diff options
| author | qingli <qingli@FreeBSD.org> | 2008-12-15 06:10:57 +0000 |
|---|---|---|
| committer | qingli <qingli@FreeBSD.org> | 2008-12-15 06:10:57 +0000 |
| commit | ec826ad5c7f97de814529d3b3bae7950f91d9a5d (patch) | |
| tree | 281ff6a89cacadf7e72f506b037ca41229a23bf6 | |
| parent | 664c3aeb0118ccccb068f485e30353a47923b4d0 (diff) | |
| download | FreeBSD-src-ec826ad5c7f97de814529d3b3bae7950f91d9a5d.zip FreeBSD-src-ec826ad5c7f97de814529d3b3bae7950f91d9a5d.tar.gz | |
This main goals of this project are:
1. separating L2 tables (ARP, NDP) from the L3 routing tables
2. removing as much locking dependencies among these layers as
possible to allow for some parallelism in the search operations
3. simplify the logic in the routing code,
The most notable end result is the obsolescent of the route
cloning (RTF_CLONING) concept, which translated into code reduction
in both IPv4 ARP and IPv6 NDP related modules, and size reduction in
struct rtentry{}. The change in design obsoletes the semantics of
RTF_CLONING, RTF_WASCLONE and RTF_LLINFO routing flags. The userland
applications such as "arp" and "ndp" have been modified to reflect
those changes. The output from "netstat -r" shows only the routing
entries.
Quite a few developers have contributed to this project in the
past: Glebius Smirnoff, Luigi Rizzo, Alessandro Cerri, and
Andre Oppermann. And most recently:
- Kip Macy revised the locking code completely, thus completing
the last piece of the puzzle, Kip has also been conducting
active functional testing
- Sam Leffler has helped me improving/refactoring the code, and
provided valuable reviews
- Julian Elischer setup the perforce tree for me and has helped
me maintaining that branch before the svn conversion
66 files changed, 1636 insertions, 2471 deletions
@@ -22,6 +22,15 @@ NOTE TO PEOPLE WHO THINK THAT FreeBSD 8.x IS SLOW: to maximize performance. (To disable malloc debugging, run ln -s aj /etc/malloc.conf.) +20081214: + __FreeBSD_version 800059 incorporates the new arp-v2 rewrite. + RTF_CLONING, RTF_LLINFO and RTF_WASCLONED flags are eliminated. + The new code reduced struct rtentry{} by 16 bytes on 32-bit + architecture and 40 bytes on 64-bit architecture. The userland + applications "arp" and "ndp" have been updated accordingly. + The output from "netstat -r" shows only routing entries and + none of the L2 information. + 20081130: __FreeBSD_version 800057 marks the switchover from the binary ath hal to source code. Users must add the line: diff --git a/contrib/bsnmp/snmp_mibII/mibII.c b/contrib/bsnmp/snmp_mibII/mibII.c index 96bd366..83e2317 100644 --- a/contrib/bsnmp/snmp_mibII/mibII.c +++ b/contrib/bsnmp/snmp_mibII/mibII.c @@ -48,8 +48,7 @@ static void *route_fd; /* if-index allocator */ static uint32_t next_if_index = 1; -/* re-fetch arp table */ -static int update_arp; +/* currently fetching the arp table */ static int in_update_arp; /* OR registrations */ @@ -911,36 +910,6 @@ mib_find_ifa(struct in_addr addr) } /* - * Process a new ARP entry - */ -static void -process_arp(const struct rt_msghdr *rtm, const struct sockaddr_dl *sdl, - const struct sockaddr_in *sa) -{ - struct mibif *ifp; - struct mibarp *at; - - /* IP arp table entry */ - if (sdl->sdl_alen == 0) { - update_arp = 1; - return; - } - if ((ifp = mib_find_if_sys(sdl->sdl_index)) == NULL) - return; - /* have a valid entry */ - if ((at = mib_find_arp(ifp, sa->sin_addr)) == NULL && - (at = mib_arp_create(ifp, sa->sin_addr, - sdl->sdl_data + sdl->sdl_nlen, sdl->sdl_alen)) == NULL) - return; - - if (rtm->rtm_rmx.rmx_expire == 0) - at->flags |= MIBARP_PERM; - else - at->flags &= ~MIBARP_PERM; - at->flags |= MIBARP_FOUND; -} - -/* * Handle a routing socket message. */ static void @@ -1080,46 +1049,12 @@ handle_rtmsg(struct rt_msghdr *rtm) } break; #endif - case RTM_GET: - mib_extract_addrs(rtm->rtm_addrs, (u_char *)(rtm + 1), addrs); - if (rtm->rtm_flags & RTF_LLINFO) { - if (addrs[RTAX_DST] == NULL || - addrs[RTAX_GATEWAY] == NULL || - addrs[RTAX_DST]->sa_family != AF_INET || - addrs[RTAX_GATEWAY]->sa_family != AF_LINK) - break; - process_arp(rtm, - (struct sockaddr_dl *)(void *)addrs[RTAX_GATEWAY], - (struct sockaddr_in *)(void *)addrs[RTAX_DST]); - } else { - if (rtm->rtm_errno == 0 && (rtm->rtm_flags & RTF_UP)) - mib_sroute_process(rtm, addrs[RTAX_GATEWAY], - addrs[RTAX_DST], addrs[RTAX_NETMASK]); - } - break; - case RTM_ADD: - mib_extract_addrs(rtm->rtm_addrs, (u_char *)(rtm + 1), addrs); - if (rtm->rtm_flags & RTF_LLINFO) { - if (addrs[RTAX_DST] == NULL || - addrs[RTAX_GATEWAY] == NULL || - addrs[RTAX_DST]->sa_family != AF_INET || - addrs[RTAX_GATEWAY]->sa_family != AF_LINK) - break; - process_arp(rtm, - (struct sockaddr_dl *)(void *)addrs[RTAX_GATEWAY], - (struct sockaddr_in *)(void *)addrs[RTAX_DST]); - } else { - if (rtm->rtm_errno == 0 && (rtm->rtm_flags & RTF_UP)) - mib_sroute_process(rtm, addrs[RTAX_GATEWAY], - addrs[RTAX_DST], addrs[RTAX_NETMASK]); - } - break; - case RTM_DELETE: mib_extract_addrs(rtm->rtm_addrs, (u_char *)(rtm + 1), addrs); - if (rtm->rtm_errno == 0 && !(rtm->rtm_flags & RTF_LLINFO)) + + if (rtm->rtm_errno == 0 && (rtm->rtm_flags & RTF_UP)) mib_sroute_process(rtm, addrs[RTAX_GATEWAY], addrs[RTAX_DST], addrs[RTAX_NETMASK]); break; @@ -1289,7 +1224,8 @@ update_ifa_info(void) /* * Update arp table - */ + * +*/ void mib_arp_update(void) { @@ -1305,11 +1241,11 @@ mib_arp_update(void) TAILQ_FOREACH(at, &mibarp_list, link) at->flags &= ~MIBARP_FOUND; - if ((buf = mib_fetch_rtab(AF_INET, NET_RT_FLAGS, RTF_LLINFO, &needed)) == NULL) { + if ((buf = mib_fetch_rtab(AF_INET, NET_RT_FLAGS, 0, &needed)) == NULL) { in_update_arp = 0; return; } - + next = buf; while (next < buf + needed) { rtm = (struct rt_msghdr *)(void *)next; @@ -1326,7 +1262,6 @@ mib_arp_update(void) at = at1; } mibarpticks = get_ticks(); - update_arp = 0; in_update_arp = 0; } @@ -1634,8 +1569,8 @@ mibII_idle(void) mib_arp_update(); mib_iflist_bad = 0; } - if (update_arp) - mib_arp_update(); + + mib_arp_update(); } diff --git a/contrib/bsnmp/snmp_mibII/mibII_route.c b/contrib/bsnmp/snmp_mibII/mibII_route.c index e17497d..b11c422 100644 --- a/contrib/bsnmp/snmp_mibII/mibII_route.c +++ b/contrib/bsnmp/snmp_mibII/mibII_route.c @@ -186,8 +186,7 @@ mib_sroute_process(struct rt_msghdr *rtm, struct sockaddr *gw, memcpy(r->index, key.index, sizeof(r->index)); r->ifindex = (ifp == NULL) ? 0 : ifp->index; - r->type = (rtm->rtm_flags & RTF_LLINFO) ? 3 : - (rtm->rtm_flags & RTF_REJECT) ? 2 : 4; + r->type = (rtm->rtm_flags & RTF_REJECT) ? 2 : 4; /* cannot really know, what protocol it runs */ r->proto = (rtm->rtm_flags & RTF_LOCAL) ? 2 : diff --git a/contrib/ipfilter/ipsend/44arp.c b/contrib/ipfilter/ipsend/44arp.c index 6ee4f1b..5c07fb6 100644 --- a/contrib/ipfilter/ipsend/44arp.c +++ b/contrib/ipfilter/ipsend/44arp.c @@ -88,7 +88,12 @@ char *addr, *eaddr; mib[2] = 0; mib[3] = AF_INET; mib[4] = NET_RT_FLAGS; +#ifdef RTF_LLINFO mib[5] = RTF_LLINFO; +#else + mib[5] = 0; +#endif + if (sysctl(mib, 6, NULL, &needed, NULL, 0) == -1) { perror("route-sysctl-estimate"); diff --git a/lib/libstand/if_ether.h b/lib/libstand/if_ether.h index 7783c41..373f95a 100644 --- a/lib/libstand/if_ether.h +++ b/lib/libstand/if_ether.h @@ -156,7 +156,7 @@ struct ifqueue arpintrq; void arpwhohas(struct arpcom *, struct in_addr *); void arpintr(void); int arpresolve(struct arpcom *, - struct rtentry *, struct mbuf *, struct sockaddr *, u_char *); + struct rtentry *, struct mbuf *, struct sockaddr *, u_char *, struct llentry **); void arp_ifinit(struct arpcom *, struct ifaddr *); void arp_rtrequest(int, struct rtentry *, struct sockaddr *); @@ -233,7 +233,7 @@ struct ether_multistep { #ifdef _KERNEL void arp_rtrequest(int, struct rtentry *, struct sockaddr *); int arpresolve(struct arpcom *, struct rtentry *, struct mbuf *, - struct sockaddr *, u_char *); + struct sockaddr *, u_char *, struct llentry **); void arpintr(void); int arpioctl(u_long, caddr_t); void arp_ifinit(struct arpcom *, struct ifaddr *); diff --git a/libexec/bootpd/rtmsg.c b/libexec/bootpd/rtmsg.c index 4584569..071a7d1 100644 --- a/libexec/bootpd/rtmsg.c +++ b/libexec/bootpd/rtmsg.c @@ -152,7 +152,6 @@ tryagain: sdl = (struct sockaddr_dl *)(sin->sin_len + (char *)sin); if (sin->sin_addr.s_addr == sin_m.sin_addr.s_addr) { if (sdl->sdl_family == AF_LINK && - (rtm->rtm_flags & RTF_LLINFO) && !(rtm->rtm_flags & RTF_GATEWAY)) switch (sdl->sdl_type) { case IFT_ETHER: case IFT_FDDI: case IFT_ISO88023: case IFT_ISO88024: case IFT_ISO88025: diff --git a/release/picobsd/tinyware/ns/ns.c b/release/picobsd/tinyware/ns/ns.c index c6e58e2..2ada2a7 100644 --- a/release/picobsd/tinyware/ns/ns.c +++ b/release/picobsd/tinyware/ns/ns.c @@ -382,10 +382,6 @@ print_routing(char *proto) rtm = (struct rt_msghdr *)next; sa = (struct sockaddr *)(rtm + 1); get_rtaddrs(rtm->rtm_addrs, sa, rti_info); - if (rtm->rtm_flags & RTF_WASCLONED) { - if ((rtm->rtm_flags & RTF_LLINFO) == 0) - continue; - } if ((sa = rti_info[RTAX_DST]) != NULL) { sprintf(fbuf, "%s", sock_ntop(sa, sa->sa_len)); if (((sa1 = rti_info[RTAX_NETMASK]) != NULL) diff --git a/sbin/route/route.c b/sbin/route/route.c index 082ac53..bae9983 100644 --- a/sbin/route/route.c +++ b/sbin/route/route.c @@ -605,9 +605,6 @@ newroute(argc, argv) case K_NOSTATIC: flags &= ~RTF_STATIC; break; - case K_LLINFO: - flags |= RTF_LLINFO; - break; case K_LOCK: locking = 1; break; @@ -632,9 +629,6 @@ newroute(argc, argv) case K_PROXY: proxy = 1; break; - case K_CLONING: - flags |= RTF_CLONING; - break; case K_XRESOLVE: flags |= RTF_XRESOLVE; break; diff --git a/sbin/routed/table.c b/sbin/routed/table.c index 491c5cf..9fd4cb8 100644 --- a/sbin/routed/table.c +++ b/sbin/routed/table.c @@ -1103,12 +1103,13 @@ flush_kern(void) || INFO_DST(&info)->sa_family != AF_INET) continue; +#if defined (RTF_LLINFO) /* ignore ARP table entries on systems with a merged route * and ARP table. */ if (rtm->rtm_flags & RTF_LLINFO) continue; - +#endif #if defined(RTF_WASCLONED) && defined(__FreeBSD__) /* ignore cloned routes */ @@ -1261,11 +1262,13 @@ read_rt(void) continue; } +#if defined(RTF_LLINFO) if (m.r.rtm.rtm_flags & RTF_LLINFO) { trace_act("ignore ARP %s", str); continue; } - +#endif + #if defined(RTF_WASCLONED) && defined(__FreeBSD__) if (m.r.rtm.rtm_flags & RTF_WASCLONED) { trace_act("ignore cloned %s", str); diff --git a/share/man/man4/route.4 b/share/man/man4/route.4 index 0e10be5..aaf59e4 100644 --- a/share/man/man4/route.4 +++ b/share/man/man4/route.4 @@ -196,7 +196,7 @@ Messages include: #define RTM_REDIRECT 0x6 /* Told to use different route */ #define RTM_MISS 0x7 /* Lookup failed on this address */ #define RTM_LOCK 0x8 /* fix specified metrics */ -#define RTM_RESOLVE 0xb /* request to resolve dst to LL addr */ +#define RTM_RESOLVE 0xb /* request to resolve dst to LL addr - unused */ #define RTM_NEWADDR 0xc /* address being added to iface */ #define RTM_DELADDR 0xd /* address being removed from iface */ #define RTM_IFINFO 0xe /* iface going up/down etc. */ @@ -308,7 +308,7 @@ Specifiers for which addresses are present in the messages are: #define RTA_DST 0x1 /* destination sockaddr present */ #define RTA_GATEWAY 0x2 /* gateway sockaddr present */ #define RTA_NETMASK 0x4 /* netmask sockaddr present */ -#define RTA_GENMASK 0x8 /* cloning mask sockaddr present */ +#define RTA_GENMASK 0x8 /* cloning mask sockaddr present - unused */ #define RTA_IFP 0x10 /* interface name sockaddr present */ #define RTA_IFA 0x20 /* interface addr sockaddr present */ #define RTA_AUTHOR 0x40 /* sockaddr for author of redirect */ diff --git a/share/man/man9/rtalloc.9 b/share/man/man9/rtalloc.9 index 4e7f818..cbd2c5e 100644 --- a/share/man/man9/rtalloc.9 +++ b/share/man/man9/rtalloc.9 @@ -27,7 +27,8 @@ .\" SUCH DAMAGE. .\" .\" $FreeBSD$ -.Dd October 11, 2004 +.\" +.Dd December 11, 2008 .Os .Dt RTALLOC 9 .Sh NAME @@ -64,21 +65,6 @@ certain protocol\- and interface-specific actions to take place. .\" XXX - -mdoc should contain a standard request for getting em and .\" en dashes. .Pp -When a route with the flag -.Dv RTF_CLONING -is retrieved, and the action of this flag is not masked, the -.Nm -facility automatically generates a new route using information in the -old route as a template, and -sends an -.Dv RTM_RESOLVE -message to the appropriate interface-address route-management routine -.Pq Fn ifa->ifa_rtrequest . -This generated route is called -.Em cloned , -and has -.Dv RTF_WASCLONED -flag set. .Dv RTF_PRCLONING flag is obsolete and thus ignored by facility. If the @@ -123,22 +109,19 @@ field. .Pp The .Fn rtalloc_ign -interface can be used when the default actions of -.Fn rtalloc -in the presence of the -.Dv RTF_CLONING -flag is undesired. +interface can be used when the caller does not want to receive +the returned +.Fa rtentry +locked. The .Fa ro argument is the same as .Fn rtalloc , but there is additionally a .Fa flags -argument, which lists the flags in the route which are to be -.Em ignored -(in most cases this is -.Dv RTF_CLONING -flag). +argument, which is now only used to pass +.Dv RTF_RNH_LOCKED +indicating that the radix tree lock is already held. Both .Fn rtalloc and @@ -163,16 +146,7 @@ directly as the argument. The second argument, .Fa report , -controls whether -.Dv RTM_RESOLVE -requests are sent to the lower layers when an -.Dv RTF_CLONING -or -.Dv RTF_PRCLONING -route is cloned. -Ordinarily a value of one should be passed, except -in the processing of those lower layers which use the cloning -facility. +controls whether the lower layers are notified when a lookup fails. The third argument, .Fa flags , is a set of flags to ignore, as in diff --git a/share/man/man9/rtentry.9 b/share/man/man9/rtentry.9 index a1c48a6..0be14f1 100644 --- a/share/man/man9/rtentry.9 +++ b/share/man/man9/rtentry.9 @@ -28,7 +28,7 @@ .\" .\" $FreeBSD$ .\" -.Dd October 7, 2004 +.Dd December 11, 2008 .Os .Dt RTENTRY 9 .Sh NAME @@ -76,8 +76,11 @@ right (some protocols will put a link-layer address here), or some intermediate stop on the way to that destination (if the .Dv RTF_GATEWAY flag is set). -.It Vt "u_long rt_flags" ; +.It Vt "int rt_flags" ; See below. +.It Vt "int rt_refcnt" ; +Route entries are reference-counted; this field indicates the number +of external (to the radix tree) references. .It Vt "struct ifnet *rt_ifp" ; .It Vt "struct ifaddr *rt_ifa" ; These two fields represent the @@ -88,48 +91,17 @@ packet to the destination or set of destinations which this route represents. .It Vt "struct rt_metrics_lite rt_rmx" ; See below. -.It Vt "long rt_refcnt" ; -Route entries are reference-counted; this field indicates the number -of external (to the radix tree) references. -If the .Dv RTF_UP flag is not present, the .Fn rtfree function will delete the route from the radix tree when the last reference drops. -.It Vt "struct sockaddr *rt_genmask" ; -When the -.Fn rtalloc -family of functions performs a cloning operation as requested by the -.Dv RTF_CLONING -flag, this field is used as the mask for the new route which is -inserted into the table. -If this field is a null pointer, then a host -route is generated. -.It Vt "caddr_t rt_llinfo" ; -When the -.Dv RTF_LLINFO -flag is set, this field contains information specific to the link -layer represented by the named interface address. -(It is normally managed by the -.Va rt_ifa->ifa_rtrequest -routine.) -Protocols such as -.Xr arp 4 -use this field to reference per-destination state internal to that -protocol. .It Vt "struct rtentry *rt_gwroute" ; This member is a reference to a route whose destination is .Va rt_gateway . It is only used for .Dv RTF_GATEWAY routes. -.It Vt "struct rtentry *rt_parent" ; -A reference to the route from which this route was cloned, or a null -pointer if this route was not generated by cloning. -See also the -.Dv RTF_WASCLONED -flag. .It Vt "struct mtx rt_mtx" ; Mutex to lock this routing entry. .El @@ -162,23 +134,12 @@ This route was modified by Used only in the .Xr route 4 protocol, indicating that the request was executed. -.It Dv RTF_CLONING -When this route is returned as a result of a lookup, automatically -create a new route using this one as a template and -.Va rt_genmask -(if present) as a mask. .It Dv RTF_XRESOLVE When this route is returned as a result of a lookup, send a report on the .Xr route 4 interface requesting that an external process perform resolution for this route. -(Used in conjunction with -.Dv RTF_CLONING . ) -.It Dv RTF_LLINFO -Indicates that this route represents information being managed by a -link layer's adaptation layer (e.g., -.Tn ARP ) . .It Dv RTF_STATIC Indicates that this route was manually added by means of the .Xr route 8 @@ -191,14 +152,6 @@ Requests that output sent via this route be discarded. Protocol-specific. .It Dv RTF_PRCLONING This flag is obsolete and simply ignored by facility. -.It Dv RTF_WASCLONED -Indicates that this route was generated as a result of cloning -requested by the -.Dv RTF_CLONING -flag. -When set, the -.Va rt_parent -field indicates the route from which this one was generated. .It Dv RTF_PINNED (Reserved for future use to indicate routes which are not to be modified by a routing protocol.) @@ -296,8 +249,3 @@ The and .Va rmx_filler fields could be named better. -.Pp -There is some disagreement over whether it is legitimate for -.Dv RTF_LLINFO -to be set by any process other than -.Va rt_ifa->ifa_rtrequest . diff --git a/sys/conf/NOTES b/sys/conf/NOTES index 4c66d14..6825dd4 100644 --- a/sys/conf/NOTES +++ b/sys/conf/NOTES @@ -1958,7 +1958,7 @@ device fatm #Fore PCA200E device hatm #Fore/Marconi HE155/622 device patm #IDT77252 cards (ProATM and IDT) device utopia #ATM PHY driver -options NATM #native ATM +#options NATM #native ATM options LIBMBPOOL #needed by patm, iatm diff --git a/sys/conf/files b/sys/conf/files index d09ead4..633b411 100644 --- a/sys/conf/files +++ b/sys/conf/files @@ -2172,6 +2172,7 @@ net/if_gre.c optional gre net/if_iso88025subr.c optional token net/if_lagg.c optional lagg net/if_loop.c optional loop +net/if_llatbl.c standard net/if_media.c standard net/if_mib.c standard net/if_ppp.c optional ppp diff --git a/sys/contrib/pf/net/pf.c b/sys/contrib/pf/net/pf.c index 2959830..d69372b 100644 --- a/sys/contrib/pf/net/pf.c +++ b/sys/contrib/pf/net/pf.c @@ -3162,7 +3162,7 @@ pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer) #ifdef RTF_PRCLONING rtalloc_ign(&ro, (RTF_CLONING | RTF_PRCLONING)); #else /* !RTF_PRCLONING */ - in_rtalloc_ign(&ro, RTF_CLONING, 0); + in_rtalloc_ign(&ro, 0, 0); #endif #else /* ! __FreeBSD__ */ rtalloc_noclone(&ro, NO_CLONING); @@ -3183,7 +3183,7 @@ pf_calc_mss(struct pf_addr *addr, sa_family_t af, u_int16_t offer) rtalloc_ign((struct route *)&ro6, (RTF_CLONING | RTF_PRCLONING)); #else /* !RTF_PRCLONING */ - rtalloc_ign((struct route *)&ro6, RTF_CLONING); + rtalloc_ign((struct route *)&ro6, 0); #endif #else /* ! __FreeBSD__ */ rtalloc_noclone((struct route *)&ro6, NO_CLONING); @@ -5986,9 +5986,9 @@ pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kif *kif) #ifdef __FreeBSD__ /* XXX MRT not always INET */ /* stick with table 0 though */ if (af == AF_INET) - in_rtalloc_ign((struct route *)&ro, RTF_CLONING, 0); + in_rtalloc_ign((struct route *)&ro, 0, 0); else - rtalloc_ign((struct route *)&ro, RTF_CLONING); + rtalloc_ign((struct route *)&ro, 0); #else /* ! __FreeBSD__ */ rtalloc_noclone((struct route *)&ro, NO_CLONING); #endif @@ -6068,9 +6068,9 @@ pf_rtlabel_match(struct pf_addr *addr, sa_family_t af, struct pf_addr_wrap *aw) rtalloc_ign((struct route *)&ro, (RTF_CLONING|RTF_PRCLONING)); # else /* !RTF_PRCLONING */ if (af == AF_INET) - in_rtalloc_ign((struct route *)&ro, RTF_CLONING, 0); + in_rtalloc_ign((struct route *)&ro, 0, 0); else - rtalloc_ign((struct route *)&ro, RTF_CLONING); + rtalloc_ign((struct route *)&ro, 0); # endif #else /* ! __FreeBSD__ */ rtalloc_noclone((struct route *)&ro, NO_CLONING); diff --git a/sys/contrib/rdma/rdma_addr.c b/sys/contrib/rdma/rdma_addr.c index e052b80..971b4de 100644 --- a/sys/contrib/rdma/rdma_addr.c +++ b/sys/contrib/rdma/rdma_addr.c @@ -163,6 +163,7 @@ static void addr_send_arp(struct sockaddr_in *dst_in) struct route iproute; struct sockaddr_in *dst = (struct sockaddr_in *)&iproute.ro_dst; char dmac[ETHER_ADDR_LEN]; + struct llentry *lle; bzero(&iproute, sizeof iproute); *dst = *dst_in; @@ -172,7 +173,7 @@ static void addr_send_arp(struct sockaddr_in *dst_in) return; arpresolve(iproute.ro_rt->rt_ifp, iproute.ro_rt, NULL, - rt_key(iproute.ro_rt), dmac); + rt_key(iproute.ro_rt), dmac, &lle); RTFREE(iproute.ro_rt); } @@ -186,6 +187,7 @@ static int addr_resolve_remote(struct sockaddr_in *src_in, struct route iproute; struct sockaddr_in *dst = (struct sockaddr_in *)&iproute.ro_dst; char dmac[ETHER_ADDR_LEN]; + struct llentry *lle; bzero(&iproute, sizeof iproute); *dst = *dst_in; @@ -202,7 +204,7 @@ static int addr_resolve_remote(struct sockaddr_in *src_in, goto put; } ret = arpresolve(iproute.ro_rt->rt_ifp, iproute.ro_rt, NULL, - rt_key(iproute.ro_rt), dmac); + rt_key(iproute.ro_rt), dmac, &lle); if (ret) { goto put; } diff --git a/sys/dev/cxgb/ulp/tom/cxgb_l2t.c b/sys/dev/cxgb/ulp/tom/cxgb_l2t.c index 67856e6..2484923 100644 --- a/sys/dev/cxgb/ulp/tom/cxgb_l2t.c +++ b/sys/dev/cxgb/ulp/tom/cxgb_l2t.c @@ -93,15 +93,15 @@ arp_hash(u32 key, int ifindex, const struct l2t_data *d) } static inline void -neigh_replace(struct l2t_entry *e, struct rtentry *rt) +neigh_replace(struct l2t_entry *e, struct llentry *neigh) { - RT_LOCK(rt); - RT_ADDREF(rt); - RT_UNLOCK(rt); + LLE_WLOCK(neigh); + LLE_ADDREF(neigh); + LLE_WUNLOCK(neigh); if (e->neigh) - RTFREE(e->neigh); - e->neigh = rt; + LLE_FREE(e->neigh); + e->neigh = neigh; } /* @@ -164,7 +164,7 @@ arpq_enqueue(struct l2t_entry *e, struct mbuf *m) int t3_l2t_send_slow(struct t3cdev *dev, struct mbuf *m, struct l2t_entry *e) { - struct rtentry *rt = e->neigh; + struct llentry *lle = e->neigh; struct sockaddr_in sin; bzero(&sin, sizeof(struct sockaddr_in)); @@ -177,7 +177,7 @@ again: switch (e->state) { case L2T_STATE_STALE: /* entry is stale, kick off revalidation */ arpresolve(rt->rt_ifp, rt, NULL, - (struct sockaddr *)&sin, e->dmac); + (struct sockaddr *)&sin, e->dmac, &lle); mtx_lock(&e->lock); if (e->state == L2T_STATE_STALE) e->state = L2T_STATE_VALID; @@ -201,7 +201,7 @@ again: * entries when there's no memory. */ if (arpresolve(rt->rt_ifp, rt, NULL, - (struct sockaddr *)&sin, e->dmac) == 0) { + (struct sockaddr *)&sin, e->dmac, &lle) == 0) { CTR6(KTR_CXGB, "mac=%x:%x:%x:%x:%x:%x\n", e->dmac[0], e->dmac[1], e->dmac[2], e->dmac[3], e->dmac[4], e->dmac[5]); @@ -222,12 +222,12 @@ again: void t3_l2t_send_event(struct t3cdev *dev, struct l2t_entry *e) { - struct rtentry *rt; struct mbuf *m0; struct sockaddr_in sin; sin.sin_family = AF_INET; sin.sin_len = sizeof(struct sockaddr_in); sin.sin_addr.s_addr = e->addr; + struct llentry *lle; if ((m0 = m_gethdr(M_NOWAIT, MT_DATA)) == NULL) return; @@ -237,7 +237,7 @@ again: switch (e->state) { case L2T_STATE_STALE: /* entry is stale, kick off revalidation */ arpresolve(rt->rt_ifp, rt, NULL, - (struct sockaddr *)&sin, e->dmac); + (struct sockaddr *)&sin, e->dmac, &lle); mtx_lock(&e->lock); if (e->state == L2T_STATE_STALE) { e->state = L2T_STATE_VALID; @@ -263,7 +263,7 @@ again: * entries when there's no memory. */ arpresolve(rt->rt_ifp, rt, NULL, - (struct sockaddr *)&sin, e->dmac); + (struct sockaddr *)&sin, e->dmac, &lle); } return; @@ -321,18 +321,18 @@ found: void t3_l2e_free(struct l2t_data *d, struct l2t_entry *e) { - struct rtentry *rt = NULL; - + struct llentry *lle; + mtx_lock(&e->lock); if (atomic_load_acq_int(&e->refcnt) == 0) { /* hasn't been recycled */ - rt = e->neigh; + lle = e->neigh; e->neigh = NULL; } mtx_unlock(&e->lock); atomic_add_int(&d->nfree, 1); - if (rt) - RTFREE(rt); + if (lle) + LLE_FREE(lle); } @@ -341,11 +341,8 @@ t3_l2e_free(struct l2t_data *d, struct l2t_entry *e) * Must be called with softirqs disabled. */ static inline void -reuse_entry(struct l2t_entry *e, struct rtentry *neigh) +reuse_entry(struct l2t_entry *e, struct llentry *neigh) { - struct llinfo_arp *la; - - la = (struct llinfo_arp *)neigh->rt_llinfo; mtx_lock(&e->lock); /* avoid race with t3_l2t_free */ if (neigh != e->neigh) @@ -362,13 +359,13 @@ reuse_entry(struct l2t_entry *e, struct rtentry *neigh) } struct l2t_entry * -t3_l2t_get(struct t3cdev *dev, struct rtentry *neigh, struct ifnet *ifp, +t3_l2t_get(struct t3cdev *dev, struct llentry *neigh, struct ifnet *ifp, struct sockaddr *sa) { struct l2t_entry *e; struct l2t_data *d = L2DATA(dev); u32 addr = ((struct sockaddr_in *)sa)->sin_addr.s_addr; - int ifidx = neigh->rt_ifp->if_index; + int ifidx = ifp->if_index; int hash = arp_hash(addr, ifidx, d); unsigned int smt_idx = ((struct port_info *)ifp->if_softc)->port_id; @@ -448,20 +445,19 @@ handle_failed_resolution(struct t3cdev *dev, struct mbuf *arpq) } void -t3_l2t_update(struct t3cdev *dev, struct rtentry *neigh, +t3_l2t_update(struct t3cdev *dev, struct llentry *neigh, uint8_t *enaddr, struct sockaddr *sa) { struct l2t_entry *e; struct mbuf *arpq = NULL; struct l2t_data *d = L2DATA(dev); u32 addr = *(u32 *) &((struct sockaddr_in *)sa)->sin_addr; - int ifidx = neigh->rt_ifp->if_index; int hash = arp_hash(addr, ifidx, d); struct llinfo_arp *la; rw_rlock(&d->lock); for (e = d->l2tab[hash].first; e; e = e->next) - if (e->addr == addr && e->ifindex == ifidx) { + if (e->addr == addr) { mtx_lock(&e->lock); goto found; } diff --git a/sys/dev/cxgb/ulp/tom/cxgb_l2t.h b/sys/dev/cxgb/ulp/tom/cxgb_l2t.h index 1d547d3..308ba66 100644 --- a/sys/dev/cxgb/ulp/tom/cxgb_l2t.h +++ b/sys/dev/cxgb/ulp/tom/cxgb_l2t.h @@ -68,7 +68,7 @@ struct l2t_entry { int ifindex; /* neighbor's net_device's ifindex */ uint16_t smt_idx; /* SMT index */ uint16_t vlan; /* VLAN TCI (id: bits 0-11, prio: 13-15 */ - struct rtentry *neigh; /* associated neighbour */ + struct llentry *neigh; /* associated neighbour */ struct l2t_entry *first; /* start of hash chain */ struct l2t_entry *next; /* next l2t_entry on chain */ struct mbuf *arpq_head; /* queue of packets awaiting resolution */ diff --git a/sys/modules/cxgb/Makefile b/sys/modules/cxgb/Makefile index 7d56988..71cbb7a 100644 --- a/sys/modules/cxgb/Makefile +++ b/sys/modules/cxgb/Makefile @@ -25,7 +25,7 @@ _toe_header = ${_sysdir}/netinet/toedev.h .if exists(${_toe_header}) _toecore = toecore -_tom = tom +#_tom = tom .endif .if ${MACHINE_ARCH} == "i386" && exists(${_toe_header}) diff --git a/sys/net/if.c b/sys/net/if.c index 5142a5c..6f829b7 100644 --- a/sys/net/if.c +++ b/sys/net/if.c @@ -60,6 +60,7 @@ #include <sys/jail.h> #include <sys/vimage.h> #include <machine/stdarg.h> +#include <vm/uma.h> #include <net/if.h> #include <net/if_arp.h> @@ -1366,6 +1367,9 @@ done: return (ifa); } +#include <net/route.h> +#include <net/if_llatbl.h> + /* * Default action when installing a route with a Link Level gateway. * Lookup an appropriate real ifa to point to. diff --git a/sys/net/if_arcsubr.c b/sys/net/if_arcsubr.c index 0f5fef3..023b647 100644 --- a/sys/net/if_arcsubr.c +++ b/sys/net/if_arcsubr.c @@ -64,6 +64,7 @@ #include <net/if_arc.h> #include <net/if_arp.h> #include <net/bpf.h> +#include <net/if_llatbl.h> #if defined(INET) || defined(INET6) #include <netinet/in.h> @@ -108,6 +109,7 @@ arc_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst, u_int8_t atype, adst; int loop_copy = 0; int isphds; + struct llentry *lle; if (!((ifp->if_flags & IFF_UP) && (ifp->if_drv_flags & IFF_DRV_RUNNING))) @@ -127,7 +129,7 @@ arc_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst, else if (ifp->if_flags & IFF_NOARP) adst = ntohl(SIN(dst)->sin_addr.s_addr) & 0xFF; else { - error = arpresolve(ifp, rt0, m, dst, &adst); + error = arpresolve(ifp, rt0, m, dst, &adst, &lle); if (error) return (error == EWOULDBLOCK ? 0 : error); } @@ -165,7 +167,7 @@ arc_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst, #endif #ifdef INET6 case AF_INET6: - error = nd6_storelladdr(ifp, rt0, m, dst, (u_char *)&adst); + error = nd6_storelladdr(ifp, rt0, m, dst, (u_char *)&adst, &lle); if (error) return (error); atype = ARCTYPE_INET6; diff --git a/sys/net/if_atmsubr.c b/sys/net/if_atmsubr.c index 9d1a7fa..7e36187 100644 --- a/sys/net/if_atmsubr.c +++ b/sys/net/if_atmsubr.c @@ -153,22 +153,11 @@ atm_output(struct ifnet *ifp, struct mbuf *m0, struct sockaddr *dst, case AF_INET: case AF_INET6: { - struct rtentry *rt = NULL; - /* - * check route - */ - if (rt0 != NULL) { - error = rt_check(&rt, &rt0, dst); - if (error) - goto bad; - RT_UNLOCK(rt); - } - if (dst->sa_family == AF_INET6) etype = ETHERTYPE_IPV6; else etype = ETHERTYPE_IP; - if (!atmresolve(rt, m, dst, &atmdst)) { + if (!atmresolve(rt0, m, dst, &atmdst)) { m = NULL; /* XXX: atmresolve already free'd it */ senderr(EHOSTUNREACH); diff --git a/sys/net/if_ethersubr.c b/sys/net/if_ethersubr.c index 4524fdd..e5978c6 100644 --- a/sys/net/if_ethersubr.c +++ b/sys/net/if_ethersubr.c @@ -64,6 +64,7 @@ #include <net/ethernet.h> #include <net/if_bridgevar.h> #include <net/if_vlan_var.h> +#include <net/if_llatbl.h> #include <net/pf_mtag.h> #include <net/vnet.h> @@ -87,6 +88,7 @@ #include <netipx/ipx.h> #include <netipx/ipx_if.h> #endif + int (*ef_inputp)(struct ifnet*, struct ether_header *eh, struct mbuf *m); int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, struct sockaddr *dst, short *tp, int *hlen); @@ -151,6 +153,7 @@ static int ether_ipfw; #endif #endif + /* * Ethernet output routine. * Encapsulate a packet of type family for the local net. @@ -164,6 +167,7 @@ ether_output(struct ifnet *ifp, struct mbuf *m, short type; int error, hdrcmplt = 0; u_char esrc[ETHER_ADDR_LEN], edst[ETHER_ADDR_LEN]; + struct llentry *lle = NULL; struct ether_header *eh; struct pf_mtag *t; int loop_copy = 1; @@ -186,7 +190,7 @@ ether_output(struct ifnet *ifp, struct mbuf *m, switch (dst->sa_family) { #ifdef INET case AF_INET: - error = arpresolve(ifp, rt0, m, dst, edst); + error = arpresolve(ifp, rt0, m, dst, edst, &lle); if (error) return (error == EWOULDBLOCK ? 0 : error); type = htons(ETHERTYPE_IP); @@ -221,7 +225,7 @@ ether_output(struct ifnet *ifp, struct mbuf *m, #endif #ifdef INET6 case AF_INET6: - error = nd6_storelladdr(ifp, rt0, m, dst, (u_char *)edst); + error = nd6_storelladdr(ifp, rt0, m, dst, (u_char *)edst, &lle); if (error) return error; type = htons(ETHERTYPE_IPV6); @@ -289,6 +293,17 @@ ether_output(struct ifnet *ifp, struct mbuf *m, senderr(EAFNOSUPPORT); } + if (lle != NULL && (lle->la_flags & LLE_IFADDR)) { + int csum_flags = 0; + if (m->m_pkthdr.csum_flags & CSUM_IP) + csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID); + if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) + csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR); + m->m_pkthdr.csum_flags |= csum_flags; + m->m_pkthdr.csum_data = 0xffff; + return (if_simloop(ifp, m, dst->sa_family, 0)); + } + /* * Add local net header. If no space in first mbuf, * allocate another. diff --git a/sys/net/if_fddisubr.c b/sys/net/if_fddisubr.c index e37d7ae..26d64e0 100644 --- a/sys/net/if_fddisubr.c +++ b/sys/net/if_fddisubr.c @@ -55,6 +55,7 @@ #include <net/if_dl.h> #include <net/if_llc.h> #include <net/if_types.h> +#include <net/if_llatbl.h> #include <net/ethernet.h> #include <net/netisr.h> @@ -120,6 +121,7 @@ fddi_output(ifp, m, dst, rt0) int loop_copy = 0, error = 0, hdrcmplt = 0; u_char esrc[FDDI_ADDR_LEN], edst[FDDI_ADDR_LEN]; struct fddi_header *fh; + struct llentry *lle; #ifdef MAC error = mac_ifnet_check_transmit(ifp, m); @@ -137,7 +139,7 @@ fddi_output(ifp, m, dst, rt0) switch (dst->sa_family) { #ifdef INET case AF_INET: { - error = arpresolve(ifp, rt0, m, dst, edst); + error = arpresolve(ifp, rt0, m, dst, edst, &lle); if (error) return (error == EWOULDBLOCK ? 0 : error); type = htons(ETHERTYPE_IP); @@ -173,7 +175,7 @@ fddi_output(ifp, m, dst, rt0) #endif /* INET */ #ifdef INET6 case AF_INET6: - error = nd6_storelladdr(ifp, rt0, m, dst, (u_char *)edst); + error = nd6_storelladdr(ifp, rt0, m, dst, (u_char *)edst, &lle); if (error) return (error); /* Something bad happened */ type = htons(ETHERTYPE_IPV6); diff --git a/sys/net/if_fwsubr.c b/sys/net/if_fwsubr.c index 77e7b70..b087370 100644 --- a/sys/net/if_fwsubr.c +++ b/sys/net/if_fwsubr.c @@ -51,6 +51,7 @@ #include <net/if_types.h> #include <net/bpf.h> #include <net/firewire.h> +#include <net/if_llatbl.h> #if defined(INET) || defined(INET6) #include <netinet/in.h> @@ -80,7 +81,6 @@ firewire_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst, { struct fw_com *fc = IFP2FWC(ifp); int error, type; - struct rtentry *rt = NULL; struct m_tag *mtag; union fw_encap *enc; struct fw_hwaddr *destfw; @@ -89,6 +89,7 @@ firewire_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst, struct mbuf *mtail; int unicast, dgl, foff; static int next_dgl; + struct llentry *lle; #ifdef MAC error = mac_ifnet_check_transmit(ifp, m); @@ -102,13 +103,6 @@ firewire_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst, goto bad; } - if (rt0 != NULL) { - error = rt_check(&rt, &rt0, dst); - if (error) - goto bad; - RT_UNLOCK(rt); - } - /* * For unicast, we make a tag to store the lladdr of the * destination. This might not be the first time we have seen @@ -144,7 +138,7 @@ firewire_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst, * doesn't fit into the arp model. */ if (unicast) { - error = arpresolve(ifp, rt, m, dst, (u_char *) destfw); + error = arpresolve(ifp, rt0, m, dst, (u_char *) destfw, &lle); if (error) return (error == EWOULDBLOCK ? 0 : error); } @@ -173,8 +167,8 @@ firewire_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst, #ifdef INET6 case AF_INET6: if (unicast) { - error = nd6_storelladdr(fc->fc_ifp, rt, m, dst, - (u_char *) destfw); + error = nd6_storelladdr(fc->fc_ifp, rt0, m, dst, + (u_char *) destfw, &lle); if (error) return (error); } diff --git a/sys/net/if_iso88025subr.c b/sys/net/if_iso88025subr.c index 00e323a..2bc336c 100644 --- a/sys/net/if_iso88025subr.c +++ b/sys/net/if_iso88025subr.c @@ -59,6 +59,7 @@ #include <net/if_dl.h> #include <net/if_llc.h> #include <net/if_types.h> +#include <net/if_llatbl.h> #include <net/ethernet.h> #include <net/netisr.h> @@ -243,7 +244,7 @@ iso88025_output(ifp, m, dst, rt0) struct iso88025_header *th; struct iso88025_header gen_th; struct sockaddr_dl *sdl = NULL; - struct rtentry *rt = NULL; + struct llentry *lle; #ifdef MAC error = mac_ifnet_check_transmit(ifp, m); @@ -260,14 +261,8 @@ iso88025_output(ifp, m, dst, rt0) /* Calculate routing info length based on arp table entry */ /* XXX any better way to do this ? */ - if (rt0 != NULL) { - error = rt_check(&rt, &rt0, dst); - if (error) - goto bad; - RT_UNLOCK(rt); - } - if (rt && (sdl = (struct sockaddr_dl *)rt->rt_gateway)) + if (rt0 && (sdl = (struct sockaddr_dl *)rt0->rt_gateway)) if (SDL_ISO88025(sdl)->trld_rcf != 0) rif_len = TR_RCF_RIFLEN(SDL_ISO88025(sdl)->trld_rcf); @@ -289,7 +284,7 @@ iso88025_output(ifp, m, dst, rt0) switch (dst->sa_family) { #ifdef INET case AF_INET: - error = arpresolve(ifp, rt0, m, dst, edst); + error = arpresolve(ifp, rt0, m, dst, edst, &lle); if (error) return (error == EWOULDBLOCK ? 0 : error); snap_type = ETHERTYPE_IP; @@ -324,7 +319,7 @@ iso88025_output(ifp, m, dst, rt0) #endif /* INET */ #ifdef INET6 case AF_INET6: - error = nd6_storelladdr(ifp, rt0, m, dst, (u_char *)edst); + error = nd6_storelladdr(ifp, rt0, m, dst, (u_char *)edst, &lle); if (error) return (error); snap_type = ETHERTYPE_IPV6; diff --git a/sys/net/if_var.h b/sys/net/if_var.h index 62c2b99..c4cb4f6 100644 --- a/sys/net/if_var.h +++ b/sys/net/if_var.h @@ -169,8 +169,6 @@ struct ifnet { void *if_bridge; /* bridge glue */ - struct lltable *lltables; /* list of L3-L2 resolution tables */ - struct label *if_label; /* interface MAC label */ /* these are only used by IPv6 */ @@ -181,6 +179,7 @@ struct ifnet { struct task if_starttask; /* task for IFF_NEEDSGIANT */ struct task if_linktask; /* task for link change events */ struct mtx if_addr_mtx; /* mutex to protect address lists */ + LIST_ENTRY(ifnet) if_clones; /* interfaces of a cloner */ TAILQ_HEAD(, ifg_list) if_groups; /* linked list of groups per if */ /* protected by if_addr_mtx */ @@ -365,6 +364,9 @@ EVENTHANDLER_DECLARE(group_change_event, group_change_event_handler_t); #define IF_AFDATA_UNLOCK(ifp) mtx_unlock(&(ifp)->if_afdata_mtx) #define IF_AFDATA_DESTROY(ifp) mtx_destroy(&(ifp)->if_afdata_mtx) +#define IF_AFDATA_LOCK_ASSERT(ifp) mtx_assert(&(ifp)->if_afdata_mtx, MA_OWNED) +#define IF_AFDATA_UNLOCK_ASSERT(ifp) mtx_assert(&(ifp)->if_afdata_mtx, MA_NOTOWNED) + #define IFF_LOCKGIANT(ifp) do { \ if ((ifp)->if_flags & IFF_NEEDSGIANT) \ mtx_lock(&Giant); \ diff --git a/sys/net/radix_mpath.c b/sys/net/radix_mpath.c index 68233fe..8d94d01 100644 --- a/sys/net/radix_mpath.c +++ b/sys/net/radix_mpath.c @@ -266,8 +266,8 @@ rtalloc_mpath_fib(struct route *ro, u_int32_t hash, u_int fibnum) * be done for sendto(3) case? */ if (ro->ro_rt && ro->ro_rt->rt_ifp && (ro->ro_rt->rt_flags & RTF_UP)) - return; /* XXX */ - ro->ro_rt = rtalloc1_fib(&ro->ro_dst, 1, RTF_CLONING, fibnum); + return; + ro->ro_rt = rtalloc1_fib(&ro->ro_dst, 1, 0, fibnum); /* if the route does not exist or it is not multipath, don't care */ if (ro->ro_rt == NULL) diff --git a/sys/net/route.c b/sys/net/route.c index be71e13..9359514 100644 --- a/sys/net/route.c +++ b/sys/net/route.c @@ -41,6 +41,7 @@ #include <sys/param.h> #include <sys/systm.h> +#include <sys/syslog.h> #include <sys/malloc.h> #include <sys/mbuf.h> #include <sys/socket.h> @@ -53,6 +54,7 @@ #include <sys/vimage.h> #include <net/if.h> +#include <net/if_dl.h> #include <net/route.h> #ifdef RADIX_MPATH @@ -269,8 +271,7 @@ rtalloc1_fib(struct sockaddr *dst, int report, u_long ignflags, struct radix_node *rn; struct rtentry *newrt; struct rt_addrinfo info; - u_long nflags; - int needresolve = 0, err = 0, msgtype = RTM_MISS; + int err = 0, msgtype = RTM_MISS; int needlock; KASSERT((fibnum < rt_numfibs), ("rtalloc1_fib: bad fibnum")); @@ -283,10 +284,9 @@ rtalloc1_fib(struct sockaddr *dst, int report, u_long ignflags, */ if (rnh == NULL) { V_rtstat.rts_unreach++; - goto miss2; + goto miss; } needlock = !(ignflags & RTF_RNH_LOCKED); -retry: if (needlock) RADIX_NODE_HEAD_RLOCK(rnh); #ifdef INVARIANTS @@ -295,103 +295,33 @@ retry: #endif rn = rnh->rnh_matchaddr(dst, rnh); if (rn && ((rn->rn_flags & RNF_ROOT) == 0)) { - newrt = rt = RNTORT(rn); - nflags = rt->rt_flags & ~ignflags; - if (report && (nflags & RTF_CLONING)) { - if (needlock && !RADIX_NODE_HEAD_LOCK_TRY_UPGRADE(rnh)) { - RADIX_NODE_HEAD_RUNLOCK(rnh); - RADIX_NODE_HEAD_LOCK(rnh); - /* - * lookup again to make sure it wasn't changed - */ - rn = rnh->rnh_matchaddr(dst, rnh); - if (!(rn && ((rn->rn_flags & RNF_ROOT) == 0))) { - RADIX_NODE_HEAD_UNLOCK(rnh); - needresolve = 0; - log(LOG_INFO, "retrying route lookup ...\n"); - goto retry; - } - } - needresolve = 1; - } else { - RT_LOCK(newrt); - RT_ADDREF(newrt); - if (needlock) - RADIX_NODE_HEAD_RUNLOCK(rnh); - goto done; - } - } + RT_LOCK(newrt); + RT_ADDREF(newrt); + if (needlock) + RADIX_NODE_HEAD_RUNLOCK(rnh); + goto done; + + } else if (needlock) + RADIX_NODE_HEAD_RUNLOCK(rnh); + /* - * if needresolve is set then we have the exclusive lock - * and we need to keep it held for the benefit of rtrequest_fib + * Either we hit the root or couldn't find any match, + * Which basically means + * "caint get there frm here" */ - if (!needresolve && needlock) - RADIX_NODE_HEAD_RUNLOCK(rnh); - - if (needresolve) { - RADIX_NODE_HEAD_WLOCK_ASSERT(rnh); + V_rtstat.rts_unreach++; +miss: + if (report) { /* - * We are apparently adding (report = 0 in delete). - * If it requires that it be cloned, do so. - * (This implies it wasn't a HOST route.) + * If required, report the failure to the supervising + * Authorities. + * For a delete, this is not an error. (report == 0) */ - err = rtrequest_fib(RTM_RESOLVE, dst, NULL, - NULL, RTF_RNH_LOCKED, &newrt, fibnum); - if (err) { - /* - * If the cloning didn't succeed, maybe - * what we have will do. Return that. - */ - newrt = rt; /* existing route */ - RT_LOCK(newrt); - RT_ADDREF(newrt); - goto miss; - } - KASSERT(newrt, ("no route and no error")); - RT_LOCK(newrt); - if (newrt->rt_flags & RTF_XRESOLVE) { - /* - * If the new route specifies it be - * externally resolved, then go do that. - */ - msgtype = RTM_RESOLVE; - goto miss; - } - /* Inform listeners of the new route. */ bzero(&info, sizeof(info)); - info.rti_info[RTAX_DST] = rt_key(newrt); - info.rti_info[RTAX_NETMASK] = rt_mask(newrt); - info.rti_info[RTAX_GATEWAY] = newrt->rt_gateway; - if (newrt->rt_ifp != NULL) { - info.rti_info[RTAX_IFP] = - newrt->rt_ifp->if_addr->ifa_addr; - info.rti_info[RTAX_IFA] = newrt->rt_ifa->ifa_addr; - } - rt_missmsg(RTM_ADD, &info, newrt->rt_flags, 0); - if (needlock) - RADIX_NODE_HEAD_UNLOCK(rnh); - } else { - /* - * Either we hit the root or couldn't find any match, - * Which basically means - * "caint get there frm here" - */ - V_rtstat.rts_unreach++; - miss: - if (needlock && needresolve) - RADIX_NODE_HEAD_UNLOCK(rnh); - miss2: if (report) { - /* - * If required, report the failure to the supervising - * Authorities. - * For a delete, this is not an error. (report == 0) - */ - bzero(&info, sizeof(info)); - info.rti_info[RTAX_DST] = dst; - rt_missmsg(msgtype, &info, 0, err); - } - } + info.rti_info[RTAX_DST] = dst; + rt_missmsg(msgtype, &info, 0, err); + } done: if (newrt) RT_LOCK_ASSERT(newrt); @@ -420,7 +350,7 @@ rtfree(struct rtentry *rt) */ RT_REMREF(rt); if (rt->rt_refcnt > 0) { - printf("%s: %p has %lu refs\n", __func__, rt, rt->rt_refcnt); + log(LOG_DEBUG, "%s: %p has %d refs\t", __func__, rt, rt->rt_refcnt); goto done; } @@ -461,8 +391,6 @@ rtfree(struct rtentry *rt) */ if (rt->rt_ifa) IFAFREE(rt->rt_ifa); - rt->rt_parent = NULL; /* NB: no refcnt on parent */ - /* * The key is separatly alloc'd so free it (see rt_setgate()). * This also frees the gateway, as they are always malloc'd @@ -715,14 +643,6 @@ ifa_ifwithroute_fib(int flags, struct sockaddr *dst, struct sockaddr *gateway, return (ifa); } -static walktree_f_t rt_fixdelete; -static walktree_f_t rt_fixchange; - -struct rtfc_arg { - struct rtentry *rt0; - struct radix_node_head *rnh; -}; - /* * Do appropriate manipulations of a routing tree given * all the bits of info needed @@ -828,8 +748,13 @@ rtexpunge(struct rtentry *rt) struct ifaddr *ifa; int error = 0; + /* + * Find the correct routing tree to use for this Address Family + */ rnh = V_rt_tables[rt->rt_fibnum][rt_key(rt)->sa_family]; RT_LOCK_ASSERT(rt); + if (rnh == NULL) + return (EAFNOSUPPORT); RADIX_NODE_HEAD_LOCK_ASSERT(rnh); #if 0 /* @@ -840,13 +765,6 @@ rtexpunge(struct rtentry *rt) KASSERT(rt->rt_refcnt <= 1, ("bogus refcnt %ld", rt->rt_refcnt)); #endif /* - * Find the correct routing tree to use for this Address Family - */ - rnh = V_rt_tables[rt->rt_fibnum][rt_key(rt)->sa_family]; - if (rnh == NULL) - return (EAFNOSUPPORT); - - /* * Remove the item from the tree; it should be there, * but when callers invoke us blindly it may not (sigh). */ @@ -863,24 +781,6 @@ rtexpunge(struct rtentry *rt) rt->rt_flags &= ~RTF_UP; /* - * Now search what's left of the subtree for any cloned - * routes which might have been formed from this node. - */ - if ((rt->rt_flags & RTF_CLONING) && rt_mask(rt)) - rnh->rnh_walktree_from(rnh, rt_key(rt), rt_mask(rt), - rt_fixdelete, rt); - - /* - * Remove any external references we may have. - * This might result in another rtentry being freed if - * we held its last reference. - */ - if (rt->rt_gwroute) { - RTFREE(rt->rt_gwroute); - rt->rt_gwroute = NULL; - } - - /* * Give the protocol a chance to keep things in sync. */ if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest) { @@ -904,12 +804,6 @@ bad: } int -rtrequest1(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt) -{ - return (rtrequest1_fib(req, info, ret_nrt, 0)); -} - -int rtrequest1_fib(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt, u_int fibnum) { @@ -941,10 +835,9 @@ rtrequest1_fib(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt, * If we are adding a host route then we don't want to put * a netmask in the tree, nor do we want to clone it. */ - if (flags & RTF_HOST) { + if (flags & RTF_HOST) netmask = NULL; - flags &= ~RTF_CLONING; - } + switch (req) { case RTM_DELETE: #ifdef RADIX_MPATH @@ -1023,26 +916,6 @@ normal_rtdel: rt->rt_flags &= ~RTF_UP; /* - * Now search what's left of the subtree for any cloned - * routes which might have been formed from this node. - */ - if ((rt->rt_flags & RTF_CLONING) && - rt_mask(rt)) { - rnh->rnh_walktree_from(rnh, dst, rt_mask(rt), - rt_fixdelete, rt); - } - - /* - * Remove any external references we may have. - * This might result in another rtentry being freed if - * we held its last reference. - */ - if (rt->rt_gwroute) { - RTFREE(rt->rt_gwroute); - rt->rt_gwroute = NULL; - } - - /* * give the protocol a chance to keep things in sync. */ if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest) @@ -1069,20 +942,12 @@ deldone: } else RTFREE_LOCKED(rt); break; - case RTM_RESOLVE: - if (ret_nrt == NULL || (rt = *ret_nrt) == NULL) - senderr(EINVAL); - ifa = rt->rt_ifa; - /* XXX locking? */ - flags = rt->rt_flags & - ~(RTF_CLONING | RTF_STATIC); - flags |= RTF_WASCLONED; - gateway = rt->rt_gateway; - if ((netmask = rt->rt_genmask) == NULL) - flags |= RTF_HOST; - goto makeroute; - + /* + * resolve was only used for route cloning + * here for compat + */ + break; case RTM_ADD: if ((flags & RTF_GATEWAY) && !gateway) senderr(EINVAL); @@ -1093,8 +958,6 @@ deldone: if (info->rti_ifa == NULL && (error = rt_getifa_fib(info, fibnum))) senderr(error); ifa = info->rti_ifa; - - makeroute: rt = uma_zalloc(rtzone, M_NOWAIT | M_ZERO); if (rt == NULL) senderr(ENOBUFS); @@ -1103,7 +966,7 @@ deldone: rt->rt_fibnum = fibnum; /* * Add the gateway. Possibly re-malloc-ing the storage for it - * also add the rt_gwroute if possible. + * */ RT_LOCK(rt); if ((error = rt_setgate(rt, dst, gateway)) != 0) { @@ -1138,8 +1001,6 @@ deldone: /* do not permit exactly the same dst/mask/gw pair */ if (rn_mpath_capable(rnh) && rt_mpath_conflict(rnh, rt, netmask)) { - if (rt->rt_gwroute) - RTFREE(rt->rt_gwroute); if (rt->rt_ifa) { IFAFREE(rt->rt_ifa); } @@ -1152,34 +1013,11 @@ deldone: /* XXX mtu manipulation will be done in rnh_addaddr -- itojun */ rn = rnh->rnh_addaddr(ndst, netmask, rnh, rt->rt_nodes); - if (rn == NULL) { - struct rtentry *rt2; - /* - * Uh-oh, we already have one of these in the tree. - * We do a special hack: if the route that's already - * there was generated by the cloning mechanism - * then we just blow it away and retry the insertion - * of the new one. - */ - rt2 = rtalloc1_fib(dst, 0, RTF_RNH_LOCKED, fibnum); - if (rt2 && rt2->rt_parent) { - rtexpunge(rt2); - RT_UNLOCK(rt2); - rn = rnh->rnh_addaddr(ndst, netmask, - rnh, rt->rt_nodes); - } else if (rt2) { - /* undo the extra ref we got */ - RTFREE_LOCKED(rt2); - } - } - /* * If it still failed to go into the tree, * then un-make it (this should be a function) */ if (rn == NULL) { - if (rt->rt_gwroute) - RTFREE(rt->rt_gwroute); if (rt->rt_ifa) IFAFREE(rt->rt_ifa); Free(rt_key(rt)); @@ -1188,33 +1026,6 @@ deldone: senderr(EEXIST); } - rt->rt_parent = NULL; - - /* - * If we got here from RESOLVE, then we are cloning - * so clone the rest, and note that we - * are a clone (and increment the parent's references) - */ - if (req == RTM_RESOLVE) { - KASSERT(ret_nrt && *ret_nrt, - ("no route to clone from")); - rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */ - rt->rt_rmx.rmx_pksent = 0; /* reset packet counter */ - if ((*ret_nrt)->rt_flags & RTF_CLONING) { - /* - * NB: We do not bump the refcnt on the parent - * entry under the assumption that it will - * remain so long as we do. This is - * important when deleting the parent route - * as this operation requires traversing - * the tree to delete all clones and futzing - * with refcnts requires us to double-lock - * parent through this back reference. - */ - rt->rt_parent = *ret_nrt; - } - } - /* * If this protocol has something to add to this then * allow it to do that as well. @@ -1223,20 +1034,6 @@ deldone: ifa->ifa_rtrequest(req, rt, info); /* - * We repeat the same procedure from rt_setgate() here because - * it doesn't fire when we call it there because the node - * hasn't been added to the tree yet. - */ - if (req == RTM_ADD && - !(rt->rt_flags & RTF_HOST) && rt_mask(rt) != NULL) { - struct rtfc_arg arg; - arg.rnh = rnh; - arg.rt0 = rt; - rnh->rnh_walktree_from(rnh, rt_key(rt), rt_mask(rt), - rt_fixchange, &arg); - } - - /* * actually return a resultant rtentry and * give the caller a single reference. */ @@ -1263,159 +1060,21 @@ bad: #undef ifpaddr #undef flags -/* - * Called from rtrequest(RTM_DELETE, ...) to fix up the route's ``family'' - * (i.e., the routes related to it by the operation of cloning). This - * routine is iterated over all potential former-child-routes by way of - * rnh->rnh_walktree_from() above, and those that actually are children of - * the late parent (passed in as VP here) are themselves deleted. - */ -static int -rt_fixdelete(struct radix_node *rn, void *vp) -{ - struct rtentry *rt = RNTORT(rn); - struct rtentry *rt0 = vp; - - if (rt->rt_parent == rt0 && - !(rt->rt_flags & (RTF_PINNED | RTF_CLONING))) { - return rtrequest_fib(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), - rt->rt_flags|RTF_RNH_LOCKED, NULL, rt->rt_fibnum); - } - return 0; -} - -/* - * This routine is called from rt_setgate() to do the analogous thing for - * adds and changes. There is the added complication in this case of a - * middle insert; i.e., insertion of a new network route between an older - * network route and (cloned) host routes. For this reason, a simple check - * of rt->rt_parent is insufficient; each candidate route must be tested - * against the (mask, value) of the new route (passed as before in vp) - * to see if the new route matches it. - * - * XXX - it may be possible to do fixdelete() for changes and reserve this - * routine just for adds. I'm not sure why I thought it was necessary to do - * changes this way. - */ - -static int -rt_fixchange(struct radix_node *rn, void *vp) -{ - struct rtentry *rt = RNTORT(rn); - struct rtfc_arg *ap = vp; - struct rtentry *rt0 = ap->rt0; - struct radix_node_head *rnh = ap->rnh; - u_char *xk1, *xm1, *xk2, *xmp; - int i, len, mlen; - - /* make sure we have a parent, and route is not pinned or cloning */ - if (!rt->rt_parent || - (rt->rt_flags & (RTF_PINNED | RTF_CLONING))) - return 0; - - if (rt->rt_parent == rt0) /* parent match */ - goto delete_rt; - /* - * There probably is a function somewhere which does this... - * if not, there should be. - */ - len = imin(rt_key(rt0)->sa_len, rt_key(rt)->sa_len); - - xk1 = (u_char *)rt_key(rt0); - xm1 = (u_char *)rt_mask(rt0); - xk2 = (u_char *)rt_key(rt); - - /* avoid applying a less specific route */ - xmp = (u_char *)rt_mask(rt->rt_parent); - mlen = rt_key(rt->rt_parent)->sa_len; - if (mlen > rt_key(rt0)->sa_len) /* less specific route */ - return 0; - for (i = rnh->rnh_treetop->rn_offset; i < mlen; i++) - if ((xmp[i] & ~(xmp[i] ^ xm1[i])) != xmp[i]) - return 0; /* less specific route */ - - for (i = rnh->rnh_treetop->rn_offset; i < len; i++) - if ((xk2[i] & xm1[i]) != xk1[i]) - return 0; /* no match */ - - /* - * OK, this node is a clone, and matches the node currently being - * changed/added under the node's mask. So, get rid of it. - */ -delete_rt: - return rtrequest_fib(RTM_DELETE, rt_key(rt), NULL, - rt_mask(rt), rt->rt_flags, NULL, rt->rt_fibnum); -} - int rt_setgate(struct rtentry *rt, struct sockaddr *dst, struct sockaddr *gate) { INIT_VNET_NET(curvnet); /* XXX dst may be overwritten, can we move this to below */ + int dlen = SA_SIZE(dst), glen = SA_SIZE(gate); +#ifdef INVARIANTS struct radix_node_head *rnh = V_rt_tables[rt->rt_fibnum][dst->sa_family]; - int dlen = SA_SIZE(dst), glen = SA_SIZE(gate); +#endif -again: RT_LOCK_ASSERT(rt); RADIX_NODE_HEAD_LOCK_ASSERT(rnh); /* - * A host route with the destination equal to the gateway - * will interfere with keeping LLINFO in the routing - * table, so disallow it. - */ - if (((rt->rt_flags & (RTF_HOST|RTF_GATEWAY|RTF_LLINFO)) == - (RTF_HOST|RTF_GATEWAY)) && - dst->sa_len == gate->sa_len && - bcmp(dst, gate, dst->sa_len) == 0) { - /* - * The route might already exist if this is an RTM_CHANGE - * or a routing redirect, so try to delete it. - */ - if (rt_key(rt)) - rtexpunge(rt); - return EADDRNOTAVAIL; - } - - /* - * Cloning loop avoidance in case of bad configuration. - */ - if (rt->rt_flags & RTF_GATEWAY) { - struct rtentry *gwrt; - - RT_UNLOCK(rt); /* XXX workaround LOR */ - gwrt = rtalloc1_fib(gate, 1, RTF_RNH_LOCKED, rt->rt_fibnum); - if (gwrt == rt) { - RT_REMREF(rt); - return (EADDRINUSE); /* failure */ - } - /* - * Try to reacquire the lock on rt, and if it fails, - * clean state and restart from scratch. - */ - if (!RT_TRYLOCK(rt)) { - RTFREE_LOCKED(gwrt); - RT_LOCK(rt); - goto again; - } - /* - * If there is already a gwroute, then drop it. If we - * are asked to replace route with itself, then do - * not leak its refcounter. - */ - if (rt->rt_gwroute != NULL) { - if (rt->rt_gwroute == gwrt) { - RT_REMREF(rt->rt_gwroute); - } else - RTFREE(rt->rt_gwroute); - } - - if ((rt->rt_gwroute = gwrt) != NULL) - RT_UNLOCK(rt->rt_gwroute); - } - - /* * Prepare to store the gateway in rt->rt_gateway. * Both dst and gateway are stored one after the other in the same * malloc'd chunk. If we have room, we can reuse the old buffer, @@ -1446,21 +1105,7 @@ again: */ bcopy(gate, rt->rt_gateway, glen); - /* - * This isn't going to do anything useful for host routes, so - * don't bother. Also make sure we have a reasonable mask - * (we don't yet have one during adds). - */ - if (!(rt->rt_flags & RTF_HOST) && rt_mask(rt) != 0) { - struct rtfc_arg arg; - - arg.rnh = rnh; - arg.rt0 = rt; - rnh->rnh_walktree_from(rnh, rt_key(rt), rt_mask(rt), - rt_fixchange, &arg); - } - - return 0; + return (0); } static void @@ -1500,6 +1145,7 @@ rtinit1(struct ifaddr *ifa, int cmd, int flags, int fibnum) char tempbuf[_SOCKADDR_TMPSIZE]; int didwork = 0; int a_failure = 0; + static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; if (flags & RTF_HOST) { dst = ifa->ifa_dstaddr; @@ -1604,7 +1250,14 @@ rtinit1(struct ifaddr *ifa, int cmd, int flags, int fibnum) info.rti_ifa = ifa; info.rti_flags = flags | ifa->ifa_flags; info.rti_info[RTAX_DST] = dst; - info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr; + /* + * doing this for compatibility reasons + */ + if (cmd == RTM_ADD) + info.rti_info[RTAX_GATEWAY] = + (struct sockaddr *)&null_sdl; + else + info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr; info.rti_info[RTAX_NETMASK] = netmask; error = rtrequest1_fib(cmd, &info, &rt, fibnum); if (error == 0 && rt != NULL) { @@ -1628,6 +1281,15 @@ rtinit1(struct ifaddr *ifa, int cmd, int flags, int fibnum) rt->rt_ifa = ifa; } #endif + /* + * doing this for compatibility reasons + */ + if (cmd == RTM_ADD) { + ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type = + rt->rt_ifp->if_type; + ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index = + rt->rt_ifp->if_index; + } rt_newaddrmsg(cmd, ifa, error, rt); if (cmd == RTM_DELETE) { /* @@ -1696,147 +1358,5 @@ rtinit(struct ifaddr *ifa, int cmd, int flags) return (rtinit1(ifa, cmd, flags, fib)); } -/* - * rt_check() is invoked on each layer 2 output path, prior to - * encapsulating outbound packets. - * - * The function is mostly used to find a routing entry for the gateway, - * which in some protocol families could also point to the link-level - * address for the gateway itself (the side effect of revalidating the - * route to the destination is rather pointless at this stage, we did it - * already a moment before in the pr_output() routine to locate the ifp - * and gateway to use). - * - * When we remove the layer-3 to layer-2 mapping tables from the - * routing table, this function can be removed. - * - * === On input === - * *dst is the address of the NEXT HOP (which coincides with the - * final destination if directly reachable); - * *lrt0 points to the cached route to the final destination; - * *lrt is not meaningful; - * (*lrt0 has no ref held on it by us so REMREF is not needed. - * Refs only account for major structural references and not usages, - * which is actually a bit of a problem.) - * - * === Operation === - * If the route is marked down try to find a new route. If the route - * to the gateway is gone, try to setup a new route. Otherwise, - * if the route is marked for packets to be rejected, enforce that. - * Note that rtalloc returns an rtentry with an extra REF that we may - * need to lose. - * - * === On return === - * *dst is unchanged; - * *lrt0 points to the (possibly new) route to the final destination - * *lrt points to the route to the next hop [LOCKED] - * - * Their values are meaningful ONLY if no error is returned. - * - * To follow this you have to remember that: - * RT_REMREF reduces the reference count by 1 but doesn't check it for 0 (!) - * RTFREE_LOCKED includes an RT_REMREF (or an rtfree if refs == 1) - * and an RT_UNLOCK - * RTFREE does an RT_LOCK and an RTFREE_LOCKED - * The gwroute pointer counts as a reference on the rtentry to which it points. - * so when we add it we use the ref that rtalloc gives us and when we lose it - * we need to remove the reference. - * RT_TEMP_UNLOCK does an RT_ADDREF before freeing the lock, and - * RT_RELOCK locks it (it can't have gone away due to the ref) and - * drops the ref, possibly freeing it and zeroing the pointer if - * the ref goes to 0 (unlocking in the process). - */ -int -rt_check(struct rtentry **lrt, struct rtentry **lrt0, struct sockaddr *dst) -{ - struct rtentry *rt; - struct rtentry *rt0; - u_int fibnum; - - KASSERT(*lrt0 != NULL, ("rt_check")); - rt0 = *lrt0; - rt = NULL; - fibnum = rt0->rt_fibnum; - - /* NB: the locking here is tortuous... */ - RT_LOCK(rt0); -retry: - if (rt0 && (rt0->rt_flags & RTF_UP) == 0) { - /* Current rt0 is useless, try get a replacement. */ - RT_UNLOCK(rt0); - rt0 = NULL; - } - if (rt0 == NULL) { - rt0 = rtalloc1_fib(dst, 1, 0UL, fibnum); - if (rt0 == NULL) { - return (EHOSTUNREACH); - } - RT_REMREF(rt0); /* don't need the reference. */ - } - - if (rt0->rt_flags & RTF_GATEWAY) { - if ((rt = rt0->rt_gwroute) != NULL) { - RT_LOCK(rt); /* NB: gwroute */ - if ((rt->rt_flags & RTF_UP) == 0) { - /* gw route is dud. ignore/lose it */ - RTFREE_LOCKED(rt); /* unref (&unlock) gwroute */ - rt = rt0->rt_gwroute = NULL; - } - } - - if (rt == NULL) { /* NOT AN ELSE CLAUSE */ - RT_TEMP_UNLOCK(rt0); /* MUST return to undo this */ - rt = rtalloc1_fib(rt0->rt_gateway, 1, 0UL, fibnum); - if ((rt == rt0) || (rt == NULL)) { - /* the best we can do is not good enough */ - if (rt) { - RT_REMREF(rt); /* assumes ref > 0 */ - RT_UNLOCK(rt); - } - RTFREE(rt0); /* lock, unref, (unlock) */ - return (ENETUNREACH); - } - /* - * Relock it and lose the added reference. - * All sorts of things could have happenned while we - * had no lock on it, so check for them. - */ - RT_RELOCK(rt0); - if (rt0 == NULL || ((rt0->rt_flags & RTF_UP) == 0)) - /* Ru-roh.. what we had is no longer any good */ - goto retry; - /* - * While we were away, someone replaced the gateway. - * Since a reference count is involved we can't just - * overwrite it. - */ - if (rt0->rt_gwroute) { - if (rt0->rt_gwroute != rt) { - RTFREE_LOCKED(rt); - goto retry; - } - } else { - rt0->rt_gwroute = rt; - } - } - RT_LOCK_ASSERT(rt); - RT_UNLOCK(rt0); - } else { - /* think of rt as having the lock from now on.. */ - rt = rt0; - } - /* XXX why are we inspecting rmx_expire? */ - if ((rt->rt_flags & RTF_REJECT) && - (rt->rt_rmx.rmx_expire == 0 || - time_uptime < rt->rt_rmx.rmx_expire)) { - RT_UNLOCK(rt); - return (rt == rt0 ? EHOSTDOWN : EHOSTUNREACH); - } - - *lrt = rt; - *lrt0 = rt0; - return (0); -} - /* This must be before ip6_init2(), which is now SI_ORDER_MIDDLE */ SYSINIT(route, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, route_init, 0); diff --git a/sys/net/route.h b/sys/net/route.h index fb8df39..5062198 100644 --- a/sys/net/route.h +++ b/sys/net/route.h @@ -135,15 +135,11 @@ struct rtentry { #define rt_key(r) (*((struct sockaddr **)(&(r)->rt_nodes->rn_key))) #define rt_mask(r) (*((struct sockaddr **)(&(r)->rt_nodes->rn_mask))) struct sockaddr *rt_gateway; /* value */ - u_long rt_flags; /* up/down?, host/net */ + int rt_flags; /* up/down?, host/net */ + int rt_refcnt; /* # held references */ struct ifnet *rt_ifp; /* the answer: interface to use */ struct ifaddr *rt_ifa; /* the answer: interface address to use */ struct rt_metrics_lite rt_rmx; /* metrics used by rx'ing protocols */ - long rt_refcnt; /* # held references */ - struct sockaddr *rt_genmask; /* for generation of cloned routes */ - caddr_t rt_llinfo; /* pointer to link level info cache */ - struct rtentry *rt_gwroute; /* implied entry for gatewayed routes */ - struct rtentry *rt_parent; /* cloning parent of this route */ u_int rt_fibnum; /* which FIB */ #ifdef _KERNEL /* XXX ugly, user apps use this definition but don't have a mtx def */ @@ -175,9 +171,9 @@ struct ortentry { #define RTF_MODIFIED 0x20 /* modified dynamically (by redirect) */ #define RTF_DONE 0x40 /* message confirmed */ /* 0x80 unused, was RTF_DELCLONE */ -#define RTF_CLONING 0x100 /* generate new routes on use */ +/* 0x100 unused, was RTF_CLONING */ #define RTF_XRESOLVE 0x200 /* external daemon resolves name */ -#define RTF_LLINFO 0x400 /* generated by link layer (e.g. ARP) */ +/* 0x400 unused, was RTF_LLINFO */ #define RTF_STATIC 0x800 /* manually added */ #define RTF_BLACKHOLE 0x1000 /* just discard pkts (during updates) */ #define RTF_PROTO2 0x4000 /* protocol specific routing flag */ @@ -188,7 +184,7 @@ struct ortentry { #define RTF_PRCLONING 0x10000 /* unused, for compatibility */ #endif -#define RTF_WASCLONED 0x20000 /* route generated through cloning */ +/* 0x20000 unused, was RTF_WASCLONED */ #define RTF_PROTO3 0x40000 /* protocol specific routing flag */ /* 0x80000 unused */ #define RTF_PINNED 0x100000 /* future use */ @@ -326,14 +322,14 @@ struct rt_addrinfo { #define RT_ADDREF(_rt) do { \ RT_LOCK_ASSERT(_rt); \ KASSERT((_rt)->rt_refcnt >= 0, \ - ("negative refcnt %ld", (_rt)->rt_refcnt)); \ + ("negative refcnt %d", (_rt)->rt_refcnt)); \ (_rt)->rt_refcnt++; \ } while (0) #define RT_REMREF(_rt) do { \ RT_LOCK_ASSERT(_rt); \ KASSERT((_rt)->rt_refcnt > 0, \ - ("bogus refcnt %ld", (_rt)->rt_refcnt)); \ + ("bogus refcnt %d", (_rt)->rt_refcnt)); \ (_rt)->rt_refcnt--; \ } while (0) @@ -409,7 +405,6 @@ void rtredirect(struct sockaddr *, struct sockaddr *, struct sockaddr *, int, struct sockaddr *); int rtrequest(int, struct sockaddr *, struct sockaddr *, struct sockaddr *, int, struct rtentry **); -int rtrequest1(int, struct rt_addrinfo *, struct rtentry **); /* defaults to "all" FIBs */ int rtinit_fib(struct ifaddr *, int, int); diff --git a/sys/net/rtsock.c b/sys/net/rtsock.c index 5003b03..30591c7 100644 --- a/sys/net/rtsock.c +++ b/sys/net/rtsock.c @@ -53,6 +53,7 @@ #include <sys/vimage.h> #include <net/if.h> +#include <net/if_llatbl.h> #include <net/netisr.h> #include <net/raw_cb.h> #include <net/route.h> @@ -496,19 +497,6 @@ route_output(struct mbuf *m, struct socket *so) (info.rti_info[RTAX_GATEWAY] != NULL && info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) senderr(EINVAL); - if (info.rti_info[RTAX_GENMASK]) { - struct radix_node *t; - t = rn_addmask((caddr_t) info.rti_info[RTAX_GENMASK], 0, 1); - if (t != NULL && - bcmp((char *)(void *)info.rti_info[RTAX_GENMASK] + 1, - (char *)(void *)t->rn_key + 1, - ((struct sockaddr *)t->rn_key)->sa_len - 1) == 0) - info.rti_info[RTAX_GENMASK] = - (struct sockaddr *)t->rn_key; - else - senderr(ENOBUFS); - } - /* * Verify that the caller has the appropriate privilege; RTM_GET * is the only operation the non-superuser is allowed. @@ -526,6 +514,11 @@ route_output(struct mbuf *m, struct socket *so) if (info.rti_info[RTAX_GATEWAY] == NULL) senderr(EINVAL); saved_nrt = NULL; + /* support for new ARP code */ + if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) { + error = lla_rt_output(rtm, &info); + break; + } error = rtrequest1_fib(RTM_ADD, &info, &saved_nrt, so->so_fibnum); if (error == 0 && saved_nrt) { @@ -534,13 +527,18 @@ route_output(struct mbuf *m, struct socket *so) &rtm->rtm_rmx, &saved_nrt->rt_rmx); rtm->rtm_index = saved_nrt->rt_ifp->if_index; RT_REMREF(saved_nrt); - saved_nrt->rt_genmask = info.rti_info[RTAX_GENMASK]; RT_UNLOCK(saved_nrt); } break; case RTM_DELETE: saved_nrt = NULL; + /* support for new ARP code */ + if (info.rti_info[RTAX_GATEWAY] && + (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK)) { + error = lla_rt_output(rtm, &info); + break; + } error = rtrequest1_fib(RTM_DELETE, &info, &saved_nrt, so->so_fibnum); if (error == 0) { @@ -612,7 +610,7 @@ route_output(struct mbuf *m, struct socket *so) info.rti_info[RTAX_DST] = rt_key(rt); info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; info.rti_info[RTAX_NETMASK] = rt_mask(rt); - info.rti_info[RTAX_GENMASK] = rt->rt_genmask; + info.rti_info[RTAX_GENMASK] = 0; if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) { ifp = rt->rt_ifp; if (ifp) { @@ -699,8 +697,7 @@ route_output(struct mbuf *m, struct socket *so) RT_UNLOCK(rt); senderr(error); } - if (!(rt->rt_flags & RTF_LLINFO)) - rt->rt_flags |= RTF_GATEWAY; + rt->rt_flags |= RTF_GATEWAY; } if (info.rti_ifa != NULL && info.rti_ifa != rt->rt_ifa) { @@ -718,8 +715,6 @@ route_output(struct mbuf *m, struct socket *so) rtm->rtm_index = rt->rt_ifp->if_index; if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest) rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, &info); - if (info.rti_info[RTAX_GENMASK]) - rt->rt_genmask = info.rti_info[RTAX_GENMASK]; /* FALLTHROUGH */ case RTM_LOCK: /* We don't support locks anymore */ @@ -1261,7 +1256,7 @@ sysctl_dumpentry(struct radix_node *rn, void *vw) info.rti_info[RTAX_DST] = rt_key(rt); info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; info.rti_info[RTAX_NETMASK] = rt_mask(rt); - info.rti_info[RTAX_GENMASK] = rt->rt_genmask; + info.rti_info[RTAX_GENMASK] = 0; if (rt->rt_ifp) { info.rti_info[RTAX_IFP] = rt->rt_ifp->if_addr->ifa_addr; info.rti_info[RTAX_IFA] = rt->rt_ifa->ifa_addr; @@ -1440,6 +1435,11 @@ sysctl_rtsock(SYSCTL_HANDLER_ARGS) RADIX_NODE_HEAD_UNLOCK(rnh); } else if (af != 0) error = EAFNOSUPPORT; + /* + * take care of llinfo entries + */ + if (w.w_op == NET_RT_FLAGS) + error = lltable_sysctl_dumparp(af, w.w_req); break; case NET_RT_IFLIST: diff --git a/sys/netgraph/netflow/netflow.c b/sys/netgraph/netflow/netflow.c index 5443629..411730a 100644 --- a/sys/netgraph/netflow/netflow.c +++ b/sys/netgraph/netflow/netflow.c @@ -270,7 +270,7 @@ hash_insert(priv_p priv, struct flow_hash_entry *hsh, struct flow_rec *r, sin.sin_family = AF_INET; sin.sin_addr = fle->f.r.r_dst; /* XXX MRT 0 as a default.. need the m here to get fib */ - rt = rtalloc1_fib((struct sockaddr *)&sin, 0, RTF_CLONING, 0); + rt = rtalloc1_fib((struct sockaddr *)&sin, 0, 0, 0); if (rt != NULL) { fle->f.fle_o_ifx = rt->rt_ifp->if_index; @@ -295,7 +295,7 @@ hash_insert(priv_p priv, struct flow_hash_entry *hsh, struct flow_rec *r, sin.sin_family = AF_INET; sin.sin_addr = fle->f.r.r_src; /* XXX MRT 0 as a default revisit. need the mbuf for fib*/ - rt = rtalloc1_fib((struct sockaddr *)&sin, 0, RTF_CLONING, 0); + rt = rtalloc1_fib((struct sockaddr *)&sin, 0, 0, 0); if (rt != NULL) { if (rt_mask(rt)) fle->f.src_mask = bitcount32(((struct sockaddr_in *) diff --git a/sys/netinet/if_atm.c b/sys/netinet/if_atm.c index 065f0c4..d534a18 100644 --- a/sys/netinet/if_atm.c +++ b/sys/netinet/if_atm.c @@ -127,10 +127,6 @@ atm_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info) break; } - if ((rt->rt_flags & RTF_CLONING) != 0) { - printf("atm_rtrequest: cloning route detected?\n"); - break; - } if (gate->sa_family != AF_LINK || gate->sa_len < sizeof(null_sdl)) { log(LOG_DEBUG, "atm_rtrequest: bad gateway value"); @@ -332,8 +328,6 @@ atmresolve(struct rtentry *rt, struct mbuf *m, struct sockaddr *dst, goto bad; /* failed */ RT_REMREF(rt); /* don't keep LL references */ if ((rt->rt_flags & RTF_GATEWAY) != 0 || - (rt->rt_flags & RTF_LLINFO) == 0 || - /* XXX: are we using LLINFO? */ rt->rt_gateway->sa_family != AF_LINK) { RT_UNLOCK(rt); goto bad; diff --git a/sys/netinet/if_ether.c b/sys/netinet/if_ether.c index da7946d..9a67712 100644 --- a/sys/netinet/if_ether.c +++ b/sys/netinet/if_ether.c @@ -65,6 +65,7 @@ __FBSDID("$FreeBSD$"); #include <netinet/in.h> #include <netinet/in_var.h> +#include <net/if_llatbl.h> #include <netinet/if_ether.h> #include <netinet/vinet.h> @@ -79,6 +80,7 @@ __FBSDID("$FreeBSD$"); #define SIN(s) ((struct sockaddr_in *)s) #define SDL(s) ((struct sockaddr_dl *)s) +#define LLTABLE(ifp) ((struct lltable *)(ifp)->if_afdata[AF_INET]) SYSCTL_DECL(_net_link_ether); SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, ""); @@ -87,23 +89,13 @@ SYSCTL_NODE(_net_link_ether, PF_INET, inet, CTLFLAG_RW, 0, ""); #ifdef VIMAGE_GLOBALS static int arpt_keep; /* once resolved, good for 20 more minutes */ static int arp_maxtries; -static int useloopback; /* use loopback interface for local traffic */ +int useloopback; /* use loopback interface for local traffic */ static int arp_proxyall; #endif SYSCTL_V_INT(V_NET, vnet_inet, _net_link_ether_inet, OID_AUTO, max_age, CTLFLAG_RW, arpt_keep, 0, "ARP entry lifetime in seconds"); -#define rt_expire rt_rmx.rmx_expire - -struct llinfo_arp { - struct callout la_timer; - struct rtentry *la_rt; - struct mbuf *la_hold; /* last packet until resolved/timeout */ - u_short la_preempt; /* countdown for pre-expiry arps */ - u_short la_asked; /* # requests sent */ -}; - static struct ifqueue arpintrq; SYSCTL_V_INT(V_NET, vnet_inet, _net_link_ether_inet, OID_AUTO, maxtries, @@ -117,196 +109,65 @@ SYSCTL_V_INT(V_NET, vnet_inet, _net_link_ether_inet, OID_AUTO, proxyall, "Enable proxy ARP for all suitable requests"); static void arp_init(void); -static void arp_rtrequest(int, struct rtentry *, struct rt_addrinfo *); -static void arprequest(struct ifnet *, +void arprequest(struct ifnet *, struct in_addr *, struct in_addr *, u_char *); static void arpintr(struct mbuf *); static void arptimer(void *); -static struct rtentry - *arplookup(u_long, int, int, int); #ifdef INET static void in_arpinput(struct mbuf *); #endif +#ifdef AF_INET +void arp_ifscrub(struct ifnet *ifp, uint32_t addr); + /* - * Timeout routine. + * called by in_ifscrub to remove entry from the table when + * the interface goes away */ -static void -arptimer(void *arg) +void +arp_ifscrub(struct ifnet *ifp, uint32_t addr) { - struct rtentry *rt = (struct rtentry *)arg; - - RT_LOCK_ASSERT(rt); - /* - * The lock is needed to close a theoretical race - * between spontaneous expiry and intentional removal. - * We still got an extra reference on rtentry, so can - * safely pass pointers to its contents. - */ - RT_UNLOCK(rt); - - in_rtrequest(RTM_DELETE, rt_key(rt), NULL, rt_mask(rt), 0, NULL, - rt->rt_fibnum); + struct sockaddr_in addr4; + + bzero((void *)&addr4, sizeof(addr4)); + addr4.sin_len = sizeof(addr4); + addr4.sin_family = AF_INET; + addr4.sin_addr.s_addr = addr; + IF_AFDATA_LOCK(ifp); + lla_lookup(LLTABLE(ifp), (LLE_DELETE | LLE_IFADDR), + (struct sockaddr *)&addr4); + IF_AFDATA_UNLOCK(ifp); } +#endif /* - * Parallel to llc_rtrequest. + * Timeout routine. Age arp_tab entries periodically. */ static void -arp_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info) +arptimer(void *arg) { - INIT_VNET_NET(curvnet); - INIT_VNET_INET(curvnet); - struct sockaddr *gate; - struct llinfo_arp *la; - static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; - struct in_ifaddr *ia; - struct ifaddr *ifa; - - RT_LOCK_ASSERT(rt); + struct ifnet *ifp; + struct llentry *lle = (struct llentry *)arg; - if (rt->rt_flags & RTF_GATEWAY) + if (lle == NULL) { + panic("%s: NULL entry!\n", __func__); return; - gate = rt->rt_gateway; - la = (struct llinfo_arp *)rt->rt_llinfo; - switch (req) { - - case RTM_ADD: - /* - * XXX: If this is a manually added route to interface - * such as older version of routed or gated might provide, - * restore cloning bit. - */ - if ((rt->rt_flags & RTF_HOST) == 0 && - rt_mask(rt) != NULL && - SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff) - rt->rt_flags |= RTF_CLONING; - if (rt->rt_flags & RTF_CLONING) { - /* - * Case 1: This route should come from a route to iface. - */ - rt_setgate(rt, rt_key(rt), - (struct sockaddr *)&null_sdl); - gate = rt->rt_gateway; - SDL(gate)->sdl_type = rt->rt_ifp->if_type; - SDL(gate)->sdl_index = rt->rt_ifp->if_index; - rt->rt_expire = time_uptime; - break; - } - /* Announce a new entry if requested. */ - if (rt->rt_flags & RTF_ANNOUNCE) - arprequest(rt->rt_ifp, - &SIN(rt_key(rt))->sin_addr, - &SIN(rt_key(rt))->sin_addr, - (u_char *)LLADDR(SDL(gate))); - /*FALLTHROUGH*/ - case RTM_RESOLVE: - if (gate->sa_family != AF_LINK || - gate->sa_len < sizeof(null_sdl)) { - log(LOG_DEBUG, "%s: bad gateway %s%s\n", __func__, - inet_ntoa(SIN(rt_key(rt))->sin_addr), - (gate->sa_family != AF_LINK) ? - " (!AF_LINK)": ""); - break; - } - SDL(gate)->sdl_type = rt->rt_ifp->if_type; - SDL(gate)->sdl_index = rt->rt_ifp->if_index; - if (la != 0) - break; /* This happens on a route change */ - /* - * Case 2: This route may come from cloning, or a manual route - * add with a LL address. - */ - R_Zalloc(la, struct llinfo_arp *, sizeof(*la)); - rt->rt_llinfo = (caddr_t)la; - if (la == 0) { - log(LOG_DEBUG, "%s: malloc failed\n", __func__); - break; - } - /* - * We are storing a route entry outside of radix tree. So, - * it can be found and accessed by other means than radix - * lookup. The routing code assumes that any rtentry detached - * from radix can be destroyed safely. To prevent this, we - * add an additional reference. - */ - RT_ADDREF(rt); - la->la_rt = rt; - rt->rt_flags |= RTF_LLINFO; - callout_init_mtx(&la->la_timer, &rt->rt_mtx, - CALLOUT_RETURNUNLOCKED); - -#ifdef INET + } + ifp = lle->lle_tbl->llt_ifp; + IF_AFDATA_LOCK(ifp); + LLE_WLOCK(lle); + if ((lle->la_flags & LLE_DELETED) || + (time_second >= lle->la_expire)) { + if (!callout_pending(&lle->la_timer) && + callout_active(&lle->la_timer)) + (void) llentry_free(lle); + } else { /* - * This keeps the multicast addresses from showing up - * in `arp -a' listings as unresolved. It's not actually - * functional. Then the same for broadcast. + * Still valid, just drop our reference */ - if (IN_MULTICAST(ntohl(SIN(rt_key(rt))->sin_addr.s_addr)) && - rt->rt_ifp->if_type != IFT_ARCNET) { - ETHER_MAP_IP_MULTICAST(&SIN(rt_key(rt))->sin_addr, - LLADDR(SDL(gate))); - SDL(gate)->sdl_alen = 6; - rt->rt_expire = 0; - } - if (in_broadcast(SIN(rt_key(rt))->sin_addr, rt->rt_ifp)) { - memcpy(LLADDR(SDL(gate)), rt->rt_ifp->if_broadcastaddr, - rt->rt_ifp->if_addrlen); - SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen; - rt->rt_expire = 0; - } -#endif - - TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { - if (ia->ia_ifp == rt->rt_ifp && - SIN(rt_key(rt))->sin_addr.s_addr == - (IA_SIN(ia))->sin_addr.s_addr) - break; - } - if (ia) { - /* - * This test used to be - * if (loif.if_flags & IFF_UP) - * It allowed local traffic to be forced - * through the hardware by configuring the loopback down. - * However, it causes problems during network configuration - * for boards that can't receive packets they send. - * It is now necessary to clear "useloopback" and remove - * the route to force traffic out to the hardware. - */ - rt->rt_expire = 0; - bcopy(IF_LLADDR(rt->rt_ifp), LLADDR(SDL(gate)), - SDL(gate)->sdl_alen = rt->rt_ifp->if_addrlen); - if (V_useloopback) { - rt->rt_ifp = V_loif; - rt->rt_rmx.rmx_mtu = V_loif->if_mtu; - } - - /* - * make sure to set rt->rt_ifa to the interface - * address we are using, otherwise we will have trouble - * with source address selection. - */ - ifa = &ia->ia_ifa; - if (ifa != rt->rt_ifa) { - IFAFREE(rt->rt_ifa); - IFAREF(ifa); - rt->rt_ifa = ifa; - } - } - break; - - case RTM_DELETE: - if (la == NULL) /* XXX: at least CARP does this. */ - break; - callout_stop(&la->la_timer); - rt->rt_llinfo = NULL; - rt->rt_flags &= ~RTF_LLINFO; - RT_REMREF(rt); - if (la->la_hold) - m_freem(la->la_hold); - Free((caddr_t)la); + LLE_FREE_LOCKED(lle); } + IF_AFDATA_UNLOCK(ifp); } /* @@ -315,14 +176,37 @@ arp_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info) * - arp header target ip address * - arp header source ethernet address */ -static void -arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip, +void +arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip, u_char *enaddr) { struct mbuf *m; struct arphdr *ah; struct sockaddr sa; + if (sip == NULL) { + /* XXX don't believe this can happen (or explain why) */ + /* + * The caller did not supply a source address, try to find + * a compatible one among those assigned to this interface. + */ + struct ifaddr *ifa; + + TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { + if (!ifa->ifa_addr || + ifa->ifa_addr->sa_family != AF_INET) + continue; + sip = &SIN(ifa->ifa_addr)->sin_addr; + if (0 == ((sip->s_addr ^ tip->s_addr) & + SIN(ifa->ifa_netmask)->sin_addr.s_addr) ) + break; /* found it. */ + } + if (sip == NULL) { + printf("%s: cannot find matching address\n", __func__); + return; + } + } + if ((m = m_gethdr(M_DONTWAIT, MT_DATA)) == NULL) return; m->m_len = sizeof(*ah) + 2*sizeof(struct in_addr) + @@ -345,8 +229,6 @@ arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip, sa.sa_len = 2; m->m_flags |= M_BCAST; (*ifp->if_output)(ifp, m, &sa, (struct rtentry *)0); - - return; } /* @@ -365,16 +247,15 @@ arprequest(struct ifnet *ifp, struct in_addr *sip, struct in_addr *tip, */ int arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m, - struct sockaddr *dst, u_char *desten) + struct sockaddr *dst, u_char *desten, struct llentry **lle) { INIT_VNET_INET(ifp->if_vnet); - struct llinfo_arp *la = NULL; - struct rtentry *rt = NULL; - struct sockaddr_dl *sdl; - int error; - int fibnum = -1; + struct llentry *la = 0; + u_int flags; + int error, renew; - if (m) { + *lle = NULL; + if (m != NULL) { if (m->m_flags & M_BCAST) { /* broadcast */ (void)memcpy(desten, @@ -386,110 +267,78 @@ arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m, ETHER_MAP_IP_MULTICAST(&SIN(dst)->sin_addr, desten); return (0); } - fibnum = M_GETFIB(m); } - if (rt0 != NULL) { - /* Look for a cached arp (ll) entry. */ - if (m == NULL) - fibnum = rt0->rt_fibnum; - error = rt_check(&rt, &rt0, dst); - if (error) { - m_freem(m); - return error; - } - la = (struct llinfo_arp *)rt->rt_llinfo; - if (la == NULL) - RT_UNLOCK(rt); - } + flags = (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) ? 0 : LLE_CREATE; - /* - * If we had no mbuf and no route, then hope the caller - * has a fib in mind because we are running out of ideas. - * I think this should not happen in current code. - * (kmacy would know). + /* XXXXX + * Since this function returns an llentry, the + * lock is held by the caller. + * XXX if caller is required to hold lock, assert it */ - if (fibnum == -1) - fibnum = curthread->td_proc->p_fibnum; /* last gasp */ - +retry: + IF_AFDATA_LOCK(ifp); + la = lla_lookup(LLTABLE(ifp), flags, dst); + IF_AFDATA_UNLOCK(ifp); if (la == NULL) { - /* - * We enter this block if rt0 was NULL, - * or if rt found by rt_check() didn't have llinfo. - * we should get a cloned route, which since it should - * come from the local interface should have a ll entry. - * It may be incomplete but that's ok. - */ - rt = arplookup(SIN(dst)->sin_addr.s_addr, 1, 0, fibnum); - if (rt == NULL) { - log(LOG_DEBUG, - "arpresolve: can't allocate route for %s\n", - inet_ntoa(SIN(dst)->sin_addr)); - m_freem(m); - return (EINVAL); /* XXX */ - } - la = (struct llinfo_arp *)rt->rt_llinfo; - if (la == NULL) { - RT_UNLOCK(rt); + if (flags & LLE_CREATE) log(LOG_DEBUG, "arpresolve: can't allocate llinfo for %s\n", inet_ntoa(SIN(dst)->sin_addr)); - m_freem(m); - return (EINVAL); /* XXX */ - } - } - sdl = SDL(rt->rt_gateway); - /* - * Check the address family and length is valid, the address - * is resolved; otherwise, try to resolve. - */ - if ((rt->rt_expire == 0 || rt->rt_expire > time_uptime) && - sdl->sdl_family == AF_LINK && sdl->sdl_alen != 0) { - - bcopy(LLADDR(sdl), desten, sdl->sdl_alen); + m_freem(m); + return (EINVAL); + } + if ((la->la_flags & LLE_VALID) && + ((la->la_flags & LLE_STATIC) || la->la_expire > time_uptime)) { + bcopy(&la->ll_addr, desten, ifp->if_addrlen); /* * If entry has an expiry time and it is approaching, - * send an ARP request. + * see if we need to send an ARP request within this + * arpt_down interval. */ - if ((rt->rt_expire != 0) && - (time_uptime + la->la_preempt > rt->rt_expire)) { - struct in_addr sin = - SIN(rt->rt_ifa->ifa_addr)->sin_addr; + if (!(la->la_flags & LLE_STATIC) && + time_uptime + la->la_preempt > la->la_expire) { + arprequest(ifp, NULL, + &SIN(dst)->sin_addr, IF_LLADDR(ifp)); la->la_preempt--; - RT_UNLOCK(rt); - arprequest(ifp, &sin, &SIN(dst)->sin_addr, - IF_LLADDR(ifp)); - return (0); - } - - RT_UNLOCK(rt); - return (0); - } - /* - * If ARP is disabled or static on this interface, stop. - * XXX - * Probably should not allocate empty llinfo struct if we are - * not going to be sending out an arp request. - */ - if (ifp->if_flags & (IFF_NOARP | IFF_STATICARP)) { - RT_UNLOCK(rt); + } + + *lle = la; + error = 0; + goto done; + } + + if (la->la_flags & LLE_STATIC) { /* should not happen! */ + log(LOG_DEBUG, "arpresolve: ouch, empty static llinfo for %s\n", + inet_ntoa(SIN(dst)->sin_addr)); m_freem(m); - return (EINVAL); + error = EINVAL; + goto done; + } + + renew = (la->la_asked == 0 || la->la_expire != time_uptime); + if ((renew || m != NULL) && (flags & LLE_EXCLUSIVE) == 0) { + flags |= LLE_EXCLUSIVE; + LLE_RUNLOCK(la); + goto retry; } /* * There is an arptab entry, but no ethernet address * response yet. Replace the held mbuf with this * latest one. */ - if (m) { - if (la->la_hold) + if (m != NULL) { + if (la->la_hold != NULL) m_freem(la->la_hold); la->la_hold = m; + if (renew == 0 && (flags & LLE_EXCLUSIVE)) { + flags &= ~LLE_EXCLUSIVE; + LLE_DOWNGRADE(la); + } + } - KASSERT(rt->rt_expire > 0, ("sending ARP request for static entry")); - /* * Return EWOULDBLOCK if we have tried less than arp_maxtries. It * will be masked by ether_output(). Return EHOSTDOWN/EHOSTUNREACH @@ -499,22 +348,24 @@ arpresolve(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m, if (la->la_asked < V_arp_maxtries) error = EWOULDBLOCK; /* First request. */ else - error = (rt == rt0) ? EHOSTDOWN : EHOSTUNREACH; + error = + (rt0->rt_flags & RTF_GATEWAY) ? EHOSTDOWN : EHOSTUNREACH; - if (la->la_asked == 0 || rt->rt_expire != time_uptime) { - struct in_addr sin = - SIN(rt->rt_ifa->ifa_addr)->sin_addr; - - rt->rt_expire = time_uptime; - callout_reset(&la->la_timer, hz, arptimer, rt); + if (renew) { + LLE_ADDREF(la); + la->la_expire = time_uptime; + callout_reset(&la->la_timer, hz, arptimer, la); la->la_asked++; - RT_UNLOCK(rt); - - arprequest(ifp, &sin, &SIN(dst)->sin_addr, + LLE_WUNLOCK(la); + arprequest(ifp, NULL, &SIN(dst)->sin_addr, IF_LLADDR(ifp)); - } else - RT_UNLOCK(rt); - + return (error); + } +done: + if (flags & LLE_EXCLUSIVE) + LLE_WUNLOCK(la); + else + LLE_RUNLOCK(la); return (error); } @@ -598,21 +449,17 @@ in_arpinput(struct mbuf *m) { struct arphdr *ah; struct ifnet *ifp = m->m_pkthdr.rcvif; - struct llinfo_arp *la; + struct llentry *la = NULL; struct rtentry *rt; struct ifaddr *ifa; struct in_ifaddr *ia; - struct sockaddr_dl *sdl; struct sockaddr sa; struct in_addr isaddr, itaddr, myaddr; - struct mbuf *hold; u_int8_t *enaddr = NULL; - int op, rif_len; + int op, flags; + struct mbuf *m0; int req_len; int bridged = 0, is_bridge = 0; - u_int fibnum; - u_int goodfib = 0; - int firstpass = 1; #ifdef DEV_CARP int carp_match = 0; #endif @@ -649,7 +496,7 @@ in_arpinput(struct mbuf *m) */ LIST_FOREACH(ia, INADDR_HASH(itaddr.s_addr), ia_hash) { if (((bridged && ia->ia_ifp->if_bridge != NULL) || - (ia->ia_ifp == ifp)) && + ia->ia_ifp == ifp) && itaddr.s_addr == ia->ia_addr.sin_addr.s_addr) goto match; #ifdef DEV_CARP @@ -663,7 +510,7 @@ in_arpinput(struct mbuf *m) } LIST_FOREACH(ia, INADDR_HASH(isaddr.s_addr), ia_hash) if (((bridged && ia->ia_ifp->if_bridge != NULL) || - (ia->ia_ifp == ifp)) && + ia->ia_ifp == ifp) && isaddr.s_addr == ia->ia_addr.sin_addr.s_addr) goto match; @@ -729,191 +576,94 @@ match: } if (ifp->if_flags & IFF_STATICARP) goto reply; - /* - * We look for any FIB that has this address to find - * the interface etc. - * For sanity checks that are FIB independent we abort the loop. - */ - for (fibnum = 0; fibnum < rt_numfibs; fibnum++) { - rt = arplookup(isaddr.s_addr, - itaddr.s_addr == myaddr.s_addr, 0, fibnum); - if (rt == NULL) - continue; - - sdl = SDL(rt->rt_gateway); - /* Only call this once */ - if (firstpass) { - sin.sin_addr.s_addr = isaddr.s_addr; - EVENTHANDLER_INVOKE(route_arp_update_event, rt, - ar_sha(ah), (struct sockaddr *)&sin); - } - - la = (struct llinfo_arp *)rt->rt_llinfo; - if (la == NULL) { - RT_UNLOCK(rt); - continue; - } - if (firstpass) { - /* The following is not an error when doing bridging. */ - if (!bridged && rt->rt_ifp != ifp + bzero(&sin, sizeof(sin)); + sin.sin_len = sizeof(struct sockaddr_in); + sin.sin_family = AF_INET; + sin.sin_addr = isaddr; + flags = (itaddr.s_addr == myaddr.s_addr) ? LLE_CREATE : 0; + flags |= LLE_EXCLUSIVE; + IF_AFDATA_LOCK(ifp); + la = lla_lookup(LLTABLE(ifp), flags, (struct sockaddr *)&sin); + IF_AFDATA_UNLOCK(ifp); + if (la != NULL) { + /* the following is not an error when doing bridging */ + if (!bridged && la->lle_tbl->llt_ifp != ifp #ifdef DEV_CARP - && (ifp->if_type != IFT_CARP || !carp_match) + && (ifp->if_type != IFT_CARP || !carp_match) #endif - ) { - if (log_arp_wrong_iface) - log(LOG_ERR, "arp: %s is on %s " - "but got reply from %*D " - "on %s\n", - inet_ntoa(isaddr), - rt->rt_ifp->if_xname, - ifp->if_addrlen, - (u_char *)ar_sha(ah), ":", - ifp->if_xname); - RT_UNLOCK(rt); - break; - } - if (sdl->sdl_alen && - bcmp(ar_sha(ah), LLADDR(sdl), sdl->sdl_alen)) { - if (rt->rt_expire) { - if (log_arp_movements) - log(LOG_INFO, - "arp: %s moved from %*D to %*D " - "on %s\n", - inet_ntoa(isaddr), - ifp->if_addrlen, - (u_char *)LLADDR(sdl), ":", - ifp->if_addrlen, - (u_char *)ar_sha(ah), ":", - ifp->if_xname); - } else { - RT_UNLOCK(rt); - if (log_arp_permanent_modify) - log(LOG_ERR, - "arp: %*D attempts to " - "modify permanent entry " - "for %s on %s\n", - ifp->if_addrlen, - (u_char *)ar_sha(ah), ":", - inet_ntoa(isaddr), - ifp->if_xname); - break; - } - } - /* - * sanity check for the address length. - * XXX this does not work for protocols - * with variable address length. -is - */ - if (sdl->sdl_alen && - sdl->sdl_alen != ah->ar_hln) { - log(LOG_WARNING, - "arp from %*D: new addr len %d, was %d", - ifp->if_addrlen, (u_char *) ar_sha(ah), - ":", ah->ar_hln, sdl->sdl_alen); + ) { + if (log_arp_wrong_iface) + log(LOG_ERR, "arp: %s is on %s " + "but got reply from %*D on %s\n", + inet_ntoa(isaddr), + la->lle_tbl->llt_ifp->if_xname, + ifp->if_addrlen, (u_char *)ar_sha(ah), ":", + ifp->if_xname); + goto reply; + } + if ((la->la_flags & LLE_VALID) && + bcmp(ar_sha(ah), &la->ll_addr, ifp->if_addrlen)) { + if (la->la_flags & LLE_STATIC) { + log(LOG_ERR, + "arp: %*D attempts to modify permanent " + "entry for %s on %s\n", + ifp->if_addrlen, (u_char *)ar_sha(ah), ":", + inet_ntoa(isaddr), ifp->if_xname); + goto reply; } - if (ifp->if_addrlen != ah->ar_hln) { - log(LOG_WARNING, - "arp from %*D: addr len: " - "new %d, i/f %d (ignored)", - ifp->if_addrlen, (u_char *) ar_sha(ah), - ":", ah->ar_hln, ifp->if_addrlen); - RT_UNLOCK(rt); - break; + if (log_arp_movements) { + log(LOG_INFO, "arp: %s moved from %*D " + "to %*D on %s\n", + inet_ntoa(isaddr), + ifp->if_addrlen, + (u_char *)&la->ll_addr, ":", + ifp->if_addrlen, (u_char *)ar_sha(ah), ":", + ifp->if_xname); } - firstpass = 0; - goodfib = fibnum; } - - /* Copy in the information received. */ - (void)memcpy(LLADDR(sdl), ar_sha(ah), - sdl->sdl_alen = ah->ar_hln); - /* - * If we receive an arp from a token-ring station over - * a token-ring nic then try to save the source routing info. - * XXXMRT Only minimal Token Ring support for MRT. - * Only do this on the first pass as if modifies the mbuf. - */ - if (ifp->if_type == IFT_ISO88025) { - struct iso88025_header *th = NULL; - struct iso88025_sockaddr_dl_data *trld; - - /* force the fib loop to end after this pass */ - fibnum = rt_numfibs - 1; - - th = (struct iso88025_header *)m->m_pkthdr.header; - trld = SDL_ISO88025(sdl); - rif_len = TR_RCF_RIFLEN(th->rcf); - if ((th->iso88025_shost[0] & TR_RII) && - (rif_len > 2)) { - trld->trld_rcf = th->rcf; - trld->trld_rcf ^= htons(TR_RCF_DIR); - memcpy(trld->trld_route, th->rd, rif_len - 2); - trld->trld_rcf &= ~htons(TR_RCF_BCST_MASK); - /* - * Set up source routing information for - * reply packet (XXX) - */ - m->m_data -= rif_len; - m->m_len += rif_len; - m->m_pkthdr.len += rif_len; - } else { - th->iso88025_shost[0] &= ~TR_RII; - trld->trld_rcf = 0; - } - m->m_data -= 8; - m->m_len += 8; - m->m_pkthdr.len += 8; - th->rcf = trld->trld_rcf; + + if (ifp->if_addrlen != ah->ar_hln) { + log(LOG_WARNING, + "arp from %*D: addr len: new %d, i/f %d (ignored)", + ifp->if_addrlen, (u_char *) ar_sha(ah), ":", + ah->ar_hln, ifp->if_addrlen); + goto reply; } + (void)memcpy(&la->ll_addr, ar_sha(ah), ifp->if_addrlen); + la->la_flags |= LLE_VALID; - if (rt->rt_expire) { - rt->rt_expire = time_uptime + V_arpt_keep; + if (!(la->la_flags & LLE_STATIC)) { + la->la_expire = time_uptime + V_arpt_keep; callout_reset(&la->la_timer, hz * V_arpt_keep, - arptimer, rt); + arptimer, la); } la->la_asked = 0; la->la_preempt = V_arp_maxtries; - hold = la->la_hold; - la->la_hold = NULL; - RT_UNLOCK(rt); - if (hold != NULL) - (*ifp->if_output)(ifp, hold, rt_key(rt), rt); - } /* end of FIB loop */ + if (la->la_hold != NULL) { + m0 = la->la_hold; + la->la_hold = 0; + memcpy(&sa, L3_ADDR(la), sizeof(sa)); + LLE_WUNLOCK(la); + + (*ifp->if_output)(ifp, m0, &sa, NULL); + return; + } + } reply: - - /* - * Decide if we have to respond to something. - */ if (op != ARPOP_REQUEST) goto drop; + if (itaddr.s_addr == myaddr.s_addr) { /* Shortcut.. the receiving interface is the target. */ (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); (void)memcpy(ar_sha(ah), enaddr, ah->ar_hln); } else { - /* It's not asking for our address. But it still may - * be something we should answer. - * - * XXX MRT - * We assume that link level info is independent of - * the table used and so we use whichever we can and don't - * have a better option. - */ - /* Have we been asked to proxy for the target. */ - rt = arplookup(itaddr.s_addr, 0, SIN_PROXY, goodfib); - if (rt == NULL) { - /* Nope, only intersted now if proxying everything. */ - struct sockaddr_in sin; - + if (la == NULL) { if (!V_arp_proxyall) goto drop; - bzero(&sin, sizeof sin); - sin.sin_family = AF_INET; - sin.sin_len = sizeof sin; sin.sin_addr = itaddr; - /* XXX MRT use table 0 for arp reply */ rt = in_rtalloc1((struct sockaddr *)&sin, 0, 0UL, 0); if (!rt) @@ -964,19 +714,18 @@ reply: * Otherwise we may conflict with the host we are * proxying for. */ - if (rt->rt_ifp != ifp && - (rt->rt_ifp->if_bridge != ifp->if_bridge || + if (la->lle_tbl->llt_ifp != ifp && + (la->lle_tbl->llt_ifp->if_bridge != ifp->if_bridge || ifp->if_bridge == NULL)) { - RT_UNLOCK(rt); goto drop; } - sdl = SDL(rt->rt_gateway); (void)memcpy(ar_tha(ah), ar_sha(ah), ah->ar_hln); - (void)memcpy(ar_sha(ah), LLADDR(sdl), ah->ar_hln); - RT_UNLOCK(rt); + (void)memcpy(ar_sha(ah), &la->ll_addr, ah->ar_hln); } } + if (la != NULL) + LLE_WUNLOCK(la); if (itaddr.s_addr == myaddr.s_addr && IN_LINKLOCAL(ntohl(itaddr.s_addr))) { /* RFC 3927 link-local IPv4; always reply by broadcast. */ @@ -1002,68 +751,34 @@ reply: return; drop: + if (la != NULL) + LLE_WUNLOCK(la); m_freem(m); } #endif -/* - * Lookup or enter a new address in arptab. - */ -static struct rtentry * -arplookup(u_long addr, int create, int proxy, int fibnum) -{ - struct rtentry *rt; - struct sockaddr_inarp sin; - const char *why = 0; - - bzero(&sin, sizeof(sin)); - sin.sin_len = sizeof(sin); - sin.sin_family = AF_INET; - sin.sin_addr.s_addr = addr; - if (proxy) - sin.sin_other = SIN_PROXY; - rt = in_rtalloc1((struct sockaddr *)&sin, create, 0UL, fibnum); - if (rt == 0) - return (0); - - if (rt->rt_flags & RTF_GATEWAY) - why = "host is not on local network"; - else if ((rt->rt_flags & RTF_LLINFO) == 0) - why = "could not allocate llinfo"; - else if (rt->rt_gateway->sa_family != AF_LINK) - why = "gateway route is not ours"; - - if (why) { -#define ISDYNCLONE(_rt) \ - (((_rt)->rt_flags & (RTF_STATIC | RTF_WASCLONED)) == RTF_WASCLONED) - if (create) - log(LOG_DEBUG, "arplookup %s failed: %s\n", - inet_ntoa(sin.sin_addr), why); - /* - * If there are no references to this Layer 2 route, - * and it is a cloned route, and not static, and - * arplookup() is creating the route, then purge - * it from the routing table as it is probably bogus. - */ - if (rt->rt_refcnt == 1 && ISDYNCLONE(rt)) - rtexpunge(rt); - RTFREE_LOCKED(rt); - return (0); -#undef ISDYNCLONE - } else { - RT_REMREF(rt); - return (rt); - } -} - void arp_ifinit(struct ifnet *ifp, struct ifaddr *ifa) { + struct llentry *lle; + if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) arprequest(ifp, &IA_SIN(ifa)->sin_addr, &IA_SIN(ifa)->sin_addr, IF_LLADDR(ifp)); - ifa->ifa_rtrequest = arp_rtrequest; - ifa->ifa_flags |= RTF_CLONING; + /* + * interface address is considered static entry + * because the output of the arp utility shows + * that L2 entry as permanent + */ + IF_AFDATA_LOCK(ifp); + lle = lla_lookup(LLTABLE(ifp), (LLE_CREATE | LLE_IFADDR | LLE_STATIC), + (struct sockaddr *)IA_SIN(ifa)); + IF_AFDATA_UNLOCK(ifp); + if (lle == NULL) + log(LOG_INFO, "arp_ifinit: cannot create arp " + "entry for interface address\n"); + LLE_RUNLOCK(lle); + ifa->ifa_rtrequest = NULL; } void @@ -1072,8 +787,7 @@ arp_ifinit2(struct ifnet *ifp, struct ifaddr *ifa, u_char *enaddr) if (ntohl(IA_SIN(ifa)->sin_addr.s_addr) != INADDR_ANY) arprequest(ifp, &IA_SIN(ifa)->sin_addr, &IA_SIN(ifa)->sin_addr, enaddr); - ifa->ifa_rtrequest = arp_rtrequest; - ifa->ifa_flags |= RTF_CLONING; + ifa->ifa_rtrequest = NULL; } static void diff --git a/sys/netinet/if_ether.h b/sys/netinet/if_ether.h index 14df15f..ba4e2f6 100644 --- a/sys/netinet/if_ether.h +++ b/sys/netinet/if_ether.h @@ -109,8 +109,11 @@ struct sockaddr_inarp { extern u_char ether_ipmulticast_min[ETHER_ADDR_LEN]; extern u_char ether_ipmulticast_max[ETHER_ADDR_LEN]; +struct llentry; + int arpresolve(struct ifnet *ifp, struct rtentry *rt, - struct mbuf *m, struct sockaddr *dst, u_char *desten); + struct mbuf *m, struct sockaddr *dst, u_char *desten, + struct llentry **lle); void arp_ifinit(struct ifnet *, struct ifaddr *); void arp_ifinit2(struct ifnet *, struct ifaddr *, u_char *); #endif diff --git a/sys/netinet/in.c b/sys/netinet/in.c index 9b6dd48..852ef6d 100644 --- a/sys/netinet/in.c +++ b/sys/netinet/in.c @@ -46,6 +46,7 @@ __FBSDID("$FreeBSD$"); #include <sys/vimage.h> #include <net/if.h> +#include <net/if_llatbl.h> #include <net/if_types.h> #include <net/route.h> @@ -871,6 +872,8 @@ in_addprefix(struct in_ifaddr *target, int flags) return (error); } +extern void arp_ifscrub(struct ifnet *ifp, uint32_t addr); + /* * If there is no other address in the system that can serve a route to the * same prefix, remove the route. Hand over the route to the new address @@ -893,6 +896,8 @@ in_scrubprefix(struct in_ifaddr *target) prefix = target->ia_addr.sin_addr; mask = target->ia_sockmask.sin_addr; prefix.s_addr &= mask.s_addr; + /* remove arp cache */ + arp_ifscrub(target->ia_ifp, IA_SIN(target)->sin_addr.s_addr); } TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { @@ -1015,3 +1020,238 @@ in_ifdetach(struct ifnet *ifp) in_pcbpurgeif0(&V_udbinfo, ifp); in_purgemaddrs(ifp); } + +#include <sys/syslog.h> +#include <net/if_dl.h> +#include <netinet/if_ether.h> + +struct in_llentry { + struct llentry base; + struct sockaddr_in l3_addr4; +}; + +static struct llentry * +in_lltable_new(const struct sockaddr *l3addr, u_int flags) +{ + struct in_llentry *lle; + + lle = malloc(sizeof(struct in_llentry), M_LLTABLE, M_DONTWAIT | M_ZERO); + if (lle == NULL) /* NB: caller generates msg */ + return NULL; + + callout_init(&lle->base.la_timer, CALLOUT_MPSAFE); + /* + * For IPv4 this will trigger "arpresolve" to generate + * an ARP request. + */ + lle->base.la_expire = time_second; /* mark expired */ + lle->l3_addr4 = *(const struct sockaddr_in *)l3addr; + lle->base.lle_refcnt = 1; + LLE_LOCK_INIT(&lle->base); + return &lle->base; +} + +/* + * Deletes an address from the address table. + * This function is called by the timer functions + * such as arptimer() and nd6_llinfo_timer(), and + * the caller does the locking. + */ +static void +in_lltable_free(struct lltable *llt, struct llentry *lle) +{ + free(lle, M_LLTABLE); +} + +static int +in_lltable_rtcheck(struct ifnet *ifp, const struct sockaddr *l3addr) +{ + struct rtentry *rt; + + KASSERT(l3addr->sa_family == AF_INET, + ("sin_family %d", l3addr->sa_family)); + + /* XXX rtalloc1 should take a const param */ + rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0); + if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) { + log(LOG_INFO, "IPv4 address: \"%s\" is not on the network\n", + inet_ntoa(((const struct sockaddr_in *)l3addr)->sin_addr)); + if (rt != NULL) + RTFREE_LOCKED(rt); + return (EINVAL); + } + RTFREE_LOCKED(rt); + return 0; +} + +/* + * Return NULL if not found or marked for deletion. + * If found return lle read locked. + */ +static struct llentry * +in_lltable_lookup(struct lltable *llt, u_int flags, const struct sockaddr *l3addr) +{ + const struct sockaddr_in *sin = (const struct sockaddr_in *)l3addr; + struct ifnet *ifp = llt->llt_ifp; + struct llentry *lle; + struct llentries *lleh; + u_int hashkey; + + IF_AFDATA_LOCK_ASSERT(ifp); + KASSERT(l3addr->sa_family == AF_INET, + ("sin_family %d", l3addr->sa_family)); + + hashkey = sin->sin_addr.s_addr; + lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)]; + LIST_FOREACH(lle, lleh, lle_next) { + if (lle->la_flags & LLE_DELETED) + continue; + if (bcmp(L3_ADDR(lle), l3addr, sizeof(struct sockaddr_in)) == 0) + break; + } + if (lle == NULL) { +#ifdef DIAGNOSTICS + if (flags & LLE_DELETE) + log(LOG_INFO, "interface address is missing from cache = %p in delete\n", lle); +#endif + if (!(flags & LLE_CREATE)) + return (NULL); + /* + * A route that covers the given address must have + * been installed 1st because we are doing a resolution, + * verify this. + */ + if (!(flags & LLE_IFADDR) && + in_lltable_rtcheck(ifp, l3addr) != 0) + goto done; + + lle = in_lltable_new(l3addr, flags); + if (lle == NULL) { + log(LOG_INFO, "lla_lookup: new lle malloc failed\n"); + goto done; + } + lle->la_flags = flags & ~LLE_CREATE; + if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) { + bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen); + lle->la_flags |= (LLE_VALID | LLE_STATIC); + } + + lle->lle_tbl = llt; + lle->lle_head = lleh; + LIST_INSERT_HEAD(lleh, lle, lle_next); + } else if (flags & LLE_DELETE) { + if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) { + LLE_WLOCK(lle); + lle->la_flags = LLE_DELETED; + LLE_WUNLOCK(lle); +#ifdef DIAGNOSTICS + log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle); +#endif + } + lle = (void *)-1; + + } + if (lle != NULL && lle != (void *)-1) { + if (flags & LLE_EXCLUSIVE) + LLE_WLOCK(lle); + else + LLE_RLOCK(lle); + } +done: + return (lle); +} + +static int +in_lltable_dump(struct lltable *llt, struct sysctl_req *wr) +{ +#define SIN(lle) ((struct sockaddr_in *) L3_ADDR(lle)) + struct ifnet *ifp = llt->llt_ifp; + struct llentry *lle; + /* XXX stack use */ + struct { + struct rt_msghdr rtm; + struct sockaddr_inarp sin; + struct sockaddr_dl sdl; + } arpc; + int error, i; + + /* XXXXX + * current IFNET_RLOCK() is mapped to IFNET_WLOCK() + * so it is okay to use this ASSERT, change it when + * IFNET lock is finalized + */ + IFNET_WLOCK_ASSERT(); + + error = 0; + for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) { + LIST_FOREACH(lle, &llt->lle_head[i], lle_next) { + struct sockaddr_dl *sdl; + + /* skip deleted entries */ + if ((lle->la_flags & (LLE_DELETED|LLE_VALID)) != LLE_VALID) + continue; + /* + * produce a msg made of: + * struct rt_msghdr; + * struct sockaddr_inarp; (IPv4) + * struct sockaddr_dl; + */ + bzero(&arpc, sizeof(arpc)); + arpc.rtm.rtm_msglen = sizeof(arpc); + arpc.sin.sin_family = AF_INET; + arpc.sin.sin_len = sizeof(arpc.sin); + arpc.sin.sin_addr.s_addr = SIN(lle)->sin_addr.s_addr; + + /* publish */ + if (lle->la_flags & LLE_PUB) { + arpc.rtm.rtm_flags |= RTF_ANNOUNCE; + /* proxy only */ + if (lle->la_flags & LLE_PROXY) + arpc.sin.sin_other = SIN_PROXY; + } + + sdl = &arpc.sdl; + sdl->sdl_family = AF_LINK; + sdl->sdl_len = sizeof(*sdl); + sdl->sdl_alen = ifp->if_addrlen; + sdl->sdl_index = ifp->if_index; + sdl->sdl_type = ifp->if_type; + bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen); + + arpc.rtm.rtm_rmx.rmx_expire = + lle->la_flags & LLE_STATIC ? 0 : lle->la_expire; + arpc.rtm.rtm_flags |= RTF_HOST; + if (lle->la_flags & LLE_STATIC) + arpc.rtm.rtm_flags |= RTF_STATIC; + arpc.rtm.rtm_index = ifp->if_index; + error = SYSCTL_OUT(wr, &arpc, sizeof(arpc)); + if (error) + break; + } + } + return error; +#undef SIN +} + +void * +in_domifattach(struct ifnet *ifp) +{ + struct lltable *llt = lltable_init(ifp, AF_INET); + + if (llt != NULL) { + llt->llt_new = in_lltable_new; + llt->llt_free = in_lltable_free; + llt->llt_rtcheck = in_lltable_rtcheck; + llt->llt_lookup = in_lltable_lookup; + llt->llt_dump = in_lltable_dump; + } + return (llt); +} + +void +in_domifdetach(struct ifnet *ifp __unused, void *aux) +{ + struct lltable *llt = (struct lltable *)aux; + + lltable_free(llt); +} diff --git a/sys/netinet/in_mcast.c b/sys/netinet/in_mcast.c index e7916dd..6f5dfe1 100644 --- a/sys/netinet/in_mcast.c +++ b/sys/netinet/in_mcast.c @@ -1036,7 +1036,7 @@ inp_join_group(struct inpcb *inp, struct sockopt *sopt) ro.ro_rt = NULL; *(struct sockaddr_in *)&ro.ro_dst = gsa->sin; - in_rtalloc_ign(&ro, RTF_CLONING, + in_rtalloc_ign(&ro, 0, inp->inp_inc.inc_fibnum); if (ro.ro_rt != NULL) { ifp = ro.ro_rt->rt_ifp; diff --git a/sys/netinet/in_pcb.c b/sys/netinet/in_pcb.c index c5515da..84bff92 100644 --- a/sys/netinet/in_pcb.c +++ b/sys/netinet/in_pcb.c @@ -570,7 +570,7 @@ in_pcbladdr(struct inpcb *inp, struct in_addr *faddr, struct in_addr *laddr, * Find out route to destination. */ if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0) - in_rtalloc_ign(&sro, RTF_CLONING, inp->inp_inc.inc_fibnum); + in_rtalloc_ign(&sro, 0, inp->inp_inc.inc_fibnum); /* * If we found a route, use the address corresponding to diff --git a/sys/netinet/in_proto.c b/sys/netinet/in_proto.c index a93f1f2..f3fbe0c 100644 --- a/sys/netinet/in_proto.c +++ b/sys/netinet/in_proto.c @@ -59,6 +59,7 @@ __FBSDID("$FreeBSD$"); #include <netinet/in.h> #include <netinet/in_systm.h> +#include <netinet/in_var.h> #include <netinet/ip.h> #include <netinet/ip_var.h> #include <netinet/ip_icmp.h> @@ -364,7 +365,9 @@ struct domain inetdomain = { .dom_rtattach = in_inithead, #endif .dom_rtoffset = 32, - .dom_maxrtkey = sizeof(struct sockaddr_in) + .dom_maxrtkey = sizeof(struct sockaddr_in), + .dom_ifattach = in_domifattach, + .dom_ifdetach = in_domifdetach }; DOMAIN_SET(inet); diff --git a/sys/netinet/in_rmx.c b/sys/netinet/in_rmx.c index e61e36a..f9e9d98 100644 --- a/sys/netinet/in_rmx.c +++ b/sys/netinet/in_rmx.c @@ -75,8 +75,8 @@ in_addroute(void *v_arg, void *n_arg, struct radix_node_head *head, { struct rtentry *rt = (struct rtentry *)treenodes; struct sockaddr_in *sin = (struct sockaddr_in *)rt_key(rt); - struct radix_node *ret; + RADIX_NODE_HEAD_WLOCK_ASSERT(head); /* * A little bit of help for both IP output and input: * For host routes, we make sure that RTF_BROADCAST @@ -106,31 +106,7 @@ in_addroute(void *v_arg, void *n_arg, struct radix_node_head *head, if (!rt->rt_rmx.rmx_mtu && rt->rt_ifp) 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) { - 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 = in_rtalloc1((struct sockaddr *)sin, 0, - RTF_CLONING|RTF_RNH_LOCKED, rt->rt_fibnum); - if (rt2) { - if (rt2->rt_flags & RTF_LLINFO && - rt2->rt_flags & RTF_HOST && - rt2->rt_gateway && - rt2->rt_gateway->sa_family == AF_LINK) { - rtexpunge(rt2); - RTFREE_LOCKED(rt2); - ret = rn_addroute(v_arg, n_arg, head, - treenodes); - } else - RTFREE_LOCKED(rt2); - } - } - - return ret; + return (rn_addroute(v_arg, n_arg, head, treenodes)); } /* @@ -187,13 +163,10 @@ in_clsroute(struct radix_node *rn, struct radix_node_head *head) if (!(rt->rt_flags & RTF_UP)) return; /* prophylactic measures */ - if ((rt->rt_flags & (RTF_LLINFO | RTF_HOST)) != RTF_HOST) - return; - if (rt->rt_flags & RTPRF_OURS) return; - if (!(rt->rt_flags & (RTF_WASCLONED | RTF_DYNAMIC))) + if (!(rt->rt_flags & RTF_DYNAMIC)) return; /* @@ -434,7 +407,6 @@ in_ifadownkill(struct radix_node *rn, void *xap) * the routes that rtrequest() would have in any case, * so that behavior is not needed there. */ - rt->rt_flags &= ~RTF_CLONING; rtexpunge(rt); } RT_UNLOCK(rt); diff --git a/sys/netinet/in_var.h b/sys/netinet/in_var.h index 7facfa8..73868ad 100644 --- a/sys/netinet/in_var.h +++ b/sys/netinet/in_var.h @@ -321,6 +321,9 @@ void ip_input(struct mbuf *); int in_ifadown(struct ifaddr *ifa, int); void in_ifscrub(struct ifnet *, struct in_ifaddr *); struct mbuf *ip_fastforward(struct mbuf *); +void *in_domifattach(struct ifnet *); +void in_domifdetach(struct ifnet *, void *); + /* XXX */ void in_rtalloc_ign(struct route *ro, u_long ignflags, u_int fibnum); diff --git a/sys/netinet/ip_carp.c b/sys/netinet/ip_carp.c index 9f6bc31..6d96163 100644 --- a/sys/netinet/ip_carp.c +++ b/sys/netinet/ip_carp.c @@ -371,14 +371,6 @@ carp_setroute(struct carp_softc *sc, int cmd) (cmd == RTM_DELETE && count == 0)) rtinit(ifa, cmd, RTF_UP | RTF_HOST); } -#ifdef INET6 - if (ifa->ifa_addr->sa_family == AF_INET6) { - if (cmd == RTM_ADD) - in6_ifaddloop(ifa); - else - in6_ifremloop(ifa); - } -#endif /* INET6 */ } splx(s); } diff --git a/sys/netinet/ip_fastfwd.c b/sys/netinet/ip_fastfwd.c index 407b08c..ea106df 100644 --- a/sys/netinet/ip_fastfwd.c +++ b/sys/netinet/ip_fastfwd.c @@ -128,7 +128,7 @@ ip_findroute(struct route *ro, struct in_addr dest, struct mbuf *m) dst->sin_family = AF_INET; dst->sin_len = sizeof(*dst); dst->sin_addr.s_addr = dest.s_addr; - in_rtalloc_ign(ro, RTF_CLONING, M_GETFIB(m)); + in_rtalloc_ign(ro, 0, M_GETFIB(m)); /* * Route there and interface still up? diff --git a/sys/netinet/ip_fw2.c b/sys/netinet/ip_fw2.c index e91fc9b..eae49c8 100644 --- a/sys/netinet/ip_fw2.c +++ b/sys/netinet/ip_fw2.c @@ -528,7 +528,7 @@ verify_path(struct in_addr src, struct ifnet *ifp, u_int fib) dst->sin_family = AF_INET; dst->sin_len = sizeof(*dst); dst->sin_addr = src; - in_rtalloc_ign(&ro, RTF_CLONING, fib); + in_rtalloc_ign(&ro, 0, fib); if (ro.ro_rt == NULL) return 0; @@ -620,7 +620,7 @@ verify_path6(struct in6_addr *src, struct ifnet *ifp) dst->sin6_len = sizeof(*dst); dst->sin6_addr = *src; /* XXX MRT 0 for ipv6 at this time */ - rtalloc_ign((struct route *)&ro, RTF_CLONING); + rtalloc_ign((struct route *)&ro, 0); if (ro.ro_rt == NULL) return 0; diff --git a/sys/netinet/ip_input.c b/sys/netinet/ip_input.c index c99c53e..4a169ff 100644 --- a/sys/netinet/ip_input.c +++ b/sys/netinet/ip_input.c @@ -1286,7 +1286,7 @@ ip_rtaddr(struct in_addr dst, u_int fibnum) sin->sin_family = AF_INET; sin->sin_len = sizeof(*sin); sin->sin_addr = dst; - in_rtalloc_ign(&sro, RTF_CLONING, fibnum); + in_rtalloc_ign(&sro, 0, fibnum); if (sro.ro_rt == NULL) return (NULL); @@ -1412,7 +1412,7 @@ ip_forward(struct mbuf *m, int srcrt) sin->sin_family = AF_INET; sin->sin_len = sizeof(*sin); sin->sin_addr = ip->ip_dst; - in_rtalloc_ign(&ro, RTF_CLONING, M_GETFIB(m)); + in_rtalloc_ign(&ro, 0, M_GETFIB(m)); rt = ro.ro_rt; diff --git a/sys/netinet/ip_output.c b/sys/netinet/ip_output.c index 03cf56f..6cda8aa 100644 --- a/sys/netinet/ip_output.c +++ b/sys/netinet/ip_output.c @@ -567,7 +567,6 @@ passout: * to avoid confusing lower layers. */ m->m_flags &= ~(M_PROTOFLAGS); - error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, ro->ro_rt); goto done; diff --git a/sys/netinet/tcp_subr.c b/sys/netinet/tcp_subr.c index 2c2337e..e7a6dff 100644 --- a/sys/netinet/tcp_subr.c +++ b/sys/netinet/tcp_subr.c @@ -1659,7 +1659,7 @@ tcp_maxmtu(struct in_conninfo *inc, int *flags) dst->sin_family = AF_INET; dst->sin_len = sizeof(*dst); dst->sin_addr = inc->inc_faddr; - in_rtalloc_ign(&sro, RTF_CLONING, inc->inc_fibnum); + in_rtalloc_ign(&sro, 0, inc->inc_fibnum); } if (sro.ro_rt != NULL) { ifp = sro.ro_rt->rt_ifp; @@ -1694,7 +1694,7 @@ tcp_maxmtu6(struct in_conninfo *inc, int *flags) sro6.ro_dst.sin6_family = AF_INET6; sro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6); sro6.ro_dst.sin6_addr = inc->inc6_faddr; - rtalloc_ign((struct route *)&sro6, RTF_CLONING); + rtalloc_ign((struct route *)&sro6, 0); } if (sro6.ro_rt != NULL) { ifp = sro6.ro_rt->rt_ifp; diff --git a/sys/netinet6/icmp6.c b/sys/netinet6/icmp6.c index 59bc95a..229aaa2 100644 --- a/sys/netinet6/icmp6.c +++ b/sys/netinet6/icmp6.c @@ -85,6 +85,7 @@ __FBSDID("$FreeBSD$"); #include <net/if.h> #include <net/if_dl.h> +#include <net/if_llatbl.h> #include <net/if_types.h> #include <net/route.h> #include <net/vnet.h> @@ -2573,32 +2574,32 @@ icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt) { /* target lladdr option */ - struct rtentry *rt_router = NULL; int len; - struct sockaddr_dl *sdl; + struct llentry *ln; struct nd_opt_hdr *nd_opt; char *lladdr; - rt_router = nd6_lookup(router_ll6, 0, ifp); - if (!rt_router) + IF_AFDATA_LOCK(ifp); + ln = nd6_lookup(router_ll6, 0, ifp); + IF_AFDATA_UNLOCK(ifp); + if (!ln) goto nolladdropt; + len = sizeof(*nd_opt) + ifp->if_addrlen; len = (len + 7) & ~7; /* round by 8 */ /* safety check */ if (len + (p - (u_char *)ip6) > maxlen) goto nolladdropt; - if (!(rt_router->rt_flags & RTF_GATEWAY) && - (rt_router->rt_flags & RTF_LLINFO) && - (rt_router->rt_gateway->sa_family == AF_LINK) && - (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) && - sdl->sdl_alen) { + + if (ln->la_flags & LLE_VALID) { nd_opt = (struct nd_opt_hdr *)p; nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR; nd_opt->nd_opt_len = len >> 3; lladdr = (char *)(nd_opt + 1); - bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen); + bcopy(&ln->ll_addr, lladdr, ifp->if_addrlen); p += len; } + LLE_RUNLOCK(ln); } nolladdropt:; diff --git a/sys/netinet6/in6.c b/sys/netinet6/in6.c index c784845..8127281 100644 --- a/sys/netinet6/in6.c +++ b/sys/netinet6/in6.c @@ -88,6 +88,7 @@ __FBSDID("$FreeBSD$"); #include <netinet/in.h> #include <netinet/in_var.h> +#include <net/if_llatbl.h> #include <netinet/if_ether.h> #include <netinet/in_systm.h> #include <netinet/ip.h> @@ -135,152 +136,7 @@ static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *); struct in6_multihead in6_multihead; /* XXX BSS initialization */ int (*faithprefix_p)(struct in6_addr *); -/* - * Subroutine for in6_ifaddloop() and in6_ifremloop(). - * This routine does actual work. - */ -static void -in6_ifloop_request(int cmd, struct ifaddr *ifa) -{ - struct sockaddr_in6 all1_sa; - struct rtentry *nrt = NULL; - int e; - char ip6buf[INET6_ADDRSTRLEN]; - - bzero(&all1_sa, sizeof(all1_sa)); - all1_sa.sin6_family = AF_INET6; - all1_sa.sin6_len = sizeof(struct sockaddr_in6); - all1_sa.sin6_addr = in6mask128; - - /* - * We specify the address itself as the gateway, and set the - * RTF_LLINFO flag, so that the corresponding host route would have - * the flag, and thus applications that assume traditional behavior - * would be happy. Note that we assume the caller of the function - * (probably implicitly) set nd6_rtrequest() to ifa->ifa_rtrequest, - * which changes the outgoing interface to the loopback interface. - */ - e = rtrequest(cmd, ifa->ifa_addr, ifa->ifa_addr, - (struct sockaddr *)&all1_sa, RTF_UP|RTF_HOST|RTF_LLINFO, &nrt); - if (e != 0) { - /* XXX need more descriptive message */ - - log(LOG_ERR, "in6_ifloop_request: " - "%s operation failed for %s (errno=%d)\n", - cmd == RTM_ADD ? "ADD" : "DELETE", - ip6_sprintf(ip6buf, - &((struct in6_ifaddr *)ifa)->ia_addr.sin6_addr), e); - } - - /* - * Report the addition/removal of the address to the routing socket. - * XXX: since we called rtinit for a p2p interface with a destination, - * we end up reporting twice in such a case. Should we rather - * omit the second report? - */ - if (nrt) { - RT_LOCK(nrt); - /* - * Make sure rt_ifa be equal to IFA, the second argument of - * the function. We need this because when we refer to - * rt_ifa->ia6_flags in ip6_input, we assume that the rt_ifa - * points to the address instead of the loopback address. - */ - if (cmd == RTM_ADD && ifa != nrt->rt_ifa) { - IFAFREE(nrt->rt_ifa); - IFAREF(ifa); - nrt->rt_ifa = ifa; - } - - rt_newaddrmsg(cmd, ifa, e, nrt); - if (cmd == RTM_DELETE) - RTFREE_LOCKED(nrt); - else { - /* the cmd must be RTM_ADD here */ - RT_REMREF(nrt); - RT_UNLOCK(nrt); - } - } -} - -/* - * Add ownaddr as loopback rtentry. We previously add the route only if - * necessary (ex. on a p2p link). However, since we now manage addresses - * separately from prefixes, we should always add the route. We can't - * rely on the cloning mechanism from the corresponding interface route - * any more. - */ -void -in6_ifaddloop(struct ifaddr *ifa) -{ - struct rtentry *rt; - int need_loop; - - /* If there is no loopback entry, allocate one. */ - rt = rtalloc1(ifa->ifa_addr, 0, 0); - need_loop = (rt == NULL || (rt->rt_flags & RTF_HOST) == 0 || - (rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0); - if (rt) - RTFREE_LOCKED(rt); - if (need_loop) - in6_ifloop_request(RTM_ADD, ifa); -} - -/* - * Remove loopback rtentry of ownaddr generated by in6_ifaddloop(), - * if it exists. - */ -void -in6_ifremloop(struct ifaddr *ifa) -{ - INIT_VNET_INET6(curvnet); - struct in6_ifaddr *ia; - struct rtentry *rt; - int ia_count = 0; - - /* - * Some of BSD variants do not remove cloned routes - * from an interface direct route, when removing the direct route - * (see comments in net/net_osdep.h). Even for variants that do remove - * cloned routes, they could fail to remove the cloned routes when - * we handle multple addresses that share a common prefix. - * So, we should remove the route corresponding to the deleted address. - */ - /* - * Delete the entry only if exact one ifa exists. More than one ifa - * can exist if we assign a same single address to multiple - * (probably p2p) interfaces. - * XXX: we should avoid such a configuration in IPv6... - */ - for (ia = V_in6_ifaddr; ia; ia = ia->ia_next) { - if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ia->ia_addr.sin6_addr)) { - ia_count++; - if (ia_count > 1) - break; - } - } - - if (ia_count == 1) { - /* - * Before deleting, check if a corresponding loopbacked host - * route surely exists. With this check, we can avoid to - * delete an interface direct route whose destination is same - * as the address being removed. This can happen when removing - * a subnet-router anycast address on an interface attahced - * to a shared medium. - */ - rt = rtalloc1(ifa->ifa_addr, 0, 0); - if (rt != NULL) { - if ((rt->rt_flags & RTF_HOST) != 0 && - (rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) { - RTFREE_LOCKED(rt); - in6_ifloop_request(RTM_DELETE, ifa); - } else - RT_UNLOCK(rt); - } - } -} int in6_mask2len(struct in6_addr *mask, u_char *lim0) @@ -1131,10 +987,9 @@ in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, } } if (!rt) { - /* XXX: we need RTF_CLONING to fake nd6_rtrequest */ error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr, (struct sockaddr *)&ia->ia_addr, - (struct sockaddr *)&mltmask, RTF_UP | RTF_CLONING, + (struct sockaddr *)&mltmask, RTF_UP, (struct rtentry **)0); if (error) goto cleanup; @@ -1208,7 +1063,7 @@ in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, if (!rt) { error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr, (struct sockaddr *)&ia->ia_addr, - (struct sockaddr *)&mltmask, RTF_UP | RTF_CLONING, + (struct sockaddr *)&mltmask, RTF_UP, (struct rtentry **)0); if (error) goto cleanup; @@ -1287,34 +1142,16 @@ in6_purgeaddr(struct ifaddr *ifa) { struct ifnet *ifp = ifa->ifa_ifp; struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa; - char ip6buf[INET6_ADDRSTRLEN]; struct in6_multi_mship *imm; /* stop DAD processing */ nd6_dad_stop(ifa); - /* - * delete route to the destination of the address being purged. - * The interface must be p2p or loopback in this case. - */ - if ((ia->ia_flags & IFA_ROUTE) != 0 && ia->ia_dstaddr.sin6_len != 0) { - int e; - - if ((e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST)) - != 0) { - log(LOG_ERR, "in6_purgeaddr: failed to remove " - "a route to the p2p destination: %s on %s, " - "errno=%d\n", - ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr), - if_name(ifp), e); - /* proceed anyway... */ - } else - ia->ia_flags &= ~IFA_ROUTE; - } - - /* Remove ownaddr's loopback rtentry, if it exists. */ - in6_ifremloop(&(ia->ia_ifa)); - + IF_AFDATA_LOCK(ifp); + lla_lookup(LLTABLE6(ifp), (LLE_DELETE | LLE_IFADDR), + (struct sockaddr *)&ia->ia_addr); + IF_AFDATA_UNLOCK(ifp); + /* * leave from multicast groups we have joined for the interface */ @@ -1688,26 +1525,15 @@ in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia, /* we could do in(6)_socktrim here, but just omit it at this moment. */ - if (newhost) { - /* - * set the rtrequest function to create llinfo. It also - * adjust outgoing interface of the route for the local - * address when called via in6_ifaddloop() below. - */ - ia->ia_ifa.ifa_rtrequest = nd6_rtrequest; - } - /* * Special case: * If a new destination address is specified for a point-to-point * interface, install a route to the destination as an interface - * direct route. In addition, if the link is expected to have neighbor - * cache entries, specify RTF_LLINFO so that a cache entry for the - * destination address will be created. - * created + * direct route. * XXX: the logic below rejects assigning multiple addresses on a p2p * interface that share the same destination. */ +#if 0 /* QL - verify */ plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */ if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 && ia->ia_dstaddr.sin6_family == AF_INET6) { @@ -1715,7 +1541,6 @@ in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia, struct rtentry *rt = NULL, **rtp = NULL; if (nd6_need_cache(ifp) != 0) { - rtflags |= RTF_LLINFO; rtp = &rt; } @@ -1744,16 +1569,36 @@ in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia, } ia->ia_flags |= IFA_ROUTE; } - if (plen < 128) { - /* - * The RTF_CLONING flag is necessary for in6_is_ifloop_auto(). - */ - ia->ia_ifa.ifa_flags |= RTF_CLONING; +#else + plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */ + if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 && + ia->ia_dstaddr.sin6_family == AF_INET6) { + if ((error = rtinit(&(ia->ia_ifa), (int)RTM_ADD, + RTF_UP | RTF_HOST)) != 0) + return (error); + ia->ia_flags |= IFA_ROUTE; } +#endif /* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */ - if (newhost) - in6_ifaddloop(&(ia->ia_ifa)); + if (newhost) { + struct llentry *ln; + + IF_AFDATA_LOCK(ifp); + ia->ia_ifa.ifa_rtrequest = NULL; + + /* XXX QL + * we need to report rt_newaddrmsg + */ + ln = lla_lookup(LLTABLE6(ifp), (LLE_CREATE | LLE_IFADDR | LLE_EXCLUSIVE), + (struct sockaddr *)&ia->ia_addr); + IF_AFDATA_UNLOCK(ifp); + if (ln) { + ln->la_expire = 0; /* for IPv6 this means permanent */ + ln->ln_state = ND6_LLINFO_REACHABLE; + LLE_WUNLOCK(ln); + } + } return (error); } @@ -2237,6 +2082,214 @@ in6_if2idlen(struct ifnet *ifp) } } +#include <sys/sysctl.h> + +struct in6_llentry { + struct llentry base; + struct sockaddr_in6 l3_addr6; +}; + +static struct llentry * +in6_lltable_new(const struct sockaddr *l3addr, u_int flags) +{ + struct in6_llentry *lle; + + lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, + M_DONTWAIT | M_ZERO); + if (lle == NULL) /* NB: caller generates msg */ + return NULL; + + callout_init(&lle->base.ln_timer_ch, CALLOUT_MPSAFE); + lle->l3_addr6 = *(const struct sockaddr_in6 *)l3addr; + lle->base.lle_refcnt = 1; + LLE_LOCK_INIT(&lle->base); + return &lle->base; +} + +/* + * Deletes an address from the address table. + * This function is called by the timer functions + * such as arptimer() and nd6_llinfo_timer(), and + * the caller does the locking. + */ +static void +in6_lltable_free(struct lltable *llt, struct llentry *lle) +{ + free(lle, M_LLTABLE); +} + +static int +in6_lltable_rtcheck(struct ifnet *ifp, const struct sockaddr *l3addr) +{ + struct rtentry *rt; + char ip6buf[INET6_ADDRSTRLEN]; + + KASSERT(l3addr->sa_family == AF_INET6, + ("sin_family %d", l3addr->sa_family)); + + /* XXX rtalloc1 should take a const param */ + rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0); + if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) { + struct ifaddr *ifa; + /* + * Create an ND6 cache for an IPv6 neighbor + * that is not covered by our own prefix. + */ + /* XXX ifaof_ifpforaddr should take a const param */ + ifa = ifaof_ifpforaddr(__DECONST(struct sockaddr *, l3addr), ifp); + if (ifa != NULL) { + if (rt != NULL) + rtfree(rt); + return 0; + } + log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n", + ip6_sprintf(ip6buf, &((const struct sockaddr_in6 *)l3addr)->sin6_addr)); + if (rt != NULL) + rtfree(rt); + return EINVAL; + } + rtfree(rt); + return 0; +} + +static struct llentry * +in6_lltable_lookup(struct lltable *llt, u_int flags, + const struct sockaddr *l3addr) +{ + const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr; + struct ifnet *ifp = llt->llt_ifp; + struct llentry *lle; + struct llentries *lleh; + u_int hashkey; + + IF_AFDATA_LOCK_ASSERT(ifp); + KASSERT(l3addr->sa_family == AF_INET6, + ("sin_family %d", l3addr->sa_family)); + + hashkey = sin6->sin6_addr.s6_addr32[3]; + lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)]; + LIST_FOREACH(lle, lleh, lle_next) { + if (lle->la_flags & LLE_DELETED) + continue; + if (bcmp(L3_ADDR(lle), l3addr, l3addr->sa_len) == 0) + break; + } + + if (lle == NULL) { + if (!(flags & LLE_CREATE)) + return (NULL); + /* + * A route that covers the given address must have + * been installed 1st because we are doing a resolution, + * verify this. + */ + if (!(flags & LLE_IFADDR) && + in6_lltable_rtcheck(ifp, l3addr) != 0) + return NULL; + + lle = in6_lltable_new(l3addr, flags); + if (lle == NULL) { + log(LOG_INFO, "lla_lookup: new lle malloc failed\n"); + return NULL; + } + lle->la_flags = flags & ~LLE_CREATE; + if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) { + bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen); + lle->la_flags |= (LLE_VALID | LLE_STATIC); + } + + lle->lle_tbl = llt; + lle->lle_head = lleh; + LIST_INSERT_HEAD(lleh, lle, lle_next); + } else if (flags & LLE_DELETE) { + LLE_WLOCK(lle); + lle->la_flags = LLE_DELETED; + LLE_WUNLOCK(lle); +#ifdef DIAGNOSTICS + log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle); +#endif + lle = (void *)-1; + } + if (LLE_IS_VALID(lle)) { + if (flags & LLE_EXCLUSIVE) + LLE_WLOCK(lle); + else + LLE_RLOCK(lle); + } + return (lle); +} + +static int +in6_lltable_dump(struct lltable *llt, struct sysctl_req *wr) +{ + struct ifnet *ifp = llt->llt_ifp; + struct llentry *lle; + /* XXX stack use */ + struct { + struct rt_msghdr rtm; + struct sockaddr_in6 sin6; + /* + * ndp.c assumes that sdl is word aligned + */ +#ifdef __LP64__ + uint32_t pad; +#endif + struct sockaddr_dl sdl; + } ndpc; + int i, error; + + /* XXXXX + * current IFNET_RLOCK() is mapped to IFNET_WLOCK() + * so it is okay to use this ASSERT, change it when + * IFNET lock is finalized + */ + IFNET_WLOCK_ASSERT(); + + error = 0; + for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) { + LIST_FOREACH(lle, &llt->lle_head[i], lle_next) { + struct sockaddr_dl *sdl; + + /* skip deleted or invalid entries */ + if ((lle->la_flags & (LLE_DELETED|LLE_VALID)) != LLE_VALID) + continue; + /* + * produce a msg made of: + * struct rt_msghdr; + * struct sockaddr_in6 (IPv6) + * struct sockaddr_dl; + */ + bzero(&ndpc, sizeof(ndpc)); + ndpc.rtm.rtm_msglen = sizeof(ndpc); + ndpc.sin6.sin6_family = AF_INET6; + ndpc.sin6.sin6_len = sizeof(ndpc.sin6); + bcopy(L3_ADDR(lle), &ndpc.sin6, L3_ADDR_LEN(lle)); + + /* publish */ + if (lle->la_flags & LLE_PUB) + ndpc.rtm.rtm_flags |= RTF_ANNOUNCE; + + sdl = &ndpc.sdl; + sdl->sdl_family = AF_LINK; + sdl->sdl_len = sizeof(*sdl); + sdl->sdl_alen = ifp->if_addrlen; + sdl->sdl_index = ifp->if_index; + sdl->sdl_type = ifp->if_type; + bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen); + ndpc.rtm.rtm_rmx.rmx_expire = + lle->la_flags & LLE_STATIC ? 0 : lle->la_expire; + ndpc.rtm.rtm_flags |= RTF_HOST; + if (lle->la_flags & LLE_STATIC) + ndpc.rtm.rtm_flags |= RTF_STATIC; + ndpc.rtm.rtm_index = ifp->if_index; + error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc)); + if (error) + break; + } + } + return error; +} + void * in6_domifattach(struct ifnet *ifp) { @@ -2256,6 +2309,14 @@ in6_domifattach(struct ifnet *ifp) ext->nd_ifinfo = nd6_ifattach(ifp); ext->scope6_id = scope6_ifattach(ifp); + ext->lltable = lltable_init(ifp, AF_INET6); + if (ext->lltable != NULL) { + ext->lltable->llt_new = in6_lltable_new; + ext->lltable->llt_free = in6_lltable_free; + ext->lltable->llt_rtcheck = in6_lltable_rtcheck; + ext->lltable->llt_lookup = in6_lltable_lookup; + ext->lltable->llt_dump = in6_lltable_dump; + } return ext; } @@ -2266,6 +2327,7 @@ in6_domifdetach(struct ifnet *ifp, void *aux) scope6_ifdetach(ext->scope6_id); nd6_ifdetach(ext->nd_ifinfo); + lltable_free(ext->lltable); free(ext->in6_ifstat, M_IFADDR); free(ext->icmp6_ifstat, M_IFADDR); free(ext, M_IFADDR); diff --git a/sys/netinet6/in6_rmx.c b/sys/netinet6/in6_rmx.c index bb54341..42980f7 100644 --- a/sys/netinet6/in6_rmx.c +++ b/sys/netinet6/in6_rmx.c @@ -124,6 +124,7 @@ 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; + RADIX_NODE_HEAD_WLOCK_ASSERT(head); if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) rt->rt_flags |= RTF_MULTICAST; @@ -153,27 +154,7 @@ in6_addroute(void *v_arg, void *n_arg, struct radix_node_head *head, rt->rt_rmx.rmx_mtu = IN6_LINKMTU(rt->rt_ifp); ret = rn_addroute(v_arg, n_arg, head, treenodes); - 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, RTF_RNH_LOCKED|RTF_CLONING); - if (rt2) { - if (rt2->rt_flags & RTF_LLINFO && - rt2->rt_flags & RTF_HOST && - rt2->rt_gateway && - rt2->rt_gateway->sa_family == AF_LINK) { - rtexpunge(rt2); - RTFREE_LOCKED(rt2); - ret = rn_addroute(v_arg, n_arg, head, - treenodes); - } else - RTFREE_LOCKED(rt2); - } - } else if (ret == NULL && rt->rt_flags & RTF_CLONING) { + if (ret == NULL) { struct rtentry *rt2; /* * We are trying to add a net route, but can't. @@ -187,10 +168,9 @@ 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, RTF_RNH_LOCKED|RTF_CLONING); + rt2 = rtalloc1((struct sockaddr *)sin6, 0, RTF_RNH_LOCKED); if (rt2) { - if ((rt2->rt_flags & (RTF_CLONING|RTF_HOST|RTF_GATEWAY)) - == RTF_CLONING + if (((rt2->rt_flags & (RTF_HOST|RTF_GATEWAY)) == 0) && rt2->rt_gateway && rt2->rt_gateway->sa_family == AF_LINK && rt2->rt_ifp == rt->rt_ifp) { @@ -199,7 +179,7 @@ in6_addroute(void *v_arg, void *n_arg, struct radix_node_head *head, RTFREE_LOCKED(rt2); } } - return ret; + return (ret); } /* @@ -255,12 +235,6 @@ in6_clsroute(struct radix_node *rn, struct radix_node_head *head) 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) - return; - /* * As requested by David Greenman: * If rtq_reallyold6 is 0, just delete the route without @@ -307,7 +281,7 @@ in6_rtqkill(struct radix_node *rn, void *rock) err = rtrequest(RTM_DELETE, (struct sockaddr *)rt_key(rt), rt->rt_gateway, rt_mask(rt), - rt->rt_flags, 0); + rt->rt_flags|RTF_RNH_LOCKED, 0); if (err) { log(LOG_WARNING, "in6_rtqkill: error %d", err); } else { diff --git a/sys/netinet6/in6_src.c b/sys/netinet6/in6_src.c index 4d7723a..c343634 100644 --- a/sys/netinet6/in6_src.c +++ b/sys/netinet6/in6_src.c @@ -87,6 +87,7 @@ __FBSDID("$FreeBSD$"); #include <net/if.h> #include <net/route.h> +#include <net/if_llatbl.h> #ifdef RADIX_MPATH #include <net/radix_mpath.h> #endif @@ -131,7 +132,7 @@ int ip6_prefer_tempaddr; static int selectroute __P((struct sockaddr_in6 *, struct ip6_pktopts *, struct ip6_moptions *, struct route_in6 *, struct ifnet **, - struct rtentry **, int, int)); + struct rtentry **, int)); static int in6_selectif __P((struct sockaddr_in6 *, struct ip6_pktopts *, struct ip6_moptions *, struct route_in6 *ro, struct ifnet **)); @@ -479,8 +480,7 @@ in6_selectsrc(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, static int selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, struct ip6_moptions *mopts, struct route_in6 *ro, - struct ifnet **retifp, struct rtentry **retrt, int clone, - int norouteok) + struct ifnet **retifp, struct rtentry **retrt, int norouteok) { INIT_VNET_INET6(curvnet); int error = 0; @@ -536,9 +536,10 @@ selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, */ if (opts && opts->ip6po_nexthop) { struct route_in6 *ron; - + struct llentry *la; + sin6_next = satosin6(opts->ip6po_nexthop); - + /* at this moment, we only support AF_INET6 next hops */ if (sin6_next->sin6_family != AF_INET6) { error = EAFNOSUPPORT; /* or should we proceed? */ @@ -550,6 +551,36 @@ selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, * by that address must be a neighbor of the sending host. */ ron = &opts->ip6po_nextroute; + /* + * XXX what do we do here? + * PLZ to be fixing + */ + + + if (ron->ro_rt == NULL) { + rtalloc((struct route *)ron); /* multi path case? */ + if (ron->ro_rt == NULL) { + if (ron->ro_rt) { + RTFREE(ron->ro_rt); + ron->ro_rt = NULL; + } + error = EHOSTUNREACH; + goto done; + } + } + + rt = ron->ro_rt; + ifp = rt->rt_ifp; + IF_AFDATA_LOCK(ifp); + la = lla_lookup(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6_next->sin6_addr); + IF_AFDATA_UNLOCK(ifp); + if (la) + LLE_RUNLOCK(la); + else { + error = EHOSTUNREACH; + goto done; + } +#if 0 if ((ron->ro_rt && (ron->ro_rt->rt_flags & (RTF_UP | RTF_LLINFO)) != (RTF_UP | RTF_LLINFO)) || @@ -573,16 +604,14 @@ selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, goto done; } } - rt = ron->ro_rt; - ifp = rt->rt_ifp; +#endif /* * When cloning is required, try to allocate a route to the * destination so that the caller can store path MTU * information. */ - if (!clone) - goto done; + goto done; } /* @@ -608,21 +637,17 @@ selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, *sa6 = *dstsock; sa6->sin6_scope_id = 0; - if (clone) { #ifdef RADIX_MPATH rtalloc_mpath((struct route *)ro, ntohl(sa6->sin6_addr.s6_addr32[3])); -#else - rtalloc((struct route *)ro); -#endif - } else { +#else ro->ro_rt = rtalloc1(&((struct route *)ro) - ->ro_dst, 0, 0UL); + ->ro_dst, 0, 0UL); if (ro->ro_rt) RT_UNLOCK(ro->ro_rt); - } +#endif } - + /* * do not care about the result if we have the nexthop * explicitly specified. @@ -693,7 +718,7 @@ in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, } if ((error = selectroute(dstsock, opts, mopts, ro, retifp, - &rt, 0, 1)) != 0) { + &rt, 1)) != 0) { if (ro == &sro && rt && rt == sro.ro_rt) RTFREE(rt); return (error); @@ -745,11 +770,11 @@ in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, int in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts, struct ip6_moptions *mopts, struct route_in6 *ro, - struct ifnet **retifp, struct rtentry **retrt, int clone) + struct ifnet **retifp, struct rtentry **retrt) { return (selectroute(dstsock, opts, mopts, ro, retifp, - retrt, clone, 0)); + retrt, 0)); } /* diff --git a/sys/netinet6/in6_var.h b/sys/netinet6/in6_var.h index 957155a..b5dba5e 100644 --- a/sys/netinet6/in6_var.h +++ b/sys/netinet6/in6_var.h @@ -88,13 +88,17 @@ struct in6_addrlifetime { struct nd_ifinfo; struct scope6_id; +struct lltable; struct in6_ifextra { struct in6_ifstat *in6_ifstat; struct icmp6_ifstat *icmp6_ifstat; struct nd_ifinfo *nd_ifinfo; struct scope6_id *scope6_id; + struct lltable *lltable; }; +#define LLTABLE6(ifp) (((struct in6_ifextra *)(ifp)->if_afdata[AF_INET6])->lltable) + struct in6_ifaddr { struct ifaddr ia_ifa; /* protocol-independent info */ #define ia_ifp ia_ifa.ifa_ifp diff --git a/sys/netinet6/ip6_input.c b/sys/netinet6/ip6_input.c index 0c259e3..475888f 100644 --- a/sys/netinet6/ip6_input.c +++ b/sys/netinet6/ip6_input.c @@ -92,6 +92,7 @@ __FBSDID("$FreeBSD$"); #include <netinet/in.h> #include <netinet/in_systm.h> +#include <net/if_llatbl.h> #ifdef INET #include <netinet/ip.h> #include <netinet/ip_icmp.h> @@ -311,9 +312,11 @@ ip6_input(struct mbuf *m) u_int32_t plen; u_int32_t rtalert = ~0; int nxt, ours = 0; - struct ifnet *deliverifp = NULL; + struct ifnet *deliverifp = NULL, *ifp = NULL; struct in6_addr odst; int srcrt = 0; + struct llentry *lle = NULL; + struct sockaddr_in6 dst6; #ifdef IPSEC /* @@ -548,6 +551,24 @@ passin: /* * Unicast check */ + + bzero(&dst6, sizeof(dst6)); + dst6.sin6_family = AF_INET6; + dst6.sin6_len = sizeof(struct sockaddr_in6); + dst6.sin6_addr = ip6->ip6_dst; + ifp = m->m_pkthdr.rcvif; + IF_AFDATA_LOCK(ifp); + lle = lla_lookup(LLTABLE6(ifp), 0, + (struct sockaddr *)&dst6); + IF_AFDATA_UNLOCK(ifp); + if ((lle != NULL) && (lle->la_flags & LLE_IFADDR)) { + ours = 1; + deliverifp = ifp; + LLE_RUNLOCK(lle); + goto hbhcheck; + } + LLE_RUNLOCK(lle); + if (V_ip6_forward_rt.ro_rt != NULL && (V_ip6_forward_rt.ro_rt->rt_flags & RTF_UP) != 0 && IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, diff --git a/sys/netinet6/ip6_output.c b/sys/netinet6/ip6_output.c index 6e5d0d0..c0f614f 100644 --- a/sys/netinet6/ip6_output.c +++ b/sys/netinet6/ip6_output.c @@ -615,7 +615,7 @@ again: dst_sa.sin6_len = sizeof(dst_sa); dst_sa.sin6_addr = ip6->ip6_dst; if ((error = in6_selectroute(&dst_sa, opt, im6o, ro, - &ifp, &rt, 0)) != 0) { + &ifp, &rt)) != 0) { switch (error) { case EHOSTUNREACH: V_ip6stat.ip6s_noroute++; diff --git a/sys/netinet6/ip6_var.h b/sys/netinet6/ip6_var.h index 5975cc9..55bc5db 100644 --- a/sys/netinet6/ip6_var.h +++ b/sys/netinet6/ip6_var.h @@ -398,7 +398,7 @@ struct in6_addr *in6_selectsrc __P((struct sockaddr_in6 *, struct ip6_pktopts *, struct ifnet **, int *)); int in6_selectroute __P((struct sockaddr_in6 *, struct ip6_pktopts *, struct ip6_moptions *, struct route_in6 *, struct ifnet **, - struct rtentry **, int)); + struct rtentry **)); u_int32_t ip6_randomid __P((void)); u_int32_t ip6_randomflowlabel __P((void)); #endif /* _KERNEL */ diff --git a/sys/netinet6/nd6.c b/sys/netinet6/nd6.c index 18e81c3..5095d23 100644 --- a/sys/netinet6/nd6.c +++ b/sys/netinet6/nd6.c @@ -48,6 +48,8 @@ __FBSDID("$FreeBSD$"); #include <sys/protosw.h> #include <sys/errno.h> #include <sys/syslog.h> +#include <sys/lock.h> +#include <sys/rwlock.h> #include <sys/queue.h> #include <sys/sysctl.h> @@ -61,6 +63,8 @@ __FBSDID("$FreeBSD$"); #include <net/vnet.h> #include <netinet/in.h> +#include <net/if_llatbl.h> +#define L3_ADDR_SIN6(le) ((struct sockaddr_in6 *) L3_ADDR(le)) #include <netinet/if_ether.h> #include <netinet6/in6_var.h> #include <netinet/ip6.h> @@ -98,8 +102,9 @@ int nd6_maxqueuelen; int nd6_debug; /* for debugging? */ +#if 0 static int nd6_inuse, nd6_allocated; -struct llinfo_nd6 llinfo_nd6; +#endif struct nd_drhead nd_defrouter; struct nd_prhead nd_prefix; @@ -114,9 +119,9 @@ static int nd6_is_new_addr_neighbor __P((struct sockaddr_in6 *, static void nd6_setmtu0(struct ifnet *, struct nd_ifinfo *); static void nd6_slowtimo(void *); static int regen_tmpaddr(struct in6_ifaddr *); -static struct llinfo_nd6 *nd6_free(struct rtentry *, int); +static struct llentry *nd6_free(struct llentry *, int); static void nd6_llinfo_timer(void *); -static void clear_llinfo_pqueue(struct llinfo_nd6 *); +static void clear_llinfo_pqueue(struct llentry *); #ifdef VIMAGE_GLOBALS struct callout nd6_slowtimo_ch; @@ -162,8 +167,13 @@ nd6_init(void) V_dad_ignore_ns = 0; /* ignore NS in DAD - specwise incorrect*/ V_dad_maxtry = 15; /* max # of *tries* to transmit DAD packet */ + /* + * XXX just to get this to compile KMM + */ +#ifdef notyet V_llinfo_nd6.ln_next = &V_llinfo_nd6; V_llinfo_nd6.ln_prev = &V_llinfo_nd6; +#endif LIST_INIT(&V_nd_prefix); V_ip6_use_tempaddr = 0; @@ -424,14 +434,23 @@ skip1: * ND6 timer routine to handle ND6 entries */ void -nd6_llinfo_settimer(struct llinfo_nd6 *ln, long tick) +nd6_llinfo_settimer_locked(struct llentry *ln, long tick) { if (tick < 0) { - ln->ln_expire = 0; + ln->la_expire = 0; ln->ln_ntick = 0; callout_stop(&ln->ln_timer_ch); + /* + * XXX - do we know that there is + * callout installed? i.e. are we + * guaranteed that we're not dropping + * a reference that we did not add? + * KMM + */ + LLE_REMREF(ln); } else { - ln->ln_expire = time_second + tick / hz; + ln->la_expire = time_second + tick / hz; + LLE_ADDREF(ln); if (tick > INT_MAX) { ln->ln_ntick = tick - INT_MAX; callout_reset(&ln->ln_timer_ch, INT_MAX, @@ -444,16 +463,34 @@ nd6_llinfo_settimer(struct llinfo_nd6 *ln, long tick) } } +void +nd6_llinfo_settimer(struct llentry *ln, long tick) +{ + + LLE_WLOCK(ln); + nd6_llinfo_settimer_locked(ln, tick); + LLE_WUNLOCK(ln); +} + static void nd6_llinfo_timer(void *arg) { - struct llinfo_nd6 *ln; - struct rtentry *rt; + struct llentry *ln; struct in6_addr *dst; struct ifnet *ifp; struct nd_ifinfo *ndi = NULL; - ln = (struct llinfo_nd6 *)arg; + ln = (struct llentry *)arg; + if (ln == NULL) { + panic("%s: NULL entry!\n", __func__); + return; + } + + if ((ifp = ((ln->lle_tbl != NULL) ? ln->lle_tbl->llt_ifp : NULL)) == NULL) + panic("ln ifp == NULL"); + + CURVNET_SET(ifp->if_vnet); + INIT_VNET_INET6(curvnet); if (ln->ln_ntick > 0) { if (ln->ln_ntick > INT_MAX) { @@ -463,52 +500,44 @@ nd6_llinfo_timer(void *arg) ln->ln_ntick = 0; nd6_llinfo_settimer(ln, ln->ln_ntick); } - return; + goto done; } - if ((rt = ln->ln_rt) == NULL) - panic("ln->ln_rt == NULL"); - if ((ifp = rt->rt_ifp) == NULL) - panic("ln->ln_rt->rt_ifp == NULL"); ndi = ND_IFINFO(ifp); + dst = &L3_ADDR_SIN6(ln)->sin6_addr; + if ((ln->la_flags & LLE_STATIC) || (ln->la_expire > time_second)) { + goto done; + } - CURVNET_SET(ifp->if_vnet); - INIT_VNET_INET6(curvnet); - - /* sanity check */ - if (rt->rt_llinfo && (struct llinfo_nd6 *)rt->rt_llinfo != ln) - panic("rt_llinfo(%p) is not equal to ln(%p)", - rt->rt_llinfo, ln); - if (rt_key(rt) == NULL) - panic("rt key is NULL in nd6_timer(ln=%p)", ln); - - dst = &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr; + if (ln->la_flags & LLE_DELETED) { + (void)nd6_free(ln, 0); + goto done; + } switch (ln->ln_state) { case ND6_LLINFO_INCOMPLETE: - if (ln->ln_asked < V_nd6_mmaxtries) { - ln->ln_asked++; + if (ln->la_asked < V_nd6_mmaxtries) { + ln->la_asked++; nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); nd6_ns_output(ifp, NULL, dst, ln, 0); } else { - struct mbuf *m = ln->ln_hold; + struct mbuf *m = ln->la_hold; if (m) { struct mbuf *m0; /* - * assuming every packet in ln_hold has the + * assuming every packet in la_hold has the * same IP header */ m0 = m->m_nextpkt; m->m_nextpkt = NULL; icmp6_error2(m, ICMP6_DST_UNREACH, - ICMP6_DST_UNREACH_ADDR, 0, rt->rt_ifp); + ICMP6_DST_UNREACH_ADDR, 0, ifp); - ln->ln_hold = m0; + ln->la_hold = m0; clear_llinfo_pqueue(ln); } - if (rt && rt->rt_llinfo) - (void)nd6_free(rt, 0); + (void)nd6_free(ln, 0); ln = NULL; } break; @@ -522,8 +551,7 @@ nd6_llinfo_timer(void *arg) case ND6_LLINFO_STALE: /* Garbage Collection(RFC 2461 5.3) */ if (!ND6_LLINFO_PERMANENT(ln)) { - if (rt && rt->rt_llinfo) - (void)nd6_free(rt, 1); + (void)nd6_free(ln, 1); ln = NULL; } break; @@ -531,7 +559,7 @@ nd6_llinfo_timer(void *arg) case ND6_LLINFO_DELAY: if (ndi && (ndi->flags & ND6_IFF_PERFORMNUD) != 0) { /* We need NUD */ - ln->ln_asked = 1; + ln->la_asked = 1; ln->ln_state = ND6_LLINFO_PROBE; nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); nd6_ns_output(ifp, dst, dst, ln, 0); @@ -541,31 +569,20 @@ nd6_llinfo_timer(void *arg) } break; case ND6_LLINFO_PROBE: - if (ln->ln_asked < V_nd6_umaxtries) { - ln->ln_asked++; + if (ln->la_asked < V_nd6_umaxtries) { + ln->la_asked++; nd6_llinfo_settimer(ln, (long)ndi->retrans * hz / 1000); nd6_ns_output(ifp, dst, dst, ln, 0); - } else if (rt->rt_ifa != NULL && - rt->rt_ifa->ifa_addr->sa_family == AF_INET6 && - (((struct in6_ifaddr *)rt->rt_ifa)->ia_flags & IFA_ROUTE)) { - /* - * This is an unreachable neighbor whose address is - * specified as the destination of a p2p interface - * (see in6_ifinit()). We should not free the entry - * since this is sort of a "static" entry generated - * via interface address configuration. - */ - ln->ln_asked = 0; - ln->ln_expire = 0; /* make it permanent */ - ln->ln_state = ND6_LLINFO_STALE; } else { - if (rt && rt->rt_llinfo) - (void)nd6_free(rt, 0); + (void)nd6_free(ln, 0); ln = NULL; } break; } CURVNET_RESTORE(); +done: + if (ln != NULL) + LLE_FREE(ln); } @@ -772,7 +789,6 @@ void nd6_purge(struct ifnet *ifp) { INIT_VNET_INET6(ifp->if_vnet); - struct llinfo_nd6 *ln, *nln; struct nd_defrouter *dr, *ndr; struct nd_prefix *pr, *npr; @@ -829,132 +845,54 @@ nd6_purge(struct ifnet *ifp) nd6_setdefaultiface(0); if (!V_ip6_forwarding && V_ip6_accept_rtadv) { /* XXX: too restrictive? */ - /* refresh default router list */ + /* refresh default router list + * + * + */ defrouter_select(); + } - /* - * Nuke neighbor cache entries for the ifp. - * Note that rt->rt_ifp may not be the same as ifp, - * due to KAME goto ours hack. See RTM_RESOLVE case in - * nd6_rtrequest(), and ip6_input(). + /* XXXXX + * We do not nuke the neighbor cache entries here any more + * because the neighbor cache is kept in if_afdata[AF_INET6]. + * nd6_purge() is invoked by in6_ifdetach() which is called + * from if_detach() where everything gets purged. So let + * in6_domifdetach() do the actual L2 table purging work. */ - ln = V_llinfo_nd6.ln_next; - while (ln && ln != &V_llinfo_nd6) { - struct rtentry *rt; - struct sockaddr_dl *sdl; - - nln = ln->ln_next; - rt = ln->ln_rt; - if (rt && rt->rt_gateway && - rt->rt_gateway->sa_family == AF_LINK) { - sdl = (struct sockaddr_dl *)rt->rt_gateway; - if (sdl->sdl_index == ifp->if_index) - nln = nd6_free(rt, 0); - } - ln = nln; - } } -struct rtentry * -nd6_lookup(struct in6_addr *addr6, int create, struct ifnet *ifp) +/* + * the caller acquires and releases the lock on the lltbls + * Returns the llentry locked + */ +struct llentry * +nd6_lookup(struct in6_addr *addr6, int flags, struct ifnet *ifp) { INIT_VNET_INET6(curvnet); - struct rtentry *rt; struct sockaddr_in6 sin6; - char ip6buf[INET6_ADDRSTRLEN]; - + struct llentry *ln; + int llflags = 0; + bzero(&sin6, sizeof(sin6)); sin6.sin6_len = sizeof(struct sockaddr_in6); sin6.sin6_family = AF_INET6; sin6.sin6_addr = *addr6; - rt = rtalloc1((struct sockaddr *)&sin6, create, 0UL); - if (rt) { - if ((rt->rt_flags & RTF_LLINFO) == 0 && create) { - /* - * This is the case for the default route. - * If we want to create a neighbor cache for the - * address, we should free the route for the - * destination and allocate an interface route. - */ - RTFREE_LOCKED(rt); - rt = NULL; - } - } - if (rt == NULL) { - if (create && ifp) { - int e; - /* - * If no route is available and create is set, - * we allocate a host route for the destination - * and treat it like an interface route. - * This hack is necessary for a neighbor which can't - * be covered by our own prefix. - */ - struct ifaddr *ifa = - ifaof_ifpforaddr((struct sockaddr *)&sin6, ifp); - if (ifa == NULL) - return (NULL); + IF_AFDATA_LOCK_ASSERT(ifp); - /* - * Create a new route. RTF_LLINFO is necessary - * to create a Neighbor Cache entry for the - * destination in nd6_rtrequest which will be - * called in rtrequest via ifa->ifa_rtrequest. - */ - if ((e = rtrequest(RTM_ADD, (struct sockaddr *)&sin6, - ifa->ifa_addr, (struct sockaddr *)&all1_sa, - (ifa->ifa_flags | RTF_HOST | RTF_LLINFO) & - ~RTF_CLONING, &rt)) != 0) { - log(LOG_ERR, - "nd6_lookup: failed to add route for a " - "neighbor(%s), errno=%d\n", - ip6_sprintf(ip6buf, addr6), e); - } - if (rt == NULL) - return (NULL); - RT_LOCK(rt); - if (rt->rt_llinfo) { - struct llinfo_nd6 *ln = - (struct llinfo_nd6 *)rt->rt_llinfo; - ln->ln_state = ND6_LLINFO_NOSTATE; - } - } else - return (NULL); - } - RT_LOCK_ASSERT(rt); - RT_REMREF(rt); - /* - * Validation for the entry. - * Note that the check for rt_llinfo is necessary because a cloned - * route from a parent route that has the L flag (e.g. the default - * route to a p2p interface) may have the flag, too, while the - * destination is not actually a neighbor. - * XXX: we can't use rt->rt_ifp to check for the interface, since - * it might be the loopback interface if the entry is for our - * own address on a non-loopback interface. Instead, we should - * use rt->rt_ifa->ifa_ifp, which would specify the REAL - * interface. - * Note also that ifa_ifp and ifp may differ when we connect two - * interfaces to a same link, install a link prefix to an interface, - * and try to install a neighbor cache on an interface that does not - * have a route to the prefix. - */ - if ((rt->rt_flags & RTF_GATEWAY) || (rt->rt_flags & RTF_LLINFO) == 0 || - rt->rt_gateway->sa_family != AF_LINK || rt->rt_llinfo == NULL || - (ifp && rt->rt_ifa->ifa_ifp != ifp)) { - if (create) { - nd6log((LOG_DEBUG, - "nd6_lookup: failed to lookup %s (if = %s)\n", - ip6_sprintf(ip6buf, addr6), - ifp ? if_name(ifp) : "unspec")); - } - RT_UNLOCK(rt); - return (NULL); + if (flags & ND6_CREATE) + llflags |= LLE_CREATE; + if (flags & ND6_EXCLUSIVE) + llflags |= LLE_EXCLUSIVE; + + ln = lla_lookup(LLTABLE6(ifp), llflags, (struct sockaddr *)&sin6); + if ((ln != NULL) && (flags & LLE_CREATE)) { + ln->ln_state = ND6_LLINFO_NOSTATE; + callout_init(&ln->ln_timer_ch, 0); } - RT_UNLOCK(rt); /* XXX not ready to return rt locked */ - return (rt); + + return (ln); } /* @@ -1040,7 +978,10 @@ nd6_is_new_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) int nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) { + struct llentry *lle; + int rc = 0; + IF_AFDATA_UNLOCK_ASSERT(ifp); if (nd6_is_new_addr_neighbor(addr, ifp)) return (1); @@ -1048,10 +989,13 @@ nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) * Even if the address matches none of our addresses, it might be * in the neighbor cache. */ - if (nd6_lookup(&addr->sin6_addr, 0, ifp) != NULL) - return (1); - - return (0); + IF_AFDATA_LOCK(ifp); + if ((lle = nd6_lookup(&addr->sin6_addr, 0, ifp)) != NULL) { + LLE_RUNLOCK(lle); + rc = 1; + } + IF_AFDATA_UNLOCK(ifp); + return (rc); } /* @@ -1060,13 +1004,13 @@ nd6_is_addr_neighbor(struct sockaddr_in6 *addr, struct ifnet *ifp) * make it global, unless you have a strong reason for the change, and are sure * that the change is safe. */ -static struct llinfo_nd6 * -nd6_free(struct rtentry *rt, int gc) +static struct llentry * +nd6_free(struct llentry *ln, int gc) { INIT_VNET_INET6(curvnet); - struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo, *next; - struct in6_addr in6 = ((struct sockaddr_in6 *)rt_key(rt))->sin6_addr; + struct llentry *next; struct nd_defrouter *dr; + struct ifnet *ifp=NULL; /* * we used to have pfctlinput(PRC_HOSTDEAD) here. @@ -1079,8 +1023,7 @@ nd6_free(struct rtentry *rt, int gc) if (!V_ip6_forwarding) { int s; s = splnet(); - dr = defrouter_lookup(&((struct sockaddr_in6 *)rt_key(rt))->sin6_addr, - rt->rt_ifp); + dr = defrouter_lookup(&L3_ADDR_SIN6(ln)->sin6_addr, ln->lle_tbl->llt_ifp); if (dr != NULL && dr->expire && ln->ln_state == ND6_LLINFO_STALE && gc) { @@ -1102,7 +1045,7 @@ nd6_free(struct rtentry *rt, int gc) else nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); splx(s); - return (ln->ln_next); + return (LIST_NEXT(ln, lle_next)); } if (ln->ln_router || dr) { @@ -1111,7 +1054,7 @@ nd6_free(struct rtentry *rt, int gc) * is in the Default Router List. * See a corresponding comment in nd6_na_input(). */ - rt6_flush(&in6, rt->rt_ifp); + rt6_flush(&L3_ADDR_SIN6(ln)->sin6_addr, ln->lle_tbl->llt_ifp); } if (dr) { @@ -1152,15 +1095,13 @@ nd6_free(struct rtentry *rt, int gc) * might have freed other entries (particularly the old next entry) as * a side effect (XXX). */ - next = ln->ln_next; + next = LIST_NEXT(ln, lle_next); - /* - * Detach the route from the routing tree and the list of neighbor - * caches, and disable the route entry not to be used in already - * cached routes. - */ - rtrequest(RTM_DELETE, rt_key(rt), (struct sockaddr *)0, - rt_mask(rt), 0, (struct rtentry **)0); + ifp = ln->lle_tbl->llt_ifp; + IF_AFDATA_LOCK(ifp); + LLE_WLOCK(ln); + llentry_free(ln); + IF_AFDATA_UNLOCK(ifp); return (next); } @@ -1174,297 +1115,42 @@ void nd6_nud_hint(struct rtentry *rt, struct in6_addr *dst6, int force) { INIT_VNET_INET6(curvnet); - struct llinfo_nd6 *ln; + struct llentry *ln; + struct ifnet *ifp; - /* - * If the caller specified "rt", use that. Otherwise, resolve the - * routing table by supplied "dst6". - */ - if (rt == NULL) { - if (dst6 == NULL) - return; - if ((rt = nd6_lookup(dst6, 0, NULL)) == NULL) - return; - } - - if ((rt->rt_flags & RTF_GATEWAY) != 0 || - (rt->rt_flags & RTF_LLINFO) == 0 || - rt->rt_llinfo == NULL || rt->rt_gateway == NULL || - rt->rt_gateway->sa_family != AF_LINK) { - /* This is not a host route. */ + if ((dst6 == NULL) || (rt == NULL)) return; - } - ln = (struct llinfo_nd6 *)rt->rt_llinfo; - if (ln->ln_state < ND6_LLINFO_REACHABLE) + ifp = rt->rt_ifp; + IF_AFDATA_LOCK(ifp); + ln = nd6_lookup(dst6, ND6_EXCLUSIVE, NULL); + IF_AFDATA_UNLOCK(ifp); + if (ln == NULL) return; + if (ln->ln_state < ND6_LLINFO_REACHABLE) + goto done; + /* * if we get upper-layer reachability confirmation many times, * it is possible we have false information. */ if (!force) { ln->ln_byhint++; - if (ln->ln_byhint > V_nd6_maxnudhint) - return; + if (ln->ln_byhint > V_nd6_maxnudhint) { + goto done; + } } - ln->ln_state = ND6_LLINFO_REACHABLE; + ln->ln_state = ND6_LLINFO_REACHABLE; if (!ND6_LLINFO_PERMANENT(ln)) { nd6_llinfo_settimer(ln, (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); } +done: + LLE_WUNLOCK(ln); } -/* - * info - XXX unused - */ -void -nd6_rtrequest(int req, struct rtentry *rt, struct rt_addrinfo *info) -{ - struct sockaddr *gate = rt->rt_gateway; - struct llinfo_nd6 *ln = (struct llinfo_nd6 *)rt->rt_llinfo; - static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; - struct ifnet *ifp = rt->rt_ifp; - struct ifaddr *ifa; - INIT_VNET_NET(ifp->if_vnet); - INIT_VNET_INET6(ifp->if_vnet); - - RT_LOCK_ASSERT(rt); - - if ((rt->rt_flags & RTF_GATEWAY) != 0) - return; - - if (nd6_need_cache(ifp) == 0 && (rt->rt_flags & RTF_HOST) == 0) { - /* - * This is probably an interface direct route for a link - * which does not need neighbor caches (e.g. fe80::%lo0/64). - * We do not need special treatment below for such a route. - * Moreover, the RTF_LLINFO flag which would be set below - * would annoy the ndp(8) command. - */ - return; - } - - if (req == RTM_RESOLVE && - (nd6_need_cache(ifp) == 0 || /* stf case */ - !nd6_is_new_addr_neighbor((struct sockaddr_in6 *)rt_key(rt), - ifp))) { - /* - * FreeBSD and BSD/OS often make a cloned host route based - * on a less-specific route (e.g. the default route). - * If the less specific route does not have a "gateway" - * (this is the case when the route just goes to a p2p or an - * stf interface), we'll mistakenly make a neighbor cache for - * the host route, and will see strange neighbor solicitation - * for the corresponding destination. In order to avoid the - * confusion, we check if the destination of the route is - * a neighbor in terms of neighbor discovery, and stop the - * process if not. Additionally, we remove the LLINFO flag - * so that ndp(8) will not try to get the neighbor information - * of the destination. - */ - rt->rt_flags &= ~RTF_LLINFO; - return; - } - - switch (req) { - case RTM_ADD: - /* - * There is no backward compatibility :) - * - * if ((rt->rt_flags & RTF_HOST) == 0 && - * SIN(rt_mask(rt))->sin_addr.s_addr != 0xffffffff) - * rt->rt_flags |= RTF_CLONING; - */ - if ((rt->rt_flags & RTF_CLONING) || - ((rt->rt_flags & RTF_LLINFO) && ln == NULL)) { - /* - * Case 1: This route should come from a route to - * interface (RTF_CLONING case) or the route should be - * treated as on-link but is currently not - * (RTF_LLINFO && ln == NULL case). - */ - rt_setgate(rt, rt_key(rt), - (struct sockaddr *)&null_sdl); - gate = rt->rt_gateway; - SDL(gate)->sdl_type = ifp->if_type; - SDL(gate)->sdl_index = ifp->if_index; - if (ln) - nd6_llinfo_settimer(ln, 0); - if ((rt->rt_flags & RTF_CLONING) != 0) - break; - } - /* - * In IPv4 code, we try to annonuce new RTF_ANNOUNCE entry here. - * We don't do that here since llinfo is not ready yet. - * - * There are also couple of other things to be discussed: - * - unsolicited NA code needs improvement beforehand - * - RFC2461 says we MAY send multicast unsolicited NA - * (7.2.6 paragraph 4), however, it also says that we - * SHOULD provide a mechanism to prevent multicast NA storm. - * we don't have anything like it right now. - * note that the mechanism needs a mutual agreement - * between proxies, which means that we need to implement - * a new protocol, or a new kludge. - * - from RFC2461 6.2.4, host MUST NOT send an unsolicited NA. - * we need to check ip6forwarding before sending it. - * (or should we allow proxy ND configuration only for - * routers? there's no mention about proxy ND from hosts) - */ - /* FALLTHROUGH */ - case RTM_RESOLVE: - if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) { - /* - * Address resolution isn't necessary for a point to - * point link, so we can skip this test for a p2p link. - */ - if (gate->sa_family != AF_LINK || - gate->sa_len < sizeof(null_sdl)) { - log(LOG_DEBUG, - "nd6_rtrequest: bad gateway value: %s\n", - if_name(ifp)); - break; - } - SDL(gate)->sdl_type = ifp->if_type; - SDL(gate)->sdl_index = ifp->if_index; - } - if (ln != NULL) - break; /* This happens on a route change */ - /* - * Case 2: This route may come from cloning, or a manual route - * add with a LL address. - */ - R_Malloc(ln, struct llinfo_nd6 *, sizeof(*ln)); - rt->rt_llinfo = (caddr_t)ln; - if (ln == NULL) { - log(LOG_DEBUG, "nd6_rtrequest: malloc failed\n"); - break; - } - V_nd6_inuse++; - V_nd6_allocated++; - bzero(ln, sizeof(*ln)); - RT_ADDREF(rt); - ln->ln_rt = rt; - callout_init(&ln->ln_timer_ch, 0); - - /* this is required for "ndp" command. - shin */ - if (req == RTM_ADD) { - /* - * gate should have some valid AF_LINK entry, - * and ln->ln_expire should have some lifetime - * which is specified by ndp command. - */ - ln->ln_state = ND6_LLINFO_REACHABLE; - ln->ln_byhint = 0; - } else { - /* - * When req == RTM_RESOLVE, rt is created and - * initialized in rtrequest(), so rt_expire is 0. - */ - ln->ln_state = ND6_LLINFO_NOSTATE; - nd6_llinfo_settimer(ln, 0); - } - rt->rt_flags |= RTF_LLINFO; - ln->ln_next = V_llinfo_nd6.ln_next; - V_llinfo_nd6.ln_next = ln; - ln->ln_prev = &V_llinfo_nd6; - ln->ln_next->ln_prev = ln; - - /* - * check if rt_key(rt) is one of my address assigned - * to the interface. - */ - ifa = (struct ifaddr *)in6ifa_ifpwithaddr(rt->rt_ifp, - &SIN6(rt_key(rt))->sin6_addr); - if (ifa) { - caddr_t macp = nd6_ifptomac(ifp); - nd6_llinfo_settimer(ln, -1); - ln->ln_state = ND6_LLINFO_REACHABLE; - ln->ln_byhint = 0; - if (macp) { - bcopy(macp, LLADDR(SDL(gate)), ifp->if_addrlen); - SDL(gate)->sdl_alen = ifp->if_addrlen; - } - if (V_nd6_useloopback) { - rt->rt_ifp = &V_loif[0]; /* XXX */ - /* - * Make sure rt_ifa be equal to the ifaddr - * corresponding to the address. - * We need this because when we refer - * rt_ifa->ia6_flags in ip6_input, we assume - * that the rt_ifa points to the address instead - * of the loopback address. - */ - if (ifa != rt->rt_ifa) { - IFAFREE(rt->rt_ifa); - IFAREF(ifa); - rt->rt_ifa = ifa; - } - } - } else if (rt->rt_flags & RTF_ANNOUNCE) { - nd6_llinfo_settimer(ln, -1); - ln->ln_state = ND6_LLINFO_REACHABLE; - ln->ln_byhint = 0; - - /* join solicited node multicast for proxy ND */ - if (ifp->if_flags & IFF_MULTICAST) { - struct in6_addr llsol; - int error; - - llsol = SIN6(rt_key(rt))->sin6_addr; - llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL; - llsol.s6_addr32[1] = 0; - llsol.s6_addr32[2] = htonl(1); - llsol.s6_addr8[12] = 0xff; - if (in6_setscope(&llsol, ifp, NULL)) - break; - if (in6_addmulti(&llsol, ifp, - &error, 0) == NULL) { - char ip6buf[INET6_ADDRSTRLEN]; - nd6log((LOG_ERR, "%s: failed to join " - "%s (errno=%d)\n", if_name(ifp), - ip6_sprintf(ip6buf, &llsol), - error)); - } - } - } - break; - - case RTM_DELETE: - if (ln == NULL) - break; - /* leave from solicited node multicast for proxy ND */ - if ((rt->rt_flags & RTF_ANNOUNCE) != 0 && - (ifp->if_flags & IFF_MULTICAST) != 0) { - struct in6_addr llsol; - struct in6_multi *in6m; - - llsol = SIN6(rt_key(rt))->sin6_addr; - llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL; - llsol.s6_addr32[1] = 0; - llsol.s6_addr32[2] = htonl(1); - llsol.s6_addr8[12] = 0xff; - if (in6_setscope(&llsol, ifp, NULL) == 0) { - IN6_LOOKUP_MULTI(llsol, ifp, in6m); - if (in6m) - in6_delmulti(in6m); - } else - ; /* XXX: should not happen. bark here? */ - } - V_nd6_inuse--; - ln->ln_next->ln_prev = ln->ln_prev; - ln->ln_prev->ln_next = ln->ln_next; - ln->ln_prev = NULL; - nd6_llinfo_settimer(ln, -1); - RT_REMREF(rt); - rt->rt_llinfo = 0; - rt->rt_flags &= ~RTF_LLINFO; - clear_llinfo_pqueue(ln); - Free((caddr_t)ln); - } -} int nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) @@ -1477,7 +1163,6 @@ nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) struct in6_ndifreq *ndif = (struct in6_ndifreq *)data; struct nd_defrouter *dr; struct nd_prefix *pr; - struct rtentry *rt; int i = 0, error = 0; int s; @@ -1667,25 +1352,25 @@ nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) } case SIOCGNBRINFO_IN6: { - struct llinfo_nd6 *ln; + struct llentry *ln; struct in6_addr nb_addr = nbi->addr; /* make local for safety */ if ((error = in6_setscope(&nb_addr, ifp, NULL)) != 0) return (error); - s = splnet(); - if ((rt = nd6_lookup(&nb_addr, 0, ifp)) == NULL) { + IF_AFDATA_LOCK(ifp); + ln = nd6_lookup(&nb_addr, 0, ifp); + IF_AFDATA_UNLOCK(ifp); + + if (ln == NULL) { error = EINVAL; - splx(s); break; } - ln = (struct llinfo_nd6 *)rt->rt_llinfo; nbi->state = ln->ln_state; - nbi->asked = ln->ln_asked; + nbi->asked = ln->la_asked; nbi->isrouter = ln->ln_router; - nbi->expire = ln->ln_expire; - splx(s); - + nbi->expire = ln->la_expire; + LLE_RUNLOCK(ln); break; } case SIOCGDEFIFACE_IN6: /* XXX: should be implemented as a sysctl? */ @@ -1703,20 +1388,27 @@ nd6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp) * * type - ICMP6 type * code - type dependent information + * + * XXXXX + * The caller of this function already acquired the ndp + * cache table lock because the cache entry is returned. */ -struct rtentry * +struct llentry * nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, int lladdrlen, int type, int code) { INIT_VNET_INET6(curvnet); - struct rtentry *rt = NULL; - struct llinfo_nd6 *ln = NULL; + struct llentry *ln = NULL; int is_newentry; - struct sockaddr_dl *sdl = NULL; int do_update; int olladdr; int llchange; + int flags = 0; int newstate = 0; + struct sockaddr_in6 sin6; + struct mbuf *chain = NULL; + + IF_AFDATA_UNLOCK_ASSERT(ifp); if (ifp == NULL) panic("ifp == NULL in nd6_cache_lladdr"); @@ -1736,40 +1428,29 @@ nd6_cache_lladdr(struct ifnet *ifp, struct in6_addr *from, char *lladdr, * Spec says nothing in sections for RA, RS and NA. There's small * description on it in NS section (RFC 2461 7.2.3). */ - - rt = nd6_lookup(from, 0, ifp); - if (rt == NULL) { - rt = nd6_lookup(from, 1, ifp); + flags |= lladdr ? ND6_EXCLUSIVE : 0; + IF_AFDATA_LOCK(ifp); + ln = nd6_lookup(from, flags, ifp); + if (ln) + IF_AFDATA_UNLOCK(ifp); + if (ln == NULL) { + flags |= LLE_EXCLUSIVE; + ln = nd6_lookup(from, flags |ND6_CREATE, ifp); + IF_AFDATA_UNLOCK(ifp); is_newentry = 1; } else { /* do nothing if static ndp is set */ - if (rt->rt_flags & RTF_STATIC) - return NULL; + if (ln->la_flags & LLE_STATIC) + goto done; is_newentry = 0; } - - if (rt == NULL) - return NULL; - if ((rt->rt_flags & (RTF_GATEWAY | RTF_LLINFO)) != RTF_LLINFO) { -fail: - (void)nd6_free(rt, 0); - return NULL; - } - ln = (struct llinfo_nd6 *)rt->rt_llinfo; if (ln == NULL) - goto fail; - if (rt->rt_gateway == NULL) - goto fail; - if (rt->rt_gateway->sa_family != AF_LINK) - goto fail; - sdl = SDL(rt->rt_gateway); - - olladdr = (sdl->sdl_alen) ? 1 : 0; + return (NULL); + + olladdr = (ln->la_flags & LLE_VALID) ? 1 : 0; if (olladdr && lladdr) { - if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen)) - llchange = 1; - else - llchange = 0; + llchange = bcmp(lladdr, &ln->ll_addr, + ifp->if_addrlen); } else llchange = 0; @@ -1789,8 +1470,8 @@ fail: * Record source link-layer address * XXX is it dependent to ifp->if_type? */ - sdl->sdl_alen = ifp->if_addrlen; - bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen); + bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen); + ln->la_flags |= LLE_VALID; } if (!is_newentry) { @@ -1821,17 +1502,17 @@ fail: * we must set the timer now, although it is actually * meaningless. */ - nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); + nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); - if (ln->ln_hold) { + if (ln->la_hold) { struct mbuf *m_hold, *m_hold_next; /* - * reset the ln_hold in advance, to explicitly - * prevent a ln_hold lookup in nd6_output() + * reset the la_hold in advance, to explicitly + * prevent a la_hold lookup in nd6_output() * (wouldn't happen, though...) */ - for (m_hold = ln->ln_hold, ln->ln_hold = NULL; + for (m_hold = ln->la_hold, ln->la_hold = NULL; m_hold; m_hold = m_hold_next) { m_hold_next = m_hold->m_nextpkt; m_hold->m_nextpkt = NULL; @@ -1841,14 +1522,14 @@ fail: * just set the 2nd argument as the * 1st one. */ - nd6_output(ifp, ifp, m_hold, - (struct sockaddr_in6 *)rt_key(rt), - rt); + nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain); } + if (chain) + memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6)); } } else if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { /* probe right away */ - nd6_llinfo_settimer((void *)ln, 0); + nd6_llinfo_settimer_locked((void *)ln, 0); } } @@ -1917,6 +1598,17 @@ fail: break; } + if (ln) { + if (flags & ND6_EXCLUSIVE) + LLE_WUNLOCK(ln); + else + LLE_RUNLOCK(ln); + if (ln->la_flags & LLE_STATIC) + ln = NULL; + } + if (chain) + nd6_output_flush(ifp, ifp, chain, &sin6, NULL); + /* * When the link-layer address of a router changes, select the * best router again. In particular, when the neighbor entry is newly @@ -1932,10 +1624,24 @@ fail: * for those are not autoconfigured hosts, we explicitly avoid such * cases for safety. */ - if (do_update && ln->ln_router && !V_ip6_forwarding && V_ip6_accept_rtadv) + if (do_update && ln->ln_router && !V_ip6_forwarding && V_ip6_accept_rtadv) { + /* + * guaranteed recursion + */ defrouter_select(); - - return rt; + } + + return (ln); +done: + if (ln) { + if (flags & ND6_EXCLUSIVE) + LLE_WUNLOCK(ln); + else + LLE_RUNLOCK(ln); + if (ln->la_flags & LLE_STATIC) + ln = NULL; + } + return (ln); } static void @@ -1969,18 +1675,45 @@ nd6_slowtimo(void *arg) CURVNET_RESTORE(); } -#define senderr(e) { error = (e); goto bad;} int nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, struct sockaddr_in6 *dst, struct rtentry *rt0) { + + return (nd6_output_lle(ifp, origifp, m0, dst, rt0, NULL, NULL)); +} + + +/* + * Note that I'm not enforcing any global serialization + * lle state or asked changes here as the logic is too + * complicated to avoid having to always acquire an exclusive + * lock + * KMM + * + */ +#define senderr(e) { error = (e); goto bad;} + +int +nd6_output_lle(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, + struct sockaddr_in6 *dst, struct rtentry *rt0, struct llentry *lle, + struct mbuf **tail) +{ INIT_VNET_INET6(curvnet); struct mbuf *m = m0; struct rtentry *rt = rt0; - struct sockaddr_in6 *gw6 = NULL; - struct llinfo_nd6 *ln = NULL; + struct llentry *ln = lle; int error = 0; + int flags = 0; + +#ifdef INVARIANTS + if (lle) { + + LLE_WLOCK_ASSERT(lle); + KASSERT(tail != NULL, (" lle locked but no tail pointer passed")); + } +#endif if (IN6_IS_ADDR_MULTICAST(&dst->sin6_addr)) goto sendpkt; @@ -1990,81 +1723,6 @@ nd6_output(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *m0, /* * next hop determination. This routine is derived from ether_output. */ - /* NB: the locking here is tortuous... */ - if (rt != NULL) - RT_LOCK(rt); -again: - if (rt != NULL) { - if ((rt->rt_flags & RTF_UP) == 0) { - RT_UNLOCK(rt); - rt0 = rt = rtalloc1((struct sockaddr *)dst, 1, 0UL); - if (rt != NULL) { - RT_REMREF(rt); - if (rt->rt_ifp != ifp) - /* - * XXX maybe we should update ifp too, - * but the original code didn't and I - * don't know what is correct here. - */ - goto again; - } else - senderr(EHOSTUNREACH); - } - - if (rt->rt_flags & RTF_GATEWAY) { - gw6 = (struct sockaddr_in6 *)rt->rt_gateway; - - /* - * We skip link-layer address resolution and NUD - * if the gateway is not a neighbor from ND point - * of view, regardless of the value of nd_ifinfo.flags. - * The second condition is a bit tricky; we skip - * if the gateway is our own address, which is - * sometimes used to install a route to a p2p link. - */ - if (!nd6_is_addr_neighbor(gw6, ifp) || - in6ifa_ifpwithaddr(ifp, &gw6->sin6_addr)) { - RT_UNLOCK(rt); - /* - * We allow this kind of tricky route only - * when the outgoing interface is p2p. - * XXX: we may need a more generic rule here. - */ - if ((ifp->if_flags & IFF_POINTOPOINT) == 0) - senderr(EHOSTUNREACH); - - goto sendpkt; - } - - if (rt->rt_gwroute == NULL) - goto lookup; - rt = rt->rt_gwroute; - RT_LOCK(rt); /* NB: gwroute */ - if ((rt->rt_flags & RTF_UP) == 0) { - RTFREE_LOCKED(rt); /* unlock gwroute */ - rt = rt0; - rt0->rt_gwroute = NULL; - lookup: - RT_UNLOCK(rt0); - rt = rtalloc1(rt->rt_gateway, 1, 0UL); - if (rt == rt0) { - RT_REMREF(rt0); - RT_UNLOCK(rt0); - senderr(EHOSTUNREACH); - } - RT_LOCK(rt0); - if (rt0->rt_gwroute != NULL) - RTFREE(rt0->rt_gwroute); - rt0->rt_gwroute = rt; - if (rt == NULL) { - RT_UNLOCK(rt0); - senderr(EHOSTUNREACH); - } - } - RT_UNLOCK(rt0); - } - RT_UNLOCK(rt); - } /* * Address resolution or Neighbor Unreachability Detection @@ -2073,20 +1731,25 @@ again: * or an anycast address(i.e. not a multicast). */ - /* Look up the neighbor cache for the nexthop */ - if (rt && (rt->rt_flags & RTF_LLINFO) != 0) - ln = (struct llinfo_nd6 *)rt->rt_llinfo; - else { - /* - * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), - * the condition below is not very efficient. But we believe - * it is tolerable, because this should be a rare case. - */ - if (nd6_is_addr_neighbor(dst, ifp) && - (rt = nd6_lookup(&dst->sin6_addr, 1, ifp)) != NULL) - ln = (struct llinfo_nd6 *)rt->rt_llinfo; - } - if (ln == NULL || rt == NULL) { + flags = ((m != NULL) || (lle != NULL)) ? LLE_EXCLUSIVE : 0; + if (ln == NULL) { + retry: + IF_AFDATA_LOCK(rt->rt_ifp); + ln = lla_lookup(LLTABLE6(ifp), flags, (struct sockaddr *)dst); + IF_AFDATA_UNLOCK(rt->rt_ifp); + if ((ln == NULL) && nd6_is_addr_neighbor(dst, ifp)) { + /* + * Since nd6_is_addr_neighbor() internally calls nd6_lookup(), + * the condition below is not very efficient. But we believe + * it is tolerable, because this should be a rare case. + */ + flags = ND6_CREATE | (m ? ND6_EXCLUSIVE : 0); + IF_AFDATA_LOCK(rt->rt_ifp); + ln = nd6_lookup(&dst->sin6_addr, flags, ifp); + IF_AFDATA_UNLOCK(rt->rt_ifp); + } + } + if (ln == NULL) { if ((ifp->if_flags & IFF_POINTOPOINT) == 0 && !(ND_IFINFO(ifp)->flags & ND6_IFF_PERFORMNUD)) { char ip6buf[INET6_ADDRSTRLEN]; @@ -2096,15 +1759,18 @@ again: ip6_sprintf(ip6buf, &dst->sin6_addr), ln, rt); senderr(EIO); /* XXX: good error? */ } - goto sendpkt; /* send anyway */ } /* We don't have to do link-layer address resolution on a p2p link. */ if ((ifp->if_flags & IFF_POINTOPOINT) != 0 && ln->ln_state < ND6_LLINFO_REACHABLE) { + if ((flags & LLE_EXCLUSIVE) == 0) { + flags |= LLE_EXCLUSIVE; + goto retry; + } ln->ln_state = ND6_LLINFO_STALE; - nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); + nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); } /* @@ -2115,9 +1781,14 @@ again: * (RFC 2461 7.3.3) */ if (ln->ln_state == ND6_LLINFO_STALE) { - ln->ln_asked = 0; + if ((flags & LLE_EXCLUSIVE) == 0) { + flags |= LLE_EXCLUSIVE; + LLE_RUNLOCK(ln); + goto retry; + } + ln->la_asked = 0; ln->ln_state = ND6_LLINFO_DELAY; - nd6_llinfo_settimer(ln, (long)V_nd6_delay * hz); + nd6_llinfo_settimer_locked(ln, (long)V_nd6_delay * hz); } /* @@ -2137,12 +1808,18 @@ again: */ if (ln->ln_state == ND6_LLINFO_NOSTATE) ln->ln_state = ND6_LLINFO_INCOMPLETE; - if (ln->ln_hold) { + + if ((flags & LLE_EXCLUSIVE) == 0) { + flags |= LLE_EXCLUSIVE; + LLE_RUNLOCK(ln); + goto retry; + } + if (ln->la_hold) { struct mbuf *m_hold; int i; - + i = 0; - for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold->m_nextpkt) { + for (m_hold = ln->la_hold; m_hold; m_hold = m_hold->m_nextpkt) { i++; if (m_hold->m_nextpkt == NULL) { m_hold->m_nextpkt = m; @@ -2150,21 +1827,32 @@ again: } } while (i >= V_nd6_maxqueuelen) { - m_hold = ln->ln_hold; - ln->ln_hold = ln->ln_hold->m_nextpkt; + m_hold = ln->la_hold; + ln->la_hold = ln->la_hold->m_nextpkt; m_freem(m_hold); i--; } } else { - ln->ln_hold = m; + ln->la_hold = m; } - + /* + * We did the lookup (no lle arg) so we + * need to do the unlock here + */ + if (lle == NULL) { + if (flags & LLE_EXCLUSIVE) + LLE_WUNLOCK(ln); + else + LLE_RUNLOCK(ln); + } + /* * If there has been no NS for the neighbor after entering the * INCOMPLETE state, send the first solicitation. */ - if (!ND6_LLINFO_PERMANENT(ln) && ln->ln_asked == 0) { - ln->ln_asked++; + if (!ND6_LLINFO_PERMANENT(ln) && ln->la_asked == 0) { + ln->la_asked++; + nd6_llinfo_settimer(ln, (long)ND_IFINFO(ifp)->retrans * hz / 1000); nd6_ns_output(ifp, NULL, &dst->sin6_addr, ln, 0); @@ -2177,23 +1865,82 @@ again: error = ENETDOWN; /* better error? */ goto bad; } + /* + * ln is valid and the caller did not pass in + * an llentry + */ + if (ln && (lle == NULL)) { + if (flags & LLE_EXCLUSIVE) + LLE_WUNLOCK(ln); + else + LLE_RUNLOCK(ln); + } #ifdef MAC mac_netinet6_nd6_send(ifp, m); #endif + if (lle != NULL) { + if (*tail == NULL) + *tail = m; + else + (*tail)->m_nextpkt = m; + return (error); + } if ((ifp->if_flags & IFF_LOOPBACK) != 0) { return ((*ifp->if_output)(origifp, m, (struct sockaddr *)dst, rt)); } - return ((*ifp->if_output)(ifp, m, (struct sockaddr *)dst, rt)); + error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, rt); + return (error); bad: + /* + * ln is valid and the caller did not pass in + * an llentry + */ + if (ln && (lle == NULL)) { + if (flags & LLE_EXCLUSIVE) + LLE_WUNLOCK(ln); + else + LLE_RUNLOCK(ln); + } if (m) m_freem(m); return (error); } #undef senderr + +int +nd6_output_flush(struct ifnet *ifp, struct ifnet *origifp, struct mbuf *chain, + struct sockaddr_in6 *dst, struct rtentry *rt) +{ + struct mbuf *m, *m_head; + struct ifnet *outifp; + int error = 0; + + m_head = chain; + if ((ifp->if_flags & IFF_LOOPBACK) != 0) + outifp = origifp; + else + outifp = ifp; + + while (m_head) { + m = m_head; + m_head = m_head->m_nextpkt; + error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst, rt); + } + + /* + * XXX + * note that intermediate errors are blindly ignored - but this is + * the same convention as used with nd6_output when called by + * nd6_cache_lladdr + */ + return (error); +} + + int nd6_need_cache(struct ifnet *ifp) { @@ -2229,14 +1976,18 @@ nd6_need_cache(struct ifnet *ifp) } } +/* + * the callers of this function need to be re-worked to drop + * the lle lock, drop here for now + */ int nd6_storelladdr(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m, - struct sockaddr *dst, u_char *desten) + struct sockaddr *dst, u_char *desten, struct llentry **lle) { - struct sockaddr_dl *sdl; - struct rtentry *rt; - int error; + struct llentry *ln; + *lle = NULL; + IF_AFDATA_UNLOCK_ASSERT(ifp); if (m->m_flags & M_MCAST) { int i; @@ -2271,48 +2022,42 @@ nd6_storelladdr(struct ifnet *ifp, struct rtentry *rt0, struct mbuf *m, } } - if (rt0 == NULL) { - /* this could happen, if we could not allocate memory */ - m_freem(m); - return (ENOMEM); - } - - error = rt_check(&rt, &rt0, dst); - if (error) { - m_freem(m); - return (error); - } - RT_UNLOCK(rt); - if (rt->rt_gateway->sa_family != AF_LINK) { - printf("nd6_storelladdr: something odd happens\n"); - m_freem(m); - return (EINVAL); - } - sdl = SDL(rt->rt_gateway); - if (sdl->sdl_alen == 0) { - /* this should be impossible, but we bark here for debugging */ - printf("nd6_storelladdr: sdl_alen == 0\n"); + /* + * the entry should have been created in nd6_store_lladdr + */ + IF_AFDATA_LOCK(ifp); + ln = lla_lookup(LLTABLE6(ifp), 0, dst); + IF_AFDATA_UNLOCK(ifp); + if ((ln == NULL) || !(ln->la_flags & LLE_VALID)) { + if (ln) + LLE_RUNLOCK(ln); + /* this could happen, if we could not allocate memory */ m_freem(m); - return (EINVAL); + return (1); } - bcopy(LLADDR(sdl), desten, sdl->sdl_alen); + bcopy(&ln->ll_addr, desten, ifp->if_addrlen); + *lle = ln; + LLE_RUNLOCK(ln); + /* + * A *small* use after free race exists here + */ return (0); } -static void -clear_llinfo_pqueue(struct llinfo_nd6 *ln) +static void +clear_llinfo_pqueue(struct llentry *ln) { struct mbuf *m_hold, *m_hold_next; - for (m_hold = ln->ln_hold; m_hold; m_hold = m_hold_next) { + for (m_hold = ln->la_hold; m_hold; m_hold = m_hold_next) { m_hold_next = m_hold->m_nextpkt; m_hold->m_nextpkt = NULL; m_freem(m_hold); } - ln->ln_hold = NULL; + ln->la_hold = NULL; return; } diff --git a/sys/netinet6/nd6.h b/sys/netinet6/nd6.h index 4d3c06b..f4ccd07 100644 --- a/sys/netinet6/nd6.h +++ b/sys/netinet6/nd6.h @@ -41,20 +41,7 @@ #include <sys/queue.h> #include <sys/callout.h> -struct llinfo_nd6 { - struct llinfo_nd6 *ln_next; - struct llinfo_nd6 *ln_prev; - struct rtentry *ln_rt; - struct mbuf *ln_hold; /* last packet until resolved/timeout */ - long ln_asked; /* number of queries already sent for this addr */ - u_long ln_expire; /* lifetime for NDP state transition */ - short ln_state; /* reachability state */ - short ln_router; /* 2^0: ND6 router bit */ - int ln_byhint; /* # of times we made it reachable by UL hint */ - - long ln_ntick; - struct callout ln_timer_ch; -}; +struct llentry; #define ND6_LLINFO_NOSTATE -2 /* @@ -72,7 +59,7 @@ struct llinfo_nd6 { #define ND6_LLINFO_PROBE 4 #define ND6_IS_LLINFO_PROBREACH(n) ((n)->ln_state > ND6_LLINFO_INCOMPLETE) -#define ND6_LLINFO_PERMANENT(n) (((n)->ln_expire == 0) && ((n)->ln_state > ND6_LLINFO_INCOMPLETE)) +#define ND6_LLINFO_PERMANENT(n) (((n)->la_expire == 0) && ((n)->ln_state > ND6_LLINFO_INCOMPLETE)) struct nd_ifinfo { u_int32_t linkmtu; /* LinkMTU */ @@ -98,6 +85,9 @@ struct nd_ifinfo { */ #define ND6_IFF_DONT_SET_IFROUTE 0x10 +#define ND6_CREATE LLE_CREATE +#define ND6_EXCLUSIVE LLE_EXCLUSIVE + #ifdef _KERNEL #define ND_IFINFO(ifp) \ (((struct in6_ifextra *)(ifp)->if_afdata[AF_INET6])->nd_ifinfo) @@ -336,7 +326,6 @@ extern int nd6_mmaxtries; extern int nd6_useloopback; extern int nd6_maxnudhint; extern int nd6_gctimer; -extern struct llinfo_nd6 llinfo_nd6; extern struct nd_drhead nd_defrouter; extern struct nd_prhead nd_prefix; extern int nd6_debug; @@ -388,23 +377,28 @@ int nd6_is_addr_neighbor __P((struct sockaddr_in6 *, struct ifnet *)); void nd6_option_init __P((void *, int, union nd_opts *)); struct nd_opt_hdr *nd6_option __P((union nd_opts *)); int nd6_options __P((union nd_opts *)); -struct rtentry *nd6_lookup __P((struct in6_addr *, int, struct ifnet *)); +struct llentry *nd6_lookup __P((struct in6_addr *, int, struct ifnet *)); void nd6_setmtu __P((struct ifnet *)); -void nd6_llinfo_settimer __P((struct llinfo_nd6 *, long)); +void nd6_llinfo_settimer __P((struct llentry *, long)); +void nd6_llinfo_settimer_locked __P((struct llentry *, long)); void nd6_timer __P((void *)); void nd6_purge __P((struct ifnet *)); void nd6_nud_hint __P((struct rtentry *, struct in6_addr *, int)); int nd6_resolve __P((struct ifnet *, struct rtentry *, struct mbuf *, struct sockaddr *, u_char *)); -void nd6_rtrequest __P((int, struct rtentry *, struct rt_addrinfo *)); int nd6_ioctl __P((u_long, caddr_t, struct ifnet *)); -struct rtentry *nd6_cache_lladdr __P((struct ifnet *, struct in6_addr *, +struct llentry *nd6_cache_lladdr __P((struct ifnet *, struct in6_addr *, char *, int, int, int)); int nd6_output __P((struct ifnet *, struct ifnet *, struct mbuf *, struct sockaddr_in6 *, struct rtentry *)); +int nd6_output_lle __P((struct ifnet *, struct ifnet *, struct mbuf *, + struct sockaddr_in6 *, struct rtentry *, struct llentry *, + struct mbuf **)); +int nd6_output_flush __P((struct ifnet *, struct ifnet *, struct mbuf *, + struct sockaddr_in6 *, struct rtentry *)); int nd6_need_cache __P((struct ifnet *)); int nd6_storelladdr __P((struct ifnet *, struct rtentry *, struct mbuf *, - struct sockaddr *, u_char *)); + struct sockaddr *, u_char *, struct llentry **)); /* nd6_nbr.c */ void nd6_na_input __P((struct mbuf *, int, int)); @@ -412,7 +406,7 @@ void nd6_na_output __P((struct ifnet *, const struct in6_addr *, const struct in6_addr *, u_long, int, struct sockaddr *)); void nd6_ns_input __P((struct mbuf *, int, int)); void nd6_ns_output __P((struct ifnet *, const struct in6_addr *, - const struct in6_addr *, struct llinfo_nd6 *, int)); + const struct in6_addr *, struct llentry *, int)); caddr_t nd6_ifptomac __P((struct ifnet *)); void nd6_dad_start __P((struct ifaddr *, int)); void nd6_dad_stop __P((struct ifaddr *)); diff --git a/sys/netinet6/nd6_nbr.c b/sys/netinet6/nd6_nbr.c index ecfad0e..2dcabdb 100644 --- a/sys/netinet6/nd6_nbr.c +++ b/sys/netinet6/nd6_nbr.c @@ -41,6 +41,8 @@ __FBSDID("$FreeBSD$"); #include <sys/param.h> #include <sys/systm.h> #include <sys/malloc.h> +#include <sys/lock.h> +#include <sys/rwlock.h> #include <sys/mbuf.h> #include <sys/socket.h> #include <sys/sockio.h> @@ -63,6 +65,8 @@ __FBSDID("$FreeBSD$"); #include <netinet/in.h> #include <netinet/in_var.h> +#include <net/if_llatbl.h> +#define L3_ADDR_SIN6(le) ((struct sockaddr_in6 *) L3_ADDR(le)) #include <netinet6/in6_var.h> #include <netinet6/in6_ifattach.h> #include <netinet/ip6.h> @@ -167,7 +171,7 @@ nd6_ns_input(struct mbuf *m, int off, int icmp6len) src_sa6.sin6_family = AF_INET6; src_sa6.sin6_len = sizeof(src_sa6); src_sa6.sin6_addr = saddr6; - if (!nd6_is_addr_neighbor(&src_sa6, ifp)) { + if (nd6_is_addr_neighbor(&src_sa6, ifp) == 0) { nd6log((LOG_INFO, "nd6_ns_input: " "NS packet from non-neighbor\n")); goto bad; @@ -378,8 +382,8 @@ nd6_ns_input(struct mbuf *m, int off, int icmp6len) * dad - duplicate address detection */ void -nd6_ns_output(struct ifnet *ifp, const struct in6_addr *daddr6, - const struct in6_addr *taddr6, struct llinfo_nd6 *ln, int dad) +nd6_ns_output(struct ifnet *ifp, const struct in6_addr *daddr6, + const struct in6_addr *taddr6, struct llentry *ln, int dad) { INIT_VNET_INET6(ifp->if_vnet); struct mbuf *m; @@ -470,14 +474,14 @@ nd6_ns_output(struct ifnet *ifp, const struct in6_addr *daddr6, struct ip6_hdr *hip6; /* hold ip6 */ struct in6_addr *hsrc = NULL; - if (ln && ln->ln_hold) { + if (ln && ln->la_hold) { /* - * assuming every packet in ln_hold has the same IP + * assuming every packet in la_hold has the same IP * header */ - hip6 = mtod(ln->ln_hold, struct ip6_hdr *); + hip6 = mtod(ln->la_hold, struct ip6_hdr *); /* XXX pullup? */ - if (sizeof(*hip6) < ln->ln_hold->m_len) + if (sizeof(*hip6) < ln->la_hold->m_len) hsrc = &hip6->ip6_src; else hsrc = NULL; @@ -600,10 +604,10 @@ nd6_na_input(struct mbuf *m, int off, int icmp6len) char *lladdr = NULL; int lladdrlen = 0; struct ifaddr *ifa; - struct llinfo_nd6 *ln; - struct rtentry *rt; - struct sockaddr_dl *sdl; + struct llentry *ln = NULL; union nd_opts ndopts; + struct mbuf *chain = NULL; + struct sockaddr_in6 sin6; char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; if (ip6->ip6_hlim != 255) { @@ -697,35 +701,37 @@ nd6_na_input(struct mbuf *m, int off, int icmp6len) * If no neighbor cache entry is found, NA SHOULD silently be * discarded. */ - rt = nd6_lookup(&taddr6, 0, ifp); - if ((rt == NULL) || - ((ln = (struct llinfo_nd6 *)rt->rt_llinfo) == NULL) || - ((sdl = SDL(rt->rt_gateway)) == NULL)) + IF_AFDATA_LOCK(ifp); + ln = nd6_lookup(&taddr6, LLE_EXCLUSIVE, ifp); + IF_AFDATA_UNLOCK(ifp); + if (ln == NULL) { goto freeit; + } if (ln->ln_state == ND6_LLINFO_INCOMPLETE) { /* * If the link-layer has address, and no lladdr option came, * discard the packet. */ - if (ifp->if_addrlen && lladdr == NULL) + if (ifp->if_addrlen && lladdr == NULL) { goto freeit; + } /* * Record link-layer address, and update the state. */ - sdl->sdl_alen = ifp->if_addrlen; - bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen); + bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen); + ln->la_flags |= LLE_VALID; if (is_solicited) { ln->ln_state = ND6_LLINFO_REACHABLE; ln->ln_byhint = 0; if (!ND6_LLINFO_PERMANENT(ln)) { - nd6_llinfo_settimer(ln, - (long)ND_IFINFO(rt->rt_ifp)->reachable * hz); + nd6_llinfo_settimer_locked(ln, + (long)ND_IFINFO(ln->lle_tbl->llt_ifp)->reachable * hz); } } else { ln->ln_state = ND6_LLINFO_STALE; - nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); + nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); } if ((ln->ln_router = is_router) != 0) { /* @@ -744,8 +750,8 @@ nd6_na_input(struct mbuf *m, int off, int icmp6len) if (lladdr == NULL) llchange = 0; else { - if (sdl->sdl_alen) { - if (bcmp(lladdr, LLADDR(sdl), ifp->if_addrlen)) + if (ln->la_flags & LLE_VALID) { + if (bcmp(lladdr, &ln->ll_addr, ifp->if_addrlen)) llchange = 1; else llchange = 0; @@ -779,7 +785,7 @@ nd6_na_input(struct mbuf *m, int off, int icmp6len) */ if (ln->ln_state == ND6_LLINFO_REACHABLE) { ln->ln_state = ND6_LLINFO_STALE; - nd6_llinfo_settimer(ln, (long)V_nd6_gctimer * hz); + nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); } goto freeit; } else if (is_override /* (2a) */ @@ -789,8 +795,8 @@ nd6_na_input(struct mbuf *m, int off, int icmp6len) * Update link-local address, if any. */ if (lladdr != NULL) { - sdl->sdl_alen = ifp->if_addrlen; - bcopy(lladdr, LLADDR(sdl), ifp->if_addrlen); + bcopy(lladdr, &ln->ll_addr, ifp->if_addrlen); + ln->la_flags |= LLE_VALID; } /* @@ -802,13 +808,13 @@ nd6_na_input(struct mbuf *m, int off, int icmp6len) ln->ln_state = ND6_LLINFO_REACHABLE; ln->ln_byhint = 0; if (!ND6_LLINFO_PERMANENT(ln)) { - nd6_llinfo_settimer(ln, + nd6_llinfo_settimer_locked(ln, (long)ND_IFINFO(ifp)->reachable * hz); } } else { if (lladdr != NULL && llchange) { ln->ln_state = ND6_LLINFO_STALE; - nd6_llinfo_settimer(ln, + nd6_llinfo_settimer_locked(ln, (long)V_nd6_gctimer * hz); } } @@ -822,9 +828,8 @@ nd6_na_input(struct mbuf *m, int off, int icmp6len) */ struct nd_defrouter *dr; struct in6_addr *in6; - int s; - in6 = &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr; + in6 = &L3_ADDR_SIN6(ln)->sin6_addr; /* * Lock to protect the default router list. @@ -832,8 +837,7 @@ nd6_na_input(struct mbuf *m, int off, int icmp6len) * is only called under the network software interrupt * context. However, we keep it just for safety. */ - s = splnet(); - dr = defrouter_lookup(in6, ifp); + dr = defrouter_lookup(in6, ln->lle_tbl->llt_ifp); if (dr) defrtrlist_del(dr); else if (!V_ip6_forwarding) { @@ -846,21 +850,23 @@ nd6_na_input(struct mbuf *m, int off, int icmp6len) */ rt6_flush(&ip6->ip6_src, ifp); } - splx(s); } ln->ln_router = is_router; } - rt->rt_flags &= ~RTF_REJECT; - ln->ln_asked = 0; - if (ln->ln_hold) { + /* XXX - QL + * Does this matter? + * rt->rt_flags &= ~RTF_REJECT; + */ + ln->la_asked = 0; + if (ln->la_hold) { struct mbuf *m_hold, *m_hold_next; /* - * reset the ln_hold in advance, to explicitly - * prevent a ln_hold lookup in nd6_output() + * reset the la_hold in advance, to explicitly + * prevent a la_hold lookup in nd6_output() * (wouldn't happen, though...) */ - for (m_hold = ln->ln_hold; + for (m_hold = ln->la_hold, ln->la_hold = NULL; m_hold; m_hold = m_hold_next) { m_hold_next = m_hold->m_nextpkt; m_hold->m_nextpkt = NULL; @@ -868,17 +874,25 @@ nd6_na_input(struct mbuf *m, int off, int icmp6len) * we assume ifp is not a loopback here, so just set * the 2nd argument as the 1st one. */ - nd6_output(ifp, ifp, m_hold, - (struct sockaddr_in6 *)rt_key(rt), rt); + nd6_output_lle(ifp, ifp, m_hold, L3_ADDR_SIN6(ln), NULL, ln, &chain); } - ln->ln_hold = NULL; } - freeit: + if (ln) { + if (chain) + memcpy(&sin6, L3_ADDR_SIN6(ln), sizeof(sin6)); + LLE_WUNLOCK(ln); + + if (chain) + nd6_output_flush(ifp, ifp, chain, &sin6, NULL); + } m_freem(m); return; bad: + if (ln) + LLE_WUNLOCK(ln); + V_icmp6stat.icp6s_badna++; m_freem(m); } diff --git a/sys/netinet6/nd6_rtr.c b/sys/netinet6/nd6_rtr.c index c9ed36d..7b4af3c 100644 --- a/sys/netinet6/nd6_rtr.c +++ b/sys/netinet6/nd6_rtr.c @@ -58,6 +58,7 @@ __FBSDID("$FreeBSD$"); #include <net/vnet.h> #include <netinet/in.h> +#include <net/if_llatbl.h> #include <netinet6/in6_var.h> #include <netinet6/in6_ifattach.h> #include <netinet/ip6.h> @@ -471,10 +472,8 @@ defrouter_addreq(struct nd_defrouter *new) (struct sockaddr *)&gate, (struct sockaddr *)&mask, RTF_GATEWAY, &newrt); if (newrt) { - RT_LOCK(newrt); nd6_rtmsg(RTM_ADD, newrt); /* tell user process */ - RT_REMREF(newrt); - RT_UNLOCK(newrt); + RTFREE(newrt); } if (error == 0) new->installed = 1; @@ -615,8 +614,7 @@ defrouter_select(void) INIT_VNET_INET6(curvnet); int s = splnet(); struct nd_defrouter *dr, *selected_dr = NULL, *installed_dr = NULL; - struct rtentry *rt = NULL; - struct llinfo_nd6 *ln = NULL; + struct llentry *ln = NULL; /* * This function should be called only when acting as an autoconfigured @@ -648,12 +646,13 @@ defrouter_select(void) */ for (dr = TAILQ_FIRST(&V_nd_defrouter); dr; dr = TAILQ_NEXT(dr, dr_entry)) { + IF_AFDATA_LOCK(dr->ifp); if (selected_dr == NULL && - (rt = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) && - (ln = (struct llinfo_nd6 *)rt->rt_llinfo) && + (ln = nd6_lookup(&dr->rtaddr, 0, dr->ifp)) && ND6_IS_LLINFO_PROBREACH(ln)) { selected_dr = dr; } + IF_AFDATA_UNLOCK(dr->ifp); if (dr->installed && installed_dr == NULL) installed_dr = dr; @@ -676,12 +675,14 @@ defrouter_select(void) selected_dr = TAILQ_FIRST(&V_nd_defrouter); else selected_dr = TAILQ_NEXT(installed_dr, dr_entry); - } else if (installed_dr && - (rt = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) && - (ln = (struct llinfo_nd6 *)rt->rt_llinfo) && - ND6_IS_LLINFO_PROBREACH(ln) && - rtpref(selected_dr) <= rtpref(installed_dr)) { - selected_dr = installed_dr; + } else if (installed_dr) { + IF_AFDATA_LOCK(installed_dr->ifp); + if ((ln = nd6_lookup(&installed_dr->rtaddr, 0, installed_dr->ifp)) && + ND6_IS_LLINFO_PROBREACH(ln) && + rtpref(selected_dr) <= rtpref(installed_dr)) { + selected_dr = installed_dr; + } + IF_AFDATA_UNLOCK(installed_dr->ifp); } /* @@ -1323,18 +1324,19 @@ static struct nd_pfxrouter * find_pfxlist_reachable_router(struct nd_prefix *pr) { struct nd_pfxrouter *pfxrtr; - struct rtentry *rt; - struct llinfo_nd6 *ln; + struct llentry *ln; for (pfxrtr = LIST_FIRST(&pr->ndpr_advrtrs); pfxrtr; pfxrtr = LIST_NEXT(pfxrtr, pfr_entry)) { - if ((rt = nd6_lookup(&pfxrtr->router->rtaddr, 0, + IF_AFDATA_LOCK(pfxrtr->router->ifp); + if ((ln = nd6_lookup(&pfxrtr->router->rtaddr, 0, pfxrtr->router->ifp)) && - (ln = (struct llinfo_nd6 *)rt->rt_llinfo) && - ND6_IS_LLINFO_PROBREACH(ln)) + ND6_IS_LLINFO_PROBREACH(ln)) { + IF_AFDATA_UNLOCK(pfxrtr->router->ifp); break; /* found */ + } + IF_AFDATA_UNLOCK(pfxrtr->router->ifp); } - return (pfxrtr); } @@ -1541,8 +1543,10 @@ nd6_prefix_onlink(struct nd_prefix *pr) struct nd_prefix *opr; u_long rtflags; int error = 0; + struct radix_node_head *rnh; struct rtentry *rt = NULL; char ip6buf[INET6_ADDRSTRLEN]; + struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK}; /* sanity check */ if ((pr->ndpr_stateflags & NDPRF_ONLINK) != 0) { @@ -1609,21 +1613,24 @@ nd6_prefix_onlink(struct nd_prefix *pr) bzero(&mask6, sizeof(mask6)); mask6.sin6_len = sizeof(mask6); mask6.sin6_addr = pr->ndpr_mask; - rtflags = ifa->ifa_flags | RTF_CLONING | RTF_UP; - if (nd6_need_cache(ifp)) { - /* explicitly set in case ifa_flags does not set the flag. */ - rtflags |= RTF_CLONING; - } else { - /* - * explicitly clear the cloning bit in case ifa_flags sets it. - */ - rtflags &= ~RTF_CLONING; - } + rtflags = ifa->ifa_flags | RTF_UP; error = rtrequest(RTM_ADD, (struct sockaddr *)&pr->ndpr_prefix, ifa->ifa_addr, (struct sockaddr *)&mask6, rtflags, &rt); if (error == 0) { - if (rt != NULL) /* this should be non NULL, though */ + if (rt != NULL) /* this should be non NULL, though */ { + rnh = V_rt_tables[rt->rt_fibnum][AF_INET6]; + RADIX_NODE_HEAD_LOCK(rnh); + RT_LOCK(rt); + if (!rt_setgate(rt, rt_key(rt), (struct sockaddr *)&null_sdl)) { + ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type = + rt->rt_ifp->if_type; + ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index = + rt->rt_ifp->if_index; + } + RADIX_NODE_HEAD_UNLOCK(rnh); nd6_rtmsg(RTM_ADD, rt); + RT_UNLOCK(rt); + } pr->ndpr_stateflags |= NDPRF_ONLINK; } else { char ip6bufg[INET6_ADDRSTRLEN], ip6bufm[INET6_ADDRSTRLEN]; diff --git a/sys/netinet6/vinet6.h b/sys/netinet6/vinet6.h index d6c3f33..e271d4f 100644 --- a/sys/netinet6/vinet6.h +++ b/sys/netinet6/vinet6.h @@ -78,7 +78,6 @@ struct vnet_inet6 { int _nd6_inuse; int _nd6_allocated; int _nd6_onlink_ns_rfc4861; - struct llinfo_nd6 _llinfo_nd6; struct nd_drhead _nd_defrouter; struct nd_prhead _nd_prefix; struct ifnet * _nd6_defifp; diff --git a/sys/sys/param.h b/sys/sys/param.h index 497b362..f8a4b2a 100644 --- a/sys/sys/param.h +++ b/sys/sys/param.h @@ -57,7 +57,7 @@ * is created, otherwise 1. */ #undef __FreeBSD_version -#define __FreeBSD_version 800058 /* Master, propagated to newvers */ +#define __FreeBSD_version 800059 /* Master, propagated to newvers */ #ifndef LOCORE #include <sys/types.h> diff --git a/usr.bin/netstat/route.c b/usr.bin/netstat/route.c index 01311e5..596fcd5 100644 --- a/usr.bin/netstat/route.c +++ b/usr.bin/netstat/route.c @@ -89,17 +89,23 @@ struct bits { { RTF_DYNAMIC, 'D' }, { RTF_MODIFIED, 'M' }, { RTF_DONE, 'd' }, /* Completed -- for routing messages only */ - { RTF_CLONING, 'C' }, { RTF_XRESOLVE, 'X' }, - { RTF_LLINFO, 'L' }, { RTF_STATIC, 'S' }, { RTF_PROTO1, '1' }, { RTF_PROTO2, '2' }, - { RTF_WASCLONED,'W' }, { RTF_PRCLONING,'c' }, { RTF_PROTO3, '3' }, { RTF_BLACKHOLE,'B' }, { RTF_BROADCAST,'b' }, +#ifdef RTF_LLINFO + { RTF_LLINFO, 'L' }, +#endif +#ifdef RTF_WASCLONED + { RTF_WASCLONED,'W' }, +#endif +#ifdef RTF_CLONING + { RTF_CLONING, 'C' }, +#endif { 0 , 0 } }; @@ -316,23 +322,12 @@ static void size_cols_rtentry(struct rtentry *rt) { static struct ifnet ifnet, *lastif; - struct rtentry parent; static char buffer[100]; const char *bp; struct sockaddr *sa; sa_u addr, mask; int len; - /* - * Don't print protocol-cloned routes unless -a. - */ - if (rt->rt_flags & RTF_WASCLONED && !aflag) { - if (kget(rt->rt_parent, parent) != 0) - return; - if (parent.rt_flags & RTF_PRCLONING) - return; - } - bzero(&addr, sizeof(addr)); if ((sa = kgetsa(rt_key(rt)))) bcopy(sa, &addr, sa->sa_len); @@ -352,7 +347,7 @@ size_cols_rtentry(struct rtentry *rt) wid_flags = MAX(len, wid_flags); if (addr.u_sa.sa_family == AF_INET || Wflag) { - len = snprintf(buffer, sizeof(buffer), "%ld", rt->rt_refcnt); + len = snprintf(buffer, sizeof(buffer), "%d", rt->rt_refcnt); wid_refs = MAX(len, wid_refs); len = snprintf(buffer, sizeof(buffer), "%lu", rt->rt_use); wid_use = MAX(len, wid_use); @@ -761,22 +756,11 @@ static void p_rtentry(struct rtentry *rt) { static struct ifnet ifnet, *lastif; - struct rtentry parent; static char buffer[128]; static char prettyname[128]; struct sockaddr *sa; sa_u addr, mask; - /* - * Don't print protocol-cloned routes unless -a. - */ - if (rt->rt_flags & RTF_WASCLONED && !aflag) { - if (kget(rt->rt_parent, parent) != 0) - return; - if (parent.rt_flags & RTF_PRCLONING) - return; - } - bzero(&addr, sizeof(addr)); if ((sa = kgetsa(rt_key(rt)))) bcopy(sa, &addr, sa->sa_len); @@ -788,7 +772,7 @@ p_rtentry(struct rtentry *rt) snprintf(buffer, sizeof(buffer), "%%-%d.%ds ", wid_flags, wid_flags); p_flags(rt->rt_flags, buffer); if (addr.u_sa.sa_family == AF_INET || Wflag) { - printf("%*ld %*lu ", wid_refs, rt->rt_refcnt, + printf("%*d %*lu ", wid_refs, rt->rt_refcnt, wid_use, rt->rt_use); if (Wflag) { if (rt->rt_rmx.rmx_mtu != 0) diff --git a/usr.sbin/arp/arp.c b/usr.sbin/arp/arp.c index c8ef3e9..4154c3a1 100644 --- a/usr.sbin/arp/arp.c +++ b/usr.sbin/arp/arp.c @@ -370,7 +370,6 @@ set(int argc, char **argv) if (addr->sin_addr.s_addr != dst->sin_addr.s_addr) break; if (sdl->sdl_family == AF_LINK && - (rtm->rtm_flags & RTF_LLINFO) && !(rtm->rtm_flags & RTF_GATEWAY) && valid_type(sdl->sdl_type) ) break; @@ -426,25 +425,46 @@ delete(char *host, int do_proxy) struct sockaddr_inarp *addr, *dst; struct rt_msghdr *rtm; struct sockaddr_dl *sdl; + struct sockaddr_dl sdl_m; dst = getaddr(host); if (dst == NULL) return (1); dst->sin_other = do_proxy; + + /* + * setup the data structure to notify the kernel + * it is the ARP entry the RTM_GET is interested + * in + */ + bzero(&sdl_m, sizeof(sdl_m)); + sdl_m.sdl_len = sizeof(sdl_m); + sdl_m.sdl_family = AF_LINK; + for (;;) { /* try twice */ - rtm = rtmsg(RTM_GET, dst, NULL); + rtm = rtmsg(RTM_GET, dst, &sdl_m); if (rtm == NULL) { warn("%s", host); return (1); } addr = (struct sockaddr_inarp *)(rtm + 1); sdl = (struct sockaddr_dl *)(SA_SIZE(addr) + (char *)addr); - if (addr->sin_addr.s_addr == dst->sin_addr.s_addr && - sdl->sdl_family == AF_LINK && - (rtm->rtm_flags & RTF_LLINFO) && + + /* + * With the new L2/L3 restructure, the route + * returned is a prefix route. The important + * piece of information from the previous + * RTM_GET is the interface index. In the + * case of ECMP, the kernel will traverse + * the route group for the given entry. + */ + if (sdl->sdl_family == AF_LINK && !(rtm->rtm_flags & RTF_GATEWAY) && - valid_type(sdl->sdl_type) ) - break; /* found it */ + valid_type(sdl->sdl_type) ) { + addr->sin_addr.s_addr = dst->sin_addr.s_addr; + break; + } + if (dst->sin_other & SIN_PROXY) { fprintf(stderr, "delete: cannot locate %s\n",host); return (1); @@ -478,7 +498,11 @@ search(u_long addr, action_fn *action) mib[2] = 0; mib[3] = AF_INET; mib[4] = NET_RT_FLAGS; +#ifdef RTF_LLINFO mib[5] = RTF_LLINFO; +#else + mib[5] = 0; +#endif if (sysctl(mib, 6, NULL, &needed, NULL, 0) < 0) err(1, "route-sysctl-estimate"); if (needed == 0) /* empty table */ @@ -563,15 +587,9 @@ print_entry(struct sockaddr_dl *sdl, printf(" permanent"); if (addr->sin_other & SIN_PROXY) printf(" published (proxy only)"); - if (rtm->rtm_addrs & RTA_NETMASK) { - addr = (struct sockaddr_inarp *) - (SA_SIZE(sdl) + (char *)sdl); - if (addr->sin_addr.s_addr == 0xffffffff) - printf(" published"); - if (addr->sin_len != 8) - printf("(weird)"); - } - switch(sdl->sdl_type) { + if (rtm->rtm_flags & RTF_ANNOUNCE) + printf(" published"); + switch(sdl->sdl_type) { case IFT_ETHER: printf(" [ethernet]"); break; diff --git a/usr.sbin/ndp/ndp.c b/usr.sbin/ndp/ndp.c index 86bd124..a158fb2 100644 --- a/usr.sbin/ndp/ndp.c +++ b/usr.sbin/ndp/ndp.c @@ -114,6 +114,11 @@ ((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long)) #define ADVANCE(x, n) (x += ROUNDUP((n)->sa_len)) +#define NEXTADDR(w, s) \ + if (rtm->rtm_addrs & (w)) { \ + bcopy((char *)&s, cp, sizeof(s)); cp += sizeof(s);} + + static pid_t pid; static int nflag; static int tflag; @@ -428,7 +433,6 @@ set(argc, argv) sdl = (struct sockaddr_dl *)(ROUNDUP(sin->sin6_len) + (char *)sin); if (IN6_ARE_ADDR_EQUAL(&sin->sin6_addr, &sin_m.sin6_addr)) { if (sdl->sdl_family == AF_LINK && - (rtm->rtm_flags & RTF_LLINFO) && !(rtm->rtm_flags & RTF_GATEWAY)) { switch (sdl->sdl_type) { case IFT_ETHER: case IFT_FDDI: case IFT_ISO88023: @@ -499,6 +503,7 @@ delete(host) { struct sockaddr_in6 *sin = &sin_m; register struct rt_msghdr *rtm = &m_rtmsg.m_rtm; + register char *cp = m_rtmsg.m_space; struct sockaddr_dl *sdl; struct addrinfo hints, *res; int gai_error; @@ -529,7 +534,6 @@ delete(host) sdl = (struct sockaddr_dl *)(ROUNDUP(sin->sin6_len) + (char *)sin); if (IN6_ARE_ADDR_EQUAL(&sin->sin6_addr, &sin_m.sin6_addr)) { if (sdl->sdl_family == AF_LINK && - (rtm->rtm_flags & RTF_LLINFO) && !(rtm->rtm_flags & RTF_GATEWAY)) { goto delete; } @@ -545,6 +549,11 @@ delete: printf("cannot locate %s\n", host); return (1); } + /* + * need to reinit the field because it has rt_key + * but we want the actual address + */ + NEXTADDR(RTA_DST, sin_m); if (rtmsg(RTM_DELETE) == 0) { struct sockaddr_in6 s6 = *sin; /* XXX: for safety */ @@ -603,7 +612,11 @@ again:; mib[2] = 0; mib[3] = AF_INET6; mib[4] = NET_RT_FLAGS; +#ifdef RTF_LLINFO mib[5] = RTF_LLINFO; +#else + mib[5] = 0; +#endif if (sysctl(mib, 6, NULL, &needed, NULL, 0) < 0) err(1, "sysctl(PF_ROUTE estimate)"); if (needed > 0) { @@ -893,9 +906,6 @@ rtmsg(cmd) case RTM_GET: rtm->rtm_addrs |= RTA_DST; } -#define NEXTADDR(w, s) \ - if (rtm->rtm_addrs & (w)) { \ - bcopy((char *)&s, cp, sizeof(s)); cp += SA_SIZE(&s);} NEXTADDR(RTA_DST, sin_m); NEXTADDR(RTA_GATEWAY, sdl_m); @@ -1616,3 +1626,5 @@ ts_print(tvp) (void)printf("%02d:%02d:%02d.%06u ", s / 3600, (s % 3600) / 60, s % 60, (u_int32_t)tvp->tv_usec); } + +#undef NEXTADDR diff --git a/usr.sbin/ppp/route.c b/usr.sbin/ppp/route.c index fdc6002..f5efe63 100644 --- a/usr.sbin/ppp/route.c +++ b/usr.sbin/ppp/route.c @@ -157,13 +157,21 @@ static struct bits { { RTF_DYNAMIC, 'D' }, { RTF_MODIFIED, 'M' }, { RTF_DONE, 'd' }, - { RTF_CLONING, 'C' }, { RTF_XRESOLVE, 'X' }, - { RTF_LLINFO, 'L' }, +#ifdef RTF_CLONING + { RTF_CLONING, 'C' }, +#endif { RTF_STATIC, 'S' }, { RTF_PROTO1, '1' }, { RTF_PROTO2, '2' }, { RTF_BLACKHOLE, 'B' }, + +#ifdef RTF_LLINFO + { RTF_LLINFO, 'L' }, +#endif +#ifdef RTF_CLONING + { RTF_CLONING, 'C' }, +#endif #ifdef RTF_WASCLONED { RTF_WASCLONED, 'W' }, #endif diff --git a/usr.sbin/route6d/route6d.c b/usr.sbin/route6d/route6d.c index 17ddb58..0205a03 100644 --- a/usr.sbin/route6d/route6d.c +++ b/usr.sbin/route6d/route6d.c @@ -1752,7 +1752,6 @@ rtrecv() continue; case RTM_LOSING: case RTM_MISS: - case RTM_RESOLVE: case RTM_GET: case RTM_LOCK: /* nothing to be done here */ @@ -1790,7 +1789,6 @@ rtrecv() case RTM_ADD: case RTM_LOSING: case RTM_MISS: - case RTM_RESOLVE: case RTM_GET: case RTM_LOCK: /* should already be handled */ @@ -2141,10 +2139,7 @@ ifrt(ifcp, again) rrt->rrt_info.rip6_tag = htons(routetag & 0xffff); rrt->rrt_info.rip6_metric = 1 + ifcp->ifc_metric; rrt->rrt_info.rip6_plen = ifa->ifa_plen; - if (ifa->ifa_plen == 128) - rrt->rrt_flags = RTF_HOST; - else - rrt->rrt_flags = RTF_CLONING; + rrt->rrt_flags = RTF_HOST; rrt->rrt_rflags |= RRTF_CHANGED; applyplen(&rrt->rrt_info.rip6_dest, ifa->ifa_plen); memset(&rrt->rrt_gw, 0, sizeof(struct in6_addr)); @@ -2433,7 +2428,6 @@ do { \ RTTYPE("LOCK", RTM_LOCK); RTTYPE("OLDADD", RTM_OLDADD); RTTYPE("OLDDEL", RTM_OLDDEL); - RTTYPE("RESOLVE", RTM_RESOLVE); RTTYPE("NEWADDR", RTM_NEWADDR); RTTYPE("DELADDR", RTM_DELADDR); RTTYPE("IFINFO", RTM_IFINFO); @@ -2484,7 +2478,9 @@ do { \ #ifdef RTF_MASK RTFLAG("m", RTF_MASK); #endif +#ifdef RTF_CLONING RTFLAG("C", RTF_CLONING); +#endif #ifdef RTF_CLONED RTFLAG("c", RTF_CLONED); #endif @@ -2495,7 +2491,9 @@ do { \ RTFLAG("W", RTF_WASCLONED); #endif RTFLAG("X", RTF_XRESOLVE); +#ifdef RTF_LLINFO RTFLAG("L", RTF_LLINFO); +#endif RTFLAG("S", RTF_STATIC); RTFLAG("B", RTF_BLACKHOLE); #ifdef RTF_PROTO3 @@ -2631,7 +2629,7 @@ rt_entry(rtm, again) sin6_dst = sin6_gw = sin6_mask = sin6_genmask = sin6_ifp = 0; if ((rtm->rtm_flags & RTF_UP) == 0 || rtm->rtm_flags & - (RTF_CLONING|RTF_XRESOLVE|RTF_LLINFO|RTF_BLACKHOLE)) { + (RTF_XRESOLVE|RTF_BLACKHOLE)) { return; /* not interested in the link route */ } /* do not look at cloned routes */ |
