X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/9bccf70c0258c7cac2dcb80011b2a964d884c552..d41d1dae2cd00cc08c7982087d1c445180cad9f5:/bsd/netinet6/in6_rmx.c diff --git a/bsd/netinet6/in6_rmx.c b/bsd/netinet6/in6_rmx.c index 805f2a9e0..d0ad6f2b1 100644 --- a/bsd/netinet6/in6_rmx.c +++ b/bsd/netinet6/in6_rmx.c @@ -1,3 +1,31 @@ +/* + * Copyright (c) 2003-2009 Apple Inc. All rights reserved. + * + * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ + * + * This file contains Original Code and/or Modifications of Original Code + * as defined in and that are subject to the Apple Public Source License + * Version 2.0 (the 'License'). You may not use this file except in + * compliance with the License. The rights granted to you under the License + * may not be used to create, or enable the creation or redistribution of, + * unlawful or unlicensed copies of an Apple operating system, or to + * circumvent, violate, or enable the circumvention or violation of, any + * terms of an Apple operating system software license agreement. + * + * Please obtain a copy of the License at + * http://www.opensource.apple.com/apsl/ and read it before using this file. + * + * The Original Code and all software distributed under the License are + * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER + * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, + * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. + * Please see the License for the specific language governing rights and + * limitations under the License. + * + * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ + */ + /* $FreeBSD: src/sys/netinet6/in6_rmx.c,v 1.1.2.2 2001/07/03 11:01:52 ume Exp $ */ /* $KAME: in6_rmx.c,v 1.10 2001/05/24 05:44:58 itojun Exp $ */ @@ -82,6 +110,7 @@ #include #include #include +#include #include #include @@ -99,12 +128,22 @@ #include #include -extern int in6_inithead __P((void **head, int off)); -static void in6_rtqtimo __P((void *rock)); -static void in6_mtutimo __P((void *rock)); +extern int in6_inithead(void **head, int off); +static void in6_rtqtimo(void *rock); +static void in6_mtutimo(void *rock); +extern int tvtohz(struct timeval *); + +static struct radix_node *in6_matroute_args(void *, struct radix_node_head *, + rn_matchf_t *, void *); #define RTPRF_OURS RTF_PROTO3 /* set on routes we manage */ +/* + * Accessed by in6_addroute(), in6_deleteroute() and in6_rtqkill(), during + * which the routing lock (rnh_lock) is held and thus protects the variable. + */ +static int in6dynroutes; + /* * Do what we need to do when inserting a route. */ @@ -116,6 +155,21 @@ in6_addroute(void *v_arg, void *n_arg, struct radix_node_head *head, struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)rt_key(rt); struct radix_node *ret; + lck_mtx_assert(rnh_lock, LCK_MTX_ASSERT_OWNED); + RT_LOCK_ASSERT_HELD(rt); + + /* + * If this is a dynamic route (which is created via Redirect) and + * we already have the maximum acceptable number of such route entries, + * reject creating a new one. We could initiate garbage collection to + * make available space right now, but the benefit would probably not + * be worth the cleaning overhead; we only have to endure a slightly + * suboptimal path even without the redirecbted route. + */ + if ((rt->rt_flags & RTF_DYNAMIC) != 0 && + ip6_maxdynroutes >= 0 && in6dynroutes >= ip6_maxdynroutes) + return (NULL); + /* * For IPv6, all unicast non-host routes are automatically cloning. */ @@ -153,30 +207,39 @@ in6_addroute(void *v_arg, void *n_arg, struct radix_node_head *head, rt->rt_rmx.rmx_mtu = rt->rt_ifp->if_mtu; ret = rn_addroute(v_arg, n_arg, head, treenodes); - if (ret == NULL && rt->rt_flags & RTF_HOST) { + if (ret == NULL && (rt->rt_flags & RTF_HOST)) { struct rtentry *rt2; /* * We are trying to add a host route, but can't. * Find out if it is because of an * ARP entry and delete it if so. */ - rt2 = rtalloc1((struct sockaddr *)sin6, 0, + rt2 = rtalloc1_locked((struct sockaddr *)sin6, 0, RTF_CLONING | RTF_PRCLONING); if (rt2) { - if (rt2->rt_flags & RTF_LLINFO && - rt2->rt_flags & RTF_HOST && - rt2->rt_gateway && - rt2->rt_gateway->sa_family == AF_LINK) { - rtrequest(RTM_DELETE, - (struct sockaddr *)rt_key(rt2), - rt2->rt_gateway, - rt_mask(rt2), rt2->rt_flags, 0); + RT_LOCK(rt2); + if ((rt2->rt_flags & RTF_LLINFO) && + (rt2->rt_flags & RTF_HOST) && + rt2->rt_gateway != NULL && + rt2->rt_gateway->sa_family == AF_LINK) { + /* + * Safe to drop rt_lock and use rt_key, + * rt_gateway, since holding rnh_lock here + * prevents another thread from calling + * rt_setgate() on this route. + */ + RT_UNLOCK(rt2); + (void) rtrequest_locked(RTM_DELETE, rt_key(rt2), + rt2->rt_gateway, rt_mask(rt2), + rt2->rt_flags, 0); ret = rn_addroute(v_arg, n_arg, head, treenodes); + } else { + RT_UNLOCK(rt2); } - rtfree(rt2); + rtfree_locked(rt2); } - } else if (ret == NULL && rt->rt_flags & RTF_CLONING) { + } else if (ret == NULL && (rt->rt_flags & RTF_CLONING)) { struct rtentry *rt2; /* * We are trying to add a net route, but can't. @@ -190,9 +253,10 @@ in6_addroute(void *v_arg, void *n_arg, struct radix_node_head *head, * net route entry, 3ffe:0501:: -> if0. * This case should not raise an error. */ - rt2 = rtalloc1((struct sockaddr *)sin6, 0, + rt2 = rtalloc1_locked((struct sockaddr *)sin6, 0, RTF_CLONING | RTF_PRCLONING); if (rt2) { + RT_LOCK(rt2); if ((rt2->rt_flags & (RTF_CLONING|RTF_HOST|RTF_GATEWAY)) == RTF_CLONING && rt2->rt_gateway @@ -200,30 +264,67 @@ in6_addroute(void *v_arg, void *n_arg, struct radix_node_head *head, && rt2->rt_ifp == rt->rt_ifp) { ret = rt2->rt_nodes; } - rtfree(rt2); + RT_UNLOCK(rt2); + rtfree_locked(rt2); } } + + if (ret != NULL && (rt->rt_flags & RTF_DYNAMIC) != 0) + in6dynroutes++; + return ret; } +static struct radix_node * +in6_deleteroute(void * v_arg, void *netmask_arg, struct radix_node_head *head) +{ + struct radix_node *rn; + + lck_mtx_assert(rnh_lock, LCK_MTX_ASSERT_OWNED); + + rn = rn_delete(v_arg, netmask_arg, head); + if (rn != NULL) { + struct rtentry *rt = (struct rtentry *)rn; + RT_LOCK_SPIN(rt); + if ((rt->rt_flags & RTF_DYNAMIC) != 0) + in6dynroutes--; + RT_UNLOCK(rt); + } + + return (rn); +} + +/* + * Similar to in6_matroute_args except without the leaf-matching parameters. + */ +static struct radix_node * +in6_matroute(void *v_arg, struct radix_node_head *head) +{ + return (in6_matroute_args(v_arg, head, NULL, NULL)); +} + /* * This code is the inverse of in6_clsroute: on first reference, if we * were managing the route, stop doing so and set the expiration timer * back off again. */ static struct radix_node * -in6_matroute(void *v_arg, struct radix_node_head *head) +in6_matroute_args(void *v_arg, struct radix_node_head *head, + rn_matchf_t *f, void *w) { - struct radix_node *rn = rn_match(v_arg, head); + struct radix_node *rn = rn_match_args(v_arg, head, f, w); struct rtentry *rt = (struct rtentry *)rn; - if (rt && rt->rt_refcnt == 0) { /* this is first reference */ - if (rt->rt_flags & RTPRF_OURS) { + /* This is first reference? */ + if (rt != NULL) { + RT_LOCK_SPIN(rt); + if (rt->rt_refcnt == 0 && (rt->rt_flags & RTPRF_OURS)) { rt->rt_flags &= ~RTPRF_OURS; rt->rt_rmx.rmx_expire = 0; } + RT_UNLOCK(rt); } - return rn; + return (rn); } SYSCTL_DECL(_net_inet6_ip6); @@ -232,50 +333,73 @@ static int rtq_reallyold = 60*60; /* one hour is ``really old'' */ SYSCTL_INT(_net_inet6_ip6, IPV6CTL_RTEXPIRE, rtexpire, CTLFLAG_RW, &rtq_reallyold , 0, ""); - + static int rtq_minreallyold = 10; /* never automatically crank down to less */ SYSCTL_INT(_net_inet6_ip6, IPV6CTL_RTMINEXPIRE, rtminexpire, CTLFLAG_RW, &rtq_minreallyold , 0, ""); - + static int rtq_toomany = 128; /* 128 cached routes is ``too many'' */ SYSCTL_INT(_net_inet6_ip6, IPV6CTL_RTMAXCACHE, rtmaxcache, CTLFLAG_RW, &rtq_toomany , 0, ""); - + /* * On last reference drop, mark the route as belong to us so that it can be * timed out. */ static void -in6_clsroute(struct radix_node *rn, struct radix_node_head *head) +in6_clsroute(struct radix_node *rn, __unused struct radix_node_head *head) { struct rtentry *rt = (struct rtentry *)rn; + lck_mtx_assert(rnh_lock, LCK_MTX_ASSERT_OWNED); + RT_LOCK_ASSERT_HELD(rt); + if (!(rt->rt_flags & RTF_UP)) return; /* prophylactic measures */ if ((rt->rt_flags & (RTF_LLINFO | RTF_HOST)) != RTF_HOST) return; - if ((rt->rt_flags & (RTF_WASCLONED | RTPRF_OURS)) - != RTF_WASCLONED) + if (rt->rt_flags & RTPRF_OURS) + return; + + if (!(rt->rt_flags & (RTF_WASCLONED | RTF_DYNAMIC))) return; /* - * As requested by David Greenman: - * If rtq_reallyold is 0, just delete the route without - * waiting for a timeout cycle to kill it. + * Delete the route immediately if RTF_DELCLONE is set or + * if route caching is disabled (rtq_reallyold set to 0). + * Otherwise, let it expire and be deleted by in6_rtqkill(). */ - if (rtq_reallyold != 0) { - rt->rt_flags |= RTPRF_OURS; - rt->rt_rmx.rmx_expire = time_second + rtq_reallyold; + if ((rt->rt_flags & RTF_DELCLONE) || rtq_reallyold == 0) { + /* + * Delete the route from the radix tree but since we are + * called when the route's reference count is 0, don't + * deallocate it until we return from this routine by + * telling rtrequest that we're interested in it. + * Safe to drop rt_lock and use rt_key, rt_gateway, + * since holding rnh_lock here prevents another thread + * from calling rt_setgate() on this route. + */ + RT_UNLOCK(rt); + if (rtrequest_locked(RTM_DELETE, rt_key(rt), + rt->rt_gateway, rt_mask(rt), rt->rt_flags, &rt) == 0) { + /* Now let the caller free it */ + RT_LOCK(rt); + RT_REMREF_LOCKED(rt); + } else { + RT_LOCK(rt); + } } else { - rtrequest(RTM_DELETE, - (struct sockaddr *)rt_key(rt), - rt->rt_gateway, rt_mask(rt), - rt->rt_flags, 0); + struct timeval timenow; + + getmicrotime(&timenow); + rt->rt_flags |= RTPRF_OURS; + rt->rt_rmx.rmx_expire = + rt_expiry(rt, timenow.tv_sec, rtq_reallyold); } } @@ -291,8 +415,10 @@ struct rtqk_arg { /* * Get rid of old routes. When draining, this deletes everything, even when - * the timeout is not expired yet. When updating, this makes sure that - * nothing has a timeout longer than the current value of rtq_reallyold. + * the timeout is not expired yet. This also applies if the route is dynamic + * and there are sufficiently large number of such routes (more than a half of + * maximum). When updating, this makes sure that nothing has a timeout longer + * than the current value of rtq_reallyold. */ static int in6_rtqkill(struct radix_node *rn, void *rock) @@ -300,33 +426,52 @@ in6_rtqkill(struct radix_node *rn, void *rock) struct rtqk_arg *ap = rock; struct rtentry *rt = (struct rtentry *)rn; int err; + struct timeval timenow; + getmicrotime(&timenow); + lck_mtx_assert(rnh_lock, LCK_MTX_ASSERT_OWNED); + + RT_LOCK(rt); if (rt->rt_flags & RTPRF_OURS) { ap->found++; - if (ap->draining || rt->rt_rmx.rmx_expire <= time_second) { + if (ap->draining || rt->rt_rmx.rmx_expire <= timenow.tv_sec || + ((rt->rt_flags & RTF_DYNAMIC) != 0 && + ip6_maxdynroutes >= 0 && + in6dynroutes > ip6_maxdynroutes / 2)) { if (rt->rt_refcnt > 0) panic("rtqkill route really not free"); - err = rtrequest(RTM_DELETE, - (struct sockaddr *)rt_key(rt), - rt->rt_gateway, rt_mask(rt), - rt->rt_flags, 0); + /* + * Delete this route since we're done with it; + * the route may be freed afterwards, so we + * can no longer refer to 'rt' upon returning + * from rtrequest(). Safe to drop rt_lock and + * use rt_key, rt_gateway, since holding rnh_lock + * here prevents another thread from calling + * rt_setgate() on this route. + */ + RT_UNLOCK(rt); + err = rtrequest_locked(RTM_DELETE, rt_key(rt), + rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0); if (err) { log(LOG_WARNING, "in6_rtqkill: error %d", err); } else { ap->killed++; } } else { - if (ap->updating - && (rt->rt_rmx.rmx_expire - time_second - > rtq_reallyold)) { - rt->rt_rmx.rmx_expire = time_second - + rtq_reallyold; + if (ap->updating && + (unsigned)(rt->rt_rmx.rmx_expire - timenow.tv_sec) > + rt_expiry(rt, 0, rtq_reallyold)) { + rt->rt_rmx.rmx_expire = rt_expiry(rt, + timenow.tv_sec, rtq_reallyold); } ap->nextstop = lmin(ap->nextstop, rt->rt_rmx.rmx_expire); + RT_UNLOCK(rt); } + } else { + RT_UNLOCK(rt); } return 0; @@ -335,19 +480,6 @@ in6_rtqkill(struct radix_node *rn, void *rock) #define RTQ_TIMEOUT 60*10 /* run no less than once every ten minutes */ static int rtq_timeout = RTQ_TIMEOUT; -static void -in6_rtqtimo_funneled(void *rock) -{ -#ifdef __APPLE__ - boolean_t funnel_state; - funnel_state = thread_funnel_set(network_flock, TRUE); - in6_rtqtimo(rock); -#endif -#ifdef __APPLE__ - (void) thread_funnel_set(network_flock, FALSE); -#endif -} - static void in6_rtqtimo(void *rock) { @@ -355,15 +487,17 @@ in6_rtqtimo(void *rock) struct rtqk_arg arg; struct timeval atv; static time_t last_adjusted_timeout = 0; - int s; + struct timeval timenow; + + lck_mtx_lock(rnh_lock); + /* Get the timestamp after we acquire the lock for better accuracy */ + getmicrotime(&timenow); arg.found = arg.killed = 0; arg.rnh = rnh; - arg.nextstop = time_second + rtq_timeout; + arg.nextstop = timenow.tv_sec + rtq_timeout; arg.draining = arg.updating = 0; - s = splnet(); rnh->rnh_walktree(rnh, in6_rtqkill, &arg); - splx(s); /* * Attempt to be somewhat dynamic about this: @@ -374,28 +508,27 @@ in6_rtqtimo(void *rock) * hard. */ if ((arg.found - arg.killed > rtq_toomany) - && (time_second - last_adjusted_timeout >= rtq_timeout) + && (timenow.tv_sec - last_adjusted_timeout >= rtq_timeout) && rtq_reallyold > rtq_minreallyold) { rtq_reallyold = 2*rtq_reallyold / 3; if (rtq_reallyold < rtq_minreallyold) { rtq_reallyold = rtq_minreallyold; } - last_adjusted_timeout = time_second; + last_adjusted_timeout = timenow.tv_sec; #if DIAGNOSTIC log(LOG_DEBUG, "in6_rtqtimo: adjusted rtq_reallyold to %d", rtq_reallyold); #endif arg.found = arg.killed = 0; arg.updating = 1; - s = splnet(); rnh->rnh_walktree(rnh, in6_rtqkill, &arg); - splx(s); } atv.tv_usec = 0; - atv.tv_sec = arg.nextstop - time_second; - timeout(in6_rtqtimo_funneled, rock, tvtohz(&atv)); + atv.tv_sec = arg.nextstop - timenow.tv_sec; + lck_mtx_unlock(rnh_lock); + timeout(in6_rtqtimo, rock, tvtohz(&atv)); } /* @@ -411,81 +544,72 @@ in6_mtuexpire(struct radix_node *rn, void *rock) { struct rtentry *rt = (struct rtentry *)rn; struct mtuex_arg *ap = rock; + struct timeval timenow; + + getmicrotime(&timenow); /* sanity */ if (!rt) panic("rt == NULL in in6_mtuexpire"); + RT_LOCK(rt); if (rt->rt_rmx.rmx_expire && !(rt->rt_flags & RTF_PROBEMTU)) { - if (rt->rt_rmx.rmx_expire <= time_second) { + if (rt->rt_rmx.rmx_expire <= timenow.tv_sec) { rt->rt_flags |= RTF_PROBEMTU; } else { ap->nextstop = lmin(ap->nextstop, rt->rt_rmx.rmx_expire); } } + RT_UNLOCK(rt); return 0; } #define MTUTIMO_DEFAULT (60*1) -static void -in6_mtutimo_funneled(void *rock) -{ -#ifdef __APPLE__ - boolean_t funnel_state; - funnel_state = thread_funnel_set(network_flock, TRUE); - in6_mtutimo(rock); -#endif -#ifdef __APPLE__ - (void) thread_funnel_set(network_flock, FALSE); -#endif -} - static void in6_mtutimo(void *rock) { struct radix_node_head *rnh = rock; struct mtuex_arg arg; struct timeval atv; - int s; + struct timeval timenow; + + getmicrotime(&timenow); arg.rnh = rnh; - arg.nextstop = time_second + MTUTIMO_DEFAULT; - s = splnet(); + arg.nextstop = timenow.tv_sec + MTUTIMO_DEFAULT; + lck_mtx_lock(rnh_lock); rnh->rnh_walktree(rnh, in6_mtuexpire, &arg); - splx(s); atv.tv_usec = 0; atv.tv_sec = arg.nextstop; - if (atv.tv_sec < time_second) { + if (atv.tv_sec < timenow.tv_sec) { #if DIAGNOSTIC log(LOG_DEBUG, "IPv6: invalid mtu expiration time on routing table\n"); #endif - arg.nextstop = time_second + 30; /*last resort*/ + arg.nextstop = timenow.tv_sec + 30; /*last resort*/ } - atv.tv_sec -= time_second; - timeout(in6_mtutimo_funneled, rock, tvtohz(&atv)); + atv.tv_sec -= timenow.tv_sec; + lck_mtx_unlock(rnh_lock); + timeout(in6_mtutimo, rock, tvtohz(&atv)); } -#if 0 void in6_rtqdrain() { struct radix_node_head *rnh = rt_tables[AF_INET6]; struct rtqk_arg arg; - int s; arg.found = arg.killed = 0; arg.rnh = rnh; arg.nextstop = 0; arg.draining = 1; arg.updating = 0; - s = splnet(); + lck_mtx_lock(rnh_lock); rnh->rnh_walktree(rnh, in6_rtqkill, &arg); - splx(s); + lck_mtx_unlock(rnh_lock); } -#endif /* * Initialize our routing tree. @@ -503,7 +627,9 @@ in6_inithead(void **head, int off) rnh = *head; rnh->rnh_addaddr = in6_addroute; + rnh->rnh_deladdr = in6_deleteroute; rnh->rnh_matchaddr = in6_matroute; + rnh->rnh_matchaddr_args = in6_matroute_args; rnh->rnh_close = in6_clsroute; in6_rtqtimo(rnh); /* kick off timeout first time */ in6_mtutimo(rnh); /* kick off timeout first time */