1/*	$OpenBSD: bpf.c,v 1.13 2004/05/05 14:28:58 deraadt Exp $	*/
2
3/* BPF socket interface code, originally contributed by Archie Cobbs. */
4
5/*
6 * Copyright (c) 1995, 1996, 1998, 1999
7 * The Internet Software Consortium.    All rights reserved.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 *
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.
21 *
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
34 * SUCH DAMAGE.
35 *
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''.
41 */
42
43#include <sys/cdefs.h>
44__FBSDID("$FreeBSD$");
45
46#include "dhcpd.h"
47#include <sys/ioctl.h>
48#include <sys/uio.h>
49
50#include <net/bpf.h>
51#include <netinet/in_systm.h>
52#include <netinet/ip.h>
53#include <netinet/udp.h>
54#include <netinet/if_ether.h>
55
56#define BPF_FORMAT "/dev/bpf%d"
57
58/*
59 * Called by get_interface_list for each interface that's discovered.
60 * Opens a packet filter for each interface and adds it to the select
61 * mask.
62 */
63int
64if_register_bpf(struct interface_info *info)
65{
66	char filename[50];
67	int sock, b;
68
69	/* Open a BPF device */
70	for (b = 0; 1; b++) {
71		snprintf(filename, sizeof(filename), BPF_FORMAT, b);
72		sock = open(filename, O_RDWR, 0);
73		if (sock < 0) {
74			if (errno == EBUSY)
75				continue;
76			else
77				error("Can't find free bpf: %m");
78		} else
79			break;
80	}
81
82	/* Set the BPF device to point at this interface. */
83	if (ioctl(sock, BIOCSETIF, info->ifp) < 0)
84		error("Can't attach interface %s to bpf device %s: %m",
85		    info->name, filename);
86
87	return (sock);
88}
89
90void
91if_register_send(struct interface_info *info)
92{
93	int sock, on = 1;
94
95	/*
96	 * If we're using the bpf API for sending and receiving, we
97	 * don't need to register this interface twice.
98	 */
99	info->wfdesc = info->rfdesc;
100
101	/*
102	 * Use raw socket for unicast send.
103	 */
104	if ((sock = socket(AF_INET, SOCK_RAW, IPPROTO_UDP)) == -1)
105		error("socket(SOCK_RAW): %m");
106	if (setsockopt(sock, IPPROTO_IP, IP_HDRINCL, &on,
107	    sizeof(on)) == -1)
108		error("setsockopt(IP_HDRINCL): %m");
109	info->ufdesc = sock;
110}
111
112/*
113 * Packet filter program...
114 *
115 * XXX: Changes to the filter program may require changes to the
116 * constant offsets used in if_register_send to patch the BPF program!
117 */
118struct bpf_insn dhcp_bpf_filter[] = {
119	/* Make sure this is an IP packet... */
120	BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 12),
121	BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, ETHERTYPE_IP, 0, 8),
122
123	/* Make sure it's a UDP packet... */
124	BPF_STMT(BPF_LD + BPF_B + BPF_ABS, 23),
125	BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, IPPROTO_UDP, 0, 6),
126
127	/* Make sure this isn't a fragment... */
128	BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 20),
129	BPF_JUMP(BPF_JMP + BPF_JSET + BPF_K, 0x1fff, 4, 0),
130
131	/* Get the IP header length... */
132	BPF_STMT(BPF_LDX + BPF_B + BPF_MSH, 14),
133
134	/* Make sure it's to the right port... */
135	BPF_STMT(BPF_LD + BPF_H + BPF_IND, 16),
136	BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, 67, 0, 1),		/* patch */
137
138	/* If we passed all the tests, ask for the whole packet. */
139	BPF_STMT(BPF_RET+BPF_K, (u_int)-1),
140
141	/* Otherwise, drop it. */
142	BPF_STMT(BPF_RET+BPF_K, 0),
143};
144
145int dhcp_bpf_filter_len = sizeof(dhcp_bpf_filter) / sizeof(struct bpf_insn);
146
147/*
148 * Packet write filter program:
149 * 'ip and udp and src port bootps and dst port (bootps or bootpc)'
150 */
151struct bpf_insn dhcp_bpf_wfilter[] = {
152	BPF_STMT(BPF_LD + BPF_B + BPF_IND, 14),
153	BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, (IPVERSION << 4) + 5, 0, 12),
154
155	/* Make sure this is an IP packet... */
156	BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 12),
157	BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, ETHERTYPE_IP, 0, 10),
158
159	/* Make sure it's a UDP packet... */
160	BPF_STMT(BPF_LD + BPF_B + BPF_ABS, 23),
161	BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, IPPROTO_UDP, 0, 8),
162
163	/* Make sure this isn't a fragment... */
164	BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 20),
165	BPF_JUMP(BPF_JMP + BPF_JSET + BPF_K, 0x1fff, 6, 0),	/* patched */
166
167	/* Get the IP header length... */
168	BPF_STMT(BPF_LDX + BPF_B + BPF_MSH, 14),
169
170	/* Make sure it's from the right port... */
171	BPF_STMT(BPF_LD + BPF_H + BPF_IND, 14),
172	BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, 68, 0, 3),
173
174	/* Make sure it is to the right ports ... */
175	BPF_STMT(BPF_LD + BPF_H + BPF_IND, 16),
176	BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, 67, 0, 1),
177
178	/* If we passed all the tests, ask for the whole packet. */
179	BPF_STMT(BPF_RET+BPF_K, (u_int)-1),
180
181	/* Otherwise, drop it. */
182	BPF_STMT(BPF_RET+BPF_K, 0),
183};
184
185int dhcp_bpf_wfilter_len = sizeof(dhcp_bpf_wfilter) / sizeof(struct bpf_insn);
186
187void
188if_register_receive(struct interface_info *info)
189{
190	struct bpf_version v;
191	struct bpf_program p;
192	int flag = 1, sz;
193
194	/* Open a BPF device and hang it on this interface... */
195	info->rfdesc = if_register_bpf(info);
196
197	/* Make sure the BPF version is in range... */
198	if (ioctl(info->rfdesc, BIOCVERSION, &v) < 0)
199		error("Can't get BPF version: %m");
200
201	if (v.bv_major != BPF_MAJOR_VERSION ||
202	    v.bv_minor < BPF_MINOR_VERSION)
203		error("Kernel BPF version out of range - recompile dhcpd!");
204
205	/*
206	 * Set immediate mode so that reads return as soon as a packet
207	 * comes in, rather than waiting for the input buffer to fill
208	 * with packets.
209	 */
210	if (ioctl(info->rfdesc, BIOCIMMEDIATE, &flag) < 0)
211		error("Can't set immediate mode on bpf device: %m");
212
213	/* Get the required BPF buffer length from the kernel. */
214	if (ioctl(info->rfdesc, BIOCGBLEN, &sz) < 0)
215		error("Can't get bpf buffer length: %m");
216	info->rbuf_max = sz;
217	info->rbuf = malloc(info->rbuf_max);
218	if (!info->rbuf)
219		error("Can't allocate %lu bytes for bpf input buffer.",
220		    (unsigned long)info->rbuf_max);
221	info->rbuf_offset = 0;
222	info->rbuf_len = 0;
223
224	/* Set up the bpf filter program structure. */
225	p.bf_len = dhcp_bpf_filter_len;
226	p.bf_insns = dhcp_bpf_filter;
227
228	/* Patch the server port into the BPF program...
229	 *
230	 * XXX: changes to filter program may require changes to the
231	 * insn number(s) used below!
232	 */
233	dhcp_bpf_filter[8].k = LOCAL_PORT;
234
235	if (ioctl(info->rfdesc, BIOCSETF, &p) < 0)
236		error("Can't install packet filter program: %m");
237
238	/* Set up the bpf write filter program structure. */
239	p.bf_len = dhcp_bpf_wfilter_len;
240	p.bf_insns = dhcp_bpf_wfilter;
241
242	if (dhcp_bpf_wfilter[7].k == 0x1fff)
243		dhcp_bpf_wfilter[7].k = htons(IP_MF|IP_OFFMASK);
244
245	if (ioctl(info->rfdesc, BIOCSETWF, &p) < 0)
246		error("Can't install write filter program: %m");
247
248	if (ioctl(info->rfdesc, BIOCLOCK, NULL) < 0)
249		error("Cannot lock bpf");
250}
251
252ssize_t
253send_packet(struct interface_info *interface, struct dhcp_packet *raw,
254    size_t len, struct in_addr from, struct sockaddr_in *to,
255    struct hardware *hto)
256{
257	unsigned char buf[256];
258	struct iovec iov[2];
259	struct msghdr msg;
260	int result, bufp = 0;
261
262	/* Assemble the headers... */
263	if (to->sin_addr.s_addr == INADDR_BROADCAST)
264		assemble_hw_header(interface, buf, &bufp, hto);
265	assemble_udp_ip_header(buf, &bufp, from.s_addr,
266	    to->sin_addr.s_addr, to->sin_port, (unsigned char *)raw, len);
267
268	iov[0].iov_base = (char *)buf;
269	iov[0].iov_len = bufp;
270	iov[1].iov_base = (char *)raw;
271	iov[1].iov_len = len;
272
273	/* Fire it off */
274	if (to->sin_addr.s_addr == INADDR_BROADCAST)
275		result = writev(interface->wfdesc, iov, 2);
276	else {
277		memset(&msg, 0, sizeof(msg));
278		msg.msg_name = (struct sockaddr *)to;
279		msg.msg_namelen = sizeof(*to);
280		msg.msg_iov = iov;
281		msg.msg_iovlen = 2;
282		result = sendmsg(interface->ufdesc, &msg, 0);
283	}
284
285	if (result < 0)
286		warning("send_packet: %m");
287	return (result);
288}
289
290ssize_t
291receive_packet(struct interface_info *interface, unsigned char *buf,
292    size_t len, struct sockaddr_in *from, struct hardware *hfrom)
293{
294	int length = 0, offset = 0;
295	struct bpf_hdr hdr;
296
297	/*
298	 * All this complexity is because BPF doesn't guarantee that
299	 * only one packet will be returned at a time.  We're getting
300	 * what we deserve, though - this is a terrible abuse of the BPF
301	 * interface.  Sigh.
302	 */
303
304	/* Process packets until we get one we can return or until we've
305	 * done a read and gotten nothing we can return...
306	 */
307	do {
308		/* If the buffer is empty, fill it. */
309		if (interface->rbuf_offset >= interface->rbuf_len) {
310			length = read(interface->rfdesc, interface->rbuf,
311			    interface->rbuf_max);
312			if (length <= 0)
313				return (length);
314			interface->rbuf_offset = 0;
315			interface->rbuf_len = length;
316		}
317
318		/*
319		 * If there isn't room for a whole bpf header, something
320		 * went wrong, but we'll ignore it and hope it goes
321		 * away... XXX
322		 */
323		if (interface->rbuf_len - interface->rbuf_offset <
324		    sizeof(hdr)) {
325			interface->rbuf_offset = interface->rbuf_len;
326			continue;
327		}
328
329		/* Copy out a bpf header... */
330		memcpy(&hdr, &interface->rbuf[interface->rbuf_offset],
331		    sizeof(hdr));
332
333		/*
334		 * If the bpf header plus data doesn't fit in what's
335		 * left of the buffer, stick head in sand yet again...
336		 */
337		if (interface->rbuf_offset + hdr.bh_hdrlen + hdr.bh_caplen >
338		    interface->rbuf_len) {
339			interface->rbuf_offset = interface->rbuf_len;
340			continue;
341		}
342
343		/* Skip over the BPF header... */
344		interface->rbuf_offset += hdr.bh_hdrlen;
345
346		/*
347		 * If the captured data wasn't the whole packet, or if
348		 * the packet won't fit in the input buffer, all we can
349		 * do is drop it.
350		 */
351		if (hdr.bh_caplen != hdr.bh_datalen) {
352			interface->rbuf_offset =
353			    BPF_WORDALIGN(interface->rbuf_offset +
354			    hdr.bh_caplen);
355			continue;
356		}
357
358		/* Decode the physical header... */
359		offset = decode_hw_header(interface->rbuf,
360		    interface->rbuf_offset, hfrom);
361
362		/*
363		 * If a physical layer checksum failed (dunno of any
364		 * physical layer that supports this, but WTH), skip
365		 * this packet.
366		 */
367		if (offset < 0) {
368			interface->rbuf_offset =
369			    BPF_WORDALIGN(interface->rbuf_offset +
370			    hdr.bh_caplen);
371			continue;
372		}
373		interface->rbuf_offset += offset;
374		hdr.bh_caplen -= offset;
375
376		/* Decode the IP and UDP headers... */
377		offset = decode_udp_ip_header(interface->rbuf,
378		    interface->rbuf_offset, from, NULL, hdr.bh_caplen);
379
380		/* If the IP or UDP checksum was bad, skip the packet... */
381		if (offset < 0) {
382			interface->rbuf_offset =
383			    BPF_WORDALIGN(interface->rbuf_offset +
384			    hdr.bh_caplen);
385			continue;
386		}
387		interface->rbuf_offset += offset;
388		hdr.bh_caplen -= offset;
389
390		/*
391		 * If there's not enough room to stash the packet data,
392		 * we have to skip it (this shouldn't happen in real
393		 * life, though).
394		 */
395		if (hdr.bh_caplen > len) {
396			interface->rbuf_offset =
397			    BPF_WORDALIGN(interface->rbuf_offset +
398			    hdr.bh_caplen);
399			continue;
400		}
401
402		/* Copy out the data in the packet... */
403		memcpy(buf, interface->rbuf + interface->rbuf_offset,
404		    hdr.bh_caplen);
405		interface->rbuf_offset =
406		    BPF_WORDALIGN(interface->rbuf_offset +
407		    hdr.bh_caplen);
408		return (hdr.bh_caplen);
409	} while (!length);
410	return (0);
411}
412