2 * Copyright (C) 2012-2013 Steven Barth <steven@midlink.org>
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License v2 as published by
6 * the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful,
9 * but WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11 * GNU General Public License for more details.
22 #include <arpa/inet.h>
23 #include <sys/socket.h>
24 #include <net/ethernet.h>
25 #include <netinet/ip6.h>
26 #include <netinet/icmp6.h>
27 #include <netpacket/packet.h>
29 #include <linux/rtnetlink.h>
30 #include <linux/filter.h>
32 #include <netlink/msg.h>
33 #include <netlink/socket.h>
34 #include <netlink/attr.h>
41 struct odhcpd_event ev
;
46 static void handle_solicit(void *addr
, void *data
, size_t len
,
47 struct interface
*iface
, void *dest
);
48 static void handle_rtnl_event(struct odhcpd_event
*ev
);
49 static int cb_rtnl_valid(struct nl_msg
*msg
, void *arg
);
50 static void catch_rtnl_err(struct odhcpd_event
*e
, int error
);
52 static int ping_socket
= -1;
53 static struct event_socket rtnl_event
= {
55 .uloop
= {.fd
= - 1, },
57 .handle_error
= catch_rtnl_err
,
58 .recv_msgs
= handle_rtnl_event
,
61 .sock_bufsize
= 133120,
64 // Filter ICMPv6 messages of type neighbor soliciation
65 static struct sock_filter bpf
[] = {
66 BPF_STMT(BPF_LD
| BPF_B
| BPF_ABS
, offsetof(struct ip6_hdr
, ip6_nxt
)),
67 BPF_JUMP(BPF_JMP
| BPF_JEQ
| BPF_K
, IPPROTO_ICMPV6
, 0, 3),
68 BPF_STMT(BPF_LD
| BPF_B
| BPF_ABS
, sizeof(struct ip6_hdr
) +
69 offsetof(struct icmp6_hdr
, icmp6_type
)),
70 BPF_JUMP(BPF_JMP
| BPF_JEQ
| BPF_K
, ND_NEIGHBOR_SOLICIT
, 0, 1),
71 BPF_STMT(BPF_RET
| BPF_K
, 0xffffffff),
72 BPF_STMT(BPF_RET
| BPF_K
, 0),
74 static const struct sock_fprog bpf_prog
= {sizeof(bpf
) / sizeof(*bpf
), bpf
};
77 // Initialize NDP-proxy
82 rtnl_event
.sock
= odhcpd_create_nl_socket(NETLINK_ROUTE
);
86 rtnl_event
.ev
.uloop
.fd
= nl_socket_get_fd(rtnl_event
.sock
);
88 if (nl_socket_set_buffer_size(rtnl_event
.sock
, rtnl_event
.sock_bufsize
, 0))
91 nl_socket_disable_seq_check(rtnl_event
.sock
);
93 nl_socket_modify_cb(rtnl_event
.sock
, NL_CB_VALID
, NL_CB_CUSTOM
,
96 // Receive IPv6 address, IPv6 routes and neighbor events
97 if (nl_socket_add_memberships(rtnl_event
.sock
, RTNLGRP_IPV6_IFADDR
,
98 RTNLGRP_IPV6_ROUTE
, RTNLGRP_NEIGH
, 0))
101 odhcpd_register(&rtnl_event
.ev
);
103 // Open ICMPv6 socket
104 ping_socket
= socket(AF_INET6
, SOCK_RAW
| SOCK_CLOEXEC
, IPPROTO_ICMPV6
);
105 if (ping_socket
< 0) {
106 syslog(LOG_ERR
, "Unable to open raw socket: %s", strerror(errno
));
110 setsockopt(ping_socket
, IPPROTO_RAW
, IPV6_CHECKSUM
, &val
, sizeof(val
));
112 // This is required by RFC 4861
114 setsockopt(ping_socket
, IPPROTO_IPV6
, IPV6_MULTICAST_HOPS
, &val
, sizeof(val
));
115 setsockopt(ping_socket
, IPPROTO_IPV6
, IPV6_UNICAST_HOPS
, &val
, sizeof(val
));
117 // Filter all packages, we only want to send
118 struct icmp6_filter filt
;
119 ICMP6_FILTER_SETBLOCKALL(&filt
);
120 setsockopt(ping_socket
, IPPROTO_ICMPV6
, ICMP6_FILTER
, &filt
, sizeof(filt
));
125 if (rtnl_event
.sock
) {
126 nl_socket_free(rtnl_event
.sock
);
127 rtnl_event
.sock
= NULL
;
128 rtnl_event
.ev
.uloop
.fd
= -1;
134 static void dump_neigh_table(const bool proxy
)
138 .ndm_family
= AF_INET6
,
139 .ndm_flags
= proxy
? NTF_PROXY
: 0,
142 msg
= nlmsg_alloc_simple(RTM_GETNEIGH
, NLM_F_REQUEST
| NLM_F_DUMP
);
146 nlmsg_append(msg
, &ndm
, sizeof(ndm
), 0);
148 nl_send_auto_complete(rtnl_event
.sock
, msg
);
153 static void dump_addr_table(void)
156 struct ifaddrmsg ifa
= {
157 .ifa_family
= AF_INET6
,
160 msg
= nlmsg_alloc_simple(RTM_GETADDR
, NLM_F_REQUEST
| NLM_F_DUMP
);
164 nlmsg_append(msg
, &ifa
, sizeof(ifa
), 0);
166 nl_send_auto_complete(rtnl_event
.sock
, msg
);
171 int setup_ndp_interface(struct interface
*iface
, bool enable
)
174 bool dump_neigh
= false;
177 snprintf(procbuf
, sizeof(procbuf
), "/proc/sys/net/ipv6/conf/%s/proxy_ndp", iface
->ifname
);
178 procfd
= open(procbuf
, O_WRONLY
);
185 if (iface
->ndp_event
.uloop
.fd
> 0) {
186 uloop_fd_delete(&iface
->ndp_event
.uloop
);
187 close(iface
->ndp_event
.uloop
.fd
);
188 iface
->ndp_event
.uloop
.fd
= -1;
190 if (!enable
|| iface
->ndp
!= RELAYD_RELAY
)
191 if (write(procfd
, "0\n", 2) < 0) {}
196 if (enable
&& (iface
->ra
== RELAYD_SERVER
||
197 iface
->dhcpv6
== RELAYD_SERVER
|| iface
->ndp
== RELAYD_RELAY
))
200 if (enable
&& iface
->ndp
== RELAYD_RELAY
) {
201 if (write(procfd
, "1\n", 2) < 0) {}
203 int sock
= socket(AF_PACKET
, SOCK_DGRAM
| SOCK_CLOEXEC
, htons(ETH_P_IPV6
));
205 syslog(LOG_ERR
, "Unable to open packet socket: %s",
211 #ifdef PACKET_RECV_TYPE
212 int pktt
= 1 << PACKET_MULTICAST
;
213 setsockopt(sock
, SOL_PACKET
, PACKET_RECV_TYPE
, &pktt
, sizeof(pktt
));
216 if (setsockopt(sock
, SOL_SOCKET
, SO_ATTACH_FILTER
,
217 &bpf_prog
, sizeof(bpf_prog
))) {
218 syslog(LOG_ERR
, "Failed to set BPF: %s", strerror(errno
));
223 struct sockaddr_ll ll
= {
224 .sll_family
= AF_PACKET
,
225 .sll_ifindex
= iface
->ifindex
,
226 .sll_protocol
= htons(ETH_P_IPV6
),
232 bind(sock
, (struct sockaddr
*)&ll
, sizeof(ll
));
234 struct packet_mreq mreq
= {iface
->ifindex
, PACKET_MR_ALLMULTI
, ETH_ALEN
, {0}};
235 setsockopt(sock
, SOL_PACKET
, PACKET_ADD_MEMBERSHIP
, &mreq
, sizeof(mreq
));
237 iface
->ndp_event
.uloop
.fd
= sock
;
238 iface
->ndp_event
.handle_dgram
= handle_solicit
;
239 odhcpd_register(&iface
->ndp_event
);
241 // If we already were enabled dump is unnecessary, if not do dump
243 dump_neigh_table(false);
249 dump_neigh_table(true);
259 // Send an ICMP-ECHO. This is less for actually pinging but for the
260 // neighbor cache to be kept up-to-date.
261 static void ping6(struct in6_addr
*addr
,
262 const struct interface
*iface
)
264 struct sockaddr_in6 dest
= { .sin6_family
= AF_INET6
, .sin6_addr
= *addr
, .sin6_scope_id
= iface
->ifindex
, };
265 struct icmp6_hdr echo
= { .icmp6_type
= ICMP6_ECHO_REQUEST
};
266 struct iovec iov
= { .iov_base
= &echo
, .iov_len
= sizeof(echo
) };
268 odhcpd_setup_route(addr
, 128, iface
, NULL
, 128, true);
269 odhcpd_send(ping_socket
, &dest
, &iov
, 1, iface
);
270 odhcpd_setup_route(addr
, 128, iface
, NULL
, 128, false);
274 // Handle solicitations
275 static void handle_solicit(void *addr
, void *data
, size_t len
,
276 struct interface
*iface
, _unused
void *dest
)
278 struct ip6_hdr
*ip6
= data
;
279 struct nd_neighbor_solicit
*req
= (struct nd_neighbor_solicit
*)&ip6
[1];
280 struct sockaddr_ll
*ll
= addr
;
282 // Solicitation is for duplicate address detection
283 bool ns_is_dad
= IN6_IS_ADDR_UNSPECIFIED(&ip6
->ip6_src
);
285 // Don't process solicit messages on non relay interfaces
286 // Don't forward any non-DAD solicitation for external ifaces
287 // TODO: check if we should even forward DADs for them
288 if (iface
->ndp
!= RELAYD_RELAY
|| (iface
->external
&& !ns_is_dad
))
291 if (len
< sizeof(*ip6
) + sizeof(*req
))
292 return; // Invalid reqicitation
294 if (IN6_IS_ADDR_LINKLOCAL(&req
->nd_ns_target
) ||
295 IN6_IS_ADDR_LOOPBACK(&req
->nd_ns_target
) ||
296 IN6_IS_ADDR_MULTICAST(&req
->nd_ns_target
))
297 return; // Invalid target
299 char ipbuf
[INET6_ADDRSTRLEN
];
300 inet_ntop(AF_INET6
, &req
->nd_ns_target
, ipbuf
, sizeof(ipbuf
));
301 syslog(LOG_DEBUG
, "Got a NS for %s", ipbuf
);
304 odhcpd_get_mac(iface
, mac
);
305 if (!memcmp(ll
->sll_addr
, mac
, sizeof(mac
)))
306 return; // Looped back
309 list_for_each_entry(c
, &interfaces
, head
)
310 if (iface
!= c
&& c
->ndp
== RELAYD_RELAY
&&
311 (ns_is_dad
|| !c
->external
))
312 ping6(&req
->nd_ns_target
, c
);
315 // Use rtnetlink to modify kernel routes
316 static void setup_route(struct in6_addr
*addr
, struct interface
*iface
, bool add
)
318 char namebuf
[INET6_ADDRSTRLEN
];
319 inet_ntop(AF_INET6
, addr
, namebuf
, sizeof(namebuf
));
320 syslog(LOG_NOTICE
, "%s about %s on %s",
321 (add
) ? "Learned" : "Forgot", namebuf
, iface
->ifname
);
323 if (iface
->learn_routes
)
324 odhcpd_setup_route(addr
, 128, iface
, NULL
, 1024, add
);
328 static int prefixcmp(const void *va
, const void *vb
)
330 const struct odhcpd_ipaddr
*a
= va
, *b
= vb
;
331 uint32_t a_pref
= IN6_IS_ADDR_ULA(&a
->addr
) ? 1 : a
->preferred
;
332 uint32_t b_pref
= IN6_IS_ADDR_ULA(&b
->addr
) ? 1 : b
->preferred
;
333 return (a_pref
< b_pref
) ? 1 : (a_pref
> b_pref
) ? -1 : 0;
336 // Check address update
337 static void check_addr_updates(struct interface
*iface
)
339 struct odhcpd_ipaddr addr
[RELAYD_MAX_ADDRS
] = {{IN6ADDR_ANY_INIT
, 0, 0, 0, 0}};
340 time_t now
= odhcpd_time();
341 ssize_t len
= odhcpd_get_interface_addresses(iface
->ifindex
, addr
, ARRAY_SIZE(addr
));
346 qsort(addr
, len
, sizeof(*addr
), prefixcmp
);
348 for (int i
= 0; i
< len
; ++i
) {
349 addr
[i
].addr
.s6_addr32
[3] = 0;
351 if (addr
[i
].preferred
< UINT32_MAX
- now
)
352 addr
[i
].preferred
+= now
;
354 if (addr
[i
].valid
< UINT32_MAX
- now
)
355 addr
[i
].valid
+= now
;
358 bool change
= len
!= (ssize_t
)iface
->ia_addr_len
;
359 for (ssize_t i
= 0; !change
&& i
< len
; ++i
)
360 if (!IN6_ARE_ADDR_EQUAL(&addr
[i
].addr
, &iface
->ia_addr
[i
].addr
) ||
361 (addr
[i
].preferred
> 0) != (iface
->ia_addr
[i
].preferred
> 0) ||
362 addr
[i
].valid
< iface
->ia_addr
[i
].valid
||
363 addr
[i
].preferred
< iface
->ia_addr
[i
].preferred
)
367 dhcpv6_ia_preupdate(iface
);
369 memcpy(iface
->ia_addr
, addr
, len
* sizeof(*addr
));
370 iface
->ia_addr_len
= len
;
373 dhcpv6_ia_postupdate(iface
, now
);
376 syslog(LOG_INFO
, "Raising SIGUSR1 due to address change on %s", iface
->ifname
);
381 void setup_addr_for_relaying(struct in6_addr
*addr
, struct interface
*iface
, bool add
)
385 list_for_each_entry(c
, &interfaces
, head
) {
386 if (iface
== c
|| (c
->ndp
!= RELAYD_RELAY
&& !add
))
389 odhcpd_setup_proxy_neigh(addr
, c
, c
->ndp
== RELAYD_RELAY
? add
: false);
393 void setup_ping6(struct in6_addr
*addr
, struct interface
*iface
)
397 list_for_each_entry(c
, &interfaces
, head
) {
398 if (iface
== c
|| c
->ndp
!= RELAYD_RELAY
||
406 static struct in6_addr last_solicited
;
408 static void handle_rtnl_event(struct odhcpd_event
*e
)
410 struct event_socket
*ev_sock
= container_of(e
, struct event_socket
, ev
);
412 nl_recvmsgs_default(ev_sock
->sock
);
416 // Handler for neighbor cache entries from the kernel. This is our source
417 // to learn and unlearn hosts on interfaces.
418 static int cb_rtnl_valid(struct nl_msg
*msg
, _unused
void *arg
)
420 struct nlmsghdr
*hdr
= nlmsg_hdr(msg
);
421 struct in6_addr
*addr
= NULL
;
422 struct interface
*iface
= NULL
;
425 switch (hdr
->nlmsg_type
) {
428 struct rtmsg
*rtm
= nlmsg_data(hdr
);
430 if (!nlmsg_valid_hdr(hdr
, sizeof(*rtm
)) ||
431 rtm
->rtm_family
!= AF_INET6
)
434 if (rtm
->rtm_dst_len
== 0) {
435 syslog(LOG_INFO
, "Raising SIGUSR1 due to default route change");
444 struct ifaddrmsg
*ifa
= nlmsg_data(hdr
);
445 struct nlattr
*nla
[__IFA_MAX
];
447 if (!nlmsg_valid_hdr(hdr
, sizeof(*ifa
)) ||
448 ifa
->ifa_family
!= AF_INET6
)
451 iface
= odhcpd_get_interface_by_index(ifa
->ifa_index
);
455 nlmsg_parse(hdr
, sizeof(*ifa
), nla
, __IFA_MAX
- 1, NULL
);
456 if (!nla
[IFA_ADDRESS
])
459 addr
= nla_data(nla
[IFA_ADDRESS
]);
460 if (!addr
|| IN6_IS_ADDR_LINKLOCAL(addr
) ||
461 IN6_IS_ADDR_MULTICAST(addr
))
464 check_addr_updates(iface
);
466 if (iface
->ndp
!= RELAYD_RELAY
)
469 /* handle the relay logic below */
470 setup_addr_for_relaying(addr
, iface
, add
);
473 dump_neigh_table(false);
480 struct ndmsg
*ndm
= nlmsg_data(hdr
);
481 struct nlattr
*nla
[__NDA_MAX
];
483 if (!nlmsg_valid_hdr(hdr
, sizeof(*ndm
)) ||
484 ndm
->ndm_family
!= AF_INET6
)
487 iface
= odhcpd_get_interface_by_index(ndm
->ndm_ifindex
);
488 if (!iface
|| iface
->ndp
!= RELAYD_RELAY
)
489 return (iface
? NL_OK
: NL_SKIP
);
491 nlmsg_parse(hdr
, sizeof(*ndm
), nla
, __NDA_MAX
- 1, NULL
);
495 addr
= nla_data(nla
[NDA_DST
]);
496 if (!addr
|| IN6_IS_ADDR_LINKLOCAL(addr
) ||
497 IN6_IS_ADDR_MULTICAST(addr
))
500 if (ndm
->ndm_flags
& NTF_PROXY
) {
501 /* Dump and flush proxy entries */
502 if (hdr
->nlmsg_type
== RTM_NEWNEIGH
) {
503 odhcpd_setup_proxy_neigh(addr
, iface
, false);
504 setup_route(addr
, iface
, false);
505 dump_neigh_table(false);
511 if (add
&& !(ndm
->ndm_state
&
512 (NUD_REACHABLE
| NUD_STALE
| NUD_DELAY
| NUD_PROBE
|
513 NUD_PERMANENT
| NUD_NOARP
))) {
514 if (!IN6_ARE_ADDR_EQUAL(&last_solicited
, addr
)) {
515 last_solicited
= *addr
;
516 setup_ping6(addr
, iface
);
522 setup_addr_for_relaying(addr
, iface
, add
);
523 setup_route(addr
, iface
, add
);
526 dump_neigh_table(false);
537 static void catch_rtnl_err(struct odhcpd_event
*e
, int error
)
539 struct event_socket
*ev_sock
= container_of(e
, struct event_socket
, ev
);
541 if (error
!= ENOBUFS
)
544 /* Double netlink event buffer size */
545 ev_sock
->sock_bufsize
*= 2;
547 if (nl_socket_set_buffer_size(ev_sock
->sock
, ev_sock
->sock_bufsize
, 0))
554 odhcpd_deregister(e
);