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 * SPDX-License-Identifier: BSD-3-Clause
8 * Copyright (c) 1995, 1996, 1998, 1999
9 * The Internet Software Consortium. All rights reserved.
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of The Internet Software Consortium nor the names
21 * of its contributors may be used to endorse or promote products derived
22 * from this software without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED BY THE INTERNET SOFTWARE CONSORTIUM AND
25 * CONTRIBUTORS ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES,
26 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28 * DISCLAIMED. IN NO EVENT SHALL THE INTERNET SOFTWARE CONSORTIUM OR
29 * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
30 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
31 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF
32 * USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
33 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
34 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
35 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38 * This software has been written for the Internet Software Consortium
39 * by Ted Lemon <mellon@fugue.com> in cooperation with Vixie
40 * Enterprises. To learn more about the Internet Software Consortium,
41 * see ``http://www.vix.com/isc''. To learn more about Vixie
42 * Enterprises, see ``http://www.vix.com''.
45 #include <sys/cdefs.h>
46 __FBSDID("$FreeBSD$");
50 #include <sys/capsicum.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 #include <capsicum_helpers.h>
62 #define BPF_FORMAT "/dev/bpf%d"
65 * Called by get_interface_list for each interface that's discovered.
66 * Opens a packet filter for each interface and adds it to the select
70 if_register_bpf(struct interface_info *info, int flags)
75 /* Open a BPF device */
77 snprintf(filename, sizeof(filename), BPF_FORMAT, b);
78 sock = open(filename, flags);
83 error("Can't find free bpf: %m");
88 /* Set the BPF device to point at this interface. */
89 if (ioctl(sock, BIOCSETIF, info->ifp) < 0)
90 error("Can't attach interface %s to bpf device %s: %m",
91 info->name, filename);
93 /* Tag the packets with the proper VLAN PCP setting. */
94 if (info->client->config->vlan_pcp != 0) {
95 if (ioctl(sock, BIOCSETVLANPCP,
96 &info->client->config->vlan_pcp) < 0)
97 error( "Can't set the VLAN PCP tag on interface %s: %m",
105 * Packet write filter program:
106 * 'ip and udp and src port bootps and dst port (bootps or bootpc)'
108 static const struct bpf_insn dhcp_bpf_wfilter[] = {
109 BPF_STMT(BPF_LD + BPF_B + BPF_IND, 14),
110 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, (IPVERSION << 4) + 5, 0, 12),
112 /* Make sure this is an IP packet... */
113 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 12),
114 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, ETHERTYPE_IP, 0, 10),
116 /* Make sure it's a UDP packet... */
117 BPF_STMT(BPF_LD + BPF_B + BPF_ABS, 23),
118 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, IPPROTO_UDP, 0, 8),
120 /* Make sure this isn't a fragment... */
121 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 20),
122 BPF_JUMP(BPF_JMP + BPF_JSET + BPF_K, IP_MF|IP_OFFMASK, 6, 0),
124 /* Get the IP header length... */
125 BPF_STMT(BPF_LDX + BPF_B + BPF_MSH, 14),
127 /* Make sure it's from the right port... */
128 BPF_STMT(BPF_LD + BPF_H + BPF_IND, 14),
129 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, LOCAL_PORT, 0, 3),
131 /* Make sure it is to the right ports ... */
132 BPF_STMT(BPF_LD + BPF_H + BPF_IND, 16),
133 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, REMOTE_PORT, 0, 1),
135 /* If we passed all the tests, ask for the whole packet. */
136 BPF_STMT(BPF_RET+BPF_K, (u_int)-1),
138 /* Otherwise, drop it. */
139 BPF_STMT(BPF_RET+BPF_K, 0),
143 if_register_send(struct interface_info *info)
146 struct bpf_version v;
147 struct bpf_program p;
150 /* Open a BPF device and hang it on this interface... */
151 info->wfdesc = if_register_bpf(info, O_WRONLY);
153 /* Make sure the BPF version is in range... */
154 if (ioctl(info->wfdesc, BIOCVERSION, &v) < 0)
155 error("Can't get BPF version: %m");
157 if (v.bv_major != BPF_MAJOR_VERSION ||
158 v.bv_minor < BPF_MINOR_VERSION)
159 error("Kernel BPF version out of range - recompile dhcpd!");
161 /* Set up the bpf write filter program structure. */
162 p.bf_insns = __DECONST(struct bpf_insn *, dhcp_bpf_wfilter);
163 p.bf_len = nitems(dhcp_bpf_wfilter);
165 if (ioctl(info->wfdesc, BIOCSETWF, &p) < 0)
166 error("Can't install write filter program: %m");
168 if (ioctl(info->wfdesc, BIOCLOCK, NULL) < 0)
169 error("Cannot lock bpf");
171 cap_rights_init(&rights, CAP_WRITE);
172 if (caph_rights_limit(info->wfdesc, &rights) < 0)
173 error("Can't limit bpf descriptor: %m");
176 * Use raw socket for unicast send.
178 if ((sock = socket(AF_INET, SOCK_RAW, IPPROTO_UDP)) == -1)
179 error("socket(SOCK_RAW): %m");
180 if (setsockopt(sock, IPPROTO_IP, IP_HDRINCL, &on,
182 error("setsockopt(IP_HDRINCL): %m");
187 * Packet filter program...
189 static const struct bpf_insn dhcp_bpf_filter[] = {
190 /* Make sure this is an IP packet... */
191 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 12),
192 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, ETHERTYPE_IP, 0, 8),
194 /* Make sure it's a UDP packet... */
195 BPF_STMT(BPF_LD + BPF_B + BPF_ABS, 23),
196 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, IPPROTO_UDP, 0, 6),
198 /* Make sure this isn't a fragment... */
199 BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 20),
200 BPF_JUMP(BPF_JMP + BPF_JSET + BPF_K, IP_MF|IP_OFFMASK, 4, 0),
202 /* Get the IP header length... */
203 BPF_STMT(BPF_LDX + BPF_B + BPF_MSH, 14),
205 /* Make sure it's to the right port... */
206 BPF_STMT(BPF_LD + BPF_H + BPF_IND, 16),
207 BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, LOCAL_PORT, 0, 1),
209 /* If we passed all the tests, ask for the whole packet. */
210 BPF_STMT(BPF_RET+BPF_K, (u_int)-1),
212 /* Otherwise, drop it. */
213 BPF_STMT(BPF_RET+BPF_K, 0),
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_insns = __DECONST(struct bpf_insn *, dhcp_bpf_filter);
257 p.bf_len = nitems(dhcp_bpf_filter);
259 if (ioctl(info->rfdesc, BIOCSETF, &p) < 0)
260 error("Can't install packet filter program: %m");
262 if (ioctl(info->rfdesc, BIOCLOCK, NULL) < 0)
263 error("Cannot lock bpf");
265 cap_rights_init(&rights, CAP_IOCTL, CAP_EVENT, CAP_READ);
266 if (caph_rights_limit(info->rfdesc, &rights) < 0)
267 error("Can't limit bpf descriptor: %m");
268 if (caph_ioctls_limit(info->rfdesc, cmds, 2) < 0)
269 error("Can't limit ioctls for bpf descriptor: %m");
273 send_packet_unpriv(int privfd, struct dhcp_packet *raw, size_t len,
274 struct in_addr from, struct in_addr to)
280 hdr.code = IMSG_SEND_PACKET;
281 hdr.len = sizeof(hdr) +
282 sizeof(size_t) + len +
283 sizeof(from) + sizeof(to);
285 if ((buf = buf_open(hdr.len)) == NULL)
286 error("buf_open: %m");
289 errs += buf_add(buf, &hdr, sizeof(hdr));
290 errs += buf_add(buf, &len, sizeof(len));
291 errs += buf_add(buf, raw, len);
292 errs += buf_add(buf, &from, sizeof(from));
293 errs += buf_add(buf, &to, sizeof(to));
295 error("buf_add: %m");
297 if (buf_close(privfd, buf) == -1)
298 error("buf_close: %m");
302 send_packet_priv(struct interface_info *interface, struct imsg_hdr *hdr, int fd)
304 unsigned char buf[256];
307 struct dhcp_packet raw;
309 struct in_addr from, to;
310 int result, bufp = 0;
312 if (hdr->len < sizeof(*hdr) + sizeof(size_t))
313 error("corrupted message received");
314 buf_read(fd, &len, sizeof(len));
315 if (hdr->len != sizeof(*hdr) + sizeof(size_t) + len +
316 sizeof(from) + sizeof(to)) {
317 error("corrupted message received");
319 if (len > sizeof(raw))
320 error("corrupted message received");
321 buf_read(fd, &raw, len);
322 buf_read(fd, &from, sizeof(from));
323 buf_read(fd, &to, sizeof(to));
325 /* Assemble the headers... */
326 if (to.s_addr == INADDR_BROADCAST)
327 assemble_hw_header(interface, buf, &bufp);
328 assemble_udp_ip_header(buf, &bufp, from.s_addr, to.s_addr,
329 htons(REMOTE_PORT), (unsigned char *)&raw, len);
331 iov[0].iov_base = buf;
332 iov[0].iov_len = bufp;
333 iov[1].iov_base = &raw;
334 iov[1].iov_len = len;
337 if (to.s_addr == INADDR_BROADCAST)
338 result = writev(interface->wfdesc, iov, 2);
340 struct sockaddr_in sato;
343 sato.sin_port = htons(REMOTE_PORT);
344 sato.sin_family = AF_INET;
345 sato.sin_len = sizeof(sato);
347 memset(&msg, 0, sizeof(msg));
348 msg.msg_name = (struct sockaddr *)&sato;
349 msg.msg_namelen = sizeof(sato);
352 result = sendmsg(interface->ufdesc, &msg, 0);
356 warning("send_packet: %m");
360 receive_packet(struct interface_info *interface, unsigned char *buf,
361 size_t len, struct sockaddr_in *from, struct hardware *hfrom)
363 int length = 0, offset = 0;
367 * All this complexity is because BPF doesn't guarantee that
368 * only one packet will be returned at a time. We're getting
369 * what we deserve, though - this is a terrible abuse of the BPF
373 /* Process packets until we get one we can return or until we've
374 * done a read and gotten nothing we can return...
377 /* If the buffer is empty, fill it. */
378 if (interface->rbuf_offset >= interface->rbuf_len) {
379 length = read(interface->rfdesc, interface->rbuf,
380 interface->rbuf_max);
383 interface->rbuf_offset = 0;
384 interface->rbuf_len = length;
388 * If there isn't room for a whole bpf header, something
389 * went wrong, but we'll ignore it and hope it goes
392 if (interface->rbuf_len - interface->rbuf_offset <
394 interface->rbuf_offset = interface->rbuf_len;
398 /* Copy out a bpf header... */
399 memcpy(&hdr, &interface->rbuf[interface->rbuf_offset],
403 * If the bpf header plus data doesn't fit in what's
404 * left of the buffer, stick head in sand yet again...
406 if (interface->rbuf_offset + hdr.bh_hdrlen + hdr.bh_caplen >
407 interface->rbuf_len) {
408 interface->rbuf_offset = interface->rbuf_len;
412 /* Skip over the BPF header... */
413 interface->rbuf_offset += hdr.bh_hdrlen;
416 * If the captured data wasn't the whole packet, or if
417 * the packet won't fit in the input buffer, all we can
420 if (hdr.bh_caplen != hdr.bh_datalen) {
421 interface->rbuf_offset =
422 BPF_WORDALIGN(interface->rbuf_offset +
427 /* Decode the physical header... */
428 offset = decode_hw_header(interface->rbuf,
429 interface->rbuf_offset, hfrom);
432 * If a physical layer checksum failed (dunno of any
433 * physical layer that supports this, but WTH), skip
437 interface->rbuf_offset =
438 BPF_WORDALIGN(interface->rbuf_offset +
442 interface->rbuf_offset += offset;
443 hdr.bh_caplen -= offset;
445 /* Decode the IP and UDP headers... */
446 offset = decode_udp_ip_header(interface->rbuf,
447 interface->rbuf_offset, from, NULL, hdr.bh_caplen);
449 /* If the IP or UDP checksum was bad, skip the packet... */
451 interface->rbuf_offset =
452 BPF_WORDALIGN(interface->rbuf_offset +
456 interface->rbuf_offset += offset;
457 hdr.bh_caplen -= offset;
460 * If there's not enough room to stash the packet data,
461 * we have to skip it (this shouldn't happen in real
464 if (hdr.bh_caplen > len) {
465 interface->rbuf_offset =
466 BPF_WORDALIGN(interface->rbuf_offset +
471 /* Copy out the data in the packet... */
472 memcpy(buf, interface->rbuf + interface->rbuf_offset,
474 interface->rbuf_offset =
475 BPF_WORDALIGN(interface->rbuf_offset +
477 return (hdr.bh_caplen);