1 // SPDX-License-Identifier: GPL-2.0-only 2 /* 3 * Packet matching code. 4 * 5 * Copyright (C) 1999 Paul `Rusty' Russell & Michael J. Neuling 6 * Copyright (C) 2000-2005 Netfilter Core Team <coreteam@netfilter.org> 7 * Copyright (C) 2006-2010 Patrick McHardy <kaber@trash.net> 8 */ 9 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 10 #include <linux/cache.h> 11 #include <linux/capability.h> 12 #include <linux/skbuff.h> 13 #include <linux/kmod.h> 14 #include <linux/vmalloc.h> 15 #include <linux/netdevice.h> 16 #include <linux/module.h> 17 #include <linux/icmp.h> 18 #include <net/ip.h> 19 #include <net/compat.h> 20 #include <linux/uaccess.h> 21 #include <linux/mutex.h> 22 #include <linux/proc_fs.h> 23 #include <linux/err.h> 24 #include <linux/cpumask.h> 25 26 #include <linux/netfilter/x_tables.h> 27 #include <linux/netfilter_ipv4/ip_tables.h> 28 #include <net/netfilter/nf_log.h> 29 #include "../../netfilter/xt_repldata.h" 30 31 MODULE_LICENSE("GPL"); 32 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>"); 33 MODULE_DESCRIPTION("IPv4 packet filter"); 34 MODULE_ALIAS("ipt_icmp"); 35 36 void *ipt_alloc_initial_table(const struct xt_table *info) 37 { 38 return xt_alloc_initial_table(ipt, IPT); 39 } 40 EXPORT_SYMBOL_GPL(ipt_alloc_initial_table); 41 42 /* Returns whether matches rule or not. */ 43 /* Performance critical - called for every packet */ 44 static inline bool 45 ip_packet_match(const struct iphdr *ip, 46 const char *indev, 47 const char *outdev, 48 const struct ipt_ip *ipinfo, 49 int isfrag) 50 { 51 unsigned long ret; 52 53 if (NF_INVF(ipinfo, IPT_INV_SRCIP, 54 (ip->saddr & ipinfo->smsk.s_addr) != ipinfo->src.s_addr) || 55 NF_INVF(ipinfo, IPT_INV_DSTIP, 56 (ip->daddr & ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr)) 57 return false; 58 59 ret = ifname_compare_aligned(indev, ipinfo->iniface, ipinfo->iniface_mask); 60 61 if (NF_INVF(ipinfo, IPT_INV_VIA_IN, ret != 0)) 62 return false; 63 64 ret = ifname_compare_aligned(outdev, ipinfo->outiface, ipinfo->outiface_mask); 65 66 if (NF_INVF(ipinfo, IPT_INV_VIA_OUT, ret != 0)) 67 return false; 68 69 /* Check specific protocol */ 70 if (ipinfo->proto && 71 NF_INVF(ipinfo, IPT_INV_PROTO, ip->protocol != ipinfo->proto)) 72 return false; 73 74 /* If we have a fragment rule but the packet is not a fragment 75 * then we return zero */ 76 if (NF_INVF(ipinfo, IPT_INV_FRAG, 77 (ipinfo->flags & IPT_F_FRAG) && !isfrag)) 78 return false; 79 80 return true; 81 } 82 83 static bool 84 ip_checkentry(const struct ipt_ip *ip) 85 { 86 if (ip->flags & ~IPT_F_MASK) 87 return false; 88 if (ip->invflags & ~IPT_INV_MASK) 89 return false; 90 return true; 91 } 92 93 static unsigned int 94 ipt_error(struct sk_buff *skb, const struct xt_action_param *par) 95 { 96 net_info_ratelimited("error: `%s'\n", (const char *)par->targinfo); 97 98 return NF_DROP; 99 } 100 101 /* Performance critical */ 102 static inline struct ipt_entry * 103 get_entry(const void *base, unsigned int offset) 104 { 105 return (struct ipt_entry *)(base + offset); 106 } 107 108 /* All zeroes == unconditional rule. */ 109 /* Mildly perf critical (only if packet tracing is on) */ 110 static inline bool unconditional(const struct ipt_entry *e) 111 { 112 static const struct ipt_ip uncond; 113 114 return e->target_offset == sizeof(struct ipt_entry) && 115 memcmp(&e->ip, &uncond, sizeof(uncond)) == 0; 116 } 117 118 /* for const-correctness */ 119 static inline const struct xt_entry_target * 120 ipt_get_target_c(const struct ipt_entry *e) 121 { 122 return ipt_get_target((struct ipt_entry *)e); 123 } 124 125 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) 126 static const char *const hooknames[] = { 127 [NF_INET_PRE_ROUTING] = "PREROUTING", 128 [NF_INET_LOCAL_IN] = "INPUT", 129 [NF_INET_FORWARD] = "FORWARD", 130 [NF_INET_LOCAL_OUT] = "OUTPUT", 131 [NF_INET_POST_ROUTING] = "POSTROUTING", 132 }; 133 134 enum nf_ip_trace_comments { 135 NF_IP_TRACE_COMMENT_RULE, 136 NF_IP_TRACE_COMMENT_RETURN, 137 NF_IP_TRACE_COMMENT_POLICY, 138 }; 139 140 static const char *const comments[] = { 141 [NF_IP_TRACE_COMMENT_RULE] = "rule", 142 [NF_IP_TRACE_COMMENT_RETURN] = "return", 143 [NF_IP_TRACE_COMMENT_POLICY] = "policy", 144 }; 145 146 static const struct nf_loginfo trace_loginfo = { 147 .type = NF_LOG_TYPE_LOG, 148 .u = { 149 .log = { 150 .level = 4, 151 .logflags = NF_LOG_DEFAULT_MASK, 152 }, 153 }, 154 }; 155 156 /* Mildly perf critical (only if packet tracing is on) */ 157 static inline int 158 get_chainname_rulenum(const struct ipt_entry *s, const struct ipt_entry *e, 159 const char *hookname, const char **chainname, 160 const char **comment, unsigned int *rulenum) 161 { 162 const struct xt_standard_target *t = (void *)ipt_get_target_c(s); 163 164 if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) { 165 /* Head of user chain: ERROR target with chainname */ 166 *chainname = t->target.data; 167 (*rulenum) = 0; 168 } else if (s == e) { 169 (*rulenum)++; 170 171 if (unconditional(s) && 172 strcmp(t->target.u.kernel.target->name, 173 XT_STANDARD_TARGET) == 0 && 174 t->verdict < 0) { 175 /* Tail of chains: STANDARD target (return/policy) */ 176 *comment = *chainname == hookname 177 ? comments[NF_IP_TRACE_COMMENT_POLICY] 178 : comments[NF_IP_TRACE_COMMENT_RETURN]; 179 } 180 return 1; 181 } else 182 (*rulenum)++; 183 184 return 0; 185 } 186 187 static void trace_packet(struct net *net, 188 const struct sk_buff *skb, 189 unsigned int hook, 190 const struct net_device *in, 191 const struct net_device *out, 192 const char *tablename, 193 const struct xt_table_info *private, 194 const struct ipt_entry *e) 195 { 196 const struct ipt_entry *root; 197 const char *hookname, *chainname, *comment; 198 const struct ipt_entry *iter; 199 unsigned int rulenum = 0; 200 201 root = get_entry(private->entries, private->hook_entry[hook]); 202 203 hookname = chainname = hooknames[hook]; 204 comment = comments[NF_IP_TRACE_COMMENT_RULE]; 205 206 xt_entry_foreach(iter, root, private->size - private->hook_entry[hook]) 207 if (get_chainname_rulenum(iter, e, hookname, 208 &chainname, &comment, &rulenum) != 0) 209 break; 210 211 nf_log_trace(net, AF_INET, hook, skb, in, out, &trace_loginfo, 212 "TRACE: %s:%s:%s:%u ", 213 tablename, chainname, comment, rulenum); 214 } 215 #endif 216 217 static inline 218 struct ipt_entry *ipt_next_entry(const struct ipt_entry *entry) 219 { 220 return (void *)entry + entry->next_offset; 221 } 222 223 /* Returns one of the generic firewall policies, like NF_ACCEPT. */ 224 unsigned int 225 ipt_do_table(struct sk_buff *skb, 226 const struct nf_hook_state *state, 227 struct xt_table *table) 228 { 229 unsigned int hook = state->hook; 230 static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long)))); 231 const struct iphdr *ip; 232 /* Initializing verdict to NF_DROP keeps gcc happy. */ 233 unsigned int verdict = NF_DROP; 234 const char *indev, *outdev; 235 const void *table_base; 236 struct ipt_entry *e, **jumpstack; 237 unsigned int stackidx, cpu; 238 const struct xt_table_info *private; 239 struct xt_action_param acpar; 240 unsigned int addend; 241 242 /* Initialization */ 243 stackidx = 0; 244 ip = ip_hdr(skb); 245 indev = state->in ? state->in->name : nulldevname; 246 outdev = state->out ? state->out->name : nulldevname; 247 /* We handle fragments by dealing with the first fragment as 248 * if it was a normal packet. All other fragments are treated 249 * normally, except that they will NEVER match rules that ask 250 * things we don't know, ie. tcp syn flag or ports). If the 251 * rule is also a fragment-specific rule, non-fragments won't 252 * match it. */ 253 acpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET; 254 acpar.thoff = ip_hdrlen(skb); 255 acpar.hotdrop = false; 256 acpar.state = state; 257 258 WARN_ON(!(table->valid_hooks & (1 << hook))); 259 local_bh_disable(); 260 addend = xt_write_recseq_begin(); 261 private = rcu_access_pointer(table->private); 262 cpu = smp_processor_id(); 263 table_base = private->entries; 264 jumpstack = (struct ipt_entry **)private->jumpstack[cpu]; 265 266 /* Switch to alternate jumpstack if we're being invoked via TEE. 267 * TEE issues XT_CONTINUE verdict on original skb so we must not 268 * clobber the jumpstack. 269 * 270 * For recursion via REJECT or SYNPROXY the stack will be clobbered 271 * but it is no problem since absolute verdict is issued by these. 272 */ 273 if (static_key_false(&xt_tee_enabled)) 274 jumpstack += private->stacksize * __this_cpu_read(nf_skb_duplicated); 275 276 e = get_entry(table_base, private->hook_entry[hook]); 277 278 do { 279 const struct xt_entry_target *t; 280 const struct xt_entry_match *ematch; 281 struct xt_counters *counter; 282 283 WARN_ON(!e); 284 if (!ip_packet_match(ip, indev, outdev, 285 &e->ip, acpar.fragoff)) { 286 no_match: 287 e = ipt_next_entry(e); 288 continue; 289 } 290 291 xt_ematch_foreach(ematch, e) { 292 acpar.match = ematch->u.kernel.match; 293 acpar.matchinfo = ematch->data; 294 if (!acpar.match->match(skb, &acpar)) 295 goto no_match; 296 } 297 298 counter = xt_get_this_cpu_counter(&e->counters); 299 ADD_COUNTER(*counter, skb->len, 1); 300 301 t = ipt_get_target_c(e); 302 WARN_ON(!t->u.kernel.target); 303 304 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) 305 /* The packet is traced: log it */ 306 if (unlikely(skb->nf_trace)) 307 trace_packet(state->net, skb, hook, state->in, 308 state->out, table->name, private, e); 309 #endif 310 /* Standard target? */ 311 if (!t->u.kernel.target->target) { 312 int v; 313 314 v = ((struct xt_standard_target *)t)->verdict; 315 if (v < 0) { 316 /* Pop from stack? */ 317 if (v != XT_RETURN) { 318 verdict = (unsigned int)(-v) - 1; 319 break; 320 } 321 if (stackidx == 0) { 322 e = get_entry(table_base, 323 private->underflow[hook]); 324 } else { 325 e = jumpstack[--stackidx]; 326 e = ipt_next_entry(e); 327 } 328 continue; 329 } 330 if (table_base + v != ipt_next_entry(e) && 331 !(e->ip.flags & IPT_F_GOTO)) { 332 if (unlikely(stackidx >= private->stacksize)) { 333 verdict = NF_DROP; 334 break; 335 } 336 jumpstack[stackidx++] = e; 337 } 338 339 e = get_entry(table_base, v); 340 continue; 341 } 342 343 acpar.target = t->u.kernel.target; 344 acpar.targinfo = t->data; 345 346 verdict = t->u.kernel.target->target(skb, &acpar); 347 if (verdict == XT_CONTINUE) { 348 /* Target might have changed stuff. */ 349 ip = ip_hdr(skb); 350 e = ipt_next_entry(e); 351 } else { 352 /* Verdict */ 353 break; 354 } 355 } while (!acpar.hotdrop); 356 357 xt_write_recseq_end(addend); 358 local_bh_enable(); 359 360 if (acpar.hotdrop) 361 return NF_DROP; 362 else return verdict; 363 } 364 365 /* Figures out from what hook each rule can be called: returns 0 if 366 there are loops. Puts hook bitmask in comefrom. */ 367 static int 368 mark_source_chains(const struct xt_table_info *newinfo, 369 unsigned int valid_hooks, void *entry0, 370 unsigned int *offsets) 371 { 372 unsigned int hook; 373 374 /* No recursion; use packet counter to save back ptrs (reset 375 to 0 as we leave), and comefrom to save source hook bitmask */ 376 for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) { 377 unsigned int pos = newinfo->hook_entry[hook]; 378 struct ipt_entry *e = entry0 + pos; 379 380 if (!(valid_hooks & (1 << hook))) 381 continue; 382 383 /* Set initial back pointer. */ 384 e->counters.pcnt = pos; 385 386 for (;;) { 387 const struct xt_standard_target *t 388 = (void *)ipt_get_target_c(e); 389 int visited = e->comefrom & (1 << hook); 390 391 if (e->comefrom & (1 << NF_INET_NUMHOOKS)) 392 return 0; 393 394 e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS)); 395 396 /* Unconditional return/END. */ 397 if ((unconditional(e) && 398 (strcmp(t->target.u.user.name, 399 XT_STANDARD_TARGET) == 0) && 400 t->verdict < 0) || visited) { 401 unsigned int oldpos, size; 402 403 /* Return: backtrack through the last 404 big jump. */ 405 do { 406 e->comefrom ^= (1<<NF_INET_NUMHOOKS); 407 oldpos = pos; 408 pos = e->counters.pcnt; 409 e->counters.pcnt = 0; 410 411 /* We're at the start. */ 412 if (pos == oldpos) 413 goto next; 414 415 e = entry0 + pos; 416 } while (oldpos == pos + e->next_offset); 417 418 /* Move along one */ 419 size = e->next_offset; 420 e = entry0 + pos + size; 421 if (pos + size >= newinfo->size) 422 return 0; 423 e->counters.pcnt = pos; 424 pos += size; 425 } else { 426 int newpos = t->verdict; 427 428 if (strcmp(t->target.u.user.name, 429 XT_STANDARD_TARGET) == 0 && 430 newpos >= 0) { 431 /* This a jump; chase it. */ 432 if (!xt_find_jump_offset(offsets, newpos, 433 newinfo->number)) 434 return 0; 435 } else { 436 /* ... this is a fallthru */ 437 newpos = pos + e->next_offset; 438 if (newpos >= newinfo->size) 439 return 0; 440 } 441 e = entry0 + newpos; 442 e->counters.pcnt = pos; 443 pos = newpos; 444 } 445 } 446 next: ; 447 } 448 return 1; 449 } 450 451 static void cleanup_match(struct xt_entry_match *m, struct net *net) 452 { 453 struct xt_mtdtor_param par; 454 455 par.net = net; 456 par.match = m->u.kernel.match; 457 par.matchinfo = m->data; 458 par.family = NFPROTO_IPV4; 459 if (par.match->destroy != NULL) 460 par.match->destroy(&par); 461 module_put(par.match->me); 462 } 463 464 static int 465 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par) 466 { 467 const struct ipt_ip *ip = par->entryinfo; 468 469 par->match = m->u.kernel.match; 470 par->matchinfo = m->data; 471 472 return xt_check_match(par, m->u.match_size - sizeof(*m), 473 ip->proto, ip->invflags & IPT_INV_PROTO); 474 } 475 476 static int 477 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par) 478 { 479 struct xt_match *match; 480 int ret; 481 482 match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name, 483 m->u.user.revision); 484 if (IS_ERR(match)) 485 return PTR_ERR(match); 486 m->u.kernel.match = match; 487 488 ret = check_match(m, par); 489 if (ret) 490 goto err; 491 492 return 0; 493 err: 494 module_put(m->u.kernel.match->me); 495 return ret; 496 } 497 498 static int check_target(struct ipt_entry *e, struct net *net, const char *name) 499 { 500 struct xt_entry_target *t = ipt_get_target(e); 501 struct xt_tgchk_param par = { 502 .net = net, 503 .table = name, 504 .entryinfo = e, 505 .target = t->u.kernel.target, 506 .targinfo = t->data, 507 .hook_mask = e->comefrom, 508 .family = NFPROTO_IPV4, 509 }; 510 511 return xt_check_target(&par, t->u.target_size - sizeof(*t), 512 e->ip.proto, e->ip.invflags & IPT_INV_PROTO); 513 } 514 515 static int 516 find_check_entry(struct ipt_entry *e, struct net *net, const char *name, 517 unsigned int size, 518 struct xt_percpu_counter_alloc_state *alloc_state) 519 { 520 struct xt_entry_target *t; 521 struct xt_target *target; 522 int ret; 523 unsigned int j; 524 struct xt_mtchk_param mtpar; 525 struct xt_entry_match *ematch; 526 527 if (!xt_percpu_counter_alloc(alloc_state, &e->counters)) 528 return -ENOMEM; 529 530 j = 0; 531 memset(&mtpar, 0, sizeof(mtpar)); 532 mtpar.net = net; 533 mtpar.table = name; 534 mtpar.entryinfo = &e->ip; 535 mtpar.hook_mask = e->comefrom; 536 mtpar.family = NFPROTO_IPV4; 537 xt_ematch_foreach(ematch, e) { 538 ret = find_check_match(ematch, &mtpar); 539 if (ret != 0) 540 goto cleanup_matches; 541 ++j; 542 } 543 544 t = ipt_get_target(e); 545 target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name, 546 t->u.user.revision); 547 if (IS_ERR(target)) { 548 ret = PTR_ERR(target); 549 goto cleanup_matches; 550 } 551 t->u.kernel.target = target; 552 553 ret = check_target(e, net, name); 554 if (ret) 555 goto err; 556 557 return 0; 558 err: 559 module_put(t->u.kernel.target->me); 560 cleanup_matches: 561 xt_ematch_foreach(ematch, e) { 562 if (j-- == 0) 563 break; 564 cleanup_match(ematch, net); 565 } 566 567 xt_percpu_counter_free(&e->counters); 568 569 return ret; 570 } 571 572 static bool check_underflow(const struct ipt_entry *e) 573 { 574 const struct xt_entry_target *t; 575 unsigned int verdict; 576 577 if (!unconditional(e)) 578 return false; 579 t = ipt_get_target_c(e); 580 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0) 581 return false; 582 verdict = ((struct xt_standard_target *)t)->verdict; 583 verdict = -verdict - 1; 584 return verdict == NF_DROP || verdict == NF_ACCEPT; 585 } 586 587 static int 588 check_entry_size_and_hooks(struct ipt_entry *e, 589 struct xt_table_info *newinfo, 590 const unsigned char *base, 591 const unsigned char *limit, 592 const unsigned int *hook_entries, 593 const unsigned int *underflows, 594 unsigned int valid_hooks) 595 { 596 unsigned int h; 597 int err; 598 599 if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 || 600 (unsigned char *)e + sizeof(struct ipt_entry) >= limit || 601 (unsigned char *)e + e->next_offset > limit) 602 return -EINVAL; 603 604 if (e->next_offset 605 < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target)) 606 return -EINVAL; 607 608 if (!ip_checkentry(&e->ip)) 609 return -EINVAL; 610 611 err = xt_check_entry_offsets(e, e->elems, e->target_offset, 612 e->next_offset); 613 if (err) 614 return err; 615 616 /* Check hooks & underflows */ 617 for (h = 0; h < NF_INET_NUMHOOKS; h++) { 618 if (!(valid_hooks & (1 << h))) 619 continue; 620 if ((unsigned char *)e - base == hook_entries[h]) 621 newinfo->hook_entry[h] = hook_entries[h]; 622 if ((unsigned char *)e - base == underflows[h]) { 623 if (!check_underflow(e)) 624 return -EINVAL; 625 626 newinfo->underflow[h] = underflows[h]; 627 } 628 } 629 630 /* Clear counters and comefrom */ 631 e->counters = ((struct xt_counters) { 0, 0 }); 632 e->comefrom = 0; 633 return 0; 634 } 635 636 static void 637 cleanup_entry(struct ipt_entry *e, struct net *net) 638 { 639 struct xt_tgdtor_param par; 640 struct xt_entry_target *t; 641 struct xt_entry_match *ematch; 642 643 /* Cleanup all matches */ 644 xt_ematch_foreach(ematch, e) 645 cleanup_match(ematch, net); 646 t = ipt_get_target(e); 647 648 par.net = net; 649 par.target = t->u.kernel.target; 650 par.targinfo = t->data; 651 par.family = NFPROTO_IPV4; 652 if (par.target->destroy != NULL) 653 par.target->destroy(&par); 654 module_put(par.target->me); 655 xt_percpu_counter_free(&e->counters); 656 } 657 658 /* Checks and translates the user-supplied table segment (held in 659 newinfo) */ 660 static int 661 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0, 662 const struct ipt_replace *repl) 663 { 664 struct xt_percpu_counter_alloc_state alloc_state = { 0 }; 665 struct ipt_entry *iter; 666 unsigned int *offsets; 667 unsigned int i; 668 int ret = 0; 669 670 newinfo->size = repl->size; 671 newinfo->number = repl->num_entries; 672 673 /* Init all hooks to impossible value. */ 674 for (i = 0; i < NF_INET_NUMHOOKS; i++) { 675 newinfo->hook_entry[i] = 0xFFFFFFFF; 676 newinfo->underflow[i] = 0xFFFFFFFF; 677 } 678 679 offsets = xt_alloc_entry_offsets(newinfo->number); 680 if (!offsets) 681 return -ENOMEM; 682 i = 0; 683 /* Walk through entries, checking offsets. */ 684 xt_entry_foreach(iter, entry0, newinfo->size) { 685 ret = check_entry_size_and_hooks(iter, newinfo, entry0, 686 entry0 + repl->size, 687 repl->hook_entry, 688 repl->underflow, 689 repl->valid_hooks); 690 if (ret != 0) 691 goto out_free; 692 if (i < repl->num_entries) 693 offsets[i] = (void *)iter - entry0; 694 ++i; 695 if (strcmp(ipt_get_target(iter)->u.user.name, 696 XT_ERROR_TARGET) == 0) 697 ++newinfo->stacksize; 698 } 699 700 ret = -EINVAL; 701 if (i != repl->num_entries) 702 goto out_free; 703 704 ret = xt_check_table_hooks(newinfo, repl->valid_hooks); 705 if (ret) 706 goto out_free; 707 708 if (!mark_source_chains(newinfo, repl->valid_hooks, entry0, offsets)) { 709 ret = -ELOOP; 710 goto out_free; 711 } 712 kvfree(offsets); 713 714 /* Finally, each sanity check must pass */ 715 i = 0; 716 xt_entry_foreach(iter, entry0, newinfo->size) { 717 ret = find_check_entry(iter, net, repl->name, repl->size, 718 &alloc_state); 719 if (ret != 0) 720 break; 721 ++i; 722 } 723 724 if (ret != 0) { 725 xt_entry_foreach(iter, entry0, newinfo->size) { 726 if (i-- == 0) 727 break; 728 cleanup_entry(iter, net); 729 } 730 return ret; 731 } 732 733 return ret; 734 out_free: 735 kvfree(offsets); 736 return ret; 737 } 738 739 static void 740 get_counters(const struct xt_table_info *t, 741 struct xt_counters counters[]) 742 { 743 struct ipt_entry *iter; 744 unsigned int cpu; 745 unsigned int i; 746 747 for_each_possible_cpu(cpu) { 748 seqcount_t *s = &per_cpu(xt_recseq, cpu); 749 750 i = 0; 751 xt_entry_foreach(iter, t->entries, t->size) { 752 struct xt_counters *tmp; 753 u64 bcnt, pcnt; 754 unsigned int start; 755 756 tmp = xt_get_per_cpu_counter(&iter->counters, cpu); 757 do { 758 start = read_seqcount_begin(s); 759 bcnt = tmp->bcnt; 760 pcnt = tmp->pcnt; 761 } while (read_seqcount_retry(s, start)); 762 763 ADD_COUNTER(counters[i], bcnt, pcnt); 764 ++i; /* macro does multi eval of i */ 765 cond_resched(); 766 } 767 } 768 } 769 770 static void get_old_counters(const struct xt_table_info *t, 771 struct xt_counters counters[]) 772 { 773 struct ipt_entry *iter; 774 unsigned int cpu, i; 775 776 for_each_possible_cpu(cpu) { 777 i = 0; 778 xt_entry_foreach(iter, t->entries, t->size) { 779 const struct xt_counters *tmp; 780 781 tmp = xt_get_per_cpu_counter(&iter->counters, cpu); 782 ADD_COUNTER(counters[i], tmp->bcnt, tmp->pcnt); 783 ++i; /* macro does multi eval of i */ 784 } 785 786 cond_resched(); 787 } 788 } 789 790 static struct xt_counters *alloc_counters(const struct xt_table *table) 791 { 792 unsigned int countersize; 793 struct xt_counters *counters; 794 const struct xt_table_info *private = xt_table_get_private_protected(table); 795 796 /* We need atomic snapshot of counters: rest doesn't change 797 (other than comefrom, which userspace doesn't care 798 about). */ 799 countersize = sizeof(struct xt_counters) * private->number; 800 counters = vzalloc(countersize); 801 802 if (counters == NULL) 803 return ERR_PTR(-ENOMEM); 804 805 get_counters(private, counters); 806 807 return counters; 808 } 809 810 static int 811 copy_entries_to_user(unsigned int total_size, 812 const struct xt_table *table, 813 void __user *userptr) 814 { 815 unsigned int off, num; 816 const struct ipt_entry *e; 817 struct xt_counters *counters; 818 const struct xt_table_info *private = xt_table_get_private_protected(table); 819 int ret = 0; 820 const void *loc_cpu_entry; 821 822 counters = alloc_counters(table); 823 if (IS_ERR(counters)) 824 return PTR_ERR(counters); 825 826 loc_cpu_entry = private->entries; 827 828 /* FIXME: use iterator macros --RR */ 829 /* ... then go back and fix counters and names */ 830 for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){ 831 unsigned int i; 832 const struct xt_entry_match *m; 833 const struct xt_entry_target *t; 834 835 e = loc_cpu_entry + off; 836 if (copy_to_user(userptr + off, e, sizeof(*e))) { 837 ret = -EFAULT; 838 goto free_counters; 839 } 840 if (copy_to_user(userptr + off 841 + offsetof(struct ipt_entry, counters), 842 &counters[num], 843 sizeof(counters[num])) != 0) { 844 ret = -EFAULT; 845 goto free_counters; 846 } 847 848 for (i = sizeof(struct ipt_entry); 849 i < e->target_offset; 850 i += m->u.match_size) { 851 m = (void *)e + i; 852 853 if (xt_match_to_user(m, userptr + off + i)) { 854 ret = -EFAULT; 855 goto free_counters; 856 } 857 } 858 859 t = ipt_get_target_c(e); 860 if (xt_target_to_user(t, userptr + off + e->target_offset)) { 861 ret = -EFAULT; 862 goto free_counters; 863 } 864 } 865 866 free_counters: 867 vfree(counters); 868 return ret; 869 } 870 871 #ifdef CONFIG_COMPAT 872 static void compat_standard_from_user(void *dst, const void *src) 873 { 874 int v = *(compat_int_t *)src; 875 876 if (v > 0) 877 v += xt_compat_calc_jump(AF_INET, v); 878 memcpy(dst, &v, sizeof(v)); 879 } 880 881 static int compat_standard_to_user(void __user *dst, const void *src) 882 { 883 compat_int_t cv = *(int *)src; 884 885 if (cv > 0) 886 cv -= xt_compat_calc_jump(AF_INET, cv); 887 return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0; 888 } 889 890 static int compat_calc_entry(const struct ipt_entry *e, 891 const struct xt_table_info *info, 892 const void *base, struct xt_table_info *newinfo) 893 { 894 const struct xt_entry_match *ematch; 895 const struct xt_entry_target *t; 896 unsigned int entry_offset; 897 int off, i, ret; 898 899 off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry); 900 entry_offset = (void *)e - base; 901 xt_ematch_foreach(ematch, e) 902 off += xt_compat_match_offset(ematch->u.kernel.match); 903 t = ipt_get_target_c(e); 904 off += xt_compat_target_offset(t->u.kernel.target); 905 newinfo->size -= off; 906 ret = xt_compat_add_offset(AF_INET, entry_offset, off); 907 if (ret) 908 return ret; 909 910 for (i = 0; i < NF_INET_NUMHOOKS; i++) { 911 if (info->hook_entry[i] && 912 (e < (struct ipt_entry *)(base + info->hook_entry[i]))) 913 newinfo->hook_entry[i] -= off; 914 if (info->underflow[i] && 915 (e < (struct ipt_entry *)(base + info->underflow[i]))) 916 newinfo->underflow[i] -= off; 917 } 918 return 0; 919 } 920 921 static int compat_table_info(const struct xt_table_info *info, 922 struct xt_table_info *newinfo) 923 { 924 struct ipt_entry *iter; 925 const void *loc_cpu_entry; 926 int ret; 927 928 if (!newinfo || !info) 929 return -EINVAL; 930 931 /* we dont care about newinfo->entries */ 932 memcpy(newinfo, info, offsetof(struct xt_table_info, entries)); 933 newinfo->initial_entries = 0; 934 loc_cpu_entry = info->entries; 935 ret = xt_compat_init_offsets(AF_INET, info->number); 936 if (ret) 937 return ret; 938 xt_entry_foreach(iter, loc_cpu_entry, info->size) { 939 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo); 940 if (ret != 0) 941 return ret; 942 } 943 return 0; 944 } 945 #endif 946 947 static int get_info(struct net *net, void __user *user, const int *len) 948 { 949 char name[XT_TABLE_MAXNAMELEN]; 950 struct xt_table *t; 951 int ret; 952 953 if (*len != sizeof(struct ipt_getinfo)) 954 return -EINVAL; 955 956 if (copy_from_user(name, user, sizeof(name)) != 0) 957 return -EFAULT; 958 959 name[XT_TABLE_MAXNAMELEN-1] = '\0'; 960 #ifdef CONFIG_COMPAT 961 if (in_compat_syscall()) 962 xt_compat_lock(AF_INET); 963 #endif 964 t = xt_request_find_table_lock(net, AF_INET, name); 965 if (!IS_ERR(t)) { 966 struct ipt_getinfo info; 967 const struct xt_table_info *private = xt_table_get_private_protected(t); 968 #ifdef CONFIG_COMPAT 969 struct xt_table_info tmp; 970 971 if (in_compat_syscall()) { 972 ret = compat_table_info(private, &tmp); 973 xt_compat_flush_offsets(AF_INET); 974 private = &tmp; 975 } 976 #endif 977 memset(&info, 0, sizeof(info)); 978 info.valid_hooks = t->valid_hooks; 979 memcpy(info.hook_entry, private->hook_entry, 980 sizeof(info.hook_entry)); 981 memcpy(info.underflow, private->underflow, 982 sizeof(info.underflow)); 983 info.num_entries = private->number; 984 info.size = private->size; 985 strcpy(info.name, name); 986 987 if (copy_to_user(user, &info, *len) != 0) 988 ret = -EFAULT; 989 else 990 ret = 0; 991 992 xt_table_unlock(t); 993 module_put(t->me); 994 } else 995 ret = PTR_ERR(t); 996 #ifdef CONFIG_COMPAT 997 if (in_compat_syscall()) 998 xt_compat_unlock(AF_INET); 999 #endif 1000 return ret; 1001 } 1002 1003 static int 1004 get_entries(struct net *net, struct ipt_get_entries __user *uptr, 1005 const int *len) 1006 { 1007 int ret; 1008 struct ipt_get_entries get; 1009 struct xt_table *t; 1010 1011 if (*len < sizeof(get)) 1012 return -EINVAL; 1013 if (copy_from_user(&get, uptr, sizeof(get)) != 0) 1014 return -EFAULT; 1015 if (*len != sizeof(struct ipt_get_entries) + get.size) 1016 return -EINVAL; 1017 get.name[sizeof(get.name) - 1] = '\0'; 1018 1019 t = xt_find_table_lock(net, AF_INET, get.name); 1020 if (!IS_ERR(t)) { 1021 const struct xt_table_info *private = xt_table_get_private_protected(t); 1022 if (get.size == private->size) 1023 ret = copy_entries_to_user(private->size, 1024 t, uptr->entrytable); 1025 else 1026 ret = -EAGAIN; 1027 1028 module_put(t->me); 1029 xt_table_unlock(t); 1030 } else 1031 ret = PTR_ERR(t); 1032 1033 return ret; 1034 } 1035 1036 static int 1037 __do_replace(struct net *net, const char *name, unsigned int valid_hooks, 1038 struct xt_table_info *newinfo, unsigned int num_counters, 1039 void __user *counters_ptr) 1040 { 1041 int ret; 1042 struct xt_table *t; 1043 struct xt_table_info *oldinfo; 1044 struct xt_counters *counters; 1045 struct ipt_entry *iter; 1046 1047 ret = 0; 1048 counters = xt_counters_alloc(num_counters); 1049 if (!counters) { 1050 ret = -ENOMEM; 1051 goto out; 1052 } 1053 1054 t = xt_request_find_table_lock(net, AF_INET, name); 1055 if (IS_ERR(t)) { 1056 ret = PTR_ERR(t); 1057 goto free_newinfo_counters_untrans; 1058 } 1059 1060 /* You lied! */ 1061 if (valid_hooks != t->valid_hooks) { 1062 ret = -EINVAL; 1063 goto put_module; 1064 } 1065 1066 oldinfo = xt_replace_table(t, num_counters, newinfo, &ret); 1067 if (!oldinfo) 1068 goto put_module; 1069 1070 /* Update module usage count based on number of rules */ 1071 if ((oldinfo->number > oldinfo->initial_entries) || 1072 (newinfo->number <= oldinfo->initial_entries)) 1073 module_put(t->me); 1074 if ((oldinfo->number > oldinfo->initial_entries) && 1075 (newinfo->number <= oldinfo->initial_entries)) 1076 module_put(t->me); 1077 1078 xt_table_unlock(t); 1079 1080 get_old_counters(oldinfo, counters); 1081 1082 /* Decrease module usage counts and free resource */ 1083 xt_entry_foreach(iter, oldinfo->entries, oldinfo->size) 1084 cleanup_entry(iter, net); 1085 1086 xt_free_table_info(oldinfo); 1087 if (copy_to_user(counters_ptr, counters, 1088 sizeof(struct xt_counters) * num_counters) != 0) { 1089 /* Silent error, can't fail, new table is already in place */ 1090 net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n"); 1091 } 1092 vfree(counters); 1093 return ret; 1094 1095 put_module: 1096 module_put(t->me); 1097 xt_table_unlock(t); 1098 free_newinfo_counters_untrans: 1099 vfree(counters); 1100 out: 1101 return ret; 1102 } 1103 1104 static int 1105 do_replace(struct net *net, sockptr_t arg, unsigned int len) 1106 { 1107 int ret; 1108 struct ipt_replace tmp; 1109 struct xt_table_info *newinfo; 1110 void *loc_cpu_entry; 1111 struct ipt_entry *iter; 1112 1113 if (copy_from_sockptr(&tmp, arg, sizeof(tmp)) != 0) 1114 return -EFAULT; 1115 1116 /* overflow check */ 1117 if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters)) 1118 return -ENOMEM; 1119 if (tmp.num_counters == 0) 1120 return -EINVAL; 1121 1122 tmp.name[sizeof(tmp.name)-1] = 0; 1123 1124 newinfo = xt_alloc_table_info(tmp.size); 1125 if (!newinfo) 1126 return -ENOMEM; 1127 1128 loc_cpu_entry = newinfo->entries; 1129 if (copy_from_sockptr_offset(loc_cpu_entry, arg, sizeof(tmp), 1130 tmp.size) != 0) { 1131 ret = -EFAULT; 1132 goto free_newinfo; 1133 } 1134 1135 ret = translate_table(net, newinfo, loc_cpu_entry, &tmp); 1136 if (ret != 0) 1137 goto free_newinfo; 1138 1139 ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo, 1140 tmp.num_counters, tmp.counters); 1141 if (ret) 1142 goto free_newinfo_untrans; 1143 return 0; 1144 1145 free_newinfo_untrans: 1146 xt_entry_foreach(iter, loc_cpu_entry, newinfo->size) 1147 cleanup_entry(iter, net); 1148 free_newinfo: 1149 xt_free_table_info(newinfo); 1150 return ret; 1151 } 1152 1153 static int 1154 do_add_counters(struct net *net, sockptr_t arg, unsigned int len) 1155 { 1156 unsigned int i; 1157 struct xt_counters_info tmp; 1158 struct xt_counters *paddc; 1159 struct xt_table *t; 1160 const struct xt_table_info *private; 1161 int ret = 0; 1162 struct ipt_entry *iter; 1163 unsigned int addend; 1164 1165 paddc = xt_copy_counters(arg, len, &tmp); 1166 if (IS_ERR(paddc)) 1167 return PTR_ERR(paddc); 1168 1169 t = xt_find_table_lock(net, AF_INET, tmp.name); 1170 if (IS_ERR(t)) { 1171 ret = PTR_ERR(t); 1172 goto free; 1173 } 1174 1175 local_bh_disable(); 1176 private = xt_table_get_private_protected(t); 1177 if (private->number != tmp.num_counters) { 1178 ret = -EINVAL; 1179 goto unlock_up_free; 1180 } 1181 1182 i = 0; 1183 addend = xt_write_recseq_begin(); 1184 xt_entry_foreach(iter, private->entries, private->size) { 1185 struct xt_counters *tmp; 1186 1187 tmp = xt_get_this_cpu_counter(&iter->counters); 1188 ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt); 1189 ++i; 1190 } 1191 xt_write_recseq_end(addend); 1192 unlock_up_free: 1193 local_bh_enable(); 1194 xt_table_unlock(t); 1195 module_put(t->me); 1196 free: 1197 vfree(paddc); 1198 1199 return ret; 1200 } 1201 1202 #ifdef CONFIG_COMPAT 1203 struct compat_ipt_replace { 1204 char name[XT_TABLE_MAXNAMELEN]; 1205 u32 valid_hooks; 1206 u32 num_entries; 1207 u32 size; 1208 u32 hook_entry[NF_INET_NUMHOOKS]; 1209 u32 underflow[NF_INET_NUMHOOKS]; 1210 u32 num_counters; 1211 compat_uptr_t counters; /* struct xt_counters * */ 1212 struct compat_ipt_entry entries[]; 1213 }; 1214 1215 static int 1216 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr, 1217 unsigned int *size, struct xt_counters *counters, 1218 unsigned int i) 1219 { 1220 struct xt_entry_target *t; 1221 struct compat_ipt_entry __user *ce; 1222 u_int16_t target_offset, next_offset; 1223 compat_uint_t origsize; 1224 const struct xt_entry_match *ematch; 1225 int ret = 0; 1226 1227 origsize = *size; 1228 ce = *dstptr; 1229 if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 || 1230 copy_to_user(&ce->counters, &counters[i], 1231 sizeof(counters[i])) != 0) 1232 return -EFAULT; 1233 1234 *dstptr += sizeof(struct compat_ipt_entry); 1235 *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry); 1236 1237 xt_ematch_foreach(ematch, e) { 1238 ret = xt_compat_match_to_user(ematch, dstptr, size); 1239 if (ret != 0) 1240 return ret; 1241 } 1242 target_offset = e->target_offset - (origsize - *size); 1243 t = ipt_get_target(e); 1244 ret = xt_compat_target_to_user(t, dstptr, size); 1245 if (ret) 1246 return ret; 1247 next_offset = e->next_offset - (origsize - *size); 1248 if (put_user(target_offset, &ce->target_offset) != 0 || 1249 put_user(next_offset, &ce->next_offset) != 0) 1250 return -EFAULT; 1251 return 0; 1252 } 1253 1254 static int 1255 compat_find_calc_match(struct xt_entry_match *m, 1256 const struct ipt_ip *ip, 1257 int *size) 1258 { 1259 struct xt_match *match; 1260 1261 match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name, 1262 m->u.user.revision); 1263 if (IS_ERR(match)) 1264 return PTR_ERR(match); 1265 1266 m->u.kernel.match = match; 1267 *size += xt_compat_match_offset(match); 1268 return 0; 1269 } 1270 1271 static void compat_release_entry(struct compat_ipt_entry *e) 1272 { 1273 struct xt_entry_target *t; 1274 struct xt_entry_match *ematch; 1275 1276 /* Cleanup all matches */ 1277 xt_ematch_foreach(ematch, e) 1278 module_put(ematch->u.kernel.match->me); 1279 t = compat_ipt_get_target(e); 1280 module_put(t->u.kernel.target->me); 1281 } 1282 1283 static int 1284 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e, 1285 struct xt_table_info *newinfo, 1286 unsigned int *size, 1287 const unsigned char *base, 1288 const unsigned char *limit) 1289 { 1290 struct xt_entry_match *ematch; 1291 struct xt_entry_target *t; 1292 struct xt_target *target; 1293 unsigned int entry_offset; 1294 unsigned int j; 1295 int ret, off; 1296 1297 if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 || 1298 (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit || 1299 (unsigned char *)e + e->next_offset > limit) 1300 return -EINVAL; 1301 1302 if (e->next_offset < sizeof(struct compat_ipt_entry) + 1303 sizeof(struct compat_xt_entry_target)) 1304 return -EINVAL; 1305 1306 if (!ip_checkentry(&e->ip)) 1307 return -EINVAL; 1308 1309 ret = xt_compat_check_entry_offsets(e, e->elems, 1310 e->target_offset, e->next_offset); 1311 if (ret) 1312 return ret; 1313 1314 off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry); 1315 entry_offset = (void *)e - (void *)base; 1316 j = 0; 1317 xt_ematch_foreach(ematch, e) { 1318 ret = compat_find_calc_match(ematch, &e->ip, &off); 1319 if (ret != 0) 1320 goto release_matches; 1321 ++j; 1322 } 1323 1324 t = compat_ipt_get_target(e); 1325 target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name, 1326 t->u.user.revision); 1327 if (IS_ERR(target)) { 1328 ret = PTR_ERR(target); 1329 goto release_matches; 1330 } 1331 t->u.kernel.target = target; 1332 1333 off += xt_compat_target_offset(target); 1334 *size += off; 1335 ret = xt_compat_add_offset(AF_INET, entry_offset, off); 1336 if (ret) 1337 goto out; 1338 1339 return 0; 1340 1341 out: 1342 module_put(t->u.kernel.target->me); 1343 release_matches: 1344 xt_ematch_foreach(ematch, e) { 1345 if (j-- == 0) 1346 break; 1347 module_put(ematch->u.kernel.match->me); 1348 } 1349 return ret; 1350 } 1351 1352 static void 1353 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr, 1354 unsigned int *size, 1355 struct xt_table_info *newinfo, unsigned char *base) 1356 { 1357 struct xt_entry_target *t; 1358 struct ipt_entry *de; 1359 unsigned int origsize; 1360 int h; 1361 struct xt_entry_match *ematch; 1362 1363 origsize = *size; 1364 de = *dstptr; 1365 memcpy(de, e, sizeof(struct ipt_entry)); 1366 memcpy(&de->counters, &e->counters, sizeof(e->counters)); 1367 1368 *dstptr += sizeof(struct ipt_entry); 1369 *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry); 1370 1371 xt_ematch_foreach(ematch, e) 1372 xt_compat_match_from_user(ematch, dstptr, size); 1373 1374 de->target_offset = e->target_offset - (origsize - *size); 1375 t = compat_ipt_get_target(e); 1376 xt_compat_target_from_user(t, dstptr, size); 1377 1378 de->next_offset = e->next_offset - (origsize - *size); 1379 1380 for (h = 0; h < NF_INET_NUMHOOKS; h++) { 1381 if ((unsigned char *)de - base < newinfo->hook_entry[h]) 1382 newinfo->hook_entry[h] -= origsize - *size; 1383 if ((unsigned char *)de - base < newinfo->underflow[h]) 1384 newinfo->underflow[h] -= origsize - *size; 1385 } 1386 } 1387 1388 static int 1389 translate_compat_table(struct net *net, 1390 struct xt_table_info **pinfo, 1391 void **pentry0, 1392 const struct compat_ipt_replace *compatr) 1393 { 1394 unsigned int i, j; 1395 struct xt_table_info *newinfo, *info; 1396 void *pos, *entry0, *entry1; 1397 struct compat_ipt_entry *iter0; 1398 struct ipt_replace repl; 1399 unsigned int size; 1400 int ret; 1401 1402 info = *pinfo; 1403 entry0 = *pentry0; 1404 size = compatr->size; 1405 info->number = compatr->num_entries; 1406 1407 j = 0; 1408 xt_compat_lock(AF_INET); 1409 ret = xt_compat_init_offsets(AF_INET, compatr->num_entries); 1410 if (ret) 1411 goto out_unlock; 1412 /* Walk through entries, checking offsets. */ 1413 xt_entry_foreach(iter0, entry0, compatr->size) { 1414 ret = check_compat_entry_size_and_hooks(iter0, info, &size, 1415 entry0, 1416 entry0 + compatr->size); 1417 if (ret != 0) 1418 goto out_unlock; 1419 ++j; 1420 } 1421 1422 ret = -EINVAL; 1423 if (j != compatr->num_entries) 1424 goto out_unlock; 1425 1426 ret = -ENOMEM; 1427 newinfo = xt_alloc_table_info(size); 1428 if (!newinfo) 1429 goto out_unlock; 1430 1431 newinfo->number = compatr->num_entries; 1432 for (i = 0; i < NF_INET_NUMHOOKS; i++) { 1433 newinfo->hook_entry[i] = compatr->hook_entry[i]; 1434 newinfo->underflow[i] = compatr->underflow[i]; 1435 } 1436 entry1 = newinfo->entries; 1437 pos = entry1; 1438 size = compatr->size; 1439 xt_entry_foreach(iter0, entry0, compatr->size) 1440 compat_copy_entry_from_user(iter0, &pos, &size, 1441 newinfo, entry1); 1442 1443 /* all module references in entry0 are now gone. 1444 * entry1/newinfo contains a 64bit ruleset that looks exactly as 1445 * generated by 64bit userspace. 1446 * 1447 * Call standard translate_table() to validate all hook_entrys, 1448 * underflows, check for loops, etc. 1449 */ 1450 xt_compat_flush_offsets(AF_INET); 1451 xt_compat_unlock(AF_INET); 1452 1453 memcpy(&repl, compatr, sizeof(*compatr)); 1454 1455 for (i = 0; i < NF_INET_NUMHOOKS; i++) { 1456 repl.hook_entry[i] = newinfo->hook_entry[i]; 1457 repl.underflow[i] = newinfo->underflow[i]; 1458 } 1459 1460 repl.num_counters = 0; 1461 repl.counters = NULL; 1462 repl.size = newinfo->size; 1463 ret = translate_table(net, newinfo, entry1, &repl); 1464 if (ret) 1465 goto free_newinfo; 1466 1467 *pinfo = newinfo; 1468 *pentry0 = entry1; 1469 xt_free_table_info(info); 1470 return 0; 1471 1472 free_newinfo: 1473 xt_free_table_info(newinfo); 1474 return ret; 1475 out_unlock: 1476 xt_compat_flush_offsets(AF_INET); 1477 xt_compat_unlock(AF_INET); 1478 xt_entry_foreach(iter0, entry0, compatr->size) { 1479 if (j-- == 0) 1480 break; 1481 compat_release_entry(iter0); 1482 } 1483 return ret; 1484 } 1485 1486 static int 1487 compat_do_replace(struct net *net, sockptr_t arg, unsigned int len) 1488 { 1489 int ret; 1490 struct compat_ipt_replace tmp; 1491 struct xt_table_info *newinfo; 1492 void *loc_cpu_entry; 1493 struct ipt_entry *iter; 1494 1495 if (copy_from_sockptr(&tmp, arg, sizeof(tmp)) != 0) 1496 return -EFAULT; 1497 1498 /* overflow check */ 1499 if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters)) 1500 return -ENOMEM; 1501 if (tmp.num_counters == 0) 1502 return -EINVAL; 1503 1504 tmp.name[sizeof(tmp.name)-1] = 0; 1505 1506 newinfo = xt_alloc_table_info(tmp.size); 1507 if (!newinfo) 1508 return -ENOMEM; 1509 1510 loc_cpu_entry = newinfo->entries; 1511 if (copy_from_sockptr_offset(loc_cpu_entry, arg, sizeof(tmp), 1512 tmp.size) != 0) { 1513 ret = -EFAULT; 1514 goto free_newinfo; 1515 } 1516 1517 ret = translate_compat_table(net, &newinfo, &loc_cpu_entry, &tmp); 1518 if (ret != 0) 1519 goto free_newinfo; 1520 1521 ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo, 1522 tmp.num_counters, compat_ptr(tmp.counters)); 1523 if (ret) 1524 goto free_newinfo_untrans; 1525 return 0; 1526 1527 free_newinfo_untrans: 1528 xt_entry_foreach(iter, loc_cpu_entry, newinfo->size) 1529 cleanup_entry(iter, net); 1530 free_newinfo: 1531 xt_free_table_info(newinfo); 1532 return ret; 1533 } 1534 1535 struct compat_ipt_get_entries { 1536 char name[XT_TABLE_MAXNAMELEN]; 1537 compat_uint_t size; 1538 struct compat_ipt_entry entrytable[]; 1539 }; 1540 1541 static int 1542 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table, 1543 void __user *userptr) 1544 { 1545 struct xt_counters *counters; 1546 const struct xt_table_info *private = xt_table_get_private_protected(table); 1547 void __user *pos; 1548 unsigned int size; 1549 int ret = 0; 1550 unsigned int i = 0; 1551 struct ipt_entry *iter; 1552 1553 counters = alloc_counters(table); 1554 if (IS_ERR(counters)) 1555 return PTR_ERR(counters); 1556 1557 pos = userptr; 1558 size = total_size; 1559 xt_entry_foreach(iter, private->entries, total_size) { 1560 ret = compat_copy_entry_to_user(iter, &pos, 1561 &size, counters, i++); 1562 if (ret != 0) 1563 break; 1564 } 1565 1566 vfree(counters); 1567 return ret; 1568 } 1569 1570 static int 1571 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr, 1572 int *len) 1573 { 1574 int ret; 1575 struct compat_ipt_get_entries get; 1576 struct xt_table *t; 1577 1578 if (*len < sizeof(get)) 1579 return -EINVAL; 1580 1581 if (copy_from_user(&get, uptr, sizeof(get)) != 0) 1582 return -EFAULT; 1583 1584 if (*len != sizeof(struct compat_ipt_get_entries) + get.size) 1585 return -EINVAL; 1586 1587 get.name[sizeof(get.name) - 1] = '\0'; 1588 1589 xt_compat_lock(AF_INET); 1590 t = xt_find_table_lock(net, AF_INET, get.name); 1591 if (!IS_ERR(t)) { 1592 const struct xt_table_info *private = xt_table_get_private_protected(t); 1593 struct xt_table_info info; 1594 ret = compat_table_info(private, &info); 1595 if (!ret && get.size == info.size) 1596 ret = compat_copy_entries_to_user(private->size, 1597 t, uptr->entrytable); 1598 else if (!ret) 1599 ret = -EAGAIN; 1600 1601 xt_compat_flush_offsets(AF_INET); 1602 module_put(t->me); 1603 xt_table_unlock(t); 1604 } else 1605 ret = PTR_ERR(t); 1606 1607 xt_compat_unlock(AF_INET); 1608 return ret; 1609 } 1610 #endif 1611 1612 static int 1613 do_ipt_set_ctl(struct sock *sk, int cmd, sockptr_t arg, unsigned int len) 1614 { 1615 int ret; 1616 1617 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) 1618 return -EPERM; 1619 1620 switch (cmd) { 1621 case IPT_SO_SET_REPLACE: 1622 #ifdef CONFIG_COMPAT 1623 if (in_compat_syscall()) 1624 ret = compat_do_replace(sock_net(sk), arg, len); 1625 else 1626 #endif 1627 ret = do_replace(sock_net(sk), arg, len); 1628 break; 1629 1630 case IPT_SO_SET_ADD_COUNTERS: 1631 ret = do_add_counters(sock_net(sk), arg, len); 1632 break; 1633 1634 default: 1635 ret = -EINVAL; 1636 } 1637 1638 return ret; 1639 } 1640 1641 static int 1642 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len) 1643 { 1644 int ret; 1645 1646 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) 1647 return -EPERM; 1648 1649 switch (cmd) { 1650 case IPT_SO_GET_INFO: 1651 ret = get_info(sock_net(sk), user, len); 1652 break; 1653 1654 case IPT_SO_GET_ENTRIES: 1655 #ifdef CONFIG_COMPAT 1656 if (in_compat_syscall()) 1657 ret = compat_get_entries(sock_net(sk), user, len); 1658 else 1659 #endif 1660 ret = get_entries(sock_net(sk), user, len); 1661 break; 1662 1663 case IPT_SO_GET_REVISION_MATCH: 1664 case IPT_SO_GET_REVISION_TARGET: { 1665 struct xt_get_revision rev; 1666 int target; 1667 1668 if (*len != sizeof(rev)) { 1669 ret = -EINVAL; 1670 break; 1671 } 1672 if (copy_from_user(&rev, user, sizeof(rev)) != 0) { 1673 ret = -EFAULT; 1674 break; 1675 } 1676 rev.name[sizeof(rev.name)-1] = 0; 1677 1678 if (cmd == IPT_SO_GET_REVISION_TARGET) 1679 target = 1; 1680 else 1681 target = 0; 1682 1683 try_then_request_module(xt_find_revision(AF_INET, rev.name, 1684 rev.revision, 1685 target, &ret), 1686 "ipt_%s", rev.name); 1687 break; 1688 } 1689 1690 default: 1691 ret = -EINVAL; 1692 } 1693 1694 return ret; 1695 } 1696 1697 static void __ipt_unregister_table(struct net *net, struct xt_table *table) 1698 { 1699 struct xt_table_info *private; 1700 void *loc_cpu_entry; 1701 struct module *table_owner = table->me; 1702 struct ipt_entry *iter; 1703 1704 private = xt_unregister_table(table); 1705 1706 /* Decrease module usage counts and free resources */ 1707 loc_cpu_entry = private->entries; 1708 xt_entry_foreach(iter, loc_cpu_entry, private->size) 1709 cleanup_entry(iter, net); 1710 if (private->number > private->initial_entries) 1711 module_put(table_owner); 1712 xt_free_table_info(private); 1713 } 1714 1715 int ipt_register_table(struct net *net, const struct xt_table *table, 1716 const struct ipt_replace *repl, 1717 const struct nf_hook_ops *ops, struct xt_table **res) 1718 { 1719 int ret; 1720 struct xt_table_info *newinfo; 1721 struct xt_table_info bootstrap = {0}; 1722 void *loc_cpu_entry; 1723 struct xt_table *new_table; 1724 1725 newinfo = xt_alloc_table_info(repl->size); 1726 if (!newinfo) 1727 return -ENOMEM; 1728 1729 loc_cpu_entry = newinfo->entries; 1730 memcpy(loc_cpu_entry, repl->entries, repl->size); 1731 1732 ret = translate_table(net, newinfo, loc_cpu_entry, repl); 1733 if (ret != 0) 1734 goto out_free; 1735 1736 new_table = xt_register_table(net, table, &bootstrap, newinfo); 1737 if (IS_ERR(new_table)) { 1738 ret = PTR_ERR(new_table); 1739 goto out_free; 1740 } 1741 1742 /* set res now, will see skbs right after nf_register_net_hooks */ 1743 WRITE_ONCE(*res, new_table); 1744 if (!ops) 1745 return 0; 1746 1747 ret = nf_register_net_hooks(net, ops, hweight32(table->valid_hooks)); 1748 if (ret != 0) { 1749 __ipt_unregister_table(net, new_table); 1750 *res = NULL; 1751 } 1752 1753 return ret; 1754 1755 out_free: 1756 xt_free_table_info(newinfo); 1757 return ret; 1758 } 1759 1760 void ipt_unregister_table_pre_exit(struct net *net, struct xt_table *table, 1761 const struct nf_hook_ops *ops) 1762 { 1763 nf_unregister_net_hooks(net, ops, hweight32(table->valid_hooks)); 1764 } 1765 1766 void ipt_unregister_table_exit(struct net *net, struct xt_table *table) 1767 { 1768 __ipt_unregister_table(net, table); 1769 } 1770 1771 void ipt_unregister_table(struct net *net, struct xt_table *table, 1772 const struct nf_hook_ops *ops) 1773 { 1774 if (ops) 1775 ipt_unregister_table_pre_exit(net, table, ops); 1776 __ipt_unregister_table(net, table); 1777 } 1778 1779 /* Returns 1 if the type and code is matched by the range, 0 otherwise */ 1780 static inline bool 1781 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code, 1782 u_int8_t type, u_int8_t code, 1783 bool invert) 1784 { 1785 return ((test_type == 0xFF) || 1786 (type == test_type && code >= min_code && code <= max_code)) 1787 ^ invert; 1788 } 1789 1790 static bool 1791 icmp_match(const struct sk_buff *skb, struct xt_action_param *par) 1792 { 1793 const struct icmphdr *ic; 1794 struct icmphdr _icmph; 1795 const struct ipt_icmp *icmpinfo = par->matchinfo; 1796 1797 /* Must not be a fragment. */ 1798 if (par->fragoff != 0) 1799 return false; 1800 1801 ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph); 1802 if (ic == NULL) { 1803 /* We've been asked to examine this packet, and we 1804 * can't. Hence, no choice but to drop. 1805 */ 1806 par->hotdrop = true; 1807 return false; 1808 } 1809 1810 return icmp_type_code_match(icmpinfo->type, 1811 icmpinfo->code[0], 1812 icmpinfo->code[1], 1813 ic->type, ic->code, 1814 !!(icmpinfo->invflags&IPT_ICMP_INV)); 1815 } 1816 1817 static int icmp_checkentry(const struct xt_mtchk_param *par) 1818 { 1819 const struct ipt_icmp *icmpinfo = par->matchinfo; 1820 1821 /* Must specify no unknown invflags */ 1822 return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0; 1823 } 1824 1825 static struct xt_target ipt_builtin_tg[] __read_mostly = { 1826 { 1827 .name = XT_STANDARD_TARGET, 1828 .targetsize = sizeof(int), 1829 .family = NFPROTO_IPV4, 1830 #ifdef CONFIG_COMPAT 1831 .compatsize = sizeof(compat_int_t), 1832 .compat_from_user = compat_standard_from_user, 1833 .compat_to_user = compat_standard_to_user, 1834 #endif 1835 }, 1836 { 1837 .name = XT_ERROR_TARGET, 1838 .target = ipt_error, 1839 .targetsize = XT_FUNCTION_MAXNAMELEN, 1840 .family = NFPROTO_IPV4, 1841 }, 1842 }; 1843 1844 static struct nf_sockopt_ops ipt_sockopts = { 1845 .pf = PF_INET, 1846 .set_optmin = IPT_BASE_CTL, 1847 .set_optmax = IPT_SO_SET_MAX+1, 1848 .set = do_ipt_set_ctl, 1849 .get_optmin = IPT_BASE_CTL, 1850 .get_optmax = IPT_SO_GET_MAX+1, 1851 .get = do_ipt_get_ctl, 1852 .owner = THIS_MODULE, 1853 }; 1854 1855 static struct xt_match ipt_builtin_mt[] __read_mostly = { 1856 { 1857 .name = "icmp", 1858 .match = icmp_match, 1859 .matchsize = sizeof(struct ipt_icmp), 1860 .checkentry = icmp_checkentry, 1861 .proto = IPPROTO_ICMP, 1862 .family = NFPROTO_IPV4, 1863 .me = THIS_MODULE, 1864 }, 1865 }; 1866 1867 static int __net_init ip_tables_net_init(struct net *net) 1868 { 1869 return xt_proto_init(net, NFPROTO_IPV4); 1870 } 1871 1872 static void __net_exit ip_tables_net_exit(struct net *net) 1873 { 1874 xt_proto_fini(net, NFPROTO_IPV4); 1875 } 1876 1877 static struct pernet_operations ip_tables_net_ops = { 1878 .init = ip_tables_net_init, 1879 .exit = ip_tables_net_exit, 1880 }; 1881 1882 static int __init ip_tables_init(void) 1883 { 1884 int ret; 1885 1886 ret = register_pernet_subsys(&ip_tables_net_ops); 1887 if (ret < 0) 1888 goto err1; 1889 1890 /* No one else will be downing sem now, so we won't sleep */ 1891 ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg)); 1892 if (ret < 0) 1893 goto err2; 1894 ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt)); 1895 if (ret < 0) 1896 goto err4; 1897 1898 /* Register setsockopt */ 1899 ret = nf_register_sockopt(&ipt_sockopts); 1900 if (ret < 0) 1901 goto err5; 1902 1903 return 0; 1904 1905 err5: 1906 xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt)); 1907 err4: 1908 xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg)); 1909 err2: 1910 unregister_pernet_subsys(&ip_tables_net_ops); 1911 err1: 1912 return ret; 1913 } 1914 1915 static void __exit ip_tables_fini(void) 1916 { 1917 nf_unregister_sockopt(&ipt_sockopts); 1918 1919 xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt)); 1920 xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg)); 1921 unregister_pernet_subsys(&ip_tables_net_ops); 1922 } 1923 1924 EXPORT_SYMBOL(ipt_register_table); 1925 EXPORT_SYMBOL(ipt_unregister_table); 1926 EXPORT_SYMBOL(ipt_unregister_table_pre_exit); 1927 EXPORT_SYMBOL(ipt_unregister_table_exit); 1928 EXPORT_SYMBOL(ipt_do_table); 1929 module_init(ip_tables_init); 1930 module_exit(ip_tables_fini); 1931