1 // SPDX-License-Identifier: GPL-2.0 2 /* Connection tracking via netlink socket. Allows for user space 3 * protocol helpers and general trouble making from userspace. 4 * 5 * (C) 2001 by Jay Schulist <jschlst@samba.org> 6 * (C) 2002-2006 by Harald Welte <laforge@gnumonks.org> 7 * (C) 2003 by Patrick Mchardy <kaber@trash.net> 8 * (C) 2005-2012 by Pablo Neira Ayuso <pablo@netfilter.org> 9 * 10 * Initial connection tracking via netlink development funded and 11 * generally made possible by Network Robots, Inc. (www.networkrobots.com) 12 * 13 * Further development of this code funded by Astaro AG (http://www.astaro.com) 14 */ 15 16 #include <linux/init.h> 17 #include <linux/module.h> 18 #include <linux/kernel.h> 19 #include <linux/rculist.h> 20 #include <linux/rculist_nulls.h> 21 #include <linux/types.h> 22 #include <linux/timer.h> 23 #include <linux/security.h> 24 #include <linux/skbuff.h> 25 #include <linux/errno.h> 26 #include <linux/netlink.h> 27 #include <linux/spinlock.h> 28 #include <linux/interrupt.h> 29 #include <linux/slab.h> 30 #include <linux/siphash.h> 31 32 #include <linux/netfilter.h> 33 #include <net/ipv6.h> 34 #include <net/netlink.h> 35 #include <net/sock.h> 36 #include <net/netfilter/nf_conntrack.h> 37 #include <net/netfilter/nf_conntrack_core.h> 38 #include <net/netfilter/nf_conntrack_expect.h> 39 #include <net/netfilter/nf_conntrack_helper.h> 40 #include <net/netfilter/nf_conntrack_seqadj.h> 41 #include <net/netfilter/nf_conntrack_l4proto.h> 42 #include <net/netfilter/nf_conntrack_tuple.h> 43 #include <net/netfilter/nf_conntrack_acct.h> 44 #include <net/netfilter/nf_conntrack_zones.h> 45 #include <net/netfilter/nf_conntrack_timestamp.h> 46 #include <net/netfilter/nf_conntrack_labels.h> 47 #include <net/netfilter/nf_conntrack_synproxy.h> 48 #if IS_ENABLED(CONFIG_NF_NAT) 49 #include <net/netfilter/nf_nat.h> 50 #include <net/netfilter/nf_nat_helper.h> 51 #endif 52 53 #include <linux/netfilter/nfnetlink.h> 54 #include <linux/netfilter/nfnetlink_conntrack.h> 55 56 #include "nf_internals.h" 57 58 MODULE_LICENSE("GPL"); 59 MODULE_DESCRIPTION("List and change connection tracking table"); 60 61 struct ctnetlink_list_dump_ctx { 62 unsigned long last_id; 63 unsigned int cpu; 64 bool done; 65 }; 66 67 static int ctnetlink_dump_tuples_proto(struct sk_buff *skb, 68 const struct nf_conntrack_tuple *tuple, 69 const struct nf_conntrack_l4proto *l4proto) 70 { 71 int ret = 0; 72 struct nlattr *nest_parms; 73 74 nest_parms = nla_nest_start(skb, CTA_TUPLE_PROTO); 75 if (!nest_parms) 76 goto nla_put_failure; 77 if (nla_put_u8(skb, CTA_PROTO_NUM, tuple->dst.protonum)) 78 goto nla_put_failure; 79 80 if (likely(l4proto->tuple_to_nlattr)) 81 ret = l4proto->tuple_to_nlattr(skb, tuple); 82 83 nla_nest_end(skb, nest_parms); 84 85 return ret; 86 87 nla_put_failure: 88 return -1; 89 } 90 91 static int ipv4_tuple_to_nlattr(struct sk_buff *skb, 92 const struct nf_conntrack_tuple *tuple) 93 { 94 if (nla_put_in_addr(skb, CTA_IP_V4_SRC, tuple->src.u3.ip) || 95 nla_put_in_addr(skb, CTA_IP_V4_DST, tuple->dst.u3.ip)) 96 return -EMSGSIZE; 97 return 0; 98 } 99 100 static int ipv6_tuple_to_nlattr(struct sk_buff *skb, 101 const struct nf_conntrack_tuple *tuple) 102 { 103 if (nla_put_in6_addr(skb, CTA_IP_V6_SRC, &tuple->src.u3.in6) || 104 nla_put_in6_addr(skb, CTA_IP_V6_DST, &tuple->dst.u3.in6)) 105 return -EMSGSIZE; 106 return 0; 107 } 108 109 static int ctnetlink_dump_tuples_ip(struct sk_buff *skb, 110 const struct nf_conntrack_tuple *tuple) 111 { 112 int ret = 0; 113 struct nlattr *nest_parms; 114 115 nest_parms = nla_nest_start(skb, CTA_TUPLE_IP); 116 if (!nest_parms) 117 goto nla_put_failure; 118 119 switch (tuple->src.l3num) { 120 case NFPROTO_IPV4: 121 ret = ipv4_tuple_to_nlattr(skb, tuple); 122 break; 123 case NFPROTO_IPV6: 124 ret = ipv6_tuple_to_nlattr(skb, tuple); 125 break; 126 } 127 128 nla_nest_end(skb, nest_parms); 129 130 return ret; 131 132 nla_put_failure: 133 return -1; 134 } 135 136 static int ctnetlink_dump_tuples(struct sk_buff *skb, 137 const struct nf_conntrack_tuple *tuple) 138 { 139 const struct nf_conntrack_l4proto *l4proto; 140 int ret; 141 142 rcu_read_lock(); 143 ret = ctnetlink_dump_tuples_ip(skb, tuple); 144 145 if (ret >= 0) { 146 l4proto = nf_ct_l4proto_find(tuple->dst.protonum); 147 ret = ctnetlink_dump_tuples_proto(skb, tuple, l4proto); 148 } 149 rcu_read_unlock(); 150 return ret; 151 } 152 153 static int ctnetlink_dump_zone_id(struct sk_buff *skb, int attrtype, 154 const struct nf_conntrack_zone *zone, int dir) 155 { 156 if (zone->id == NF_CT_DEFAULT_ZONE_ID || zone->dir != dir) 157 return 0; 158 if (nla_put_be16(skb, attrtype, htons(zone->id))) 159 goto nla_put_failure; 160 return 0; 161 162 nla_put_failure: 163 return -1; 164 } 165 166 static int ctnetlink_dump_status(struct sk_buff *skb, const struct nf_conn *ct) 167 { 168 if (nla_put_be32(skb, CTA_STATUS, htonl(ct->status))) 169 goto nla_put_failure; 170 return 0; 171 172 nla_put_failure: 173 return -1; 174 } 175 176 static int ctnetlink_dump_timeout(struct sk_buff *skb, const struct nf_conn *ct, 177 bool skip_zero) 178 { 179 long timeout; 180 181 if (nf_ct_is_confirmed(ct)) 182 timeout = nf_ct_expires(ct) / HZ; 183 else 184 timeout = ct->timeout / HZ; 185 186 if (skip_zero && timeout == 0) 187 return 0; 188 189 if (nla_put_be32(skb, CTA_TIMEOUT, htonl(timeout))) 190 goto nla_put_failure; 191 return 0; 192 193 nla_put_failure: 194 return -1; 195 } 196 197 static int ctnetlink_dump_protoinfo(struct sk_buff *skb, struct nf_conn *ct, 198 bool destroy) 199 { 200 const struct nf_conntrack_l4proto *l4proto; 201 struct nlattr *nest_proto; 202 int ret; 203 204 l4proto = nf_ct_l4proto_find(nf_ct_protonum(ct)); 205 if (!l4proto->to_nlattr) 206 return 0; 207 208 nest_proto = nla_nest_start(skb, CTA_PROTOINFO); 209 if (!nest_proto) 210 goto nla_put_failure; 211 212 ret = l4proto->to_nlattr(skb, nest_proto, ct, destroy); 213 214 nla_nest_end(skb, nest_proto); 215 216 return ret; 217 218 nla_put_failure: 219 return -1; 220 } 221 222 static int ctnetlink_dump_helpinfo(struct sk_buff *skb, 223 const struct nf_conn *ct) 224 { 225 struct nlattr *nest_helper; 226 const struct nf_conn_help *help = nfct_help(ct); 227 struct nf_conntrack_helper *helper; 228 229 if (!help) 230 return 0; 231 232 rcu_read_lock(); 233 helper = rcu_dereference(help->helper); 234 if (!helper) 235 goto out; 236 237 nest_helper = nla_nest_start(skb, CTA_HELP); 238 if (!nest_helper) 239 goto nla_put_failure; 240 if (nla_put_string(skb, CTA_HELP_NAME, helper->name)) 241 goto nla_put_failure; 242 243 if (rcu_access_pointer(helper->help) && 244 helper->to_nlattr) 245 helper->to_nlattr(skb, ct); 246 247 nla_nest_end(skb, nest_helper); 248 out: 249 rcu_read_unlock(); 250 return 0; 251 252 nla_put_failure: 253 rcu_read_unlock(); 254 return -1; 255 } 256 257 static int 258 dump_counters(struct sk_buff *skb, struct nf_conn_acct *acct, 259 enum ip_conntrack_dir dir, int type) 260 { 261 enum ctattr_type attr = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG; 262 struct nf_conn_counter *counter = acct->counter; 263 struct nlattr *nest_count; 264 u64 pkts, bytes; 265 266 if (type == IPCTNL_MSG_CT_GET_CTRZERO) { 267 pkts = atomic64_xchg(&counter[dir].packets, 0); 268 bytes = atomic64_xchg(&counter[dir].bytes, 0); 269 } else { 270 pkts = atomic64_read(&counter[dir].packets); 271 bytes = atomic64_read(&counter[dir].bytes); 272 } 273 274 nest_count = nla_nest_start(skb, attr); 275 if (!nest_count) 276 goto nla_put_failure; 277 278 if (nla_put_be64(skb, CTA_COUNTERS_PACKETS, cpu_to_be64(pkts), 279 CTA_COUNTERS_PAD) || 280 nla_put_be64(skb, CTA_COUNTERS_BYTES, cpu_to_be64(bytes), 281 CTA_COUNTERS_PAD)) 282 goto nla_put_failure; 283 284 nla_nest_end(skb, nest_count); 285 286 return 0; 287 288 nla_put_failure: 289 return -1; 290 } 291 292 static int 293 ctnetlink_dump_acct(struct sk_buff *skb, const struct nf_conn *ct, int type) 294 { 295 struct nf_conn_acct *acct = nf_conn_acct_find(ct); 296 297 if (!acct) 298 return 0; 299 300 if (dump_counters(skb, acct, IP_CT_DIR_ORIGINAL, type) < 0) 301 return -1; 302 if (dump_counters(skb, acct, IP_CT_DIR_REPLY, type) < 0) 303 return -1; 304 305 return 0; 306 } 307 308 static int 309 ctnetlink_dump_timestamp(struct sk_buff *skb, const struct nf_conn *ct) 310 { 311 struct nlattr *nest_count; 312 const struct nf_conn_tstamp *tstamp; 313 314 tstamp = nf_conn_tstamp_find(ct); 315 if (!tstamp) 316 return 0; 317 318 nest_count = nla_nest_start(skb, CTA_TIMESTAMP); 319 if (!nest_count) 320 goto nla_put_failure; 321 322 if (nla_put_be64(skb, CTA_TIMESTAMP_START, cpu_to_be64(tstamp->start), 323 CTA_TIMESTAMP_PAD) || 324 (tstamp->stop != 0 && nla_put_be64(skb, CTA_TIMESTAMP_STOP, 325 cpu_to_be64(tstamp->stop), 326 CTA_TIMESTAMP_PAD))) 327 goto nla_put_failure; 328 nla_nest_end(skb, nest_count); 329 330 return 0; 331 332 nla_put_failure: 333 return -1; 334 } 335 336 #ifdef CONFIG_NF_CONNTRACK_MARK 337 static int ctnetlink_dump_mark(struct sk_buff *skb, const struct nf_conn *ct, 338 bool dump) 339 { 340 u32 mark = READ_ONCE(ct->mark); 341 342 if (!mark && !dump) 343 return 0; 344 345 if (nla_put_be32(skb, CTA_MARK, htonl(mark))) 346 goto nla_put_failure; 347 return 0; 348 349 nla_put_failure: 350 return -1; 351 } 352 #else 353 #define ctnetlink_dump_mark(a, b, c) (0) 354 #endif 355 356 #ifdef CONFIG_NF_CONNTRACK_SECMARK 357 static int ctnetlink_dump_secctx(struct sk_buff *skb, const struct nf_conn *ct) 358 { 359 struct nlattr *nest_secctx; 360 struct lsm_context ctx; 361 int ret; 362 363 ret = security_secid_to_secctx(ct->secmark, &ctx); 364 if (ret < 0) 365 return 0; 366 367 ret = -1; 368 nest_secctx = nla_nest_start(skb, CTA_SECCTX); 369 if (!nest_secctx) 370 goto nla_put_failure; 371 372 if (nla_put_string(skb, CTA_SECCTX_NAME, ctx.context)) 373 goto nla_put_failure; 374 nla_nest_end(skb, nest_secctx); 375 376 ret = 0; 377 nla_put_failure: 378 security_release_secctx(&ctx); 379 return ret; 380 } 381 #else 382 #define ctnetlink_dump_secctx(a, b) (0) 383 #endif 384 385 #ifdef CONFIG_NF_CONNTRACK_EVENTS 386 static int 387 ctnetlink_dump_event_timestamp(struct sk_buff *skb, const struct nf_conn *ct) 388 { 389 #ifdef CONFIG_NF_CONNTRACK_TIMESTAMP 390 const struct nf_conntrack_ecache *e = nf_ct_ecache_find(ct); 391 392 if (e) { 393 u64 ts = local64_read(&e->timestamp); 394 395 if (ts) 396 return nla_put_be64(skb, CTA_TIMESTAMP_EVENT, 397 cpu_to_be64(ts), CTA_TIMESTAMP_PAD); 398 } 399 #endif 400 return 0; 401 } 402 403 static inline int ctnetlink_label_size(const struct nf_conn *ct) 404 { 405 struct nf_conn_labels *labels = nf_ct_labels_find(ct); 406 407 if (!labels) 408 return 0; 409 return nla_total_size(sizeof(labels->bits)); 410 } 411 #endif 412 413 static int 414 ctnetlink_dump_labels(struct sk_buff *skb, const struct nf_conn *ct) 415 { 416 struct nf_conn_labels *labels = nf_ct_labels_find(ct); 417 unsigned int i; 418 419 if (!labels) 420 return 0; 421 422 i = 0; 423 do { 424 if (labels->bits[i] != 0) 425 return nla_put(skb, CTA_LABELS, sizeof(labels->bits), 426 labels->bits); 427 i++; 428 } while (i < ARRAY_SIZE(labels->bits)); 429 430 return 0; 431 } 432 433 #define master_tuple(ct) &(ct->master->tuplehash[IP_CT_DIR_ORIGINAL].tuple) 434 435 static int ctnetlink_dump_master(struct sk_buff *skb, const struct nf_conn *ct) 436 { 437 struct nlattr *nest_parms; 438 439 if (!(ct->status & IPS_EXPECTED)) 440 return 0; 441 442 nest_parms = nla_nest_start(skb, CTA_TUPLE_MASTER); 443 if (!nest_parms) 444 goto nla_put_failure; 445 if (ctnetlink_dump_tuples(skb, master_tuple(ct)) < 0) 446 goto nla_put_failure; 447 nla_nest_end(skb, nest_parms); 448 449 return 0; 450 451 nla_put_failure: 452 return -1; 453 } 454 455 static int 456 dump_ct_seq_adj(struct sk_buff *skb, const struct nf_ct_seqadj *seq, int type) 457 { 458 struct nlattr *nest_parms; 459 460 nest_parms = nla_nest_start(skb, type); 461 if (!nest_parms) 462 goto nla_put_failure; 463 464 if (nla_put_be32(skb, CTA_SEQADJ_CORRECTION_POS, 465 htonl(seq->correction_pos)) || 466 nla_put_be32(skb, CTA_SEQADJ_OFFSET_BEFORE, 467 htonl(seq->offset_before)) || 468 nla_put_be32(skb, CTA_SEQADJ_OFFSET_AFTER, 469 htonl(seq->offset_after))) 470 goto nla_put_failure; 471 472 nla_nest_end(skb, nest_parms); 473 474 return 0; 475 476 nla_put_failure: 477 return -1; 478 } 479 480 static int ctnetlink_dump_ct_seq_adj(struct sk_buff *skb, struct nf_conn *ct) 481 { 482 struct nf_conn_seqadj *seqadj = nfct_seqadj(ct); 483 struct nf_ct_seqadj *seq; 484 485 if (!(ct->status & IPS_SEQ_ADJUST) || !seqadj) 486 return 0; 487 488 spin_lock_bh(&ct->lock); 489 seq = &seqadj->seq[IP_CT_DIR_ORIGINAL]; 490 if (dump_ct_seq_adj(skb, seq, CTA_SEQ_ADJ_ORIG) == -1) 491 goto err; 492 493 seq = &seqadj->seq[IP_CT_DIR_REPLY]; 494 if (dump_ct_seq_adj(skb, seq, CTA_SEQ_ADJ_REPLY) == -1) 495 goto err; 496 497 spin_unlock_bh(&ct->lock); 498 return 0; 499 err: 500 spin_unlock_bh(&ct->lock); 501 return -1; 502 } 503 504 static int ctnetlink_dump_ct_synproxy(struct sk_buff *skb, struct nf_conn *ct) 505 { 506 struct nf_conn_synproxy *synproxy = nfct_synproxy(ct); 507 struct nlattr *nest_parms; 508 509 if (!synproxy) 510 return 0; 511 512 nest_parms = nla_nest_start(skb, CTA_SYNPROXY); 513 if (!nest_parms) 514 goto nla_put_failure; 515 516 if (nla_put_be32(skb, CTA_SYNPROXY_ISN, htonl(synproxy->isn)) || 517 nla_put_be32(skb, CTA_SYNPROXY_ITS, htonl(synproxy->its)) || 518 nla_put_be32(skb, CTA_SYNPROXY_TSOFF, htonl(synproxy->tsoff))) 519 goto nla_put_failure; 520 521 nla_nest_end(skb, nest_parms); 522 523 return 0; 524 525 nla_put_failure: 526 return -1; 527 } 528 529 static int ctnetlink_dump_id(struct sk_buff *skb, const struct nf_conn *ct) 530 { 531 __be32 id = (__force __be32)nf_ct_get_id(ct); 532 533 if (nla_put_be32(skb, CTA_ID, id)) 534 goto nla_put_failure; 535 return 0; 536 537 nla_put_failure: 538 return -1; 539 } 540 541 static int ctnetlink_dump_use(struct sk_buff *skb, const struct nf_conn *ct) 542 { 543 if (nla_put_be32(skb, CTA_USE, htonl(refcount_read(&ct->ct_general.use)))) 544 goto nla_put_failure; 545 return 0; 546 547 nla_put_failure: 548 return -1; 549 } 550 551 /* all these functions access ct->ext. Caller must either hold a reference 552 * on ct or prevent its deletion by holding either the bucket spinlock or 553 * pcpu dying list lock. 554 */ 555 static int ctnetlink_dump_extinfo(struct sk_buff *skb, 556 struct nf_conn *ct, u32 type) 557 { 558 if (ctnetlink_dump_acct(skb, ct, type) < 0 || 559 ctnetlink_dump_timestamp(skb, ct) < 0 || 560 ctnetlink_dump_helpinfo(skb, ct) < 0 || 561 ctnetlink_dump_labels(skb, ct) < 0 || 562 ctnetlink_dump_ct_seq_adj(skb, ct) < 0 || 563 ctnetlink_dump_ct_synproxy(skb, ct) < 0) 564 return -1; 565 566 return 0; 567 } 568 569 static int ctnetlink_dump_info(struct sk_buff *skb, struct nf_conn *ct) 570 { 571 if (ctnetlink_dump_status(skb, ct) < 0 || 572 ctnetlink_dump_mark(skb, ct, true) < 0 || 573 ctnetlink_dump_secctx(skb, ct) < 0 || 574 ctnetlink_dump_id(skb, ct) < 0 || 575 ctnetlink_dump_use(skb, ct) < 0 || 576 ctnetlink_dump_master(skb, ct) < 0) 577 return -1; 578 579 if (!test_bit(IPS_OFFLOAD_BIT, &ct->status) && 580 (ctnetlink_dump_timeout(skb, ct, false) < 0 || 581 ctnetlink_dump_protoinfo(skb, ct, false) < 0)) 582 return -1; 583 584 return 0; 585 } 586 587 static int 588 ctnetlink_fill_info(struct sk_buff *skb, u32 portid, u32 seq, u32 type, 589 struct nf_conn *ct, bool extinfo, unsigned int flags) 590 { 591 const struct nf_conntrack_zone *zone; 592 struct nlmsghdr *nlh; 593 struct nlattr *nest_parms; 594 unsigned int event; 595 596 if (portid) 597 flags |= NLM_F_MULTI; 598 event = nfnl_msg_type(NFNL_SUBSYS_CTNETLINK, IPCTNL_MSG_CT_NEW); 599 nlh = nfnl_msg_put(skb, portid, seq, event, flags, nf_ct_l3num(ct), 600 NFNETLINK_V0, 0); 601 if (!nlh) 602 goto nlmsg_failure; 603 604 zone = nf_ct_zone(ct); 605 606 nest_parms = nla_nest_start(skb, CTA_TUPLE_ORIG); 607 if (!nest_parms) 608 goto nla_put_failure; 609 if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_ORIGINAL)) < 0) 610 goto nla_put_failure; 611 if (ctnetlink_dump_zone_id(skb, CTA_TUPLE_ZONE, zone, 612 NF_CT_ZONE_DIR_ORIG) < 0) 613 goto nla_put_failure; 614 nla_nest_end(skb, nest_parms); 615 616 nest_parms = nla_nest_start(skb, CTA_TUPLE_REPLY); 617 if (!nest_parms) 618 goto nla_put_failure; 619 if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_REPLY)) < 0) 620 goto nla_put_failure; 621 if (ctnetlink_dump_zone_id(skb, CTA_TUPLE_ZONE, zone, 622 NF_CT_ZONE_DIR_REPL) < 0) 623 goto nla_put_failure; 624 nla_nest_end(skb, nest_parms); 625 626 if (ctnetlink_dump_zone_id(skb, CTA_ZONE, zone, 627 NF_CT_DEFAULT_ZONE_DIR) < 0) 628 goto nla_put_failure; 629 630 if (ctnetlink_dump_info(skb, ct) < 0) 631 goto nla_put_failure; 632 if (extinfo && ctnetlink_dump_extinfo(skb, ct, type) < 0) 633 goto nla_put_failure; 634 635 nlmsg_end(skb, nlh); 636 return skb->len; 637 638 nlmsg_failure: 639 nla_put_failure: 640 nlmsg_cancel(skb, nlh); 641 return -1; 642 } 643 644 static const struct nla_policy cta_ip_nla_policy[CTA_IP_MAX + 1] = { 645 [CTA_IP_V4_SRC] = { .type = NLA_U32 }, 646 [CTA_IP_V4_DST] = { .type = NLA_U32 }, 647 [CTA_IP_V6_SRC] = { .len = sizeof(__be32) * 4 }, 648 [CTA_IP_V6_DST] = { .len = sizeof(__be32) * 4 }, 649 }; 650 651 #if defined(CONFIG_NETFILTER_NETLINK_GLUE_CT) || defined(CONFIG_NF_CONNTRACK_EVENTS) 652 static size_t ctnetlink_proto_size(const struct nf_conn *ct) 653 { 654 const struct nf_conntrack_l4proto *l4proto; 655 size_t len, len4 = 0; 656 657 len = nla_policy_len(cta_ip_nla_policy, CTA_IP_MAX + 1); 658 len *= 3u; /* ORIG, REPLY, MASTER */ 659 660 l4proto = nf_ct_l4proto_find(nf_ct_protonum(ct)); 661 len += l4proto->nlattr_size; 662 if (l4proto->nlattr_tuple_size) { 663 len4 = l4proto->nlattr_tuple_size(); 664 len4 *= 3u; /* ORIG, REPLY, MASTER */ 665 } 666 667 return len + len4; 668 } 669 670 static inline size_t ctnetlink_acct_size(const struct nf_conn *ct) 671 { 672 if (!nf_ct_ext_exist(ct, NF_CT_EXT_ACCT)) 673 return 0; 674 return 2 * nla_total_size(0) /* CTA_COUNTERS_ORIG|REPL */ 675 + 2 * nla_total_size_64bit(sizeof(uint64_t)) /* CTA_COUNTERS_PACKETS */ 676 + 2 * nla_total_size_64bit(sizeof(uint64_t)) /* CTA_COUNTERS_BYTES */ 677 ; 678 } 679 680 static inline int ctnetlink_secctx_size(const struct nf_conn *ct) 681 { 682 #ifdef CONFIG_NF_CONNTRACK_SECMARK 683 int ret; 684 685 ret = security_secid_to_secctx(ct->secmark, NULL); 686 if (ret < 0) 687 return 0; 688 689 return nla_total_size(0) /* CTA_SECCTX */ 690 + nla_total_size(sizeof(char) * ret); /* CTA_SECCTX_NAME */ 691 #else 692 return 0; 693 #endif 694 } 695 696 static inline size_t ctnetlink_timestamp_size(const struct nf_conn *ct) 697 { 698 #ifdef CONFIG_NF_CONNTRACK_TIMESTAMP 699 if (!nf_ct_ext_exist(ct, NF_CT_EXT_TSTAMP)) 700 return 0; 701 return nla_total_size(0) + 2 * nla_total_size_64bit(sizeof(uint64_t)); 702 #else 703 return 0; 704 #endif 705 } 706 #endif 707 708 #ifdef CONFIG_NF_CONNTRACK_EVENTS 709 static size_t ctnetlink_nlmsg_size(const struct nf_conn *ct) 710 { 711 return NLMSG_ALIGN(sizeof(struct nfgenmsg)) 712 + 3 * nla_total_size(0) /* CTA_TUPLE_ORIG|REPL|MASTER */ 713 + 3 * nla_total_size(0) /* CTA_TUPLE_IP */ 714 + 3 * nla_total_size(0) /* CTA_TUPLE_PROTO */ 715 + 3 * nla_total_size(sizeof(u_int8_t)) /* CTA_PROTO_NUM */ 716 + nla_total_size(sizeof(u_int32_t)) /* CTA_ID */ 717 + nla_total_size(sizeof(u_int32_t)) /* CTA_STATUS */ 718 + ctnetlink_acct_size(ct) 719 + ctnetlink_timestamp_size(ct) 720 + nla_total_size(sizeof(u_int32_t)) /* CTA_TIMEOUT */ 721 + nla_total_size(0) /* CTA_PROTOINFO */ 722 + nla_total_size(0) /* CTA_HELP */ 723 + nla_total_size(NF_CT_HELPER_NAME_LEN) /* CTA_HELP_NAME */ 724 + ctnetlink_secctx_size(ct) 725 #if IS_ENABLED(CONFIG_NF_NAT) 726 + 2 * nla_total_size(0) /* CTA_NAT_SEQ_ADJ_ORIG|REPL */ 727 + 6 * nla_total_size(sizeof(u_int32_t)) /* CTA_NAT_SEQ_OFFSET */ 728 #endif 729 #ifdef CONFIG_NF_CONNTRACK_MARK 730 + nla_total_size(sizeof(u_int32_t)) /* CTA_MARK */ 731 #endif 732 #ifdef CONFIG_NF_CONNTRACK_ZONES 733 + nla_total_size(sizeof(u_int16_t)) /* CTA_ZONE|CTA_TUPLE_ZONE */ 734 #endif 735 + ctnetlink_proto_size(ct) 736 + ctnetlink_label_size(ct) 737 #ifdef CONFIG_NF_CONNTRACK_TIMESTAMP 738 + nla_total_size(sizeof(u64)) /* CTA_TIMESTAMP_EVENT */ 739 #endif 740 ; 741 } 742 743 static int 744 ctnetlink_conntrack_event(unsigned int events, const struct nf_ct_event *item) 745 { 746 const struct nf_conntrack_zone *zone; 747 struct net *net; 748 struct nlmsghdr *nlh; 749 struct nlattr *nest_parms; 750 struct nf_conn *ct = item->ct; 751 struct sk_buff *skb; 752 unsigned int type; 753 unsigned int flags = 0, group; 754 int err; 755 756 if (events & (1 << IPCT_DESTROY)) { 757 type = IPCTNL_MSG_CT_DELETE; 758 group = NFNLGRP_CONNTRACK_DESTROY; 759 } else if (events & ((1 << IPCT_NEW) | (1 << IPCT_RELATED))) { 760 type = IPCTNL_MSG_CT_NEW; 761 flags = NLM_F_CREATE|NLM_F_EXCL; 762 group = NFNLGRP_CONNTRACK_NEW; 763 } else if (events) { 764 type = IPCTNL_MSG_CT_NEW; 765 group = NFNLGRP_CONNTRACK_UPDATE; 766 } else 767 return 0; 768 769 net = nf_ct_net(ct); 770 if (!item->report && !nfnetlink_has_listeners(net, group)) 771 return 0; 772 773 skb = nlmsg_new(ctnetlink_nlmsg_size(ct), GFP_ATOMIC); 774 if (skb == NULL) 775 goto errout; 776 777 type = nfnl_msg_type(NFNL_SUBSYS_CTNETLINK, type); 778 nlh = nfnl_msg_put(skb, item->portid, 0, type, flags, nf_ct_l3num(ct), 779 NFNETLINK_V0, 0); 780 if (!nlh) 781 goto nlmsg_failure; 782 783 zone = nf_ct_zone(ct); 784 785 nest_parms = nla_nest_start(skb, CTA_TUPLE_ORIG); 786 if (!nest_parms) 787 goto nla_put_failure; 788 if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_ORIGINAL)) < 0) 789 goto nla_put_failure; 790 if (ctnetlink_dump_zone_id(skb, CTA_TUPLE_ZONE, zone, 791 NF_CT_ZONE_DIR_ORIG) < 0) 792 goto nla_put_failure; 793 nla_nest_end(skb, nest_parms); 794 795 nest_parms = nla_nest_start(skb, CTA_TUPLE_REPLY); 796 if (!nest_parms) 797 goto nla_put_failure; 798 if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_REPLY)) < 0) 799 goto nla_put_failure; 800 if (ctnetlink_dump_zone_id(skb, CTA_TUPLE_ZONE, zone, 801 NF_CT_ZONE_DIR_REPL) < 0) 802 goto nla_put_failure; 803 nla_nest_end(skb, nest_parms); 804 805 if (ctnetlink_dump_zone_id(skb, CTA_ZONE, zone, 806 NF_CT_DEFAULT_ZONE_DIR) < 0) 807 goto nla_put_failure; 808 809 if (ctnetlink_dump_id(skb, ct) < 0) 810 goto nla_put_failure; 811 812 if (ctnetlink_dump_status(skb, ct) < 0) 813 goto nla_put_failure; 814 815 if (events & (1 << IPCT_DESTROY)) { 816 if (ctnetlink_dump_timeout(skb, ct, true) < 0) 817 goto nla_put_failure; 818 819 if (ctnetlink_dump_acct(skb, ct, type) < 0 || 820 ctnetlink_dump_timestamp(skb, ct) < 0 || 821 ctnetlink_dump_protoinfo(skb, ct, true) < 0) 822 goto nla_put_failure; 823 } else { 824 if (ctnetlink_dump_timeout(skb, ct, false) < 0) 825 goto nla_put_failure; 826 827 if (events & (1 << IPCT_PROTOINFO) && 828 ctnetlink_dump_protoinfo(skb, ct, false) < 0) 829 goto nla_put_failure; 830 831 if ((events & (1 << IPCT_HELPER) || nfct_help(ct)) 832 && ctnetlink_dump_helpinfo(skb, ct) < 0) 833 goto nla_put_failure; 834 835 #ifdef CONFIG_NF_CONNTRACK_SECMARK 836 if ((events & (1 << IPCT_SECMARK) || ct->secmark) 837 && ctnetlink_dump_secctx(skb, ct) < 0) 838 goto nla_put_failure; 839 #endif 840 if (events & (1 << IPCT_LABEL) && 841 ctnetlink_dump_labels(skb, ct) < 0) 842 goto nla_put_failure; 843 844 if (events & (1 << IPCT_RELATED) && 845 ctnetlink_dump_master(skb, ct) < 0) 846 goto nla_put_failure; 847 848 if (events & (1 << IPCT_SEQADJ) && 849 ctnetlink_dump_ct_seq_adj(skb, ct) < 0) 850 goto nla_put_failure; 851 852 if (events & (1 << IPCT_SYNPROXY) && 853 ctnetlink_dump_ct_synproxy(skb, ct) < 0) 854 goto nla_put_failure; 855 } 856 857 #ifdef CONFIG_NF_CONNTRACK_MARK 858 if (ctnetlink_dump_mark(skb, ct, events & (1 << IPCT_MARK))) 859 goto nla_put_failure; 860 #endif 861 862 if (ctnetlink_dump_event_timestamp(skb, ct)) 863 goto nla_put_failure; 864 865 nlmsg_end(skb, nlh); 866 err = nfnetlink_send(skb, net, item->portid, group, item->report, 867 GFP_ATOMIC); 868 if (err == -ENOBUFS || err == -EAGAIN) 869 return -ENOBUFS; 870 871 return 0; 872 873 nla_put_failure: 874 nlmsg_cancel(skb, nlh); 875 nlmsg_failure: 876 kfree_skb(skb); 877 errout: 878 if (nfnetlink_set_err(net, 0, group, -ENOBUFS) > 0) 879 return -ENOBUFS; 880 881 return 0; 882 } 883 #endif /* CONFIG_NF_CONNTRACK_EVENTS */ 884 885 static int ctnetlink_done(struct netlink_callback *cb) 886 { 887 kfree(cb->data); 888 return 0; 889 } 890 891 struct ctnetlink_filter_u32 { 892 u32 val; 893 u32 mask; 894 }; 895 896 struct ctnetlink_filter { 897 u8 family; 898 bool zone_filter; 899 900 u_int32_t orig_flags; 901 u_int32_t reply_flags; 902 903 struct nf_conntrack_tuple orig; 904 struct nf_conntrack_tuple reply; 905 struct nf_conntrack_zone zone; 906 907 struct ctnetlink_filter_u32 mark; 908 struct ctnetlink_filter_u32 status; 909 }; 910 911 static const struct nla_policy cta_filter_nla_policy[CTA_FILTER_MAX + 1] = { 912 [CTA_FILTER_ORIG_FLAGS] = NLA_POLICY_MASK(NLA_U32, CTA_FILTER_F_ALL), 913 [CTA_FILTER_REPLY_FLAGS] = NLA_POLICY_MASK(NLA_U32, CTA_FILTER_F_ALL), 914 }; 915 916 static int ctnetlink_parse_filter(const struct nlattr *attr, 917 struct ctnetlink_filter *filter) 918 { 919 struct nlattr *tb[CTA_FILTER_MAX + 1]; 920 int ret = 0; 921 922 ret = nla_parse_nested(tb, CTA_FILTER_MAX, attr, cta_filter_nla_policy, 923 NULL); 924 if (ret) 925 return ret; 926 927 if (tb[CTA_FILTER_ORIG_FLAGS]) 928 filter->orig_flags = nla_get_u32(tb[CTA_FILTER_ORIG_FLAGS]); 929 930 if (tb[CTA_FILTER_REPLY_FLAGS]) 931 filter->reply_flags = nla_get_u32(tb[CTA_FILTER_REPLY_FLAGS]); 932 933 return 0; 934 } 935 936 static int ctnetlink_parse_zone(const struct nlattr *attr, 937 struct nf_conntrack_zone *zone); 938 static int ctnetlink_parse_tuple_filter(const struct nlattr * const cda[], 939 struct nf_conntrack_tuple *tuple, 940 u32 type, u_int8_t l3num, 941 struct nf_conntrack_zone *zone, 942 u_int32_t flags); 943 944 static int ctnetlink_filter_parse_mark(struct ctnetlink_filter_u32 *mark, 945 const struct nlattr * const cda[]) 946 { 947 #ifdef CONFIG_NF_CONNTRACK_MARK 948 if (cda[CTA_MARK]) { 949 mark->val = ntohl(nla_get_be32(cda[CTA_MARK])); 950 951 if (cda[CTA_MARK_MASK]) 952 mark->mask = ntohl(nla_get_be32(cda[CTA_MARK_MASK])); 953 else 954 mark->mask = 0xffffffff; 955 } else if (cda[CTA_MARK_MASK]) { 956 return -EINVAL; 957 } 958 #endif 959 return 0; 960 } 961 962 static int ctnetlink_filter_parse_status(struct ctnetlink_filter_u32 *status, 963 const struct nlattr * const cda[]) 964 { 965 if (cda[CTA_STATUS]) { 966 status->val = ntohl(nla_get_be32(cda[CTA_STATUS])); 967 if (cda[CTA_STATUS_MASK]) 968 status->mask = ntohl(nla_get_be32(cda[CTA_STATUS_MASK])); 969 else 970 status->mask = status->val; 971 972 /* status->val == 0? always true, else always false. */ 973 if (status->mask == 0) 974 return -EINVAL; 975 } else if (cda[CTA_STATUS_MASK]) { 976 return -EINVAL; 977 } 978 979 /* CTA_STATUS is NLA_U32, if this fires UAPI needs to be extended */ 980 BUILD_BUG_ON(__IPS_MAX_BIT >= 32); 981 return 0; 982 } 983 984 static struct ctnetlink_filter * 985 ctnetlink_alloc_filter(const struct nlattr * const cda[], u8 family) 986 { 987 struct ctnetlink_filter *filter; 988 int err; 989 990 #ifndef CONFIG_NF_CONNTRACK_MARK 991 if (cda[CTA_MARK] || cda[CTA_MARK_MASK]) 992 return ERR_PTR(-EOPNOTSUPP); 993 #endif 994 995 filter = kzalloc_obj(*filter); 996 if (filter == NULL) 997 return ERR_PTR(-ENOMEM); 998 999 filter->family = family; 1000 1001 err = ctnetlink_filter_parse_mark(&filter->mark, cda); 1002 if (err) 1003 goto err_filter; 1004 1005 err = ctnetlink_filter_parse_status(&filter->status, cda); 1006 if (err) 1007 goto err_filter; 1008 1009 if (cda[CTA_ZONE]) { 1010 err = ctnetlink_parse_zone(cda[CTA_ZONE], &filter->zone); 1011 if (err < 0) 1012 goto err_filter; 1013 filter->zone_filter = true; 1014 } 1015 1016 if (!cda[CTA_FILTER]) 1017 return filter; 1018 1019 err = ctnetlink_parse_filter(cda[CTA_FILTER], filter); 1020 if (err < 0) 1021 goto err_filter; 1022 1023 if (filter->orig_flags) { 1024 if (!cda[CTA_TUPLE_ORIG]) { 1025 err = -EINVAL; 1026 goto err_filter; 1027 } 1028 1029 err = ctnetlink_parse_tuple_filter(cda, &filter->orig, 1030 CTA_TUPLE_ORIG, 1031 filter->family, 1032 &filter->zone, 1033 filter->orig_flags); 1034 if (err < 0) 1035 goto err_filter; 1036 } 1037 1038 if (filter->reply_flags) { 1039 if (!cda[CTA_TUPLE_REPLY]) { 1040 err = -EINVAL; 1041 goto err_filter; 1042 } 1043 1044 err = ctnetlink_parse_tuple_filter(cda, &filter->reply, 1045 CTA_TUPLE_REPLY, 1046 filter->family, 1047 &filter->zone, 1048 filter->reply_flags); 1049 if (err < 0) 1050 goto err_filter; 1051 } 1052 1053 return filter; 1054 1055 err_filter: 1056 kfree(filter); 1057 1058 return ERR_PTR(err); 1059 } 1060 1061 static bool ctnetlink_needs_filter(u8 family, const struct nlattr * const *cda) 1062 { 1063 return family || cda[CTA_MARK] || cda[CTA_FILTER] || cda[CTA_STATUS] || cda[CTA_ZONE]; 1064 } 1065 1066 static int ctnetlink_start(struct netlink_callback *cb) 1067 { 1068 const struct nlattr * const *cda = cb->data; 1069 struct ctnetlink_filter *filter = NULL; 1070 struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh); 1071 u8 family = nfmsg->nfgen_family; 1072 1073 if (ctnetlink_needs_filter(family, cda)) { 1074 filter = ctnetlink_alloc_filter(cda, family); 1075 if (IS_ERR(filter)) 1076 return PTR_ERR(filter); 1077 } 1078 1079 cb->data = filter; 1080 return 0; 1081 } 1082 1083 static int ctnetlink_filter_match_tuple(struct nf_conntrack_tuple *filter_tuple, 1084 struct nf_conntrack_tuple *ct_tuple, 1085 u_int32_t flags, int family) 1086 { 1087 switch (family) { 1088 case NFPROTO_IPV4: 1089 if ((flags & CTA_FILTER_FLAG(CTA_IP_SRC)) && 1090 filter_tuple->src.u3.ip != ct_tuple->src.u3.ip) 1091 return 0; 1092 1093 if ((flags & CTA_FILTER_FLAG(CTA_IP_DST)) && 1094 filter_tuple->dst.u3.ip != ct_tuple->dst.u3.ip) 1095 return 0; 1096 break; 1097 case NFPROTO_IPV6: 1098 if ((flags & CTA_FILTER_FLAG(CTA_IP_SRC)) && 1099 !ipv6_addr_cmp(&filter_tuple->src.u3.in6, 1100 &ct_tuple->src.u3.in6)) 1101 return 0; 1102 1103 if ((flags & CTA_FILTER_FLAG(CTA_IP_DST)) && 1104 !ipv6_addr_cmp(&filter_tuple->dst.u3.in6, 1105 &ct_tuple->dst.u3.in6)) 1106 return 0; 1107 break; 1108 } 1109 1110 if ((flags & CTA_FILTER_FLAG(CTA_PROTO_NUM)) && 1111 filter_tuple->dst.protonum != ct_tuple->dst.protonum) 1112 return 0; 1113 1114 switch (ct_tuple->dst.protonum) { 1115 case IPPROTO_TCP: 1116 case IPPROTO_UDP: 1117 if ((flags & CTA_FILTER_FLAG(CTA_PROTO_SRC_PORT)) && 1118 filter_tuple->src.u.tcp.port != ct_tuple->src.u.tcp.port) 1119 return 0; 1120 1121 if ((flags & CTA_FILTER_FLAG(CTA_PROTO_DST_PORT)) && 1122 filter_tuple->dst.u.tcp.port != ct_tuple->dst.u.tcp.port) 1123 return 0; 1124 break; 1125 case IPPROTO_ICMP: 1126 if ((flags & CTA_FILTER_FLAG(CTA_PROTO_ICMP_TYPE)) && 1127 filter_tuple->dst.u.icmp.type != ct_tuple->dst.u.icmp.type) 1128 return 0; 1129 if ((flags & CTA_FILTER_FLAG(CTA_PROTO_ICMP_CODE)) && 1130 filter_tuple->dst.u.icmp.code != ct_tuple->dst.u.icmp.code) 1131 return 0; 1132 if ((flags & CTA_FILTER_FLAG(CTA_PROTO_ICMP_ID)) && 1133 filter_tuple->src.u.icmp.id != ct_tuple->src.u.icmp.id) 1134 return 0; 1135 break; 1136 case IPPROTO_ICMPV6: 1137 if ((flags & CTA_FILTER_FLAG(CTA_PROTO_ICMPV6_TYPE)) && 1138 filter_tuple->dst.u.icmp.type != ct_tuple->dst.u.icmp.type) 1139 return 0; 1140 if ((flags & CTA_FILTER_FLAG(CTA_PROTO_ICMPV6_CODE)) && 1141 filter_tuple->dst.u.icmp.code != ct_tuple->dst.u.icmp.code) 1142 return 0; 1143 if ((flags & CTA_FILTER_FLAG(CTA_PROTO_ICMPV6_ID)) && 1144 filter_tuple->src.u.icmp.id != ct_tuple->src.u.icmp.id) 1145 return 0; 1146 break; 1147 } 1148 1149 return 1; 1150 } 1151 1152 static int ctnetlink_filter_match(struct nf_conn *ct, void *data) 1153 { 1154 struct ctnetlink_filter *filter = data; 1155 struct nf_conntrack_tuple *tuple; 1156 u32 status; 1157 1158 if (filter == NULL) 1159 goto out; 1160 1161 /* Match entries of a given L3 protocol number. 1162 * If it is not specified, ie. l3proto == 0, 1163 * then match everything. 1164 */ 1165 if (filter->family && nf_ct_l3num(ct) != filter->family) 1166 goto ignore_entry; 1167 1168 if (filter->zone_filter && 1169 !nf_ct_zone_equal_any(ct, &filter->zone)) 1170 goto ignore_entry; 1171 1172 if (filter->orig_flags) { 1173 tuple = nf_ct_tuple(ct, IP_CT_DIR_ORIGINAL); 1174 if (!ctnetlink_filter_match_tuple(&filter->orig, tuple, 1175 filter->orig_flags, 1176 filter->family)) 1177 goto ignore_entry; 1178 } 1179 1180 if (filter->reply_flags) { 1181 tuple = nf_ct_tuple(ct, IP_CT_DIR_REPLY); 1182 if (!ctnetlink_filter_match_tuple(&filter->reply, tuple, 1183 filter->reply_flags, 1184 filter->family)) 1185 goto ignore_entry; 1186 } 1187 1188 #ifdef CONFIG_NF_CONNTRACK_MARK 1189 if ((READ_ONCE(ct->mark) & filter->mark.mask) != filter->mark.val) 1190 goto ignore_entry; 1191 #endif 1192 status = (u32)READ_ONCE(ct->status); 1193 if ((status & filter->status.mask) != filter->status.val) 1194 goto ignore_entry; 1195 1196 out: 1197 return 1; 1198 1199 ignore_entry: 1200 return 0; 1201 } 1202 1203 static unsigned long ctnetlink_get_id(const struct nf_conn *ct) 1204 { 1205 unsigned long id = nf_ct_get_id(ct); 1206 1207 return id ? id : 1; 1208 } 1209 1210 static int 1211 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb) 1212 { 1213 unsigned int flags = cb->data ? NLM_F_DUMP_FILTERED : 0; 1214 struct net *net = sock_net(skb->sk); 1215 unsigned long last_id = cb->args[1]; 1216 struct nf_conntrack_tuple_hash *h; 1217 struct hlist_nulls_node *n; 1218 struct nf_conn *nf_ct_evict[8]; 1219 struct nf_conn *ct; 1220 int res, i; 1221 spinlock_t *lockp; 1222 1223 i = 0; 1224 1225 local_bh_disable(); 1226 for (; cb->args[0] < nf_conntrack_htable_size; cb->args[0]++) { 1227 restart: 1228 while (i) { 1229 i--; 1230 if (nf_ct_should_gc(nf_ct_evict[i])) 1231 nf_ct_kill(nf_ct_evict[i]); 1232 nf_ct_put(nf_ct_evict[i]); 1233 } 1234 1235 lockp = &nf_conntrack_locks[cb->args[0] % CONNTRACK_LOCKS]; 1236 nf_conntrack_lock(lockp); 1237 if (cb->args[0] >= nf_conntrack_htable_size) { 1238 spin_unlock(lockp); 1239 goto out; 1240 } 1241 hlist_nulls_for_each_entry(h, n, &nf_conntrack_hash[cb->args[0]], 1242 hnnode) { 1243 ct = nf_ct_tuplehash_to_ctrack(h); 1244 if (nf_ct_is_expired(ct)) { 1245 /* need to defer nf_ct_kill() until lock is released */ 1246 if (i < ARRAY_SIZE(nf_ct_evict) && 1247 refcount_inc_not_zero(&ct->ct_general.use)) 1248 nf_ct_evict[i++] = ct; 1249 continue; 1250 } 1251 1252 if (!net_eq(net, nf_ct_net(ct))) 1253 continue; 1254 1255 if (NF_CT_DIRECTION(h) != IP_CT_DIR_ORIGINAL) 1256 continue; 1257 1258 if (cb->args[1]) { 1259 if (ctnetlink_get_id(ct) != last_id) 1260 continue; 1261 cb->args[1] = 0; 1262 } 1263 if (!ctnetlink_filter_match(ct, cb->data)) 1264 continue; 1265 1266 res = 1267 ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).portid, 1268 cb->nlh->nlmsg_seq, 1269 NFNL_MSG_TYPE(cb->nlh->nlmsg_type), 1270 ct, true, flags); 1271 if (res < 0) { 1272 cb->args[1] = ctnetlink_get_id(ct); 1273 spin_unlock(lockp); 1274 goto out; 1275 } 1276 } 1277 spin_unlock(lockp); 1278 if (cb->args[1]) { 1279 cb->args[1] = 0; 1280 goto restart; 1281 } 1282 } 1283 out: 1284 local_bh_enable(); 1285 if (last_id) { 1286 /* nf ct hash resize happened, now clear the leftover. */ 1287 if (cb->args[1] == last_id) 1288 cb->args[1] = 0; 1289 } 1290 1291 while (i) { 1292 i--; 1293 if (nf_ct_should_gc(nf_ct_evict[i])) 1294 nf_ct_kill(nf_ct_evict[i]); 1295 nf_ct_put(nf_ct_evict[i]); 1296 } 1297 1298 return skb->len; 1299 } 1300 1301 static int ipv4_nlattr_to_tuple(struct nlattr *tb[], 1302 struct nf_conntrack_tuple *t, 1303 u_int32_t flags) 1304 { 1305 if (flags & CTA_FILTER_FLAG(CTA_IP_SRC)) { 1306 if (!tb[CTA_IP_V4_SRC]) 1307 return -EINVAL; 1308 1309 t->src.u3.ip = nla_get_in_addr(tb[CTA_IP_V4_SRC]); 1310 } 1311 1312 if (flags & CTA_FILTER_FLAG(CTA_IP_DST)) { 1313 if (!tb[CTA_IP_V4_DST]) 1314 return -EINVAL; 1315 1316 t->dst.u3.ip = nla_get_in_addr(tb[CTA_IP_V4_DST]); 1317 } 1318 1319 return 0; 1320 } 1321 1322 static int ipv6_nlattr_to_tuple(struct nlattr *tb[], 1323 struct nf_conntrack_tuple *t, 1324 u_int32_t flags) 1325 { 1326 if (flags & CTA_FILTER_FLAG(CTA_IP_SRC)) { 1327 if (!tb[CTA_IP_V6_SRC]) 1328 return -EINVAL; 1329 1330 t->src.u3.in6 = nla_get_in6_addr(tb[CTA_IP_V6_SRC]); 1331 } 1332 1333 if (flags & CTA_FILTER_FLAG(CTA_IP_DST)) { 1334 if (!tb[CTA_IP_V6_DST]) 1335 return -EINVAL; 1336 1337 t->dst.u3.in6 = nla_get_in6_addr(tb[CTA_IP_V6_DST]); 1338 } 1339 1340 return 0; 1341 } 1342 1343 static int ctnetlink_parse_tuple_ip(struct nlattr *attr, 1344 struct nf_conntrack_tuple *tuple, 1345 u_int32_t flags) 1346 { 1347 struct nlattr *tb[CTA_IP_MAX+1]; 1348 int ret = 0; 1349 1350 ret = nla_parse_nested_deprecated(tb, CTA_IP_MAX, attr, 1351 cta_ip_nla_policy, NULL); 1352 if (ret < 0) 1353 return ret; 1354 1355 switch (tuple->src.l3num) { 1356 case NFPROTO_IPV4: 1357 ret = ipv4_nlattr_to_tuple(tb, tuple, flags); 1358 break; 1359 case NFPROTO_IPV6: 1360 ret = ipv6_nlattr_to_tuple(tb, tuple, flags); 1361 break; 1362 } 1363 1364 return ret; 1365 } 1366 1367 static const struct nla_policy proto_nla_policy[CTA_PROTO_MAX+1] = { 1368 [CTA_PROTO_NUM] = { .type = NLA_U8 }, 1369 }; 1370 1371 static int ctnetlink_parse_tuple_proto(struct nlattr *attr, 1372 struct nf_conntrack_tuple *tuple, 1373 u_int32_t flags) 1374 { 1375 const struct nf_conntrack_l4proto *l4proto; 1376 struct nlattr *tb[CTA_PROTO_MAX+1]; 1377 int ret = 0; 1378 1379 ret = nla_parse_nested_deprecated(tb, CTA_PROTO_MAX, attr, 1380 proto_nla_policy, NULL); 1381 if (ret < 0) 1382 return ret; 1383 1384 if (!(flags & CTA_FILTER_FLAG(CTA_PROTO_NUM))) 1385 return 0; 1386 1387 if (!tb[CTA_PROTO_NUM]) 1388 return -EINVAL; 1389 1390 tuple->dst.protonum = nla_get_u8(tb[CTA_PROTO_NUM]); 1391 1392 rcu_read_lock(); 1393 l4proto = nf_ct_l4proto_find(tuple->dst.protonum); 1394 1395 if (likely(l4proto->nlattr_to_tuple)) { 1396 ret = nla_validate_nested_deprecated(attr, CTA_PROTO_MAX, 1397 l4proto->nla_policy, 1398 NULL); 1399 if (ret == 0) 1400 ret = l4proto->nlattr_to_tuple(tb, tuple, flags); 1401 } 1402 1403 rcu_read_unlock(); 1404 1405 return ret; 1406 } 1407 1408 static int 1409 ctnetlink_parse_zone(const struct nlattr *attr, 1410 struct nf_conntrack_zone *zone) 1411 { 1412 nf_ct_zone_init(zone, NF_CT_DEFAULT_ZONE_ID, 1413 NF_CT_DEFAULT_ZONE_DIR, 0); 1414 #ifdef CONFIG_NF_CONNTRACK_ZONES 1415 if (attr) 1416 zone->id = ntohs(nla_get_be16(attr)); 1417 #else 1418 if (attr) 1419 return -EOPNOTSUPP; 1420 #endif 1421 return 0; 1422 } 1423 1424 static int 1425 ctnetlink_parse_tuple_zone(struct nlattr *attr, enum ctattr_type type, 1426 struct nf_conntrack_zone *zone) 1427 { 1428 int ret; 1429 1430 if (zone->id != NF_CT_DEFAULT_ZONE_ID) 1431 return -EINVAL; 1432 1433 ret = ctnetlink_parse_zone(attr, zone); 1434 if (ret < 0) 1435 return ret; 1436 1437 if (type == CTA_TUPLE_REPLY) 1438 zone->dir = NF_CT_ZONE_DIR_REPL; 1439 else 1440 zone->dir = NF_CT_ZONE_DIR_ORIG; 1441 1442 return 0; 1443 } 1444 1445 static const struct nla_policy tuple_nla_policy[CTA_TUPLE_MAX+1] = { 1446 [CTA_TUPLE_IP] = { .type = NLA_NESTED }, 1447 [CTA_TUPLE_PROTO] = { .type = NLA_NESTED }, 1448 [CTA_TUPLE_ZONE] = { .type = NLA_U16 }, 1449 }; 1450 1451 #define CTA_FILTER_F_ALL_CTA_PROTO \ 1452 (CTA_FILTER_F_CTA_PROTO_SRC_PORT | \ 1453 CTA_FILTER_F_CTA_PROTO_DST_PORT | \ 1454 CTA_FILTER_F_CTA_PROTO_ICMP_TYPE | \ 1455 CTA_FILTER_F_CTA_PROTO_ICMP_CODE | \ 1456 CTA_FILTER_F_CTA_PROTO_ICMP_ID | \ 1457 CTA_FILTER_F_CTA_PROTO_ICMPV6_TYPE | \ 1458 CTA_FILTER_F_CTA_PROTO_ICMPV6_CODE | \ 1459 CTA_FILTER_F_CTA_PROTO_ICMPV6_ID) 1460 1461 static int 1462 ctnetlink_parse_tuple_filter(const struct nlattr * const cda[], 1463 struct nf_conntrack_tuple *tuple, u32 type, 1464 u_int8_t l3num, struct nf_conntrack_zone *zone, 1465 u_int32_t flags) 1466 { 1467 struct nlattr *tb[CTA_TUPLE_MAX+1]; 1468 int err; 1469 1470 memset(tuple, 0, sizeof(*tuple)); 1471 1472 err = nla_parse_nested_deprecated(tb, CTA_TUPLE_MAX, cda[type], 1473 tuple_nla_policy, NULL); 1474 if (err < 0) 1475 return err; 1476 1477 if (l3num != NFPROTO_IPV4 && l3num != NFPROTO_IPV6) 1478 return -EOPNOTSUPP; 1479 tuple->src.l3num = l3num; 1480 1481 if (flags & CTA_FILTER_FLAG(CTA_IP_DST) || 1482 flags & CTA_FILTER_FLAG(CTA_IP_SRC)) { 1483 if (!tb[CTA_TUPLE_IP]) 1484 return -EINVAL; 1485 1486 err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP], tuple, flags); 1487 if (err < 0) 1488 return err; 1489 } 1490 1491 if (flags & CTA_FILTER_FLAG(CTA_PROTO_NUM)) { 1492 if (!tb[CTA_TUPLE_PROTO]) 1493 return -EINVAL; 1494 1495 err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO], tuple, flags); 1496 if (err < 0) 1497 return err; 1498 } else if (flags & CTA_FILTER_FLAG(ALL_CTA_PROTO)) { 1499 /* Can't manage proto flags without a protonum */ 1500 return -EINVAL; 1501 } 1502 1503 if ((flags & CTA_FILTER_FLAG(CTA_TUPLE_ZONE)) && tb[CTA_TUPLE_ZONE]) { 1504 if (!zone) 1505 return -EINVAL; 1506 1507 err = ctnetlink_parse_tuple_zone(tb[CTA_TUPLE_ZONE], 1508 type, zone); 1509 if (err < 0) 1510 return err; 1511 } 1512 1513 /* orig and expect tuples get DIR_ORIGINAL */ 1514 if (type == CTA_TUPLE_REPLY) 1515 tuple->dst.dir = IP_CT_DIR_REPLY; 1516 else 1517 tuple->dst.dir = IP_CT_DIR_ORIGINAL; 1518 1519 return 0; 1520 } 1521 1522 static int 1523 ctnetlink_parse_tuple(const struct nlattr * const cda[], 1524 struct nf_conntrack_tuple *tuple, u32 type, 1525 u_int8_t l3num, struct nf_conntrack_zone *zone) 1526 { 1527 return ctnetlink_parse_tuple_filter(cda, tuple, type, l3num, zone, 1528 CTA_FILTER_FLAG(ALL)); 1529 } 1530 1531 static const struct nla_policy help_nla_policy[CTA_HELP_MAX+1] = { 1532 [CTA_HELP_NAME] = { .type = NLA_NUL_STRING, 1533 .len = NF_CT_HELPER_NAME_LEN - 1 }, 1534 }; 1535 1536 static int ctnetlink_parse_help(const struct nlattr *attr, char **helper_name, 1537 struct nlattr **helpinfo) 1538 { 1539 int err; 1540 struct nlattr *tb[CTA_HELP_MAX+1]; 1541 1542 err = nla_parse_nested_deprecated(tb, CTA_HELP_MAX, attr, 1543 help_nla_policy, NULL); 1544 if (err < 0) 1545 return err; 1546 1547 if (!tb[CTA_HELP_NAME]) 1548 return -EINVAL; 1549 1550 *helper_name = nla_data(tb[CTA_HELP_NAME]); 1551 1552 if (tb[CTA_HELP_INFO]) 1553 *helpinfo = tb[CTA_HELP_INFO]; 1554 1555 return 0; 1556 } 1557 1558 static const struct nla_policy ct_nla_policy[CTA_MAX+1] = { 1559 [CTA_TUPLE_ORIG] = { .type = NLA_NESTED }, 1560 [CTA_TUPLE_REPLY] = { .type = NLA_NESTED }, 1561 [CTA_STATUS] = { .type = NLA_U32 }, 1562 [CTA_PROTOINFO] = { .type = NLA_NESTED }, 1563 [CTA_HELP] = { .type = NLA_NESTED }, 1564 [CTA_NAT_SRC] = { .type = NLA_NESTED }, 1565 [CTA_TIMEOUT] = { .type = NLA_U32 }, 1566 [CTA_MARK] = { .type = NLA_U32 }, 1567 [CTA_ID] = { .type = NLA_U32 }, 1568 [CTA_NAT_DST] = { .type = NLA_NESTED }, 1569 [CTA_TUPLE_MASTER] = { .type = NLA_NESTED }, 1570 [CTA_NAT_SEQ_ADJ_ORIG] = { .type = NLA_NESTED }, 1571 [CTA_NAT_SEQ_ADJ_REPLY] = { .type = NLA_NESTED }, 1572 [CTA_ZONE] = { .type = NLA_U16 }, 1573 [CTA_MARK_MASK] = { .type = NLA_U32 }, 1574 [CTA_LABELS] = { .type = NLA_BINARY, 1575 .len = NF_CT_LABELS_MAX_SIZE }, 1576 [CTA_LABELS_MASK] = { .type = NLA_BINARY, 1577 .len = NF_CT_LABELS_MAX_SIZE }, 1578 [CTA_FILTER] = { .type = NLA_NESTED }, 1579 [CTA_STATUS_MASK] = { .type = NLA_U32 }, 1580 [CTA_TIMESTAMP_EVENT] = { .type = NLA_REJECT }, 1581 }; 1582 1583 static int ctnetlink_flush_iterate(struct nf_conn *ct, void *data) 1584 { 1585 return ctnetlink_filter_match(ct, data); 1586 } 1587 1588 static int ctnetlink_flush_conntrack(struct net *net, 1589 const struct nlattr * const cda[], 1590 u32 portid, int report, u8 family) 1591 { 1592 struct ctnetlink_filter *filter = NULL; 1593 struct nf_ct_iter_data iter = { 1594 .net = net, 1595 .portid = portid, 1596 .report = report, 1597 }; 1598 1599 if (ctnetlink_needs_filter(family, cda)) { 1600 filter = ctnetlink_alloc_filter(cda, family); 1601 if (IS_ERR(filter)) 1602 return PTR_ERR(filter); 1603 1604 iter.data = filter; 1605 } 1606 1607 nf_ct_iterate_cleanup_net(ctnetlink_flush_iterate, &iter); 1608 kfree(filter); 1609 1610 return 0; 1611 } 1612 1613 static int ctnetlink_del_conntrack(struct sk_buff *skb, 1614 const struct nfnl_info *info, 1615 const struct nlattr * const cda[]) 1616 { 1617 u8 family = info->nfmsg->nfgen_family; 1618 struct nf_conntrack_tuple_hash *h; 1619 struct nf_conntrack_tuple tuple; 1620 struct nf_conntrack_zone zone; 1621 struct nf_conn *ct; 1622 int err; 1623 1624 err = ctnetlink_parse_zone(cda[CTA_ZONE], &zone); 1625 if (err < 0) 1626 return err; 1627 1628 if (cda[CTA_TUPLE_ORIG] && !cda[CTA_FILTER]) 1629 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG, 1630 family, &zone); 1631 else if (cda[CTA_TUPLE_REPLY] && !cda[CTA_FILTER]) 1632 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY, 1633 family, &zone); 1634 else { 1635 u8 u3 = info->nfmsg->version || cda[CTA_FILTER] ? family : AF_UNSPEC; 1636 1637 return ctnetlink_flush_conntrack(info->net, cda, 1638 NETLINK_CB(skb).portid, 1639 nlmsg_report(info->nlh), u3); 1640 } 1641 1642 if (err < 0) 1643 return err; 1644 1645 h = nf_conntrack_find_get(info->net, &zone, &tuple); 1646 if (!h) 1647 return -ENOENT; 1648 1649 ct = nf_ct_tuplehash_to_ctrack(h); 1650 1651 if (cda[CTA_ID]) { 1652 __be32 id = nla_get_be32(cda[CTA_ID]); 1653 1654 if (id != (__force __be32)nf_ct_get_id(ct)) { 1655 nf_ct_put(ct); 1656 return -ENOENT; 1657 } 1658 } 1659 1660 nf_ct_delete(ct, NETLINK_CB(skb).portid, nlmsg_report(info->nlh)); 1661 nf_ct_put(ct); 1662 1663 return 0; 1664 } 1665 1666 static int ctnetlink_get_conntrack(struct sk_buff *skb, 1667 const struct nfnl_info *info, 1668 const struct nlattr * const cda[]) 1669 { 1670 u_int8_t u3 = info->nfmsg->nfgen_family; 1671 struct nf_conntrack_tuple_hash *h; 1672 struct nf_conntrack_tuple tuple; 1673 struct nf_conntrack_zone zone; 1674 struct sk_buff *skb2; 1675 struct nf_conn *ct; 1676 int err; 1677 1678 if (info->nlh->nlmsg_flags & NLM_F_DUMP) { 1679 struct netlink_dump_control c = { 1680 .start = ctnetlink_start, 1681 .dump = ctnetlink_dump_table, 1682 .done = ctnetlink_done, 1683 .data = (void *)cda, 1684 }; 1685 1686 return netlink_dump_start(info->sk, skb, info->nlh, &c); 1687 } 1688 1689 err = ctnetlink_parse_zone(cda[CTA_ZONE], &zone); 1690 if (err < 0) 1691 return err; 1692 1693 if (cda[CTA_TUPLE_ORIG]) 1694 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG, 1695 u3, &zone); 1696 else if (cda[CTA_TUPLE_REPLY]) 1697 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY, 1698 u3, &zone); 1699 else 1700 return -EINVAL; 1701 1702 if (err < 0) 1703 return err; 1704 1705 h = nf_conntrack_find_get(info->net, &zone, &tuple); 1706 if (!h) 1707 return -ENOENT; 1708 1709 ct = nf_ct_tuplehash_to_ctrack(h); 1710 1711 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); 1712 if (!skb2) { 1713 nf_ct_put(ct); 1714 return -ENOMEM; 1715 } 1716 1717 err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).portid, 1718 info->nlh->nlmsg_seq, 1719 NFNL_MSG_TYPE(info->nlh->nlmsg_type), ct, 1720 true, 0); 1721 nf_ct_put(ct); 1722 if (err <= 0) { 1723 kfree_skb(skb2); 1724 return -ENOMEM; 1725 } 1726 1727 return nfnetlink_unicast(skb2, info->net, NETLINK_CB(skb).portid); 1728 } 1729 1730 #ifdef CONFIG_NF_CONNTRACK_EVENTS 1731 static int ctnetlink_dump_one_entry(struct sk_buff *skb, 1732 struct netlink_callback *cb, 1733 struct nf_conn *ct, 1734 bool dying) 1735 { 1736 struct ctnetlink_list_dump_ctx *ctx = (void *)cb->ctx; 1737 struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh); 1738 u8 l3proto = nfmsg->nfgen_family; 1739 int res; 1740 1741 if (l3proto && nf_ct_l3num(ct) != l3proto) 1742 return 0; 1743 1744 if (ctx->last_id) { 1745 if (ctnetlink_get_id(ct) != ctx->last_id) 1746 return 0; 1747 1748 ctx->last_id = 0; 1749 } 1750 1751 /* We can't dump extension info for the unconfirmed 1752 * list because unconfirmed conntracks can have 1753 * ct->ext reallocated (and thus freed). 1754 * 1755 * In the dying list case ct->ext can't be free'd 1756 * until after we drop pcpu->lock. 1757 */ 1758 res = ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).portid, 1759 cb->nlh->nlmsg_seq, 1760 NFNL_MSG_TYPE(cb->nlh->nlmsg_type), 1761 ct, dying, 0); 1762 if (res < 0) 1763 ctx->last_id = ctnetlink_get_id(ct); 1764 1765 return res; 1766 } 1767 #endif 1768 1769 static int 1770 ctnetlink_dump_unconfirmed(struct sk_buff *skb, struct netlink_callback *cb) 1771 { 1772 return 0; 1773 } 1774 1775 static int 1776 ctnetlink_dump_dying(struct sk_buff *skb, struct netlink_callback *cb) 1777 { 1778 struct ctnetlink_list_dump_ctx *ctx = (void *)cb->ctx; 1779 #ifdef CONFIG_NF_CONNTRACK_EVENTS 1780 const struct net *net = sock_net(skb->sk); 1781 struct nf_conntrack_net_ecache *ecache_net; 1782 unsigned long last_id = ctx->last_id; 1783 struct nf_conntrack_tuple_hash *h; 1784 struct hlist_nulls_node *n; 1785 #endif 1786 1787 if (ctx->done) 1788 return 0; 1789 1790 ctx->last_id = 0; 1791 1792 #ifdef CONFIG_NF_CONNTRACK_EVENTS 1793 ecache_net = nf_conn_pernet_ecache(net); 1794 spin_lock_bh(&ecache_net->dying_lock); 1795 1796 hlist_nulls_for_each_entry(h, n, &ecache_net->dying_list, hnnode) { 1797 struct nf_conn *ct; 1798 int res; 1799 1800 ct = nf_ct_tuplehash_to_ctrack(h); 1801 if (last_id && last_id != ctnetlink_get_id(ct)) 1802 continue; 1803 1804 res = ctnetlink_dump_one_entry(skb, cb, ct, true); 1805 if (res < 0) { 1806 spin_unlock_bh(&ecache_net->dying_lock); 1807 return skb->len; 1808 } 1809 1810 last_id = 0; 1811 } 1812 1813 spin_unlock_bh(&ecache_net->dying_lock); 1814 #endif 1815 ctx->done = true; 1816 1817 return skb->len; 1818 } 1819 1820 static int ctnetlink_get_ct_dying(struct sk_buff *skb, 1821 const struct nfnl_info *info, 1822 const struct nlattr * const cda[]) 1823 { 1824 if (info->nlh->nlmsg_flags & NLM_F_DUMP) { 1825 struct netlink_dump_control c = { 1826 .dump = ctnetlink_dump_dying, 1827 }; 1828 return netlink_dump_start(info->sk, skb, info->nlh, &c); 1829 } 1830 1831 return -EOPNOTSUPP; 1832 } 1833 1834 static int ctnetlink_get_ct_unconfirmed(struct sk_buff *skb, 1835 const struct nfnl_info *info, 1836 const struct nlattr * const cda[]) 1837 { 1838 if (info->nlh->nlmsg_flags & NLM_F_DUMP) { 1839 struct netlink_dump_control c = { 1840 .dump = ctnetlink_dump_unconfirmed, 1841 }; 1842 return netlink_dump_start(info->sk, skb, info->nlh, &c); 1843 } 1844 1845 return -EOPNOTSUPP; 1846 } 1847 1848 #if IS_ENABLED(CONFIG_NF_NAT) 1849 static int 1850 ctnetlink_parse_nat_setup(struct nf_conn *ct, 1851 enum nf_nat_manip_type manip, 1852 const struct nlattr *attr) 1853 __must_hold(RCU) 1854 { 1855 const struct nf_nat_hook *nat_hook; 1856 int err; 1857 1858 nat_hook = rcu_dereference(nf_nat_hook); 1859 if (!nat_hook) { 1860 #ifdef CONFIG_MODULES 1861 rcu_read_unlock(); 1862 nfnl_unlock(NFNL_SUBSYS_CTNETLINK); 1863 if (request_module("nf-nat") < 0) { 1864 nfnl_lock(NFNL_SUBSYS_CTNETLINK); 1865 rcu_read_lock(); 1866 return -EOPNOTSUPP; 1867 } 1868 nfnl_lock(NFNL_SUBSYS_CTNETLINK); 1869 rcu_read_lock(); 1870 nat_hook = rcu_dereference(nf_nat_hook); 1871 if (nat_hook) 1872 return -EAGAIN; 1873 #endif 1874 return -EOPNOTSUPP; 1875 } 1876 1877 err = nat_hook->parse_nat_setup(ct, manip, attr); 1878 if (err == -EAGAIN) { 1879 #ifdef CONFIG_MODULES 1880 rcu_read_unlock(); 1881 nfnl_unlock(NFNL_SUBSYS_CTNETLINK); 1882 if (request_module("nf-nat-%u", nf_ct_l3num(ct)) < 0) { 1883 nfnl_lock(NFNL_SUBSYS_CTNETLINK); 1884 rcu_read_lock(); 1885 return -EOPNOTSUPP; 1886 } 1887 nfnl_lock(NFNL_SUBSYS_CTNETLINK); 1888 rcu_read_lock(); 1889 #else 1890 err = -EOPNOTSUPP; 1891 #endif 1892 } 1893 return err; 1894 } 1895 #endif 1896 1897 static int 1898 ctnetlink_change_status(struct nf_conn *ct, const struct nlattr * const cda[]) 1899 { 1900 return nf_ct_change_status_common(ct, ntohl(nla_get_be32(cda[CTA_STATUS]))); 1901 } 1902 1903 static int 1904 ctnetlink_setup_nat(struct nf_conn *ct, const struct nlattr * const cda[]) 1905 { 1906 #if IS_ENABLED(CONFIG_NF_NAT) 1907 int ret; 1908 1909 if (!cda[CTA_NAT_DST] && !cda[CTA_NAT_SRC]) 1910 return 0; 1911 1912 ret = ctnetlink_parse_nat_setup(ct, NF_NAT_MANIP_DST, 1913 cda[CTA_NAT_DST]); 1914 if (ret < 0) 1915 return ret; 1916 1917 return ctnetlink_parse_nat_setup(ct, NF_NAT_MANIP_SRC, 1918 cda[CTA_NAT_SRC]); 1919 #else 1920 if (!cda[CTA_NAT_DST] && !cda[CTA_NAT_SRC]) 1921 return 0; 1922 return -EOPNOTSUPP; 1923 #endif 1924 } 1925 1926 static int ctnetlink_change_helper(struct nf_conn *ct, 1927 const struct nlattr * const cda[]) 1928 { 1929 struct nf_conntrack_helper *helper; 1930 struct nf_conn_help *help = nfct_help(ct); 1931 char *helpname = NULL; 1932 struct nlattr *helpinfo = NULL; 1933 int err; 1934 1935 err = ctnetlink_parse_help(cda[CTA_HELP], &helpname, &helpinfo); 1936 if (err < 0) 1937 return err; 1938 1939 rcu_read_lock(); 1940 /* don't change helper of sibling connections */ 1941 if (ct->master) { 1942 /* If we try to change the helper to the same thing twice, 1943 * treat the second attempt as a no-op instead of returning 1944 * an error. 1945 */ 1946 err = -EBUSY; 1947 if (help) { 1948 helper = rcu_dereference(help->helper); 1949 if (helper && !strcmp(helper->name, helpname)) 1950 err = 0; 1951 } 1952 rcu_read_unlock(); 1953 return err; 1954 } 1955 1956 helper = __nf_conntrack_helper_find(helpname, nf_ct_l3num(ct), 1957 nf_ct_protonum(ct)); 1958 if (helper == NULL) { 1959 rcu_read_unlock(); 1960 return -EOPNOTSUPP; 1961 } 1962 1963 if (help) { 1964 if (rcu_access_pointer(help->helper) == helper) { 1965 /* update private helper data if allowed. */ 1966 if (rcu_access_pointer(helper->help) && 1967 helper->from_nlattr) 1968 helper->from_nlattr(helpinfo, ct); 1969 err = 0; 1970 } else 1971 err = -EBUSY; 1972 } else { 1973 /* we cannot set a helper for an existing conntrack */ 1974 err = -EOPNOTSUPP; 1975 } 1976 1977 rcu_read_unlock(); 1978 return err; 1979 } 1980 1981 static int ctnetlink_change_timeout(struct nf_conn *ct, 1982 const struct nlattr * const cda[]) 1983 { 1984 return __nf_ct_change_timeout(ct, (u64)ntohl(nla_get_be32(cda[CTA_TIMEOUT])) * HZ); 1985 } 1986 1987 #if defined(CONFIG_NF_CONNTRACK_MARK) 1988 static void ctnetlink_change_mark(struct nf_conn *ct, 1989 const struct nlattr * const cda[]) 1990 { 1991 u32 mark, newmark, mask = 0; 1992 1993 if (cda[CTA_MARK_MASK]) 1994 mask = ~ntohl(nla_get_be32(cda[CTA_MARK_MASK])); 1995 1996 mark = ntohl(nla_get_be32(cda[CTA_MARK])); 1997 newmark = (READ_ONCE(ct->mark) & mask) ^ mark; 1998 if (newmark != READ_ONCE(ct->mark)) 1999 WRITE_ONCE(ct->mark, newmark); 2000 } 2001 #endif 2002 2003 static const struct nla_policy protoinfo_policy[CTA_PROTOINFO_MAX+1] = { 2004 [CTA_PROTOINFO_TCP] = { .type = NLA_NESTED }, 2005 [CTA_PROTOINFO_SCTP] = { .type = NLA_NESTED }, 2006 }; 2007 2008 static int ctnetlink_change_protoinfo(struct nf_conn *ct, 2009 const struct nlattr * const cda[]) 2010 { 2011 const struct nlattr *attr = cda[CTA_PROTOINFO]; 2012 const struct nf_conntrack_l4proto *l4proto; 2013 struct nlattr *tb[CTA_PROTOINFO_MAX+1]; 2014 int err = 0; 2015 2016 err = nla_parse_nested_deprecated(tb, CTA_PROTOINFO_MAX, attr, 2017 protoinfo_policy, NULL); 2018 if (err < 0) 2019 return err; 2020 2021 l4proto = nf_ct_l4proto_find(nf_ct_protonum(ct)); 2022 if (l4proto->from_nlattr) 2023 err = l4proto->from_nlattr(tb, ct); 2024 2025 return err; 2026 } 2027 2028 static const struct nla_policy seqadj_policy[CTA_SEQADJ_MAX+1] = { 2029 [CTA_SEQADJ_CORRECTION_POS] = { .type = NLA_U32 }, 2030 [CTA_SEQADJ_OFFSET_BEFORE] = { .type = NLA_U32 }, 2031 [CTA_SEQADJ_OFFSET_AFTER] = { .type = NLA_U32 }, 2032 }; 2033 2034 static int change_seq_adj(struct nf_ct_seqadj *seq, 2035 const struct nlattr * const attr) 2036 { 2037 int err; 2038 struct nlattr *cda[CTA_SEQADJ_MAX+1]; 2039 2040 err = nla_parse_nested_deprecated(cda, CTA_SEQADJ_MAX, attr, 2041 seqadj_policy, NULL); 2042 if (err < 0) 2043 return err; 2044 2045 if (!cda[CTA_SEQADJ_CORRECTION_POS]) 2046 return -EINVAL; 2047 2048 seq->correction_pos = 2049 ntohl(nla_get_be32(cda[CTA_SEQADJ_CORRECTION_POS])); 2050 2051 if (!cda[CTA_SEQADJ_OFFSET_BEFORE]) 2052 return -EINVAL; 2053 2054 seq->offset_before = 2055 ntohl(nla_get_be32(cda[CTA_SEQADJ_OFFSET_BEFORE])); 2056 2057 if (!cda[CTA_SEQADJ_OFFSET_AFTER]) 2058 return -EINVAL; 2059 2060 seq->offset_after = 2061 ntohl(nla_get_be32(cda[CTA_SEQADJ_OFFSET_AFTER])); 2062 2063 return 0; 2064 } 2065 2066 static int 2067 ctnetlink_change_seq_adj(struct nf_conn *ct, 2068 const struct nlattr * const cda[]) 2069 { 2070 struct nf_conn_seqadj *seqadj = nfct_seqadj(ct); 2071 int ret = 0; 2072 2073 if (!seqadj) 2074 return 0; 2075 2076 spin_lock_bh(&ct->lock); 2077 if (cda[CTA_SEQ_ADJ_ORIG]) { 2078 ret = change_seq_adj(&seqadj->seq[IP_CT_DIR_ORIGINAL], 2079 cda[CTA_SEQ_ADJ_ORIG]); 2080 if (ret < 0) 2081 goto err; 2082 2083 set_bit(IPS_SEQ_ADJUST_BIT, &ct->status); 2084 } 2085 2086 if (cda[CTA_SEQ_ADJ_REPLY]) { 2087 ret = change_seq_adj(&seqadj->seq[IP_CT_DIR_REPLY], 2088 cda[CTA_SEQ_ADJ_REPLY]); 2089 if (ret < 0) 2090 goto err; 2091 2092 set_bit(IPS_SEQ_ADJUST_BIT, &ct->status); 2093 } 2094 2095 spin_unlock_bh(&ct->lock); 2096 return 0; 2097 err: 2098 spin_unlock_bh(&ct->lock); 2099 return ret; 2100 } 2101 2102 static const struct nla_policy synproxy_policy[CTA_SYNPROXY_MAX + 1] = { 2103 [CTA_SYNPROXY_ISN] = { .type = NLA_U32 }, 2104 [CTA_SYNPROXY_ITS] = { .type = NLA_U32 }, 2105 [CTA_SYNPROXY_TSOFF] = { .type = NLA_U32 }, 2106 }; 2107 2108 static int ctnetlink_change_synproxy(struct nf_conn *ct, 2109 const struct nlattr * const cda[]) 2110 { 2111 struct nf_conn_synproxy *synproxy = nfct_synproxy(ct); 2112 struct nlattr *tb[CTA_SYNPROXY_MAX + 1]; 2113 int err; 2114 2115 if (!synproxy) 2116 return 0; 2117 2118 err = nla_parse_nested_deprecated(tb, CTA_SYNPROXY_MAX, 2119 cda[CTA_SYNPROXY], synproxy_policy, 2120 NULL); 2121 if (err < 0) 2122 return err; 2123 2124 if (!tb[CTA_SYNPROXY_ISN] || 2125 !tb[CTA_SYNPROXY_ITS] || 2126 !tb[CTA_SYNPROXY_TSOFF]) 2127 return -EINVAL; 2128 2129 synproxy->isn = ntohl(nla_get_be32(tb[CTA_SYNPROXY_ISN])); 2130 synproxy->its = ntohl(nla_get_be32(tb[CTA_SYNPROXY_ITS])); 2131 synproxy->tsoff = ntohl(nla_get_be32(tb[CTA_SYNPROXY_TSOFF])); 2132 2133 return 0; 2134 } 2135 2136 static int 2137 ctnetlink_attach_labels(struct nf_conn *ct, const struct nlattr * const cda[]) 2138 { 2139 #ifdef CONFIG_NF_CONNTRACK_LABELS 2140 size_t len = nla_len(cda[CTA_LABELS]); 2141 const void *mask = cda[CTA_LABELS_MASK]; 2142 2143 if (len & (sizeof(u32)-1)) /* must be multiple of u32 */ 2144 return -EINVAL; 2145 2146 if (mask) { 2147 if (nla_len(cda[CTA_LABELS_MASK]) == 0 || 2148 nla_len(cda[CTA_LABELS_MASK]) != len) 2149 return -EINVAL; 2150 mask = nla_data(cda[CTA_LABELS_MASK]); 2151 } 2152 2153 len /= sizeof(u32); 2154 2155 return nf_connlabels_replace(ct, nla_data(cda[CTA_LABELS]), mask, len); 2156 #else 2157 return -EOPNOTSUPP; 2158 #endif 2159 } 2160 2161 static int 2162 ctnetlink_change_conntrack(struct nf_conn *ct, 2163 const struct nlattr * const cda[]) 2164 { 2165 int err; 2166 2167 /* only allow NAT changes and master assignation for new conntracks */ 2168 if (cda[CTA_NAT_SRC] || cda[CTA_NAT_DST] || cda[CTA_TUPLE_MASTER]) 2169 return -EOPNOTSUPP; 2170 2171 if (cda[CTA_HELP]) { 2172 err = ctnetlink_change_helper(ct, cda); 2173 if (err < 0) 2174 return err; 2175 } 2176 2177 if (cda[CTA_TIMEOUT]) { 2178 err = ctnetlink_change_timeout(ct, cda); 2179 if (err < 0) 2180 return err; 2181 } 2182 2183 if (cda[CTA_STATUS]) { 2184 err = ctnetlink_change_status(ct, cda); 2185 if (err < 0) 2186 return err; 2187 } 2188 2189 if (cda[CTA_PROTOINFO]) { 2190 err = ctnetlink_change_protoinfo(ct, cda); 2191 if (err < 0) 2192 return err; 2193 } 2194 2195 #if defined(CONFIG_NF_CONNTRACK_MARK) 2196 if (cda[CTA_MARK]) 2197 ctnetlink_change_mark(ct, cda); 2198 #endif 2199 2200 if (cda[CTA_SEQ_ADJ_ORIG] || cda[CTA_SEQ_ADJ_REPLY]) { 2201 err = ctnetlink_change_seq_adj(ct, cda); 2202 if (err < 0) 2203 return err; 2204 } 2205 2206 if (cda[CTA_SYNPROXY]) { 2207 err = ctnetlink_change_synproxy(ct, cda); 2208 if (err < 0) 2209 return err; 2210 } 2211 2212 if (cda[CTA_LABELS]) { 2213 err = ctnetlink_attach_labels(ct, cda); 2214 if (err < 0) 2215 return err; 2216 } 2217 2218 return 0; 2219 } 2220 2221 static struct nf_conn * 2222 ctnetlink_create_conntrack(struct net *net, 2223 const struct nf_conntrack_zone *zone, 2224 const struct nlattr * const cda[], 2225 struct nf_conntrack_tuple *otuple, 2226 struct nf_conntrack_tuple *rtuple, 2227 u8 u3) 2228 { 2229 struct nf_conn *ct; 2230 int err = -EINVAL; 2231 struct nf_conntrack_helper *helper; 2232 struct nf_conn_tstamp *tstamp; 2233 u64 timeout; 2234 2235 ct = nf_conntrack_alloc(net, zone, otuple, rtuple, GFP_ATOMIC); 2236 if (IS_ERR(ct)) 2237 return ERR_PTR(-ENOMEM); 2238 2239 if (!cda[CTA_TIMEOUT]) 2240 goto err1; 2241 2242 rcu_read_lock(); 2243 if (cda[CTA_HELP]) { 2244 char *helpname = NULL; 2245 struct nlattr *helpinfo = NULL; 2246 2247 err = ctnetlink_parse_help(cda[CTA_HELP], &helpname, &helpinfo); 2248 if (err < 0) 2249 goto err2; 2250 2251 helper = __nf_conntrack_helper_find(helpname, nf_ct_l3num(ct), 2252 nf_ct_protonum(ct)); 2253 if (helper == NULL) { 2254 rcu_read_unlock(); 2255 #ifdef CONFIG_MODULES 2256 if (request_module("nfct-helper-%s", helpname) < 0) { 2257 err = -EOPNOTSUPP; 2258 goto err1; 2259 } 2260 2261 rcu_read_lock(); 2262 helper = __nf_conntrack_helper_find(helpname, 2263 nf_ct_l3num(ct), 2264 nf_ct_protonum(ct)); 2265 if (helper) { 2266 err = -EAGAIN; 2267 goto err2; 2268 } 2269 rcu_read_unlock(); 2270 #endif 2271 err = -EOPNOTSUPP; 2272 goto err1; 2273 } else { 2274 struct nf_conn_help *help; 2275 2276 help = nf_ct_helper_ext_add(ct, GFP_ATOMIC); 2277 if (help == NULL) { 2278 err = -ENOMEM; 2279 goto err2; 2280 } 2281 /* set private helper data if allowed. */ 2282 if (rcu_access_pointer(helper->help) && 2283 helper->from_nlattr) 2284 helper->from_nlattr(helpinfo, ct); 2285 2286 /* disable helper auto-assignment for this entry */ 2287 ct->status |= IPS_HELPER; 2288 if (!refcount_inc_not_zero(&helper->ct_refcnt)) { 2289 err = -ENOENT; 2290 goto err2; 2291 } 2292 RCU_INIT_POINTER(help->helper, helper); 2293 } 2294 } 2295 2296 err = ctnetlink_setup_nat(ct, cda); 2297 if (err < 0) 2298 goto err2; 2299 2300 nf_ct_acct_ext_add(ct, GFP_ATOMIC); 2301 nf_ct_tstamp_ext_add(ct, GFP_ATOMIC); 2302 nf_ct_ecache_ext_add(ct, 0, 0, GFP_ATOMIC); 2303 nf_ct_labels_ext_add(ct); 2304 nfct_seqadj_ext_add(ct); 2305 nfct_synproxy_ext_add(ct); 2306 2307 /* we must add conntrack extensions before confirmation. */ 2308 ct->status |= IPS_CONFIRMED; 2309 2310 timeout = (u64)ntohl(nla_get_be32(cda[CTA_TIMEOUT])) * HZ; 2311 __nf_ct_set_timeout(ct, timeout); 2312 2313 if (cda[CTA_STATUS]) { 2314 err = ctnetlink_change_status(ct, cda); 2315 if (err < 0) 2316 goto err2; 2317 } 2318 2319 if (cda[CTA_SEQ_ADJ_ORIG] || cda[CTA_SEQ_ADJ_REPLY]) { 2320 err = ctnetlink_change_seq_adj(ct, cda); 2321 if (err < 0) 2322 goto err2; 2323 } 2324 2325 memset(&ct->proto, 0, sizeof(ct->proto)); 2326 if (cda[CTA_PROTOINFO]) { 2327 err = ctnetlink_change_protoinfo(ct, cda); 2328 if (err < 0) 2329 goto err2; 2330 } 2331 2332 if (cda[CTA_SYNPROXY]) { 2333 err = ctnetlink_change_synproxy(ct, cda); 2334 if (err < 0) 2335 goto err2; 2336 } 2337 2338 #if defined(CONFIG_NF_CONNTRACK_MARK) 2339 if (cda[CTA_MARK]) 2340 ctnetlink_change_mark(ct, cda); 2341 #endif 2342 2343 /* setup master conntrack: this is a confirmed expectation */ 2344 if (cda[CTA_TUPLE_MASTER]) { 2345 struct nf_conntrack_tuple master; 2346 struct nf_conntrack_tuple_hash *master_h; 2347 struct nf_conn *master_ct; 2348 2349 err = ctnetlink_parse_tuple(cda, &master, CTA_TUPLE_MASTER, 2350 u3, NULL); 2351 if (err < 0) 2352 goto err2; 2353 2354 master_h = nf_conntrack_find_get(net, zone, &master); 2355 if (master_h == NULL) { 2356 err = -ENOENT; 2357 goto err2; 2358 } 2359 master_ct = nf_ct_tuplehash_to_ctrack(master_h); 2360 __set_bit(IPS_EXPECTED_BIT, &ct->status); 2361 ct->master = master_ct; 2362 } 2363 tstamp = nf_conn_tstamp_find(ct); 2364 if (tstamp) 2365 tstamp->start = ktime_get_real_ns(); 2366 2367 err = nf_conntrack_hash_check_insert(ct); 2368 if (err < 0) 2369 goto err3; 2370 2371 rcu_read_unlock(); 2372 2373 return ct; 2374 2375 err3: 2376 if (ct->master) 2377 nf_ct_put(ct->master); 2378 err2: 2379 rcu_read_unlock(); 2380 err1: 2381 nf_conntrack_free(ct); 2382 return ERR_PTR(err); 2383 } 2384 2385 static int ctnetlink_new_conntrack(struct sk_buff *skb, 2386 const struct nfnl_info *info, 2387 const struct nlattr * const cda[]) 2388 { 2389 struct nf_conntrack_tuple otuple, rtuple; 2390 struct nf_conntrack_tuple_hash *h = NULL; 2391 u_int8_t u3 = info->nfmsg->nfgen_family; 2392 struct nf_conntrack_zone zone; 2393 struct nf_conn *ct; 2394 int err; 2395 2396 err = ctnetlink_parse_zone(cda[CTA_ZONE], &zone); 2397 if (err < 0) 2398 return err; 2399 2400 if (cda[CTA_TUPLE_ORIG]) { 2401 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG, 2402 u3, &zone); 2403 if (err < 0) 2404 return err; 2405 } 2406 2407 if (cda[CTA_TUPLE_REPLY]) { 2408 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY, 2409 u3, &zone); 2410 if (err < 0) 2411 return err; 2412 } 2413 2414 if (cda[CTA_TUPLE_ORIG]) 2415 h = nf_conntrack_find_get(info->net, &zone, &otuple); 2416 else if (cda[CTA_TUPLE_REPLY]) 2417 h = nf_conntrack_find_get(info->net, &zone, &rtuple); 2418 2419 if (h == NULL) { 2420 err = -ENOENT; 2421 if (info->nlh->nlmsg_flags & NLM_F_CREATE) { 2422 enum ip_conntrack_events events; 2423 2424 if (!cda[CTA_TUPLE_ORIG] || !cda[CTA_TUPLE_REPLY]) 2425 return -EINVAL; 2426 if (otuple.dst.protonum != rtuple.dst.protonum) 2427 return -EINVAL; 2428 2429 ct = ctnetlink_create_conntrack(info->net, &zone, cda, 2430 &otuple, &rtuple, u3); 2431 if (IS_ERR(ct)) 2432 return PTR_ERR(ct); 2433 2434 err = 0; 2435 if (test_bit(IPS_EXPECTED_BIT, &ct->status)) 2436 events = 1 << IPCT_RELATED; 2437 else 2438 events = 1 << IPCT_NEW; 2439 2440 if (cda[CTA_LABELS] && 2441 ctnetlink_attach_labels(ct, cda) == 0) 2442 events |= (1 << IPCT_LABEL); 2443 2444 nf_conntrack_eventmask_report((1 << IPCT_REPLY) | 2445 (1 << IPCT_ASSURED) | 2446 (1 << IPCT_HELPER) | 2447 (1 << IPCT_PROTOINFO) | 2448 (1 << IPCT_SEQADJ) | 2449 (1 << IPCT_MARK) | 2450 (1 << IPCT_SYNPROXY) | 2451 events, 2452 ct, NETLINK_CB(skb).portid, 2453 nlmsg_report(info->nlh)); 2454 nf_ct_put(ct); 2455 } 2456 2457 return err; 2458 } 2459 /* implicit 'else' */ 2460 2461 err = -EEXIST; 2462 ct = nf_ct_tuplehash_to_ctrack(h); 2463 if (!(info->nlh->nlmsg_flags & NLM_F_EXCL)) { 2464 err = ctnetlink_change_conntrack(ct, cda); 2465 if (err == 0) { 2466 nf_conntrack_eventmask_report((1 << IPCT_REPLY) | 2467 (1 << IPCT_ASSURED) | 2468 (1 << IPCT_HELPER) | 2469 (1 << IPCT_LABEL) | 2470 (1 << IPCT_PROTOINFO) | 2471 (1 << IPCT_SEQADJ) | 2472 (1 << IPCT_MARK) | 2473 (1 << IPCT_SYNPROXY), 2474 ct, NETLINK_CB(skb).portid, 2475 nlmsg_report(info->nlh)); 2476 } 2477 } 2478 2479 nf_ct_put(ct); 2480 return err; 2481 } 2482 2483 static int 2484 ctnetlink_ct_stat_cpu_fill_info(struct sk_buff *skb, u32 portid, u32 seq, 2485 __u16 cpu, const struct ip_conntrack_stat *st) 2486 { 2487 struct nlmsghdr *nlh; 2488 unsigned int flags = portid ? NLM_F_MULTI : 0, event; 2489 2490 event = nfnl_msg_type(NFNL_SUBSYS_CTNETLINK, 2491 IPCTNL_MSG_CT_GET_STATS_CPU); 2492 nlh = nfnl_msg_put(skb, portid, seq, event, flags, AF_UNSPEC, 2493 NFNETLINK_V0, htons(cpu)); 2494 if (!nlh) 2495 goto nlmsg_failure; 2496 2497 if (nla_put_be32(skb, CTA_STATS_FOUND, htonl(st->found)) || 2498 nla_put_be32(skb, CTA_STATS_INVALID, htonl(st->invalid)) || 2499 nla_put_be32(skb, CTA_STATS_INSERT, htonl(st->insert)) || 2500 nla_put_be32(skb, CTA_STATS_INSERT_FAILED, 2501 htonl(st->insert_failed)) || 2502 nla_put_be32(skb, CTA_STATS_DROP, htonl(st->drop)) || 2503 nla_put_be32(skb, CTA_STATS_EARLY_DROP, htonl(st->early_drop)) || 2504 nla_put_be32(skb, CTA_STATS_ERROR, htonl(st->error)) || 2505 nla_put_be32(skb, CTA_STATS_SEARCH_RESTART, 2506 htonl(st->search_restart)) || 2507 nla_put_be32(skb, CTA_STATS_CLASH_RESOLVE, 2508 htonl(st->clash_resolve)) || 2509 nla_put_be32(skb, CTA_STATS_CHAIN_TOOLONG, 2510 htonl(st->chaintoolong))) 2511 goto nla_put_failure; 2512 2513 nlmsg_end(skb, nlh); 2514 return skb->len; 2515 2516 nla_put_failure: 2517 nlmsg_failure: 2518 nlmsg_cancel(skb, nlh); 2519 return -1; 2520 } 2521 2522 static int 2523 ctnetlink_ct_stat_cpu_dump(struct sk_buff *skb, struct netlink_callback *cb) 2524 { 2525 int cpu; 2526 struct net *net = sock_net(skb->sk); 2527 2528 if (cb->args[0] == nr_cpu_ids) 2529 return 0; 2530 2531 for (cpu = cb->args[0]; cpu < nr_cpu_ids; cpu++) { 2532 const struct ip_conntrack_stat *st; 2533 2534 if (!cpu_possible(cpu)) 2535 continue; 2536 2537 st = per_cpu_ptr(net->ct.stat, cpu); 2538 if (ctnetlink_ct_stat_cpu_fill_info(skb, 2539 NETLINK_CB(cb->skb).portid, 2540 cb->nlh->nlmsg_seq, 2541 cpu, st) < 0) 2542 break; 2543 } 2544 cb->args[0] = cpu; 2545 2546 return skb->len; 2547 } 2548 2549 static int ctnetlink_stat_ct_cpu(struct sk_buff *skb, 2550 const struct nfnl_info *info, 2551 const struct nlattr * const cda[]) 2552 { 2553 if (info->nlh->nlmsg_flags & NLM_F_DUMP) { 2554 struct netlink_dump_control c = { 2555 .dump = ctnetlink_ct_stat_cpu_dump, 2556 }; 2557 return netlink_dump_start(info->sk, skb, info->nlh, &c); 2558 } 2559 2560 return 0; 2561 } 2562 2563 static int 2564 ctnetlink_stat_ct_fill_info(struct sk_buff *skb, u32 portid, u32 seq, u32 type, 2565 struct net *net) 2566 { 2567 unsigned int flags = portid ? NLM_F_MULTI : 0, event; 2568 unsigned int nr_conntracks; 2569 struct nlmsghdr *nlh; 2570 2571 event = nfnl_msg_type(NFNL_SUBSYS_CTNETLINK, IPCTNL_MSG_CT_GET_STATS); 2572 nlh = nfnl_msg_put(skb, portid, seq, event, flags, AF_UNSPEC, 2573 NFNETLINK_V0, 0); 2574 if (!nlh) 2575 goto nlmsg_failure; 2576 2577 nr_conntracks = nf_conntrack_count(net); 2578 if (nla_put_be32(skb, CTA_STATS_GLOBAL_ENTRIES, htonl(nr_conntracks))) 2579 goto nla_put_failure; 2580 2581 if (nla_put_be32(skb, CTA_STATS_GLOBAL_MAX_ENTRIES, htonl(nf_conntrack_max))) 2582 goto nla_put_failure; 2583 2584 nlmsg_end(skb, nlh); 2585 return skb->len; 2586 2587 nla_put_failure: 2588 nlmsg_failure: 2589 nlmsg_cancel(skb, nlh); 2590 return -1; 2591 } 2592 2593 static int ctnetlink_stat_ct(struct sk_buff *skb, const struct nfnl_info *info, 2594 const struct nlattr * const cda[]) 2595 { 2596 struct sk_buff *skb2; 2597 int err; 2598 2599 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); 2600 if (skb2 == NULL) 2601 return -ENOMEM; 2602 2603 err = ctnetlink_stat_ct_fill_info(skb2, NETLINK_CB(skb).portid, 2604 info->nlh->nlmsg_seq, 2605 NFNL_MSG_TYPE(info->nlh->nlmsg_type), 2606 sock_net(skb->sk)); 2607 if (err <= 0) { 2608 kfree_skb(skb2); 2609 return -ENOMEM; 2610 } 2611 2612 return nfnetlink_unicast(skb2, info->net, NETLINK_CB(skb).portid); 2613 } 2614 2615 static const struct nla_policy exp_nla_policy[CTA_EXPECT_MAX+1] = { 2616 [CTA_EXPECT_MASTER] = { .type = NLA_NESTED }, 2617 [CTA_EXPECT_TUPLE] = { .type = NLA_NESTED }, 2618 [CTA_EXPECT_MASK] = { .type = NLA_NESTED }, 2619 [CTA_EXPECT_TIMEOUT] = { .type = NLA_U32 }, 2620 [CTA_EXPECT_ID] = { .type = NLA_U32 }, 2621 [CTA_EXPECT_HELP_NAME] = { .type = NLA_NUL_STRING, 2622 .len = NF_CT_HELPER_NAME_LEN - 1 }, 2623 [CTA_EXPECT_ZONE] = { .type = NLA_U16 }, 2624 [CTA_EXPECT_FLAGS] = NLA_POLICY_MASK(NLA_BE32, NF_CT_EXPECT_MASK), 2625 [CTA_EXPECT_CLASS] = { .type = NLA_U32 }, 2626 [CTA_EXPECT_NAT] = { .type = NLA_NESTED }, 2627 [CTA_EXPECT_FN] = { .type = NLA_NUL_STRING }, 2628 }; 2629 2630 static struct nf_conntrack_expect * 2631 ctnetlink_alloc_expect(const struct nlattr *const cda[], struct nf_conn *ct, 2632 const struct nf_conntrack_helper *assign_helper, 2633 struct nf_conntrack_tuple *tuple, 2634 struct nf_conntrack_tuple *mask); 2635 2636 #ifdef CONFIG_NETFILTER_NETLINK_GLUE_CT 2637 static size_t 2638 ctnetlink_glue_build_size(const struct nf_conn *ct) 2639 { 2640 return 3 * nla_total_size(0) /* CTA_TUPLE_ORIG|REPL|MASTER */ 2641 + 3 * nla_total_size(0) /* CTA_TUPLE_IP */ 2642 + 3 * nla_total_size(0) /* CTA_TUPLE_PROTO */ 2643 + 3 * nla_total_size(sizeof(u_int8_t)) /* CTA_PROTO_NUM */ 2644 + nla_total_size(sizeof(u_int32_t)) /* CTA_ID */ 2645 + nla_total_size(sizeof(u_int32_t)) /* CTA_STATUS */ 2646 + nla_total_size(sizeof(u_int32_t)) /* CTA_TIMEOUT */ 2647 + nla_total_size(0) /* CTA_PROTOINFO */ 2648 + nla_total_size(0) /* CTA_HELP */ 2649 + nla_total_size(NF_CT_HELPER_NAME_LEN) /* CTA_HELP_NAME */ 2650 + ctnetlink_secctx_size(ct) 2651 + ctnetlink_acct_size(ct) 2652 + ctnetlink_timestamp_size(ct) 2653 #if IS_ENABLED(CONFIG_NF_NAT) 2654 + 2 * nla_total_size(0) /* CTA_NAT_SEQ_ADJ_ORIG|REPL */ 2655 + 6 * nla_total_size(sizeof(u_int32_t)) /* CTA_NAT_SEQ_OFFSET */ 2656 #endif 2657 #ifdef CONFIG_NF_CONNTRACK_MARK 2658 + nla_total_size(sizeof(u_int32_t)) /* CTA_MARK */ 2659 #endif 2660 #ifdef CONFIG_NF_CONNTRACK_ZONES 2661 + nla_total_size(sizeof(u_int16_t)) /* CTA_ZONE|CTA_TUPLE_ZONE */ 2662 #endif 2663 + ctnetlink_proto_size(ct) 2664 ; 2665 } 2666 2667 static int __ctnetlink_glue_build(struct sk_buff *skb, struct nf_conn *ct) 2668 { 2669 const struct nf_conntrack_zone *zone; 2670 struct nlattr *nest_parms; 2671 2672 zone = nf_ct_zone(ct); 2673 2674 nest_parms = nla_nest_start(skb, CTA_TUPLE_ORIG); 2675 if (!nest_parms) 2676 goto nla_put_failure; 2677 if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_ORIGINAL)) < 0) 2678 goto nla_put_failure; 2679 if (ctnetlink_dump_zone_id(skb, CTA_TUPLE_ZONE, zone, 2680 NF_CT_ZONE_DIR_ORIG) < 0) 2681 goto nla_put_failure; 2682 nla_nest_end(skb, nest_parms); 2683 2684 nest_parms = nla_nest_start(skb, CTA_TUPLE_REPLY); 2685 if (!nest_parms) 2686 goto nla_put_failure; 2687 if (ctnetlink_dump_tuples(skb, nf_ct_tuple(ct, IP_CT_DIR_REPLY)) < 0) 2688 goto nla_put_failure; 2689 if (ctnetlink_dump_zone_id(skb, CTA_TUPLE_ZONE, zone, 2690 NF_CT_ZONE_DIR_REPL) < 0) 2691 goto nla_put_failure; 2692 nla_nest_end(skb, nest_parms); 2693 2694 if (ctnetlink_dump_zone_id(skb, CTA_ZONE, zone, 2695 NF_CT_DEFAULT_ZONE_DIR) < 0) 2696 goto nla_put_failure; 2697 2698 if (ctnetlink_dump_id(skb, ct) < 0) 2699 goto nla_put_failure; 2700 2701 if (ctnetlink_dump_status(skb, ct) < 0) 2702 goto nla_put_failure; 2703 2704 if (ctnetlink_dump_timeout(skb, ct, false) < 0) 2705 goto nla_put_failure; 2706 2707 if (ctnetlink_dump_protoinfo(skb, ct, false) < 0) 2708 goto nla_put_failure; 2709 2710 if (ctnetlink_dump_acct(skb, ct, IPCTNL_MSG_CT_GET) < 0 || 2711 ctnetlink_dump_timestamp(skb, ct) < 0) 2712 goto nla_put_failure; 2713 2714 if (ctnetlink_dump_helpinfo(skb, ct) < 0) 2715 goto nla_put_failure; 2716 2717 #ifdef CONFIG_NF_CONNTRACK_SECMARK 2718 if (ct->secmark && ctnetlink_dump_secctx(skb, ct) < 0) 2719 goto nla_put_failure; 2720 #endif 2721 if (ct->master && ctnetlink_dump_master(skb, ct) < 0) 2722 goto nla_put_failure; 2723 2724 if ((ct->status & IPS_SEQ_ADJUST) && 2725 ctnetlink_dump_ct_seq_adj(skb, ct) < 0) 2726 goto nla_put_failure; 2727 2728 if (ctnetlink_dump_ct_synproxy(skb, ct) < 0) 2729 goto nla_put_failure; 2730 2731 #ifdef CONFIG_NF_CONNTRACK_MARK 2732 if (ctnetlink_dump_mark(skb, ct, true) < 0) 2733 goto nla_put_failure; 2734 #endif 2735 if (ctnetlink_dump_labels(skb, ct) < 0) 2736 goto nla_put_failure; 2737 return 0; 2738 2739 nla_put_failure: 2740 return -ENOSPC; 2741 } 2742 2743 static int 2744 ctnetlink_glue_build(struct sk_buff *skb, struct nf_conn *ct, 2745 enum ip_conntrack_info ctinfo, 2746 u_int16_t ct_attr, u_int16_t ct_info_attr) 2747 { 2748 struct nlattr *nest_parms; 2749 2750 nest_parms = nla_nest_start(skb, ct_attr); 2751 if (!nest_parms) 2752 goto nla_put_failure; 2753 2754 if (__ctnetlink_glue_build(skb, ct) < 0) 2755 goto nla_put_failure; 2756 2757 nla_nest_end(skb, nest_parms); 2758 2759 if (nla_put_be32(skb, ct_info_attr, htonl(ctinfo))) 2760 goto nla_put_failure; 2761 2762 return 0; 2763 2764 nla_put_failure: 2765 return -ENOSPC; 2766 } 2767 2768 static int 2769 ctnetlink_update_status(struct nf_conn *ct, const struct nlattr * const cda[]) 2770 { 2771 unsigned int status = ntohl(nla_get_be32(cda[CTA_STATUS])); 2772 unsigned long d = ct->status ^ status; 2773 2774 if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY)) 2775 /* SEEN_REPLY bit can only be set */ 2776 return -EBUSY; 2777 2778 if (d & IPS_ASSURED && !(status & IPS_ASSURED)) 2779 /* ASSURED bit can only be set */ 2780 return -EBUSY; 2781 2782 /* This check is less strict than ctnetlink_change_status() 2783 * because callers often flip IPS_EXPECTED bits when sending 2784 * an NFQA_CT attribute to the kernel. So ignore the 2785 * unchangeable bits but do not error out. Also user programs 2786 * are allowed to clear the bits that they are allowed to change. 2787 */ 2788 __nf_ct_change_status(ct, status, ~status); 2789 return 0; 2790 } 2791 2792 static int 2793 ctnetlink_glue_parse_ct(const struct nlattr *cda[], struct nf_conn *ct) 2794 { 2795 int err; 2796 2797 if (cda[CTA_TIMEOUT]) { 2798 err = ctnetlink_change_timeout(ct, cda); 2799 if (err < 0) 2800 return err; 2801 } 2802 if (cda[CTA_STATUS]) { 2803 err = ctnetlink_update_status(ct, cda); 2804 if (err < 0) 2805 return err; 2806 } 2807 if (cda[CTA_HELP]) { 2808 err = ctnetlink_change_helper(ct, cda); 2809 if (err < 0) 2810 return err; 2811 } 2812 if (cda[CTA_LABELS]) { 2813 err = ctnetlink_attach_labels(ct, cda); 2814 if (err < 0) 2815 return err; 2816 } 2817 #if defined(CONFIG_NF_CONNTRACK_MARK) 2818 if (cda[CTA_MARK]) { 2819 ctnetlink_change_mark(ct, cda); 2820 } 2821 #endif 2822 return 0; 2823 } 2824 2825 static int 2826 ctnetlink_glue_parse(const struct nlattr *attr, struct nf_conn *ct) 2827 { 2828 struct nlattr *cda[CTA_MAX+1]; 2829 int ret; 2830 2831 ret = nla_parse_nested_deprecated(cda, CTA_MAX, attr, ct_nla_policy, 2832 NULL); 2833 if (ret < 0) 2834 return ret; 2835 2836 return ctnetlink_glue_parse_ct((const struct nlattr **)cda, ct); 2837 } 2838 2839 static int ctnetlink_glue_exp_parse(const struct nlattr * const *cda, 2840 const struct nf_conn *ct, 2841 struct nf_conntrack_tuple *tuple, 2842 struct nf_conntrack_tuple *mask) 2843 { 2844 int err; 2845 2846 err = ctnetlink_parse_tuple(cda, tuple, CTA_EXPECT_TUPLE, 2847 nf_ct_l3num(ct), NULL); 2848 if (err < 0) 2849 return err; 2850 2851 return ctnetlink_parse_tuple(cda, mask, CTA_EXPECT_MASK, 2852 nf_ct_l3num(ct), NULL); 2853 } 2854 2855 static int 2856 ctnetlink_glue_attach_expect(const struct nlattr *attr, struct nf_conn *ct, 2857 u32 portid, u32 report) 2858 { 2859 struct nf_conntrack_helper *assign_helper = NULL; 2860 struct nlattr *cda[CTA_EXPECT_MAX+1]; 2861 struct nf_conntrack_tuple tuple, mask; 2862 struct nf_conntrack_expect *exp; 2863 int err; 2864 2865 err = nla_parse_nested_deprecated(cda, CTA_EXPECT_MAX, attr, 2866 exp_nla_policy, NULL); 2867 if (err < 0) 2868 return err; 2869 2870 if (!cda[CTA_EXPECT_TUPLE] || !cda[CTA_EXPECT_MASK]) 2871 return -EINVAL; 2872 2873 err = ctnetlink_glue_exp_parse((const struct nlattr * const *)cda, 2874 ct, &tuple, &mask); 2875 if (err < 0) 2876 return err; 2877 2878 if (cda[CTA_EXPECT_HELP_NAME]) { 2879 const char *helpname = nla_data(cda[CTA_EXPECT_HELP_NAME]); 2880 2881 assign_helper = __nf_conntrack_helper_find(helpname, 2882 nf_ct_l3num(ct), 2883 tuple.dst.protonum); 2884 if (!assign_helper) 2885 return -EOPNOTSUPP; 2886 } 2887 2888 exp = ctnetlink_alloc_expect((const struct nlattr * const *)cda, ct, 2889 assign_helper, &tuple, &mask); 2890 if (IS_ERR(exp)) 2891 return PTR_ERR(exp); 2892 2893 err = nf_ct_expect_related_report(exp, portid, report, 0); 2894 nf_ct_expect_put(exp); 2895 return err; 2896 } 2897 2898 static void ctnetlink_glue_seqadj(struct sk_buff *skb, struct nf_conn *ct, 2899 enum ip_conntrack_info ctinfo, int diff) 2900 { 2901 if (!(ct->status & IPS_NAT_MASK)) 2902 return; 2903 2904 nf_ct_tcp_seqadj_set(skb, ct, ctinfo, diff); 2905 } 2906 2907 static const struct nfnl_ct_hook ctnetlink_glue_hook = { 2908 .build_size = ctnetlink_glue_build_size, 2909 .build = ctnetlink_glue_build, 2910 .parse = ctnetlink_glue_parse, 2911 .attach_expect = ctnetlink_glue_attach_expect, 2912 .seq_adjust = ctnetlink_glue_seqadj, 2913 }; 2914 #endif /* CONFIG_NETFILTER_NETLINK_GLUE_CT */ 2915 2916 /*********************************************************************** 2917 * EXPECT 2918 ***********************************************************************/ 2919 2920 static int ctnetlink_exp_dump_tuple(struct sk_buff *skb, 2921 const struct nf_conntrack_tuple *tuple, 2922 u32 type) 2923 { 2924 struct nlattr *nest_parms; 2925 2926 nest_parms = nla_nest_start(skb, type); 2927 if (!nest_parms) 2928 goto nla_put_failure; 2929 if (ctnetlink_dump_tuples(skb, tuple) < 0) 2930 goto nla_put_failure; 2931 nla_nest_end(skb, nest_parms); 2932 2933 return 0; 2934 2935 nla_put_failure: 2936 return -1; 2937 } 2938 2939 static int ctnetlink_exp_dump_mask(struct sk_buff *skb, 2940 const struct nf_conntrack_tuple *tuple, 2941 const struct nf_conntrack_tuple_mask *mask) 2942 { 2943 const struct nf_conntrack_l4proto *l4proto; 2944 struct nf_conntrack_tuple m; 2945 struct nlattr *nest_parms; 2946 int ret; 2947 2948 memset(&m, 0xFF, sizeof(m)); 2949 memcpy(&m.src.u3, &mask->src.u3, sizeof(m.src.u3)); 2950 m.src.u.all = mask->src.u.all; 2951 m.src.l3num = tuple->src.l3num; 2952 m.dst.protonum = tuple->dst.protonum; 2953 2954 nest_parms = nla_nest_start(skb, CTA_EXPECT_MASK); 2955 if (!nest_parms) 2956 goto nla_put_failure; 2957 2958 rcu_read_lock(); 2959 ret = ctnetlink_dump_tuples_ip(skb, &m); 2960 if (ret >= 0) { 2961 l4proto = nf_ct_l4proto_find(tuple->dst.protonum); 2962 ret = ctnetlink_dump_tuples_proto(skb, &m, l4proto); 2963 } 2964 rcu_read_unlock(); 2965 2966 if (unlikely(ret < 0)) 2967 goto nla_put_failure; 2968 2969 nla_nest_end(skb, nest_parms); 2970 2971 return 0; 2972 2973 nla_put_failure: 2974 return -1; 2975 } 2976 2977 #if IS_ENABLED(CONFIG_NF_NAT) 2978 static const union nf_inet_addr any_addr; 2979 #endif 2980 2981 static __be32 nf_expect_get_id(const struct nf_conntrack_expect *exp) 2982 { 2983 static siphash_aligned_key_t exp_id_seed; 2984 unsigned long a, b, c, d; 2985 2986 net_get_random_once(&exp_id_seed, sizeof(exp_id_seed)); 2987 2988 a = (unsigned long)exp; 2989 b = (unsigned long)exp->helper; 2990 c = (unsigned long)exp->master; 2991 d = (unsigned long)siphash(&exp->tuple, sizeof(exp->tuple), &exp_id_seed); 2992 2993 #ifdef CONFIG_64BIT 2994 return (__force __be32)siphash_4u64((u64)a, (u64)b, (u64)c, (u64)d, &exp_id_seed); 2995 #else 2996 return (__force __be32)siphash_4u32((u32)a, (u32)b, (u32)c, (u32)d, &exp_id_seed); 2997 #endif 2998 } 2999 3000 static int 3001 ctnetlink_exp_dump_expect(struct sk_buff *skb, 3002 const struct nf_conntrack_expect *exp) 3003 { 3004 __s32 timeout = (__s32)(READ_ONCE(exp->timeout) - nfct_time_stamp) / HZ; 3005 struct nf_conntrack_helper *helper; 3006 #if IS_ENABLED(CONFIG_NF_NAT) 3007 struct nlattr *nest_parms; 3008 struct nf_conntrack_tuple nat_tuple = {}; 3009 #endif 3010 struct nf_ct_helper_expectfn *expfn; 3011 3012 if (timeout < 0) 3013 timeout = 0; 3014 3015 if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0) 3016 goto nla_put_failure; 3017 if (ctnetlink_exp_dump_mask(skb, &exp->tuple, &exp->mask) < 0) 3018 goto nla_put_failure; 3019 if (ctnetlink_exp_dump_tuple(skb, &exp->master_tuple, CTA_EXPECT_MASTER) < 0) 3020 goto nla_put_failure; 3021 3022 #if IS_ENABLED(CONFIG_NF_NAT) 3023 if (!nf_inet_addr_cmp(&exp->saved_addr, &any_addr) || 3024 exp->saved_proto.all) { 3025 nest_parms = nla_nest_start(skb, CTA_EXPECT_NAT); 3026 if (!nest_parms) 3027 goto nla_put_failure; 3028 3029 if (nla_put_be32(skb, CTA_EXPECT_NAT_DIR, htonl(exp->dir))) 3030 goto nla_put_failure; 3031 3032 nat_tuple.src.l3num = exp->master_tuple.src.l3num; 3033 nat_tuple.src.u3 = exp->saved_addr; 3034 nat_tuple.dst.protonum = exp->master_tuple.dst.protonum; 3035 nat_tuple.src.u = exp->saved_proto; 3036 3037 if (ctnetlink_exp_dump_tuple(skb, &nat_tuple, 3038 CTA_EXPECT_NAT_TUPLE) < 0) 3039 goto nla_put_failure; 3040 nla_nest_end(skb, nest_parms); 3041 } 3042 #endif 3043 if (nla_put_be32(skb, CTA_EXPECT_TIMEOUT, htonl(timeout)) || 3044 nla_put_be32(skb, CTA_EXPECT_ID, nf_expect_get_id(exp)) || 3045 nla_put_be32(skb, CTA_EXPECT_FLAGS, htonl(exp->flags)) || 3046 nla_put_be32(skb, CTA_EXPECT_CLASS, htonl(exp->class))) 3047 goto nla_put_failure; 3048 3049 helper = rcu_dereference(exp->helper); 3050 if (helper && 3051 nla_put_string(skb, CTA_EXPECT_HELP_NAME, helper->name)) 3052 goto nla_put_failure; 3053 3054 expfn = nf_ct_helper_expectfn_find_by_symbol(exp->expectfn); 3055 if (expfn != NULL && 3056 nla_put_string(skb, CTA_EXPECT_FN, expfn->name)) 3057 goto nla_put_failure; 3058 3059 return 0; 3060 3061 nla_put_failure: 3062 return -1; 3063 } 3064 3065 static int 3066 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 portid, u32 seq, 3067 int event, const struct nf_conntrack_expect *exp) 3068 { 3069 struct nlmsghdr *nlh; 3070 unsigned int flags = portid ? NLM_F_MULTI : 0; 3071 3072 event = nfnl_msg_type(NFNL_SUBSYS_CTNETLINK_EXP, event); 3073 nlh = nfnl_msg_put(skb, portid, seq, event, flags, 3074 exp->tuple.src.l3num, NFNETLINK_V0, 0); 3075 if (!nlh) 3076 goto nlmsg_failure; 3077 3078 if (ctnetlink_exp_dump_expect(skb, exp) < 0) 3079 goto nla_put_failure; 3080 3081 nlmsg_end(skb, nlh); 3082 return skb->len; 3083 3084 nlmsg_failure: 3085 nla_put_failure: 3086 nlmsg_cancel(skb, nlh); 3087 return -1; 3088 } 3089 3090 #ifdef CONFIG_NF_CONNTRACK_EVENTS 3091 static int 3092 ctnetlink_expect_event(unsigned int events, const struct nf_exp_event *item) 3093 { 3094 struct nf_conntrack_expect *exp = item->exp; 3095 struct net *net = nf_ct_exp_net(exp); 3096 struct nlmsghdr *nlh; 3097 struct sk_buff *skb; 3098 unsigned int type, group; 3099 int flags = 0; 3100 3101 if (events & (1 << IPEXP_DESTROY)) { 3102 type = IPCTNL_MSG_EXP_DELETE; 3103 group = NFNLGRP_CONNTRACK_EXP_DESTROY; 3104 } else if (events & (1 << IPEXP_NEW)) { 3105 type = IPCTNL_MSG_EXP_NEW; 3106 flags = NLM_F_CREATE|NLM_F_EXCL; 3107 group = NFNLGRP_CONNTRACK_EXP_NEW; 3108 } else 3109 return 0; 3110 3111 if (!item->report && !nfnetlink_has_listeners(net, group)) 3112 return 0; 3113 3114 skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC); 3115 if (skb == NULL) 3116 goto errout; 3117 3118 type = nfnl_msg_type(NFNL_SUBSYS_CTNETLINK_EXP, type); 3119 nlh = nfnl_msg_put(skb, item->portid, 0, type, flags, 3120 exp->tuple.src.l3num, NFNETLINK_V0, 0); 3121 if (!nlh) 3122 goto nlmsg_failure; 3123 3124 if (ctnetlink_exp_dump_expect(skb, exp) < 0) 3125 goto nla_put_failure; 3126 3127 nlmsg_end(skb, nlh); 3128 nfnetlink_send(skb, net, item->portid, group, item->report, GFP_ATOMIC); 3129 return 0; 3130 3131 nla_put_failure: 3132 nlmsg_cancel(skb, nlh); 3133 nlmsg_failure: 3134 kfree_skb(skb); 3135 errout: 3136 nfnetlink_set_err(net, 0, 0, -ENOBUFS); 3137 return 0; 3138 } 3139 #endif 3140 3141 static unsigned long ctnetlink_exp_id(const struct nf_conntrack_expect *exp) 3142 { 3143 unsigned long id = (unsigned long)exp; 3144 3145 id += nf_ct_get_id(exp->master); 3146 id += exp->class; 3147 3148 return id ? id : 1; 3149 } 3150 3151 static int 3152 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb) 3153 { 3154 struct net *net = sock_net(skb->sk); 3155 struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh); 3156 u_int8_t l3proto = nfmsg->nfgen_family; 3157 unsigned long last_id = cb->args[1]; 3158 struct nf_conntrack_expect *exp; 3159 3160 rcu_read_lock(); 3161 for (; cb->args[0] < nf_ct_expect_hsize; cb->args[0]++) { 3162 restart: 3163 hlist_for_each_entry_rcu(exp, &nf_ct_expect_hash[cb->args[0]], 3164 hnode) { 3165 if (nf_ct_exp_is_expired(exp)) 3166 continue; 3167 3168 if (l3proto && exp->tuple.src.l3num != l3proto) 3169 continue; 3170 3171 if (!net_eq(nf_ct_exp_net(exp), net)) 3172 continue; 3173 3174 if (cb->args[1]) { 3175 if (ctnetlink_exp_id(exp) != last_id) 3176 continue; 3177 cb->args[1] = 0; 3178 } 3179 if (ctnetlink_exp_fill_info(skb, 3180 NETLINK_CB(cb->skb).portid, 3181 cb->nlh->nlmsg_seq, 3182 IPCTNL_MSG_EXP_NEW, 3183 exp) < 0) { 3184 cb->args[1] = ctnetlink_exp_id(exp); 3185 goto out; 3186 } 3187 } 3188 if (cb->args[1]) { 3189 cb->args[1] = 0; 3190 goto restart; 3191 } 3192 } 3193 out: 3194 rcu_read_unlock(); 3195 return skb->len; 3196 } 3197 3198 static int 3199 ctnetlink_exp_ct_dump_table(struct sk_buff *skb, struct netlink_callback *cb) 3200 { 3201 struct nfgenmsg *nfmsg = nlmsg_data(cb->nlh); 3202 struct nf_conn *ct = cb->data; 3203 struct nf_conn_help *help; 3204 u_int8_t l3proto = nfmsg->nfgen_family; 3205 unsigned long last_id = cb->args[1]; 3206 struct nf_conntrack_expect *exp; 3207 3208 if (cb->args[0]) 3209 return 0; 3210 3211 help = nfct_help(ct); 3212 if (!help) 3213 return 0; 3214 3215 rcu_read_lock(); 3216 3217 restart: 3218 hlist_for_each_entry_rcu(exp, &help->expectations, lnode) { 3219 if (l3proto && exp->tuple.src.l3num != l3proto) 3220 continue; 3221 if (cb->args[1]) { 3222 if (ctnetlink_exp_id(exp) != last_id) 3223 continue; 3224 cb->args[1] = 0; 3225 } 3226 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).portid, 3227 cb->nlh->nlmsg_seq, 3228 IPCTNL_MSG_EXP_NEW, 3229 exp) < 0) { 3230 cb->args[1] = ctnetlink_exp_id(exp); 3231 goto out; 3232 } 3233 } 3234 if (cb->args[1]) { 3235 cb->args[1] = 0; 3236 goto restart; 3237 } 3238 cb->args[0] = 1; 3239 out: 3240 rcu_read_unlock(); 3241 return skb->len; 3242 } 3243 3244 static int ctnetlink_dump_exp_ct_start(struct netlink_callback *cb) 3245 { 3246 struct nf_conn *ct = cb->data; 3247 3248 if (!refcount_inc_not_zero(&ct->ct_general.use)) 3249 return -ENOENT; 3250 return 0; 3251 } 3252 3253 static int ctnetlink_dump_exp_ct_done(struct netlink_callback *cb) 3254 { 3255 struct nf_conn *ct = cb->data; 3256 3257 if (ct) 3258 nf_ct_put(ct); 3259 return 0; 3260 } 3261 3262 static int ctnetlink_dump_exp_ct(struct net *net, struct sock *ctnl, 3263 struct sk_buff *skb, 3264 const struct nlmsghdr *nlh, 3265 const struct nlattr * const cda[], 3266 struct netlink_ext_ack *extack) 3267 { 3268 int err; 3269 struct nfgenmsg *nfmsg = nlmsg_data(nlh); 3270 u_int8_t u3 = nfmsg->nfgen_family; 3271 struct nf_conntrack_tuple tuple; 3272 struct nf_conntrack_tuple_hash *h; 3273 struct nf_conn *ct; 3274 struct nf_conntrack_zone zone; 3275 struct netlink_dump_control c = { 3276 .dump = ctnetlink_exp_ct_dump_table, 3277 .start = ctnetlink_dump_exp_ct_start, 3278 .done = ctnetlink_dump_exp_ct_done, 3279 }; 3280 3281 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER, 3282 u3, NULL); 3283 if (err < 0) 3284 return err; 3285 3286 err = ctnetlink_parse_zone(cda[CTA_EXPECT_ZONE], &zone); 3287 if (err < 0) 3288 return err; 3289 3290 h = nf_conntrack_find_get(net, &zone, &tuple); 3291 if (!h) 3292 return -ENOENT; 3293 3294 ct = nf_ct_tuplehash_to_ctrack(h); 3295 /* No expectation linked to this connection tracking. */ 3296 if (!nfct_help(ct)) { 3297 nf_ct_put(ct); 3298 return 0; 3299 } 3300 3301 c.data = ct; 3302 3303 err = netlink_dump_start(ctnl, skb, nlh, &c); 3304 nf_ct_put(ct); 3305 3306 return err; 3307 } 3308 3309 static int ctnetlink_get_expect(struct sk_buff *skb, 3310 const struct nfnl_info *info, 3311 const struct nlattr * const cda[]) 3312 { 3313 u_int8_t u3 = info->nfmsg->nfgen_family; 3314 struct nf_conntrack_tuple tuple; 3315 struct nf_conntrack_expect *exp; 3316 struct nf_conntrack_zone zone; 3317 struct sk_buff *skb2; 3318 int err; 3319 3320 if (info->nlh->nlmsg_flags & NLM_F_DUMP) { 3321 if (cda[CTA_EXPECT_MASTER]) 3322 return ctnetlink_dump_exp_ct(info->net, info->sk, skb, 3323 info->nlh, cda, 3324 info->extack); 3325 else { 3326 struct netlink_dump_control c = { 3327 .dump = ctnetlink_exp_dump_table, 3328 }; 3329 return netlink_dump_start(info->sk, skb, info->nlh, &c); 3330 } 3331 } 3332 3333 err = ctnetlink_parse_zone(cda[CTA_EXPECT_ZONE], &zone); 3334 if (err < 0) 3335 return err; 3336 3337 if (cda[CTA_EXPECT_TUPLE]) 3338 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, 3339 u3, NULL); 3340 else if (cda[CTA_EXPECT_MASTER]) 3341 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER, 3342 u3, NULL); 3343 else 3344 return -EINVAL; 3345 3346 if (err < 0) 3347 return err; 3348 3349 skb2 = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL); 3350 if (!skb2) 3351 return -ENOMEM; 3352 3353 spin_lock_bh(&nf_conntrack_expect_lock); 3354 exp = nf_ct_expect_find_get(info->net, &zone, &tuple); 3355 if (!exp) { 3356 spin_unlock_bh(&nf_conntrack_expect_lock); 3357 kfree_skb(skb2); 3358 return -ENOENT; 3359 } 3360 3361 if (cda[CTA_EXPECT_ID]) { 3362 __be32 id = nla_get_be32(cda[CTA_EXPECT_ID]); 3363 3364 if (id != nf_expect_get_id(exp)) { 3365 nf_ct_expect_put(exp); 3366 spin_unlock_bh(&nf_conntrack_expect_lock); 3367 kfree_skb(skb2); 3368 return -ENOENT; 3369 } 3370 } 3371 3372 rcu_read_lock(); 3373 err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).portid, 3374 info->nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW, 3375 exp); 3376 rcu_read_unlock(); 3377 nf_ct_expect_put(exp); 3378 spin_unlock_bh(&nf_conntrack_expect_lock); 3379 3380 if (err <= 0) { 3381 kfree_skb(skb2); 3382 return -ENOMEM; 3383 } 3384 3385 return nfnetlink_unicast(skb2, info->net, NETLINK_CB(skb).portid); 3386 } 3387 3388 static bool expect_iter_name(struct nf_conntrack_expect *exp, void *data) 3389 { 3390 struct nf_conntrack_helper *helper; 3391 const char *name = data; 3392 3393 helper = rcu_dereference(exp->helper); 3394 if (!helper) 3395 return false; 3396 3397 return strcmp(helper->name, name) == 0; 3398 } 3399 3400 static bool expect_iter_all(struct nf_conntrack_expect *exp, void *data) 3401 { 3402 return true; 3403 } 3404 3405 static int ctnetlink_del_expect(struct sk_buff *skb, 3406 const struct nfnl_info *info, 3407 const struct nlattr * const cda[]) 3408 { 3409 u_int8_t u3 = info->nfmsg->nfgen_family; 3410 struct nf_conntrack_expect *exp; 3411 struct nf_conntrack_tuple tuple; 3412 struct nf_conntrack_zone zone; 3413 int err; 3414 3415 if (cda[CTA_EXPECT_TUPLE]) { 3416 /* delete a single expect by tuple */ 3417 err = ctnetlink_parse_zone(cda[CTA_EXPECT_ZONE], &zone); 3418 if (err < 0) 3419 return err; 3420 3421 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, 3422 u3, NULL); 3423 if (err < 0) 3424 return err; 3425 3426 spin_lock_bh(&nf_conntrack_expect_lock); 3427 3428 /* bump usage count to 2 */ 3429 exp = nf_ct_expect_find_get(info->net, &zone, &tuple); 3430 if (!exp) { 3431 spin_unlock_bh(&nf_conntrack_expect_lock); 3432 return -ENOENT; 3433 } 3434 3435 if (cda[CTA_EXPECT_ID]) { 3436 __be32 id = nla_get_be32(cda[CTA_EXPECT_ID]); 3437 3438 if (id != nf_expect_get_id(exp)) { 3439 nf_ct_expect_put(exp); 3440 spin_unlock_bh(&nf_conntrack_expect_lock); 3441 return -ENOENT; 3442 } 3443 } 3444 3445 /* after list removal, usage count == 1 */ 3446 nf_ct_unlink_expect_report(exp, NETLINK_CB(skb).portid, 3447 nlmsg_report(info->nlh)); 3448 spin_unlock_bh(&nf_conntrack_expect_lock); 3449 /* have to put what we 'get' above. 3450 * after this line usage count == 0 */ 3451 nf_ct_expect_put(exp); 3452 } else if (cda[CTA_EXPECT_HELP_NAME]) { 3453 char *name = nla_data(cda[CTA_EXPECT_HELP_NAME]); 3454 3455 nf_ct_expect_iterate_net(info->net, expect_iter_name, name, 3456 NETLINK_CB(skb).portid, 3457 nlmsg_report(info->nlh)); 3458 } else { 3459 /* This basically means we have to flush everything*/ 3460 nf_ct_expect_iterate_net(info->net, expect_iter_all, NULL, 3461 NETLINK_CB(skb).portid, 3462 nlmsg_report(info->nlh)); 3463 } 3464 3465 return 0; 3466 } 3467 static int 3468 ctnetlink_change_expect(struct nf_conntrack_expect *x, 3469 const struct nlattr * const cda[]) 3470 { 3471 if (cda[CTA_EXPECT_TIMEOUT]) 3472 WRITE_ONCE(x->timeout, nfct_time_stamp + 3473 ntohl(nla_get_be32(cda[CTA_EXPECT_TIMEOUT])) * HZ); 3474 3475 return 0; 3476 } 3477 3478 #if IS_ENABLED(CONFIG_NF_NAT) 3479 static const struct nla_policy exp_nat_nla_policy[CTA_EXPECT_NAT_MAX+1] = { 3480 [CTA_EXPECT_NAT_DIR] = NLA_POLICY_MAX(NLA_BE32, IP_CT_DIR_REPLY), 3481 [CTA_EXPECT_NAT_TUPLE] = { .type = NLA_NESTED }, 3482 }; 3483 #endif 3484 3485 static int 3486 ctnetlink_parse_expect_nat(const struct nlattr *attr, 3487 struct nf_conntrack_expect *exp, 3488 u_int8_t u3) 3489 { 3490 #if IS_ENABLED(CONFIG_NF_NAT) 3491 struct nlattr *tb[CTA_EXPECT_NAT_MAX+1]; 3492 struct nf_conntrack_tuple nat_tuple = {}; 3493 int err; 3494 3495 err = nla_parse_nested_deprecated(tb, CTA_EXPECT_NAT_MAX, attr, 3496 exp_nat_nla_policy, NULL); 3497 if (err < 0) 3498 return err; 3499 3500 if (!tb[CTA_EXPECT_NAT_DIR] || !tb[CTA_EXPECT_NAT_TUPLE]) 3501 return -EINVAL; 3502 3503 err = ctnetlink_parse_tuple((const struct nlattr * const *)tb, 3504 &nat_tuple, CTA_EXPECT_NAT_TUPLE, 3505 u3, NULL); 3506 if (err < 0) 3507 return err; 3508 3509 exp->saved_addr = nat_tuple.src.u3; 3510 exp->saved_proto = nat_tuple.src.u; 3511 exp->dir = ntohl(nla_get_be32(tb[CTA_EXPECT_NAT_DIR])); 3512 3513 return 0; 3514 #else 3515 return -EOPNOTSUPP; 3516 #endif 3517 } 3518 3519 static struct nf_conntrack_expect * 3520 ctnetlink_alloc_expect(const struct nlattr * const cda[], struct nf_conn *ct, 3521 const struct nf_conntrack_helper *assign_helper, 3522 struct nf_conntrack_tuple *tuple, 3523 struct nf_conntrack_tuple *mask) 3524 { 3525 struct net *net = read_pnet(&ct->ct_net); 3526 struct nf_conntrack_helper *helper; 3527 struct nf_conntrack_expect *exp; 3528 struct nf_conn_help *help; 3529 u32 class = 0; 3530 int err; 3531 3532 help = nfct_help(ct); 3533 if (!help) 3534 return ERR_PTR(-EOPNOTSUPP); 3535 3536 helper = rcu_dereference(help->helper); 3537 if (!helper) 3538 return ERR_PTR(-EOPNOTSUPP); 3539 3540 if (cda[CTA_EXPECT_CLASS]) { 3541 class = ntohl(nla_get_be32(cda[CTA_EXPECT_CLASS])); 3542 if (class > helper->expect_class_max) 3543 return ERR_PTR(-EINVAL); 3544 } 3545 exp = nf_ct_expect_alloc(ct); 3546 if (!exp) 3547 return ERR_PTR(-ENOMEM); 3548 3549 if (cda[CTA_EXPECT_FLAGS]) { 3550 exp->flags = ntohl(nla_get_be32(cda[CTA_EXPECT_FLAGS])); 3551 exp->flags &= ~NF_CT_EXPECT_USERSPACE; 3552 } else { 3553 exp->flags = 0; 3554 } 3555 if (cda[CTA_EXPECT_FN]) { 3556 const char *name = nla_data(cda[CTA_EXPECT_FN]); 3557 struct nf_ct_helper_expectfn *expfn; 3558 3559 expfn = nf_ct_helper_expectfn_find_by_name(name); 3560 if (expfn == NULL) { 3561 err = -EINVAL; 3562 goto err_out; 3563 } 3564 exp->expectfn = expfn->expectfn; 3565 } else 3566 exp->expectfn = NULL; 3567 3568 exp->class = class; 3569 exp->master = ct; 3570 write_pnet(&exp->net, net); 3571 #ifdef CONFIG_NF_CONNTRACK_ZONES 3572 exp->zone = ct->zone; 3573 #endif 3574 rcu_assign_pointer(exp->helper, helper); 3575 rcu_assign_pointer(exp->assign_helper, assign_helper); 3576 exp->master_tuple = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple; 3577 exp->tuple = *tuple; 3578 exp->mask.src.u3 = mask->src.u3; 3579 exp->mask.src.u.all = mask->src.u.all; 3580 3581 if (cda[CTA_EXPECT_NAT]) { 3582 err = ctnetlink_parse_expect_nat(cda[CTA_EXPECT_NAT], 3583 exp, nf_ct_l3num(ct)); 3584 if (err < 0) 3585 goto err_out; 3586 #if IS_ENABLED(CONFIG_NF_NAT) 3587 } else { 3588 memset(&exp->saved_addr, 0, sizeof(exp->saved_addr)); 3589 memset(&exp->saved_proto, 0, sizeof(exp->saved_proto)); 3590 exp->dir = 0; 3591 #endif 3592 } 3593 return exp; 3594 err_out: 3595 nf_ct_expect_put(exp); 3596 return ERR_PTR(err); 3597 } 3598 3599 static int 3600 ctnetlink_create_expect(struct net *net, 3601 const struct nf_conntrack_zone *zone, 3602 const struct nlattr * const cda[], 3603 u_int8_t u3, u32 portid, int report) 3604 { 3605 struct nf_conntrack_tuple tuple, mask, master_tuple; 3606 struct nf_conntrack_tuple_hash *h = NULL; 3607 struct nf_conntrack_expect *exp; 3608 struct nf_conn *ct; 3609 int err; 3610 3611 /* caller guarantees that those three CTA_EXPECT_* exist */ 3612 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, 3613 u3, NULL); 3614 if (err < 0) 3615 return err; 3616 err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK, 3617 u3, NULL); 3618 if (err < 0) 3619 return err; 3620 err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER, 3621 u3, NULL); 3622 if (err < 0) 3623 return err; 3624 3625 /* Look for master conntrack of this expectation */ 3626 h = nf_conntrack_find_get(net, zone, &master_tuple); 3627 if (!h) 3628 return -ENOENT; 3629 ct = nf_ct_tuplehash_to_ctrack(h); 3630 3631 rcu_read_lock(); 3632 exp = ctnetlink_alloc_expect(cda, ct, NULL, &tuple, &mask); 3633 if (IS_ERR(exp)) { 3634 err = PTR_ERR(exp); 3635 goto err_rcu; 3636 } 3637 3638 err = nf_ct_expect_related_report(exp, portid, report, 0); 3639 nf_ct_expect_put(exp); 3640 err_rcu: 3641 rcu_read_unlock(); 3642 nf_ct_put(ct); 3643 3644 return err; 3645 } 3646 3647 static int ctnetlink_new_expect(struct sk_buff *skb, 3648 const struct nfnl_info *info, 3649 const struct nlattr * const cda[]) 3650 { 3651 u_int8_t u3 = info->nfmsg->nfgen_family; 3652 struct nf_conntrack_tuple tuple; 3653 struct nf_conntrack_expect *exp; 3654 struct nf_conntrack_zone zone; 3655 int err; 3656 3657 if (!cda[CTA_EXPECT_TUPLE] 3658 || !cda[CTA_EXPECT_MASK] 3659 || !cda[CTA_EXPECT_MASTER]) 3660 return -EINVAL; 3661 3662 err = ctnetlink_parse_zone(cda[CTA_EXPECT_ZONE], &zone); 3663 if (err < 0) 3664 return err; 3665 3666 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE, 3667 u3, NULL); 3668 if (err < 0) 3669 return err; 3670 3671 spin_lock_bh(&nf_conntrack_expect_lock); 3672 exp = __nf_ct_expect_find(info->net, &zone, &tuple); 3673 if (!exp) { 3674 spin_unlock_bh(&nf_conntrack_expect_lock); 3675 err = -ENOENT; 3676 if (info->nlh->nlmsg_flags & NLM_F_CREATE) { 3677 err = ctnetlink_create_expect(info->net, &zone, cda, u3, 3678 NETLINK_CB(skb).portid, 3679 nlmsg_report(info->nlh)); 3680 } 3681 return err; 3682 } 3683 3684 err = -EEXIST; 3685 if (!(info->nlh->nlmsg_flags & NLM_F_EXCL)) 3686 err = ctnetlink_change_expect(exp, cda); 3687 spin_unlock_bh(&nf_conntrack_expect_lock); 3688 3689 return err; 3690 } 3691 3692 static int 3693 ctnetlink_exp_stat_fill_info(struct sk_buff *skb, u32 portid, u32 seq, int cpu, 3694 const struct ip_conntrack_stat *st) 3695 { 3696 struct nlmsghdr *nlh; 3697 unsigned int flags = portid ? NLM_F_MULTI : 0, event; 3698 3699 event = nfnl_msg_type(NFNL_SUBSYS_CTNETLINK, 3700 IPCTNL_MSG_EXP_GET_STATS_CPU); 3701 nlh = nfnl_msg_put(skb, portid, seq, event, flags, AF_UNSPEC, 3702 NFNETLINK_V0, htons(cpu)); 3703 if (!nlh) 3704 goto nlmsg_failure; 3705 3706 if (nla_put_be32(skb, CTA_STATS_EXP_NEW, htonl(st->expect_new)) || 3707 nla_put_be32(skb, CTA_STATS_EXP_CREATE, htonl(st->expect_create)) || 3708 nla_put_be32(skb, CTA_STATS_EXP_DELETE, htonl(st->expect_delete))) 3709 goto nla_put_failure; 3710 3711 nlmsg_end(skb, nlh); 3712 return skb->len; 3713 3714 nla_put_failure: 3715 nlmsg_failure: 3716 nlmsg_cancel(skb, nlh); 3717 return -1; 3718 } 3719 3720 static int 3721 ctnetlink_exp_stat_cpu_dump(struct sk_buff *skb, struct netlink_callback *cb) 3722 { 3723 int cpu; 3724 struct net *net = sock_net(skb->sk); 3725 3726 if (cb->args[0] == nr_cpu_ids) 3727 return 0; 3728 3729 for (cpu = cb->args[0]; cpu < nr_cpu_ids; cpu++) { 3730 const struct ip_conntrack_stat *st; 3731 3732 if (!cpu_possible(cpu)) 3733 continue; 3734 3735 st = per_cpu_ptr(net->ct.stat, cpu); 3736 if (ctnetlink_exp_stat_fill_info(skb, NETLINK_CB(cb->skb).portid, 3737 cb->nlh->nlmsg_seq, 3738 cpu, st) < 0) 3739 break; 3740 } 3741 cb->args[0] = cpu; 3742 3743 return skb->len; 3744 } 3745 3746 static int ctnetlink_stat_exp_cpu(struct sk_buff *skb, 3747 const struct nfnl_info *info, 3748 const struct nlattr * const cda[]) 3749 { 3750 if (info->nlh->nlmsg_flags & NLM_F_DUMP) { 3751 struct netlink_dump_control c = { 3752 .dump = ctnetlink_exp_stat_cpu_dump, 3753 }; 3754 return netlink_dump_start(info->sk, skb, info->nlh, &c); 3755 } 3756 3757 return 0; 3758 } 3759 3760 #ifdef CONFIG_NF_CONNTRACK_EVENTS 3761 static struct nf_ct_event_notifier ctnl_notifier = { 3762 .ct_event = ctnetlink_conntrack_event, 3763 .exp_event = ctnetlink_expect_event, 3764 }; 3765 #endif 3766 3767 static const struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = { 3768 [IPCTNL_MSG_CT_NEW] = { 3769 .call = ctnetlink_new_conntrack, 3770 .type = NFNL_CB_MUTEX, 3771 .attr_count = CTA_MAX, 3772 .policy = ct_nla_policy 3773 }, 3774 [IPCTNL_MSG_CT_GET] = { 3775 .call = ctnetlink_get_conntrack, 3776 .type = NFNL_CB_MUTEX, 3777 .attr_count = CTA_MAX, 3778 .policy = ct_nla_policy 3779 }, 3780 [IPCTNL_MSG_CT_DELETE] = { 3781 .call = ctnetlink_del_conntrack, 3782 .type = NFNL_CB_MUTEX, 3783 .attr_count = CTA_MAX, 3784 .policy = ct_nla_policy 3785 }, 3786 [IPCTNL_MSG_CT_GET_CTRZERO] = { 3787 .call = ctnetlink_get_conntrack, 3788 .type = NFNL_CB_MUTEX, 3789 .attr_count = CTA_MAX, 3790 .policy = ct_nla_policy 3791 }, 3792 [IPCTNL_MSG_CT_GET_STATS_CPU] = { 3793 .call = ctnetlink_stat_ct_cpu, 3794 .type = NFNL_CB_MUTEX, 3795 }, 3796 [IPCTNL_MSG_CT_GET_STATS] = { 3797 .call = ctnetlink_stat_ct, 3798 .type = NFNL_CB_MUTEX, 3799 }, 3800 [IPCTNL_MSG_CT_GET_DYING] = { 3801 .call = ctnetlink_get_ct_dying, 3802 .type = NFNL_CB_MUTEX, 3803 }, 3804 [IPCTNL_MSG_CT_GET_UNCONFIRMED] = { 3805 .call = ctnetlink_get_ct_unconfirmed, 3806 .type = NFNL_CB_MUTEX, 3807 }, 3808 }; 3809 3810 static const struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = { 3811 [IPCTNL_MSG_EXP_GET] = { 3812 .call = ctnetlink_get_expect, 3813 .type = NFNL_CB_MUTEX, 3814 .attr_count = CTA_EXPECT_MAX, 3815 .policy = exp_nla_policy 3816 }, 3817 [IPCTNL_MSG_EXP_NEW] = { 3818 .call = ctnetlink_new_expect, 3819 .type = NFNL_CB_MUTEX, 3820 .attr_count = CTA_EXPECT_MAX, 3821 .policy = exp_nla_policy 3822 }, 3823 [IPCTNL_MSG_EXP_DELETE] = { 3824 .call = ctnetlink_del_expect, 3825 .type = NFNL_CB_MUTEX, 3826 .attr_count = CTA_EXPECT_MAX, 3827 .policy = exp_nla_policy 3828 }, 3829 [IPCTNL_MSG_EXP_GET_STATS_CPU] = { 3830 .call = ctnetlink_stat_exp_cpu, 3831 .type = NFNL_CB_MUTEX, 3832 }, 3833 }; 3834 3835 static const struct nfnetlink_subsystem ctnl_subsys = { 3836 .name = "conntrack", 3837 .subsys_id = NFNL_SUBSYS_CTNETLINK, 3838 .cb_count = IPCTNL_MSG_MAX, 3839 .cb = ctnl_cb, 3840 }; 3841 3842 static const struct nfnetlink_subsystem ctnl_exp_subsys = { 3843 .name = "conntrack_expect", 3844 .subsys_id = NFNL_SUBSYS_CTNETLINK_EXP, 3845 .cb_count = IPCTNL_MSG_EXP_MAX, 3846 .cb = ctnl_exp_cb, 3847 }; 3848 3849 MODULE_ALIAS("ip_conntrack_netlink"); 3850 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK); 3851 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK_EXP); 3852 3853 static int __net_init ctnetlink_net_init(struct net *net) 3854 { 3855 #ifdef CONFIG_NF_CONNTRACK_EVENTS 3856 nf_conntrack_register_notifier(net, &ctnl_notifier); 3857 #endif 3858 return 0; 3859 } 3860 3861 static void ctnetlink_net_pre_exit(struct net *net) 3862 { 3863 #ifdef CONFIG_NF_CONNTRACK_EVENTS 3864 nf_conntrack_unregister_notifier(net); 3865 #endif 3866 } 3867 3868 static struct pernet_operations ctnetlink_net_ops = { 3869 .init = ctnetlink_net_init, 3870 .pre_exit = ctnetlink_net_pre_exit, 3871 }; 3872 3873 static int __init ctnetlink_init(void) 3874 { 3875 int ret; 3876 3877 NL_ASSERT_CTX_FITS(struct ctnetlink_list_dump_ctx); 3878 3879 ret = nfnetlink_subsys_register(&ctnl_subsys); 3880 if (ret < 0) { 3881 pr_err("ctnetlink_init: cannot register with nfnetlink.\n"); 3882 goto err_out; 3883 } 3884 3885 ret = nfnetlink_subsys_register(&ctnl_exp_subsys); 3886 if (ret < 0) { 3887 pr_err("ctnetlink_init: cannot register exp with nfnetlink.\n"); 3888 goto err_unreg_subsys; 3889 } 3890 3891 ret = register_pernet_subsys(&ctnetlink_net_ops); 3892 if (ret < 0) { 3893 pr_err("ctnetlink_init: cannot register pernet operations\n"); 3894 goto err_unreg_exp_subsys; 3895 } 3896 #ifdef CONFIG_NETFILTER_NETLINK_GLUE_CT 3897 /* setup interaction between nf_queue and nf_conntrack_netlink. */ 3898 RCU_INIT_POINTER(nfnl_ct_hook, &ctnetlink_glue_hook); 3899 #endif 3900 return 0; 3901 3902 err_unreg_exp_subsys: 3903 nfnetlink_subsys_unregister(&ctnl_exp_subsys); 3904 err_unreg_subsys: 3905 nfnetlink_subsys_unregister(&ctnl_subsys); 3906 err_out: 3907 return ret; 3908 } 3909 3910 static void __exit ctnetlink_exit(void) 3911 { 3912 unregister_pernet_subsys(&ctnetlink_net_ops); 3913 nfnetlink_subsys_unregister(&ctnl_exp_subsys); 3914 nfnetlink_subsys_unregister(&ctnl_subsys); 3915 #ifdef CONFIG_NETFILTER_NETLINK_GLUE_CT 3916 RCU_INIT_POINTER(nfnl_ct_hook, NULL); 3917 #endif 3918 synchronize_rcu(); 3919 } 3920 3921 module_init(ctnetlink_init); 3922 module_exit(ctnetlink_exit); 3923