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>
40 struct odhcpd_event ev
;
45 static void handle_solicit(void *addr
, void *data
, size_t len
,
46 struct interface
*iface
, void *dest
);
47 static void handle_rtnl_event(struct odhcpd_event
*ev
);
48 static int cb_rtnl_valid(struct nl_msg
*msg
, void *arg
);
49 static void catch_rtnl_err(struct odhcpd_event
*e
, int error
);
51 static int addr6_dump_rqs
= 0;
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_addr6_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 void ndp_handle_addr6_dump(void)
180 inline void ndp_rqs_addr6_dump(void)
185 int setup_ndp_interface(struct interface
*iface
, bool enable
)
188 bool dump_neigh
= false;
191 snprintf(procbuf
, sizeof(procbuf
), "/proc/sys/net/ipv6/conf/%s/proxy_ndp", iface
->ifname
);
192 procfd
= open(procbuf
, O_WRONLY
);
199 if (iface
->ndp_event
.uloop
.fd
> 0) {
200 uloop_fd_delete(&iface
->ndp_event
.uloop
);
201 close(iface
->ndp_event
.uloop
.fd
);
202 iface
->ndp_event
.uloop
.fd
= -1;
204 if (!enable
|| iface
->ndp
!= RELAYD_RELAY
)
205 if (write(procfd
, "0\n", 2) < 0) {}
210 if (enable
&& iface
->ndp
== RELAYD_RELAY
) {
211 if (write(procfd
, "1\n", 2) < 0) {}
213 int sock
= socket(AF_PACKET
, SOCK_DGRAM
| SOCK_CLOEXEC
, htons(ETH_P_IPV6
));
215 syslog(LOG_ERR
, "Unable to open packet socket: %s",
221 #ifdef PACKET_RECV_TYPE
222 int pktt
= 1 << PACKET_MULTICAST
;
223 setsockopt(sock
, SOL_PACKET
, PACKET_RECV_TYPE
, &pktt
, sizeof(pktt
));
226 if (setsockopt(sock
, SOL_SOCKET
, SO_ATTACH_FILTER
,
227 &bpf_prog
, sizeof(bpf_prog
))) {
228 syslog(LOG_ERR
, "Failed to set BPF: %s", strerror(errno
));
233 struct sockaddr_ll ll
= {
234 .sll_family
= AF_PACKET
,
235 .sll_ifindex
= iface
->ifindex
,
236 .sll_protocol
= htons(ETH_P_IPV6
),
242 bind(sock
, (struct sockaddr
*)&ll
, sizeof(ll
));
244 struct packet_mreq mreq
= {iface
->ifindex
, PACKET_MR_ALLMULTI
, ETH_ALEN
, {0}};
245 setsockopt(sock
, SOL_PACKET
, PACKET_ADD_MEMBERSHIP
, &mreq
, sizeof(mreq
));
247 iface
->ndp_event
.uloop
.fd
= sock
;
248 iface
->ndp_event
.handle_dgram
= handle_solicit
;
249 odhcpd_register(&iface
->ndp_event
);
251 // If we already were enabled dump is unnecessary, if not do dump
253 dump_neigh_table(false);
257 ndp_rqs_addr6_dump();
261 dump_neigh_table(true);
271 // Send an ICMP-ECHO. This is less for actually pinging but for the
272 // neighbor cache to be kept up-to-date.
273 static void ping6(struct in6_addr
*addr
,
274 const struct interface
*iface
)
276 struct sockaddr_in6 dest
= { .sin6_family
= AF_INET6
, .sin6_addr
= *addr
, .sin6_scope_id
= iface
->ifindex
, };
277 struct icmp6_hdr echo
= { .icmp6_type
= ICMP6_ECHO_REQUEST
};
278 struct iovec iov
= { .iov_base
= &echo
, .iov_len
= sizeof(echo
) };
279 char ipbuf
[INET6_ADDRSTRLEN
];
281 inet_ntop(AF_INET6
, addr
, ipbuf
, sizeof(ipbuf
));
282 syslog(LOG_NOTICE
, "Pinging for %s%%%s", ipbuf
, iface
->ifname
);
284 odhcpd_setup_route(addr
, 128, iface
, NULL
, 128, true);
285 odhcpd_send(ping_socket
, &dest
, &iov
, 1, iface
);
286 odhcpd_setup_route(addr
, 128, iface
, NULL
, 128, false);
289 // Handle solicitations
290 static void handle_solicit(void *addr
, void *data
, size_t len
,
291 struct interface
*iface
, _unused
void *dest
)
293 struct ip6_hdr
*ip6
= data
;
294 struct nd_neighbor_solicit
*req
= (struct nd_neighbor_solicit
*)&ip6
[1];
295 struct sockaddr_ll
*ll
= addr
;
296 char ipbuf
[INET6_ADDRSTRLEN
];
299 // Solicitation is for duplicate address detection
300 bool ns_is_dad
= IN6_IS_ADDR_UNSPECIFIED(&ip6
->ip6_src
);
302 // Don't process solicit messages on non relay interfaces
303 // Don't forward any non-DAD solicitation for external ifaces
304 // TODO: check if we should even forward DADs for them
305 if (iface
->ndp
!= RELAYD_RELAY
|| (iface
->external
&& !ns_is_dad
))
308 if (len
< sizeof(*ip6
) + sizeof(*req
))
309 return; // Invalid reqicitation
311 if (IN6_IS_ADDR_LINKLOCAL(&req
->nd_ns_target
) ||
312 IN6_IS_ADDR_LOOPBACK(&req
->nd_ns_target
) ||
313 IN6_IS_ADDR_MULTICAST(&req
->nd_ns_target
))
314 return; // Invalid target
316 inet_ntop(AF_INET6
, &req
->nd_ns_target
, ipbuf
, sizeof(ipbuf
));
317 syslog(LOG_DEBUG
, "Got a NS for %s%%%s", ipbuf
, iface
->ifname
);
319 odhcpd_get_mac(iface
, mac
);
320 if (!memcmp(ll
->sll_addr
, mac
, sizeof(mac
)))
321 return; // Looped back
324 list_for_each_entry(c
, &interfaces
, head
)
325 if (iface
!= c
&& c
->ndp
== RELAYD_RELAY
&&
326 (ns_is_dad
|| !c
->external
))
327 ping6(&req
->nd_ns_target
, c
);
330 // Use rtnetlink to modify kernel routes
331 static void setup_route(struct in6_addr
*addr
, struct interface
*iface
, bool add
)
333 char ipbuf
[INET6_ADDRSTRLEN
];
335 inet_ntop(AF_INET6
, addr
, ipbuf
, sizeof(ipbuf
));
336 syslog(LOG_NOTICE
, "%s about %s%%%s",
337 (add
) ? "Learned" : "Forgot", ipbuf
, iface
->ifname
);
339 if (iface
->learn_routes
)
340 odhcpd_setup_route(addr
, 128, iface
, NULL
, 1024, add
);
344 static int prefixcmp(const void *va
, const void *vb
)
346 const struct odhcpd_ipaddr
*a
= va
, *b
= vb
;
347 uint32_t a_pref
= IN6_IS_ADDR_ULA(&a
->addr
) ? 1 : a
->preferred
;
348 uint32_t b_pref
= IN6_IS_ADDR_ULA(&b
->addr
) ? 1 : b
->preferred
;
349 return (a_pref
< b_pref
) ? 1 : (a_pref
> b_pref
) ? -1 : 0;
352 // Check address update
353 static void check_addr_updates(struct interface
*iface
)
355 struct odhcpd_ipaddr
*addr
= NULL
;
356 time_t now
= odhcpd_time();
357 ssize_t len
= odhcpd_get_interface_addresses(iface
->ifindex
, &addr
);
362 qsort(addr
, len
, sizeof(*addr
), prefixcmp
);
364 for (int i
= 0; i
< len
; ++i
) {
365 if (addr
[i
].preferred
< UINT32_MAX
- now
)
366 addr
[i
].preferred
+= now
;
368 if (addr
[i
].valid
< UINT32_MAX
- now
)
369 addr
[i
].valid
+= now
;
372 bool change
= len
!= (ssize_t
)iface
->ia_addr_len
;
373 for (ssize_t i
= 0; !change
&& i
< len
; ++i
)
374 if (!IN6_ARE_ADDR_EQUAL(&addr
[i
].addr
, &iface
->ia_addr
[i
].addr
) ||
375 (addr
[i
].preferred
> 0) != (iface
->ia_addr
[i
].preferred
> 0) ||
376 addr
[i
].valid
< iface
->ia_addr
[i
].valid
||
377 addr
[i
].preferred
< iface
->ia_addr
[i
].preferred
)
381 dhcpv6_ia_preupdate(iface
);
383 free(iface
->ia_addr
);
384 iface
->ia_addr
= addr
;
385 iface
->ia_addr_len
= len
;
388 dhcpv6_ia_postupdate(iface
, now
);
391 syslog(LOG_INFO
, "Raising SIGUSR1 due to address change on %s", iface
->ifname
);
396 static void setup_addr_for_relaying(struct in6_addr
*addr
, struct interface
*iface
, bool add
)
399 char ipbuf
[INET6_ADDRSTRLEN
];
401 inet_ntop(AF_INET6
, addr
, ipbuf
, sizeof(ipbuf
));
403 list_for_each_entry(c
, &interfaces
, head
) {
404 if (iface
== c
|| (c
->ndp
!= RELAYD_RELAY
&& !add
))
407 bool neigh_add
= (c
->ndp
== RELAYD_RELAY
? add
: false);
409 if (odhcpd_setup_proxy_neigh(addr
, c
, neigh_add
))
410 syslog(LOG_DEBUG
, "Failed to %s proxy neighbour entry %s%%%s",
411 neigh_add
? "add" : "delete", ipbuf
, c
->ifname
);
413 syslog(LOG_DEBUG
, "%s proxy neighbour entry %s%%%s",
414 neigh_add
? "Added" : "Deleted", ipbuf
, c
->ifname
);
418 static void handle_rtnl_event(struct odhcpd_event
*e
)
420 struct event_socket
*ev_sock
= container_of(e
, struct event_socket
, ev
);
422 nl_recvmsgs_default(ev_sock
->sock
);
426 // Handler for neighbor cache entries from the kernel. This is our source
427 // to learn and unlearn hosts on interfaces.
428 static int cb_rtnl_valid(struct nl_msg
*msg
, _unused
void *arg
)
430 struct nlmsghdr
*hdr
= nlmsg_hdr(msg
);
431 struct in6_addr
*addr
= NULL
;
432 struct interface
*iface
= NULL
;
434 char ipbuf
[INET6_ADDRSTRLEN
];
436 switch (hdr
->nlmsg_type
) {
439 struct rtmsg
*rtm
= nlmsg_data(hdr
);
441 if (!nlmsg_valid_hdr(hdr
, sizeof(*rtm
)) ||
442 rtm
->rtm_family
!= AF_INET6
)
445 if (rtm
->rtm_dst_len
== 0) {
446 syslog(LOG_INFO
, "Raising SIGUSR1 due to default route change");
456 struct ifaddrmsg
*ifa
= nlmsg_data(hdr
);
457 struct nlattr
*nla
[__IFA_MAX
];
459 if (!nlmsg_valid_hdr(hdr
, sizeof(*ifa
)) ||
460 ifa
->ifa_family
!= AF_INET6
)
463 iface
= odhcpd_get_interface_by_index(ifa
->ifa_index
);
467 nlmsg_parse(hdr
, sizeof(*ifa
), nla
, __IFA_MAX
- 1, NULL
);
468 if (!nla
[IFA_ADDRESS
])
471 addr
= nla_data(nla
[IFA_ADDRESS
]);
472 if (!addr
|| IN6_IS_ADDR_LINKLOCAL(addr
) ||
473 IN6_IS_ADDR_MULTICAST(addr
))
476 inet_ntop(AF_INET6
, addr
, ipbuf
, sizeof(ipbuf
));
477 syslog(LOG_DEBUG
, "Netlink %s %s%%%s", true ? "newaddr" : "deladdr",
478 ipbuf
, iface
->ifname
);
480 check_addr_updates(iface
);
482 if (iface
->ndp
!= RELAYD_RELAY
)
485 /* handle the relay logic below */
486 setup_addr_for_relaying(addr
, iface
, add
);
489 dump_neigh_table(false);
497 struct ndmsg
*ndm
= nlmsg_data(hdr
);
498 struct nlattr
*nla
[__NDA_MAX
];
500 if (!nlmsg_valid_hdr(hdr
, sizeof(*ndm
)) ||
501 ndm
->ndm_family
!= AF_INET6
)
504 iface
= odhcpd_get_interface_by_index(ndm
->ndm_ifindex
);
505 if (!iface
|| iface
->ndp
!= RELAYD_RELAY
)
506 return (iface
? NL_OK
: NL_SKIP
);
508 nlmsg_parse(hdr
, sizeof(*ndm
), nla
, __NDA_MAX
- 1, NULL
);
512 addr
= nla_data(nla
[NDA_DST
]);
513 if (!addr
|| IN6_IS_ADDR_LINKLOCAL(addr
) ||
514 IN6_IS_ADDR_MULTICAST(addr
))
517 inet_ntop(AF_INET6
, addr
, ipbuf
, sizeof(ipbuf
));
518 syslog(LOG_DEBUG
, "Netlink %s %s%%%s", true ? "newneigh" : "delneigh",
519 ipbuf
, iface
->ifname
);
521 if (ndm
->ndm_flags
& NTF_PROXY
) {
522 /* Dump and flush proxy entries */
523 if (hdr
->nlmsg_type
== RTM_NEWNEIGH
) {
524 odhcpd_setup_proxy_neigh(addr
, iface
, false);
525 setup_route(addr
, iface
, false);
526 dump_neigh_table(false);
532 if (add
&& !(ndm
->ndm_state
&
533 (NUD_REACHABLE
| NUD_STALE
| NUD_DELAY
| NUD_PROBE
|
534 NUD_PERMANENT
| NUD_NOARP
)))
537 setup_addr_for_relaying(addr
, iface
, add
);
538 setup_route(addr
, iface
, add
);
541 dump_neigh_table(false);
552 static void catch_rtnl_err(struct odhcpd_event
*e
, int error
)
554 struct event_socket
*ev_sock
= container_of(e
, struct event_socket
, ev
);
556 if (error
!= ENOBUFS
)
559 /* Double netlink event buffer size */
560 ev_sock
->sock_bufsize
*= 2;
562 if (nl_socket_set_buffer_size(ev_sock
->sock
, ev_sock
->sock_bufsize
, 0))
569 odhcpd_deregister(e
);