1 // SPDX-License-Identifier: GPL-2.0-only 2 /* 3 * This is a module which is used for queueing packets and communicating with 4 * userspace via nfnetlink. 5 * 6 * (C) 2005 by Harald Welte <[email protected]> 7 * (C) 2007 by Patrick McHardy <[email protected]> 8 * 9 * Based on the old ipv4-only ip_queue.c: 10 * (C) 2000-2002 James Morris <[email protected]> 11 * (C) 2003-2005 Netfilter Core Team <[email protected]> 12 */ 13 14 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt 15 16 #include <linux/module.h> 17 #include <linux/skbuff.h> 18 #include <linux/init.h> 19 #include <linux/spinlock.h> 20 #include <linux/slab.h> 21 #include <linux/notifier.h> 22 #include <linux/netdevice.h> 23 #include <linux/netfilter.h> 24 #include <linux/proc_fs.h> 25 #include <linux/netfilter_ipv4.h> 26 #include <linux/netfilter_ipv6.h> 27 #include <linux/netfilter_bridge.h> 28 #include <linux/netfilter/nfnetlink.h> 29 #include <linux/netfilter/nfnetlink_queue.h> 30 #include <linux/netfilter/nf_conntrack_common.h> 31 #include <linux/list.h> 32 #include <net/sock.h> 33 #include <net/tcp_states.h> 34 #include <net/netfilter/nf_queue.h> 35 #include <net/netns/generic.h> 36 37 #include <linux/atomic.h> 38 39 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 40 #include "../bridge/br_private.h" 41 #endif 42 43 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 44 #include <net/netfilter/nf_conntrack.h> 45 #endif 46 47 #define NFQNL_QMAX_DEFAULT 1024 48 49 /* We're using struct nlattr which has 16bit nla_len. Note that nla_len 50 * includes the header length. Thus, the maximum packet length that we 51 * support is 65531 bytes. We send truncated packets if the specified length 52 * is larger than that. Userspace can check for presence of NFQA_CAP_LEN 53 * attribute to detect truncation. 54 */ 55 #define NFQNL_MAX_COPY_RANGE (0xffff - NLA_HDRLEN) 56 57 struct nfqnl_instance { 58 struct hlist_node hlist; /* global list of queues */ 59 struct rcu_head rcu; 60 61 u32 peer_portid; 62 unsigned int queue_maxlen; 63 unsigned int copy_range; 64 unsigned int queue_dropped; 65 unsigned int queue_user_dropped; 66 67 68 u_int16_t queue_num; /* number of this queue */ 69 u_int8_t copy_mode; 70 u_int32_t flags; /* Set using NFQA_CFG_FLAGS */ 71 /* 72 * Following fields are dirtied for each queued packet, 73 * keep them in same cache line if possible. 74 */ 75 spinlock_t lock ____cacheline_aligned_in_smp; 76 unsigned int queue_total; 77 unsigned int id_sequence; /* 'sequence' of pkt ids */ 78 struct list_head queue_list; /* packets in queue */ 79 }; 80 81 typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long); 82 83 static unsigned int nfnl_queue_net_id __read_mostly; 84 85 #define INSTANCE_BUCKETS 16 86 struct nfnl_queue_net { 87 spinlock_t instances_lock; 88 struct hlist_head instance_table[INSTANCE_BUCKETS]; 89 }; 90 91 static struct nfnl_queue_net *nfnl_queue_pernet(struct net *net) 92 { 93 return net_generic(net, nfnl_queue_net_id); 94 } 95 96 static inline u_int8_t instance_hashfn(u_int16_t queue_num) 97 { 98 return ((queue_num >> 8) ^ queue_num) % INSTANCE_BUCKETS; 99 } 100 101 static struct nfqnl_instance * 102 instance_lookup(struct nfnl_queue_net *q, u_int16_t queue_num) 103 { 104 struct hlist_head *head; 105 struct nfqnl_instance *inst; 106 107 head = &q->instance_table[instance_hashfn(queue_num)]; 108 hlist_for_each_entry_rcu(inst, head, hlist) { 109 if (inst->queue_num == queue_num) 110 return inst; 111 } 112 return NULL; 113 } 114 115 static struct nfqnl_instance * 116 instance_create(struct nfnl_queue_net *q, u_int16_t queue_num, u32 portid) 117 { 118 struct nfqnl_instance *inst; 119 unsigned int h; 120 int err; 121 122 spin_lock(&q->instances_lock); 123 if (instance_lookup(q, queue_num)) { 124 err = -EEXIST; 125 goto out_unlock; 126 } 127 128 inst = kzalloc(sizeof(*inst), GFP_ATOMIC); 129 if (!inst) { 130 err = -ENOMEM; 131 goto out_unlock; 132 } 133 134 inst->queue_num = queue_num; 135 inst->peer_portid = portid; 136 inst->queue_maxlen = NFQNL_QMAX_DEFAULT; 137 inst->copy_range = NFQNL_MAX_COPY_RANGE; 138 inst->copy_mode = NFQNL_COPY_NONE; 139 spin_lock_init(&inst->lock); 140 INIT_LIST_HEAD(&inst->queue_list); 141 142 if (!try_module_get(THIS_MODULE)) { 143 err = -EAGAIN; 144 goto out_free; 145 } 146 147 h = instance_hashfn(queue_num); 148 hlist_add_head_rcu(&inst->hlist, &q->instance_table[h]); 149 150 spin_unlock(&q->instances_lock); 151 152 return inst; 153 154 out_free: 155 kfree(inst); 156 out_unlock: 157 spin_unlock(&q->instances_lock); 158 return ERR_PTR(err); 159 } 160 161 static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, 162 unsigned long data); 163 164 static void 165 instance_destroy_rcu(struct rcu_head *head) 166 { 167 struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance, 168 rcu); 169 170 nfqnl_flush(inst, NULL, 0); 171 kfree(inst); 172 module_put(THIS_MODULE); 173 } 174 175 static void 176 __instance_destroy(struct nfqnl_instance *inst) 177 { 178 hlist_del_rcu(&inst->hlist); 179 call_rcu(&inst->rcu, instance_destroy_rcu); 180 } 181 182 static void 183 instance_destroy(struct nfnl_queue_net *q, struct nfqnl_instance *inst) 184 { 185 spin_lock(&q->instances_lock); 186 __instance_destroy(inst); 187 spin_unlock(&q->instances_lock); 188 } 189 190 static inline void 191 __enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry) 192 { 193 list_add_tail(&entry->list, &queue->queue_list); 194 queue->queue_total++; 195 } 196 197 static void 198 __dequeue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry) 199 { 200 list_del(&entry->list); 201 queue->queue_total--; 202 } 203 204 static struct nf_queue_entry * 205 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id) 206 { 207 struct nf_queue_entry *entry = NULL, *i; 208 209 spin_lock_bh(&queue->lock); 210 211 list_for_each_entry(i, &queue->queue_list, list) { 212 if (i->id == id) { 213 entry = i; 214 break; 215 } 216 } 217 218 if (entry) 219 __dequeue_entry(queue, entry); 220 221 spin_unlock_bh(&queue->lock); 222 223 return entry; 224 } 225 226 static void nfqnl_reinject(struct nf_queue_entry *entry, unsigned int verdict) 227 { 228 struct nf_ct_hook *ct_hook; 229 int err; 230 231 if (verdict == NF_ACCEPT || 232 verdict == NF_REPEAT || 233 verdict == NF_STOP) { 234 rcu_read_lock(); 235 ct_hook = rcu_dereference(nf_ct_hook); 236 if (ct_hook) { 237 err = ct_hook->update(entry->state.net, entry->skb); 238 if (err < 0) 239 verdict = NF_DROP; 240 } 241 rcu_read_unlock(); 242 } 243 nf_reinject(entry, verdict); 244 } 245 246 static void 247 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data) 248 { 249 struct nf_queue_entry *entry, *next; 250 251 spin_lock_bh(&queue->lock); 252 list_for_each_entry_safe(entry, next, &queue->queue_list, list) { 253 if (!cmpfn || cmpfn(entry, data)) { 254 list_del(&entry->list); 255 queue->queue_total--; 256 nfqnl_reinject(entry, NF_DROP); 257 } 258 } 259 spin_unlock_bh(&queue->lock); 260 } 261 262 static int 263 nfqnl_put_packet_info(struct sk_buff *nlskb, struct sk_buff *packet, 264 bool csum_verify) 265 { 266 __u32 flags = 0; 267 268 if (packet->ip_summed == CHECKSUM_PARTIAL) 269 flags = NFQA_SKB_CSUMNOTREADY; 270 else if (csum_verify) 271 flags = NFQA_SKB_CSUM_NOTVERIFIED; 272 273 if (skb_is_gso(packet)) 274 flags |= NFQA_SKB_GSO; 275 276 return flags ? nla_put_be32(nlskb, NFQA_SKB_INFO, htonl(flags)) : 0; 277 } 278 279 static int nfqnl_put_sk_uidgid(struct sk_buff *skb, struct sock *sk) 280 { 281 const struct cred *cred; 282 283 if (!sk_fullsock(sk)) 284 return 0; 285 286 read_lock_bh(&sk->sk_callback_lock); 287 if (sk->sk_socket && sk->sk_socket->file) { 288 cred = sk->sk_socket->file->f_cred; 289 if (nla_put_be32(skb, NFQA_UID, 290 htonl(from_kuid_munged(&init_user_ns, cred->fsuid)))) 291 goto nla_put_failure; 292 if (nla_put_be32(skb, NFQA_GID, 293 htonl(from_kgid_munged(&init_user_ns, cred->fsgid)))) 294 goto nla_put_failure; 295 } 296 read_unlock_bh(&sk->sk_callback_lock); 297 return 0; 298 299 nla_put_failure: 300 read_unlock_bh(&sk->sk_callback_lock); 301 return -1; 302 } 303 304 static u32 nfqnl_get_sk_secctx(struct sk_buff *skb, char **secdata) 305 { 306 u32 seclen = 0; 307 #if IS_ENABLED(CONFIG_NETWORK_SECMARK) 308 if (!skb || !sk_fullsock(skb->sk)) 309 return 0; 310 311 read_lock_bh(&skb->sk->sk_callback_lock); 312 313 if (skb->secmark) 314 security_secid_to_secctx(skb->secmark, secdata, &seclen); 315 316 read_unlock_bh(&skb->sk->sk_callback_lock); 317 #endif 318 return seclen; 319 } 320 321 static u32 nfqnl_get_bridge_size(struct nf_queue_entry *entry) 322 { 323 struct sk_buff *entskb = entry->skb; 324 u32 nlalen = 0; 325 326 if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb)) 327 return 0; 328 329 if (skb_vlan_tag_present(entskb)) 330 nlalen += nla_total_size(nla_total_size(sizeof(__be16)) + 331 nla_total_size(sizeof(__be16))); 332 333 if (entskb->network_header > entskb->mac_header) 334 nlalen += nla_total_size((entskb->network_header - 335 entskb->mac_header)); 336 337 return nlalen; 338 } 339 340 static int nfqnl_put_bridge(struct nf_queue_entry *entry, struct sk_buff *skb) 341 { 342 struct sk_buff *entskb = entry->skb; 343 344 if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb)) 345 return 0; 346 347 if (skb_vlan_tag_present(entskb)) { 348 struct nlattr *nest; 349 350 nest = nla_nest_start(skb, NFQA_VLAN); 351 if (!nest) 352 goto nla_put_failure; 353 354 if (nla_put_be16(skb, NFQA_VLAN_TCI, htons(entskb->vlan_tci)) || 355 nla_put_be16(skb, NFQA_VLAN_PROTO, entskb->vlan_proto)) 356 goto nla_put_failure; 357 358 nla_nest_end(skb, nest); 359 } 360 361 if (entskb->mac_header < entskb->network_header) { 362 int len = (int)(entskb->network_header - entskb->mac_header); 363 364 if (nla_put(skb, NFQA_L2HDR, len, skb_mac_header(entskb))) 365 goto nla_put_failure; 366 } 367 368 return 0; 369 370 nla_put_failure: 371 return -1; 372 } 373 374 static struct sk_buff * 375 nfqnl_build_packet_message(struct net *net, struct nfqnl_instance *queue, 376 struct nf_queue_entry *entry, 377 __be32 **packet_id_ptr) 378 { 379 size_t size; 380 size_t data_len = 0, cap_len = 0; 381 unsigned int hlen = 0; 382 struct sk_buff *skb; 383 struct nlattr *nla; 384 struct nfqnl_msg_packet_hdr *pmsg; 385 struct nlmsghdr *nlh; 386 struct sk_buff *entskb = entry->skb; 387 struct net_device *indev; 388 struct net_device *outdev; 389 struct nf_conn *ct = NULL; 390 enum ip_conntrack_info ctinfo; 391 struct nfnl_ct_hook *nfnl_ct; 392 bool csum_verify; 393 char *secdata = NULL; 394 u32 seclen = 0; 395 396 size = nlmsg_total_size(sizeof(struct nfgenmsg)) 397 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr)) 398 + nla_total_size(sizeof(u_int32_t)) /* ifindex */ 399 + nla_total_size(sizeof(u_int32_t)) /* ifindex */ 400 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 401 + nla_total_size(sizeof(u_int32_t)) /* ifindex */ 402 + nla_total_size(sizeof(u_int32_t)) /* ifindex */ 403 #endif 404 + nla_total_size(sizeof(u_int32_t)) /* mark */ 405 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw)) 406 + nla_total_size(sizeof(u_int32_t)) /* skbinfo */ 407 + nla_total_size(sizeof(u_int32_t)); /* cap_len */ 408 409 if (entskb->tstamp) 410 size += nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp)); 411 412 size += nfqnl_get_bridge_size(entry); 413 414 if (entry->state.hook <= NF_INET_FORWARD || 415 (entry->state.hook == NF_INET_POST_ROUTING && entskb->sk == NULL)) 416 csum_verify = !skb_csum_unnecessary(entskb); 417 else 418 csum_verify = false; 419 420 outdev = entry->state.out; 421 422 switch ((enum nfqnl_config_mode)READ_ONCE(queue->copy_mode)) { 423 case NFQNL_COPY_META: 424 case NFQNL_COPY_NONE: 425 break; 426 427 case NFQNL_COPY_PACKET: 428 if (!(queue->flags & NFQA_CFG_F_GSO) && 429 entskb->ip_summed == CHECKSUM_PARTIAL && 430 skb_checksum_help(entskb)) 431 return NULL; 432 433 data_len = READ_ONCE(queue->copy_range); 434 if (data_len > entskb->len) 435 data_len = entskb->len; 436 437 hlen = skb_zerocopy_headlen(entskb); 438 hlen = min_t(unsigned int, hlen, data_len); 439 size += sizeof(struct nlattr) + hlen; 440 cap_len = entskb->len; 441 break; 442 } 443 444 nfnl_ct = rcu_dereference(nfnl_ct_hook); 445 446 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 447 if (queue->flags & NFQA_CFG_F_CONNTRACK) { 448 if (nfnl_ct != NULL) { 449 ct = nf_ct_get(entskb, &ctinfo); 450 if (ct != NULL) 451 size += nfnl_ct->build_size(ct); 452 } 453 } 454 #endif 455 456 if (queue->flags & NFQA_CFG_F_UID_GID) { 457 size += (nla_total_size(sizeof(u_int32_t)) /* uid */ 458 + nla_total_size(sizeof(u_int32_t))); /* gid */ 459 } 460 461 if ((queue->flags & NFQA_CFG_F_SECCTX) && entskb->sk) { 462 seclen = nfqnl_get_sk_secctx(entskb, &secdata); 463 if (seclen) 464 size += nla_total_size(seclen); 465 } 466 467 skb = alloc_skb(size, GFP_ATOMIC); 468 if (!skb) { 469 skb_tx_error(entskb); 470 goto nlmsg_failure; 471 } 472 473 nlh = nfnl_msg_put(skb, 0, 0, 474 nfnl_msg_type(NFNL_SUBSYS_QUEUE, NFQNL_MSG_PACKET), 475 0, entry->state.pf, NFNETLINK_V0, 476 htons(queue->queue_num)); 477 if (!nlh) { 478 skb_tx_error(entskb); 479 kfree_skb(skb); 480 goto nlmsg_failure; 481 } 482 483 nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg)); 484 pmsg = nla_data(nla); 485 pmsg->hw_protocol = entskb->protocol; 486 pmsg->hook = entry->state.hook; 487 *packet_id_ptr = &pmsg->packet_id; 488 489 indev = entry->state.in; 490 if (indev) { 491 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 492 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex))) 493 goto nla_put_failure; 494 #else 495 if (entry->state.pf == PF_BRIDGE) { 496 /* Case 1: indev is physical input device, we need to 497 * look for bridge group (when called from 498 * netfilter_bridge) */ 499 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV, 500 htonl(indev->ifindex)) || 501 /* this is the bridge group "brX" */ 502 /* rcu_read_lock()ed by __nf_queue */ 503 nla_put_be32(skb, NFQA_IFINDEX_INDEV, 504 htonl(br_port_get_rcu(indev)->br->dev->ifindex))) 505 goto nla_put_failure; 506 } else { 507 int physinif; 508 509 /* Case 2: indev is bridge group, we need to look for 510 * physical device (when called from ipv4) */ 511 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, 512 htonl(indev->ifindex))) 513 goto nla_put_failure; 514 515 physinif = nf_bridge_get_physinif(entskb); 516 if (physinif && 517 nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV, 518 htonl(physinif))) 519 goto nla_put_failure; 520 } 521 #endif 522 } 523 524 if (outdev) { 525 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 526 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex))) 527 goto nla_put_failure; 528 #else 529 if (entry->state.pf == PF_BRIDGE) { 530 /* Case 1: outdev is physical output device, we need to 531 * look for bridge group (when called from 532 * netfilter_bridge) */ 533 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV, 534 htonl(outdev->ifindex)) || 535 /* this is the bridge group "brX" */ 536 /* rcu_read_lock()ed by __nf_queue */ 537 nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, 538 htonl(br_port_get_rcu(outdev)->br->dev->ifindex))) 539 goto nla_put_failure; 540 } else { 541 int physoutif; 542 543 /* Case 2: outdev is bridge group, we need to look for 544 * physical output device (when called from ipv4) */ 545 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, 546 htonl(outdev->ifindex))) 547 goto nla_put_failure; 548 549 physoutif = nf_bridge_get_physoutif(entskb); 550 if (physoutif && 551 nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV, 552 htonl(physoutif))) 553 goto nla_put_failure; 554 } 555 #endif 556 } 557 558 if (entskb->mark && 559 nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark))) 560 goto nla_put_failure; 561 562 if (indev && entskb->dev && 563 entskb->mac_header != entskb->network_header) { 564 struct nfqnl_msg_packet_hw phw; 565 int len; 566 567 memset(&phw, 0, sizeof(phw)); 568 len = dev_parse_header(entskb, phw.hw_addr); 569 if (len) { 570 phw.hw_addrlen = htons(len); 571 if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw)) 572 goto nla_put_failure; 573 } 574 } 575 576 if (nfqnl_put_bridge(entry, skb) < 0) 577 goto nla_put_failure; 578 579 if (entry->state.hook <= NF_INET_FORWARD && entskb->tstamp) { 580 struct nfqnl_msg_packet_timestamp ts; 581 struct timespec64 kts = ktime_to_timespec64(entskb->tstamp); 582 583 ts.sec = cpu_to_be64(kts.tv_sec); 584 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC); 585 586 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts)) 587 goto nla_put_failure; 588 } 589 590 if ((queue->flags & NFQA_CFG_F_UID_GID) && entskb->sk && 591 nfqnl_put_sk_uidgid(skb, entskb->sk) < 0) 592 goto nla_put_failure; 593 594 if (seclen && nla_put(skb, NFQA_SECCTX, seclen, secdata)) 595 goto nla_put_failure; 596 597 if (ct && nfnl_ct->build(skb, ct, ctinfo, NFQA_CT, NFQA_CT_INFO) < 0) 598 goto nla_put_failure; 599 600 if (cap_len > data_len && 601 nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len))) 602 goto nla_put_failure; 603 604 if (nfqnl_put_packet_info(skb, entskb, csum_verify)) 605 goto nla_put_failure; 606 607 if (data_len) { 608 struct nlattr *nla; 609 610 if (skb_tailroom(skb) < sizeof(*nla) + hlen) 611 goto nla_put_failure; 612 613 nla = skb_put(skb, sizeof(*nla)); 614 nla->nla_type = NFQA_PAYLOAD; 615 nla->nla_len = nla_attr_size(data_len); 616 617 if (skb_zerocopy(skb, entskb, data_len, hlen)) 618 goto nla_put_failure; 619 } 620 621 nlh->nlmsg_len = skb->len; 622 if (seclen) 623 security_release_secctx(secdata, seclen); 624 return skb; 625 626 nla_put_failure: 627 skb_tx_error(entskb); 628 kfree_skb(skb); 629 net_err_ratelimited("nf_queue: error creating packet message\n"); 630 nlmsg_failure: 631 if (seclen) 632 security_release_secctx(secdata, seclen); 633 return NULL; 634 } 635 636 static bool nf_ct_drop_unconfirmed(const struct nf_queue_entry *entry) 637 { 638 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 639 static const unsigned long flags = IPS_CONFIRMED | IPS_DYING; 640 const struct nf_conn *ct = (void *)skb_nfct(entry->skb); 641 642 if (ct && ((ct->status & flags) == IPS_DYING)) 643 return true; 644 #endif 645 return false; 646 } 647 648 static int 649 __nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue, 650 struct nf_queue_entry *entry) 651 { 652 struct sk_buff *nskb; 653 int err = -ENOBUFS; 654 __be32 *packet_id_ptr; 655 int failopen = 0; 656 657 nskb = nfqnl_build_packet_message(net, queue, entry, &packet_id_ptr); 658 if (nskb == NULL) { 659 err = -ENOMEM; 660 goto err_out; 661 } 662 spin_lock_bh(&queue->lock); 663 664 if (nf_ct_drop_unconfirmed(entry)) 665 goto err_out_free_nskb; 666 667 if (queue->queue_total >= queue->queue_maxlen) { 668 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) { 669 failopen = 1; 670 err = 0; 671 } else { 672 queue->queue_dropped++; 673 net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n", 674 queue->queue_total); 675 } 676 goto err_out_free_nskb; 677 } 678 entry->id = ++queue->id_sequence; 679 *packet_id_ptr = htonl(entry->id); 680 681 /* nfnetlink_unicast will either free the nskb or add it to a socket */ 682 err = nfnetlink_unicast(nskb, net, queue->peer_portid); 683 if (err < 0) { 684 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) { 685 failopen = 1; 686 err = 0; 687 } else { 688 queue->queue_user_dropped++; 689 } 690 goto err_out_unlock; 691 } 692 693 __enqueue_entry(queue, entry); 694 695 spin_unlock_bh(&queue->lock); 696 return 0; 697 698 err_out_free_nskb: 699 kfree_skb(nskb); 700 err_out_unlock: 701 spin_unlock_bh(&queue->lock); 702 if (failopen) 703 nfqnl_reinject(entry, NF_ACCEPT); 704 err_out: 705 return err; 706 } 707 708 static struct nf_queue_entry * 709 nf_queue_entry_dup(struct nf_queue_entry *e) 710 { 711 struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC); 712 if (entry) 713 nf_queue_entry_get_refs(entry); 714 return entry; 715 } 716 717 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 718 /* When called from bridge netfilter, skb->data must point to MAC header 719 * before calling skb_gso_segment(). Else, original MAC header is lost 720 * and segmented skbs will be sent to wrong destination. 721 */ 722 static void nf_bridge_adjust_skb_data(struct sk_buff *skb) 723 { 724 if (nf_bridge_info_get(skb)) 725 __skb_push(skb, skb->network_header - skb->mac_header); 726 } 727 728 static void nf_bridge_adjust_segmented_data(struct sk_buff *skb) 729 { 730 if (nf_bridge_info_get(skb)) 731 __skb_pull(skb, skb->network_header - skb->mac_header); 732 } 733 #else 734 #define nf_bridge_adjust_skb_data(s) do {} while (0) 735 #define nf_bridge_adjust_segmented_data(s) do {} while (0) 736 #endif 737 738 static int 739 __nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue, 740 struct sk_buff *skb, struct nf_queue_entry *entry) 741 { 742 int ret = -ENOMEM; 743 struct nf_queue_entry *entry_seg; 744 745 nf_bridge_adjust_segmented_data(skb); 746 747 if (skb->next == NULL) { /* last packet, no need to copy entry */ 748 struct sk_buff *gso_skb = entry->skb; 749 entry->skb = skb; 750 ret = __nfqnl_enqueue_packet(net, queue, entry); 751 if (ret) 752 entry->skb = gso_skb; 753 return ret; 754 } 755 756 skb_mark_not_on_list(skb); 757 758 entry_seg = nf_queue_entry_dup(entry); 759 if (entry_seg) { 760 entry_seg->skb = skb; 761 ret = __nfqnl_enqueue_packet(net, queue, entry_seg); 762 if (ret) 763 nf_queue_entry_free(entry_seg); 764 } 765 return ret; 766 } 767 768 static int 769 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum) 770 { 771 unsigned int queued; 772 struct nfqnl_instance *queue; 773 struct sk_buff *skb, *segs, *nskb; 774 int err = -ENOBUFS; 775 struct net *net = entry->state.net; 776 struct nfnl_queue_net *q = nfnl_queue_pernet(net); 777 778 /* rcu_read_lock()ed by nf_hook_thresh */ 779 queue = instance_lookup(q, queuenum); 780 if (!queue) 781 return -ESRCH; 782 783 if (queue->copy_mode == NFQNL_COPY_NONE) 784 return -EINVAL; 785 786 skb = entry->skb; 787 788 switch (entry->state.pf) { 789 case NFPROTO_IPV4: 790 skb->protocol = htons(ETH_P_IP); 791 break; 792 case NFPROTO_IPV6: 793 skb->protocol = htons(ETH_P_IPV6); 794 break; 795 } 796 797 if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(skb)) 798 return __nfqnl_enqueue_packet(net, queue, entry); 799 800 nf_bridge_adjust_skb_data(skb); 801 segs = skb_gso_segment(skb, 0); 802 /* Does not use PTR_ERR to limit the number of error codes that can be 803 * returned by nf_queue. For instance, callers rely on -ESRCH to 804 * mean 'ignore this hook'. 805 */ 806 if (IS_ERR_OR_NULL(segs)) 807 goto out_err; 808 queued = 0; 809 err = 0; 810 skb_list_walk_safe(segs, segs, nskb) { 811 if (err == 0) 812 err = __nfqnl_enqueue_packet_gso(net, queue, 813 segs, entry); 814 if (err == 0) 815 queued++; 816 else 817 kfree_skb(segs); 818 } 819 820 if (queued) { 821 if (err) /* some segments are already queued */ 822 nf_queue_entry_free(entry); 823 kfree_skb(skb); 824 return 0; 825 } 826 out_err: 827 nf_bridge_adjust_segmented_data(skb); 828 return err; 829 } 830 831 static int 832 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e, int diff) 833 { 834 struct sk_buff *nskb; 835 836 if (diff < 0) { 837 if (pskb_trim(e->skb, data_len)) 838 return -ENOMEM; 839 } else if (diff > 0) { 840 if (data_len > 0xFFFF) 841 return -EINVAL; 842 if (diff > skb_tailroom(e->skb)) { 843 nskb = skb_copy_expand(e->skb, skb_headroom(e->skb), 844 diff, GFP_ATOMIC); 845 if (!nskb) 846 return -ENOMEM; 847 kfree_skb(e->skb); 848 e->skb = nskb; 849 } 850 skb_put(e->skb, diff); 851 } 852 if (skb_ensure_writable(e->skb, data_len)) 853 return -ENOMEM; 854 skb_copy_to_linear_data(e->skb, data, data_len); 855 e->skb->ip_summed = CHECKSUM_NONE; 856 return 0; 857 } 858 859 static int 860 nfqnl_set_mode(struct nfqnl_instance *queue, 861 unsigned char mode, unsigned int range) 862 { 863 int status = 0; 864 865 spin_lock_bh(&queue->lock); 866 switch (mode) { 867 case NFQNL_COPY_NONE: 868 case NFQNL_COPY_META: 869 queue->copy_mode = mode; 870 queue->copy_range = 0; 871 break; 872 873 case NFQNL_COPY_PACKET: 874 queue->copy_mode = mode; 875 if (range == 0 || range > NFQNL_MAX_COPY_RANGE) 876 queue->copy_range = NFQNL_MAX_COPY_RANGE; 877 else 878 queue->copy_range = range; 879 break; 880 881 default: 882 status = -EINVAL; 883 884 } 885 spin_unlock_bh(&queue->lock); 886 887 return status; 888 } 889 890 static int 891 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex) 892 { 893 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER) 894 int physinif, physoutif; 895 896 physinif = nf_bridge_get_physinif(entry->skb); 897 physoutif = nf_bridge_get_physoutif(entry->skb); 898 899 if (physinif == ifindex || physoutif == ifindex) 900 return 1; 901 #endif 902 if (entry->state.in) 903 if (entry->state.in->ifindex == ifindex) 904 return 1; 905 if (entry->state.out) 906 if (entry->state.out->ifindex == ifindex) 907 return 1; 908 909 return 0; 910 } 911 912 /* drop all packets with either indev or outdev == ifindex from all queue 913 * instances */ 914 static void 915 nfqnl_dev_drop(struct net *net, int ifindex) 916 { 917 int i; 918 struct nfnl_queue_net *q = nfnl_queue_pernet(net); 919 920 rcu_read_lock(); 921 922 for (i = 0; i < INSTANCE_BUCKETS; i++) { 923 struct nfqnl_instance *inst; 924 struct hlist_head *head = &q->instance_table[i]; 925 926 hlist_for_each_entry_rcu(inst, head, hlist) 927 nfqnl_flush(inst, dev_cmp, ifindex); 928 } 929 930 rcu_read_unlock(); 931 } 932 933 static int 934 nfqnl_rcv_dev_event(struct notifier_block *this, 935 unsigned long event, void *ptr) 936 { 937 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 938 939 /* Drop any packets associated with the downed device */ 940 if (event == NETDEV_DOWN) 941 nfqnl_dev_drop(dev_net(dev), dev->ifindex); 942 return NOTIFY_DONE; 943 } 944 945 static struct notifier_block nfqnl_dev_notifier = { 946 .notifier_call = nfqnl_rcv_dev_event, 947 }; 948 949 static void nfqnl_nf_hook_drop(struct net *net) 950 { 951 struct nfnl_queue_net *q = nfnl_queue_pernet(net); 952 int i; 953 954 for (i = 0; i < INSTANCE_BUCKETS; i++) { 955 struct nfqnl_instance *inst; 956 struct hlist_head *head = &q->instance_table[i]; 957 958 hlist_for_each_entry_rcu(inst, head, hlist) 959 nfqnl_flush(inst, NULL, 0); 960 } 961 } 962 963 static int 964 nfqnl_rcv_nl_event(struct notifier_block *this, 965 unsigned long event, void *ptr) 966 { 967 struct netlink_notify *n = ptr; 968 struct nfnl_queue_net *q = nfnl_queue_pernet(n->net); 969 970 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) { 971 int i; 972 973 /* destroy all instances for this portid */ 974 spin_lock(&q->instances_lock); 975 for (i = 0; i < INSTANCE_BUCKETS; i++) { 976 struct hlist_node *t2; 977 struct nfqnl_instance *inst; 978 struct hlist_head *head = &q->instance_table[i]; 979 980 hlist_for_each_entry_safe(inst, t2, head, hlist) { 981 if (n->portid == inst->peer_portid) 982 __instance_destroy(inst); 983 } 984 } 985 spin_unlock(&q->instances_lock); 986 } 987 return NOTIFY_DONE; 988 } 989 990 static struct notifier_block nfqnl_rtnl_notifier = { 991 .notifier_call = nfqnl_rcv_nl_event, 992 }; 993 994 static const struct nla_policy nfqa_vlan_policy[NFQA_VLAN_MAX + 1] = { 995 [NFQA_VLAN_TCI] = { .type = NLA_U16}, 996 [NFQA_VLAN_PROTO] = { .type = NLA_U16}, 997 }; 998 999 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = { 1000 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) }, 1001 [NFQA_MARK] = { .type = NLA_U32 }, 1002 [NFQA_PAYLOAD] = { .type = NLA_UNSPEC }, 1003 [NFQA_CT] = { .type = NLA_UNSPEC }, 1004 [NFQA_EXP] = { .type = NLA_UNSPEC }, 1005 [NFQA_VLAN] = { .type = NLA_NESTED }, 1006 }; 1007 1008 static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = { 1009 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) }, 1010 [NFQA_MARK] = { .type = NLA_U32 }, 1011 }; 1012 1013 static struct nfqnl_instance * 1014 verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, u32 nlportid) 1015 { 1016 struct nfqnl_instance *queue; 1017 1018 queue = instance_lookup(q, queue_num); 1019 if (!queue) 1020 return ERR_PTR(-ENODEV); 1021 1022 if (queue->peer_portid != nlportid) 1023 return ERR_PTR(-EPERM); 1024 1025 return queue; 1026 } 1027 1028 static struct nfqnl_msg_verdict_hdr* 1029 verdicthdr_get(const struct nlattr * const nfqa[]) 1030 { 1031 struct nfqnl_msg_verdict_hdr *vhdr; 1032 unsigned int verdict; 1033 1034 if (!nfqa[NFQA_VERDICT_HDR]) 1035 return NULL; 1036 1037 vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]); 1038 verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK; 1039 if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN) 1040 return NULL; 1041 return vhdr; 1042 } 1043 1044 static int nfq_id_after(unsigned int id, unsigned int max) 1045 { 1046 return (int)(id - max) > 0; 1047 } 1048 1049 static int nfqnl_recv_verdict_batch(struct sk_buff *skb, 1050 const struct nfnl_info *info, 1051 const struct nlattr * const nfqa[]) 1052 { 1053 struct nfnl_queue_net *q = nfnl_queue_pernet(info->net); 1054 struct nfgenmsg *nfmsg = nlmsg_data(info->nlh); 1055 u16 queue_num = ntohs(nfmsg->res_id); 1056 struct nf_queue_entry *entry, *tmp; 1057 struct nfqnl_msg_verdict_hdr *vhdr; 1058 struct nfqnl_instance *queue; 1059 unsigned int verdict, maxid; 1060 LIST_HEAD(batch_list); 1061 1062 queue = verdict_instance_lookup(q, queue_num, 1063 NETLINK_CB(skb).portid); 1064 if (IS_ERR(queue)) 1065 return PTR_ERR(queue); 1066 1067 vhdr = verdicthdr_get(nfqa); 1068 if (!vhdr) 1069 return -EINVAL; 1070 1071 verdict = ntohl(vhdr->verdict); 1072 maxid = ntohl(vhdr->id); 1073 1074 spin_lock_bh(&queue->lock); 1075 1076 list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) { 1077 if (nfq_id_after(entry->id, maxid)) 1078 break; 1079 __dequeue_entry(queue, entry); 1080 list_add_tail(&entry->list, &batch_list); 1081 } 1082 1083 spin_unlock_bh(&queue->lock); 1084 1085 if (list_empty(&batch_list)) 1086 return -ENOENT; 1087 1088 list_for_each_entry_safe(entry, tmp, &batch_list, list) { 1089 if (nfqa[NFQA_MARK]) 1090 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK])); 1091 1092 nfqnl_reinject(entry, verdict); 1093 } 1094 return 0; 1095 } 1096 1097 static struct nf_conn *nfqnl_ct_parse(struct nfnl_ct_hook *nfnl_ct, 1098 const struct nlmsghdr *nlh, 1099 const struct nlattr * const nfqa[], 1100 struct nf_queue_entry *entry, 1101 enum ip_conntrack_info *ctinfo) 1102 { 1103 #if IS_ENABLED(CONFIG_NF_CONNTRACK) 1104 struct nf_conn *ct; 1105 1106 ct = nf_ct_get(entry->skb, ctinfo); 1107 if (ct == NULL) 1108 return NULL; 1109 1110 if (nfnl_ct->parse(nfqa[NFQA_CT], ct) < 0) 1111 return NULL; 1112 1113 if (nfqa[NFQA_EXP]) 1114 nfnl_ct->attach_expect(nfqa[NFQA_EXP], ct, 1115 NETLINK_CB(entry->skb).portid, 1116 nlmsg_report(nlh)); 1117 return ct; 1118 #else 1119 return NULL; 1120 #endif 1121 } 1122 1123 static int nfqa_parse_bridge(struct nf_queue_entry *entry, 1124 const struct nlattr * const nfqa[]) 1125 { 1126 if (nfqa[NFQA_VLAN]) { 1127 struct nlattr *tb[NFQA_VLAN_MAX + 1]; 1128 int err; 1129 1130 err = nla_parse_nested_deprecated(tb, NFQA_VLAN_MAX, 1131 nfqa[NFQA_VLAN], 1132 nfqa_vlan_policy, NULL); 1133 if (err < 0) 1134 return err; 1135 1136 if (!tb[NFQA_VLAN_TCI] || !tb[NFQA_VLAN_PROTO]) 1137 return -EINVAL; 1138 1139 __vlan_hwaccel_put_tag(entry->skb, 1140 nla_get_be16(tb[NFQA_VLAN_PROTO]), 1141 ntohs(nla_get_be16(tb[NFQA_VLAN_TCI]))); 1142 } 1143 1144 if (nfqa[NFQA_L2HDR]) { 1145 int mac_header_len = entry->skb->network_header - 1146 entry->skb->mac_header; 1147 1148 if (mac_header_len != nla_len(nfqa[NFQA_L2HDR])) 1149 return -EINVAL; 1150 else if (mac_header_len > 0) 1151 memcpy(skb_mac_header(entry->skb), 1152 nla_data(nfqa[NFQA_L2HDR]), 1153 mac_header_len); 1154 } 1155 1156 return 0; 1157 } 1158 1159 static int nfqnl_recv_verdict(struct sk_buff *skb, const struct nfnl_info *info, 1160 const struct nlattr * const nfqa[]) 1161 { 1162 struct nfnl_queue_net *q = nfnl_queue_pernet(info->net); 1163 struct nfgenmsg *nfmsg = nlmsg_data(info->nlh); 1164 u_int16_t queue_num = ntohs(nfmsg->res_id); 1165 struct nfqnl_msg_verdict_hdr *vhdr; 1166 enum ip_conntrack_info ctinfo; 1167 struct nfqnl_instance *queue; 1168 struct nf_queue_entry *entry; 1169 struct nfnl_ct_hook *nfnl_ct; 1170 struct nf_conn *ct = NULL; 1171 unsigned int verdict; 1172 int err; 1173 1174 queue = verdict_instance_lookup(q, queue_num, 1175 NETLINK_CB(skb).portid); 1176 if (IS_ERR(queue)) 1177 return PTR_ERR(queue); 1178 1179 vhdr = verdicthdr_get(nfqa); 1180 if (!vhdr) 1181 return -EINVAL; 1182 1183 verdict = ntohl(vhdr->verdict); 1184 1185 entry = find_dequeue_entry(queue, ntohl(vhdr->id)); 1186 if (entry == NULL) 1187 return -ENOENT; 1188 1189 /* rcu lock already held from nfnl->call_rcu. */ 1190 nfnl_ct = rcu_dereference(nfnl_ct_hook); 1191 1192 if (nfqa[NFQA_CT]) { 1193 if (nfnl_ct != NULL) 1194 ct = nfqnl_ct_parse(nfnl_ct, info->nlh, nfqa, entry, 1195 &ctinfo); 1196 } 1197 1198 if (entry->state.pf == PF_BRIDGE) { 1199 err = nfqa_parse_bridge(entry, nfqa); 1200 if (err < 0) 1201 return err; 1202 } 1203 1204 if (nfqa[NFQA_PAYLOAD]) { 1205 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]); 1206 int diff = payload_len - entry->skb->len; 1207 1208 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]), 1209 payload_len, entry, diff) < 0) 1210 verdict = NF_DROP; 1211 1212 if (ct && diff) 1213 nfnl_ct->seq_adjust(entry->skb, ct, ctinfo, diff); 1214 } 1215 1216 if (nfqa[NFQA_MARK]) 1217 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK])); 1218 1219 nfqnl_reinject(entry, verdict); 1220 return 0; 1221 } 1222 1223 static int nfqnl_recv_unsupp(struct sk_buff *skb, const struct nfnl_info *info, 1224 const struct nlattr * const cda[]) 1225 { 1226 return -ENOTSUPP; 1227 } 1228 1229 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = { 1230 [NFQA_CFG_CMD] = { .len = sizeof(struct nfqnl_msg_config_cmd) }, 1231 [NFQA_CFG_PARAMS] = { .len = sizeof(struct nfqnl_msg_config_params) }, 1232 [NFQA_CFG_QUEUE_MAXLEN] = { .type = NLA_U32 }, 1233 [NFQA_CFG_MASK] = { .type = NLA_U32 }, 1234 [NFQA_CFG_FLAGS] = { .type = NLA_U32 }, 1235 }; 1236 1237 static const struct nf_queue_handler nfqh = { 1238 .outfn = nfqnl_enqueue_packet, 1239 .nf_hook_drop = nfqnl_nf_hook_drop, 1240 }; 1241 1242 static int nfqnl_recv_config(struct sk_buff *skb, const struct nfnl_info *info, 1243 const struct nlattr * const nfqa[]) 1244 { 1245 struct nfnl_queue_net *q = nfnl_queue_pernet(info->net); 1246 struct nfgenmsg *nfmsg = nlmsg_data(info->nlh); 1247 u_int16_t queue_num = ntohs(nfmsg->res_id); 1248 struct nfqnl_msg_config_cmd *cmd = NULL; 1249 struct nfqnl_instance *queue; 1250 __u32 flags = 0, mask = 0; 1251 int ret = 0; 1252 1253 if (nfqa[NFQA_CFG_CMD]) { 1254 cmd = nla_data(nfqa[NFQA_CFG_CMD]); 1255 1256 /* Obsolete commands without queue context */ 1257 switch (cmd->command) { 1258 case NFQNL_CFG_CMD_PF_BIND: return 0; 1259 case NFQNL_CFG_CMD_PF_UNBIND: return 0; 1260 } 1261 } 1262 1263 /* Check if we support these flags in first place, dependencies should 1264 * be there too not to break atomicity. 1265 */ 1266 if (nfqa[NFQA_CFG_FLAGS]) { 1267 if (!nfqa[NFQA_CFG_MASK]) { 1268 /* A mask is needed to specify which flags are being 1269 * changed. 1270 */ 1271 return -EINVAL; 1272 } 1273 1274 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS])); 1275 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK])); 1276 1277 if (flags >= NFQA_CFG_F_MAX) 1278 return -EOPNOTSUPP; 1279 1280 #if !IS_ENABLED(CONFIG_NETWORK_SECMARK) 1281 if (flags & mask & NFQA_CFG_F_SECCTX) 1282 return -EOPNOTSUPP; 1283 #endif 1284 if ((flags & mask & NFQA_CFG_F_CONNTRACK) && 1285 !rcu_access_pointer(nfnl_ct_hook)) { 1286 #ifdef CONFIG_MODULES 1287 nfnl_unlock(NFNL_SUBSYS_QUEUE); 1288 request_module("ip_conntrack_netlink"); 1289 nfnl_lock(NFNL_SUBSYS_QUEUE); 1290 if (rcu_access_pointer(nfnl_ct_hook)) 1291 return -EAGAIN; 1292 #endif 1293 return -EOPNOTSUPP; 1294 } 1295 } 1296 1297 rcu_read_lock(); 1298 queue = instance_lookup(q, queue_num); 1299 if (queue && queue->peer_portid != NETLINK_CB(skb).portid) { 1300 ret = -EPERM; 1301 goto err_out_unlock; 1302 } 1303 1304 if (cmd != NULL) { 1305 switch (cmd->command) { 1306 case NFQNL_CFG_CMD_BIND: 1307 if (queue) { 1308 ret = -EBUSY; 1309 goto err_out_unlock; 1310 } 1311 queue = instance_create(q, queue_num, 1312 NETLINK_CB(skb).portid); 1313 if (IS_ERR(queue)) { 1314 ret = PTR_ERR(queue); 1315 goto err_out_unlock; 1316 } 1317 break; 1318 case NFQNL_CFG_CMD_UNBIND: 1319 if (!queue) { 1320 ret = -ENODEV; 1321 goto err_out_unlock; 1322 } 1323 instance_destroy(q, queue); 1324 goto err_out_unlock; 1325 case NFQNL_CFG_CMD_PF_BIND: 1326 case NFQNL_CFG_CMD_PF_UNBIND: 1327 break; 1328 default: 1329 ret = -ENOTSUPP; 1330 goto err_out_unlock; 1331 } 1332 } 1333 1334 if (!queue) { 1335 ret = -ENODEV; 1336 goto err_out_unlock; 1337 } 1338 1339 if (nfqa[NFQA_CFG_PARAMS]) { 1340 struct nfqnl_msg_config_params *params = 1341 nla_data(nfqa[NFQA_CFG_PARAMS]); 1342 1343 nfqnl_set_mode(queue, params->copy_mode, 1344 ntohl(params->copy_range)); 1345 } 1346 1347 if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) { 1348 __be32 *queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]); 1349 1350 spin_lock_bh(&queue->lock); 1351 queue->queue_maxlen = ntohl(*queue_maxlen); 1352 spin_unlock_bh(&queue->lock); 1353 } 1354 1355 if (nfqa[NFQA_CFG_FLAGS]) { 1356 spin_lock_bh(&queue->lock); 1357 queue->flags &= ~mask; 1358 queue->flags |= flags & mask; 1359 spin_unlock_bh(&queue->lock); 1360 } 1361 1362 err_out_unlock: 1363 rcu_read_unlock(); 1364 return ret; 1365 } 1366 1367 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = { 1368 [NFQNL_MSG_PACKET] = { 1369 .call = nfqnl_recv_unsupp, 1370 .type = NFNL_CB_RCU, 1371 .attr_count = NFQA_MAX, 1372 }, 1373 [NFQNL_MSG_VERDICT] = { 1374 .call = nfqnl_recv_verdict, 1375 .type = NFNL_CB_RCU, 1376 .attr_count = NFQA_MAX, 1377 .policy = nfqa_verdict_policy 1378 }, 1379 [NFQNL_MSG_CONFIG] = { 1380 .call = nfqnl_recv_config, 1381 .type = NFNL_CB_MUTEX, 1382 .attr_count = NFQA_CFG_MAX, 1383 .policy = nfqa_cfg_policy 1384 }, 1385 [NFQNL_MSG_VERDICT_BATCH] = { 1386 .call = nfqnl_recv_verdict_batch, 1387 .type = NFNL_CB_RCU, 1388 .attr_count = NFQA_MAX, 1389 .policy = nfqa_verdict_batch_policy 1390 }, 1391 }; 1392 1393 static const struct nfnetlink_subsystem nfqnl_subsys = { 1394 .name = "nf_queue", 1395 .subsys_id = NFNL_SUBSYS_QUEUE, 1396 .cb_count = NFQNL_MSG_MAX, 1397 .cb = nfqnl_cb, 1398 }; 1399 1400 #ifdef CONFIG_PROC_FS 1401 struct iter_state { 1402 struct seq_net_private p; 1403 unsigned int bucket; 1404 }; 1405 1406 static struct hlist_node *get_first(struct seq_file *seq) 1407 { 1408 struct iter_state *st = seq->private; 1409 struct net *net; 1410 struct nfnl_queue_net *q; 1411 1412 if (!st) 1413 return NULL; 1414 1415 net = seq_file_net(seq); 1416 q = nfnl_queue_pernet(net); 1417 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) { 1418 if (!hlist_empty(&q->instance_table[st->bucket])) 1419 return q->instance_table[st->bucket].first; 1420 } 1421 return NULL; 1422 } 1423 1424 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h) 1425 { 1426 struct iter_state *st = seq->private; 1427 struct net *net = seq_file_net(seq); 1428 1429 h = h->next; 1430 while (!h) { 1431 struct nfnl_queue_net *q; 1432 1433 if (++st->bucket >= INSTANCE_BUCKETS) 1434 return NULL; 1435 1436 q = nfnl_queue_pernet(net); 1437 h = q->instance_table[st->bucket].first; 1438 } 1439 return h; 1440 } 1441 1442 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos) 1443 { 1444 struct hlist_node *head; 1445 head = get_first(seq); 1446 1447 if (head) 1448 while (pos && (head = get_next(seq, head))) 1449 pos--; 1450 return pos ? NULL : head; 1451 } 1452 1453 static void *seq_start(struct seq_file *s, loff_t *pos) 1454 __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock) 1455 { 1456 spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock); 1457 return get_idx(s, *pos); 1458 } 1459 1460 static void *seq_next(struct seq_file *s, void *v, loff_t *pos) 1461 { 1462 (*pos)++; 1463 return get_next(s, v); 1464 } 1465 1466 static void seq_stop(struct seq_file *s, void *v) 1467 __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock) 1468 { 1469 spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock); 1470 } 1471 1472 static int seq_show(struct seq_file *s, void *v) 1473 { 1474 const struct nfqnl_instance *inst = v; 1475 1476 seq_printf(s, "%5u %6u %5u %1u %5u %5u %5u %8u %2d\n", 1477 inst->queue_num, 1478 inst->peer_portid, inst->queue_total, 1479 inst->copy_mode, inst->copy_range, 1480 inst->queue_dropped, inst->queue_user_dropped, 1481 inst->id_sequence, 1); 1482 return 0; 1483 } 1484 1485 static const struct seq_operations nfqnl_seq_ops = { 1486 .start = seq_start, 1487 .next = seq_next, 1488 .stop = seq_stop, 1489 .show = seq_show, 1490 }; 1491 #endif /* PROC_FS */ 1492 1493 static int __net_init nfnl_queue_net_init(struct net *net) 1494 { 1495 unsigned int i; 1496 struct nfnl_queue_net *q = nfnl_queue_pernet(net); 1497 1498 for (i = 0; i < INSTANCE_BUCKETS; i++) 1499 INIT_HLIST_HEAD(&q->instance_table[i]); 1500 1501 spin_lock_init(&q->instances_lock); 1502 1503 #ifdef CONFIG_PROC_FS 1504 if (!proc_create_net("nfnetlink_queue", 0440, net->nf.proc_netfilter, 1505 &nfqnl_seq_ops, sizeof(struct iter_state))) 1506 return -ENOMEM; 1507 #endif 1508 nf_register_queue_handler(net, &nfqh); 1509 return 0; 1510 } 1511 1512 static void __net_exit nfnl_queue_net_exit(struct net *net) 1513 { 1514 struct nfnl_queue_net *q = nfnl_queue_pernet(net); 1515 unsigned int i; 1516 1517 nf_unregister_queue_handler(net); 1518 #ifdef CONFIG_PROC_FS 1519 remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter); 1520 #endif 1521 for (i = 0; i < INSTANCE_BUCKETS; i++) 1522 WARN_ON_ONCE(!hlist_empty(&q->instance_table[i])); 1523 } 1524 1525 static void nfnl_queue_net_exit_batch(struct list_head *net_exit_list) 1526 { 1527 synchronize_rcu(); 1528 } 1529 1530 static struct pernet_operations nfnl_queue_net_ops = { 1531 .init = nfnl_queue_net_init, 1532 .exit = nfnl_queue_net_exit, 1533 .exit_batch = nfnl_queue_net_exit_batch, 1534 .id = &nfnl_queue_net_id, 1535 .size = sizeof(struct nfnl_queue_net), 1536 }; 1537 1538 static int __init nfnetlink_queue_init(void) 1539 { 1540 int status; 1541 1542 status = register_pernet_subsys(&nfnl_queue_net_ops); 1543 if (status < 0) { 1544 pr_err("failed to register pernet ops\n"); 1545 goto out; 1546 } 1547 1548 netlink_register_notifier(&nfqnl_rtnl_notifier); 1549 status = nfnetlink_subsys_register(&nfqnl_subsys); 1550 if (status < 0) { 1551 pr_err("failed to create netlink socket\n"); 1552 goto cleanup_netlink_notifier; 1553 } 1554 1555 status = register_netdevice_notifier(&nfqnl_dev_notifier); 1556 if (status < 0) { 1557 pr_err("failed to register netdevice notifier\n"); 1558 goto cleanup_netlink_subsys; 1559 } 1560 1561 return status; 1562 1563 cleanup_netlink_subsys: 1564 nfnetlink_subsys_unregister(&nfqnl_subsys); 1565 cleanup_netlink_notifier: 1566 netlink_unregister_notifier(&nfqnl_rtnl_notifier); 1567 unregister_pernet_subsys(&nfnl_queue_net_ops); 1568 out: 1569 return status; 1570 } 1571 1572 static void __exit nfnetlink_queue_fini(void) 1573 { 1574 unregister_netdevice_notifier(&nfqnl_dev_notifier); 1575 nfnetlink_subsys_unregister(&nfqnl_subsys); 1576 netlink_unregister_notifier(&nfqnl_rtnl_notifier); 1577 unregister_pernet_subsys(&nfnl_queue_net_ops); 1578 1579 rcu_barrier(); /* Wait for completion of call_rcu()'s */ 1580 } 1581 1582 MODULE_DESCRIPTION("netfilter packet queue handler"); 1583 MODULE_AUTHOR("Harald Welte <[email protected]>"); 1584 MODULE_LICENSE("GPL"); 1585 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE); 1586 1587 module_init(nfnetlink_queue_init); 1588 module_exit(nfnetlink_queue_fini); 1589