_DROP(q);
                   1528:        }
1.53      thorpej  1529: }
                   1530:
                   1531: /*
                   1532:  * Callback for a radix tree walk to delete all references to an
                   1533:  * ifnet.
                   1534:  */
1.163     thorpej  1535: static int
1.362     ozaki-r  1536: if_delroute_matcher(struct rtentry *rt, void *v)
1.53      thorpej  1537: {
1.55      itojun   1538:        struct ifnet *ifp = (struct ifnet *)v;
1.53      thorpej  1539:
1.362     ozaki-r  1540:        if (rt->rt_ifp == ifp)
                   1541:                return 1;
                   1542:        else
1.185     dyoung   1543:                return 0;
1.53      thorpej  1544: }
                   1545:
1.1       cgd      1546: /*
1.63      thorpej  1547:  * Create a clone network interface.
                   1548:  */
1.302     ozaki-r  1549: static int
1.163     thorpej  1550: if_clone_create(const char *name)
1.63      thorpej  1551: {
                   1552:        struct if_clone *ifc;
                   1553:        int unit;
1.336     ozaki-r  1554:        struct ifnet *ifp;
                   1555:        struct psref psref;
1.63      thorpej  1556:
1.372     ozaki-r  1557:        KASSERT(mutex_owned(&if_clone_mtx));
                   1558:
1.63      thorpej  1559:        ifc = if_clone_lookup(name, &unit);
                   1560:        if (ifc == NULL)
1.185     dyoung   1561:                return EINVAL;
1.63      thorpej  1562:
1.336     ozaki-r  1563:        ifp = if_get(name, &psref);
                   1564:        if (ifp != NULL) {
                   1565:                if_put(ifp, &psref);
1.185     dyoung   1566:                return EEXIST;
1.336     ozaki-r  1567:        }
1.63      thorpej  1568:
1.185     dyoung   1569:        return (*ifc->ifc_create)(ifc, unit);
1.63      thorpej  1570: }
                   1571:
                   1572: /*
                   1573:  * Destroy a clone network interface.
                   1574:  */
1.302     ozaki-r  1575: static int
1.163     thorpej  1576: if_clone_destroy(const char *name)
1.63      thorpej  1577: {
                   1578:        struct if_clone *ifc;
                   1579:        struct ifnet *ifp;
1.336     ozaki-r  1580:        struct psref psref;
1.63      thorpej  1581:
1.372     ozaki-r  1582:        KASSERT(mutex_owned(&if_clone_mtx));
                   1583:
1.63      thorpej  1584:        ifc = if_clone_lookup(name, NULL);
                   1585:        if (ifc == NULL)
1.185     dyoung   1586:                return EINVAL;
1.63      thorpej  1587:
1.336     ozaki-r  1588:        if (ifc->ifc_destroy == NULL)
                   1589:                return EOPNOTSUPP;
                   1590:
                   1591:        ifp = if_get(name, &psref);
1.63      thorpej  1592:        if (ifp == NULL)
1.185     dyoung   1593:                return ENXIO;
1.63      thorpej  1594:
1.336     ozaki-r  1595:        /* We have to disable ioctls here */
1.413     ozaki-r  1596:        IFNET_LOCK(ifp);
1.336     ozaki-r  1597:        ifp->if_ioctl = if_nullioctl;
1.413     ozaki-r  1598:        IFNET_UNLOCK(ifp);
1.336     ozaki-r  1599:
                   1600:        /*
                   1601:         * We cannot call ifc_destroy with holding ifp.
                   1602:         * Releasing ifp here is safe thanks to if_clone_mtx.
                   1603:         */
                   1604:        if_put(ifp, &psref);
1.63      thorpej  1605:
1.185     dyoung   1606:        return (*ifc->ifc_destroy)(ifp);
1.63      thorpej  1607: }
                   1608:
1.379     knakahar 1609: static bool
                   1610: if_is_unit(const char *name)
                   1611: {
                   1612:
                   1613:        while(*name != '\0') {
                   1614:                if (*name < '0' || *name > '9')
                   1615:                        return false;
                   1616:                name++;
                   1617:        }
                   1618:
                   1619:        return true;
                   1620: }
                   1621:
1.63      thorpej  1622: /*
                   1623:  * Look up a network interface cloner.
                   1624:  */
1.163     thorpej  1625: static struct if_clone *
                   1626: if_clone_lookup(const char *name, int *unitp)
1.63      thorpej  1627: {
                   1628:        struct if_clone *ifc;
                   1629:        const char *cp;
1.262     christos 1630:        char *dp, ifname[IFNAMSIZ + 3];
1.128     itojun   1631:        int unit;
1.63      thorpej  1632:
1.372     ozaki-r  1633:        KASSERT(mutex_owned(&if_clone_mtx));
                   1634:
1.262     christos 1635:        strcpy(ifname, "if_");
1.128     itojun   1636:        /* separate interface name from unit */
1.379     knakahar 1637:        /* TODO: search unit number from backward */
1.262     christos 1638:        for (dp = ifname + 3, cp = name; cp - name < IFNAMSIZ &&
1.379     knakahar 1639:            *cp && !if_is_unit(cp);)
1.262     christos 1640:                *dp++ = *cp++;
1.128     itojun   1641:
                   1642:        if (cp == name || cp - name == IFNAMSIZ || !*cp)
1.185     dyoung   1643:                return NULL;    /* No name or unit number */
1.262     christos 1644:        *dp++ = '\0';
1.128     itojun   1645:
1.262     christos 1646: again:
1.128     itojun   1647:        LIST_FOREACH(ifc, &if_cloners, ifc_list) {
1.262     christos 1648:                if (strcmp(ifname + 3, ifc->ifc_name) == 0)
1.128     itojun   1649:                        break;
1.63      thorpej  1650:        }
                   1651:
1.262     christos 1652:        if (ifc == NULL) {
1.375     christos 1653:                int error;
                   1654:                if (*ifname == '\0')
                   1655:                        return NULL;
                   1656:                mutex_exit(&if_clone_mtx);
                   1657:                error = module_autoload(ifname, MODULE_CLASS_DRIVER);
                   1658:                mutex_enter(&if_clone_mtx);
                   1659:                if (error)
1.262     christos 1660:                        return NULL;
                   1661:                *ifname = '\0';
                   1662:                goto again;
                   1663:        }
1.63      thorpej  1664:
1.128     itojun   1665:        unit = 0;
1.129     itojun   1666:        while (cp - name < IFNAMSIZ && *cp) {
1.245     christos 1667:                if (*cp < '0' || *cp > '9' || unit >= INT_MAX / 10) {
1.63      thorpej  1668:                        /* Bogus unit number. */
1.185     dyoung   1669:                        return NULL;
1.63      thorpej  1670:                }
1.128     itojun   1671:                unit = (unit * 10) + (*cp++ - '0');
1.63      thorpej  1672:        }
                   1673:
                   1674:        if (unitp != NULL)
1.128     itojun   1675:                *unitp = unit;
1.185     dyoung   1676:        return ifc;
1.63      thorpej  1677: }
                   1678:
                   1679: /*
                   1680:  * Register a network interface cloner.
                   1681:  */
                   1682: void
1.163     thorpej  1683: if_clone_attach(struct if_clone *ifc)
1.63      thorpej  1684: {
                   1685:
1.372     ozaki-r  1686:        mutex_enter(&if_clone_mtx);
1.63      thorpej  1687:        LIST_INSERT_HEAD(&if_cloners, ifc, ifc_list);
1.67      thorpej  1688:        if_cloners_count++;
1.372     ozaki-r  1689:        mutex_exit(&if_clone_mtx);
1.63      thorpej  1690: }
                   1691:
                   1692: /*
                   1693:  * Unregister a network interface cloner.
                   1694:  */
                   1695: void
1.163     thorpej  1696: if_clone_detach(struct if_clone *ifc)
1.63      thorpej  1697: {
                   1698:
1.377     christos 1699:        mutex_enter(&if_clone_mtx);
1.63      thorpej  1700:        LIST_REMOVE(ifc, ifc_list);
1.67      thorpej  1701:        if_cloners_count--;
1.377     christos 1702:        mutex_exit(&if_clone_mtx);
1.67      thorpej  1703: }
                   1704:
                   1705: /*
                   1706:  * Provide list of interface cloners to userspace.
                   1707:  */
1.315     martin   1708: int
                   1709: if_clone_list(int buf_count, char *buffer, int *total)
1.67      thorpej  1710: {
                   1711:        char outbuf[IFNAMSIZ], *dst;
                   1712:        struct if_clone *ifc;
                   1713:        int count, error = 0;
                   1714:
1.372     ozaki-r  1715:        mutex_enter(&if_clone_mtx);
1.315     martin   1716:        *total = if_cloners_count;
                   1717:        if ((dst = buffer) == NULL) {
1.67      thorpej  1718:                /* Just asking how many there are. */
1.372     ozaki-r  1719:                goto out;
1.67      thorpej  1720:        }
                   1721:
1.372     ozaki-r  1722:        if (buf_count < 0) {
                   1723:                error = EINVAL;
                   1724:                goto out;
                   1725:        }
1.67      thorpej  1726:
1.315     martin   1727:        count = (if_cloners_count < buf_count) ?
                   1728:            if_cloners_count : buf_count;
1.67      thorpej  1729:
                   1730:        for (ifc = LIST_FIRST(&if_cloners); ifc != NULL && count != 0;
                   1731:             ifc = LIST_NEXT(ifc, ifc_list), count--, dst += IFNAMSIZ) {
1.175     christos 1732:                (void)strncpy(outbuf, ifc->ifc_name, sizeof(outbuf));
1.372     ozaki-r  1733:                if (outbuf[sizeof(outbuf) - 1] != '\0') {
                   1734:                        error = ENAMETOOLONG;
                   1735:                        goto out;
                   1736:                }
1.172     christos 1737:                error = copyout(outbuf, dst, sizeof(outbuf));
1.185     dyoung   1738:                if (error != 0)
1.67      thorpej  1739:                        break;
                   1740:        }
                   1741:
1.372     ozaki-r  1742: out:
                   1743:        mutex_exit(&if_clone_mtx);
1.185     dyoung   1744:        return error;
1.63      thorpej  1745: }
                   1746:
1.207     dyoung   1747: void
1.357     ozaki-r  1748: ifa_psref_init(struct ifaddr *ifa)
                   1749: {
                   1750:
                   1751:        psref_target_init(&ifa->ifa_psref, ifa_psref_class);
                   1752: }
                   1753:
                   1754: void
1.291     rmind    1755: ifaref(struct ifaddr *ifa)
                   1756: {
1.366     ozaki-r  1757:        KASSERT(!ISSET(ifa->ifa_flags, IFA_DESTROYING));
1.291     rmind    1758:        ifa->ifa_refcnt++;
                   1759: }
                   1760:
                   1761: void
                   1762: ifafree(struct ifaddr *ifa)
                   1763: {
                   1764:        KASSERT(ifa != NULL);
                   1765:        KASSERT(ifa->ifa_refcnt > 0);
                   1766:
                   1767:        if (--ifa->ifa_refcnt == 0) {
                   1768:                free(ifa, M_IFADDR);
                   1769:        }
                   1770: }
                   1771:
1.366     ozaki-r  1772: bool
                   1773: ifa_is_destroying(struct ifaddr *ifa)
                   1774: {
                   1775:
                   1776:        return ISSET(ifa->ifa_flags, IFA_DESTROYING);
                   1777: }
                   1778:
1.291     rmind    1779: void
1.207     dyoung   1780: ifa_insert(struct ifnet *ifp, struct ifaddr *ifa)
                   1781: {
1.357     ozaki-r  1782:
1.207     dyoung   1783:        ifa->ifa_ifp = ifp;
1.357     ozaki-r  1784:
1.416     ozaki-r  1785:        /*
1.418     ozaki-r  1786:         * Check MP-safety for IFEF_MPSAFE drivers.
1.416     ozaki-r  1787:         * Check !IFF_RUNNING for initialization routines that normally don't
                   1788:         * take IFNET_LOCK but it's safe because there is no competitor.
                   1789:         * XXX there are false positive cases because IFF_RUNNING can be off on
                   1790:         * if_stop.
                   1791:         */
1.418     ozaki-r  1792:        KASSERT(!if_is_mpsafe(ifp) || !ISSET(ifp->if_flags, IFF_RUNNING) ||
1.416     ozaki-r  1793:            IFNET_LOCKED(ifp));
                   1794:
1.208     dyoung   1795:        TAILQ_INSERT_TAIL(&ifp->if_addrlist, ifa, ifa_list);
1.354     ozaki-r  1796:        IFADDR_ENTRY_INIT(ifa);
                   1797:        IFADDR_WRITER_INSERT_TAIL(ifp, ifa);
1.357     ozaki-r  1798:
1.291     rmind    1799:        ifaref(ifa);
1.207     dyoung   1800: }
                   1801:
                   1802: void
                   1803: ifa_remove(struct ifnet *ifp, struct ifaddr *ifa)
                   1804: {
1.357     ozaki-r  1805:
1.207     dyoung   1806:        KASSERT(ifa->ifa_ifp == ifp);
1.416     ozaki-r  1807:        /*
1.418     ozaki-r  1808:         * Check MP-safety for IFEF_MPSAFE drivers.
1.416     ozaki-r  1809:         * if_is_deactivated indicates ifa_remove is called form if_detach
                   1810:         * where is safe even if IFNET_LOCK isn't held.
                   1811:         */
1.418     ozaki-r  1812:        KASSERT(!if_is_mpsafe(ifp) || if_is_deactivated(ifp) || IFNET_LOCKED(ifp));
1.357     ozaki-r  1813:
1.207     dyoung   1814:        TAILQ_REMOVE(&ifp->if_addrlist, ifa, ifa_list);
1.354     ozaki-r  1815:        IFADDR_WRITER_REMOVE(ifa);
1.359     ozaki-r  1816: #ifdef NET_MPSAFE
1.416     ozaki-r  1817:        IFNET_GLOBAL_LOCK();
1.357     ozaki-r  1818:        pserialize_perform(ifnet_psz);
1.416     ozaki-r  1819:        IFNET_GLOBAL_UNLOCK();
1.357     ozaki-r  1820: #endif
                   1821:
1.359     ozaki-r  1822: #ifdef NET_MPSAFE
1.357     ozaki-r  1823:        psref_target_destroy(&ifa->ifa_psref, ifa_psref_class);
                   1824: #endif
1.360     ozaki-r  1825:        IFADDR_ENTRY_DESTROY(ifa);
1.291     rmind    1826:        ifafree(ifa);
1.207     dyoung   1827: }
                   1828:
1.357     ozaki-r  1829: void
                   1830: ifa_acquire(struct ifaddr *ifa, struct psref *psref)
                   1831: {
                   1832:
                   1833:        psref_acquire(psref, &ifa->ifa_psref, ifa_psref_class);
                   1834: }
                   1835:
                   1836: void
                   1837: ifa_release(struct ifaddr *ifa, struct psref *psref)
                   1838: {
                   1839:
                   1840:        if (ifa == NULL)
                   1841:                return;
                   1842:
                   1843:        psref_release(psref, &ifa->ifa_psref, ifa_psref_class);
                   1844: }
                   1845:
                   1846: bool
                   1847: ifa_held(struct ifaddr *ifa)
                   1848: {
                   1849:
                   1850:        return psref_held(&ifa->ifa_psref, ifa_psref_class);
                   1851: }
                   1852:
1.194     dyoung   1853: static inline int
                   1854: equal(const struct sockaddr *sa1, const struct sockaddr *sa2)
                   1855: {
                   1856:        return sockaddr_cmp(sa1, sa2) == 0;
                   1857: }
                   1858:
1.63      thorpej  1859: /*
1.1       cgd      1860:  * Locate an interface based on a complete address.
                   1861:  */
                   1862: /*ARGSUSED*/
                   1863: struct ifaddr *
1.163     thorpej  1864: ifa_ifwithaddr(const struct sockaddr *addr)
1.1       cgd      1865: {
1.61      augustss 1866:        struct ifnet *ifp;
                   1867:        struct ifaddr *ifa;
1.1       cgd      1868:
1.334     ozaki-r  1869:        IFNET_READER_FOREACH(ifp) {
1.348     ozaki-r  1870:                if (if_is_deactivated(ifp))
1.1       cgd      1871:                        continue;
1.354     ozaki-r  1872:                IFADDR_READER_FOREACH(ifa, ifp) {
1.53      thorpej  1873:                        if (ifa->ifa_addr->sa_family != addr->sa_family)
                   1874:                                continue;
                   1875:                        if (equal(addr, ifa->ifa_addr))
1.185     dyoung   1876:                                return ifa;
1.53      thorpej  1877:                        if ((ifp->if_flags & IFF_BROADCAST) &&
                   1878:                            ifa->ifa_broadaddr &&
                   1879:                            /* IP6 doesn't have broadcast */
                   1880:                            ifa->ifa_broadaddr->sa_len != 0 &&
                   1881:                            equal(ifa->ifa_broadaddr, addr))
1.185     dyoung   1882:                                return ifa;
1.53      thorpej  1883:                }
1.1       cgd      1884:        }
1.357     ozaki-r  1885:        return NULL;
                   1886: }
                   1887:
                   1888: struct ifaddr *
                   1889: ifa_ifwithaddr_psref(const struct sockaddr *addr, struct psref *psref)
                   1890: {
                   1891:        struct ifaddr *ifa;
                   1892:        int s = pserialize_read_enter();
                   1893:
                   1894:        ifa = ifa_ifwithaddr(addr);
                   1895:        if (ifa != NULL)
                   1896:                ifa_acquire(ifa, psref);
1.334     ozaki-r  1897:        pserialize_read_exit(s);
1.357     ozaki-r  1898:
                   1899:        return ifa;
1.1       cgd      1900: }
1.49      itojun   1901:
1.1       cgd      1902: /*
                   1903:  * Locate the point to point interface with a given destination address.
                   1904:  */
                   1905: /*ARGSUSED*/
                   1906: struct ifaddr *
1.163     thorpej  1907: ifa_ifwithdstaddr(const struct sockaddr *addr)
1.1       cgd      1908: {
1.61      augustss 1909:        struct ifnet *ifp;
                   1910:        struct ifaddr *ifa;
1.1       cgd      1911:
1.334     ozaki-r  1912:        IFNET_READER_FOREACH(ifp) {
1.348     ozaki-r  1913:                if (if_is_deactivated(ifp))
1.53      thorpej  1914:                        continue;
1.185     dyoung   1915:                if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
                   1916:                        continue;
1.354     ozaki-r  1917:                IFADDR_READER_FOREACH(ifa, ifp) {
1.185     dyoung   1918:                        if (ifa->ifa_addr->sa_family != addr->sa_family ||
                   1919:                            ifa->ifa_dstaddr == NULL)
                   1920:                                continue;
                   1921:                        if (equal(addr, ifa->ifa_dstaddr))
                   1922:                                return ifa;
1.53      thorpej  1923:                }
1.1       cgd      1924:        }
1.357     ozaki-r  1925:
                   1926:        return NULL;
                   1927: }
                   1928:
                   1929: struct ifaddr *
                   1930: ifa_ifwithdstaddr_psref(const struct sockaddr *addr, struct psref *psref)
                   1931: {
                   1932:        struct ifaddr *ifa;
                   1933:        int s;
                   1934:
                   1935:        s = pserialize_read_enter();
                   1936:        ifa = ifa_ifwithdstaddr(addr);
                   1937:        if (ifa != NULL)
                   1938:                ifa_acquire(ifa, psref);
1.334     ozaki-r  1939:        pserialize_read_exit(s);
1.357     ozaki-r  1940:
                   1941:        return ifa;
1.1       cgd      1942: }
                   1943:
                   1944: /*
                   1945:  * Find an interface on a specific network.  If many, choice
1.15      mycroft  1946:  * is most specific found.
1.1       cgd      1947:  */
                   1948: struct ifaddr *
1.163     thorpej  1949: ifa_ifwithnet(const struct sockaddr *addr)
1.1       cgd      1950: {
1.61      augustss 1951:        struct ifnet *ifp;
1.357     ozaki-r  1952:        struct ifaddr *ifa, *ifa_maybe = NULL;
1.140     matt     1953:        const struct sockaddr_dl *sdl;
1.1       cgd      1954:        u_int af = addr->sa_family;
1.171     pooka    1955:        const char *addr_data = addr->sa_data, *cplim;
1.1       cgd      1956:
                   1957:        if (af == AF_LINK) {
1.195     dyoung   1958:                sdl = satocsdl(addr);
1.137     itojun   1959:                if (sdl->sdl_index && sdl->sdl_index < if_indexlim &&
                   1960:                    ifindex2ifnet[sdl->sdl_index] &&
1.348     ozaki-r  1961:                    !if_is_deactivated(ifindex2ifnet[sdl->sdl_index])) {
1.316     ozaki-r  1962:                        return ifindex2ifnet[sdl->sdl_index]->if_dl;
                   1963:                }
1.1       cgd      1964:        }
1.51      bouyer   1965: #ifdef NETATALK
                   1966:        if (af == AF_APPLETALK) {
1.140     matt     1967:                const struct sockaddr_at *sat, *sat2;
1.158     christos 1968:                sat = (const struct sockaddr_at *)addr;
1.334     ozaki-r  1969:                IFNET_READER_FOREACH(ifp) {
1.348     ozaki-r  1970:                        if (if_is_deactivated(ifp))
1.53      thorpej  1971:                                continue;
1.158     christos 1972:                        ifa = at_ifawithnet((const struct sockaddr_at *)addr, ifp);
1.62      bouyer   1973:                        if (ifa == NULL)
                   1974:                                continue;
                   1975:                        sat2 = (struct sockaddr_at *)ifa->ifa_addr;
                   1976:                        if (sat2->sat_addr.s_net == sat->sat_addr.s_net)
1.185     dyoung   1977:                                return ifa; /* exact match */
1.62      bouyer   1978:                        if (ifa_maybe == NULL) {
1.112     wiz      1979:                                /* else keep the if with the right range */
1.62      bouyer   1980:                                ifa_maybe = ifa;
                   1981:                        }
1.51      bouyer   1982:                }
1.185     dyoung   1983:                return ifa_maybe;
1.51      bouyer   1984:        }
                   1985: #endif
1.334     ozaki-r  1986:        IFNET_READER_FOREACH(ifp) {
1.348     ozaki-r  1987:                if (if_is_deactivated(ifp))
1.53      thorpej  1988:                        continue;
1.354     ozaki-r  1989:                IFADDR_READER_FOREACH(ifa, ifp) {
1.171     pooka    1990:                        const char *cp, *cp2, *cp3;
1.15      mycroft  1991:
                   1992:                        if (ifa->ifa_addr->sa_family != af ||
1.185     dyoung   1993:                            ifa->ifa_netmask == NULL)
1.53      thorpej  1994:  next:                         continue;
1.15      mycroft  1995:                        cp = addr_data;
                   1996:                        cp2 = ifa->ifa_addr->sa_data;
                   1997:                        cp3 = ifa->ifa_netmask->sa_data;
1.171     pooka    1998:                        cplim = (const char *)ifa->ifa_netmask +
1.53      thorpej  1999:                            ifa->ifa_netmask->sa_len;
                   2000:                        while (cp3 < cplim) {
                   2001:                                if ((*cp++ ^ *cp2++) & *cp3++) {
                   2002:                                        /* want to continue for() loop */
1.32      mrg      2003:                                        goto next;
1.53      thorpej  2004:                                }
                   2005:                        }
1.185     dyoung   2006:                        if (ifa_maybe == NULL ||
1.329     ozaki-r  2007:                            rt_refines(ifa->ifa_netmask,
                   2008:                                       ifa_maybe->ifa_netmask))
1.15      mycroft  2009:                                ifa_maybe = ifa;
                   2010:                }
1.53      thorpej  2011:        }
1.357     ozaki-r  2012:        return ifa_maybe;
                   2013: }
                   2014:
                   2015: struct ifaddr *
                   2016: ifa_ifwithnet_psref(const struct sockaddr *addr, struct psref *psref)
                   2017: {
                   2018:        struct ifaddr *ifa;
                   2019:        int s;
                   2020:
                   2021:        s = pserialize_read_enter();
                   2022:        ifa = ifa_ifwithnet(addr);
                   2023:        if (ifa != NULL)
                   2024:                ifa_acquire(ifa, psref);
1.334     ozaki-r  2025:        pserialize_read_exit(s);
1.357     ozaki-r  2026:
                   2027:        return ifa;
1.26      mrg      2028: }
1.53      thorpej  2029:
1.26      mrg      2030: /*
                   2031:  * Find the interface of the addresss.
                   2032:  */
                   2033: struct ifaddr *
1.163     thorpej  2034: ifa_ifwithladdr(const struct sockaddr *addr)
1.26      mrg      2035: {
                   2036:        struct ifaddr *ia;
                   2037:
1.53      thorpej  2038:        if ((ia = ifa_ifwithaddr(addr)) || (ia = ifa_ifwithdstaddr(addr)) ||
                   2039:            (ia = ifa_ifwithnet(addr)))
1.185     dyoung   2040:                return ia;
                   2041:        return NULL;
1.1       cgd      2042: }
                   2043:
1.357     ozaki-r  2044: struct ifaddr *
                   2045: ifa_ifwithladdr_psref(const struct sockaddr *addr, struct psref *psref)
                   2046: {
                   2047:        struct ifaddr *ifa;
                   2048:        int s;
                   2049:
                   2050:        s = pserialize_read_enter();
                   2051:        ifa = ifa_ifwithladdr(addr);
                   2052:        if (ifa != NULL)
                   2053:                ifa_acquire(ifa, psref);
                   2054:        pserialize_read_exit(s);
                   2055:
                   2056:        return ifa;
                   2057: }
                   2058:
1.1       cgd      2059: /*
                   2060:  * Find an interface using a specific address family
                   2061:  */
                   2062: struct ifaddr *
1.163     thorpej  2063: ifa_ifwithaf(int af)
1.1       cgd      2064: {
1.61      augustss 2065:        struct ifnet *ifp;
1.334     ozaki-r  2066:        struct ifaddr *ifa = NULL;
                   2067:        int s;
1.1       cgd      2068:
1.334     ozaki-r  2069:        s = pserialize_read_enter();
                   2070:        IFNET_READER_FOREACH(ifp) {
1.348     ozaki-r  2071:                if (if_is_deactivated(ifp))
1.53      thorpej  2072:                        continue;
1.354     ozaki-r  2073:                IFADDR_READER_FOREACH(ifa, ifp) {
1.21      mycroft  2074:                        if (ifa->ifa_addr->sa_family == af)
1.334     ozaki-r  2075:                                goto out;
1.53      thorpej  2076:                }
                   2077:        }
1.334     ozaki-r  2078: out:
                   2079:        pserialize_read_exit(s);
                   2080:        return ifa;
1.1       cgd      2081: }
                   2082:
                   2083: /*
                   2084:  * Find an interface address specific to an interface best matching
                   2085:  * a given address.
                   2086:  */
                   2087: struct ifaddr *
1.163     thorpej  2088: ifaof_ifpforaddr(const struct sockaddr *addr, struct ifnet *ifp)
1.1       cgd      2089: {
1.61      augustss 2090:        struct ifaddr *ifa;
1.140     matt     2091:        const char *cp, *cp2, *cp3;
                   2092:        const char *cplim;
1.1       cgd      2093:        struct ifaddr *ifa_maybe = 0;
                   2094:        u_int af = addr->sa_family;
                   2095:
1.348     ozaki-r  2096:        if (if_is_deactivated(ifp))
1.185     dyoung   2097:                return NULL;
1.53      thorpej  2098:
1.1       cgd      2099:        if (af >= AF_MAX)
1.185     dyoung   2100:                return NULL;
1.53      thorpej  2101:
1.354     ozaki-r  2102:        IFADDR_READER_FOREACH(ifa, ifp) {
1.1       cgd      2103:                if (ifa->ifa_addr->sa_family != af)
                   2104:                        continue;
                   2105:                ifa_maybe = ifa;
1.185     dyoung   2106:                if (ifa->ifa_netmask == NULL) {
1.1       cgd      2107:                        if (equal(addr, ifa->ifa_addr) ||
1.53      thorpej  2108:                            (ifa->ifa_dstaddr &&
                   2109:                             equal(addr, ifa->ifa_dstaddr)))
1.185     dyoung   2110:                                return ifa;
1.1       cgd      2111:                        continue;
                   2112:                }
                   2113:                cp = addr->sa_data;
                   2114:                cp2 = ifa->ifa_addr->sa_data;
                   2115:                cp3 = ifa->ifa_netmask->sa_data;
                   2116:                cplim = ifa->ifa_netmask->sa_len + (char *)ifa->ifa_netmask;
1.53      thorpej  2117:                for (; cp3 < cplim; cp3++) {
1.1       cgd      2118:                        if ((*cp++ ^ *cp2++) & *cp3)
                   2119:                                break;
1.53      thorpej  2120:                }
1.1       cgd      2121:                if (cp3 == cplim)
1.185     dyoung   2122:                        return ifa;
1.1       cgd      2123:        }
1.185     dyoung   2124:        return ifa_maybe;
1.1       cgd      2125: }
1.9       mycroft  2126:
1.357     ozaki-r  2127: struct ifaddr *
                   2128: ifaof_ifpforaddr_psref(const struct sockaddr *addr, struct ifnet *ifp,
                   2129:     struct psref *psref)
                   2130: {
                   2131:        struct ifaddr *ifa;
                   2132:        int s;
                   2133:
                   2134:        s = pserialize_read_enter();
                   2135:        ifa = ifaof_ifpforaddr(addr, ifp);
                   2136:        if (ifa != NULL)
                   2137:                ifa_acquire(ifa, psref);
                   2138:        pserialize_read_exit(s);
                   2139:
                   2140:        return ifa;
                   2141: }
                   2142:
1.1       cgd      2143: /*
                   2144:  * Default action when installing a route with a Link Level gateway.
                   2145:  * Lookup an appropriate real ifa to point to.
                   2146:  * This should be moved to /sys/net/link.c eventually.
                   2147:  */
1.15      mycroft  2148: void
1.228     dyoung   2149: link_rtrequest(int cmd, struct rtentry *rt, const struct rt_addrinfo *info)
1.1       cgd      2150: {
1.61      augustss 2151:        struct ifaddr *ifa;
1.194     dyoung   2152:        const struct sockaddr *dst;
1.15      mycroft  2153:        struct ifnet *ifp;
1.357     ozaki-r  2154:        struct psref psref;
1.1       cgd      2155:
1.225     dyoung   2156:        if (cmd != RTM_ADD || (ifa = rt->rt_ifa) == NULL ||
                   2157:            (ifp = ifa->ifa_ifp) == NULL || (dst = rt_getkey(rt)) == NULL)
1.1       cgd      2158:                return;
1.357     ozaki-r  2159:        if ((ifa = ifaof_ifpforaddr_psref(dst, ifp, &psref)) != NULL) {
1.176     dyoung   2160:                rt_replace_ifa(rt, ifa);
1.1       cgd      2161:                if (ifa->ifa_rtrequest && ifa->ifa_rtrequest != link_rtrequest)
1.82      itojun   2162:                        ifa->ifa_rtrequest(cmd, rt, info);
1.357     ozaki-r  2163:                ifa_release(ifa, &psref);
1.1       cgd      2164:        }
                   2165: }
                   2166:
                   2167: /*
1.325     roy      2168:  * bitmask macros to manage a densely packed link_state change queue.
                   2169:  * Because we need to store LINK_STATE_UNKNOWN(0), LINK_STATE_DOWN(1) and
                   2170:  * LINK_STATE_UP(2) we need 2 bits for each state change.
                   2171:  * As a state change to store is 0, treat all bits set as an unset item.
                   2172:  */
                   2173: #define LQ_ITEM_BITS           2
                   2174: #define LQ_ITEM_MASK           ((1 << LQ_ITEM_BITS) - 1)
                   2175: #define LQ_MASK(i)             (LQ_ITEM_MASK << (i) * LQ_ITEM_BITS)
                   2176: #define LINK_STATE_UNSET       LQ_ITEM_MASK
                   2177: #define LQ_ITEM(q, i)          (((q) & LQ_MASK((i))) >> (i) * LQ_ITEM_BITS)
                   2178: #define LQ_STORE(q, i, v)                                                    \
                   2179:        do {                                                                  \
                   2180:                (q) &= ~LQ_MASK((i));                                         \
                   2181:                (q) |= (v) << (i) * LQ_ITEM_BITS;                             \
                   2182:        } while (0 /* CONSTCOND */)
                   2183: #define LQ_MAX(q)              ((sizeof((q)) * NBBY) / LQ_ITEM_BITS)
                   2184: #define LQ_POP(q, v)                                                         \
                   2185:        do {                                                                  \
                   2186:                (v) = LQ_ITEM((q), 0);                                        \
                   2187:                (q) >>= LQ_ITEM_BITS;                                         \
                   2188:                (q) |= LINK_STATE_UNSET << (LQ_MAX((q)) - 1) * LQ_ITEM_BITS;  \
                   2189:        } while (0 /* CONSTCOND */)
                   2190: #define LQ_PUSH(q, v)                                                        \
                   2191:        do {                                                                  \
                   2192:                (q) >>= LQ_ITEM_BITS;                                         \
                   2193:                (q) |= (v) << (LQ_MAX((q)) - 1) * LQ_ITEM_BITS;               \
                   2194:        } while (0 /* CONSTCOND */)
                   2195: #define LQ_FIND_UNSET(q, i)                                                  \
                   2196:        for ((i) = 0; i < LQ_MAX((q)); (i)++) {                               \
                   2197:                if (LQ_ITEM((q), (i)) == LINK_STATE_UNSET)                    \
                   2198:                        break;                                                \
                   2199:        }
1.406     ozaki-r  2200:
                   2201: /*
                   2202:  * XXX reusing (ifp)->if_snd->ifq_lock rather than having another spin mutex
                   2203:  * for each ifnet.  It doesn't matter because:
                   2204:  * - if IFEF_MPSAFE is enabled, if_snd isn't used and lock contentions on
                   2205:  *   ifq_lock don't happen
                   2206:  * - if IFEF_MPSAFE is disabled, there is no lock contention on ifq_lock
                   2207:  *   because if_snd, if_link_state_change and if_link_state_change_softint
                   2208:  *   are all called with KERNEL_LOCK
                   2209:  */
                   2210: #define IF_LINK_STATE_CHANGE_LOCK(ifp)         \
                   2211:        mutex_enter((ifp)->if_snd.ifq_lock)
                   2212: #define IF_LINK_STATE_CHANGE_UNLOCK(ifp)       \
                   2213:        mutex_exit((ifp)->if_snd.ifq_lock)
                   2214:
1.325     roy      2215: /*
                   2216:  * Handle a change in the interface link state and
                   2217:  * queue notifications.
1.159     dyoung   2218:  */
                   2219: void
                   2220: if_link_state_change(struct ifnet *ifp, int link_state)
                   2221: {
1.406     ozaki-r  2222:        int idx;
1.263     roy      2223:
1.347     knakahar 2224:        KASSERTMSG(if_is_link_state_changeable(ifp),
1.346     knakahar 2225:            "%s: IFEF_NO_LINK_STATE_CHANGE must not be set, but if_extflags=0x%x",
                   2226:            ifp->if_xname, ifp->if_extflags);
                   2227:
1.325     roy      2228:        /* Ensure change is to a valid state */
                   2229:        switch (link_state) {
                   2230:        case LINK_STATE_UNKNOWN:        /* FALLTHROUGH */
                   2231:        case LINK_STATE_DOWN:           /* FALLTHROUGH */
                   2232:        case LINK_STATE_UP:
                   2233:                break;
                   2234:        default:
                   2235: #ifdef DEBUG
                   2236:                printf("%s: invalid link state %d\n",
                   2237:                    ifp->if_xname, link_state);
                   2238: #endif
1.185     dyoung   2239:                return;
1.264     roy      2240:        }
1.263     roy      2241:
1.406     ozaki-r  2242:        IF_LINK_STATE_CHANGE_LOCK(ifp);
1.325     roy      2243:
                   2244:        /* Find the last unset event in the queue. */
                   2245:        LQ_FIND_UNSET(ifp->if_link_queue, idx);
                   2246:
                   2247:        /*
                   2248:         * Ensure link_state doesn't match the last event in the queue.
                   2249:         * ifp->if_link_state is not checked and set here because
                   2250:         * that would present an inconsistent picture to the system.
                   2251:         */
                   2252:        if (idx != 0 &&
                   2253:            LQ_ITEM(ifp->if_link_queue, idx - 1) == (uint8_t)link_state)
                   2254:                goto out;
                   2255:
                   2256:        /* Handle queue overflow. */
                   2257:        if (idx == LQ_MAX(ifp->if_link_queue)) {
                   2258:                uint8_t lost;
                   2259:
                   2260:                /*
                   2261:                 * The DOWN state must be protected from being pushed off
                   2262:                 * the queue to ensure that userland will always be
                   2263:                 * in a sane state.
                   2264:                 * Because DOWN is protected, there is no need to protect
                   2265:                 * UNKNOWN.
                   2266:                 * It should be invalid to change from any other state to
                   2267:                 * UNKNOWN anyway ...
                   2268:                 */
                   2269:                lost = LQ_ITEM(ifp->if_link_queue, 0);
                   2270:                LQ_PUSH(ifp->if_link_queue, (uint8_t)link_state);
                   2271:                if (lost == LINK_STATE_DOWN) {
                   2272:                        lost = LQ_ITEM(ifp->if_link_queue, 0);
                   2273:                        LQ_STORE(ifp->if_link_queue, 0, LINK_STATE_DOWN);
                   2274:                }
                   2275:                printf("%s: lost link state change %s\n",
                   2276:                    ifp->if_xname,
                   2277:                    lost == LINK_STATE_UP ? "UP" :
                   2278:                    lost == LINK_STATE_DOWN ? "DOWN" :
                   2279:                    "UNKNOWN");
                   2280:        } else
                   2281:                LQ_STORE(ifp->if_link_queue, idx, (uint8_t)link_state);
                   2282:
1.324     ozaki-r  2283:        softint_schedule(ifp->if_link_si);
                   2284:
1.325     roy      2285: out:
1.406     ozaki-r  2286:        IF_LINK_STATE_CHANGE_UNLOCK(ifp);
1.324     ozaki-r  2287: }
                   2288:
1.325     roy      2289: /*
                   2290:  * Handle interface link state change notifications.
                   2291:  */
1.393     ozaki-r  2292: void
                   2293: if_link_state_change_softint(struct ifnet *ifp, int link_state)
1.324     ozaki-r  2294: {
                   2295:        struct domain *dp;
1.393     ozaki-r  2296:        int s = splnet();
1.406     ozaki-r  2297:        bool notify;
1.393     ozaki-r  2298:
                   2299:        KASSERT(!cpu_intr_p());
1.324     ozaki-r  2300:
1.406     ozaki-r  2301:        IF_LINK_STATE_CHANGE_LOCK(ifp);
                   2302:
1.325     roy      2303:        /* Ensure the change is still valid. */
1.393     ozaki-r  2304:        if (ifp->if_link_state == link_state) {
1.406     ozaki-r  2305:                IF_LINK_STATE_CHANGE_UNLOCK(ifp);
1.325     roy      2306:                return;
1.393     ozaki-r  2307:        }
1.324     ozaki-r  2308:
1.263     roy      2309: #ifdef DEBUG
                   2310:        log(LOG_DEBUG, "%s: link state %s (was %s)\n", ifp->if_xname,
                   2311:                link_state == LINK_STATE_UP ? "UP" :
                   2312:                link_state == LINK_STATE_DOWN ? "DOWN" :
                   2313:                "UNKNOWN",
1.325     roy      2314:                ifp->if_link_state == LINK_STATE_UP ? "UP" :
                   2315:                ifp->if_link_state == LINK_STATE_DOWN ? "DOWN" :
1.263     roy      2316:                "UNKNOWN");
                   2317: #endif
                   2318:
                   2319:        /*
                   2320:         * When going from UNKNOWN to UP, we need to mark existing
1.314     roy      2321:         * addresses as tentative and restart DAD as we may have
1.263     roy      2322:         * erroneously not found a duplicate.
                   2323:         *
                   2324:         * This needs to happen before rt_ifmsg to avoid a race where
                   2325:         * listeners would have an address and expect it to work right
                   2326:         * away.
                   2327:         */
1.406     ozaki-r  2328:        notify = (link_state == LINK_STATE_UP &&
                   2329:            ifp->if_link_state == LINK_STATE_UNKNOWN);
                   2330:        ifp->if_link_state = link_state;
                   2331:        /* The following routines may sleep so release the spin mutex */
                   2332:        IF_LINK_STATE_CHANGE_UNLOCK(ifp);
                   2333:
                   2334:        KERNEL_LOCK_UNLESS_NET_MPSAFE();
                   2335:        if (notify) {
1.312     roy      2336:                DOMAIN_FOREACH(dp) {
                   2337:                        if (dp->dom_if_link_state_change != NULL)
                   2338:                                dp->dom_if_link_state_change(ifp,
                   2339:                                    LINK_STATE_DOWN);
                   2340:                }
                   2341:        }
1.263     roy      2342:
1.159     dyoung   2343:        /* Notify that the link state has changed. */
1.185     dyoung   2344:        rt_ifmsg(ifp);
1.263     roy      2345:
1.166     liamjfoy 2346: #if NCARP > 0
1.185     dyoung   2347:        if (ifp->if_carp)
                   2348:                carp_carpdev_state(ifp);
1.166     liamjfoy 2349: #endif
1.263     roy      2350:
1.312     roy      2351:        DOMAIN_FOREACH(dp) {
                   2352:                if (dp->dom_if_link_state_change != NULL)
                   2353:                        dp->dom_if_link_state_change(ifp, link_state);
1.270     pooka    2354:        }
1.406     ozaki-r  2355:        KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
1.393     ozaki-r  2356:        splx(s);
1.325     roy      2357: }
                   2358:
                   2359: /*
                   2360:  * Process the interface link state change queue.
                   2361:  */
                   2362: static void
                   2363: if_link_state_change_si(void *arg)
                   2364: {
                   2365:        struct ifnet *ifp = arg;
                   2366:        int s;
                   2367:        uint8_t state;
1.406     ozaki-r  2368:        bool schedule;
1.325     roy      2369:
1.397     ozaki-r  2370:        SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE();
1.325     roy      2371:        s = splnet();
                   2372:
                   2373:        /* Pop a link state change from the queue and process it. */
1.406     ozaki-r  2374:        IF_LINK_STATE_CHANGE_LOCK(ifp);
1.325     roy      2375:        LQ_POP(ifp->if_link_queue, state);
1.406     ozaki-r  2376:        IF_LINK_STATE_CHANGE_UNLOCK(ifp);
                   2377:
1.393     ozaki-r  2378:        if_link_state_change_softint(ifp, state);
1.325     roy      2379:
                   2380:        /* If there is a link state change to come, schedule it. */
1.406     ozaki-r  2381:        IF_LINK_STATE_CHANGE_LOCK(ifp);
                   2382:        schedule = (LQ_ITEM(ifp->if_link_queue, 0) != LINK_STATE_UNSET);
                   2383:        IF_LINK_STATE_CHANGE_UNLOCK(ifp);
                   2384:        if (schedule)
1.325     roy      2385:                softint_schedule(ifp->if_link_si);
1.264     roy      2386:
                   2387:        splx(s);
1.397     ozaki-r  2388:        SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
1.159     dyoung   2389: }
                   2390:
                   2391: /*
1.310     roy      2392:  * Default action when installing a local route on a point-to-point
                   2393:  * interface.
                   2394:  */
                   2395: void
                   2396: p2p_rtrequest(int req, struct rtentry *rt,
                   2397:     __unused const struct rt_addrinfo *info)
                   2398: {
                   2399:        struct ifnet *ifp = rt->rt_ifp;
                   2400:        struct ifaddr *ifa, *lo0ifa;
1.357     ozaki-r  2401:        int s = pserialize_read_enter();
1.310     roy      2402:
                   2403:        switch (req) {
                   2404:        case RTM_ADD:
                   2405:                if ((rt->rt_flags & RTF_LOCAL) == 0)
                   2406:                        break;
                   2407:
1.351     ozaki-r  2408:                rt->rt_ifp = lo0ifp;
                   2409:
1.354     ozaki-r  2410:                IFADDR_READER_FOREACH(ifa, ifp) {
1.310     roy      2411:                        if (equal(rt_getkey(rt), ifa->ifa_addr))
                   2412:                                break;
                   2413:                }
                   2414:                if (ifa == NULL)
                   2415:                        break;
                   2416:
                   2417:                /*
                   2418:                 * Ensure lo0 has an address of the same family.
                   2419:                 */
1.354     ozaki-r  2420:                IFADDR_READER_FOREACH(lo0ifa, lo0ifp) {
1.310     roy      2421:                        if (lo0ifa->ifa_addr->sa_family ==
                   2422:                            ifa->ifa_addr->sa_family)
                   2423:                                break;
                   2424:                }
                   2425:                if (lo0ifa == NULL)
                   2426:                        break;
                   2427:
                   2428:                /*
                   2429:                 * Make sure to set rt->rt_ifa to the interface
                   2430:                 * address we are using, otherwise we will have trouble
                   2431:                 * with source address selection.
                   2432:                 */
                   2433:                if (ifa != rt->rt_ifa)
                   2434:                        rt_replace_ifa(rt, ifa);
                   2435:                break;
                   2436:        case RTM_DELETE:
                   2437:        default:
                   2438:                break;
                   2439:        }
1.357     ozaki-r  2440:        pserialize_read_exit(s);
1.310     roy      2441: }
                   2442:
1.403     ozaki-r  2443: static void
                   2444: _if_down(struct ifnet *ifp)
1.1       cgd      2445: {
1.61      augustss 2446:        struct ifaddr *ifa;
1.312     roy      2447:        struct domain *dp;
1.357     ozaki-r  2448:        int s, bound;
                   2449:        struct psref psref;
1.1       cgd      2450:
                   2451:        ifp->if_flags &= ~IFF_UP;
1.232     christos 2452:        nanotime(&ifp->if_lastchange);
1.357     ozaki-r  2453:
                   2454:        bound = curlwp_bind();
                   2455:        s = pserialize_read_enter();
                   2456:        IFADDR_READER_FOREACH(ifa, ifp) {
                   2457:                ifa_acquire(ifa, &psref);
                   2458:                pserialize_read_exit(s);
                   2459:
1.1       cgd      2460:                pfctlinput(PRC_IFDOWN, ifa->ifa_addr);
1.357     ozaki-r  2461:
                   2462:                s = pserialize_read_enter();
                   2463:                ifa_release(ifa, &psref);
                   2464:        }
                   2465:        pserialize_read_exit(s);
                   2466:        curlwp_bindx(bound);
                   2467:
1.78      thorpej  2468:        IFQ_PURGE(&ifp->if_snd);
1.166     liamjfoy 2469: #if NCARP > 0
                   2470:        if (ifp->if_carp)
                   2471:                carp_carpdev_state(ifp);
                   2472: #endif
1.15      mycroft  2473:        rt_ifmsg(ifp);
1.312     roy      2474:        DOMAIN_FOREACH(dp) {
                   2475:                if (dp->dom_if_down)
                   2476:                        dp->dom_if_down(ifp);
                   2477:        }
1.15      mycroft  2478: }
                   2479:
1.403     ozaki-r  2480: static void
                   2481: if_down_deactivated(struct ifnet *ifp)
                   2482: {
                   2483:
                   2484:        KASSERT(if_is_deactivated(ifp));
                   2485:        _if_down(ifp);
                   2486: }
                   2487:
                   2488: void
                   2489: if_down_locked(struct ifnet *ifp)
                   2490: {
                   2491:
1.413     ozaki-r  2492:        KASSERT(IFNET_LOCKED(ifp));
1.403     ozaki-r  2493:        _if_down(ifp);
                   2494: }
                   2495:
1.15      mycroft  2496: /*
1.403     ozaki-r  2497:  * Mark an interface down and notify protocols of
1.15      mycroft  2498:  * the transition.
1.23      mycroft  2499:  * NOTE: must be called at splsoftnet or equivalent.
1.15      mycroft  2500:  */
                   2501: void
1.403     ozaki-r  2502: if_down(struct ifnet *ifp)
                   2503: {
                   2504:
1.413     ozaki-r  2505:        IFNET_LOCK(ifp);
1.403     ozaki-r  2506:        if_down_locked(ifp);
1.413     ozaki-r  2507:        IFNET_UNLOCK(ifp);
1.403     ozaki-r  2508: }
                   2509:
                   2510: /*
                   2511:  * Must be called with holding if_ioctl_lock.
                   2512:  */
                   2513: static void
                   2514: if_up_locked(struct ifnet *ifp)
1.15      mycroft  2515: {
1.24      christos 2516: #ifdef notyet
1.61      augustss 2517:        struct ifaddr *ifa;
1.24      christos 2518: #endif
1.312     roy      2519:        struct domain *dp;
1.15      mycroft  2520:
1.413     ozaki-r  2521:        KASSERT(IFNET_LOCKED(ifp));
1.403     ozaki-r  2522:
                   2523:        KASSERT(!if_is_deactivated(ifp));
1.15      mycroft  2524:        ifp->if_flags |= IFF_UP;
1.232     christos 2525:        nanotime(&ifp->if_lastchange);
1.15      mycroft  2526: #ifdef notyet
                   2527:        /* this has no effect on IP, and will kill all ISO connections XXX */
1.354     ozaki-r  2528:        IFADDR_READER_FOREACH(ifa, ifp)
1.15      mycroft  2529:                pfctlinput(PRC_IFUP, ifa->ifa_addr);
                   2530: #endif
1.166     liamjfoy 2531: #if NCARP > 0
                   2532:        if (ifp->if_carp)
                   2533:                carp_carpdev_state(ifp);
                   2534: #endif
1.15      mycroft  2535:        rt_ifmsg(ifp);
1.312     roy      2536:        DOMAIN_FOREACH(dp) {
1.313     roy      2537:                if (dp->dom_if_up)
                   2538:                        dp->dom_if_up(ifp);
1.312     roy      2539:        }
1.1       cgd      2540: }
                   2541:
                   2542: /*
1.296     ozaki-r  2543:  * Handle interface slowtimo timer routine.  Called
                   2544:  * from softclock, we decrement timer (if set) and
1.1       cgd      2545:  * call the appropriate interface routine on expiration.
                   2546:  */
1.294     ozaki-r  2547: static void
1.177     christos 2548: if_slowtimo(void *arg)
1.1       cgd      2549: {
1.306     martin   2550:        void (*slowtimo)(struct ifnet *);
1.296     ozaki-r  2551:        struct ifnet *ifp = arg;
1.305     martin   2552:        int s;
1.1       cgd      2553:
1.306     martin   2554:        slowtimo = ifp->if_slowtimo;
                   2555:        if (__predict_false(slowtimo == NULL))
1.305     martin   2556:                return;
1.296     ozaki-r  2557:
1.374     ozaki-r  2558:        s = splnet();
1.411     ozaki-r  2559:        if (ifp->if_timer != 0 && --ifp->if_timer == 0)
1.306     martin   2560:                (*slowtimo)(ifp);
1.411     ozaki-r  2561:
1.1       cgd      2562:        splx(s);
1.305     martin   2563:
                   2564:        if (__predict_true(ifp->if_slowtimo != NULL))
                   2565:                callout_schedule(ifp->if_slowtimo_ch, hz / IFNET_SLOWHZ);
1.65      thorpej  2566: }
                   2567:
                   2568: /*
1.403     ozaki-r  2569:  * Mark an interface up and notify protocols of
                   2570:  * the transition.
                   2571:  * NOTE: must be called at splsoftnet or equivalent.
                   2572:  */
                   2573: void
                   2574: if_up(struct ifnet *ifp)
                   2575: {
                   2576:
1.413     ozaki-r  2577:        IFNET_LOCK(ifp);
1.403     ozaki-r  2578:        if_up_locked(ifp);
1.413     ozaki-r  2579:        IFNET_UNLOCK(ifp);
1.403     ozaki-r  2580: }
                   2581:
                   2582: /*
1.65      thorpej  2583:  * Set/clear promiscuous mode on interface ifp based on the truth value
                   2584:  * of pswitch.  The calls are reference counted so that only the first
                   2585:  * "on" request actually has an effect, as does the final "off" request.
                   2586:  * Results are undefined if the "off" and "on" requests are not matched.
                   2587:  */
                   2588: int
1.402     ozaki-r  2589: ifpromisc_locked(struct ifnet *ifp, int pswitch)
1.65      thorpej  2590: {
1.401     ozaki-r  2591:        int pcount, ret = 0;
1.259     dyoung   2592:        short nflags;
1.65      thorpej  2593:
1.413     ozaki-r  2594:        KASSERT(IFNET_LOCKED(ifp));
1.401     ozaki-r  2595:
1.65      thorpej  2596:        pcount = ifp->if_pcount;
                   2597:        if (pswitch) {
                   2598:                /*
1.71      thorpej  2599:                 * Allow the device to be "placed" into promiscuous
                   2600:                 * mode even if it is not configured up.  It will
1.242     mbalmer  2601:                 * consult IFF_PROMISC when it is brought up.
1.65      thorpej  2602:                 */
1.68      pk       2603:                if (ifp->if_pcount++ != 0)
1.401     ozaki-r  2604:                        goto out;
1.252     dyoung   2605:                nflags = ifp->if_flags | IFF_PROMISC;
1.65      thorpej  2606:        } else {
                   2607:                if (--ifp->if_pcount > 0)
1.401     ozaki-r  2608:                        goto out;
1.252     dyoung   2609:                nflags = ifp->if_flags & ~IFF_PROMISC;
1.65      thorpej  2610:        }
1.252     dyoung   2611:        ret = if_flags_set(ifp, nflags);
1.65      thorpej  2612:        /* Restore interface state if not successful. */
                   2613:        if (ret != 0) {
                   2614:                ifp->if_pcount = pcount;
                   2615:        }
1.401     ozaki-r  2616: out:
1.402     ozaki-r  2617:        return ret;
                   2618: }
                   2619:
                   2620: int
                   2621: ifpromisc(struct ifnet *ifp, int pswitch)
                   2622: {
                   2623:        int e;
                   2624:
1.413     ozaki-r  2625:        IFNET_LOCK(ifp);
1.402     ozaki-r  2626:        e = ifpromisc_locked(ifp, pswitch);
1.413     ozaki-r  2627:        IFNET_UNLOCK(ifp);
1.402     ozaki-r  2628:
                   2629:        return e;
1.1       cgd      2630: }
                   2631:
                   2632: /*
                   2633:  * Map interface name to
                   2634:  * interface structure pointer.
                   2635:  */
                   2636: struct ifnet *
1.163     thorpej  2637: ifunit(const char *name)
1.1       cgd      2638: {
1.61      augustss 2639:        struct ifnet *ifp;
1.105     matt     2640:        const char *cp = name;
                   2641:        u_int unit = 0;
                   2642:        u_int i;
1.334     ozaki-r  2643:        int s;
1.105     matt     2644:
                   2645:        /*
                   2646:         * If the entire name is a number, treat it as an ifindex.
                   2647:         */
                   2648:        for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) {
                   2649:                unit = unit * 10 + (*cp - '0');
                   2650:        }
                   2651:
                   2652:        /*
                   2653:         * If the number took all of the name, then it's a valid ifindex.
                   2654:         */
1.387     ozaki-r  2655:        if (i == IFNAMSIZ || (cp != name && *cp == '\0'))
                   2656:                return if_byindex(unit);
1.34      thorpej  2657:
1.334     ozaki-r  2658:        ifp = NULL;
                   2659:        s = pserialize_read_enter();
                   2660:        IFNET_READER_FOREACH(ifp) {
1.348     ozaki-r  2661:                if (if_is_deactivated(ifp))
1.53      thorpej  2662:                        continue;
                   2663:                if (strcmp(ifp->if_xname, name) == 0)
1.334     ozaki-r  2664:                        goto out;
1.53      thorpej  2665:        }
1.334     ozaki-r  2666: out:
                   2667:        pserialize_read_exit(s);
                   2668:        return ifp;
1.1       cgd      2669: }
1.49      itojun   2670:
1.335     ozaki-r  2671: /*
                   2672:  * Get a reference of an ifnet object by an interface name.
                   2673:  * The returned reference is protected by psref(9). The caller
                   2674:  * must release a returned reference by if_put after use.
                   2675:  */
                   2676: struct ifnet *
                   2677: if_get(const char *name, struct psref *psref)
                   2678: {
                   2679:        struct ifnet *ifp;
                   2680:        const char *cp = name;
                   2681:        u_int unit = 0;
                   2682:        u_int i;
                   2683:        int s;
                   2684:
                   2685:        /*
                   2686:         * If the entire name is a number, treat it as an ifindex.
                   2687:         */
                   2688:        for (i = 0; i < IFNAMSIZ && *cp >= '0' && *cp <= '9'; i++, cp++) {
                   2689:                unit = unit * 10 + (*cp - '0');
                   2690:        }
                   2691:
                   2692:        /*
                   2693:         * If the number took all of the name, then it's a valid ifindex.
                   2694:         */
1.387     ozaki-r  2695:        if (i == IFNAMSIZ || (cp != name && *cp == '\0'))
                   2696:                return if_get_byindex(unit, psref);
1.335     ozaki-r  2697:
                   2698:        ifp = NULL;
                   2699:        s = pserialize_read_enter();
                   2700:        IFNET_READER_FOREACH(ifp) {
1.348     ozaki-r  2701:                if (if_is_deactivated(ifp))
1.335     ozaki-r  2702:                        continue;
                   2703:                if (strcmp(ifp->if_xname, name) == 0) {
                   2704:                        psref_acquire(psref, &ifp->if_psref,
                   2705:                            ifnet_psref_class);
                   2706:                        goto out;
                   2707:                }
                   2708:        }
                   2709: out:
                   2710:        pserialize_read_exit(s);
                   2711:        return ifp;
                   2712: }
                   2713:
                   2714: /*
1.395     roy      2715:  * Release a reference of an ifnet object given by if_get, if_get_byindex
                   2716:  * or if_get_bylla.
1.335     ozaki-r  2717:  */
                   2718: void
                   2719: if_put(const struct ifnet *ifp, struct psref *psref)
                   2720: {
                   2721:
1.344     ozaki-r  2722:        if (ifp == NULL)
                   2723:                return;
                   2724:
1.335     ozaki-r  2725:        psref_release(psref, &ifp->if_psref, ifnet_psref_class);
                   2726: }
                   2727:
1.415     ozaki-r  2728: /*
                   2729:  * Return ifp having idx. Return NULL if not found.  Normally if_byindex
                   2730:  * should be used.
                   2731:  */
                   2732: ifnet_t *
                   2733: _if_byindex(u_int idx)
                   2734: {
                   2735:
                   2736:        return (__predict_true(idx < if_indexlim)) ? ifindex2ifnet[idx] : NULL;
                   2737: }
                   2738:
                   2739: /*
                   2740:  * Return ifp having idx. Return NULL if not found or the found ifp is
                   2741:  * already deactivated.
                   2742:  */
1.250     rmind    2743: ifnet_t *
                   2744: if_byindex(u_int idx)
                   2745: {
1.357     ozaki-r  2746:        ifnet_t *ifp;
                   2747:
1.415     ozaki-r  2748:        ifp = _if_byindex(idx);
1.357     ozaki-r  2749:        if (ifp != NULL && if_is_deactivated(ifp))
                   2750:                ifp = NULL;
                   2751:        return ifp;
1.250     rmind    2752: }
                   2753:
1.335     ozaki-r  2754: /*
                   2755:  * Get a reference of an ifnet object by an interface index.
                   2756:  * The returned reference is protected by psref(9). The caller
                   2757:  * must release a returned reference by if_put after use.
                   2758:  */
                   2759: ifnet_t *
                   2760: if_get_byindex(u_int idx, struct psref *psref)
                   2761: {
                   2762:        ifnet_t *ifp;
                   2763:        int s;
                   2764:
                   2765:        s = pserialize_read_enter();
1.387     ozaki-r  2766:        ifp = if_byindex(idx);
1.337     ozaki-r  2767:        if (__predict_true(ifp != NULL))
1.335     ozaki-r  2768:                psref_acquire(psref, &ifp->if_psref, ifnet_psref_class);
                   2769:        pserialize_read_exit(s);
                   2770:
                   2771:        return ifp;
                   2772: }
                   2773:
1.395     roy      2774: ifnet_t *
                   2775: if_get_bylla(const void *lla, unsigned char lla_len, struct psref *psref)
                   2776: {
                   2777:        ifnet_t *ifp;
                   2778:        int s;
                   2779:
                   2780:        s = pserialize_read_enter();
                   2781:        IFNET_READER_FOREACH(ifp) {
                   2782:                if (if_is_deactivated(ifp))
                   2783:                        continue;
                   2784:                if (ifp->if_addrlen != lla_len)
                   2785:                        continue;
                   2786:                if (memcmp(lla, CLLADDR(ifp->if_sadl), lla_len) == 0) {
                   2787:                        psref_acquire(psref, &ifp->if_psref,
                   2788:                            ifnet_psref_class);
                   2789:                        break;
                   2790:                }
                   2791:        }
                   2792:        pserialize_read_exit(s);
                   2793:
                   2794:        return ifp;
                   2795: }
                   2796:
1.338     ozaki-r  2797: /*
1.380     ozaki-r  2798:  * Note that it's safe only if the passed ifp is guaranteed to not be freed,
                   2799:  * for example using pserialize or the ifp is already held or some other
                   2800:  * object is held which guarantes the ifp to not be freed indirectly.
1.338     ozaki-r  2801:  */
                   2802: void
1.380     ozaki-r  2803: if_acquire(struct ifnet *ifp, struct psref *psref)
1.338     ozaki-r  2804: {
                   2805:
                   2806:        KASSERT(ifp->if_index != 0);
                   2807:        psref_acquire(psref, &ifp->if_psref, ifnet_psref_class);
                   2808: }
                   2809:
                   2810: bool
                   2811: if_held(struct ifnet *ifp)
                   2812: {
                   2813:
                   2814:        return psref_held(&ifp->if_psref, ifnet_psref_class);
                   2815: }
                   2816:
1.404     knakahar 2817: /*
                   2818:  * Some tunnel interfaces can nest, e.g. IPv4 over IPv4 gif(4) tunnel over IPv4.
                   2819:  * Check the tunnel nesting count.
                   2820:  * Return > 0, if tunnel nesting count is more than limit.
                   2821:  * Return 0, if tunnel nesting count is equal or less than limit.
                   2822:  */
                   2823: int
                   2824: if_tunnel_check_nesting(struct ifnet *ifp, struct mbuf *m, int limit)
                   2825: {
                   2826:        struct m_tag *mtag;
                   2827:        int *count;
                   2828:
                   2829:        mtag = m_tag_find(m, PACKET_TAG_TUNNEL_INFO, NULL);
                   2830:        if (mtag != NULL) {
                   2831:                count = (int *)(mtag + 1);
                   2832:                if (++(*count) > limit) {
                   2833:                        log(LOG_NOTICE,
                   2834:                            "%s: recursively called too many times(%d)\n",
                   2835:                            ifp->if_xname, *count);
                   2836:                        return EIO;
                   2837:                }
                   2838:        } else {
                   2839:                mtag = m_tag_get(PACKET_TAG_TUNNEL_INFO, sizeof(*count),
                   2840:                    M_NOWAIT);
                   2841:                if (mtag != NULL) {
                   2842:                        m_tag_prepend(m, mtag);
                   2843:                        count = (int *)(mtag + 1);
                   2844:                        *count = 0;
                   2845:                } else {
                   2846:                        log(LOG_DEBUG,
                   2847:                            "%s: m_tag_get() failed, recursion calls are not prevented.\n",
                   2848:                            ifp->if_xname);
                   2849:                }
                   2850:        }
                   2851:
                   2852:        return 0;
                   2853: }
1.338     ozaki-r  2854:
1.211     dyoung   2855: /* common */
1.215     dyoung   2856: int
                   2857: ifioctl_common(struct ifnet *ifp, u_long cmd, void *data)
1.211     dyoung   2858: {
1.224     dyoung   2859:        int s;
1.215     dyoung   2860:        struct ifreq *ifr;
                   2861:        struct ifcapreq *ifcr;
                   2862:        struct ifdatareq *ifdr;
1.211     dyoung   2863:
                   2864:        switch (cmd) {
                   2865:        case SIOCSIFCAP:
1.215     dyoung   2866:                ifcr = data;
1.211     dyoung   2867:                if ((ifcr->ifcr_capenable & ~ifp->if_capabilities) != 0)
                   2868:                        return EINVAL;
                   2869:
1.213     dyoung   2870:                if (ifcr->ifcr_capenable == ifp->if_capenable)
                   2871:                        return 0;
1.211     dyoung   2872:
1.213     dyoung   2873:                ifp->if_capenable = ifcr->ifcr_capenable;
1.211     dyoung   2874:
1.213     dyoung   2875:                /* Pre-compute the checksum flags mask. */
                   2876:                ifp->if_csum_flags_tx = 0;
                   2877:                ifp->if_csum_flags_rx = 0;
                   2878:                if (ifp->if_capenable & IFCAP_CSUM_IPv4_Tx) {
                   2879:                        ifp->if_csum_flags_tx |= M_CSUM_IPv4;
                   2880:                }
                   2881:                if (ifp->if_capenable & IFCAP_CSUM_IPv4_Rx) {
                   2882:                        ifp->if_csum_flags_rx |= M_CSUM_IPv4;
                   2883:                }
1.211     dyoung   2884:
1.213     dyoung   2885:                if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Tx) {
                   2886:                        ifp->if_csum_flags_tx |= M_CSUM_TCPv4;
                   2887:                }
                   2888:                if (ifp->if_capenable & IFCAP_CSUM_TCPv4_Rx) {
                   2889:                        ifp->if_csum_flags_rx |= M_CSUM_TCPv4;
                   2890:                }
1.211     dyoung   2891:
1.213     dyoung   2892:                if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Tx) {
                   2893:                        ifp->if_csum_flags_tx |= M_CSUM_UDPv4;
                   2894:                }
                   2895:                if (ifp->if_capenable & IFCAP_CSUM_UDPv4_Rx) {
                   2896:                        ifp->if_csum_flags_rx |= M_CSUM_UDPv4;
                   2897:                }
1.211     dyoung   2898:
1.213     dyoung   2899:                if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Tx) {
                   2900:                        ifp->if_csum_flags_tx |= M_CSUM_TCPv6;
                   2901:                }
                   2902:                if (ifp->if_capenable & IFCAP_CSUM_TCPv6_Rx) {
                   2903:                        ifp->if_csum_flags_rx |= M_CSUM_TCPv6;
                   2904:                }
1.211     dyoung   2905:
1.213     dyoung   2906:                if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Tx) {
                   2907:                        ifp->if_csum_flags_tx |= M_CSUM_UDPv6;
1.211     dyoung   2908:                }
1.213     dyoung   2909:                if (ifp->if_capenable & IFCAP_CSUM_UDPv6_Rx) {
                   2910:                        ifp->if_csum_flags_rx |= M_CSUM_UDPv6;
                   2911:                }
1.215     dyoung   2912:                if (ifp->if_flags & IFF_UP)
                   2913:                        return ENETRESET;
                   2914:                return 0;
1.211     dyoung   2915:        case SIOCSIFFLAGS:
1.215     dyoung   2916:                ifr = data;
1.399     ozaki-r  2917:                /*
                   2918:                 * If if_is_mpsafe(ifp), KERNEL_LOCK isn't held here, but if_up
                   2919:                 * and if_down aren't MP-safe yet, so we must hold the lock.
                   2920:                 */
                   2921:                KERNEL_LOCK_IF_IFP_MPSAFE(ifp);
1.211     dyoung   2922:                if (ifp->if_flags & IFF_UP && (ifr->ifr_flags & IFF_UP) == 0) {
1.373     ozaki-r  2923:                        s = splsoftnet();
1.403     ozaki-r  2924:                        if_down_locked(ifp);
1.211     dyoung   2925:                        splx(s);
                   2926:                }
                   2927:                if (ifr->ifr_flags & IFF_UP && (ifp->if_flags & IFF_UP) == 0) {
1.373     ozaki-r  2928:                        s = splsoftnet();
1.403     ozaki-r  2929:                        if_up_locked(ifp);
1.211     dyoung   2930:                        splx(s);
                   2931:                }
1.399     ozaki-r  2932:                KERNEL_UNLOCK_IF_IFP_MPSAFE(ifp);
1.211     dyoung   2933:                ifp->if_flags = (ifp->if_flags & IFF_CANTCHANGE) |
                   2934:                        (ifr->ifr_flags &~ IFF_CANTCHANGE);
                   2935:                break;
                   2936:        case SIOCGIFFLAGS:
1.215     dyoung   2937:                ifr = data;
1.211     dyoung   2938:                ifr->ifr_flags = ifp->if_flags;
                   2939:                break;
                   2940:
                   2941:        case SIOCGIFMETRIC:
1.215     dyoung   2942:                ifr = data;
1.211     dyoung   2943:                ifr->ifr_metric = ifp->if_metric;
                   2944:                break;
                   2945:
                   2946:        case SIOCGIFMTU:
1.215     dyoung   2947:                ifr = data;
1.211     dyoung   2948:                ifr->ifr_mtu = ifp->if_mtu;
                   2949:                break;
                   2950:
                   2951:        case SIOCGIFDLT:
1.215     dyoung   2952:                ifr = data;
1.211     dyoung   2953:                ifr->ifr_dlt = ifp->if_dlt;
                   2954:                break;
                   2955:
                   2956:        case SIOCGIFCAP:
1.215     dyoung   2957:                ifcr = data;
1.211     dyoung   2958:                ifcr->ifcr_capabilities = ifp->if_capabilities;
                   2959:                ifcr->ifcr_capenable = ifp->if_capenable;
                   2960:                break;
                   2961:
                   2962:        case SIOCSIFMETRIC:
1.215     dyoung   2963:                ifr = data;
1.211     dyoung   2964:                ifp->if_metric = ifr->ifr_metric;
                   2965:                break;
                   2966:
                   2967:        case SIOCGIFDATA:
1.215     dyoung   2968:                ifdr = data;
1.211     dyoung   2969:                ifdr->ifdr_data = ifp->if_data;
                   2970:                break;
                   2971:
1.266     christos 2972:        case SIOCGIFINDEX:
                   2973:                ifr = data;
                   2974:                ifr->ifr_index = ifp->if_index;
                   2975:                break;
                   2976:
1.211     dyoung   2977:        case SIOCZIFDATA:
1.215     dyoung   2978:                ifdr = data;
1.211     dyoung   2979:                ifdr->ifdr_data = ifp->if_data;
                   2980:                /*
                   2981:                 * Assumes that the volatile counters that can be
                   2982:                 * zero'ed are at the end of if_data.
                   2983:                 */
                   2984:                memset(&ifp->if_data.ifi_ipackets, 0, sizeof(ifp->if_data) -
                   2985:                    offsetof(struct if_data, ifi_ipackets));
1.261     msaitoh  2986:                /*
                   2987:                 * The memset() clears to the bottm of if_data. In the area,
                   2988:                 * if_lastchange is included. Please be careful if new entry
                   2989:                 * will be added into if_data or rewite this.
                   2990:                 *
                   2991:                 * And also, update if_lastchnage.
                   2992:                 */
                   2993:                getnanotime(&ifp->if_lastchange);
1.211     dyoung   2994:                break;
1.215     dyoung   2995:        case SIOCSIFMTU:
                   2996:                ifr = data;
                   2997:                if (ifp->if_mtu == ifr->ifr_mtu)
                   2998:                        break;
                   2999:                ifp->if_mtu = ifr->ifr_mtu;
                   3000:                /*
                   3001:                 * If the link MTU changed, do network layer specific procedure.
                   3002:                 */
                   3003: #ifdef INET6
1.399     ozaki-r  3004:                KERNEL_LOCK_UNLESS_NET_MPSAFE();
1.271     pooka    3005:                if (in6_present)
                   3006:                        nd6_setmtu(ifp);
1.399     ozaki-r  3007:                KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
1.215     dyoung   3008: #endif
                   3009:                return ENETRESET;
1.211     dyoung   3010:        default:
1.223     dyoung   3011:                return ENOTTY;
1.211     dyoung   3012:        }
                   3013:        return 0;
                   3014: }
                   3015:
1.235     dyoung   3016: int
1.284     rtr      3017: ifaddrpref_ioctl(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
1.235     dyoung   3018: {
                   3019:        struct if_addrprefreq *ifap = (struct if_addrprefreq *)data;
                   3020:        struct ifaddr *ifa;
                   3021:        const struct sockaddr *any, *sa;
                   3022:        union {
                   3023:                struct sockaddr sa;
                   3024:                struct sockaddr_storage ss;
1.236     jakllsch 3025:        } u, v;
1.357     ozaki-r  3026:        int s, error = 0;
1.235     dyoung   3027:
                   3028:        switch (cmd) {
                   3029:        case SIOCSIFADDRPREF:
1.284     rtr      3030:                if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_INTERFACE,
1.235     dyoung   3031:                    KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd,
                   3032:                    NULL) != 0)
                   3033:                        return EPERM;
                   3034:        case SIOCGIFADDRPREF:
                   3035:                break;
                   3036:        default:
                   3037:                return EOPNOTSUPP;
                   3038:        }
                   3039:
                   3040:        /* sanity checks */
                   3041:        if (data == NULL || ifp == NULL) {
                   3042:                panic("invalid argument to %s", __func__);
                   3043:                /*NOTREACHED*/
                   3044:        }
                   3045:
                   3046:        /* address must be specified on ADD and DELETE */
                   3047:        sa = sstocsa(&ifap->ifap_addr);
                   3048:        if (sa->sa_family != sofamily(so))
                   3049:                return EINVAL;
                   3050:        if ((any = sockaddr_any(sa)) == NULL || sa->sa_len != any->sa_len)
                   3051:                return EINVAL;
                   3052:
1.236     jakllsch 3053:        sockaddr_externalize(&v.sa, sizeof(v.ss), sa);
                   3054:
1.357     ozaki-r  3055:        s = pserialize_read_enter();
1.354     ozaki-r  3056:        IFADDR_READER_FOREACH(ifa, ifp) {
1.235     dyoung   3057:                if (ifa->ifa_addr->sa_family != sa->sa_family)
                   3058:                        continue;
                   3059:                sockaddr_externalize(&u.sa, sizeof(u.ss), ifa->ifa_addr);
1.236     jakllsch 3060:                if (sockaddr_cmp(&u.sa, &v.sa) == 0)
1.235     dyoung   3061:                        break;
                   3062:        }
1.357     ozaki-r  3063:        if (ifa == NULL) {
                   3064:                error = EADDRNOTAVAIL;
                   3065:                goto out;
                   3066:        }
1.235     dyoung   3067:
                   3068:        switch (cmd) {
                   3069:        case SIOCSIFADDRPREF:
                   3070:                ifa->ifa_preference = ifap->ifap_preference;
1.357     ozaki-r  3071:                goto out;
1.235     dyoung   3072:        case SIOCGIFADDRPREF:
                   3073:                /* fill in the if_laddrreq structure */
                   3074:                (void)sockaddr_copy(sstosa(&ifap->ifap_addr),
                   3075:                    sizeof(ifap->ifap_addr), ifa->ifa_addr);
                   3076:                ifap->ifap_preference = ifa->ifa_preference;
1.357     ozaki-r  3077:                goto out;
1.235     dyoung   3078:        default:
1.357     ozaki-r  3079:                error = EOPNOTSUPP;
1.235     dyoung   3080:        }
1.357     ozaki-r  3081: out:
                   3082:        pserialize_read_exit(s);
                   3083:        return error;
1.235     dyoung   3084: }
                   3085:
1.1       cgd      3086: /*
                   3087:  * Interface ioctls.
                   3088:  */
1.273     pooka    3089: static int
                   3090: doifioctl(struct socket *so, u_long cmd, void *data, struct lwp *l)
1.1       cgd      3091: {
1.61      augustss 3092:        struct ifnet *ifp;
                   3093:        struct ifreq *ifr;
1.217     martin   3094:        int error = 0;
1.191     christos 3095: #if defined(COMPAT_OSOCK) || defined(COMPAT_OIFREQ)
                   3096:        u_long ocmd = cmd;
                   3097: #endif
1.49      itojun   3098:        short oif_flags;
1.186     christos 3099: #ifdef COMPAT_OIFREQ
                   3100:        struct ifreq ifrb;
1.187     xtraeme  3101:        struct oifreq *oifr = NULL;
1.186     christos 3102: #endif
1.292     christos 3103:        int r;
1.336     ozaki-r  3104:        struct psref psref;
1.339     ozaki-r  3105:        int bound;
1.1       cgd      3106:
                   3107:        switch (cmd) {
                   3108:        case SIOCGIFCONF:
1.185     dyoung   3109:                return ifconf(cmd, data);
1.231     dyoung   3110:        case SIOCINITIFADDR:
                   3111:                return EPERM;
1.420     christos 3112:        default:
                   3113:                error = (*vec_compat_ifconf)(l, cmd, data);
                   3114:                if (error != ENOSYS)
                   3115:                        return error;
                   3116:                error = (*vec_compat_ifdatareq)(l, cmd, data);
                   3117:                if (error != ENOSYS)
                   3118:                        return error;
                   3119:                break;
1.1       cgd      3120:        }
1.191     christos 3121:
1.420     christos 3122:        ifr = data;
1.186     christos 3123: #ifdef COMPAT_OIFREQ
1.420     christos 3124:        if (vec_compat_cvtcmd) {
                   3125:                cmd = (*vec_compat_cvtcmd)(cmd);
                   3126:                if (cmd != ocmd) {
                   3127:                        oifr = data;
                   3128:                        data = ifr = &ifrb;
                   3129:                        ifreqo2n(oifr, ifr);
                   3130:                }
                   3131:        }
1.186     christos 3132: #endif
1.63      thorpej  3133:
                   3134:        switch (cmd) {
                   3135:        case SIOCIFCREATE:
                   3136:        case SIOCIFDESTROY:
1.339     ozaki-r  3137:                bound = curlwp_bind();
1.185     dyoung   3138:                if (l != NULL) {
1.336     ozaki-r  3139:                        ifp = if_get(ifr->ifr_name, &psref);
1.174     elad     3140:                        error = kauth_authorize_network(l->l_cred,
                   3141:                            KAUTH_NETWORK_INTERFACE,
                   3142:                            KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
                   3143:                            (void *)cmd, NULL);
1.336     ozaki-r  3144:                        if (ifp != NULL)
                   3145:                                if_put(ifp, &psref);
                   3146:                        if (error != 0) {
1.339     ozaki-r  3147:                                curlwp_bindx(bound);
1.151     yamt     3148:                                return error;
1.336     ozaki-r  3149:                        }
1.151     yamt     3150:                }
1.399     ozaki-r  3151:                KERNEL_LOCK_UNLESS_NET_MPSAFE();
1.292     christos 3152:                mutex_enter(&if_clone_mtx);
                   3153:                r = (cmd == SIOCIFCREATE) ?
1.64      thorpej  3154:                        if_clone_create(ifr->ifr_name) :
1.185     dyoung   3155:                        if_clone_destroy(ifr->ifr_name);
1.292     christos 3156:                mutex_exit(&if_clone_mtx);
1.399     ozaki-r  3157:                KERNEL_UNLOCK_UNLESS_NET_MPSAFE();
1.339     ozaki-r  3158:                curlwp_bindx(bound);
1.292     christos 3159:                return r;
1.67      thorpej  3160:
                   3161:        case SIOCIFGCLONERS:
1.315     martin   3162:                {
                   3163:                        struct if_clonereq *req = (struct if_clonereq *)data;
                   3164:                        return if_clone_list(req->ifcr_count, req->ifcr_buffer,
                   3165:                            &req->ifcr_total);
                   3166:                }
1.63      thorpej  3167:        }
                   3168:
1.339     ozaki-r  3169:        bound = curlwp_bind();
1.336     ozaki-r  3170:        ifp = if_get(ifr->ifr_name, &psref);
                   3171:        if (ifp == NULL) {
1.339     ozaki-r  3172:                curlwp_bindx(bound);
1.185     dyoung   3173:                return ENXIO;
1.336     ozaki-r  3174:        }
1.151     yamt     3175:
                   3176:        switch (cmd) {
1.233     christos 3177:        case SIOCALIFADDR:
                   3178:        case SIOCDLIFADDR:
                   3179:        case SIOCSIFADDRPREF:
1.151     yamt     3180:        case SIOCSIFFLAGS:
                   3181:        case SIOCSIFCAP:
                   3182:        case SIOCSIFMETRIC:
                   3183:        case SIOCZIFDATA:
                   3184:        case SIOCSIFMTU:
                   3185:        case SIOCSIFPHYADDR:
                   3186:        case SIOCDIFPHYADDR:
                   3187: #ifdef INET6
                   3188:        case SIOCSIFPHYADDR_IN6:
                   3189: #endif
                   3190:        case SIOCSLIFPHYADDR:
                   3191:        case SIOCADDMULTI:
                   3192:        case SIOCDELMULTI:
                   3193:        case SIOCSIFMEDIA:
1.154     perry    3194:        case SIOCSDRVSPEC:
1.196     skd      3195:        case SIOCG80211:
                   3196:        case SIOCS80211:
1.151     yamt     3197:        case SIOCS80211NWID:
                   3198:        case SIOCS80211NWKEY:
                   3199:        case SIOCS80211POWER:
                   3200:        case SIOCS80211BSSID:
                   3201:        case SIOCS80211CHANNEL:
1.249     pooka    3202:        case SIOCSLINKSTR:
1.185     dyoung   3203:                if (l != NULL) {
1.174     elad     3204:                        error = kauth_authorize_network(l->l_cred,
                   3205:                            KAUTH_NETWORK_INTERFACE,
                   3206:                            KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp,
                   3207:                            (void *)cmd, NULL);
1.185     dyoung   3208:                        if (error != 0)
1.336     ozaki-r  3209:                                goto out;
1.151     yamt     3210:                }
                   3211:        }
                   3212:
1.49      itojun   3213:        oif_flags = ifp->if_flags;
1.1       cgd      3214:
1.399     ozaki-r  3215:        KERNEL_LOCK_UNLESS_IFP_MPSAFE(ifp);
1.413     ozaki-r  3216:        IFNET_LOCK(ifp);
1.336     ozaki-r  3217:
1.231     dyoung   3218:        error = (*ifp->if_ioctl)(ifp, cmd, data);
                   3219:        if (error != ENOTTY)
                   3220:                ;
                   3221:        else if (so->so_proto == NULL)
1.252     dyoung   3222:                error = EOPNOTSUPP;
1.231     dyoung   3223:        else {
1.399     ozaki-r  3224:                KERNEL_LOCK_IF_IFP_MPSAFE(ifp);
1.161     christos 3225: #ifdef COMPAT_OSOCK
1.361     pgoyette 3226:                if (vec_compat_ifioctl != NULL)
                   3227:                        error = (*vec_compat_ifioctl)(so, ocmd, cmd, data, l);
                   3228:                else
1.1       cgd      3229: #endif
1.361     pgoyette 3230:                        error = (*so->so_proto->pr_usrreqs->pr_ioctl)(so,
                   3231:                            cmd, data, ifp);
1.399     ozaki-r  3232:                KERNEL_UNLOCK_IF_IFP_MPSAFE(ifp);
1.49      itojun   3233:        }
1.1       cgd      3234:
1.49      itojun   3235:        if (((oif_flags ^ ifp->if_flags) & IFF_UP) != 0) {
1.312     roy      3236:                if ((ifp->if_flags & IFF_UP) != 0) {
1.373     ozaki-r  3237:                        int s = splsoftnet();
1.403     ozaki-r  3238:                        if_up_locked(ifp);
1.49      itojun   3239:                        splx(s);
                   3240:                }
1.1       cgd      3241:        }
1.186     christos 3242: #ifdef COMPAT_OIFREQ
                   3243:        if (cmd != ocmd)
1.246     christos 3244:                ifreqn2o(oifr, ifr);
1.186     christos 3245: #endif
1.49      itojun   3246:
1.413     ozaki-r  3247:        IFNET_UNLOCK(ifp);
1.399     ozaki-r  3248:        KERNEL_UNLOCK_UNLESS_IFP_MPSAFE(ifp);
1.336     ozaki-r  3249: out:
                   3250:        if_put(ifp, &psref);
1.339     ozaki-r  3251:        curlwp_bindx(bound);
1.185     dyoung   3252:        return error;
1.1       cgd      3253: }
                   3254:
                   3255: /*
                   3256:  * Return interface configuration
                   3257:  * of system.  List may be used
                   3258:  * in later ioctl's (above) to get
                   3259:  * other information.
1.200     gdt      3260:  *
                   3261:  * Each record is a struct ifreq.  Before the addition of
                   3262:  * sockaddr_storage, the API rule was that sockaddr flavors that did
                   3263:  * not fit would extend beyond the struct ifreq, with the next struct
                   3264:  * ifreq starting sa_len beyond the struct sockaddr.  Because the
                   3265:  * union in struct ifreq includes struct sockaddr_storage, every kind
                   3266:  * of sockaddr must fit.  Thus, there are no longer any overlength
                   3267:  * records.
                   3268:  *
                   3269:  * Records are added to the user buffer if they fit, and ifc_len is
                   3270:  * adjusted to the length that was written.  Thus, the user is only
                   3271:  * assured of getting the complete list if ifc_len on return is at
                   3272:  * least sizeof(struct ifreq) less than it was on entry.
                   3273:  *
                   3274:  * If the user buffer pointer is NULL, this routine copies no data and
                   3275:  * returns the amount of space that would be needed.
                   3276:  *
                   3277:  * Invariants:
                   3278:  * ifrp points to the next part of the user's buffer to be used.  If
                   3279:  * ifrp != NULL, space holds the number of bytes remaining that we may
                   3280:  * write at ifrp.  Otherwise, space holds the number of bytes that
                   3281:  * would have been written had there been adequate space.
1.1       cgd      3282:  */
                   3283: /*ARGSUSED*/
1.302     ozaki-r  3284: static int
1.183     christos 3285: ifconf(u_long cmd, void *data)
1.1       cgd      3286: {
1.61      augustss 3287:        struct ifconf *ifc = (struct ifconf *)data;
                   3288:        struct ifnet *ifp;
                   3289:        struct ifaddr *ifa;
1.304     ozaki-r  3290:        struct ifreq ifr, *ifrp = NULL;
                   3291:        int space = 0, error = 0;
1.200     gdt      3292:        const int sz = (int)sizeof(struct ifreq);
1.304     ozaki-r  3293:        const bool docopy = ifc->ifc_req != NULL;
1.334     ozaki-r  3294:        int s;
1.339     ozaki-r  3295:        int bound;
1.334     ozaki-r  3296:        struct psref psref;
1.1       cgd      3297:
1.304     ozaki-r  3298:        if (docopy) {
1.190     enami    3299:                space = ifc->ifc_len;
1.304     ozaki-r  3300:                ifrp = ifc->ifc_req;
                   3301:        }
                   3302:
1.339     ozaki-r  3303:        bound = curlwp_bind();
1.334     ozaki-r  3304:        s = pserialize_read_enter();
                   3305:        IFNET_READER_FOREACH(ifp) {
                   3306:                psref_acquire(&psref, &ifp->if_psref, ifnet_psref_class);
                   3307:                pserialize_read_exit(s);
                   3308:
1.175     christos 3309:                (void)strncpy(ifr.ifr_name, ifp->if_xname,
1.173     christos 3310:                    sizeof(ifr.ifr_name));
1.334     ozaki-r  3311:                if (ifr.ifr_name[sizeof(ifr.ifr_name) - 1] != '\0') {
                   3312:                        error = ENAMETOOLONG;
                   3313:                        goto release_exit;
                   3314:                }
1.354     ozaki-r  3315:                if (IFADDR_READER_EMPTY(ifp)) {
1.200     gdt      3316:                        /* Interface with no addresses - send zero sockaddr. */
1.127     christos 3317:                        memset(&ifr.ifr_addr, 0, sizeof(ifr.ifr_addr));
1.304     ozaki-r  3318:                        if (!docopy) {
1.218     dyoung   3319:                                space += sz;
1.400     ozaki-r  3320:                                goto next;
1.218     dyoung   3321:                        }
                   3322:                        if (space >= sz) {
                   3323:                                error = copyout(&ifr, ifrp, sz);
                   3324:                                if (error != 0)
1.334     ozaki-r  3325:                                        goto release_exit;
1.218     dyoung   3326:                                ifrp++;
                   3327:                                space -= sz;
1.70      mellon   3328:                        }
1.127     christos 3329:                }
                   3330:
1.400     ozaki-r  3331:                s = pserialize_read_enter();
1.354     ozaki-r  3332:                IFADDR_READER_FOREACH(ifa, ifp) {
1.61      augustss 3333:                        struct sockaddr *sa = ifa->ifa_addr;
1.200     gdt      3334:                        /* all sockaddrs must fit in sockaddr_storage */
                   3335:                        KASSERT(sa->sa_len <= sizeof(ifr.ifr_ifru));
                   3336:
1.304     ozaki-r  3337:                        if (!docopy) {
1.218     dyoung   3338:                                space += sz;
                   3339:                                continue;
                   3340:                        }
                   3341:                        memcpy(&ifr.ifr_space, sa, sa->sa_len);
1.400     ozaki-r  3342:                        pserialize_read_exit(s);
                   3343:
1.218     dyoung   3344:                        if (space >= sz) {
                   3345:                                error = copyout(&ifr, ifrp, sz);
                   3346:                                if (error != 0)
1.334     ozaki-r  3347:                                        goto release_exit;
1.218     dyoung   3348:                                ifrp++; space -= sz;
1.1       cgd      3349:                        }
1.400     ozaki-r  3350:                        s = pserialize_read_enter();
1.1       cgd      3351:                }
1.400     ozaki-r  3352:                pserialize_read_exit(s);
1.334     ozaki-r  3353:
1.400     ozaki-r  3354:         next:
1.334     ozaki-r  3355:                s = pserialize_read_enter();
                   3356:                psref_release(&psref, &ifp->if_psref, ifnet_psref_class);
1.1       cgd      3357:        }
1.334     ozaki-r  3358:        pserialize_read_exit(s);
1.339     ozaki-r  3359:        curlwp_bindx(bound);
1.334     ozaki-r  3360:
1.304     ozaki-r  3361:        if (docopy) {
1.200     gdt      3362:                KASSERT(0 <= space && space <= ifc->ifc_len);
1.127     christos 3363:                ifc->ifc_len -= space;
1.218     dyoung   3364:        } else {
1.200     gdt      3365:                KASSERT(space >= 0);
                   3366:                ifc->ifc_len = space;
                   3367:        }
1.190     enami    3368:        return (0);
1.334     ozaki-r  3369:
                   3370: release_exit:
                   3371:        psref_release(&psref, &ifp->if_psref, ifnet_psref_class);
1.339     ozaki-r  3372:        curlwp_bindx(bound);
1.334     ozaki-r  3373:        return error;
1.1       cgd      3374: }
1.133     jonathan 3375:
1.198     dyoung   3376: int
1.247     christos 3377: ifreq_setaddr(u_long cmd, struct ifreq *ifr, const struct sockaddr *sa)
1.198     dyoung   3378: {
1.420     christos 3379:        uint8_t len = sizeof(ifr->ifr_ifru.ifru_space);
1.247     christos 3380: #ifdef COMPAT_OIFREQ
                   3381:        struct ifreq ifrb;
                   3382:        struct oifreq *oifr = NULL;
                   3383:        u_long ocmd = cmd;
1.420     christos 3384:
                   3385:        if (vec_compat_cvtcmd) {
                   3386:                    cmd = (*vec_compat_cvtcmd)(cmd);
                   3387:                    if (cmd != ocmd) {
                   3388:                            oifr = (struct oifreq *)(void *)ifr;
                   3389:                            ifr = &ifrb;
                   3390:                            ifreqo2n(oifr, ifr);
                   3391:                            len = sizeof(oifr->ifr_addr);
                   3392:                    }
                   3393:        }
1.247     christos 3394: #endif
1.198     dyoung   3395:        if (len < sa->sa_len)
                   3396:                return EFBIG;
1.247     christos 3397:
1.241     joerg    3398:        memset(&ifr->ifr_addr, 0, len);
1.202     dyoung   3399:        sockaddr_copy(&ifr->ifr_addr, len, sa);
1.247     christos 3400:
                   3401: #ifdef COMPAT_OIFREQ
                   3402:        if (cmd != ocmd)
                   3403:                ifreqn2o(oifr, ifr);
                   3404: #endif
1.198     dyoung   3405:        return 0;
                   3406: }
                   3407:
1.155     christos 3408: /*
1.332     knakahar 3409:  * wrapper function for the drivers which doesn't have if_transmit().
1.155     christos 3410:  */
1.345     knakahar 3411: static int
1.332     knakahar 3412: if_transmit(struct ifnet *ifp, struct mbuf *m)
1.155     christos 3413: {
1.332     knakahar 3414:        int s, error;
1.389     ozaki-r  3415:        size_t pktlen = m->m_pkthdr.len;
                   3416:        bool mcast = (m->m_flags & M_MCAST) != 0;
1.332     knakahar 3417:
                   3418:        s = splnet();
1.155     christos 3419:
1.330     knakahar 3420:        IFQ_ENQUEUE(&ifp->if_snd, m, error);
1.332     knakahar 3421:        if (error != 0) {
                   3422:                /* mbuf is already freed */
1.185     dyoung   3423:                goto out;
1.332     knakahar 3424:        }
                   3425:
1.389     ozaki-r  3426:        ifp->if_obytes += pktlen;
                   3427:        if (mcast)
1.155     christos 3428:                ifp->if_omcasts++;
1.332     knakahar 3429:
1.155     christos 3430:        if ((ifp->if_flags & IFF_OACTIVE) == 0)
1.343     knakahar 3431:                if_start_lock(ifp);
1.185     dyoung   3432: out:
1.155     christos 3433:        splx(s);
1.332     knakahar 3434:
1.155     christos 3435:        return error;
                   3436: }
                   3437:
1.345     knakahar 3438: int
                   3439: if_transmit_lock(struct ifnet *ifp, struct mbuf *m)
                   3440: {
                   3441:        int error;
                   3442:
                   3443: #ifdef ALTQ
                   3444:        KERNEL_LOCK(1, NULL);
                   3445:        if (ALTQ_IS_ENABLED(&ifp->if_snd)) {
                   3446:                error = if_transmit(ifp, m);
                   3447:                KERNEL_UNLOCK_ONE(NULL);
                   3448:        } else {
                   3449:                KERNEL_UNLOCK_ONE(NULL);
                   3450:                error = (*ifp->if_transmit)(ifp, m);
1.383     knakahar 3451:                /* mbuf is alredy freed */
1.345     knakahar 3452:        }
                   3453: #else /* !ALTQ */
                   3454:        error = (*ifp->if_transmit)(ifp, m);
1.383     knakahar 3455:        /* mbuf is alredy freed */
1.345     knakahar 3456: #endif /* !ALTQ */
                   3457:
                   3458:        return error;
                   3459: }
                   3460:
1.155     christos 3461: /*
1.332     knakahar 3462:  * Queue message on interface, and start output if interface
                   3463:  * not yet active.
                   3464:  */
                   3465: int
                   3466: ifq_enqueue(struct ifnet *ifp, struct mbuf *m)
                   3467: {
                   3468:
1.345     knakahar 3469:        return if_transmit_lock(ifp, m);
1.332     knakahar 3470: }
                   3471:
                   3472: /*
1.155     christos 3473:  * Queue message on interface, possibly using a second fast queue
                   3474:  */
                   3475: int
1.330     knakahar 3476: ifq_enqueue2(struct ifnet *ifp, struct ifqueue *ifq, struct mbuf *m)
1.155     christos 3477: {
                   3478:        int error = 0;
                   3479:
                   3480:        if (ifq != NULL
                   3481: #ifdef ALTQ
                   3482:            && ALTQ_IS_ENABLED(&ifp->if_snd) == 0
                   3483: #endif
                   3484:            ) {
                   3485:                if (IF_QFULL(ifq)) {
                   3486:                        IF_DROP(&ifp->if_snd);
                   3487:                        m_freem(m);
                   3488:                        if (error == 0)
                   3489:                                error = ENOBUFS;
1.185     dyoung   3490:                } else
1.155     christos 3491:                        IF_ENQUEUE(ifq, m);
                   3492:        } else
1.330     knakahar 3493:                IFQ_ENQUEUE(&ifp->if_snd, m, error);
1.155     christos 3494:        if (error != 0) {
                   3495:                ++ifp->if_oerrors;
                   3496:                return error;
                   3497:        }
                   3498:        return 0;
                   3499: }
                   3500:
1.252     dyoung   3501: int
                   3502: if_addr_init(ifnet_t *ifp, struct ifaddr *ifa, const bool src)
                   3503: {
                   3504:        int rc;
                   3505:
1.413     ozaki-r  3506:        KASSERT(IFNET_LOCKED(ifp));
1.252     dyoung   3507:        if (ifp->if_initaddr != NULL)
                   3508:                rc = (*ifp->if_initaddr)(ifp, ifa, src);
                   3509:        else if (src ||
                   3510:                 (rc = (*ifp->if_ioctl)(ifp, SIOCSIFDSTADDR, ifa)) == ENOTTY)
                   3511:                rc = (*ifp->if_ioctl)(ifp, SIOCINITIFADDR, ifa);
                   3512:
                   3513:        return rc;
                   3514: }
                   3515:
                   3516: int
1.309     roy      3517: if_do_dad(struct ifnet *ifp)
                   3518: {
                   3519:        if ((ifp->if_flags & IFF_LOOPBACK) != 0)
                   3520:                return 0;
                   3521:
                   3522:        switch (ifp->if_type) {
                   3523:        case IFT_FAITH:
                   3524:                /*
                   3525:                 * These interfaces do not have the IFF_LOOPBACK flag,
                   3526:                 * but loop packets back.  We do not have to do DAD on such
                   3527:                 * interfaces.  We should even omit it, because loop-backed
                   3528:                 * responses would confuse the DAD procedure.
                   3529:                 */
                   3530:                return 0;
                   3531:        default:
                   3532:                /*
                   3533:                 * Our DAD routine requires the interface up and running.
                   3534:                 * However, some interfaces can be up before the RUNNING
                   3535:                 * status.  Additionaly, users may try to assign addresses
                   3536:                 * before the interface becomes up (or running).
                   3537:                 * We simply skip DAD in such a case as a work around.
                   3538:                 * XXX: we should rather mark "tentative" on such addresses,
                   3539:                 * and do DAD after the interface becomes ready.
                   3540:                 */
                   3541:                if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) !=
                   3542:                    (IFF_UP|IFF_RUNNING))
                   3543:                        return 0;
                   3544:
                   3545:                return 1;
                   3546:        }
                   3547: }
                   3548:
                   3549: int
1.252     dyoung   3550: if_flags_set(ifnet_t *ifp, const short flags)
                   3551: {
                   3552:        int rc;
                   3553:
1.413     ozaki-r  3554:        KASSERT(IFNET_LOCKED(ifp));
1.401     ozaki-r  3555:
1.252     dyoung   3556:        if (ifp->if_setflags != NULL)
                   3557:                rc = (*ifp->if_setflags)(ifp, flags);
                   3558:        else {
1.259     dyoung   3559:                short cantflags, chgdflags;
1.256     dyoung   3560:                struct ifreq ifr;
                   3561:
1.259     dyoung   3562:                chgdflags = ifp->if_flags ^ flags;
                   3563:                cantflags = chgdflags & IFF_CANTCHANGE;
1.256     dyoung   3564:
                   3565:                if (cantflags != 0)
                   3566:                        ifp->if_flags ^= cantflags;
                   3567:
1.259     dyoung   3568:                 /* Traditionally, we do not call if_ioctl after
                   3569:                  * setting/clearing only IFF_PROMISC if the interface
                   3570:                  * isn't IFF_UP.  Uphold that tradition.
                   3571:                 */
1.391     ozaki-r  3572:                if (chgdflags == IFF_PROMISC && (ifp->if_flags & IFF_UP) == 0)
                   3573:                        return 0;
1.259     dyoung   3574:
                   3575:                memset(&ifr, 0, sizeof(ifr));
                   3576:
1.256     dyoung   3577:                ifr.ifr_flags = flags & ~IFF_CANTCHANGE;
1.252     dyoung   3578:                rc = (*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, &ifr);
1.256     dyoung   3579:
                   3580:                if (rc != 0 && cantflags != 0)
                   3581:                        ifp->if_flags ^= cantflags;
1.252     dyoung   3582:        }
                   3583:
                   3584:        return rc;
                   3585: }
                   3586:
                   3587: int
                   3588: if_mcast_op(ifnet_t *ifp, const unsigned long cmd, const struct sockaddr *sa)
                   3589: {
                   3590:        int rc;
                   3591:        struct ifreq ifr;
                   3592:
1.417     ozaki-r  3593:        /* CARP and MROUTING still don't deal with the lock yet */
                   3594: #if (!defined(NCARP) || (NCARP == 0)) && !defined(MROUTING)
1.415     ozaki-r  3595:        KASSERT(IFNET_LOCKED(ifp));
                   3596: #endif
1.252     dyoung   3597:        if (ifp->if_mcastop != NULL)
                   3598:                rc = (*ifp->if_mcastop)(ifp, cmd, sa);
                   3599:        else {
                   3600:                ifreq_setaddr(cmd, &ifr, sa);
                   3601:                rc = (*ifp->if_ioctl)(ifp, cmd, &ifr);
                   3602:        }
                   3603:
                   3604:        return rc;
                   3605: }
1.155     christos 3606:
1.234     dyoung   3607: static void
                   3608: sysctl_sndq_setup(struct sysctllog **clog, const char *ifname,
                   3609:     struct ifaltq *ifq)
                   3610: {
                   3611:        const struct sysctlnode *cnode, *rnode;
                   3612:
                   3613:        if (sysctl_createv(clog, 0, NULL, &rnode,
                   3614:                       CTLFLAG_PERMANENT,
                   3615:                       CTLTYPE_NODE, "interfaces",
                   3616:                       SYSCTL_DESCR("Per-interface controls"),
                   3617:                       NULL, 0, NULL, 0,
1.272     pooka    3618:                       CTL_NET, CTL_CREATE, CTL_EOL) != 0)
1.234     dyoung   3619:                goto bad;
                   3620:
                   3621:        if (sysctl_createv(clog, 0, &rnode, &rnode,
                   3622:                       CTLFLAG_PERMANENT,
                   3623:                       CTLTYPE_NODE, ifname,
                   3624:                       SYSCTL_DESCR("Interface controls"),
                   3625:                       NULL, 0, NULL, 0,
                   3626:                       CTL_CREATE, CTL_EOL) != 0)
                   3627:                goto bad;
                   3628:
                   3629:        if (sysctl_createv(clog, 0, &rnode, &rnode,
                   3630:                       CTLFLAG_PERMANENT,
                   3631:                       CTLTYPE_NODE, "sndq",
                   3632:                       SYSCTL_DESCR("Interface output queue controls"),
                   3633:                       NULL, 0, NULL, 0,
                   3634:                       CTL_CREATE, CTL_EOL) != 0)
                   3635:                goto bad;
                   3636:
                   3637:        if (sysctl_createv(clog, 0, &rnode, &cnode,
                   3638:                       CTLFLAG_PERMANENT,
                   3639:                       CTLTYPE_INT, "len",
                   3640:                       SYSCTL_DESCR("Current output queue length"),
                   3641:                       NULL, 0, &ifq->ifq_len, 0,
                   3642:                       CTL_CREATE, CTL_EOL) != 0)
                   3643:                goto bad;
                   3644:
                   3645:        if (sysctl_createv(clog, 0, &rnode, &cnode,
                   3646:                       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
                   3647:                       CTLTYPE_INT, "maxlen",
                   3648:                       SYSCTL_DESCR("Maximum allowed output queue length"),
                   3649:                       NULL, 0, &ifq->ifq_maxlen, 0,
                   3650:                       CTL_CREATE, CTL_EOL) != 0)
                   3651:                goto bad;
                   3652:
                   3653:        if (sysctl_createv(clog, 0, &rnode, &cnode,
                   3654:                       CTLFLAG_PERMANENT,
                   3655:                       CTLTYPE_INT, "drops",
                   3656:                       SYSCTL_DESCR("Packets dropped due to full output queue"),
                   3657:                       NULL, 0, &ifq->ifq_drops, 0,
                   3658:                       CTL_CREATE, CTL_EOL) != 0)
                   3659:                goto bad;
                   3660:
                   3661:        return;
                   3662: bad:
                   3663:        printf("%s: could not attach sysctl nodes\n", ifname);
                   3664:        return;
                   3665: }
                   3666:
1.138     drochner 3667: #if defined(INET) || defined(INET6)
1.276     rmind    3668:
                   3669: #define        SYSCTL_NET_PKTQ(q, cn, c)                                       \
                   3670:        static int                                                      \
                   3671:        sysctl_net_##q##_##cn(SYSCTLFN_ARGS)                            \
                   3672:        {                                                               \
                   3673:                return sysctl_pktq_count(SYSCTLFN_CALL(rnode), q, c);   \
                   3674:        }
                   3675:
                   3676: #if defined(INET)
1.279     rmind    3677: static int
                   3678: sysctl_net_ip_pktq_maxlen(SYSCTLFN_ARGS)
                   3679: {
                   3680:        return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip_pktq);
                   3681: }
1.276     rmind    3682: SYSCTL_NET_PKTQ(ip_pktq, items, PKTQ_NITEMS)
                   3683: SYSCTL_NET_PKTQ(ip_pktq, drops, PKTQ_DROPS)
                   3684: #endif
1.279     rmind    3685:
1.276     rmind    3686: #if defined(INET6)
1.279     rmind    3687: static int
                   3688: sysctl_net_ip6_pktq_maxlen(SYSCTLFN_ARGS)
                   3689: {
                   3690:        return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), ip6_pktq);
                   3691: }
1.276     rmind    3692: SYSCTL_NET_PKTQ(ip6_pktq, items, PKTQ_NITEMS)
                   3693: SYSCTL_NET_PKTQ(ip6_pktq, drops, PKTQ_DROPS)
                   3694: #endif
                   3695:
1.136     atatat   3696: static void
1.276     rmind    3697: sysctl_net_pktq_setup(struct sysctllog **clog, int pf)
1.136     atatat   3698: {
1.276     rmind    3699:        sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL;
                   3700:        const char *pfname = NULL, *ipname = NULL;
                   3701:        int ipn = 0, qid = 0;
                   3702:
                   3703:        switch (pf) {
                   3704: #if defined(INET)
                   3705:        case PF_INET:
                   3706:                len_func = sysctl_net_ip_pktq_items;
                   3707:                maxlen_func = sysctl_net_ip_pktq_maxlen;
                   3708:                drops_func = sysctl_net_ip_pktq_drops;
                   3709:                pfname = "inet", ipn = IPPROTO_IP;
                   3710:                ipname = "ip", qid = IPCTL_IFQ;
                   3711:                break;
                   3712: #endif
                   3713: #if defined(INET6)
                   3714:        case PF_INET6:
                   3715:                len_func = sysctl_net_ip6_pktq_items;
                   3716:                maxlen_func = sysctl_net_ip6_pktq_maxlen;
                   3717:                drops_func = sysctl_net_ip6_pktq_drops;
                   3718:                pfname = "inet6", ipn = IPPROTO_IPV6;
                   3719:                ipname = "ip6", qid = IPV6CTL_IFQ;
                   3720:                break;
                   3721: #endif
                   3722:        default:
                   3723:                KASSERT(false);
                   3724:        }
1.136     atatat   3725:
1.139     atatat   3726:        sysctl_createv(clog, 0, NULL, NULL,
                   3727:                       CTLFLAG_PERMANENT,
1.136     atatat   3728:                       CTLTYPE_NODE, pfname, NULL,
                   3729:                       NULL, 0, NULL, 0,
                   3730:                       CTL_NET, pf, CTL_EOL);
1.139     atatat   3731:        sysctl_createv(clog, 0, NULL, NULL,
                   3732:                       CTLFLAG_PERMANENT,
1.136     atatat   3733:                       CTLTYPE_NODE, ipname, NULL,
                   3734:                       NULL, 0, NULL, 0,
                   3735:                       CTL_NET, pf, ipn, CTL_EOL);
1.139     atatat   3736:        sysctl_createv(clog, 0, NULL, NULL,
                   3737:                       CTLFLAG_PERMANENT,
1.142     atatat   3738:                       CTLTYPE_NODE, "ifq",
                   3739:                       SYSCTL_DESCR("Protocol input queue controls"),
1.139     atatat   3740:                       NULL, 0, NULL, 0,
                   3741:                       CTL_NET, pf, ipn, qid, CTL_EOL);
1.136     atatat   3742:
1.139     atatat   3743:        sysctl_createv(clog, 0, NULL, NULL,
                   3744:                       CTLFLAG_PERMANENT,
1.142     atatat   3745:                       CTLTYPE_INT, "len",
                   3746:                       SYSCTL_DESCR("Current input queue length"),
1.276     rmind    3747:                       len_func, 0, NULL, 0,
1.136     atatat   3748:                       CTL_NET, pf, ipn, qid, IFQCTL_LEN, CTL_EOL);
1.139     atatat   3749:        sysctl_createv(clog, 0, NULL, NULL,
                   3750:                       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
1.142     atatat   3751:                       CTLTYPE_INT, "maxlen",
                   3752:                       SYSCTL_DESCR("Maximum allowed input queue length"),
1.276     rmind    3753:                       maxlen_func, 0, NULL, 0,
1.136     atatat   3754:                       CTL_NET, pf, ipn, qid, IFQCTL_MAXLEN, CTL_EOL);
1.139     atatat   3755:        sysctl_createv(clog, 0, NULL, NULL,
                   3756:                       CTLFLAG_PERMANENT,
1.142     atatat   3757:                       CTLTYPE_INT, "drops",
                   3758:                       SYSCTL_DESCR("Packets dropped due to full input queue"),
1.276     rmind    3759:                       drops_func, 0, NULL, 0,
1.136     atatat   3760:                       CTL_NET, pf, ipn, qid, IFQCTL_DROPS, CTL_EOL);
                   3761: }
1.138     drochner 3762: #endif /* INET || INET6 */
1.280     joerg    3763:
                   3764: static int
                   3765: if_sdl_sysctl(SYSCTLFN_ARGS)
                   3766: {
                   3767:        struct ifnet *ifp;
                   3768:        const struct sockaddr_dl *sdl;
1.340     ozaki-r  3769:        struct psref psref;
                   3770:        int error = 0;
                   3771:        int bound;
1.280     joerg    3772:
                   3773:        if (namelen != 1)
                   3774:                return EINVAL;
                   3775:
1.340     ozaki-r  3776:        bound = curlwp_bind();
                   3777:        ifp = if_get_byindex(name[0], &psref);
                   3778:        if (ifp == NULL) {
                   3779:                error = ENODEV;
1.341     riastrad 3780:                goto out0;
1.340     ozaki-r  3781:        }
1.280     joerg    3782:
                   3783:        sdl = ifp->if_sadl;
                   3784:        if (sdl == NULL) {
                   3785:                *oldlenp = 0;
1.341     riastrad 3786:                goto out1;
1.280     joerg    3787:        }
                   3788:
                   3789:        if (oldp == NULL) {
                   3790:                *oldlenp = sdl->sdl_alen;
1.341     riastrad 3791:                goto out1;
1.280     joerg    3792:        }
                   3793:
                   3794:        if (*oldlenp >= sdl->sdl_alen)
                   3795:                *oldlenp = sdl->sdl_alen;
1.340     ozaki-r  3796:        error = sysctl_copyout(l, &sdl->sdl_data[sdl->sdl_nlen], oldp, *oldlenp);
1.341     riastrad 3797: out1:
1.340     ozaki-r  3798:        if_put(ifp, &psref);
1.341     riastrad 3799: out0:
1.340     ozaki-r  3800:        curlwp_bindx(bound);
                   3801:        return error;
1.280     joerg    3802: }
                   3803:
1.370     ozaki-r  3804: static void
                   3805: if_sysctl_setup(struct sysctllog **clog)
1.280     joerg    3806: {
                   3807:        const struct sysctlnode *rnode = NULL;
                   3808:
                   3809:        sysctl_createv(clog, 0, NULL, &rnode,
                   3810:                       CTLFLAG_PERMANENT,
                   3811:                       CTLTYPE_NODE, "sdl",
                   3812:                       SYSCTL_DESCR("Get active link-layer address"),
                   3813:                       if_sdl_sysctl, 0, NULL, 0,
                   3814:                       CTL_NET, CTL_CREATE, CTL_EOL);
1.370     ozaki-r  3815:
                   3816: #if defined(INET)
                   3817:        sysctl_net_pktq_setup(NULL, PF_INET);
                   3818: #endif
                   3819: #ifdef INET6
                   3820:        if (in6_present)
                   3821:                sysctl_net_pktq_setup(NULL, PF_INET6);
                   3822: #endif
1.280     joerg    3823: }

CVSweb <webmaster@jp.NetBSD.org>