1 /* $OpenBSD: bpf.c,v 1.13 2004/05/05 14:28:58 deraadt Exp $ */
3 /* BPF socket interface code, originally contributed by Archie Cobbs. */
6 * Copyright (c) 1995, 1996, 1998, 1999
7 * The Internet Software Consortium. All rights reserved.
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
18 * 3. Neither the name of The Internet Software Consortium nor the names
19 * of its contributors may be used to endorse or promote products derived
20 * from this software without specific prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE INTERNET SOFTWARE CONSORTIUM AND
23 * CONTRIBUTORS ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES,
24 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
25 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
26 * DISCLAIMED. IN NO EVENT SHALL THE INTERNET SOFTWARE CONSORTIUM OR
27 * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
28 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
29 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF
30 * USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
31 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
32 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
33 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * This software has been written for the Internet Software Consortium
37 * by Ted Lemon <mellon@fugue.com> in cooperation with Vixie
38 * Enterprises. To learn more about the Internet Software Consortium,
39 * see ``http://www.vix.com/isc''. To learn more about Vixie
40 * Enterprises, see ``http://www.vix.com''.
43 #include <sys/cdefs.h>
44 __FBSDID("$FreeBSD$");
46 #include <sys/capability.h>
50 #include <sys/capability.h>
51 #include <sys/ioctl.h>
55 #include <netinet/in_systm.h>
56 #include <netinet/ip.h>
57 #include <netinet/udp.h>
58 #include <netinet/if_ether.h>
60 #define BPF_FORMAT "/dev/bpf%d"
63 * Called by get_interface_list for each interface that's discovered.
64 * Opens a packet filter for each interface and adds it to the select
68 if_register_bpf(struct interface_info *info, int flags)
73 /* Open a BPF device */
75 snprintf(filename, sizeof(filename), BPF_FORMAT, b);
76 sock = open(filename, flags);
81 error("Can't find free bpf: %m");
86 /* Set the BPF device to point at this interface. */
87 if (ioctl(sock, BIOCSETIF, info->ifp) < 0)
88 error("Can't attach interface %s to bpf device %s: %m",
89 info->name, filename);
95 * Packet write filter program:
96 * 'ip and udp and src port bootps and dst port (bootps or bootpc)'
98 struct bpf_insn dhcp_bpf_wfilter[] = {
99 BPF_STMT(BPF_LD + BPF_B + BPF_IND, 14),
100 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, (IPVERSION << 4) + 5, 0, 12),
102 /* Make sure this is an IP packet... */
103 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 12),
104 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, ETHERTYPE_IP, 0, 10),
106 /* Make sure it's a UDP packet... */
107 BPF_STMT(BPF_LD + BPF_B + BPF_ABS, 23),
108 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, IPPROTO_UDP, 0, 8),
110 /* Make sure this isn't a fragment... */
111 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 20),
112 BPF_JUMP(BPF_JMP + BPF_JSET + BPF_K, 0x1fff, 6, 0), /* patched */
114 /* Get the IP header length... */
115 BPF_STMT(BPF_LDX + BPF_B + BPF_MSH, 14),
117 /* Make sure it's from the right port... */
118 BPF_STMT(BPF_LD + BPF_H + BPF_IND, 14),
119 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, 68, 0, 3),
121 /* Make sure it is to the right ports ... */
122 BPF_STMT(BPF_LD + BPF_H + BPF_IND, 16),
123 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, 67, 0, 1),
125 /* If we passed all the tests, ask for the whole packet. */
126 BPF_STMT(BPF_RET+BPF_K, (u_int)-1),
128 /* Otherwise, drop it. */
129 BPF_STMT(BPF_RET+BPF_K, 0),
132 int dhcp_bpf_wfilter_len = sizeof(dhcp_bpf_wfilter) / sizeof(struct bpf_insn);
135 if_register_send(struct interface_info *info)
138 struct bpf_version v;
139 struct bpf_program p;
142 /* Open a BPF device and hang it on this interface... */
143 info->wfdesc = if_register_bpf(info, O_WRONLY);
145 /* Make sure the BPF version is in range... */
146 if (ioctl(info->wfdesc, BIOCVERSION, &v) < 0)
147 error("Can't get BPF version: %m");
149 if (v.bv_major != BPF_MAJOR_VERSION ||
150 v.bv_minor < BPF_MINOR_VERSION)
151 error("Kernel BPF version out of range - recompile dhcpd!");
153 /* Set up the bpf write filter program structure. */
154 p.bf_len = dhcp_bpf_wfilter_len;
155 p.bf_insns = dhcp_bpf_wfilter;
157 if (dhcp_bpf_wfilter[7].k == 0x1fff)
158 dhcp_bpf_wfilter[7].k = htons(IP_MF|IP_OFFMASK);
160 if (ioctl(info->wfdesc, BIOCSETWF, &p) < 0)
161 error("Can't install write filter program: %m");
163 if (ioctl(info->wfdesc, BIOCLOCK, NULL) < 0)
164 error("Cannot lock bpf");
166 cap_rights_init(&rights, CAP_WRITE);
167 if (cap_rights_limit(info->wfdesc, &rights) < 0 && errno != ENOSYS)
168 error("Can't limit bpf descriptor: %m");
171 * Use raw socket for unicast send.
173 if ((sock = socket(AF_INET, SOCK_RAW, IPPROTO_UDP)) == -1)
174 error("socket(SOCK_RAW): %m");
175 if (setsockopt(sock, IPPROTO_IP, IP_HDRINCL, &on,
177 error("setsockopt(IP_HDRINCL): %m");
182 * Packet filter program...
184 * XXX: Changes to the filter program may require changes to the
185 * constant offsets used in if_register_send to patch the BPF program!
187 struct bpf_insn dhcp_bpf_filter[] = {
188 /* Make sure this is an IP packet... */
189 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 12),
190 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, ETHERTYPE_IP, 0, 8),
192 /* Make sure it's a UDP packet... */
193 BPF_STMT(BPF_LD + BPF_B + BPF_ABS, 23),
194 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, IPPROTO_UDP, 0, 6),
196 /* Make sure this isn't a fragment... */
197 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 20),
198 BPF_JUMP(BPF_JMP + BPF_JSET + BPF_K, 0x1fff, 4, 0),
200 /* Get the IP header length... */
201 BPF_STMT(BPF_LDX + BPF_B + BPF_MSH, 14),
203 /* Make sure it's to the right port... */
204 BPF_STMT(BPF_LD + BPF_H + BPF_IND, 16),
205 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, 67, 0, 1), /* patch */
207 /* If we passed all the tests, ask for the whole packet. */
208 BPF_STMT(BPF_RET+BPF_K, (u_int)-1),
210 /* Otherwise, drop it. */
211 BPF_STMT(BPF_RET+BPF_K, 0),
214 int dhcp_bpf_filter_len = sizeof(dhcp_bpf_filter) / sizeof(struct bpf_insn);
217 if_register_receive(struct interface_info *info)
219 static const unsigned long cmds[2] = { SIOCGIFFLAGS, SIOCGIFMEDIA };
221 struct bpf_version v;
222 struct bpf_program p;
225 /* Open a BPF device and hang it on this interface... */
226 info->rfdesc = if_register_bpf(info, O_RDONLY);
228 /* Make sure the BPF version is in range... */
229 if (ioctl(info->rfdesc, BIOCVERSION, &v) < 0)
230 error("Can't get BPF version: %m");
232 if (v.bv_major != BPF_MAJOR_VERSION ||
233 v.bv_minor < BPF_MINOR_VERSION)
234 error("Kernel BPF version out of range - recompile dhcpd!");
237 * Set immediate mode so that reads return as soon as a packet
238 * comes in, rather than waiting for the input buffer to fill
241 if (ioctl(info->rfdesc, BIOCIMMEDIATE, &flag) < 0)
242 error("Can't set immediate mode on bpf device: %m");
244 /* Get the required BPF buffer length from the kernel. */
245 if (ioctl(info->rfdesc, BIOCGBLEN, &sz) < 0)
246 error("Can't get bpf buffer length: %m");
248 info->rbuf = malloc(info->rbuf_max);
250 error("Can't allocate %lu bytes for bpf input buffer.",
251 (unsigned long)info->rbuf_max);
252 info->rbuf_offset = 0;
255 /* Set up the bpf filter program structure. */
256 p.bf_len = dhcp_bpf_filter_len;
257 p.bf_insns = dhcp_bpf_filter;
259 /* Patch the server port into the BPF program...
261 * XXX: changes to filter program may require changes to the
262 * insn number(s) used below!
264 dhcp_bpf_filter[8].k = LOCAL_PORT;
266 if (ioctl(info->rfdesc, BIOCSETF, &p) < 0)
267 error("Can't install packet filter program: %m");
269 if (ioctl(info->rfdesc, BIOCLOCK, NULL) < 0)
270 error("Cannot lock bpf");
272 cap_rights_init(&rights, CAP_IOCTL, CAP_EVENT, CAP_READ);
273 if (cap_rights_limit(info->rfdesc, &rights) < 0 && errno != ENOSYS)
274 error("Can't limit bpf descriptor: %m");
275 if (cap_ioctls_limit(info->rfdesc, cmds, 2) < 0 && errno != ENOSYS)
276 error("Can't limit ioctls for bpf descriptor: %m");
280 send_packet_unpriv(int privfd, struct dhcp_packet *raw, size_t len,
281 struct in_addr from, struct in_addr to)
287 hdr.code = IMSG_SEND_PACKET;
288 hdr.len = sizeof(hdr) +
289 sizeof(size_t) + len +
290 sizeof(from) + sizeof(to);
292 if ((buf = buf_open(hdr.len)) == NULL)
293 error("buf_open: %m");
296 errs += buf_add(buf, &hdr, sizeof(hdr));
297 errs += buf_add(buf, &len, sizeof(len));
298 errs += buf_add(buf, raw, len);
299 errs += buf_add(buf, &from, sizeof(from));
300 errs += buf_add(buf, &to, sizeof(to));
302 error("buf_add: %m");
304 if (buf_close(privfd, buf) == -1)
305 error("buf_close: %m");
309 send_packet_priv(struct interface_info *interface, struct imsg_hdr *hdr, int fd)
311 unsigned char buf[256];
314 struct dhcp_packet raw;
316 struct in_addr from, to;
317 int result, bufp = 0;
319 if (hdr->len < sizeof(*hdr) + sizeof(size_t))
320 error("corrupted message received");
321 buf_read(fd, &len, sizeof(len));
322 if (hdr->len != sizeof(*hdr) + sizeof(size_t) + len +
323 sizeof(from) + sizeof(to)) {
324 error("corrupted message received");
326 if (len > sizeof(raw))
327 error("corrupted message received");
328 buf_read(fd, &raw, len);
329 buf_read(fd, &from, sizeof(from));
330 buf_read(fd, &to, sizeof(to));
332 /* Assemble the headers... */
333 if (to.s_addr == INADDR_BROADCAST)
334 assemble_hw_header(interface, buf, &bufp);
335 assemble_udp_ip_header(buf, &bufp, from.s_addr, to.s_addr,
336 htons(REMOTE_PORT), (unsigned char *)&raw, len);
338 iov[0].iov_base = buf;
339 iov[0].iov_len = bufp;
340 iov[1].iov_base = &raw;
341 iov[1].iov_len = len;
344 if (to.s_addr == INADDR_BROADCAST)
345 result = writev(interface->wfdesc, iov, 2);
347 struct sockaddr_in sato;
350 sato.sin_port = htons(REMOTE_PORT);
351 sato.sin_family = AF_INET;
352 sato.sin_len = sizeof(sato);
354 memset(&msg, 0, sizeof(msg));
355 msg.msg_name = (struct sockaddr *)&sato;
356 msg.msg_namelen = sizeof(sato);
359 result = sendmsg(interface->ufdesc, &msg, 0);
363 warning("send_packet: %m");
367 receive_packet(struct interface_info *interface, unsigned char *buf,
368 size_t len, struct sockaddr_in *from, struct hardware *hfrom)
370 int length = 0, offset = 0;
374 * All this complexity is because BPF doesn't guarantee that
375 * only one packet will be returned at a time. We're getting
376 * what we deserve, though - this is a terrible abuse of the BPF
380 /* Process packets until we get one we can return or until we've
381 * done a read and gotten nothing we can return...
384 /* If the buffer is empty, fill it. */
385 if (interface->rbuf_offset >= interface->rbuf_len) {
386 length = read(interface->rfdesc, interface->rbuf,
387 interface->rbuf_max);
390 interface->rbuf_offset = 0;
391 interface->rbuf_len = length;
395 * If there isn't room for a whole bpf header, something
396 * went wrong, but we'll ignore it and hope it goes
399 if (interface->rbuf_len - interface->rbuf_offset <
401 interface->rbuf_offset = interface->rbuf_len;
405 /* Copy out a bpf header... */
406 memcpy(&hdr, &interface->rbuf[interface->rbuf_offset],
410 * If the bpf header plus data doesn't fit in what's
411 * left of the buffer, stick head in sand yet again...
413 if (interface->rbuf_offset + hdr.bh_hdrlen + hdr.bh_caplen >
414 interface->rbuf_len) {
415 interface->rbuf_offset = interface->rbuf_len;
419 /* Skip over the BPF header... */
420 interface->rbuf_offset += hdr.bh_hdrlen;
423 * If the captured data wasn't the whole packet, or if
424 * the packet won't fit in the input buffer, all we can
427 if (hdr.bh_caplen != hdr.bh_datalen) {
428 interface->rbuf_offset =
429 BPF_WORDALIGN(interface->rbuf_offset +
434 /* Decode the physical header... */
435 offset = decode_hw_header(interface->rbuf,
436 interface->rbuf_offset, hfrom);
439 * If a physical layer checksum failed (dunno of any
440 * physical layer that supports this, but WTH), skip
444 interface->rbuf_offset =
445 BPF_WORDALIGN(interface->rbuf_offset +
449 interface->rbuf_offset += offset;
450 hdr.bh_caplen -= offset;
452 /* Decode the IP and UDP headers... */
453 offset = decode_udp_ip_header(interface->rbuf,
454 interface->rbuf_offset, from, NULL, hdr.bh_caplen);
456 /* If the IP or UDP checksum was bad, skip the packet... */
458 interface->rbuf_offset =
459 BPF_WORDALIGN(interface->rbuf_offset +
463 interface->rbuf_offset += offset;
464 hdr.bh_caplen -= offset;
467 * If there's not enough room to stash the packet data,
468 * we have to skip it (this shouldn't happen in real
471 if (hdr.bh_caplen > len) {
472 interface->rbuf_offset =
473 BPF_WORDALIGN(interface->rbuf_offset +
478 /* Copy out the data in the packet... */
479 memcpy(buf, interface->rbuf + interface->rbuf_offset,
481 interface->rbuf_offset =
482 BPF_WORDALIGN(interface->rbuf_offset +
484 return (hdr.bh_caplen);