bpf.c revision 147073
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 "dhcpd.h"
44#include <sys/ioctl.h>
45#include <sys/uio.h>
46
47#include <net/bpf.h>
48#include <netinet/in_systm.h>
49#include <netinet/ip.h>
50#include <netinet/udp.h>
51#include <netinet/if_ether.h>
52
53#define BPF_FORMAT "/dev/bpf%d"
54
55/*
56 * Called by get_interface_list for each interface that's discovered.
57 * Opens a packet filter for each interface and adds it to the select
58 * mask.
59 */
60int
61if_register_bpf(struct interface_info *info)
62{
63	char filename[50];
64	int sock, b;
65
66	/* Open a BPF device */
67	for (b = 0; 1; b++) {
68		snprintf(filename, sizeof(filename), BPF_FORMAT, b);
69		sock = open(filename, O_RDWR, 0);
70		if (sock < 0) {
71			if (errno == EBUSY)
72				continue;
73			else
74				error("Can't find free bpf: %m");
75		} else
76			break;
77	}
78
79	/* Set the BPF device to point at this interface. */
80	if (ioctl(sock, BIOCSETIF, info->ifp) < 0)
81		error("Can't attach interface %s to bpf device %s: %m",
82		    info->name, filename);
83
84	return (sock);
85}
86
87void
88if_register_send(struct interface_info *info)
89{
90	/*
91	 * If we're using the bpf API for sending and receiving, we
92	 * don't need to register this interface twice.
93	 */
94	info->wfdesc = info->rfdesc;
95}
96
97/*
98 * Packet filter program...
99 *
100 * XXX: Changes to the filter program may require changes to the
101 * constant offsets used in if_register_send to patch the BPF program!
102 */
103struct bpf_insn dhcp_bpf_filter[] = {
104	/* Make sure this is an IP packet... */
105	BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 12),
106	BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, ETHERTYPE_IP, 0, 8),
107
108	/* Make sure it's a UDP packet... */
109	BPF_STMT(BPF_LD + BPF_B + BPF_ABS, 23),
110	BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, IPPROTO_UDP, 0, 6),
111
112	/* Make sure this isn't a fragment... */
113	BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 20),
114	BPF_JUMP(BPF_JMP + BPF_JSET + BPF_K, 0x1fff, 4, 0),
115
116	/* Get the IP header length... */
117	BPF_STMT(BPF_LDX + BPF_B + BPF_MSH, 14),
118
119	/* Make sure it's to the right port... */
120	BPF_STMT(BPF_LD + BPF_H + BPF_IND, 16),
121	BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, 67, 0, 1),		/* patch */
122
123	/* If we passed all the tests, ask for the whole packet. */
124	BPF_STMT(BPF_RET+BPF_K, (u_int)-1),
125
126	/* Otherwise, drop it. */
127	BPF_STMT(BPF_RET+BPF_K, 0),
128};
129
130int dhcp_bpf_filter_len = sizeof(dhcp_bpf_filter) / sizeof(struct bpf_insn);
131
132/*
133 * Packet write filter program:
134 * 'ip and udp and src port bootps and dst port (bootps or bootpc)'
135 */
136struct bpf_insn dhcp_bpf_wfilter[] = {
137	BPF_STMT(BPF_LD + BPF_B + BPF_IND, 14),
138	BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, (IPVERSION << 4) + 5, 0, 12),
139
140	/* Make sure this is an IP packet... */
141	BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 12),
142	BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, ETHERTYPE_IP, 0, 10),
143
144	/* Make sure it's a UDP packet... */
145	BPF_STMT(BPF_LD + BPF_B + BPF_ABS, 23),
146	BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, IPPROTO_UDP, 0, 8),
147
148	/* Make sure this isn't a fragment... */
149	BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 20),
150	BPF_JUMP(BPF_JMP + BPF_JSET + BPF_K, 0x1fff, 6, 0),	/* patched */
151
152	/* Get the IP header length... */
153	BPF_STMT(BPF_LDX + BPF_B + BPF_MSH, 14),
154
155	/* Make sure it's from the right port... */
156	BPF_STMT(BPF_LD + BPF_H + BPF_IND, 14),
157	BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, 68, 0, 3),
158
159	/* Make sure it is to the right ports ... */
160	BPF_STMT(BPF_LD + BPF_H + BPF_IND, 16),
161	BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, 67, 0, 1),
162
163	/* If we passed all the tests, ask for the whole packet. */
164	BPF_STMT(BPF_RET+BPF_K, (u_int)-1),
165
166	/* Otherwise, drop it. */
167	BPF_STMT(BPF_RET+BPF_K, 0),
168};
169
170int dhcp_bpf_wfilter_len = sizeof(dhcp_bpf_wfilter) / sizeof(struct bpf_insn);
171
172void
173if_register_receive(struct interface_info *info)
174{
175	struct bpf_version v;
176	struct bpf_program p;
177	int flag = 1, sz;
178
179	/* Open a BPF device and hang it on this interface... */
180	info->rfdesc = if_register_bpf(info);
181
182	/* Make sure the BPF version is in range... */
183	if (ioctl(info->rfdesc, BIOCVERSION, &v) < 0)
184		error("Can't get BPF version: %m");
185
186	if (v.bv_major != BPF_MAJOR_VERSION ||
187	    v.bv_minor < BPF_MINOR_VERSION)
188		error("Kernel BPF version out of range - recompile dhcpd!");
189
190	/*
191	 * Set immediate mode so that reads return as soon as a packet
192	 * comes in, rather than waiting for the input buffer to fill
193	 * with packets.
194	 */
195	if (ioctl(info->rfdesc, BIOCIMMEDIATE, &flag) < 0)
196		error("Can't set immediate mode on bpf device: %m");
197
198	/* Get the required BPF buffer length from the kernel. */
199	if (ioctl(info->rfdesc, BIOCGBLEN, &sz) < 0)
200		error("Can't get bpf buffer length: %m");
201	info->rbuf_max = sz;
202	info->rbuf = malloc(info->rbuf_max);
203	if (!info->rbuf)
204		error("Can't allocate %lu bytes for bpf input buffer.",
205		    (unsigned long)info->rbuf_max);
206	info->rbuf_offset = 0;
207	info->rbuf_len = 0;
208
209	/* Set up the bpf filter program structure. */
210	p.bf_len = dhcp_bpf_filter_len;
211	p.bf_insns = dhcp_bpf_filter;
212
213	/* Patch the server port into the BPF program...
214	 *
215	 * XXX: changes to filter program may require changes to the
216	 * insn number(s) used below!
217	 */
218	dhcp_bpf_filter[8].k = LOCAL_PORT;
219
220	if (ioctl(info->rfdesc, BIOCSETF, &p) < 0)
221		error("Can't install packet filter program: %m");
222
223	/* Set up the bpf write filter program structure. */
224	p.bf_len = dhcp_bpf_wfilter_len;
225	p.bf_insns = dhcp_bpf_wfilter;
226
227	if (dhcp_bpf_wfilter[7].k == 0x1fff)
228		dhcp_bpf_wfilter[7].k = htons(IP_MF|IP_OFFMASK);
229
230	if (ioctl(info->rfdesc, BIOCSETWF, &p) < 0)
231		error("Can't install write filter program: %m");
232
233	if (ioctl(info->rfdesc, BIOCLOCK, NULL) < 0)
234		error("Cannot lock bpf");
235}
236
237ssize_t
238send_packet(struct interface_info *interface, struct dhcp_packet *raw,
239    size_t len, struct in_addr from, struct sockaddr_in *to,
240    struct hardware *hto)
241{
242	unsigned char buf[256];
243	struct iovec iov[2];
244	int result, bufp = 0;
245
246	/* Assemble the headers... */
247	assemble_hw_header(interface, buf, &bufp, hto);
248	assemble_udp_ip_header(buf, &bufp, from.s_addr,
249	    to->sin_addr.s_addr, to->sin_port, (unsigned char *)raw, len);
250
251	/* Fire it off */
252	iov[0].iov_base = (char *)buf;
253	iov[0].iov_len = bufp;
254	iov[1].iov_base = (char *)raw;
255	iov[1].iov_len = len;
256
257	result = writev(interface->wfdesc, iov, 2);
258	if (result < 0)
259		warning("send_packet: %m");
260	return (result);
261}
262
263ssize_t
264receive_packet(struct interface_info *interface, unsigned char *buf,
265    size_t len, struct sockaddr_in *from, struct hardware *hfrom)
266{
267	int length = 0, offset = 0;
268	struct bpf_hdr hdr;
269
270	/*
271	 * All this complexity is because BPF doesn't guarantee that
272	 * only one packet will be returned at a time.  We're getting
273	 * what we deserve, though - this is a terrible abuse of the BPF
274	 * interface.  Sigh.
275	 */
276
277	/* Process packets until we get one we can return or until we've
278	 * done a read and gotten nothing we can return...
279	 */
280	do {
281		/* If the buffer is empty, fill it. */
282		if (interface->rbuf_offset == interface->rbuf_len) {
283			length = read(interface->rfdesc, interface->rbuf,
284			    interface->rbuf_max);
285			if (length <= 0)
286				return (length);
287			interface->rbuf_offset = 0;
288			interface->rbuf_len = length;
289		}
290
291		/*
292		 * If there isn't room for a whole bpf header, something
293		 * went wrong, but we'll ignore it and hope it goes
294		 * away... XXX
295		 */
296		if (interface->rbuf_len - interface->rbuf_offset <
297		    sizeof(hdr)) {
298			interface->rbuf_offset = interface->rbuf_len;
299			continue;
300		}
301
302		/* Copy out a bpf header... */
303		memcpy(&hdr, &interface->rbuf[interface->rbuf_offset],
304		    sizeof(hdr));
305
306		/*
307		 * If the bpf header plus data doesn't fit in what's
308		 * left of the buffer, stick head in sand yet again...
309		 */
310		if (interface->rbuf_offset + hdr.bh_hdrlen + hdr.bh_caplen >
311		    interface->rbuf_len) {
312			interface->rbuf_offset = interface->rbuf_len;
313			continue;
314		}
315
316		/*
317		 * If the captured data wasn't the whole packet, or if
318		 * the packet won't fit in the input buffer, all we can
319		 * do is drop it.
320		 */
321		if (hdr.bh_caplen != hdr.bh_datalen) {
322			interface->rbuf_offset += hdr.bh_hdrlen = hdr.bh_caplen;
323			continue;
324		}
325
326		/* Skip over the BPF header... */
327		interface->rbuf_offset += hdr.bh_hdrlen;
328
329		/* Decode the physical header... */
330		offset = decode_hw_header(interface->rbuf,
331		    interface->rbuf_offset, hfrom);
332
333		/*
334		 * If a physical layer checksum failed (dunno of any
335		 * physical layer that supports this, but WTH), skip
336		 * this packet.
337		 */
338		if (offset < 0) {
339			interface->rbuf_offset += hdr.bh_caplen;
340			continue;
341		}
342		interface->rbuf_offset += offset;
343		hdr.bh_caplen -= offset;
344
345		/* Decode the IP and UDP headers... */
346		offset = decode_udp_ip_header(interface->rbuf,
347		    interface->rbuf_offset, from, NULL, hdr.bh_caplen);
348
349		/* If the IP or UDP checksum was bad, skip the packet... */
350		if (offset < 0) {
351			interface->rbuf_offset += hdr.bh_caplen;
352			continue;
353		}
354		interface->rbuf_offset += offset;
355		hdr.bh_caplen -= offset;
356
357		/*
358		 * If there's not enough room to stash the packet data,
359		 * we have to skip it (this shouldn't happen in real
360		 * life, though).
361		 */
362		if (hdr.bh_caplen > len) {
363			interface->rbuf_offset += hdr.bh_caplen;
364			continue;
365		}
366
367		/* Copy out the data in the packet... */
368		memcpy(buf, interface->rbuf + interface->rbuf_offset,
369		    hdr.bh_caplen);
370		interface->rbuf_offset += hdr.bh_caplen;
371		return (hdr.bh_caplen);
372	} while (!length);
373	return (0);
374}
375