| 1 | /* |
| 2 | * Linux Socket Filter - Kernel level socket filtering |
| 3 | * |
| 4 | * Author: |
| 5 | * Jay Schulist <jschlst@samba.org> |
| 6 | * |
| 7 | * Based on the design of: |
| 8 | * - The Berkeley Packet Filter |
| 9 | * |
| 10 | * This program is free software; you can redistribute it and/or |
| 11 | * modify it under the terms of the GNU General Public License |
| 12 | * as published by the Free Software Foundation; either version |
| 13 | * 2 of the License, or (at your option) any later version. |
| 14 | * |
| 15 | * Andi Kleen - Fix a few bad bugs and races. |
| 16 | * Kris Katterjohn - Added many additional checks in sk_chk_filter() |
| 17 | */ |
| 18 | |
| 19 | #include <linux/module.h> |
| 20 | #include <linux/types.h> |
| 21 | #include <linux/mm.h> |
| 22 | #include <linux/fcntl.h> |
| 23 | #include <linux/socket.h> |
| 24 | #include <linux/in.h> |
| 25 | #include <linux/inet.h> |
| 26 | #include <linux/netdevice.h> |
| 27 | #include <linux/if_packet.h> |
| 28 | #include <linux/gfp.h> |
| 29 | #include <net/ip.h> |
| 30 | #include <net/protocol.h> |
| 31 | #include <net/netlink.h> |
| 32 | #include <linux/skbuff.h> |
| 33 | #include <net/sock.h> |
| 34 | #include <linux/errno.h> |
| 35 | #include <linux/timer.h> |
| 36 | #include <asm/system.h> |
| 37 | #include <asm/uaccess.h> |
| 38 | #include <asm/unaligned.h> |
| 39 | #include <linux/filter.h> |
| 40 | #include <linux/reciprocal_div.h> |
| 41 | |
| 42 | enum { |
| 43 | BPF_S_RET_K = 1, |
| 44 | BPF_S_RET_A, |
| 45 | BPF_S_ALU_ADD_K, |
| 46 | BPF_S_ALU_ADD_X, |
| 47 | BPF_S_ALU_SUB_K, |
| 48 | BPF_S_ALU_SUB_X, |
| 49 | BPF_S_ALU_MUL_K, |
| 50 | BPF_S_ALU_MUL_X, |
| 51 | BPF_S_ALU_DIV_X, |
| 52 | BPF_S_ALU_AND_K, |
| 53 | BPF_S_ALU_AND_X, |
| 54 | BPF_S_ALU_OR_K, |
| 55 | BPF_S_ALU_OR_X, |
| 56 | BPF_S_ALU_LSH_K, |
| 57 | BPF_S_ALU_LSH_X, |
| 58 | BPF_S_ALU_RSH_K, |
| 59 | BPF_S_ALU_RSH_X, |
| 60 | BPF_S_ALU_NEG, |
| 61 | BPF_S_LD_W_ABS, |
| 62 | BPF_S_LD_H_ABS, |
| 63 | BPF_S_LD_B_ABS, |
| 64 | BPF_S_LD_W_LEN, |
| 65 | BPF_S_LD_W_IND, |
| 66 | BPF_S_LD_H_IND, |
| 67 | BPF_S_LD_B_IND, |
| 68 | BPF_S_LD_IMM, |
| 69 | BPF_S_LDX_W_LEN, |
| 70 | BPF_S_LDX_B_MSH, |
| 71 | BPF_S_LDX_IMM, |
| 72 | BPF_S_MISC_TAX, |
| 73 | BPF_S_MISC_TXA, |
| 74 | BPF_S_ALU_DIV_K, |
| 75 | BPF_S_LD_MEM, |
| 76 | BPF_S_LDX_MEM, |
| 77 | BPF_S_ST, |
| 78 | BPF_S_STX, |
| 79 | BPF_S_JMP_JA, |
| 80 | BPF_S_JMP_JEQ_K, |
| 81 | BPF_S_JMP_JEQ_X, |
| 82 | BPF_S_JMP_JGE_K, |
| 83 | BPF_S_JMP_JGE_X, |
| 84 | BPF_S_JMP_JGT_K, |
| 85 | BPF_S_JMP_JGT_X, |
| 86 | BPF_S_JMP_JSET_K, |
| 87 | BPF_S_JMP_JSET_X, |
| 88 | /* Ancillary data */ |
| 89 | BPF_S_ANC_PROTOCOL, |
| 90 | BPF_S_ANC_PKTTYPE, |
| 91 | BPF_S_ANC_IFINDEX, |
| 92 | BPF_S_ANC_NLATTR, |
| 93 | BPF_S_ANC_NLATTR_NEST, |
| 94 | BPF_S_ANC_MARK, |
| 95 | BPF_S_ANC_QUEUE, |
| 96 | BPF_S_ANC_HATYPE, |
| 97 | BPF_S_ANC_RXHASH, |
| 98 | BPF_S_ANC_CPU, |
| 99 | }; |
| 100 | |
| 101 | /* No hurry in this branch */ |
| 102 | static void *__load_pointer(const struct sk_buff *skb, int k, unsigned int size) |
| 103 | { |
| 104 | u8 *ptr = NULL; |
| 105 | |
| 106 | if (k >= SKF_NET_OFF) |
| 107 | ptr = skb_network_header(skb) + k - SKF_NET_OFF; |
| 108 | else if (k >= SKF_LL_OFF) |
| 109 | ptr = skb_mac_header(skb) + k - SKF_LL_OFF; |
| 110 | |
| 111 | if (ptr >= skb->head && ptr + size <= skb_tail_pointer(skb)) |
| 112 | return ptr; |
| 113 | return NULL; |
| 114 | } |
| 115 | |
| 116 | static inline void *load_pointer(const struct sk_buff *skb, int k, |
| 117 | unsigned int size, void *buffer) |
| 118 | { |
| 119 | if (k >= 0) |
| 120 | return skb_header_pointer(skb, k, size, buffer); |
| 121 | return __load_pointer(skb, k, size); |
| 122 | } |
| 123 | |
| 124 | /** |
| 125 | * sk_filter - run a packet through a socket filter |
| 126 | * @sk: sock associated with &sk_buff |
| 127 | * @skb: buffer to filter |
| 128 | * |
| 129 | * Run the filter code and then cut skb->data to correct size returned by |
| 130 | * sk_run_filter. If pkt_len is 0 we toss packet. If skb->len is smaller |
| 131 | * than pkt_len we keep whole skb->data. This is the socket level |
| 132 | * wrapper to sk_run_filter. It returns 0 if the packet should |
| 133 | * be accepted or -EPERM if the packet should be tossed. |
| 134 | * |
| 135 | */ |
| 136 | int sk_filter(struct sock *sk, struct sk_buff *skb) |
| 137 | { |
| 138 | int err; |
| 139 | struct sk_filter *filter; |
| 140 | |
| 141 | err = security_sock_rcv_skb(sk, skb); |
| 142 | if (err) |
| 143 | return err; |
| 144 | |
| 145 | rcu_read_lock_bh(); |
| 146 | filter = rcu_dereference_bh(sk->sk_filter); |
| 147 | if (filter) { |
| 148 | unsigned int pkt_len = sk_run_filter(skb, filter->insns); |
| 149 | |
| 150 | err = pkt_len ? pskb_trim(skb, pkt_len) : -EPERM; |
| 151 | } |
| 152 | rcu_read_unlock_bh(); |
| 153 | |
| 154 | return err; |
| 155 | } |
| 156 | EXPORT_SYMBOL(sk_filter); |
| 157 | |
| 158 | /** |
| 159 | * sk_run_filter - run a filter on a socket |
| 160 | * @skb: buffer to run the filter on |
| 161 | * @filter: filter to apply |
| 162 | * |
| 163 | * Decode and apply filter instructions to the skb->data. |
| 164 | * Return length to keep, 0 for none. @skb is the data we are |
| 165 | * filtering, @filter is the array of filter instructions. |
| 166 | * Because all jumps are guaranteed to be before last instruction, |
| 167 | * and last instruction guaranteed to be a RET, we dont need to check |
| 168 | * flen. (We used to pass to this function the length of filter) |
| 169 | */ |
| 170 | unsigned int sk_run_filter(const struct sk_buff *skb, |
| 171 | const struct sock_filter *fentry) |
| 172 | { |
| 173 | void *ptr; |
| 174 | u32 A = 0; /* Accumulator */ |
| 175 | u32 X = 0; /* Index Register */ |
| 176 | u32 mem[BPF_MEMWORDS]; /* Scratch Memory Store */ |
| 177 | u32 tmp; |
| 178 | int k; |
| 179 | |
| 180 | /* |
| 181 | * Process array of filter instructions. |
| 182 | */ |
| 183 | for (;; fentry++) { |
| 184 | #if defined(CONFIG_X86_32) |
| 185 | #define K (fentry->k) |
| 186 | #else |
| 187 | const u32 K = fentry->k; |
| 188 | #endif |
| 189 | |
| 190 | switch (fentry->code) { |
| 191 | case BPF_S_ALU_ADD_X: |
| 192 | A += X; |
| 193 | continue; |
| 194 | case BPF_S_ALU_ADD_K: |
| 195 | A += K; |
| 196 | continue; |
| 197 | case BPF_S_ALU_SUB_X: |
| 198 | A -= X; |
| 199 | continue; |
| 200 | case BPF_S_ALU_SUB_K: |
| 201 | A -= K; |
| 202 | continue; |
| 203 | case BPF_S_ALU_MUL_X: |
| 204 | A *= X; |
| 205 | continue; |
| 206 | case BPF_S_ALU_MUL_K: |
| 207 | A *= K; |
| 208 | continue; |
| 209 | case BPF_S_ALU_DIV_X: |
| 210 | if (X == 0) |
| 211 | return 0; |
| 212 | A /= X; |
| 213 | continue; |
| 214 | case BPF_S_ALU_DIV_K: |
| 215 | A = reciprocal_divide(A, K); |
| 216 | continue; |
| 217 | case BPF_S_ALU_AND_X: |
| 218 | A &= X; |
| 219 | continue; |
| 220 | case BPF_S_ALU_AND_K: |
| 221 | A &= K; |
| 222 | continue; |
| 223 | case BPF_S_ALU_OR_X: |
| 224 | A |= X; |
| 225 | continue; |
| 226 | case BPF_S_ALU_OR_K: |
| 227 | A |= K; |
| 228 | continue; |
| 229 | case BPF_S_ALU_LSH_X: |
| 230 | A <<= X; |
| 231 | continue; |
| 232 | case BPF_S_ALU_LSH_K: |
| 233 | A <<= K; |
| 234 | continue; |
| 235 | case BPF_S_ALU_RSH_X: |
| 236 | A >>= X; |
| 237 | continue; |
| 238 | case BPF_S_ALU_RSH_K: |
| 239 | A >>= K; |
| 240 | continue; |
| 241 | case BPF_S_ALU_NEG: |
| 242 | A = -A; |
| 243 | continue; |
| 244 | case BPF_S_JMP_JA: |
| 245 | fentry += K; |
| 246 | continue; |
| 247 | case BPF_S_JMP_JGT_K: |
| 248 | fentry += (A > K) ? fentry->jt : fentry->jf; |
| 249 | continue; |
| 250 | case BPF_S_JMP_JGE_K: |
| 251 | fentry += (A >= K) ? fentry->jt : fentry->jf; |
| 252 | continue; |
| 253 | case BPF_S_JMP_JEQ_K: |
| 254 | fentry += (A == K) ? fentry->jt : fentry->jf; |
| 255 | continue; |
| 256 | case BPF_S_JMP_JSET_K: |
| 257 | fentry += (A & K) ? fentry->jt : fentry->jf; |
| 258 | continue; |
| 259 | case BPF_S_JMP_JGT_X: |
| 260 | fentry += (A > X) ? fentry->jt : fentry->jf; |
| 261 | continue; |
| 262 | case BPF_S_JMP_JGE_X: |
| 263 | fentry += (A >= X) ? fentry->jt : fentry->jf; |
| 264 | continue; |
| 265 | case BPF_S_JMP_JEQ_X: |
| 266 | fentry += (A == X) ? fentry->jt : fentry->jf; |
| 267 | continue; |
| 268 | case BPF_S_JMP_JSET_X: |
| 269 | fentry += (A & X) ? fentry->jt : fentry->jf; |
| 270 | continue; |
| 271 | case BPF_S_LD_W_ABS: |
| 272 | k = K; |
| 273 | load_w: |
| 274 | ptr = load_pointer(skb, k, 4, &tmp); |
| 275 | if (ptr != NULL) { |
| 276 | A = get_unaligned_be32(ptr); |
| 277 | continue; |
| 278 | } |
| 279 | return 0; |
| 280 | case BPF_S_LD_H_ABS: |
| 281 | k = K; |
| 282 | load_h: |
| 283 | ptr = load_pointer(skb, k, 2, &tmp); |
| 284 | if (ptr != NULL) { |
| 285 | A = get_unaligned_be16(ptr); |
| 286 | continue; |
| 287 | } |
| 288 | return 0; |
| 289 | case BPF_S_LD_B_ABS: |
| 290 | k = K; |
| 291 | load_b: |
| 292 | ptr = load_pointer(skb, k, 1, &tmp); |
| 293 | if (ptr != NULL) { |
| 294 | A = *(u8 *)ptr; |
| 295 | continue; |
| 296 | } |
| 297 | return 0; |
| 298 | case BPF_S_LD_W_LEN: |
| 299 | A = skb->len; |
| 300 | continue; |
| 301 | case BPF_S_LDX_W_LEN: |
| 302 | X = skb->len; |
| 303 | continue; |
| 304 | case BPF_S_LD_W_IND: |
| 305 | k = X + K; |
| 306 | goto load_w; |
| 307 | case BPF_S_LD_H_IND: |
| 308 | k = X + K; |
| 309 | goto load_h; |
| 310 | case BPF_S_LD_B_IND: |
| 311 | k = X + K; |
| 312 | goto load_b; |
| 313 | case BPF_S_LDX_B_MSH: |
| 314 | ptr = load_pointer(skb, K, 1, &tmp); |
| 315 | if (ptr != NULL) { |
| 316 | X = (*(u8 *)ptr & 0xf) << 2; |
| 317 | continue; |
| 318 | } |
| 319 | return 0; |
| 320 | case BPF_S_LD_IMM: |
| 321 | A = K; |
| 322 | continue; |
| 323 | case BPF_S_LDX_IMM: |
| 324 | X = K; |
| 325 | continue; |
| 326 | case BPF_S_LD_MEM: |
| 327 | A = mem[K]; |
| 328 | continue; |
| 329 | case BPF_S_LDX_MEM: |
| 330 | X = mem[K]; |
| 331 | continue; |
| 332 | case BPF_S_MISC_TAX: |
| 333 | X = A; |
| 334 | continue; |
| 335 | case BPF_S_MISC_TXA: |
| 336 | A = X; |
| 337 | continue; |
| 338 | case BPF_S_RET_K: |
| 339 | return K; |
| 340 | case BPF_S_RET_A: |
| 341 | return A; |
| 342 | case BPF_S_ST: |
| 343 | mem[K] = A; |
| 344 | continue; |
| 345 | case BPF_S_STX: |
| 346 | mem[K] = X; |
| 347 | continue; |
| 348 | case BPF_S_ANC_PROTOCOL: |
| 349 | A = ntohs(skb->protocol); |
| 350 | continue; |
| 351 | case BPF_S_ANC_PKTTYPE: |
| 352 | A = skb->pkt_type; |
| 353 | continue; |
| 354 | case BPF_S_ANC_IFINDEX: |
| 355 | if (!skb->dev) |
| 356 | return 0; |
| 357 | A = skb->dev->ifindex; |
| 358 | continue; |
| 359 | case BPF_S_ANC_MARK: |
| 360 | A = skb->mark; |
| 361 | continue; |
| 362 | case BPF_S_ANC_QUEUE: |
| 363 | A = skb->queue_mapping; |
| 364 | continue; |
| 365 | case BPF_S_ANC_HATYPE: |
| 366 | if (!skb->dev) |
| 367 | return 0; |
| 368 | A = skb->dev->type; |
| 369 | continue; |
| 370 | case BPF_S_ANC_RXHASH: |
| 371 | A = skb->rxhash; |
| 372 | continue; |
| 373 | case BPF_S_ANC_CPU: |
| 374 | A = raw_smp_processor_id(); |
| 375 | continue; |
| 376 | case BPF_S_ANC_NLATTR: { |
| 377 | struct nlattr *nla; |
| 378 | |
| 379 | if (skb_is_nonlinear(skb)) |
| 380 | return 0; |
| 381 | if (A > skb->len - sizeof(struct nlattr)) |
| 382 | return 0; |
| 383 | |
| 384 | nla = nla_find((struct nlattr *)&skb->data[A], |
| 385 | skb->len - A, X); |
| 386 | if (nla) |
| 387 | A = (void *)nla - (void *)skb->data; |
| 388 | else |
| 389 | A = 0; |
| 390 | continue; |
| 391 | } |
| 392 | case BPF_S_ANC_NLATTR_NEST: { |
| 393 | struct nlattr *nla; |
| 394 | |
| 395 | if (skb_is_nonlinear(skb)) |
| 396 | return 0; |
| 397 | if (A > skb->len - sizeof(struct nlattr)) |
| 398 | return 0; |
| 399 | |
| 400 | nla = (struct nlattr *)&skb->data[A]; |
| 401 | if (nla->nla_len > A - skb->len) |
| 402 | return 0; |
| 403 | |
| 404 | nla = nla_find_nested(nla, X); |
| 405 | if (nla) |
| 406 | A = (void *)nla - (void *)skb->data; |
| 407 | else |
| 408 | A = 0; |
| 409 | continue; |
| 410 | } |
| 411 | default: |
| 412 | WARN_ON(1); |
| 413 | return 0; |
| 414 | } |
| 415 | } |
| 416 | |
| 417 | return 0; |
| 418 | } |
| 419 | EXPORT_SYMBOL(sk_run_filter); |
| 420 | |
| 421 | /* |
| 422 | * Security : |
| 423 | * A BPF program is able to use 16 cells of memory to store intermediate |
| 424 | * values (check u32 mem[BPF_MEMWORDS] in sk_run_filter()) |
| 425 | * As we dont want to clear mem[] array for each packet going through |
| 426 | * sk_run_filter(), we check that filter loaded by user never try to read |
| 427 | * a cell if not previously written, and we check all branches to be sure |
| 428 | * a malicious user doesnt try to abuse us. |
| 429 | */ |
| 430 | static int check_load_and_stores(struct sock_filter *filter, int flen) |
| 431 | { |
| 432 | u16 *masks, memvalid = 0; /* one bit per cell, 16 cells */ |
| 433 | int pc, ret = 0; |
| 434 | |
| 435 | BUILD_BUG_ON(BPF_MEMWORDS > 16); |
| 436 | masks = kmalloc(flen * sizeof(*masks), GFP_KERNEL); |
| 437 | if (!masks) |
| 438 | return -ENOMEM; |
| 439 | memset(masks, 0xff, flen * sizeof(*masks)); |
| 440 | |
| 441 | for (pc = 0; pc < flen; pc++) { |
| 442 | memvalid &= masks[pc]; |
| 443 | |
| 444 | switch (filter[pc].code) { |
| 445 | case BPF_S_ST: |
| 446 | case BPF_S_STX: |
| 447 | memvalid |= (1 << filter[pc].k); |
| 448 | break; |
| 449 | case BPF_S_LD_MEM: |
| 450 | case BPF_S_LDX_MEM: |
| 451 | if (!(memvalid & (1 << filter[pc].k))) { |
| 452 | ret = -EINVAL; |
| 453 | goto error; |
| 454 | } |
| 455 | break; |
| 456 | case BPF_S_JMP_JA: |
| 457 | /* a jump must set masks on target */ |
| 458 | masks[pc + 1 + filter[pc].k] &= memvalid; |
| 459 | memvalid = ~0; |
| 460 | break; |
| 461 | case BPF_S_JMP_JEQ_K: |
| 462 | case BPF_S_JMP_JEQ_X: |
| 463 | case BPF_S_JMP_JGE_K: |
| 464 | case BPF_S_JMP_JGE_X: |
| 465 | case BPF_S_JMP_JGT_K: |
| 466 | case BPF_S_JMP_JGT_X: |
| 467 | case BPF_S_JMP_JSET_X: |
| 468 | case BPF_S_JMP_JSET_K: |
| 469 | /* a jump must set masks on targets */ |
| 470 | masks[pc + 1 + filter[pc].jt] &= memvalid; |
| 471 | masks[pc + 1 + filter[pc].jf] &= memvalid; |
| 472 | memvalid = ~0; |
| 473 | break; |
| 474 | } |
| 475 | } |
| 476 | error: |
| 477 | kfree(masks); |
| 478 | return ret; |
| 479 | } |
| 480 | |
| 481 | /** |
| 482 | * sk_chk_filter - verify socket filter code |
| 483 | * @filter: filter to verify |
| 484 | * @flen: length of filter |
| 485 | * |
| 486 | * Check the user's filter code. If we let some ugly |
| 487 | * filter code slip through kaboom! The filter must contain |
| 488 | * no references or jumps that are out of range, no illegal |
| 489 | * instructions, and must end with a RET instruction. |
| 490 | * |
| 491 | * All jumps are forward as they are not signed. |
| 492 | * |
| 493 | * Returns 0 if the rule set is legal or -EINVAL if not. |
| 494 | */ |
| 495 | int sk_chk_filter(struct sock_filter *filter, int flen) |
| 496 | { |
| 497 | /* |
| 498 | * Valid instructions are initialized to non-0. |
| 499 | * Invalid instructions are initialized to 0. |
| 500 | */ |
| 501 | static const u8 codes[] = { |
| 502 | [BPF_ALU|BPF_ADD|BPF_K] = BPF_S_ALU_ADD_K, |
| 503 | [BPF_ALU|BPF_ADD|BPF_X] = BPF_S_ALU_ADD_X, |
| 504 | [BPF_ALU|BPF_SUB|BPF_K] = BPF_S_ALU_SUB_K, |
| 505 | [BPF_ALU|BPF_SUB|BPF_X] = BPF_S_ALU_SUB_X, |
| 506 | [BPF_ALU|BPF_MUL|BPF_K] = BPF_S_ALU_MUL_K, |
| 507 | [BPF_ALU|BPF_MUL|BPF_X] = BPF_S_ALU_MUL_X, |
| 508 | [BPF_ALU|BPF_DIV|BPF_X] = BPF_S_ALU_DIV_X, |
| 509 | [BPF_ALU|BPF_AND|BPF_K] = BPF_S_ALU_AND_K, |
| 510 | [BPF_ALU|BPF_AND|BPF_X] = BPF_S_ALU_AND_X, |
| 511 | [BPF_ALU|BPF_OR|BPF_K] = BPF_S_ALU_OR_K, |
| 512 | [BPF_ALU|BPF_OR|BPF_X] = BPF_S_ALU_OR_X, |
| 513 | [BPF_ALU|BPF_LSH|BPF_K] = BPF_S_ALU_LSH_K, |
| 514 | [BPF_ALU|BPF_LSH|BPF_X] = BPF_S_ALU_LSH_X, |
| 515 | [BPF_ALU|BPF_RSH|BPF_K] = BPF_S_ALU_RSH_K, |
| 516 | [BPF_ALU|BPF_RSH|BPF_X] = BPF_S_ALU_RSH_X, |
| 517 | [BPF_ALU|BPF_NEG] = BPF_S_ALU_NEG, |
| 518 | [BPF_LD|BPF_W|BPF_ABS] = BPF_S_LD_W_ABS, |
| 519 | [BPF_LD|BPF_H|BPF_ABS] = BPF_S_LD_H_ABS, |
| 520 | [BPF_LD|BPF_B|BPF_ABS] = BPF_S_LD_B_ABS, |
| 521 | [BPF_LD|BPF_W|BPF_LEN] = BPF_S_LD_W_LEN, |
| 522 | [BPF_LD|BPF_W|BPF_IND] = BPF_S_LD_W_IND, |
| 523 | [BPF_LD|BPF_H|BPF_IND] = BPF_S_LD_H_IND, |
| 524 | [BPF_LD|BPF_B|BPF_IND] = BPF_S_LD_B_IND, |
| 525 | [BPF_LD|BPF_IMM] = BPF_S_LD_IMM, |
| 526 | [BPF_LDX|BPF_W|BPF_LEN] = BPF_S_LDX_W_LEN, |
| 527 | [BPF_LDX|BPF_B|BPF_MSH] = BPF_S_LDX_B_MSH, |
| 528 | [BPF_LDX|BPF_IMM] = BPF_S_LDX_IMM, |
| 529 | [BPF_MISC|BPF_TAX] = BPF_S_MISC_TAX, |
| 530 | [BPF_MISC|BPF_TXA] = BPF_S_MISC_TXA, |
| 531 | [BPF_RET|BPF_K] = BPF_S_RET_K, |
| 532 | [BPF_RET|BPF_A] = BPF_S_RET_A, |
| 533 | [BPF_ALU|BPF_DIV|BPF_K] = BPF_S_ALU_DIV_K, |
| 534 | [BPF_LD|BPF_MEM] = BPF_S_LD_MEM, |
| 535 | [BPF_LDX|BPF_MEM] = BPF_S_LDX_MEM, |
| 536 | [BPF_ST] = BPF_S_ST, |
| 537 | [BPF_STX] = BPF_S_STX, |
| 538 | [BPF_JMP|BPF_JA] = BPF_S_JMP_JA, |
| 539 | [BPF_JMP|BPF_JEQ|BPF_K] = BPF_S_JMP_JEQ_K, |
| 540 | [BPF_JMP|BPF_JEQ|BPF_X] = BPF_S_JMP_JEQ_X, |
| 541 | [BPF_JMP|BPF_JGE|BPF_K] = BPF_S_JMP_JGE_K, |
| 542 | [BPF_JMP|BPF_JGE|BPF_X] = BPF_S_JMP_JGE_X, |
| 543 | [BPF_JMP|BPF_JGT|BPF_K] = BPF_S_JMP_JGT_K, |
| 544 | [BPF_JMP|BPF_JGT|BPF_X] = BPF_S_JMP_JGT_X, |
| 545 | [BPF_JMP|BPF_JSET|BPF_K] = BPF_S_JMP_JSET_K, |
| 546 | [BPF_JMP|BPF_JSET|BPF_X] = BPF_S_JMP_JSET_X, |
| 547 | }; |
| 548 | int pc; |
| 549 | |
| 550 | if (flen == 0 || flen > BPF_MAXINSNS) |
| 551 | return -EINVAL; |
| 552 | |
| 553 | /* check the filter code now */ |
| 554 | for (pc = 0; pc < flen; pc++) { |
| 555 | struct sock_filter *ftest = &filter[pc]; |
| 556 | u16 code = ftest->code; |
| 557 | |
| 558 | if (code >= ARRAY_SIZE(codes)) |
| 559 | return -EINVAL; |
| 560 | code = codes[code]; |
| 561 | if (!code) |
| 562 | return -EINVAL; |
| 563 | /* Some instructions need special checks */ |
| 564 | switch (code) { |
| 565 | case BPF_S_ALU_DIV_K: |
| 566 | /* check for division by zero */ |
| 567 | if (ftest->k == 0) |
| 568 | return -EINVAL; |
| 569 | ftest->k = reciprocal_value(ftest->k); |
| 570 | break; |
| 571 | case BPF_S_LD_MEM: |
| 572 | case BPF_S_LDX_MEM: |
| 573 | case BPF_S_ST: |
| 574 | case BPF_S_STX: |
| 575 | /* check for invalid memory addresses */ |
| 576 | if (ftest->k >= BPF_MEMWORDS) |
| 577 | return -EINVAL; |
| 578 | break; |
| 579 | case BPF_S_JMP_JA: |
| 580 | /* |
| 581 | * Note, the large ftest->k might cause loops. |
| 582 | * Compare this with conditional jumps below, |
| 583 | * where offsets are limited. --ANK (981016) |
| 584 | */ |
| 585 | if (ftest->k >= (unsigned)(flen-pc-1)) |
| 586 | return -EINVAL; |
| 587 | break; |
| 588 | case BPF_S_JMP_JEQ_K: |
| 589 | case BPF_S_JMP_JEQ_X: |
| 590 | case BPF_S_JMP_JGE_K: |
| 591 | case BPF_S_JMP_JGE_X: |
| 592 | case BPF_S_JMP_JGT_K: |
| 593 | case BPF_S_JMP_JGT_X: |
| 594 | case BPF_S_JMP_JSET_X: |
| 595 | case BPF_S_JMP_JSET_K: |
| 596 | /* for conditionals both must be safe */ |
| 597 | if (pc + ftest->jt + 1 >= flen || |
| 598 | pc + ftest->jf + 1 >= flen) |
| 599 | return -EINVAL; |
| 600 | break; |
| 601 | case BPF_S_LD_W_ABS: |
| 602 | case BPF_S_LD_H_ABS: |
| 603 | case BPF_S_LD_B_ABS: |
| 604 | #define ANCILLARY(CODE) case SKF_AD_OFF + SKF_AD_##CODE: \ |
| 605 | code = BPF_S_ANC_##CODE; \ |
| 606 | break |
| 607 | switch (ftest->k) { |
| 608 | ANCILLARY(PROTOCOL); |
| 609 | ANCILLARY(PKTTYPE); |
| 610 | ANCILLARY(IFINDEX); |
| 611 | ANCILLARY(NLATTR); |
| 612 | ANCILLARY(NLATTR_NEST); |
| 613 | ANCILLARY(MARK); |
| 614 | ANCILLARY(QUEUE); |
| 615 | ANCILLARY(HATYPE); |
| 616 | ANCILLARY(RXHASH); |
| 617 | ANCILLARY(CPU); |
| 618 | } |
| 619 | } |
| 620 | ftest->code = code; |
| 621 | } |
| 622 | |
| 623 | /* last instruction must be a RET code */ |
| 624 | switch (filter[flen - 1].code) { |
| 625 | case BPF_S_RET_K: |
| 626 | case BPF_S_RET_A: |
| 627 | return check_load_and_stores(filter, flen); |
| 628 | } |
| 629 | return -EINVAL; |
| 630 | } |
| 631 | EXPORT_SYMBOL(sk_chk_filter); |
| 632 | |
| 633 | /** |
| 634 | * sk_filter_release_rcu - Release a socket filter by rcu_head |
| 635 | * @rcu: rcu_head that contains the sk_filter to free |
| 636 | */ |
| 637 | void sk_filter_release_rcu(struct rcu_head *rcu) |
| 638 | { |
| 639 | struct sk_filter *fp = container_of(rcu, struct sk_filter, rcu); |
| 640 | |
| 641 | kfree(fp); |
| 642 | } |
| 643 | EXPORT_SYMBOL(sk_filter_release_rcu); |
| 644 | |
| 645 | /** |
| 646 | * sk_attach_filter - attach a socket filter |
| 647 | * @fprog: the filter program |
| 648 | * @sk: the socket to use |
| 649 | * |
| 650 | * Attach the user's filter code. We first run some sanity checks on |
| 651 | * it to make sure it does not explode on us later. If an error |
| 652 | * occurs or there is insufficient memory for the filter a negative |
| 653 | * errno code is returned. On success the return is zero. |
| 654 | */ |
| 655 | int sk_attach_filter(struct sock_fprog *fprog, struct sock *sk) |
| 656 | { |
| 657 | struct sk_filter *fp, *old_fp; |
| 658 | unsigned int fsize = sizeof(struct sock_filter) * fprog->len; |
| 659 | int err; |
| 660 | |
| 661 | /* Make sure new filter is there and in the right amounts. */ |
| 662 | if (fprog->filter == NULL) |
| 663 | return -EINVAL; |
| 664 | |
| 665 | fp = sock_kmalloc(sk, fsize+sizeof(*fp), GFP_KERNEL); |
| 666 | if (!fp) |
| 667 | return -ENOMEM; |
| 668 | if (copy_from_user(fp->insns, fprog->filter, fsize)) { |
| 669 | sock_kfree_s(sk, fp, fsize+sizeof(*fp)); |
| 670 | return -EFAULT; |
| 671 | } |
| 672 | |
| 673 | atomic_set(&fp->refcnt, 1); |
| 674 | fp->len = fprog->len; |
| 675 | |
| 676 | err = sk_chk_filter(fp->insns, fp->len); |
| 677 | if (err) { |
| 678 | sk_filter_uncharge(sk, fp); |
| 679 | return err; |
| 680 | } |
| 681 | |
| 682 | old_fp = rcu_dereference_protected(sk->sk_filter, |
| 683 | sock_owned_by_user(sk)); |
| 684 | rcu_assign_pointer(sk->sk_filter, fp); |
| 685 | |
| 686 | if (old_fp) |
| 687 | sk_filter_uncharge(sk, old_fp); |
| 688 | return 0; |
| 689 | } |
| 690 | EXPORT_SYMBOL_GPL(sk_attach_filter); |
| 691 | |
| 692 | int sk_detach_filter(struct sock *sk) |
| 693 | { |
| 694 | int ret = -ENOENT; |
| 695 | struct sk_filter *filter; |
| 696 | |
| 697 | filter = rcu_dereference_protected(sk->sk_filter, |
| 698 | sock_owned_by_user(sk)); |
| 699 | if (filter) { |
| 700 | rcu_assign_pointer(sk->sk_filter, NULL); |
| 701 | sk_filter_uncharge(sk, filter); |
| 702 | ret = 0; |
| 703 | } |
| 704 | return ret; |
| 705 | } |
| 706 | EXPORT_SYMBOL_GPL(sk_detach_filter); |