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

Annotation of src/sys/net/route.c, Revision 1.126.6.3

1.126.6.1  tls         1: /*     $NetBSD$        */
1.18      kml         2:
                      3: /*-
1.106     ad          4:  * Copyright (c) 1998, 2008 The NetBSD Foundation, Inc.
1.18      kml         5:  * All rights reserved.
                      6:  *
                      7:  * This code is derived from software contributed to The NetBSD Foundation
                      8:  * by Kevin M. Lahey of the Numerical Aerospace Simulation Facility,
                      9:  * NASA Ames Research Center.
                     10:  *
                     11:  * Redistribution and use in source and binary forms, with or without
                     12:  * modification, are permitted provided that the following conditions
                     13:  * are met:
                     14:  * 1. Redistributions of source code must retain the above copyright
                     15:  *    notice, this list of conditions and the following disclaimer.
                     16:  * 2. Redistributions in binary form must reproduce the above copyright
                     17:  *    notice, this list of conditions and the following disclaimer in the
                     18:  *    documentation and/or other materials provided with the distribution.
                     19:  *
                     20:  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
                     21:  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
                     22:  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
                     23:  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
                     24:  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
                     25:  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
                     26:  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
                     27:  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
                     28:  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
                     29:  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
                     30:  * POSSIBILITY OF SUCH DAMAGE.
                     31:  */
1.11      cgd        32:
1.1       cgd        33: /*
1.25      itojun     34:  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
                     35:  * All rights reserved.
1.65      perry      36:  *
1.25      itojun     37:  * Redistribution and use in source and binary forms, with or without
                     38:  * modification, are permitted provided that the following conditions
                     39:  * are met:
                     40:  * 1. Redistributions of source code must retain the above copyright
                     41:  *    notice, this list of conditions and the following disclaimer.
                     42:  * 2. Redistributions in binary form must reproduce the above copyright
                     43:  *    notice, this list of conditions and the following disclaimer in the
                     44:  *    documentation and/or other materials provided with the distribution.
                     45:  * 3. Neither the name of the project nor the names of its contributors
                     46:  *    may be used to endorse or promote products derived from this software
                     47:  *    without specific prior written permission.
1.65      perry      48:  *
1.25      itojun     49:  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
                     50:  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
                     51:  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
                     52:  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
                     53:  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
                     54:  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
                     55:  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
                     56:  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
                     57:  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
                     58:  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
                     59:  * SUCH DAMAGE.
                     60:  */
                     61:
                     62: /*
1.10      mycroft    63:  * Copyright (c) 1980, 1986, 1991, 1993
                     64:  *     The Regents of the University of California.  All rights reserved.
1.1       cgd        65:  *
                     66:  * Redistribution and use in source and binary forms, with or without
                     67:  * modification, are permitted provided that the following conditions
                     68:  * are met:
                     69:  * 1. Redistributions of source code must retain the above copyright
                     70:  *    notice, this list of conditions and the following disclaimer.
                     71:  * 2. Redistributions in binary form must reproduce the above copyright
                     72:  *    notice, this list of conditions and the following disclaimer in the
                     73:  *    documentation and/or other materials provided with the distribution.
1.58      agc        74:  * 3. Neither the name of the University nor the names of its contributors
1.1       cgd        75:  *    may be used to endorse or promote products derived from this software
                     76:  *    without specific prior written permission.
                     77:  *
                     78:  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
                     79:  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
                     80:  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
                     81:  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
                     82:  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
                     83:  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
                     84:  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
                     85:  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
                     86:  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
                     87:  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
                     88:  * SUCH DAMAGE.
                     89:  *
1.17      christos   90:  *     @(#)route.c     8.3 (Berkeley) 1/9/95
1.1       cgd        91:  */
1.50      lukem      92:
1.126.6.3! jdolecek   93: #ifdef _KERNEL_OPT
        !            94: #include "opt_inet.h"
1.90      dyoung     95: #include "opt_route.h"
1.126.6.3! jdolecek   96: #include "opt_net_mpsafe.h"
        !            97: #endif
1.90      dyoung     98:
1.50      lukem      99: #include <sys/cdefs.h>
1.126.6.1  tls       100: __KERNEL_RCSID(0, "$NetBSD$");
1.2       cgd       101:
1.5       mycroft   102: #include <sys/param.h>
1.126.6.3! jdolecek  103: #ifdef RTFLUSH_DEBUG
1.90      dyoung    104: #include <sys/sysctl.h>
1.126.6.3! jdolecek  105: #endif
1.5       mycroft   106: #include <sys/systm.h>
1.35      thorpej   107: #include <sys/callout.h>
1.5       mycroft   108: #include <sys/proc.h>
                    109: #include <sys/mbuf.h>
                    110: #include <sys/socket.h>
                    111: #include <sys/socketvar.h>
                    112: #include <sys/domain.h>
1.18      kml       113: #include <sys/kernel.h>
1.5       mycroft   114: #include <sys/ioctl.h>
1.22      thorpej   115: #include <sys/pool.h>
1.119     elad      116: #include <sys/kauth.h>
1.126.6.3! jdolecek  117: #include <sys/workqueue.h>
        !           118: #include <sys/syslog.h>
        !           119: #include <sys/rwlock.h>
        !           120: #include <sys/mutex.h>
        !           121: #include <sys/cpu.h>
1.1       cgd       122:
1.5       mycroft   123: #include <net/if.h>
1.114     dyoung    124: #include <net/if_dl.h>
1.5       mycroft   125: #include <net/route.h>
1.126.6.3! jdolecek  126: #if defined(INET) || defined(INET6)
        !           127: #include <net/if_llatbl.h>
        !           128: #endif
1.1       cgd       129:
1.5       mycroft   130: #include <netinet/in.h>
                    131: #include <netinet/in_var.h>
1.1       cgd       132:
1.90      dyoung    133: #ifdef RTFLUSH_DEBUG
                    134: #define        rtcache_debug() __predict_false(_rtcache_debug)
                    135: #else /* RTFLUSH_DEBUG */
                    136: #define        rtcache_debug() 0
                    137: #endif /* RTFLUSH_DEBUG */
1.5       mycroft   138:
1.126.6.3! jdolecek  139: #ifdef RT_DEBUG
        !           140: #define RT_REFCNT_TRACE(rt)    printf("%s:%d: rt=%p refcnt=%d\n", \
        !           141:                                    __func__, __LINE__, (rt), (rt)->rt_refcnt)
        !           142: #else
        !           143: #define RT_REFCNT_TRACE(rt)    do {} while (0)
        !           144: #endif
        !           145:
        !           146: #ifdef DEBUG
        !           147: #define dlog(level, fmt, args...)      log(level, fmt, ##args)
        !           148: #else
        !           149: #define dlog(level, fmt, args...)      do {} while (0)
        !           150: #endif
        !           151:
        !           152: struct rtstat          rtstat;
        !           153:
        !           154: static int             rttrash;        /* routes not in table but not freed */
        !           155:
        !           156: static struct pool     rtentry_pool;
        !           157: static struct pool     rttimer_pool;
        !           158:
        !           159: static struct callout  rt_timer_ch; /* callout for rt_timer_timer() */
        !           160: static struct workqueue        *rt_timer_wq;
        !           161: static struct work     rt_timer_wk;
        !           162:
        !           163: static void    rt_timer_init(void);
        !           164: static void    rt_timer_queue_remove_all(struct rttimer_queue *);
        !           165: static void    rt_timer_remove_all(struct rtentry *);
        !           166: static void    rt_timer_timer(void *);
        !           167:
        !           168: /*
        !           169:  * Locking notes:
        !           170:  * - The routing table is protected by a global rwlock
        !           171:  *   - API: RT_RLOCK and friends
        !           172:  * - rtcaches are NOT protected by the framework
        !           173:  *   - Callers must guarantee a rtcache isn't accessed simultaneously
        !           174:  *   - How the constraint is guranteed in the wild
        !           175:  *     - Protect a rtcache by a mutex (e.g., inp_route)
        !           176:  *     - Make rtcache per-CPU and allow only accesses from softint
        !           177:  *       (e.g., ipforward_rt_percpu)
        !           178:  * - References to a rtentry is managed by reference counting and psref
        !           179:  *   - Reference couting is used for temporal reference when a rtentry
        !           180:  *     is fetched from the routing table
        !           181:  *   - psref is used for temporal reference when a rtentry is fetched
        !           182:  *     from a rtcache
        !           183:  *     - struct route (rtcache) has struct psref, so we cannot obtain
        !           184:  *       a reference twice on the same struct route
        !           185:  *   - Befere destroying or updating a rtentry, we have to wait for
        !           186:  *     all references left (see below for details)
        !           187:  *   - APIs
        !           188:  *     - An obtained rtentry via rtalloc1 or rtrequest* must be
        !           189:  *       unreferenced by rt_unref
        !           190:  *     - An obtained rtentry via rtcache_* must be unreferenced by
        !           191:  *       rtcache_unref
        !           192:  *   - TODO: once we get a lockless routing table, we should use only
        !           193:  *           psref for rtentries
        !           194:  * - rtentry destruction
        !           195:  *   - A rtentry is destroyed (freed) only when we call rtrequest(RTM_DELETE)
        !           196:  *   - If a caller of rtrequest grabs a reference of a rtentry, the caller
        !           197:  *     has a responsibility to destroy the rtentry by itself by calling
        !           198:  *     rt_free
        !           199:  *     - If not, rtrequest itself does that
        !           200:  *   - If rt_free is called in softint, the actual destruction routine is
        !           201:  *     deferred to a workqueue
        !           202:  * - rtentry update
        !           203:  *   - When updating a rtentry, RTF_UPDATING flag is set
        !           204:  *   - If a rtentry is set RTF_UPDATING, fetching the rtentry from
        !           205:  *     the routing table or a rtcache results in either of the following
        !           206:  *     cases:
        !           207:  *     - if the caller runs in softint, the caller fails to fetch
        !           208:  *     - otherwise, the caller waits for the update completed and retries
        !           209:  *       to fetch (probably succeed to fetch for the second time)
        !           210:  * - rtcache invalidation
        !           211:  *   - There is a global generation counter that is incremented when
        !           212:  *     any routes have been added or deleted
        !           213:  *   - When a rtcache caches a rtentry into itself, it also stores
        !           214:  *     a snapshot of the generation counter
        !           215:  *   - If the snapshot equals to the global counter, the cache is valid,
        !           216:  *     otherwise the cache is invalidated
        !           217:  */
        !           218:
        !           219: /*
        !           220:  * Global lock for the routing table.
        !           221:  */
        !           222: static krwlock_t               rt_lock __cacheline_aligned;
        !           223: #ifdef NET_MPSAFE
        !           224: #define RT_RLOCK()             rw_enter(&rt_lock, RW_READER)
        !           225: #define RT_WLOCK()             rw_enter(&rt_lock, RW_WRITER)
        !           226: #define RT_UNLOCK()            rw_exit(&rt_lock)
        !           227: #define RT_LOCKED()            rw_lock_held(&rt_lock)
        !           228: #define        RT_ASSERT_WLOCK()       KASSERT(rw_write_held(&rt_lock))
        !           229: #else
        !           230: #define RT_RLOCK()             do {} while (0)
        !           231: #define RT_WLOCK()             do {} while (0)
        !           232: #define RT_UNLOCK()            do {} while (0)
        !           233: #define RT_LOCKED()            false
        !           234: #define        RT_ASSERT_WLOCK()       do {} while (0)
        !           235: #endif
        !           236:
        !           237: static uint64_t rtcache_generation;
1.1       cgd       238:
1.126.6.3! jdolecek  239: /*
        !           240:  * mutex and cv that are used to wait for references to a rtentry left
        !           241:  * before updating the rtentry.
        !           242:  */
        !           243: static struct {
        !           244:        kmutex_t                lock;
        !           245:        kcondvar_t              cv;
        !           246:        bool                    ongoing;
        !           247:        const struct lwp        *lwp;
        !           248: } rt_update_global __cacheline_aligned;
1.1       cgd       249:
1.126.6.3! jdolecek  250: /*
        !           251:  * A workqueue and stuff that are used to defer the destruction routine
        !           252:  * of rtentries.
        !           253:  */
        !           254: static struct {
        !           255:        struct workqueue        *wq;
        !           256:        struct work             wk;
        !           257:        kmutex_t                lock;
        !           258:        struct rtentry          *queue[10];
        !           259: } rt_free_global __cacheline_aligned;
1.22      thorpej   260:
1.126.6.3! jdolecek  261: /* psref for rtentry */
        !           262: static struct psref_class *rt_psref_class __read_mostly;
1.35      thorpej   263:
1.90      dyoung    264: #ifdef RTFLUSH_DEBUG
                    265: static int _rtcache_debug = 0;
                    266: #endif /* RTFLUSH_DEBUG */
                    267:
1.119     elad      268: static kauth_listener_t route_listener;
                    269:
1.60      matt      270: static int rtdeletemsg(struct rtentry *);
1.126.6.3! jdolecek  271:
        !           272: static void rt_maskedcopy(const struct sockaddr *,
        !           273:     struct sockaddr *, const struct sockaddr *);
        !           274:
        !           275: static void rtcache_invalidate(void);
        !           276:
        !           277: static void rt_ref(struct rtentry *);
        !           278:
        !           279: static struct rtentry *
        !           280:     rtalloc1_locked(const struct sockaddr *, int, bool, bool);
        !           281:
        !           282: static void rtcache_ref(struct rtentry *, struct route *);
        !           283:
        !           284: #ifdef NET_MPSAFE
        !           285: static void rt_update_wait(void);
        !           286: #endif
        !           287:
        !           288: static bool rt_wait_ok(void);
        !           289: static void rt_wait_refcnt(const char *, struct rtentry *, int);
        !           290: static void rt_wait_psref(struct rtentry *);
        !           291:
        !           292: #ifdef DDB
        !           293: static void db_print_sa(const struct sockaddr *);
        !           294: static void db_print_ifa(struct ifaddr *);
        !           295: static int db_show_rtentry(struct rtentry *, void *);
        !           296: #endif
1.40      itojun    297:
1.90      dyoung    298: #ifdef RTFLUSH_DEBUG
1.118     pooka     299: static void sysctl_net_rtcache_setup(struct sysctllog **);
                    300: static void
                    301: sysctl_net_rtcache_setup(struct sysctllog **clog)
1.90      dyoung    302: {
                    303:        const struct sysctlnode *rnode;
                    304:
                    305:        if (sysctl_createv(clog, 0, NULL, &rnode, CTLFLAG_PERMANENT,
                    306:            CTLTYPE_NODE,
                    307:            "rtcache", SYSCTL_DESCR("Route cache related settings"),
1.126.6.2  tls       308:            NULL, 0, NULL, 0, CTL_NET, CTL_CREATE, CTL_EOL) != 0)
1.90      dyoung    309:                return;
                    310:        if (sysctl_createv(clog, 0, &rnode, &rnode,
                    311:            CTLFLAG_PERMANENT|CTLFLAG_READWRITE, CTLTYPE_INT,
                    312:            "debug", SYSCTL_DESCR("Debug route caches"),
                    313:            NULL, 0, &_rtcache_debug, 0, CTL_CREATE, CTL_EOL) != 0)
                    314:                return;
                    315: }
                    316: #endif /* RTFLUSH_DEBUG */
                    317:
1.126.6.3! jdolecek  318: static inline void
        !           319: rt_destroy(struct rtentry *rt)
        !           320: {
        !           321:        if (rt->_rt_key != NULL)
        !           322:                sockaddr_free(rt->_rt_key);
        !           323:        if (rt->rt_gateway != NULL)
        !           324:                sockaddr_free(rt->rt_gateway);
        !           325:        if (rt_gettag(rt) != NULL)
        !           326:                sockaddr_free(rt_gettag(rt));
        !           327:        rt->_rt_key = rt->rt_gateway = rt->rt_tag = NULL;
        !           328: }
        !           329:
        !           330: static inline const struct sockaddr *
        !           331: rt_setkey(struct rtentry *rt, const struct sockaddr *key, int flags)
        !           332: {
        !           333:        if (rt->_rt_key == key)
        !           334:                goto out;
        !           335:
        !           336:        if (rt->_rt_key != NULL)
        !           337:                sockaddr_free(rt->_rt_key);
        !           338:        rt->_rt_key = sockaddr_dup(key, flags);
        !           339: out:
        !           340:        rt->rt_nodes->rn_key = (const char *)rt->_rt_key;
        !           341:        return rt->_rt_key;
        !           342: }
        !           343:
1.81      joerg     344: struct ifaddr *
                    345: rt_get_ifa(struct rtentry *rt)
                    346: {
                    347:        struct ifaddr *ifa;
                    348:
                    349:        if ((ifa = rt->rt_ifa) == NULL)
                    350:                return ifa;
                    351:        else if (ifa->ifa_getifa == NULL)
                    352:                return ifa;
                    353: #if 0
                    354:        else if (ifa->ifa_seqno != NULL && *ifa->ifa_seqno == rt->rt_ifa_seqno)
                    355:                return ifa;
                    356: #endif
                    357:        else {
1.94      dyoung    358:                ifa = (*ifa->ifa_getifa)(ifa, rt_getkey(rt));
1.126.6.3! jdolecek  359:                if (ifa == NULL)
        !           360:                        return NULL;
1.81      joerg     361:                rt_replace_ifa(rt, ifa);
                    362:                return ifa;
                    363:        }
                    364: }
                    365:
1.80      joerg     366: static void
                    367: rt_set_ifa1(struct rtentry *rt, struct ifaddr *ifa)
                    368: {
                    369:        rt->rt_ifa = ifa;
                    370:        if (ifa->ifa_seqno != NULL)
                    371:                rt->rt_ifa_seqno = *ifa->ifa_seqno;
                    372: }
                    373:
1.116     roy       374: /*
                    375:  * Is this route the connected route for the ifa?
                    376:  */
                    377: static int
                    378: rt_ifa_connected(const struct rtentry *rt, const struct ifaddr *ifa)
                    379: {
                    380:        const struct sockaddr *key, *dst, *odst;
                    381:        struct sockaddr_storage maskeddst;
                    382:
                    383:        key = rt_getkey(rt);
                    384:        dst = rt->rt_flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr;
                    385:        if (dst == NULL ||
                    386:            dst->sa_family != key->sa_family ||
                    387:            dst->sa_len != key->sa_len)
                    388:                return 0;
                    389:        if ((rt->rt_flags & RTF_HOST) == 0 && ifa->ifa_netmask) {
                    390:                odst = dst;
                    391:                dst = (struct sockaddr *)&maskeddst;
                    392:                rt_maskedcopy(odst, (struct sockaddr *)&maskeddst,
                    393:                    ifa->ifa_netmask);
                    394:        }
                    395:        return (memcmp(dst, key, dst->sa_len) == 0);
                    396: }
                    397:
1.80      joerg     398: void
                    399: rt_replace_ifa(struct rtentry *rt, struct ifaddr *ifa)
                    400: {
1.116     roy       401:        if (rt->rt_ifa &&
                    402:            rt->rt_ifa != ifa &&
                    403:            rt->rt_ifa->ifa_flags & IFA_ROUTE &&
                    404:            rt_ifa_connected(rt, rt->rt_ifa))
                    405:        {
                    406:                RT_DPRINTF("rt->_rt_key = %p, ifa = %p, "
                    407:                    "replace deleted IFA_ROUTE\n",
                    408:                    (void *)rt->_rt_key, (void *)rt->rt_ifa);
                    409:                rt->rt_ifa->ifa_flags &= ~IFA_ROUTE;
                    410:                if (rt_ifa_connected(rt, ifa)) {
                    411:                        RT_DPRINTF("rt->_rt_key = %p, ifa = %p, "
                    412:                            "replace added IFA_ROUTE\n",
                    413:                            (void *)rt->_rt_key, (void *)ifa);
                    414:                        ifa->ifa_flags |= IFA_ROUTE;
                    415:                }
                    416:        }
                    417:
1.126.6.3! jdolecek  418:        ifaref(ifa);
        !           419:        ifafree(rt->rt_ifa);
1.80      joerg     420:        rt_set_ifa1(rt, ifa);
                    421: }
                    422:
                    423: static void
                    424: rt_set_ifa(struct rtentry *rt, struct ifaddr *ifa)
                    425: {
1.126.6.3! jdolecek  426:        ifaref(ifa);
1.80      joerg     427:        rt_set_ifa1(rt, ifa);
                    428: }
                    429:
1.119     elad      430: static int
                    431: route_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
                    432:     void *arg0, void *arg1, void *arg2, void *arg3)
                    433: {
                    434:        struct rt_msghdr *rtm;
                    435:        int result;
                    436:
                    437:        result = KAUTH_RESULT_DEFER;
                    438:        rtm = arg1;
                    439:
1.120     elad      440:        if (action != KAUTH_NETWORK_ROUTE)
                    441:                return result;
                    442:
1.119     elad      443:        if (rtm->rtm_type == RTM_GET)
                    444:                result = KAUTH_RESULT_ALLOW;
                    445:
                    446:        return result;
                    447: }
                    448:
1.126.6.3! jdolecek  449: static void rt_free_work(struct work *, void *);
        !           450:
1.9       mycroft   451: void
1.124     matt      452: rt_init(void)
1.1       cgd       453: {
1.126.6.3! jdolecek  454:        int error;
1.22      thorpej   455:
1.118     pooka     456: #ifdef RTFLUSH_DEBUG
                    457:        sysctl_net_rtcache_setup(NULL);
                    458: #endif
                    459:
1.126.6.3! jdolecek  460:        mutex_init(&rt_free_global.lock, MUTEX_DEFAULT, IPL_SOFTNET);
        !           461:        rt_psref_class = psref_class_create("rtentry", IPL_SOFTNET);
        !           462:
        !           463:        error = workqueue_create(&rt_free_global.wq, "rt_free",
        !           464:            rt_free_work, NULL, PRI_SOFTNET, IPL_SOFTNET, WQ_MPSAFE);
        !           465:        if (error)
        !           466:                panic("%s: workqueue_create failed (%d)\n", __func__, error);
        !           467:
        !           468:        mutex_init(&rt_update_global.lock, MUTEX_DEFAULT, IPL_SOFTNET);
        !           469:        cv_init(&rt_update_global.cv, "rt_update");
        !           470:
1.113     pooka     471:        pool_init(&rtentry_pool, sizeof(struct rtentry), 0, 0, 0, "rtentpl",
                    472:            NULL, IPL_SOFTNET);
                    473:        pool_init(&rttimer_pool, sizeof(struct rttimer), 0, 0, 0, "rttmrpl",
                    474:            NULL, IPL_SOFTNET);
                    475:
1.10      mycroft   476:        rn_init();      /* initialize all zeroes, all ones, mask table */
1.125     dyoung    477:        rtbl_init();
1.119     elad      478:
                    479:        route_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
                    480:            route_listener_cb, NULL);
1.1       cgd       481: }
                    482:
1.126.6.3! jdolecek  483: static void
        !           484: rtcache_invalidate(void)
1.82      dyoung    485: {
1.126.6.3! jdolecek  486:
        !           487:        RT_ASSERT_WLOCK();
1.90      dyoung    488:
                    489:        if (rtcache_debug())
                    490:                printf("%s: enter\n", __func__);
                    491:
1.126.6.3! jdolecek  492:        rtcache_generation++;
1.82      dyoung    493: }
                    494:
1.126.6.3! jdolecek  495: #ifdef RT_DEBUG
1.126.6.2  tls       496: static void
1.126.6.3! jdolecek  497: dump_rt(const struct rtentry *rt)
1.82      dyoung    498: {
1.126.6.3! jdolecek  499:        char buf[512];
1.82      dyoung    500:
1.126.6.3! jdolecek  501:        aprint_normal("rt: ");
        !           502:        aprint_normal("p=%p ", rt);
        !           503:        if (rt->_rt_key == NULL) {
        !           504:                aprint_normal("dst=(NULL) ");
        !           505:        } else {
        !           506:                sockaddr_format(rt->_rt_key, buf, sizeof(buf));
        !           507:                aprint_normal("dst=%s ", buf);
        !           508:        }
        !           509:        if (rt->rt_gateway == NULL) {
        !           510:                aprint_normal("gw=(NULL) ");
        !           511:        } else {
        !           512:                sockaddr_format(rt->_rt_key, buf, sizeof(buf));
        !           513:                aprint_normal("gw=%s ", buf);
        !           514:        }
        !           515:        aprint_normal("flags=%x ", rt->rt_flags);
        !           516:        if (rt->rt_ifp == NULL) {
        !           517:                aprint_normal("if=(NULL) ");
        !           518:        } else {
        !           519:                aprint_normal("if=%s ", rt->rt_ifp->if_xname);
        !           520:        }
        !           521:        aprint_normal("\n");
1.82      dyoung    522: }
1.126.6.3! jdolecek  523: #endif /* RT_DEBUG */
1.82      dyoung    524:
1.1       cgd       525: /*
1.126.6.3! jdolecek  526:  * Packet routing routines. If success, refcnt of a returned rtentry
        !           527:  * will be incremented. The caller has to rtfree it by itself.
1.1       cgd       528:  */
                    529: struct rtentry *
1.126.6.3! jdolecek  530: rtalloc1_locked(const struct sockaddr *dst, int report, bool wait_ok,
        !           531:     bool wlock)
1.1       cgd       532: {
1.126.6.3! jdolecek  533:        rtbl_t *rtbl;
1.36      augustss  534:        struct rtentry *rt;
1.126.6.3! jdolecek  535:        int s;
1.1       cgd       536:
1.126.6.3! jdolecek  537: #ifdef NET_MPSAFE
        !           538: retry:
        !           539: #endif
        !           540:        s = splsoftnet();
        !           541:        rtbl = rt_gettable(dst->sa_family);
        !           542:        if (rtbl == NULL)
        !           543:                goto miss;
        !           544:
        !           545:        rt = rt_matchaddr(rtbl, dst);
        !           546:        if (rt == NULL)
        !           547:                goto miss;
        !           548:
        !           549:        if (!ISSET(rt->rt_flags, RTF_UP))
        !           550:                goto miss;
        !           551:
        !           552: #ifdef NET_MPSAFE
        !           553:        if (ISSET(rt->rt_flags, RTF_UPDATING) &&
        !           554:            /* XXX updater should be always able to acquire */
        !           555:            curlwp != rt_update_global.lwp) {
        !           556:                if (!wait_ok || !rt_wait_ok())
        !           557:                        goto miss;
        !           558:                RT_UNLOCK();
        !           559:                splx(s);
        !           560:
        !           561:                /* We can wait until the update is complete */
        !           562:                rt_update_wait();
        !           563:
        !           564:                if (wlock)
        !           565:                        RT_WLOCK();
        !           566:                else
        !           567:                        RT_RLOCK();
        !           568:                goto retry;
        !           569:        }
        !           570: #endif /* NET_MPSAFE */
        !           571:
        !           572:        rt_ref(rt);
        !           573:        RT_REFCNT_TRACE(rt);
        !           574:
        !           575:        splx(s);
        !           576:        return rt;
        !           577: miss:
        !           578:        rtstat.rts_unreach++;
        !           579:        if (report) {
        !           580:                struct rt_addrinfo info;
        !           581:
        !           582:                memset(&info, 0, sizeof(info));
        !           583:                info.rti_info[RTAX_DST] = dst;
        !           584:                rt_missmsg(RTM_MISS, &info, 0, 0);
1.1       cgd       585:        }
                    586:        splx(s);
1.126.6.3! jdolecek  587:        return NULL;
        !           588: }
        !           589:
        !           590: struct rtentry *
        !           591: rtalloc1(const struct sockaddr *dst, int report)
        !           592: {
        !           593:        struct rtentry *rt;
        !           594:
        !           595:        RT_RLOCK();
        !           596:        rt = rtalloc1_locked(dst, report, true, false);
        !           597:        RT_UNLOCK();
        !           598:
        !           599:        return rt;
        !           600: }
        !           601:
        !           602: static void
        !           603: rt_ref(struct rtentry *rt)
        !           604: {
        !           605:
        !           606:        KASSERT(rt->rt_refcnt >= 0);
        !           607:        atomic_inc_uint(&rt->rt_refcnt);
1.1       cgd       608: }
                    609:
1.9       mycroft   610: void
1.126.6.3! jdolecek  611: rt_unref(struct rtentry *rt)
1.1       cgd       612: {
1.10      mycroft   613:
1.126.6.2  tls       614:        KASSERT(rt != NULL);
1.126.6.3! jdolecek  615:        KASSERTMSG(rt->rt_refcnt > 0, "refcnt=%d", rt->rt_refcnt);
        !           616:
        !           617:        atomic_dec_uint(&rt->rt_refcnt);
        !           618:        if (!ISSET(rt->rt_flags, RTF_UP) || ISSET(rt->rt_flags, RTF_UPDATING)) {
        !           619:                mutex_enter(&rt_free_global.lock);
        !           620:                cv_broadcast(&rt->rt_cv);
        !           621:                mutex_exit(&rt_free_global.lock);
        !           622:        }
        !           623: }
        !           624:
        !           625: static bool
        !           626: rt_wait_ok(void)
        !           627: {
        !           628:
        !           629:        KASSERT(!cpu_intr_p());
        !           630:        return !cpu_softintr_p();
        !           631: }
1.126.6.2  tls       632:
1.126.6.3! jdolecek  633: void
        !           634: rt_wait_refcnt(const char *title, struct rtentry *rt, int cnt)
        !           635: {
        !           636:        mutex_enter(&rt_free_global.lock);
        !           637:        while (rt->rt_refcnt > cnt) {
        !           638:                dlog(LOG_DEBUG, "%s: %s waiting (refcnt=%d)\n",
        !           639:                    __func__, title, rt->rt_refcnt);
        !           640:                cv_wait(&rt->rt_cv, &rt_free_global.lock);
        !           641:                dlog(LOG_DEBUG, "%s: %s waited (refcnt=%d)\n",
        !           642:                    __func__, title, rt->rt_refcnt);
1.1       cgd       643:        }
1.126.6.3! jdolecek  644:        mutex_exit(&rt_free_global.lock);
1.1       cgd       645: }
                    646:
1.10      mycroft   647: void
1.126.6.3! jdolecek  648: rt_wait_psref(struct rtentry *rt)
1.10      mycroft   649: {
1.30      thorpej   650:
1.126.6.3! jdolecek  651:        psref_target_destroy(&rt->rt_psref, rt_psref_class);
        !           652:        psref_target_init(&rt->rt_psref, rt_psref_class);
        !           653: }
        !           654:
        !           655: static void
        !           656: _rt_free(struct rtentry *rt)
        !           657: {
        !           658:        struct ifaddr *ifa;
        !           659:
        !           660:        /*
        !           661:         * Need to avoid a deadlock on rt_wait_refcnt of update
        !           662:         * and a conflict on psref_target_destroy of update.
        !           663:         */
        !           664: #ifdef NET_MPSAFE
        !           665:        rt_update_wait();
        !           666: #endif
        !           667:
        !           668:        RT_REFCNT_TRACE(rt);
        !           669:        KASSERTMSG(rt->rt_refcnt >= 0, "refcnt=%d", rt->rt_refcnt);
        !           670:        rt_wait_refcnt("free", rt, 0);
        !           671: #ifdef NET_MPSAFE
        !           672:        psref_target_destroy(&rt->rt_psref, rt_psref_class);
1.30      thorpej   673: #endif
1.126.6.3! jdolecek  674:
        !           675:        rt_assert_inactive(rt);
        !           676:        rttrash--;
        !           677:        ifa = rt->rt_ifa;
        !           678:        rt->rt_ifa = NULL;
        !           679:        ifafree(ifa);
        !           680:        rt->rt_ifp = NULL;
        !           681:        cv_destroy(&rt->rt_cv);
        !           682:        rt_destroy(rt);
        !           683:        pool_put(&rtentry_pool, rt);
        !           684: }
        !           685:
        !           686: static void
        !           687: rt_free_work(struct work *wk, void *arg)
        !           688: {
        !           689:        int i;
        !           690:        struct rtentry *rt;
        !           691:
        !           692: restart:
        !           693:        mutex_enter(&rt_free_global.lock);
        !           694:        for (i = 0; i < sizeof(rt_free_global.queue); i++) {
        !           695:                if (rt_free_global.queue[i] == NULL)
        !           696:                        continue;
        !           697:                rt = rt_free_global.queue[i];
        !           698:                rt_free_global.queue[i] = NULL;
        !           699:                mutex_exit(&rt_free_global.lock);
        !           700:
        !           701:                atomic_dec_uint(&rt->rt_refcnt);
        !           702:                _rt_free(rt);
        !           703:                goto restart;
        !           704:        }
        !           705:        mutex_exit(&rt_free_global.lock);
        !           706: }
        !           707:
        !           708: void
        !           709: rt_free(struct rtentry *rt)
        !           710: {
        !           711:
        !           712:        KASSERT(rt->rt_refcnt > 0);
        !           713:        if (!rt_wait_ok()) {
        !           714:                int i;
        !           715:                mutex_enter(&rt_free_global.lock);
        !           716:                for (i = 0; i < sizeof(rt_free_global.queue); i++) {
        !           717:                        if (rt_free_global.queue[i] == NULL) {
        !           718:                                rt_free_global.queue[i] = rt;
        !           719:                                break;
        !           720:                        }
        !           721:                }
        !           722:                KASSERT(i < sizeof(rt_free_global.queue));
        !           723:                rt_ref(rt);
        !           724:                mutex_exit(&rt_free_global.lock);
        !           725:                workqueue_enqueue(rt_free_global.wq, &rt_free_global.wk, NULL);
        !           726:        } else {
        !           727:                atomic_dec_uint(&rt->rt_refcnt);
        !           728:                _rt_free(rt);
        !           729:        }
        !           730: }
        !           731:
        !           732: #ifdef NET_MPSAFE
        !           733: static void
        !           734: rt_update_wait(void)
        !           735: {
        !           736:
        !           737:        mutex_enter(&rt_update_global.lock);
        !           738:        while (rt_update_global.ongoing) {
        !           739:                dlog(LOG_DEBUG, "%s: waiting lwp=%p\n", __func__, curlwp);
        !           740:                cv_wait(&rt_update_global.cv, &rt_update_global.lock);
        !           741:                dlog(LOG_DEBUG, "%s: waited lwp=%p\n", __func__, curlwp);
        !           742:        }
        !           743:        mutex_exit(&rt_update_global.lock);
        !           744: }
1.31      thorpej   745: #endif
1.126.6.3! jdolecek  746:
        !           747: int
        !           748: rt_update_prepare(struct rtentry *rt)
        !           749: {
        !           750:
        !           751:        dlog(LOG_DEBUG, "%s: updating rt=%p lwp=%p\n", __func__, rt, curlwp);
        !           752:
        !           753:        RT_WLOCK();
        !           754:        /* If the entry is being destroyed, don't proceed the update. */
        !           755:        if (!ISSET(rt->rt_flags, RTF_UP)) {
        !           756:                RT_UNLOCK();
        !           757:                return -1;
        !           758:        }
        !           759:        rt->rt_flags |= RTF_UPDATING;
        !           760:        RT_UNLOCK();
        !           761:
        !           762:        mutex_enter(&rt_update_global.lock);
        !           763:        while (rt_update_global.ongoing) {
        !           764:                dlog(LOG_DEBUG, "%s: waiting ongoing updating rt=%p lwp=%p\n",
        !           765:                    __func__, rt, curlwp);
        !           766:                cv_wait(&rt_update_global.cv, &rt_update_global.lock);
        !           767:                dlog(LOG_DEBUG, "%s: waited ongoing updating rt=%p lwp=%p\n",
        !           768:                    __func__, rt, curlwp);
        !           769:        }
        !           770:        rt_update_global.ongoing = true;
        !           771:        /* XXX need it to avoid rt_update_wait by updater itself. */
        !           772:        rt_update_global.lwp = curlwp;
        !           773:        mutex_exit(&rt_update_global.lock);
        !           774:
        !           775:        rt_wait_refcnt("update", rt, 1);
        !           776:        rt_wait_psref(rt);
        !           777:
        !           778:        return 0;
        !           779: }
        !           780:
        !           781: void
        !           782: rt_update_finish(struct rtentry *rt)
        !           783: {
        !           784:
        !           785:        RT_WLOCK();
        !           786:        rt->rt_flags &= ~RTF_UPDATING;
        !           787:        RT_UNLOCK();
        !           788:
        !           789:        mutex_enter(&rt_update_global.lock);
        !           790:        rt_update_global.ongoing = false;
        !           791:        rt_update_global.lwp = NULL;
        !           792:        cv_broadcast(&rt_update_global.cv);
        !           793:        mutex_exit(&rt_update_global.lock);
        !           794:
        !           795:        dlog(LOG_DEBUG, "%s: updated rt=%p lwp=%p\n", __func__, rt, curlwp);
1.10      mycroft   796: }
                    797:
1.1       cgd       798: /*
                    799:  * Force a routing table entry to the specified
                    800:  * destination to go through the given gateway.
                    801:  * Normally called as a result of a routing redirect
                    802:  * message from the network layer.
                    803:  *
1.13      mycroft   804:  * N.B.: must be called at splsoftnet
1.1       cgd       805:  */
1.14      christos  806: void
1.60      matt      807: rtredirect(const struct sockaddr *dst, const struct sockaddr *gateway,
                    808:        const struct sockaddr *netmask, int flags, const struct sockaddr *src,
                    809:        struct rtentry **rtp)
1.1       cgd       810: {
1.36      augustss  811:        struct rtentry *rt;
1.1       cgd       812:        int error = 0;
1.121     dyoung    813:        uint64_t *stat = NULL;
1.10      mycroft   814:        struct rt_addrinfo info;
                    815:        struct ifaddr *ifa;
1.126.6.3! jdolecek  816:        struct psref psref;
1.1       cgd       817:
                    818:        /* verify the gateway is directly reachable */
1.126.6.3! jdolecek  819:        if ((ifa = ifa_ifwithnet_psref(gateway, &psref)) == NULL) {
1.1       cgd       820:                error = ENETUNREACH;
1.8       cgd       821:                goto out;
1.1       cgd       822:        }
                    823:        rt = rtalloc1(dst, 0);
                    824:        /*
                    825:         * If the redirect isn't from our current router for this dst,
                    826:         * it's either old or wrong.  If it redirects us to ourselves,
                    827:         * we have a routing loop, perhaps as a result of an interface
                    828:         * going down recently.
                    829:         */
1.10      mycroft   830:        if (!(flags & RTF_DONE) && rt &&
1.115     yamt      831:             (sockaddr_cmp(src, rt->rt_gateway) != 0 || rt->rt_ifa != ifa))
1.1       cgd       832:                error = EINVAL;
1.126.6.3! jdolecek  833:        else {
        !           834:                int s = pserialize_read_enter();
        !           835:                struct ifaddr *_ifa;
        !           836:
        !           837:                _ifa = ifa_ifwithaddr(gateway);
        !           838:                if (_ifa != NULL)
        !           839:                        error = EHOSTUNREACH;
        !           840:                pserialize_read_exit(s);
        !           841:        }
1.1       cgd       842:        if (error)
                    843:                goto done;
                    844:        /*
                    845:         * Create a new entry if we just got back a wildcard entry
1.33      soren     846:         * or the lookup failed.  This is necessary for hosts
1.1       cgd       847:         * which use routing redirects generated by smart gateways
                    848:         * to dynamically build the routing tables.
                    849:         */
1.95      dyoung    850:        if (rt == NULL || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
1.1       cgd       851:                goto create;
                    852:        /*
                    853:         * Don't listen to the redirect if it's
1.65      perry     854:         * for a route to an interface.
1.1       cgd       855:         */
                    856:        if (rt->rt_flags & RTF_GATEWAY) {
                    857:                if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
                    858:                        /*
                    859:                         * Changing from route to net => route to host.
                    860:                         * Create new route, rather than smashing route to net.
                    861:                         */
                    862:                create:
1.95      dyoung    863:                        if (rt != NULL)
1.126.6.3! jdolecek  864:                                rt_unref(rt);
1.1       cgd       865:                        flags |=  RTF_GATEWAY | RTF_DYNAMIC;
1.122     kefren    866:                        memset(&info, 0, sizeof(info));
1.39      itojun    867:                        info.rti_info[RTAX_DST] = dst;
                    868:                        info.rti_info[RTAX_GATEWAY] = gateway;
                    869:                        info.rti_info[RTAX_NETMASK] = netmask;
                    870:                        info.rti_ifa = ifa;
                    871:                        info.rti_flags = flags;
                    872:                        rt = NULL;
                    873:                        error = rtrequest1(RTM_ADD, &info, &rt);
                    874:                        if (rt != NULL)
                    875:                                flags = rt->rt_flags;
1.1       cgd       876:                        stat = &rtstat.rts_dynamic;
                    877:                } else {
                    878:                        /*
                    879:                         * Smash the current notion of the gateway to
                    880:                         * this destination.  Should check about netmask!!!
                    881:                         */
1.126.6.3! jdolecek  882: #ifdef NET_MPSAFE
        !           883:                        KASSERT(!cpu_softintr_p());
        !           884:
        !           885:                        error = rt_update_prepare(rt);
        !           886:                        if (error == 0) {
        !           887: #endif
        !           888:                                error = rt_setgate(rt, gateway);
        !           889:                                if (error == 0) {
        !           890:                                        rt->rt_flags |= RTF_MODIFIED;
        !           891:                                        flags |= RTF_MODIFIED;
        !           892:                                }
        !           893: #ifdef NET_MPSAFE
        !           894:                                rt_update_finish(rt);
        !           895:                        } else {
        !           896:                                /*
        !           897:                                 * If error != 0, the rtentry is being
        !           898:                                 * destroyed, so doing nothing doesn't
        !           899:                                 * matter.
        !           900:                                 */
        !           901:                        }
        !           902: #endif
1.10      mycroft   903:                        stat = &rtstat.rts_newgateway;
1.1       cgd       904:                }
                    905:        } else
                    906:                error = EHOSTUNREACH;
                    907: done:
                    908:        if (rt) {
1.95      dyoung    909:                if (rtp != NULL && !error)
1.1       cgd       910:                        *rtp = rt;
                    911:                else
1.126.6.3! jdolecek  912:                        rt_unref(rt);
1.1       cgd       913:        }
1.8       cgd       914: out:
1.1       cgd       915:        if (error)
                    916:                rtstat.rts_badredirect++;
1.8       cgd       917:        else if (stat != NULL)
                    918:                (*stat)++;
1.95      dyoung    919:        memset(&info, 0, sizeof(info));
1.10      mycroft   920:        info.rti_info[RTAX_DST] = dst;
                    921:        info.rti_info[RTAX_GATEWAY] = gateway;
                    922:        info.rti_info[RTAX_NETMASK] = netmask;
                    923:        info.rti_info[RTAX_AUTHOR] = src;
                    924:        rt_missmsg(RTM_REDIRECT, &info, flags, error);
1.126.6.3! jdolecek  925:        ifa_release(ifa, &psref);
1.1       cgd       926: }
                    927:
                    928: /*
1.126.6.3! jdolecek  929:  * Delete a route and generate a message.
        !           930:  * It doesn't free a passed rt.
1.40      itojun    931:  */
                    932: static int
1.60      matt      933: rtdeletemsg(struct rtentry *rt)
1.40      itojun    934: {
                    935:        int error;
                    936:        struct rt_addrinfo info;
1.126.6.3! jdolecek  937:        struct rtentry *retrt;
1.40      itojun    938:
                    939:        /*
                    940:         * Request the new route so that the entry is not actually
                    941:         * deleted.  That will allow the information being reported to
                    942:         * be accurate (and consistent with route_output()).
                    943:         */
1.95      dyoung    944:        memset(&info, 0, sizeof(info));
1.94      dyoung    945:        info.rti_info[RTAX_DST] = rt_getkey(rt);
1.40      itojun    946:        info.rti_info[RTAX_NETMASK] = rt_mask(rt);
                    947:        info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
                    948:        info.rti_flags = rt->rt_flags;
1.126.6.3! jdolecek  949:        error = rtrequest1(RTM_DELETE, &info, &retrt);
1.40      itojun    950:
                    951:        rt_missmsg(RTM_DELETE, &info, info.rti_flags, error);
                    952:
1.95      dyoung    953:        return error;
1.40      itojun    954: }
                    955:
1.1       cgd       956: struct ifaddr *
1.126.6.3! jdolecek  957: ifa_ifwithroute_psref(int flags, const struct sockaddr *dst,
        !           958:        const struct sockaddr *gateway, struct psref *psref)
1.1       cgd       959: {
1.126.6.3! jdolecek  960:        struct ifaddr *ifa = NULL;
        !           961:
1.1       cgd       962:        if ((flags & RTF_GATEWAY) == 0) {
                    963:                /*
                    964:                 * If we are adding a route to an interface,
                    965:                 * and the interface is a pt to pt link
                    966:                 * we should search for the destination
                    967:                 * as our clue to the interface.  Otherwise
                    968:                 * we can use the local address.
                    969:                 */
1.126.6.1  tls       970:                if ((flags & RTF_HOST) && gateway->sa_family != AF_LINK)
1.126.6.3! jdolecek  971:                        ifa = ifa_ifwithdstaddr_psref(dst, psref);
1.68      christos  972:                if (ifa == NULL)
1.126.6.3! jdolecek  973:                        ifa = ifa_ifwithaddr_psref(gateway, psref);
1.1       cgd       974:        } else {
                    975:                /*
                    976:                 * If we are adding a route to a remote net
                    977:                 * or host, the gateway may still be on the
                    978:                 * other end of a pt to pt link.
                    979:                 */
1.126.6.3! jdolecek  980:                ifa = ifa_ifwithdstaddr_psref(gateway, psref);
1.1       cgd       981:        }
1.68      christos  982:        if (ifa == NULL)
1.126.6.3! jdolecek  983:                ifa = ifa_ifwithnet_psref(gateway, psref);
1.68      christos  984:        if (ifa == NULL) {
1.126.6.3! jdolecek  985:                int s;
        !           986:                struct rtentry *rt;
        !           987:
        !           988:                /* XXX we cannot call rtalloc1 if holding the rt lock */
        !           989:                if (RT_LOCKED())
        !           990:                        rt = rtalloc1_locked(gateway, 0, true, true);
        !           991:                else
        !           992:                        rt = rtalloc1(gateway, 0);
1.68      christos  993:                if (rt == NULL)
                    994:                        return NULL;
1.126.6.3! jdolecek  995:                if (rt->rt_flags & RTF_GATEWAY) {
        !           996:                        rt_unref(rt);
        !           997:                        return NULL;
        !           998:                }
        !           999:                /*
        !          1000:                 * Just in case. May not need to do this workaround.
        !          1001:                 * Revisit when working on rtentry MP-ification.
        !          1002:                 */
        !          1003:                s = pserialize_read_enter();
        !          1004:                IFADDR_READER_FOREACH(ifa, rt->rt_ifp) {
        !          1005:                        if (ifa == rt->rt_ifa)
        !          1006:                                break;
        !          1007:                }
        !          1008:                if (ifa != NULL)
        !          1009:                        ifa_acquire(ifa, psref);
        !          1010:                pserialize_read_exit(s);
        !          1011:                rt_unref(rt);
        !          1012:                if (ifa == NULL)
1.68      christos 1013:                        return NULL;
1.1       cgd      1014:        }
                   1015:        if (ifa->ifa_addr->sa_family != dst->sa_family) {
1.126.6.3! jdolecek 1016:                struct ifaddr *nifa;
        !          1017:                int s;
        !          1018:
        !          1019:                s = pserialize_read_enter();
        !          1020:                nifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
        !          1021:                if (nifa != NULL) {
        !          1022:                        ifa_release(ifa, psref);
        !          1023:                        ifa_acquire(nifa, psref);
        !          1024:                        ifa = nifa;
        !          1025:                }
        !          1026:                pserialize_read_exit(s);
1.1       cgd      1027:        }
1.95      dyoung   1028:        return ifa;
1.1       cgd      1029: }
                   1030:
1.126.6.3! jdolecek 1031: /*
        !          1032:  * If it suceeds and ret_nrt isn't NULL, refcnt of ret_nrt is incremented.
        !          1033:  * The caller has to rtfree it by itself.
        !          1034:  */
1.9       mycroft  1035: int
1.60      matt     1036: rtrequest(int req, const struct sockaddr *dst, const struct sockaddr *gateway,
                   1037:        const struct sockaddr *netmask, int flags, struct rtentry **ret_nrt)
1.1       cgd      1038: {
1.39      itojun   1039:        struct rt_addrinfo info;
                   1040:
1.44      thorpej  1041:        memset(&info, 0, sizeof(info));
1.39      itojun   1042:        info.rti_flags = flags;
                   1043:        info.rti_info[RTAX_DST] = dst;
                   1044:        info.rti_info[RTAX_GATEWAY] = gateway;
                   1045:        info.rti_info[RTAX_NETMASK] = netmask;
                   1046:        return rtrequest1(req, &info, ret_nrt);
                   1047: }
                   1048:
1.126.6.3! jdolecek 1049: /*
        !          1050:  * It's a utility function to add/remove a route to/from the routing table
        !          1051:  * and tell user processes the addition/removal on success.
        !          1052:  */
1.39      itojun   1053: int
1.126.6.3! jdolecek 1054: rtrequest_newmsg(const int req, const struct sockaddr *dst,
        !          1055:        const struct sockaddr *gateway, const struct sockaddr *netmask,
        !          1056:        const int flags)
        !          1057: {
        !          1058:        int error;
        !          1059:        struct rtentry *ret_nrt = NULL;
        !          1060:
        !          1061:        KASSERT(req == RTM_ADD || req == RTM_DELETE);
        !          1062:
        !          1063:        error = rtrequest(req, dst, gateway, netmask, flags, &ret_nrt);
        !          1064:        if (error != 0)
        !          1065:                return error;
        !          1066:
        !          1067:        KASSERT(ret_nrt != NULL);
        !          1068:
        !          1069:        rt_newmsg(req, ret_nrt); /* tell user process */
        !          1070:        if (req == RTM_DELETE)
        !          1071:                rt_free(ret_nrt);
        !          1072:        else
        !          1073:                rt_unref(ret_nrt);
        !          1074:
        !          1075:        return 0;
        !          1076: }
        !          1077:
        !          1078: struct ifnet *
        !          1079: rt_getifp(struct rt_addrinfo *info, struct psref *psref)
1.39      itojun   1080: {
1.68      christos 1081:        const struct sockaddr *ifpaddr = info->rti_info[RTAX_IFP];
1.39      itojun   1082:
1.126.6.3! jdolecek 1083:        if (info->rti_ifp != NULL)
        !          1084:                return NULL;
1.39      itojun   1085:        /*
                   1086:         * ifp may be specified by sockaddr_dl when protocol address
                   1087:         * is ambiguous
                   1088:         */
1.126.6.3! jdolecek 1089:        if (ifpaddr != NULL && ifpaddr->sa_family == AF_LINK) {
        !          1090:                struct ifaddr *ifa;
        !          1091:                int s = pserialize_read_enter();
        !          1092:
        !          1093:                ifa = ifa_ifwithnet(ifpaddr);
        !          1094:                if (ifa != NULL)
        !          1095:                        info->rti_ifp = if_get_byindex(ifa->ifa_ifp->if_index,
        !          1096:                            psref);
        !          1097:                pserialize_read_exit(s);
        !          1098:        }
        !          1099:
        !          1100:        return info->rti_ifp;
        !          1101: }
        !          1102:
        !          1103: struct ifaddr *
        !          1104: rt_getifa(struct rt_addrinfo *info, struct psref *psref)
        !          1105: {
        !          1106:        struct ifaddr *ifa = NULL;
        !          1107:        const struct sockaddr *dst = info->rti_info[RTAX_DST];
        !          1108:        const struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY];
        !          1109:        const struct sockaddr *ifaaddr = info->rti_info[RTAX_IFA];
        !          1110:        int flags = info->rti_flags;
        !          1111:        const struct sockaddr *sa;
        !          1112:
        !          1113:        if (info->rti_ifa == NULL && ifaaddr != NULL) {
        !          1114:                ifa = ifa_ifwithaddr_psref(ifaaddr, psref);
        !          1115:                if (ifa != NULL)
        !          1116:                        goto got;
        !          1117:        }
        !          1118:
        !          1119:        sa = ifaaddr != NULL ? ifaaddr :
        !          1120:            (gateway != NULL ? gateway : dst);
        !          1121:        if (sa != NULL && info->rti_ifp != NULL)
        !          1122:                ifa = ifaof_ifpforaddr_psref(sa, info->rti_ifp, psref);
        !          1123:        else if (dst != NULL && gateway != NULL)
        !          1124:                ifa = ifa_ifwithroute_psref(flags, dst, gateway, psref);
        !          1125:        else if (sa != NULL)
        !          1126:                ifa = ifa_ifwithroute_psref(flags, sa, sa, psref);
        !          1127:        if (ifa == NULL)
        !          1128:                return NULL;
        !          1129: got:
        !          1130:        if (ifa->ifa_getifa != NULL) {
        !          1131:                /* FIXME ifa_getifa is NOMPSAFE */
        !          1132:                ifa = (*ifa->ifa_getifa)(ifa, dst);
        !          1133:                if (ifa == NULL)
        !          1134:                        return NULL;
        !          1135:                ifa_acquire(ifa, psref);
        !          1136:        }
        !          1137:        info->rti_ifa = ifa;
1.74      dyoung   1138:        if (info->rti_ifp == NULL)
                   1139:                info->rti_ifp = ifa->ifa_ifp;
1.126.6.3! jdolecek 1140:        return ifa;
1.39      itojun   1141: }
                   1142:
1.126.6.3! jdolecek 1143: /*
        !          1144:  * If it suceeds and ret_nrt isn't NULL, refcnt of ret_nrt is incremented.
        !          1145:  * The caller has to rtfree it by itself.
        !          1146:  */
1.39      itojun   1147: int
1.60      matt     1148: rtrequest1(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt)
1.39      itojun   1149: {
1.126.6.3! jdolecek 1150:        int s = splsoftnet(), ss;
1.125     dyoung   1151:        int error = 0, rc;
1.126.6.3! jdolecek 1152:        struct rtentry *rt;
1.125     dyoung   1153:        rtbl_t *rtbl;
1.126.6.3! jdolecek 1154:        struct ifaddr *ifa = NULL;
1.94      dyoung   1155:        struct sockaddr_storage maskeddst;
1.68      christos 1156:        const struct sockaddr *dst = info->rti_info[RTAX_DST];
                   1157:        const struct sockaddr *gateway = info->rti_info[RTAX_GATEWAY];
                   1158:        const struct sockaddr *netmask = info->rti_info[RTAX_NETMASK];
                   1159:        int flags = info->rti_flags;
1.126.6.3! jdolecek 1160:        struct psref psref_ifp, psref_ifa;
        !          1161:        int bound = 0;
        !          1162:        struct ifnet *ifp = NULL;
        !          1163:        bool need_to_release_ifa = true;
        !          1164:        bool need_unlock = true;
1.1       cgd      1165: #define senderr(x) { error = x ; goto bad; }
                   1166:
1.126.6.3! jdolecek 1167:        RT_WLOCK();
        !          1168:
        !          1169:        bound = curlwp_bind();
1.125     dyoung   1170:        if ((rtbl = rt_gettable(dst->sa_family)) == NULL)
1.1       cgd      1171:                senderr(ESRCH);
                   1172:        if (flags & RTF_HOST)
1.68      christos 1173:                netmask = NULL;
1.1       cgd      1174:        switch (req) {
                   1175:        case RTM_DELETE:
1.63      christos 1176:                if (netmask) {
1.94      dyoung   1177:                        rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
                   1178:                            netmask);
                   1179:                        dst = (struct sockaddr *)&maskeddst;
1.63      christos 1180:                }
1.125     dyoung   1181:                if ((rt = rt_lookup(rtbl, dst, netmask)) == NULL)
1.41      itojun   1182:                        senderr(ESRCH);
1.125     dyoung   1183:                if ((rt = rt_deladdr(rtbl, dst, netmask)) == NULL)
1.1       cgd      1184:                        senderr(ESRCH);
1.28      erh      1185:                rt->rt_flags &= ~RTF_UP;
1.116     roy      1186:                if ((ifa = rt->rt_ifa)) {
                   1187:                        if (ifa->ifa_flags & IFA_ROUTE &&
                   1188:                            rt_ifa_connected(rt, ifa)) {
                   1189:                                RT_DPRINTF("rt->_rt_key = %p, ifa = %p, "
                   1190:                                    "deleted IFA_ROUTE\n",
                   1191:                                    (void *)rt->_rt_key, (void *)ifa);
                   1192:                                ifa->ifa_flags &= ~IFA_ROUTE;
                   1193:                        }
                   1194:                        if (ifa->ifa_rtrequest)
                   1195:                                ifa->ifa_rtrequest(RTM_DELETE, rt, info);
1.126.6.3! jdolecek 1196:                        ifa = NULL;
1.116     roy      1197:                }
1.1       cgd      1198:                rttrash++;
1.126.6.3! jdolecek 1199:                if (ret_nrt) {
1.10      mycroft  1200:                        *ret_nrt = rt;
1.126.6.3! jdolecek 1201:                        rt_ref(rt);
        !          1202:                        RT_REFCNT_TRACE(rt);
        !          1203:                }
        !          1204:                rtcache_invalidate();
        !          1205:                RT_UNLOCK();
        !          1206:                need_unlock = false;
        !          1207:                rt_timer_remove_all(rt);
        !          1208: #if defined(INET) || defined(INET6)
        !          1209:                if (netmask != NULL)
        !          1210:                        lltable_prefix_free(dst->sa_family, dst, netmask, 0);
        !          1211: #endif
        !          1212:                if (ret_nrt == NULL) {
        !          1213:                        /* Adjust the refcount */
        !          1214:                        rt_ref(rt);
        !          1215:                        RT_REFCNT_TRACE(rt);
        !          1216:                        rt_free(rt);
1.10      mycroft  1217:                }
1.1       cgd      1218:                break;
                   1219:
                   1220:        case RTM_ADD:
1.126.6.3! jdolecek 1221:                if (info->rti_ifa == NULL) {
        !          1222:                        ifp = rt_getifp(info, &psref_ifp);
        !          1223:                        ifa = rt_getifa(info, &psref_ifa);
        !          1224:                        if (ifa == NULL)
        !          1225:                                senderr(ENETUNREACH);
        !          1226:                } else {
        !          1227:                        /* Caller should have a reference of ifa */
        !          1228:                        ifa = info->rti_ifa;
        !          1229:                        need_to_release_ifa = false;
        !          1230:                }
1.22      thorpej  1231:                rt = pool_get(&rtentry_pool, PR_NOWAIT);
1.68      christos 1232:                if (rt == NULL)
1.1       cgd      1233:                        senderr(ENOBUFS);
1.109     dyoung   1234:                memset(rt, 0, sizeof(*rt));
1.10      mycroft  1235:                rt->rt_flags = RTF_UP | flags;
1.18      kml      1236:                LIST_INIT(&rt->rt_timer);
1.126.6.3! jdolecek 1237:
1.110     dyoung   1238:                RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
1.1       cgd      1239:                if (netmask) {
1.94      dyoung   1240:                        rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
                   1241:                            netmask);
1.96      dyoung   1242:                        rt_setkey(rt, (struct sockaddr *)&maskeddst, M_NOWAIT);
1.94      dyoung   1243:                } else {
1.96      dyoung   1244:                        rt_setkey(rt, dst, M_NOWAIT);
1.94      dyoung   1245:                }
1.126.6.3! jdolecek 1246:                RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
        !          1247:                if (rt_getkey(rt) == NULL ||
        !          1248:                    rt_setgate(rt, gateway) != 0) {
        !          1249:                        pool_put(&rtentry_pool, rt);
        !          1250:                        senderr(ENOBUFS);
        !          1251:                }
        !          1252:
1.74      dyoung   1253:                rt_set_ifa(rt, ifa);
1.126.6.3! jdolecek 1254:                if (info->rti_info[RTAX_TAG] != NULL) {
        !          1255:                        const struct sockaddr *tag;
        !          1256:                        tag = rt_settag(rt, info->rti_info[RTAX_TAG]);
        !          1257:                        if (tag == NULL)
        !          1258:                                senderr(ENOBUFS);
        !          1259:                }
1.110     dyoung   1260:                RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
1.126.6.3! jdolecek 1261:
        !          1262:                ss = pserialize_read_enter();
        !          1263:                if (info->rti_info[RTAX_IFP] != NULL) {
        !          1264:                        struct ifaddr *ifa2;
        !          1265:                        ifa2 = ifa_ifwithnet(info->rti_info[RTAX_IFP]);
        !          1266:                        if (ifa2 != NULL)
        !          1267:                                rt->rt_ifp = ifa2->ifa_ifp;
        !          1268:                        else
        !          1269:                                rt->rt_ifp = ifa->ifa_ifp;
        !          1270:                } else
1.122     kefren   1271:                        rt->rt_ifp = ifa->ifa_ifp;
1.126.6.3! jdolecek 1272:                pserialize_read_exit(ss);
        !          1273:                cv_init(&rt->rt_cv, "rtentry");
        !          1274:                psref_target_init(&rt->rt_psref, rt_psref_class);
        !          1275:
1.110     dyoung   1276:                RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
1.125     dyoung   1277:                rc = rt_addaddr(rtbl, rt, netmask);
1.110     dyoung   1278:                RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
1.125     dyoung   1279:                if (rc != 0) {
1.126.6.3! jdolecek 1280:                        ifafree(ifa); /* for rt_set_ifa above */
        !          1281:                        cv_destroy(&rt->rt_cv);
1.94      dyoung   1282:                        rt_destroy(rt);
1.40      itojun   1283:                        pool_put(&rtentry_pool, rt);
1.125     dyoung   1284:                        senderr(rc);
1.27      matt     1285:                }
1.110     dyoung   1286:                RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
1.1       cgd      1287:                if (ifa->ifa_rtrequest)
1.39      itojun   1288:                        ifa->ifa_rtrequest(req, rt, info);
1.126.6.3! jdolecek 1289:                if (need_to_release_ifa)
        !          1290:                        ifa_release(ifa, &psref_ifa);
        !          1291:                ifa = NULL;
        !          1292:                if_put(ifp, &psref_ifp);
        !          1293:                ifp = NULL;
1.110     dyoung   1294:                RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
1.1       cgd      1295:                if (ret_nrt) {
                   1296:                        *ret_nrt = rt;
1.126.6.3! jdolecek 1297:                        rt_ref(rt);
        !          1298:                        RT_REFCNT_TRACE(rt);
1.1       cgd      1299:                }
1.126.6.3! jdolecek 1300:                rtcache_invalidate();
        !          1301:                RT_UNLOCK();
        !          1302:                need_unlock = false;
1.1       cgd      1303:                break;
1.92      dyoung   1304:        case RTM_GET:
1.94      dyoung   1305:                if (netmask != NULL) {
                   1306:                        rt_maskedcopy(dst, (struct sockaddr *)&maskeddst,
                   1307:                            netmask);
                   1308:                        dst = (struct sockaddr *)&maskeddst;
                   1309:                }
1.125     dyoung   1310:                if ((rt = rt_lookup(rtbl, dst, netmask)) == NULL)
1.92      dyoung   1311:                        senderr(ESRCH);
                   1312:                if (ret_nrt != NULL) {
                   1313:                        *ret_nrt = rt;
1.126.6.3! jdolecek 1314:                        rt_ref(rt);
        !          1315:                        RT_REFCNT_TRACE(rt);
1.92      dyoung   1316:                }
                   1317:                break;
1.1       cgd      1318:        }
                   1319: bad:
1.126.6.3! jdolecek 1320:        if (need_to_release_ifa)
        !          1321:                ifa_release(ifa, &psref_ifa);
        !          1322:        if_put(ifp, &psref_ifp);
        !          1323:        curlwp_bindx(bound);
        !          1324:        if (need_unlock)
        !          1325:                RT_UNLOCK();
1.1       cgd      1326:        splx(s);
1.95      dyoung   1327:        return error;
1.1       cgd      1328: }
                   1329:
1.10      mycroft  1330: int
1.94      dyoung   1331: rt_setgate(struct rtentry *rt, const struct sockaddr *gate)
1.10      mycroft  1332: {
1.126.6.3! jdolecek 1333:        struct sockaddr *new, *old;
1.94      dyoung   1334:
                   1335:        KASSERT(rt->_rt_key != NULL);
1.110     dyoung   1336:        RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
1.94      dyoung   1337:
1.126.6.3! jdolecek 1338:        new = sockaddr_dup(gate, M_ZERO | M_NOWAIT);
        !          1339:        if (new == NULL)
1.94      dyoung   1340:                return ENOMEM;
1.126.6.3! jdolecek 1341:
        !          1342:        old = rt->rt_gateway;
        !          1343:        rt->rt_gateway = new;
        !          1344:        if (old != NULL)
        !          1345:                sockaddr_free(old);
        !          1346:
1.94      dyoung   1347:        KASSERT(rt->_rt_key != NULL);
1.110     dyoung   1348:        RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
1.94      dyoung   1349:
1.10      mycroft  1350:        if (rt->rt_flags & RTF_GATEWAY) {
1.126.6.3! jdolecek 1351:                struct rtentry *gwrt;
        !          1352:
        !          1353:                /* XXX we cannot call rtalloc1 if holding the rt lock */
        !          1354:                if (RT_LOCKED())
        !          1355:                        gwrt = rtalloc1_locked(gate, 1, false, true);
        !          1356:                else
        !          1357:                        gwrt = rtalloc1(gate, 1);
1.27      matt     1358:                /*
                   1359:                 * If we switched gateways, grab the MTU from the new
1.47      itojun   1360:                 * gateway route if the current MTU, if the current MTU is
                   1361:                 * greater than the MTU of gateway.
                   1362:                 * Note that, if the MTU of gateway is 0, we will reset the
                   1363:                 * MTU of the route to run PMTUD again from scratch. XXX
1.27      matt     1364:                 */
1.126.6.3! jdolecek 1365:                if (gwrt != NULL) {
        !          1366:                        KASSERT(gwrt->_rt_key != NULL);
        !          1367:                        RT_DPRINTF("gwrt->_rt_key = %p\n", gwrt->_rt_key);
        !          1368:                        if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0 &&
        !          1369:                            rt->rt_rmx.rmx_mtu &&
        !          1370:                            rt->rt_rmx.rmx_mtu > gwrt->rt_rmx.rmx_mtu) {
        !          1371:                                rt->rt_rmx.rmx_mtu = gwrt->rt_rmx.rmx_mtu;
        !          1372:                        }
        !          1373:                        rt_unref(gwrt);
1.27      matt     1374:                }
1.10      mycroft  1375:        }
1.94      dyoung   1376:        KASSERT(rt->_rt_key != NULL);
1.110     dyoung   1377:        RT_DPRINTF("rt->_rt_key = %p\n", (void *)rt->_rt_key);
1.10      mycroft  1378:        return 0;
                   1379: }
                   1380:
1.126.6.3! jdolecek 1381: static void
1.60      matt     1382: rt_maskedcopy(const struct sockaddr *src, struct sockaddr *dst,
                   1383:        const struct sockaddr *netmask)
1.1       cgd      1384: {
1.94      dyoung   1385:        const char *netmaskp = &netmask->sa_data[0],
                   1386:                   *srcp = &src->sa_data[0];
                   1387:        char *dstp = &dst->sa_data[0];
1.126     christos 1388:        const char *maskend = (char *)dst + MIN(netmask->sa_len, src->sa_len);
                   1389:        const char *srcend = (char *)dst + src->sa_len;
1.94      dyoung   1390:
                   1391:        dst->sa_len = src->sa_len;
                   1392:        dst->sa_family = src->sa_family;
                   1393:
                   1394:        while (dstp < maskend)
                   1395:                *dstp++ = *srcp++ & *netmaskp++;
                   1396:        if (dstp < srcend)
                   1397:                memset(dstp, 0, (size_t)(srcend - dstp));
1.1       cgd      1398: }
1.10      mycroft  1399:
1.1       cgd      1400: /*
1.126.6.3! jdolecek 1401:  * Inform the routing socket of a route change.
        !          1402:  */
        !          1403: void
        !          1404: rt_newmsg(const int cmd, const struct rtentry *rt)
        !          1405: {
        !          1406:        struct rt_addrinfo info;
        !          1407:
        !          1408:        memset((void *)&info, 0, sizeof(info));
        !          1409:        info.rti_info[RTAX_DST] = rt_getkey(rt);
        !          1410:        info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
        !          1411:        info.rti_info[RTAX_NETMASK] = rt_mask(rt);
        !          1412:        if (rt->rt_ifp) {
        !          1413:                info.rti_info[RTAX_IFP] = rt->rt_ifp->if_dl->ifa_addr;
        !          1414:                info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr;
        !          1415:        }
        !          1416:
        !          1417:        rt_missmsg(cmd, &info, rt->rt_flags, 0);
        !          1418: }
        !          1419:
        !          1420: /*
1.29      sommerfe 1421:  * Set up or tear down a routing table entry, normally
1.1       cgd      1422:  * for an interface.
                   1423:  */
1.9       mycroft  1424: int
1.60      matt     1425: rtinit(struct ifaddr *ifa, int cmd, int flags)
1.1       cgd      1426: {
1.36      augustss 1427:        struct rtentry *rt;
                   1428:        struct sockaddr *dst, *odst;
1.94      dyoung   1429:        struct sockaddr_storage maskeddst;
1.68      christos 1430:        struct rtentry *nrt = NULL;
1.1       cgd      1431:        int error;
1.39      itojun   1432:        struct rt_addrinfo info;
1.1       cgd      1433:
                   1434:        dst = flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr;
                   1435:        if (cmd == RTM_DELETE) {
                   1436:                if ((flags & RTF_HOST) == 0 && ifa->ifa_netmask) {
1.29      sommerfe 1437:                        /* Delete subnet route for this interface */
                   1438:                        odst = dst;
1.94      dyoung   1439:                        dst = (struct sockaddr *)&maskeddst;
1.29      sommerfe 1440:                        rt_maskedcopy(odst, dst, ifa->ifa_netmask);
1.1       cgd      1441:                }
1.14      christos 1442:                if ((rt = rtalloc1(dst, 0)) != NULL) {
1.126.6.3! jdolecek 1443:                        if (rt->rt_ifa != ifa) {
        !          1444:                                rt_unref(rt);
1.85      dyoung   1445:                                return (flags & RTF_HOST) ? EHOSTUNREACH
                   1446:                                                        : ENETUNREACH;
1.126.6.3! jdolecek 1447:                        }
        !          1448:                        rt_unref(rt);
1.1       cgd      1449:                }
                   1450:        }
1.44      thorpej  1451:        memset(&info, 0, sizeof(info));
1.39      itojun   1452:        info.rti_ifa = ifa;
                   1453:        info.rti_flags = flags | ifa->ifa_flags;
                   1454:        info.rti_info[RTAX_DST] = dst;
                   1455:        info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
1.126.6.3! jdolecek 1456:
1.39      itojun   1457:        /*
                   1458:         * XXX here, it seems that we are assuming that ifa_netmask is NULL
                   1459:         * for RTF_HOST.  bsdi4 passes NULL explicitly (via intermediate
                   1460:         * variable) when RTF_HOST is 1.  still not sure if i can safely
                   1461:         * change it to meet bsdi4 behavior.
                   1462:         */
1.114     dyoung   1463:        if (cmd != RTM_LLINFO_UPD)
                   1464:                info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask;
                   1465:        error = rtrequest1((cmd == RTM_LLINFO_UPD) ? RTM_GET : cmd, &info,
                   1466:            &nrt);
1.126.6.3! jdolecek 1467:        if (error != 0)
        !          1468:                return error;
        !          1469:
        !          1470:        rt = nrt;
        !          1471:        RT_REFCNT_TRACE(rt);
        !          1472:        switch (cmd) {
1.114     dyoung   1473:        case RTM_DELETE:
1.126.6.3! jdolecek 1474:                rt_newmsg(cmd, rt);
        !          1475:                rt_free(rt);
1.114     dyoung   1476:                break;
                   1477:        case RTM_LLINFO_UPD:
                   1478:                if (cmd == RTM_LLINFO_UPD && ifa->ifa_rtrequest != NULL)
                   1479:                        ifa->ifa_rtrequest(RTM_LLINFO_UPD, rt, &info);
1.126.6.3! jdolecek 1480:                rt_newmsg(RTM_CHANGE, rt);
        !          1481:                rt_unref(rt);
1.114     dyoung   1482:                break;
                   1483:        case RTM_ADD:
1.126.6.3! jdolecek 1484:                /*
        !          1485:                 * XXX it looks just reverting rt_ifa replaced by ifa_rtrequest
        !          1486:                 * called via rtrequest1. Can we just prevent the replacement
        !          1487:                 * somehow and remove the following code? And also doesn't
        !          1488:                 * calling ifa_rtrequest(RTM_ADD) replace rt_ifa again?
        !          1489:                 */
1.10      mycroft  1490:                if (rt->rt_ifa != ifa) {
1.17      christos 1491:                        printf("rtinit: wrong ifa (%p) was (%p)\n", ifa,
                   1492:                                rt->rt_ifa);
1.126.6.3! jdolecek 1493: #ifdef NET_MPSAFE
        !          1494:                        KASSERT(!cpu_softintr_p());
        !          1495:
        !          1496:                        error = rt_update_prepare(rt);
        !          1497:                        if (error == 0) {
        !          1498: #endif
        !          1499:                                if (rt->rt_ifa->ifa_rtrequest != NULL) {
        !          1500:                                        rt->rt_ifa->ifa_rtrequest(RTM_DELETE,
        !          1501:                                            rt, &info);
        !          1502:                                }
        !          1503:                                rt_replace_ifa(rt, ifa);
        !          1504:                                rt->rt_ifp = ifa->ifa_ifp;
        !          1505:                                if (ifa->ifa_rtrequest != NULL)
        !          1506:                                        ifa->ifa_rtrequest(RTM_ADD, rt, &info);
        !          1507: #ifdef NET_MPSAFE
        !          1508:                                rt_update_finish(rt);
        !          1509:                        } else {
        !          1510:                                /*
        !          1511:                                 * If error != 0, the rtentry is being
        !          1512:                                 * destroyed, so doing nothing doesn't
        !          1513:                                 * matter.
        !          1514:                                 */
1.114     dyoung   1515:                        }
1.126.6.3! jdolecek 1516: #endif
1.10      mycroft  1517:                }
1.126.6.3! jdolecek 1518:                rt_newmsg(cmd, rt);
        !          1519:                rt_unref(rt);
        !          1520:                RT_REFCNT_TRACE(rt);
1.114     dyoung   1521:                break;
1.1       cgd      1522:        }
1.85      dyoung   1523:        return error;
1.18      kml      1524: }
                   1525:
                   1526: /*
1.126.6.3! jdolecek 1527:  * Create a local route entry for the address.
        !          1528:  * Announce the addition of the address and the route to the routing socket.
        !          1529:  */
        !          1530: int
        !          1531: rt_ifa_addlocal(struct ifaddr *ifa)
        !          1532: {
        !          1533:        struct rtentry *rt;
        !          1534:        int e;
        !          1535:
        !          1536:        /* If there is no loopback entry, allocate one. */
        !          1537:        rt = rtalloc1(ifa->ifa_addr, 0);
        !          1538: #ifdef RT_DEBUG
        !          1539:        if (rt != NULL)
        !          1540:                dump_rt(rt);
        !          1541: #endif
        !          1542:        if (rt == NULL || (rt->rt_flags & RTF_HOST) == 0 ||
        !          1543:            (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0)
        !          1544:        {
        !          1545:                struct rt_addrinfo info;
        !          1546:                struct rtentry *nrt;
        !          1547:
        !          1548:                memset(&info, 0, sizeof(info));
        !          1549:                info.rti_flags = RTF_HOST | RTF_LOCAL;
        !          1550:                info.rti_info[RTAX_DST] = ifa->ifa_addr;
        !          1551:                info.rti_info[RTAX_GATEWAY] =
        !          1552:                    (const struct sockaddr *)ifa->ifa_ifp->if_sadl;
        !          1553:                info.rti_ifa = ifa;
        !          1554:                nrt = NULL;
        !          1555:                e = rtrequest1(RTM_ADD, &info, &nrt);
        !          1556:                if (nrt && ifa != nrt->rt_ifa)
        !          1557:                        rt_replace_ifa(nrt, ifa);
        !          1558:                rt_newaddrmsg(RTM_ADD, ifa, e, nrt);
        !          1559:                if (nrt != NULL) {
        !          1560: #ifdef RT_DEBUG
        !          1561:                        dump_rt(nrt);
        !          1562: #endif
        !          1563:                        rt_unref(nrt);
        !          1564:                        RT_REFCNT_TRACE(nrt);
        !          1565:                }
        !          1566:        } else {
        !          1567:                e = 0;
        !          1568:                rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
        !          1569:        }
        !          1570:        if (rt != NULL)
        !          1571:                rt_unref(rt);
        !          1572:        return e;
        !          1573: }
        !          1574:
        !          1575: /*
        !          1576:  * Remove the local route entry for the address.
        !          1577:  * Announce the removal of the address and the route to the routing socket.
        !          1578:  */
        !          1579: int
        !          1580: rt_ifa_remlocal(struct ifaddr *ifa, struct ifaddr *alt_ifa)
        !          1581: {
        !          1582:        struct rtentry *rt;
        !          1583:        int e = 0;
        !          1584:
        !          1585:        rt = rtalloc1(ifa->ifa_addr, 0);
        !          1586:
        !          1587:        /*
        !          1588:         * Before deleting, check if a corresponding loopbacked
        !          1589:         * host route surely exists.  With this check, we can avoid
        !          1590:         * deleting an interface direct route whose destination is
        !          1591:         * the same as the address being removed.  This can happen
        !          1592:         * when removing a subnet-router anycast address on an
        !          1593:         * interface attached to a shared medium.
        !          1594:         */
        !          1595:        if (rt != NULL &&
        !          1596:            (rt->rt_flags & RTF_HOST) &&
        !          1597:            (rt->rt_ifp->if_flags & IFF_LOOPBACK))
        !          1598:        {
        !          1599:                /* If we cannot replace the route's ifaddr with the equivalent
        !          1600:                 * ifaddr of another interface, I believe it is safest to
        !          1601:                 * delete the route.
        !          1602:                 */
        !          1603:                if (alt_ifa == NULL) {
        !          1604:                        e = rtdeletemsg(rt);
        !          1605:                        if (e == 0) {
        !          1606:                                rt_unref(rt);
        !          1607:                                rt_free(rt);
        !          1608:                                rt = NULL;
        !          1609:                        }
        !          1610:                        rt_newaddrmsg(RTM_DELADDR, ifa, 0, NULL);
        !          1611:                } else {
        !          1612:                        rt_replace_ifa(rt, alt_ifa);
        !          1613:                        rt_newmsg(RTM_CHANGE, rt);
        !          1614:                }
        !          1615:        } else
        !          1616:                rt_newaddrmsg(RTM_DELADDR, ifa, 0, NULL);
        !          1617:        if (rt != NULL)
        !          1618:                rt_unref(rt);
        !          1619:        return e;
        !          1620: }
        !          1621:
        !          1622: /*
1.18      kml      1623:  * Route timer routines.  These routes allow functions to be called
                   1624:  * for various routes at any time.  This is useful in supporting
                   1625:  * path MTU discovery and redirect route deletion.
                   1626:  *
                   1627:  * This is similar to some BSDI internal functions, but it provides
                   1628:  * for multiple queues for efficiency's sake...
                   1629:  */
                   1630:
                   1631: LIST_HEAD(, rttimer_queue) rttimer_queue_head;
                   1632: static int rt_init_done = 0;
                   1633:
1.65      perry    1634: /*
1.18      kml      1635:  * Some subtle order problems with domain initialization mean that
                   1636:  * we cannot count on this being run from rt_init before various
                   1637:  * protocol initializations are done.  Therefore, we make sure
                   1638:  * that this is run when the first queue is added...
                   1639:  */
                   1640:
1.126.6.3! jdolecek 1641: static void rt_timer_work(struct work *, void *);
        !          1642:
        !          1643: static void
1.60      matt     1644: rt_timer_init(void)
1.18      kml      1645: {
1.126.6.3! jdolecek 1646:        int error;
        !          1647:
1.18      kml      1648:        assert(rt_init_done == 0);
                   1649:
1.126.6.3! jdolecek 1650:        /* XXX should be in rt_init */
        !          1651:        rw_init(&rt_lock);
        !          1652:
1.18      kml      1653:        LIST_INIT(&rttimer_queue_head);
1.126.6.3! jdolecek 1654:        callout_init(&rt_timer_ch, CALLOUT_MPSAFE);
        !          1655:        error = workqueue_create(&rt_timer_wq, "rt_timer",
        !          1656:            rt_timer_work, NULL, PRI_SOFTNET, IPL_SOFTNET, WQ_MPSAFE);
        !          1657:        if (error)
        !          1658:                panic("%s: workqueue_create failed (%d)\n", __func__, error);
1.35      thorpej  1659:        callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL);
1.18      kml      1660:        rt_init_done = 1;
                   1661: }
                   1662:
                   1663: struct rttimer_queue *
1.60      matt     1664: rt_timer_queue_create(u_int timeout)
1.18      kml      1665: {
                   1666:        struct rttimer_queue *rtq;
                   1667:
                   1668:        if (rt_init_done == 0)
                   1669:                rt_timer_init();
                   1670:
                   1671:        R_Malloc(rtq, struct rttimer_queue *, sizeof *rtq);
                   1672:        if (rtq == NULL)
1.85      dyoung   1673:                return NULL;
1.109     dyoung   1674:        memset(rtq, 0, sizeof(*rtq));
1.18      kml      1675:
                   1676:        rtq->rtq_timeout = timeout;
1.24      thorpej  1677:        TAILQ_INIT(&rtq->rtq_head);
1.126.6.3! jdolecek 1678:        RT_WLOCK();
1.18      kml      1679:        LIST_INSERT_HEAD(&rttimer_queue_head, rtq, rtq_link);
1.126.6.3! jdolecek 1680:        RT_UNLOCK();
1.18      kml      1681:
1.85      dyoung   1682:        return rtq;
1.18      kml      1683: }
                   1684:
                   1685: void
1.60      matt     1686: rt_timer_queue_change(struct rttimer_queue *rtq, long timeout)
1.18      kml      1687: {
1.24      thorpej  1688:
1.18      kml      1689:        rtq->rtq_timeout = timeout;
                   1690: }
                   1691:
1.126.6.3! jdolecek 1692: static void
        !          1693: rt_timer_queue_remove_all(struct rttimer_queue *rtq)
1.18      kml      1694: {
1.24      thorpej  1695:        struct rttimer *r;
1.18      kml      1696:
1.126.6.3! jdolecek 1697:        RT_ASSERT_WLOCK();
        !          1698:
1.24      thorpej  1699:        while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL) {
1.18      kml      1700:                LIST_REMOVE(r, rtt_link);
1.24      thorpej  1701:                TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
1.126.6.3! jdolecek 1702:                rt_ref(r->rtt_rt); /* XXX */
        !          1703:                RT_REFCNT_TRACE(r->rtt_rt);
        !          1704:                RT_UNLOCK();
        !          1705:                (*r->rtt_func)(r->rtt_rt, r);
1.22      thorpej  1706:                pool_put(&rttimer_pool, r);
1.126.6.3! jdolecek 1707:                RT_WLOCK();
1.37      itojun   1708:                if (rtq->rtq_count > 0)
                   1709:                        rtq->rtq_count--;
                   1710:                else
1.55      itojun   1711:                        printf("rt_timer_queue_remove_all: "
                   1712:                            "rtq_count reached 0\n");
1.18      kml      1713:        }
1.55      itojun   1714: }
                   1715:
                   1716: void
1.126.6.3! jdolecek 1717: rt_timer_queue_destroy(struct rttimer_queue *rtq)
1.55      itojun   1718: {
                   1719:
1.126.6.3! jdolecek 1720:        RT_WLOCK();
        !          1721:        rt_timer_queue_remove_all(rtq);
1.18      kml      1722:        LIST_REMOVE(rtq, rtq_link);
1.126.6.3! jdolecek 1723:        RT_UNLOCK();
1.22      thorpej  1724:
                   1725:        /*
                   1726:         * Caller is responsible for freeing the rttimer_queue structure.
                   1727:         */
1.18      kml      1728: }
                   1729:
1.37      itojun   1730: unsigned long
1.60      matt     1731: rt_timer_count(struct rttimer_queue *rtq)
1.37      itojun   1732: {
                   1733:        return rtq->rtq_count;
                   1734: }
                   1735:
1.126.6.3! jdolecek 1736: static void
        !          1737: rt_timer_remove_all(struct rtentry *rt)
1.18      kml      1738: {
1.24      thorpej  1739:        struct rttimer *r;
1.18      kml      1740:
1.126.6.3! jdolecek 1741:        RT_WLOCK();
1.24      thorpej  1742:        while ((r = LIST_FIRST(&rt->rt_timer)) != NULL) {
1.18      kml      1743:                LIST_REMOVE(r, rtt_link);
1.24      thorpej  1744:                TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
1.37      itojun   1745:                if (r->rtt_queue->rtq_count > 0)
                   1746:                        r->rtt_queue->rtq_count--;
                   1747:                else
                   1748:                        printf("rt_timer_remove_all: rtq_count reached 0\n");
1.38      itojun   1749:                pool_put(&rttimer_pool, r);
1.18      kml      1750:        }
1.126.6.3! jdolecek 1751:        RT_UNLOCK();
1.18      kml      1752: }
                   1753:
1.65      perry    1754: int
1.60      matt     1755: rt_timer_add(struct rtentry *rt,
                   1756:        void (*func)(struct rtentry *, struct rttimer *),
                   1757:        struct rttimer_queue *queue)
1.18      kml      1758: {
1.24      thorpej  1759:        struct rttimer *r;
1.18      kml      1760:
1.126.6.3! jdolecek 1761:        KASSERT(func != NULL);
        !          1762:        RT_WLOCK();
1.24      thorpej  1763:        /*
                   1764:         * If there's already a timer with this action, destroy it before
                   1765:         * we add a new one.
                   1766:         */
1.85      dyoung   1767:        LIST_FOREACH(r, &rt->rt_timer, rtt_link) {
                   1768:                if (r->rtt_func == func)
                   1769:                        break;
                   1770:        }
                   1771:        if (r != NULL) {
                   1772:                LIST_REMOVE(r, rtt_link);
                   1773:                TAILQ_REMOVE(&r->rtt_queue->rtq_head, r, rtt_next);
                   1774:                if (r->rtt_queue->rtq_count > 0)
                   1775:                        r->rtt_queue->rtq_count--;
                   1776:                else
                   1777:                        printf("rt_timer_add: rtq_count reached 0\n");
                   1778:        } else {
                   1779:                r = pool_get(&rttimer_pool, PR_NOWAIT);
1.126.6.3! jdolecek 1780:                if (r == NULL) {
        !          1781:                        RT_UNLOCK();
1.85      dyoung   1782:                        return ENOBUFS;
1.126.6.3! jdolecek 1783:                }
1.18      kml      1784:        }
                   1785:
1.85      dyoung   1786:        memset(r, 0, sizeof(*r));
1.24      thorpej  1787:
                   1788:        r->rtt_rt = rt;
1.70      kardel   1789:        r->rtt_time = time_uptime;
1.24      thorpej  1790:        r->rtt_func = func;
                   1791:        r->rtt_queue = queue;
                   1792:        LIST_INSERT_HEAD(&rt->rt_timer, r, rtt_link);
                   1793:        TAILQ_INSERT_TAIL(&queue->rtq_head, r, rtt_next);
1.37      itojun   1794:        r->rtt_queue->rtq_count++;
1.65      perry    1795:
1.126.6.3! jdolecek 1796:        RT_UNLOCK();
        !          1797:
1.95      dyoung   1798:        return 0;
1.18      kml      1799: }
                   1800:
1.126.6.3! jdolecek 1801: static void
        !          1802: rt_timer_work(struct work *wk, void *arg)
1.18      kml      1803: {
1.24      thorpej  1804:        struct rttimer_queue *rtq;
                   1805:        struct rttimer *r;
1.21      kml      1806:
1.126.6.3! jdolecek 1807:        RT_WLOCK();
1.85      dyoung   1808:        LIST_FOREACH(rtq, &rttimer_queue_head, rtq_link) {
1.24      thorpej  1809:                while ((r = TAILQ_FIRST(&rtq->rtq_head)) != NULL &&
1.70      kardel   1810:                    (r->rtt_time + rtq->rtq_timeout) < time_uptime) {
1.24      thorpej  1811:                        LIST_REMOVE(r, rtt_link);
                   1812:                        TAILQ_REMOVE(&rtq->rtq_head, r, rtt_next);
1.126.6.3! jdolecek 1813:                        rt_ref(r->rtt_rt); /* XXX */
        !          1814:                        RT_REFCNT_TRACE(r->rtt_rt);
        !          1815:                        RT_UNLOCK();
        !          1816:                        (*r->rtt_func)(r->rtt_rt, r);
1.24      thorpej  1817:                        pool_put(&rttimer_pool, r);
1.126.6.3! jdolecek 1818:                        RT_WLOCK();
1.37      itojun   1819:                        if (rtq->rtq_count > 0)
                   1820:                                rtq->rtq_count--;
                   1821:                        else
                   1822:                                printf("rt_timer_timer: rtq_count reached 0\n");
1.18      kml      1823:                }
                   1824:        }
1.126.6.3! jdolecek 1825:        RT_UNLOCK();
1.18      kml      1826:
1.35      thorpej  1827:        callout_reset(&rt_timer_ch, hz, rt_timer_timer, NULL);
1.1       cgd      1828: }
1.83      joerg    1829:
1.126.6.3! jdolecek 1830: static void
        !          1831: rt_timer_timer(void *arg)
        !          1832: {
        !          1833:
        !          1834:        workqueue_enqueue(rt_timer_wq, &rt_timer_wk, NULL);
        !          1835: }
        !          1836:
1.102     dyoung   1837: static struct rtentry *
1.84      joerg    1838: _rtcache_init(struct route *ro, int flag)
                   1839: {
1.126.6.3! jdolecek 1840:        struct rtentry *rt;
        !          1841:
1.114     dyoung   1842:        rtcache_invariants(ro);
1.99      dyoung   1843:        KASSERT(ro->_ro_rt == NULL);
1.84      joerg    1844:
1.90      dyoung   1845:        if (rtcache_getdst(ro) == NULL)
1.102     dyoung   1846:                return NULL;
1.126.6.3! jdolecek 1847:        rt = rtalloc1(rtcache_getdst(ro), flag);
        !          1848:        if (rt != NULL) {
        !          1849:                RT_RLOCK();
        !          1850:                if (ISSET(rt->rt_flags, RTF_UP)) {
        !          1851:                        ro->_ro_rt = rt;
        !          1852:                        ro->ro_rtcache_generation = rtcache_generation;
        !          1853:                        rtcache_ref(rt, ro);
        !          1854:                }
        !          1855:                RT_UNLOCK();
        !          1856:                rt_unref(rt);
        !          1857:        }
1.103     dyoung   1858:
1.114     dyoung   1859:        rtcache_invariants(ro);
1.102     dyoung   1860:        return ro->_ro_rt;
1.84      joerg    1861: }
                   1862:
1.102     dyoung   1863: struct rtentry *
1.83      joerg    1864: rtcache_init(struct route *ro)
                   1865: {
1.126.6.3! jdolecek 1866:
1.102     dyoung   1867:        return _rtcache_init(ro, 1);
1.83      joerg    1868: }
                   1869:
1.102     dyoung   1870: struct rtentry *
1.83      joerg    1871: rtcache_init_noclone(struct route *ro)
                   1872: {
1.126.6.3! jdolecek 1873:
1.102     dyoung   1874:        return _rtcache_init(ro, 0);
1.83      joerg    1875: }
1.90      dyoung   1876:
1.102     dyoung   1877: struct rtentry *
1.90      dyoung   1878: rtcache_update(struct route *ro, int clone)
                   1879: {
1.126.6.3! jdolecek 1880:
        !          1881:        ro->_ro_rt = NULL;
1.102     dyoung   1882:        return _rtcache_init(ro, clone);
1.90      dyoung   1883: }
1.83      joerg    1884:
                   1885: void
1.126.6.3! jdolecek 1886: rtcache_copy(struct route *new_ro, struct route *old_ro)
1.83      joerg    1887: {
1.103     dyoung   1888:        struct rtentry *rt;
1.126.6.3! jdolecek 1889:        int ret;
1.103     dyoung   1890:
                   1891:        KASSERT(new_ro != old_ro);
1.114     dyoung   1892:        rtcache_invariants(new_ro);
                   1893:        rtcache_invariants(old_ro);
1.103     dyoung   1894:
1.126.6.3! jdolecek 1895:        rt = rtcache_validate(old_ro);
1.103     dyoung   1896:
1.126.6.3! jdolecek 1897:        if (rtcache_getdst(old_ro) == NULL)
        !          1898:                goto out;
        !          1899:        ret = rtcache_setdst(new_ro, rtcache_getdst(old_ro));
        !          1900:        if (ret != 0)
        !          1901:                goto out;
1.103     dyoung   1902:
1.126.6.3! jdolecek 1903:        RT_RLOCK();
        !          1904:        new_ro->_ro_rt = rt;
        !          1905:        new_ro->ro_rtcache_generation = rtcache_generation;
        !          1906:        RT_UNLOCK();
1.114     dyoung   1907:        rtcache_invariants(new_ro);
1.126.6.3! jdolecek 1908: out:
        !          1909:        rtcache_unref(rt, old_ro);
        !          1910:        return;
1.83      joerg    1911: }
                   1912:
1.126.6.3! jdolecek 1913: #if defined(RT_DEBUG) && defined(NET_MPSAFE)
        !          1914: static void
        !          1915: rtcache_trace(const char *func, struct rtentry *rt, struct route *ro)
        !          1916: {
        !          1917:        char dst[64];
1.105     dyoung   1918:
1.126.6.3! jdolecek 1919:        sockaddr_format(ro->ro_sa, dst, 64);
        !          1920:        printf("trace: %s:\tdst=%s cpu=%d lwp=%p psref=%p target=%p\n", func, dst,
        !          1921:            cpu_index(curcpu()), curlwp, &ro->ro_psref, &rt->rt_psref);
        !          1922: }
        !          1923: #define RTCACHE_PSREF_TRACE(rt, ro)    rtcache_trace(__func__, (rt), (ro))
        !          1924: #else
        !          1925: #define RTCACHE_PSREF_TRACE(rt, ro)    do {} while (0)
        !          1926: #endif
        !          1927:
        !          1928: static void
        !          1929: rtcache_ref(struct rtentry *rt, struct route *ro)
1.83      joerg    1930: {
1.99      dyoung   1931:
1.126.6.3! jdolecek 1932:        KASSERT(rt != NULL);
        !          1933:
        !          1934: #ifdef NET_MPSAFE
        !          1935:        RTCACHE_PSREF_TRACE(rt, ro);
        !          1936:        ro->ro_bound = curlwp_bind();
        !          1937:        psref_acquire(&ro->ro_psref, &rt->rt_psref, rt_psref_class);
        !          1938: #endif
1.105     dyoung   1939: }
                   1940:
                   1941: void
1.126.6.3! jdolecek 1942: rtcache_unref(struct rtentry *rt, struct route *ro)
1.105     dyoung   1943: {
1.126.6.3! jdolecek 1944:
        !          1945:        if (rt == NULL)
1.105     dyoung   1946:                return;
                   1947:
1.126.6.3! jdolecek 1948: #ifdef NET_MPSAFE
        !          1949:        psref_release(&ro->ro_psref, &rt->rt_psref, rt_psref_class);
        !          1950:        curlwp_bindx(ro->ro_bound);
        !          1951:        RTCACHE_PSREF_TRACE(rt, ro);
        !          1952: #endif
        !          1953: }
1.105     dyoung   1954:
1.126.6.3! jdolecek 1955: struct rtentry *
        !          1956: rtcache_validate(struct route *ro)
        !          1957: {
        !          1958:        struct rtentry *rt = NULL;
        !          1959:
        !          1960: #ifdef NET_MPSAFE
        !          1961: retry:
        !          1962: #endif
1.114     dyoung   1963:        rtcache_invariants(ro);
1.126.6.3! jdolecek 1964:        RT_RLOCK();
        !          1965:        if (ro->ro_rtcache_generation != rtcache_generation) {
        !          1966:                /* The cache is invalidated */
        !          1967:                rt = NULL;
        !          1968:                goto out;
        !          1969:        }
        !          1970:
        !          1971:        rt = ro->_ro_rt;
        !          1972:        if (rt == NULL)
        !          1973:                goto out;
        !          1974:
        !          1975:        if ((rt->rt_flags & RTF_UP) == 0) {
        !          1976:                rt = NULL;
        !          1977:                goto out;
        !          1978:        }
        !          1979: #ifdef NET_MPSAFE
        !          1980:        if (ISSET(rt->rt_flags, RTF_UPDATING)) {
        !          1981:                if (rt_wait_ok()) {
        !          1982:                        RT_UNLOCK();
        !          1983:
        !          1984:                        /* We can wait until the update is complete */
        !          1985:                        rt_update_wait();
        !          1986:                        goto retry;
        !          1987:                } else {
        !          1988:                        rt = NULL;
        !          1989:                }
        !          1990:        } else
        !          1991: #endif
        !          1992:                rtcache_ref(rt, ro);
        !          1993: out:
        !          1994:        RT_UNLOCK();
        !          1995:        return rt;
1.83      joerg    1996: }
                   1997:
1.90      dyoung   1998: struct rtentry *
1.126.6.3! jdolecek 1999: rtcache_lookup2(struct route *ro, const struct sockaddr *dst,
        !          2000:     int clone, int *hitp)
1.90      dyoung   2001: {
                   2002:        const struct sockaddr *odst;
1.104     dyoung   2003:        struct rtentry *rt = NULL;
1.90      dyoung   2004:
                   2005:        odst = rtcache_getdst(ro);
                   2006:        if (odst == NULL)
1.126.6.3! jdolecek 2007:                goto miss;
        !          2008:
        !          2009:        if (sockaddr_cmp(odst, dst) != 0) {
1.90      dyoung   2010:                rtcache_free(ro);
1.126.6.3! jdolecek 2011:                goto miss;
        !          2012:        }
1.90      dyoung   2013:
1.126.6.3! jdolecek 2014:        rt = rtcache_validate(ro);
1.104     dyoung   2015:        if (rt == NULL) {
1.126.6.3! jdolecek 2016:                ro->_ro_rt = NULL;
        !          2017:                goto miss;
        !          2018:        }
        !          2019:
        !          2020:        rtcache_invariants(ro);
        !          2021:
        !          2022:        if (hitp != NULL)
1.91      dyoung   2023:                *hitp = 1;
1.126.6.3! jdolecek 2024:        return rt;
        !          2025: miss:
        !          2026:        if (hitp != NULL)
        !          2027:                *hitp = 0;
        !          2028:        if (rtcache_setdst(ro, dst) == 0)
        !          2029:                rt = _rtcache_init(ro, clone);
1.90      dyoung   2030:
1.114     dyoung   2031:        rtcache_invariants(ro);
                   2032:
1.104     dyoung   2033:        return rt;
1.90      dyoung   2034: }
                   2035:
1.83      joerg    2036: void
1.86      dyoung   2037: rtcache_free(struct route *ro)
                   2038: {
1.126.6.3! jdolecek 2039:
        !          2040:        ro->_ro_rt = NULL;
1.86      dyoung   2041:        if (ro->ro_sa != NULL) {
                   2042:                sockaddr_free(ro->ro_sa);
                   2043:                ro->ro_sa = NULL;
                   2044:        }
1.114     dyoung   2045:        rtcache_invariants(ro);
1.86      dyoung   2046: }
                   2047:
1.90      dyoung   2048: int
                   2049: rtcache_setdst(struct route *ro, const struct sockaddr *sa)
1.83      joerg    2050: {
1.90      dyoung   2051:        KASSERT(sa != NULL);
                   2052:
1.114     dyoung   2053:        rtcache_invariants(ro);
1.126.6.3! jdolecek 2054:        if (ro->ro_sa != NULL) {
        !          2055:                if (ro->ro_sa->sa_family == sa->sa_family) {
        !          2056:                        ro->_ro_rt = NULL;
        !          2057:                        sockaddr_copy(ro->ro_sa, ro->ro_sa->sa_len, sa);
1.114     dyoung   2058:                        rtcache_invariants(ro);
1.96      dyoung   2059:                        return 0;
1.114     dyoung   2060:                }
1.126.6.3! jdolecek 2061:                /* free ro_sa, wrong family */
        !          2062:                rtcache_free(ro);
        !          2063:        }
1.90      dyoung   2064:
1.107     dyoung   2065:        KASSERT(ro->_ro_rt == NULL);
                   2066:
1.126.6.3! jdolecek 2067:        if ((ro->ro_sa = sockaddr_dup(sa, M_ZERO | M_NOWAIT)) == NULL) {
1.114     dyoung   2068:                rtcache_invariants(ro);
1.90      dyoung   2069:                return ENOMEM;
1.107     dyoung   2070:        }
1.114     dyoung   2071:        rtcache_invariants(ro);
1.90      dyoung   2072:        return 0;
1.83      joerg    2073: }
1.92      dyoung   2074:
1.123     kefren   2075: const struct sockaddr *
                   2076: rt_settag(struct rtentry *rt, const struct sockaddr *tag)
                   2077: {
                   2078:        if (rt->rt_tag != tag) {
                   2079:                if (rt->rt_tag != NULL)
                   2080:                        sockaddr_free(rt->rt_tag);
1.126.6.3! jdolecek 2081:                rt->rt_tag = sockaddr_dup(tag, M_ZERO | M_NOWAIT);
1.123     kefren   2082:        }
1.126.6.3! jdolecek 2083:        return rt->rt_tag;
1.123     kefren   2084: }
                   2085:
                   2086: struct sockaddr *
1.126.6.3! jdolecek 2087: rt_gettag(const struct rtentry *rt)
1.123     kefren   2088: {
                   2089:        return rt->rt_tag;
                   2090: }
1.126.6.3! jdolecek 2091:
        !          2092: int
        !          2093: rt_check_reject_route(const struct rtentry *rt, const struct ifnet *ifp)
        !          2094: {
        !          2095:
        !          2096:        if ((rt->rt_flags & RTF_REJECT) != 0) {
        !          2097:                /* Mimic looutput */
        !          2098:                if (ifp->if_flags & IFF_LOOPBACK)
        !          2099:                        return (rt->rt_flags & RTF_HOST) ?
        !          2100:                            EHOSTUNREACH : ENETUNREACH;
        !          2101:                else if (rt->rt_rmx.rmx_expire == 0 ||
        !          2102:                    time_uptime < rt->rt_rmx.rmx_expire)
        !          2103:                        return (rt->rt_flags & RTF_GATEWAY) ?
        !          2104:                            EHOSTUNREACH : EHOSTDOWN;
        !          2105:        }
        !          2106:
        !          2107:        return 0;
        !          2108: }
        !          2109:
        !          2110: void
        !          2111: rt_delete_matched_entries(sa_family_t family, int (*f)(struct rtentry *, void *),
        !          2112:     void *v)
        !          2113: {
        !          2114:
        !          2115:        for (;;) {
        !          2116:                int s;
        !          2117:                int error;
        !          2118:                struct rtentry *rt, *retrt = NULL;
        !          2119:
        !          2120:                RT_RLOCK();
        !          2121:                s = splsoftnet();
        !          2122:                rt = rtbl_search_matched_entry(family, f, v);
        !          2123:                if (rt == NULL) {
        !          2124:                        splx(s);
        !          2125:                        RT_UNLOCK();
        !          2126:                        return;
        !          2127:                }
        !          2128:                rt->rt_refcnt++;
        !          2129:                splx(s);
        !          2130:                RT_UNLOCK();
        !          2131:
        !          2132:                error = rtrequest(RTM_DELETE, rt_getkey(rt), rt->rt_gateway,
        !          2133:                    rt_mask(rt), rt->rt_flags, &retrt);
        !          2134:                if (error == 0) {
        !          2135:                        KASSERT(retrt == rt);
        !          2136:                        KASSERT((retrt->rt_flags & RTF_UP) == 0);
        !          2137:                        retrt->rt_ifp = NULL;
        !          2138:                        rt_unref(rt);
        !          2139:                        rt_free(retrt);
        !          2140:                } else if (error == ESRCH) {
        !          2141:                        /* Someone deleted the entry already. */
        !          2142:                        rt_unref(rt);
        !          2143:                } else {
        !          2144:                        log(LOG_ERR, "%s: unable to delete rtentry @ %p, "
        !          2145:                            "error = %d\n", rt->rt_ifp->if_xname, rt, error);
        !          2146:                        /* XXX how to treat this case? */
        !          2147:                }
        !          2148:        }
        !          2149: }
        !          2150:
        !          2151: int
        !          2152: rt_walktree(sa_family_t family, int (*f)(struct rtentry *, void *), void *v)
        !          2153: {
        !          2154:        int error;
        !          2155:
        !          2156:        RT_RLOCK();
        !          2157:        error = rtbl_walktree(family, f, v);
        !          2158:        RT_UNLOCK();
        !          2159:
        !          2160:        return error;
        !          2161: }
        !          2162:
        !          2163: #ifdef DDB
        !          2164:
        !          2165: #include <machine/db_machdep.h>
        !          2166: #include <ddb/db_interface.h>
        !          2167: #include <ddb/db_output.h>
        !          2168:
        !          2169: #define        rt_expire rt_rmx.rmx_expire
        !          2170:
        !          2171: static void
        !          2172: db_print_sa(const struct sockaddr *sa)
        !          2173: {
        !          2174:        int len;
        !          2175:        const u_char *p;
        !          2176:
        !          2177:        if (sa == NULL) {
        !          2178:                db_printf("[NULL]");
        !          2179:                return;
        !          2180:        }
        !          2181:
        !          2182:        p = (const u_char *)sa;
        !          2183:        len = sa->sa_len;
        !          2184:        db_printf("[");
        !          2185:        while (len > 0) {
        !          2186:                db_printf("%d", *p);
        !          2187:                p++; len--;
        !          2188:                if (len) db_printf(",");
        !          2189:        }
        !          2190:        db_printf("]\n");
        !          2191: }
        !          2192:
        !          2193: static void
        !          2194: db_print_ifa(struct ifaddr *ifa)
        !          2195: {
        !          2196:        if (ifa == NULL)
        !          2197:                return;
        !          2198:        db_printf("  ifa_addr=");
        !          2199:        db_print_sa(ifa->ifa_addr);
        !          2200:        db_printf("  ifa_dsta=");
        !          2201:        db_print_sa(ifa->ifa_dstaddr);
        !          2202:        db_printf("  ifa_mask=");
        !          2203:        db_print_sa(ifa->ifa_netmask);
        !          2204:        db_printf("  flags=0x%x,refcnt=%d,metric=%d\n",
        !          2205:                          ifa->ifa_flags,
        !          2206:                          ifa->ifa_refcnt,
        !          2207:                          ifa->ifa_metric);
        !          2208: }
        !          2209:
        !          2210: /*
        !          2211:  * Function to pass to rt_walktree().
        !          2212:  * Return non-zero error to abort walk.
        !          2213:  */
        !          2214: static int
        !          2215: db_show_rtentry(struct rtentry *rt, void *w)
        !          2216: {
        !          2217:        db_printf("rtentry=%p", rt);
        !          2218:
        !          2219:        db_printf(" flags=0x%x refcnt=%d use=%"PRId64" expire=%"PRId64"\n",
        !          2220:                          rt->rt_flags, rt->rt_refcnt,
        !          2221:                          rt->rt_use, (uint64_t)rt->rt_expire);
        !          2222:
        !          2223:        db_printf(" key="); db_print_sa(rt_getkey(rt));
        !          2224:        db_printf(" mask="); db_print_sa(rt_mask(rt));
        !          2225:        db_printf(" gw="); db_print_sa(rt->rt_gateway);
        !          2226:
        !          2227:        db_printf(" ifp=%p ", rt->rt_ifp);
        !          2228:        if (rt->rt_ifp)
        !          2229:                db_printf("(%s)", rt->rt_ifp->if_xname);
        !          2230:        else
        !          2231:                db_printf("(NULL)");
        !          2232:
        !          2233:        db_printf(" ifa=%p\n", rt->rt_ifa);
        !          2234:        db_print_ifa(rt->rt_ifa);
        !          2235:
        !          2236:        db_printf(" gwroute=%p llinfo=%p\n",
        !          2237:                          rt->rt_gwroute, rt->rt_llinfo);
        !          2238:
        !          2239:        return 0;
        !          2240: }
        !          2241:
        !          2242: /*
        !          2243:  * Function to print all the route trees.
        !          2244:  * Use this from ddb:  "show routes"
        !          2245:  */
        !          2246: void
        !          2247: db_show_routes(db_expr_t addr, bool have_addr,
        !          2248:     db_expr_t count, const char *modif)
        !          2249: {
        !          2250:        rt_walktree(AF_INET, db_show_rtentry, NULL);
        !          2251: }
        !          2252: #endif

CVSweb <webmaster@jp.NetBSD.org>