1*15035272Skrw /* $OpenBSD: bpf.c,v 1.6 2005/07/29 17:26:28 krw Exp $ */ 2e853bc5dShenning 3e853bc5dShenning /* BPF socket interface code, originally contributed by Archie Cobbs. */ 4e853bc5dShenning 5e853bc5dShenning /* 6e853bc5dShenning * Copyright (c) 1995, 1996, 1998, 1999 7e853bc5dShenning * The Internet Software Consortium. All rights reserved. 8e853bc5dShenning * 9e853bc5dShenning * Redistribution and use in source and binary forms, with or without 10e853bc5dShenning * modification, are permitted provided that the following conditions 11e853bc5dShenning * are met: 12e853bc5dShenning * 13e853bc5dShenning * 1. Redistributions of source code must retain the above copyright 14e853bc5dShenning * notice, this list of conditions and the following disclaimer. 15e853bc5dShenning * 2. Redistributions in binary form must reproduce the above copyright 16e853bc5dShenning * notice, this list of conditions and the following disclaimer in the 17e853bc5dShenning * documentation and/or other materials provided with the distribution. 18e853bc5dShenning * 3. Neither the name of The Internet Software Consortium nor the names 19e853bc5dShenning * of its contributors may be used to endorse or promote products derived 20e853bc5dShenning * from this software without specific prior written permission. 21e853bc5dShenning * 22e853bc5dShenning * THIS SOFTWARE IS PROVIDED BY THE INTERNET SOFTWARE CONSORTIUM AND 23e853bc5dShenning * CONTRIBUTORS ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, 24e853bc5dShenning * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF 25e853bc5dShenning * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE 26e853bc5dShenning * DISCLAIMED. IN NO EVENT SHALL THE INTERNET SOFTWARE CONSORTIUM OR 27e853bc5dShenning * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 28e853bc5dShenning * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 29e853bc5dShenning * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF 30e853bc5dShenning * USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND 31e853bc5dShenning * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 32e853bc5dShenning * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT 33e853bc5dShenning * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34e853bc5dShenning * SUCH DAMAGE. 35e853bc5dShenning * 36e853bc5dShenning * This software has been written for the Internet Software Consortium 37e853bc5dShenning * by Ted Lemon <mellon@fugue.com> in cooperation with Vixie 38e853bc5dShenning * Enterprises. To learn more about the Internet Software Consortium, 39e853bc5dShenning * see ``http://www.vix.com/isc''. To learn more about Vixie 40e853bc5dShenning * Enterprises, see ``http://www.vix.com''. 41e853bc5dShenning */ 42e853bc5dShenning 43e853bc5dShenning #include "dhcpd.h" 44e853bc5dShenning #include <sys/ioctl.h> 45e853bc5dShenning #include <sys/uio.h> 46e853bc5dShenning 47e853bc5dShenning #include <net/bpf.h> 48e853bc5dShenning #include <netinet/in_systm.h> 49e853bc5dShenning #include <netinet/ip.h> 50e853bc5dShenning #include <netinet/udp.h> 51e853bc5dShenning #include <netinet/if_ether.h> 52e853bc5dShenning 53e853bc5dShenning #define BPF_FORMAT "/dev/bpf%d" 54e853bc5dShenning 55e853bc5dShenning /* 56e853bc5dShenning * Called by get_interface_list for each interface that's discovered. 57e853bc5dShenning * Opens a packet filter for each interface and adds it to the select 58e853bc5dShenning * mask. 59e853bc5dShenning */ 60e853bc5dShenning int 61e853bc5dShenning if_register_bpf(struct interface_info *info) 62e853bc5dShenning { 63e853bc5dShenning char filename[50]; 64e853bc5dShenning int sock, b; 65e853bc5dShenning 66e853bc5dShenning /* Open a BPF device */ 67e853bc5dShenning for (b = 0; 1; b++) { 68e853bc5dShenning snprintf(filename, sizeof(filename), BPF_FORMAT, b); 69e853bc5dShenning sock = open(filename, O_RDWR, 0); 70e853bc5dShenning if (sock < 0) { 71e853bc5dShenning if (errno == EBUSY) 72e853bc5dShenning continue; 73e853bc5dShenning else 74e853bc5dShenning error("Can't find free bpf: %m"); 75e853bc5dShenning } else 76e853bc5dShenning break; 77e853bc5dShenning } 78e853bc5dShenning 79e853bc5dShenning /* Set the BPF device to point at this interface. */ 80e853bc5dShenning if (ioctl(sock, BIOCSETIF, info->ifp) < 0) 81e853bc5dShenning error("Can't attach interface %s to bpf device %s: %m", 82e853bc5dShenning info->name, filename); 83e853bc5dShenning 84e853bc5dShenning return (sock); 85e853bc5dShenning } 86e853bc5dShenning 87e853bc5dShenning void 88e853bc5dShenning if_register_send(struct interface_info *info) 89e853bc5dShenning { 90e853bc5dShenning /* 91e853bc5dShenning * If we're using the bpf API for sending and receiving, we 92e853bc5dShenning * don't need to register this interface twice. 93e853bc5dShenning */ 94e853bc5dShenning info->wfdesc = info->rfdesc; 95e853bc5dShenning } 96e853bc5dShenning 97e853bc5dShenning /* 98390956b7Scanacar * Packet read filter program: 'ip and udp and dst port bootps' 99e853bc5dShenning */ 100e853bc5dShenning struct bpf_insn dhcp_bpf_filter[] = { 101e853bc5dShenning /* Make sure this is an IP packet... */ 102e853bc5dShenning BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 12), 103e853bc5dShenning BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, ETHERTYPE_IP, 0, 8), 104e853bc5dShenning 105e853bc5dShenning /* Make sure it's a UDP packet... */ 106e853bc5dShenning BPF_STMT(BPF_LD + BPF_B + BPF_ABS, 23), 107e853bc5dShenning BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, IPPROTO_UDP, 0, 6), 108e853bc5dShenning 109e853bc5dShenning /* Make sure this isn't a fragment... */ 110e853bc5dShenning BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 20), 111e853bc5dShenning BPF_JUMP(BPF_JMP + BPF_JSET + BPF_K, 0x1fff, 4, 0), 112e853bc5dShenning 113e853bc5dShenning /* Get the IP header length... */ 114e853bc5dShenning BPF_STMT(BPF_LDX + BPF_B + BPF_MSH, 14), 115e853bc5dShenning 116e853bc5dShenning /* Make sure it's to the right port... */ 117e853bc5dShenning BPF_STMT(BPF_LD + BPF_H + BPF_IND, 16), 118390956b7Scanacar BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, SERVER_PORT, 0, 1), 119e853bc5dShenning 120e853bc5dShenning /* If we passed all the tests, ask for the whole packet. */ 121e853bc5dShenning BPF_STMT(BPF_RET+BPF_K, (u_int)-1), 122e853bc5dShenning 123e853bc5dShenning /* Otherwise, drop it. */ 124e853bc5dShenning BPF_STMT(BPF_RET+BPF_K, 0), 125e853bc5dShenning }; 126e853bc5dShenning 127e853bc5dShenning int dhcp_bpf_filter_len = sizeof(dhcp_bpf_filter) / sizeof(struct bpf_insn); 128e853bc5dShenning 129390956b7Scanacar 130390956b7Scanacar /* 131390956b7Scanacar * Packet write filter program: 132390956b7Scanacar * 'ip and udp and src port bootps and dst port (bootps or bootpc)' 133390956b7Scanacar */ 134390956b7Scanacar struct bpf_insn dhcp_bpf_wfilter[] = { 135390956b7Scanacar /* Make sure this is an IP packet... */ 136390956b7Scanacar BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 12), 137390956b7Scanacar BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, ETHERTYPE_IP, 0, 11), 138390956b7Scanacar 139390956b7Scanacar /* Make sure it's a UDP packet... */ 140390956b7Scanacar BPF_STMT(BPF_LD + BPF_B + BPF_ABS, 23), 141390956b7Scanacar BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, IPPROTO_UDP, 0, 9), 142390956b7Scanacar 143390956b7Scanacar /* Make sure this isn't a fragment... */ 144390956b7Scanacar BPF_STMT(BPF_LD + BPF_H + BPF_ABS, 20), 145390956b7Scanacar BPF_JUMP(BPF_JMP + BPF_JSET + BPF_K, 0x1fff, 7, 0), 146390956b7Scanacar 147390956b7Scanacar /* Get the IP header length... */ 148390956b7Scanacar BPF_STMT(BPF_LDX + BPF_B + BPF_MSH, 14), 149390956b7Scanacar 150390956b7Scanacar /* Make sure it's from the right port... */ 151390956b7Scanacar BPF_STMT(BPF_LD + BPF_H + BPF_IND, 14), 152390956b7Scanacar BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, SERVER_PORT, 0, 4), 153390956b7Scanacar 154390956b7Scanacar /* Make sure it is to the right ports ... */ 155390956b7Scanacar BPF_STMT(BPF_LD + BPF_H + BPF_IND, 16), 156390956b7Scanacar BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, CLIENT_PORT, 1, 0), 157390956b7Scanacar BPF_JUMP(BPF_JMP + BPF_JEQ + BPF_K, SERVER_PORT, 0, 1), 158390956b7Scanacar 159390956b7Scanacar /* If we passed all the tests, ask for the whole packet. */ 160390956b7Scanacar BPF_STMT(BPF_RET+BPF_K, (u_int)-1), 161390956b7Scanacar 162390956b7Scanacar /* Otherwise, drop it. */ 163390956b7Scanacar BPF_STMT(BPF_RET+BPF_K, 0), 164390956b7Scanacar }; 165390956b7Scanacar 166390956b7Scanacar int dhcp_bpf_wfilter_len = sizeof(dhcp_bpf_wfilter) / sizeof(struct bpf_insn); 167390956b7Scanacar 168e853bc5dShenning void 169e853bc5dShenning if_register_receive(struct interface_info *info) 170e853bc5dShenning { 171e853bc5dShenning struct bpf_version v; 172e853bc5dShenning struct bpf_program p; 173390956b7Scanacar int flag = 1, sz, cmplt = 0; 174e853bc5dShenning 175e853bc5dShenning /* Open a BPF device and hang it on this interface... */ 176e853bc5dShenning info->rfdesc = if_register_bpf(info); 177e853bc5dShenning 178e853bc5dShenning /* Make sure the BPF version is in range... */ 179e853bc5dShenning if (ioctl(info->rfdesc, BIOCVERSION, &v) < 0) 180e853bc5dShenning error("Can't get BPF version: %m"); 181e853bc5dShenning 182e853bc5dShenning if (v.bv_major != BPF_MAJOR_VERSION || 183e853bc5dShenning v.bv_minor < BPF_MINOR_VERSION) 184e853bc5dShenning error("Kernel BPF version out of range - recompile dhcpd!"); 185e853bc5dShenning 186e853bc5dShenning /* 187e853bc5dShenning * Set immediate mode so that reads return as soon as a packet 188e853bc5dShenning * comes in, rather than waiting for the input buffer to fill 189e853bc5dShenning * with packets. 190e853bc5dShenning */ 191e853bc5dShenning if (ioctl(info->rfdesc, BIOCIMMEDIATE, &flag) < 0) 192e853bc5dShenning error("Can't set immediate mode on bpf device: %m"); 193e853bc5dShenning 194390956b7Scanacar /* make sure kernel fills in the source ethernet address */ 195390956b7Scanacar if (ioctl(info->rfdesc, BIOCSHDRCMPLT, &cmplt) < 0) 196390956b7Scanacar error("Can't set header complete flag on bpf device: %m"); 197390956b7Scanacar 198e853bc5dShenning /* Get the required BPF buffer length from the kernel. */ 199e853bc5dShenning if (ioctl(info->rfdesc, BIOCGBLEN, &sz) < 0) 200e853bc5dShenning error("Can't get bpf buffer length: %m"); 201e853bc5dShenning info->rbuf_max = sz; 202e853bc5dShenning info->rbuf = malloc(info->rbuf_max); 203e853bc5dShenning if (!info->rbuf) 204e853bc5dShenning error("Can't allocate %lu bytes for bpf input buffer.", 205e853bc5dShenning (unsigned long)info->rbuf_max); 206e853bc5dShenning info->rbuf_offset = 0; 207e853bc5dShenning info->rbuf_len = 0; 208e853bc5dShenning 209e853bc5dShenning /* Set up the bpf filter program structure. */ 210e853bc5dShenning p.bf_len = dhcp_bpf_filter_len; 211e853bc5dShenning p.bf_insns = dhcp_bpf_filter; 212e853bc5dShenning 213e853bc5dShenning if (ioctl(info->rfdesc, BIOCSETF, &p) < 0) 214e853bc5dShenning error("Can't install packet filter program: %m"); 215390956b7Scanacar 216390956b7Scanacar /* Set up the bpf write filter program structure. */ 217390956b7Scanacar p.bf_len = dhcp_bpf_wfilter_len; 218390956b7Scanacar p.bf_insns = dhcp_bpf_wfilter; 219390956b7Scanacar 220390956b7Scanacar if (ioctl(info->rfdesc, BIOCSETWF, &p) < 0) 221390956b7Scanacar error("Can't install write filter program: %m"); 222390956b7Scanacar 223390956b7Scanacar /* make sure these settings cannot be changed after dropping privs */ 224390956b7Scanacar if (ioctl(info->rfdesc, BIOCLOCK) < 0) 225390956b7Scanacar error("Failed to lock bpf descriptor: %m"); 226e853bc5dShenning } 227e853bc5dShenning 228e853bc5dShenning ssize_t 229285f06efSderaadt send_packet(struct interface_info *interface, struct dhcp_packet *raw, 230285f06efSderaadt size_t len, struct in_addr from, struct sockaddr_in *to, 231285f06efSderaadt struct hardware *hto) 232e853bc5dShenning { 233e853bc5dShenning unsigned char buf[256]; 234e853bc5dShenning struct iovec iov[2]; 235e853bc5dShenning int result, bufp = 0; 236e853bc5dShenning 237e853bc5dShenning /* Assemble the headers... */ 238e853bc5dShenning assemble_hw_header(interface, buf, &bufp, hto); 239e853bc5dShenning assemble_udp_ip_header(interface, buf, &bufp, from.s_addr, 240e853bc5dShenning to->sin_addr.s_addr, to->sin_port, (unsigned char *)raw, len); 241e853bc5dShenning 242e853bc5dShenning /* Fire it off */ 243e853bc5dShenning iov[0].iov_base = (char *)buf; 244e853bc5dShenning iov[0].iov_len = bufp; 245e853bc5dShenning iov[1].iov_base = (char *)raw; 246e853bc5dShenning iov[1].iov_len = len; 247e853bc5dShenning 248e853bc5dShenning result = writev(interface->wfdesc, iov, 2); 249e853bc5dShenning if (result < 0) 2500795b389Sderaadt warning("send_packet: %m"); 251e853bc5dShenning return (result); 252e853bc5dShenning } 253e853bc5dShenning 254e853bc5dShenning ssize_t 255e853bc5dShenning receive_packet(struct interface_info *interface, unsigned char *buf, 256e853bc5dShenning size_t len, struct sockaddr_in *from, struct hardware *hfrom) 257e853bc5dShenning { 258e853bc5dShenning int length = 0, offset = 0; 259e853bc5dShenning struct bpf_hdr hdr; 260e853bc5dShenning 261e853bc5dShenning /* 262e853bc5dShenning * All this complexity is because BPF doesn't guarantee that 263e853bc5dShenning * only one packet will be returned at a time. We're getting 264e853bc5dShenning * what we deserve, though - this is a terrible abuse of the BPF 265e853bc5dShenning * interface. Sigh. 266e853bc5dShenning */ 267e853bc5dShenning 268e853bc5dShenning /* Process packets until we get one we can return or until we've 269e853bc5dShenning * done a read and gotten nothing we can return... 270e853bc5dShenning */ 271e853bc5dShenning do { 272e853bc5dShenning /* If the buffer is empty, fill it. */ 273e853bc5dShenning if (interface->rbuf_offset == interface->rbuf_len) { 274e853bc5dShenning length = read(interface->rfdesc, interface->rbuf, 275e853bc5dShenning interface->rbuf_max); 276e853bc5dShenning if (length <= 0) 277e853bc5dShenning return (length); 278e853bc5dShenning interface->rbuf_offset = 0; 279*15035272Skrw interface->rbuf_len = BPF_WORDALIGN(length); 280e853bc5dShenning } 281e853bc5dShenning 282e853bc5dShenning /* 283e853bc5dShenning * If there isn't room for a whole bpf header, something 284e853bc5dShenning * went wrong, but we'll ignore it and hope it goes 285e853bc5dShenning * away... XXX 286e853bc5dShenning */ 287e853bc5dShenning if (interface->rbuf_len - interface->rbuf_offset < 288e853bc5dShenning sizeof(hdr)) { 289e853bc5dShenning interface->rbuf_offset = interface->rbuf_len; 290e853bc5dShenning continue; 291e853bc5dShenning } 292e853bc5dShenning 293e853bc5dShenning /* Copy out a bpf header... */ 294e853bc5dShenning memcpy(&hdr, &interface->rbuf[interface->rbuf_offset], 295e853bc5dShenning sizeof(hdr)); 296e853bc5dShenning 297e853bc5dShenning /* 298e853bc5dShenning * If the bpf header plus data doesn't fit in what's 299e853bc5dShenning * left of the buffer, stick head in sand yet again... 300e853bc5dShenning */ 301e853bc5dShenning if (interface->rbuf_offset + hdr.bh_hdrlen + hdr.bh_caplen > 302e853bc5dShenning interface->rbuf_len) { 303e853bc5dShenning interface->rbuf_offset = interface->rbuf_len; 304e853bc5dShenning continue; 305e853bc5dShenning } 306e853bc5dShenning 307e853bc5dShenning /* 308e853bc5dShenning * If the captured data wasn't the whole packet, or if 309e853bc5dShenning * the packet won't fit in the input buffer, all we can 310e853bc5dShenning * do is drop it. 311e853bc5dShenning */ 312e853bc5dShenning if (hdr.bh_caplen != hdr.bh_datalen) { 313*15035272Skrw interface->rbuf_offset = BPF_WORDALIGN( 314*15035272Skrw interface->rbuf_offset + hdr.bh_hdrlen + 315*15035272Skrw hdr.bh_caplen); 316e853bc5dShenning continue; 317e853bc5dShenning } 318e853bc5dShenning 319e853bc5dShenning /* Skip over the BPF header... */ 320e853bc5dShenning interface->rbuf_offset += hdr.bh_hdrlen; 321e853bc5dShenning 322e853bc5dShenning /* Decode the physical header... */ 323e853bc5dShenning offset = decode_hw_header(interface, 324e853bc5dShenning interface->rbuf, interface->rbuf_offset, hfrom); 325e853bc5dShenning 326e853bc5dShenning /* 327e853bc5dShenning * If a physical layer checksum failed (dunno of any 328e853bc5dShenning * physical layer that supports this, but WTH), skip 329e853bc5dShenning * this packet. 330e853bc5dShenning */ 331e853bc5dShenning if (offset < 0) { 332*15035272Skrw interface->rbuf_offset = BPF_WORDALIGN( 333*15035272Skrw interface->rbuf_offset + hdr.bh_caplen); 334e853bc5dShenning continue; 335e853bc5dShenning } 336e853bc5dShenning interface->rbuf_offset += offset; 337e853bc5dShenning hdr.bh_caplen -= offset; 338e853bc5dShenning 339e853bc5dShenning /* Decode the IP and UDP headers... */ 340e853bc5dShenning offset = decode_udp_ip_header(interface, interface->rbuf, 341e853bc5dShenning interface->rbuf_offset, from, NULL, hdr.bh_caplen); 342e853bc5dShenning 343e853bc5dShenning /* If the IP or UDP checksum was bad, skip the packet... */ 344e853bc5dShenning if (offset < 0) { 345*15035272Skrw interface->rbuf_offset = BPF_WORDALIGN( 346*15035272Skrw interface->rbuf_offset + hdr.bh_caplen); 347e853bc5dShenning continue; 348e853bc5dShenning } 349e853bc5dShenning interface->rbuf_offset += offset; 350e853bc5dShenning hdr.bh_caplen -= offset; 351e853bc5dShenning 352e853bc5dShenning /* 353e853bc5dShenning * If there's not enough room to stash the packet data, 354e853bc5dShenning * we have to skip it (this shouldn't happen in real 355e853bc5dShenning * life, though). 356e853bc5dShenning */ 357e853bc5dShenning if (hdr.bh_caplen > len) { 358*15035272Skrw interface->rbuf_offset = BPF_WORDALIGN( 359*15035272Skrw interface->rbuf_offset + hdr.bh_caplen); 360e853bc5dShenning continue; 361e853bc5dShenning } 362e853bc5dShenning 363e853bc5dShenning /* Copy out the data in the packet... */ 364e853bc5dShenning memcpy(buf, interface->rbuf + interface->rbuf_offset, 365e853bc5dShenning hdr.bh_caplen); 366*15035272Skrw interface->rbuf_offset = BPF_WORDALIGN(interface->rbuf_offset + 367*15035272Skrw hdr.bh_caplen); 368e853bc5dShenning return (hdr.bh_caplen); 369e853bc5dShenning } while (!length); 370e853bc5dShenning return (0); 371e853bc5dShenning } 372