net/netfilter/nfnetlink_queue_core.c

   1 /*
   2  * This is a module which is used for queueing packets and communicating with
   3  * userspace via nfnetlink.
   4  *
   5  * (C) 2005 by Harald Welte <laforge@netfilter.org>
   6  * (C) 2007 by Patrick McHardy <kaber@trash.net>
   7  *
   8  * Based on the old ipv4-only ip_queue.c:
   9  * (C) 2000-2002 James Morris <jmorris@intercode.com.au>
  10  * (C) 2003-2005 Netfilter Core Team <coreteam@netfilter.org>
  11  *
  12  * This program is free software; you can redistribute it and/or modify
  13  * it under the terms of the GNU General Public License version 2 as
  14  * published by the Free Software Foundation.
  15  *
  16  */
  17 #include <linux/module.h>
  18 #include <linux/skbuff.h>
  19 #include <linux/init.h>
  20 #include <linux/spinlock.h>
  21 #include <linux/slab.h>
  22 #include <linux/notifier.h>
  23 #include <linux/netdevice.h>
  24 #include <linux/netfilter.h>
  25 #include <linux/proc_fs.h>
  26 #include <linux/netfilter_ipv4.h>
  27 #include <linux/netfilter_ipv6.h>
  28 #include <linux/netfilter/nfnetlink.h>
  29 #include <linux/netfilter/nfnetlink_queue.h>
  30 #include <linux/list.h>
  31 #include <net/sock.h>
  32 #include <net/netfilter/nf_queue.h>
  33 #include <net/netns/generic.h>
  34 #include <net/netfilter/nfnetlink_queue.h>
  35
  36 #include <linux/atomic.h>
  37
  38 #ifdef CONFIG_BRIDGE_NETFILTER
  39 #include "../bridge/br_private.h"
  40 #endif
  41
  42 #define NFQNL_QMAX_DEFAULT 1024
  43
  44 /* We're using struct nlattr which has 16bit nla_len. Note that nla_len
  45  * includes the header length. Thus, the maximum packet length that we
  46  * support is 65531 bytes. We send truncated packets if the specified length
  47  * is larger than that.  Userspace can check for presence of NFQA_CAP_LEN
  48  * attribute to detect truncation.
  49  */
  50 #define NFQNL_MAX_COPY_RANGE (0xffff - NLA_HDRLEN)
  51
  52 struct nfqnl_instance {
  53         struct hlist_node hlist;                /* global list of queues */
  54         struct rcu_head rcu;
  55
  56         int peer_portid;
  57         unsigned int queue_maxlen;
  58         unsigned int copy_range;
  59         unsigned int queue_dropped;
  60         unsigned int queue_user_dropped;
  61
  62
  63         u_int16_t queue_num;                    /* number of this queue */
  64         u_int8_t copy_mode;
  65         u_int32_t flags;                        /* Set using NFQA_CFG_FLAGS */
  66 /*
  67  * Following fields are dirtied for each queued packet,
  68  * keep them in same cache line if possible.
  69  */
  70         spinlock_t      lock;
  71         unsigned int    queue_total;
  72         unsigned int    id_sequence;            /* 'sequence' of pkt ids */
  73         struct list_head queue_list;            /* packets in queue */
  74 };
  75
  76 typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
  77
  78 static int nfnl_queue_net_id __read_mostly;
  79
  80 #define INSTANCE_BUCKETS        16
  81 struct nfnl_queue_net {
  82         spinlock_t instances_lock;
  83         struct hlist_head instance_table[INSTANCE_BUCKETS];
  84 };
  85
  86 static struct nfnl_queue_net *nfnl_queue_pernet(struct net *net)
  87 {
  88         return net_generic(net, nfnl_queue_net_id);
  89 }
  90
  91 static inline u_int8_t instance_hashfn(u_int16_t queue_num)
  92 {
  93         return ((queue_num >> 8) ^ queue_num) % INSTANCE_BUCKETS;
  94 }
  95
  96 static struct nfqnl_instance *
  97 instance_lookup(struct nfnl_queue_net *q, u_int16_t queue_num)
  98 {
  99         struct hlist_head *head;
 100         struct nfqnl_instance *inst;
 101
 102         head = &q->instance_table[instance_hashfn(queue_num)];
 103         hlist_for_each_entry_rcu(inst, head, hlist) {
 104                 if (inst->queue_num == queue_num)
 105                         return inst;
 106         }
 107         return NULL;
 108 }
 109
 110 static struct nfqnl_instance *
 111 instance_create(struct nfnl_queue_net *q, u_int16_t queue_num,
 112                 int portid)
 113 {
 114         struct nfqnl_instance *inst;
 115         unsigned int h;
 116         int err;
 117
 118         spin_lock(&q->instances_lock);
 119         if (instance_lookup(q, queue_num)) {
 120                 err = -EEXIST;
 121                 goto out_unlock;
 122         }
 123
 124         inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
 125         if (!inst) {
 126                 err = -ENOMEM;
 127                 goto out_unlock;
 128         }
 129
 130         inst->queue_num = queue_num;
 131         inst->peer_portid = portid;
 132         inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
 133         inst->copy_range = NFQNL_MAX_COPY_RANGE;
 134         inst->copy_mode = NFQNL_COPY_NONE;
 135         spin_lock_init(&inst->lock);
 136         INIT_LIST_HEAD(&inst->queue_list);
 137
 138         if (!try_module_get(THIS_MODULE)) {
 139                 err = -EAGAIN;
 140                 goto out_free;
 141         }
 142
 143         h = instance_hashfn(queue_num);
 144         hlist_add_head_rcu(&inst->hlist, &q->instance_table[h]);
 145
 146         spin_unlock(&q->instances_lock);
 147
 148         return inst;
 149
 150 out_free:
 151         kfree(inst);
 152 out_unlock:
 153         spin_unlock(&q->instances_lock);
 154         return ERR_PTR(err);
 155 }
 156
 157 static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
 158                         unsigned long data);
 159
 160 static void
 161 instance_destroy_rcu(struct rcu_head *head)
 162 {
 163         struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
 164                                                    rcu);
 165
 166         nfqnl_flush(inst, NULL, 0);
 167         kfree(inst);
 168         module_put(THIS_MODULE);
 169 }
 170
 171 static void
 172 __instance_destroy(struct nfqnl_instance *inst)
 173 {
 174         hlist_del_rcu(&inst->hlist);
 175         call_rcu(&inst->rcu, instance_destroy_rcu);
 176 }
 177
 178 static void
 179 instance_destroy(struct nfnl_queue_net *q, struct nfqnl_instance *inst)
 180 {
 181         spin_lock(&q->instances_lock);
 182         __instance_destroy(inst);
 183         spin_unlock(&q->instances_lock);
 184 }
 185
 186 static inline void
 187 __enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
 188 {
 189        list_add_tail(&entry->list, &queue->queue_list);
 190        queue->queue_total++;
 191 }
 192
 193 static void
 194 __dequeue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
 195 {
 196         list_del(&entry->list);
 197         queue->queue_total--;
 198 }
 199
 200 static struct nf_queue_entry *
 201 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
 202 {
 203         struct nf_queue_entry *entry = NULL, *i;
 204
 205         spin_lock_bh(&queue->lock);
 206
 207         list_for_each_entry(i, &queue->queue_list, list) {
 208                 if (i->id == id) {
 209                         entry = i;
 210                         break;
 211                 }
 212         }
 213
 214         if (entry)
 215                 __dequeue_entry(queue, entry);
 216
 217         spin_unlock_bh(&queue->lock);
 218
 219         return entry;
 220 }
 221
 222 static void
 223 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
 224 {
 225         struct nf_queue_entry *entry, *next;
 226
 227         spin_lock_bh(&queue->lock);
 228         list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
 229                 if (!cmpfn || cmpfn(entry, data)) {
 230                         list_del(&entry->list);
 231                         queue->queue_total--;
 232                         nf_reinject(entry, NF_DROP);
 233                 }
 234         }
 235         spin_unlock_bh(&queue->lock);
 236 }
 237
 238 static void
 239 nfqnl_zcopy(struct sk_buff *to, const struct sk_buff *from, int len, int hlen)
 240 {
 241         int i, j = 0;
 242         int plen = 0; /* length of skb->head fragment */
 243         struct page *page;
 244         unsigned int offset;
 245
 246         /* dont bother with small payloads */
 247         if (len <= skb_tailroom(to)) {
 248                 skb_copy_bits(from, 0, skb_put(to, len), len);
 249                 return;
 250         }
 251
 252         if (hlen) {
 253                 skb_copy_bits(from, 0, skb_put(to, hlen), hlen);
 254                 len -= hlen;
 255         } else {
 256                 plen = min_t(int, skb_headlen(from), len);
 257                 if (plen) {
 258                         page = virt_to_head_page(from->head);
 259                         offset = from->data - (unsigned char *)page_address(page);
 260                         __skb_fill_page_desc(to, 0, page, offset, plen);
 261                         get_page(page);
 262                         j = 1;
 263                         len -= plen;
 264                 }
 265         }
 266
 267         to->truesize += len + plen;
 268         to->len += len + plen;
 269         to->data_len += len + plen;
 270
 271         for (i = 0; i < skb_shinfo(from)->nr_frags; i++) {
 272                 if (!len)
 273                         break;
 274                 skb_shinfo(to)->frags[j] = skb_shinfo(from)->frags[i];
 275                 skb_shinfo(to)->frags[j].size = min_t(int, skb_shinfo(to)->frags[j].size, len);
 276                 len -= skb_shinfo(to)->frags[j].size;
 277                 skb_frag_ref(to, j);
 278                 j++;
 279         }
 280         skb_shinfo(to)->nr_frags = j;
 281 }
 282
 283 static int
 284 nfqnl_put_packet_info(struct sk_buff *nlskb, struct sk_buff *packet,
 285                       bool csum_verify)
 286 {
 287         __u32 flags = 0;
 288
 289         if (packet->ip_summed == CHECKSUM_PARTIAL)
 290                 flags = NFQA_SKB_CSUMNOTREADY;
 291         else if (csum_verify)
 292                 flags = NFQA_SKB_CSUM_NOTVERIFIED;
 293
 294         if (skb_is_gso(packet))
 295                 flags |= NFQA_SKB_GSO;
 296
 297         return flags ? nla_put_be32(nlskb, NFQA_SKB_INFO, htonl(flags)) : 0;
 298 }
 299
 300 static struct sk_buff *
 301 nfqnl_build_packet_message(struct nfqnl_instance *queue,
 302                            struct nf_queue_entry *entry,
 303                            __be32 **packet_id_ptr)
 304 {
 305         size_t size;
 306         size_t data_len = 0, cap_len = 0;
 307         int hlen = 0;
 308         struct sk_buff *skb;
 309         struct nlattr *nla;
 310         struct nfqnl_msg_packet_hdr *pmsg;
 311         struct nlmsghdr *nlh;
 312         struct nfgenmsg *nfmsg;
 313         struct sk_buff *entskb = entry->skb;
 314         struct net_device *indev;
 315         struct net_device *outdev;
 316         struct nf_conn *ct = NULL;
 317         enum ip_conntrack_info uninitialized_var(ctinfo);
 318         bool csum_verify;
 319
 320         size =    nlmsg_total_size(sizeof(struct nfgenmsg))
 321                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
 322                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 323                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 324 #ifdef CONFIG_BRIDGE_NETFILTER
 325                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 326                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 327 #endif
 328                 + nla_total_size(sizeof(u_int32_t))     /* mark */
 329                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
 330                 + nla_total_size(sizeof(u_int32_t))     /* skbinfo */
 331                 + nla_total_size(sizeof(u_int32_t));    /* cap_len */
 332
 333         if (entskb->tstamp.tv64)
 334                 size += nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
 335
 336         if (entry->hook <= NF_INET_FORWARD ||
 337            (entry->hook == NF_INET_POST_ROUTING && entskb->sk == NULL))
 338                 csum_verify = !skb_csum_unnecessary(entskb);
 339         else
 340                 csum_verify = false;
 341
 342         outdev = entry->outdev;
 343
 344         switch ((enum nfqnl_config_mode)ACCESS_ONCE(queue->copy_mode)) {
 345         case NFQNL_COPY_META:
 346         case NFQNL_COPY_NONE:
 347                 break;
 348
 349         case NFQNL_COPY_PACKET:
 350                 if (!(queue->flags & NFQA_CFG_F_GSO) &&
 351                     entskb->ip_summed == CHECKSUM_PARTIAL &&
 352                     skb_checksum_help(entskb))
 353                         return NULL;
 354
 355                 data_len = ACCESS_ONCE(queue->copy_range);
 356                 if (data_len > entskb->len)
 357                         data_len = entskb->len;
 358
 359                 if (!entskb->head_frag ||
 360                     skb_headlen(entskb) < L1_CACHE_BYTES ||
 361                     skb_shinfo(entskb)->nr_frags >= MAX_SKB_FRAGS)
 362                         hlen = skb_headlen(entskb);
 363
 364                 if (skb_has_frag_list(entskb))
 365                         hlen = entskb->len;
 366                 hlen = min_t(int, data_len, hlen);
 367                 size += sizeof(struct nlattr) + hlen;
 368                 cap_len = entskb->len;
 369                 break;
 370         }
 371
 372         if (queue->flags & NFQA_CFG_F_CONNTRACK)
 373                 ct = nfqnl_ct_get(entskb, &size, &ctinfo);
 374
 375         skb = nfnetlink_alloc_skb(&init_net, size, queue->peer_portid,
 376                                   GFP_ATOMIC);
 377         if (!skb)
 378                 return NULL;
 379
 380         nlh = nlmsg_put(skb, 0, 0,
 381                         NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
 382                         sizeof(struct nfgenmsg), 0);
 383         if (!nlh) {
 384                 kfree_skb(skb);
 385                 return NULL;
 386         }
 387         nfmsg = nlmsg_data(nlh);
 388         nfmsg->nfgen_family = entry->pf;
 389         nfmsg->version = NFNETLINK_V0;
 390         nfmsg->res_id = htons(queue->queue_num);
 391
 392         nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
 393         pmsg = nla_data(nla);
 394         pmsg->hw_protocol       = entskb->protocol;
 395         pmsg->hook              = entry->hook;
 396         *packet_id_ptr          = &pmsg->packet_id;
 397
 398         indev = entry->indev;
 399         if (indev) {
 400 #ifndef CONFIG_BRIDGE_NETFILTER
 401                 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
 402                         goto nla_put_failure;
 403 #else
 404                 if (entry->pf == PF_BRIDGE) {
 405                         /* Case 1: indev is physical input device, we need to
 406                          * look for bridge group (when called from
 407                          * netfilter_bridge) */
 408                         if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
 409                                          htonl(indev->ifindex)) ||
 410                         /* this is the bridge group "brX" */
 411                         /* rcu_read_lock()ed by __nf_queue */
 412                             nla_put_be32(skb, NFQA_IFINDEX_INDEV,
 413                                          htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
 414                                 goto nla_put_failure;
 415                 } else {
 416                         /* Case 2: indev is bridge group, we need to look for
 417                          * physical device (when called from ipv4) */
 418                         if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
 419                                          htonl(indev->ifindex)))
 420                                 goto nla_put_failure;
 421                         if (entskb->nf_bridge && entskb->nf_bridge->physindev &&
 422                             nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
 423                                          htonl(entskb->nf_bridge->physindev->ifindex)))
 424                                 goto nla_put_failure;
 425                 }
 426 #endif
 427         }
 428
 429         if (outdev) {
 430 #ifndef CONFIG_BRIDGE_NETFILTER
 431                 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
 432                         goto nla_put_failure;
 433 #else
 434                 if (entry->pf == PF_BRIDGE) {
 435                         /* Case 1: outdev is physical output device, we need to
 436                          * look for bridge group (when called from
 437                          * netfilter_bridge) */
 438                         if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 439                                          htonl(outdev->ifindex)) ||
 440                         /* this is the bridge group "brX" */
 441                         /* rcu_read_lock()ed by __nf_queue */
 442                             nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
 443                                          htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
 444                                 goto nla_put_failure;
 445                 } else {
 446                         /* Case 2: outdev is bridge group, we need to look for
 447                          * physical output device (when called from ipv4) */
 448                         if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
 449                                          htonl(outdev->ifindex)))
 450                                 goto nla_put_failure;
 451                         if (entskb->nf_bridge && entskb->nf_bridge->physoutdev &&
 452                             nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 453                                          htonl(entskb->nf_bridge->physoutdev->ifindex)))
 454                                 goto nla_put_failure;
 455                 }
 456 #endif
 457         }
 458
 459         if (entskb->mark &&
 460             nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
 461                 goto nla_put_failure;
 462
 463         if (indev && entskb->dev &&
 464             entskb->mac_header != entskb->network_header) {
 465                 struct nfqnl_msg_packet_hw phw;
 466                 int len;
 467
 468                 memset(&phw, 0, sizeof(phw));
 469                 len = dev_parse_header(entskb, phw.hw_addr);
 470                 if (len) {
 471                         phw.hw_addrlen = htons(len);
 472                         if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
 473                                 goto nla_put_failure;
 474                 }
 475         }
 476
 477         if (entskb->tstamp.tv64) {
 478                 struct nfqnl_msg_packet_timestamp ts;
 479                 struct timeval tv = ktime_to_timeval(entskb->tstamp);
 480                 ts.sec = cpu_to_be64(tv.tv_sec);
 481                 ts.usec = cpu_to_be64(tv.tv_usec);
 482
 483                 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
 484                         goto nla_put_failure;
 485         }
 486
 487         if (ct && nfqnl_ct_put(skb, ct, ctinfo) < 0)
 488                 goto nla_put_failure;
 489
 490         if (cap_len > data_len &&
 491             nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
 492                 goto nla_put_failure;
 493
 494         if (nfqnl_put_packet_info(skb, entskb, csum_verify))
 495                 goto nla_put_failure;
 496
 497         if (data_len) {
 498                 struct nlattr *nla;
 499
 500                 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
 501                         goto nla_put_failure;
 502
 503                 nla = (struct nlattr *)skb_put(skb, sizeof(*nla));
 504                 nla->nla_type = NFQA_PAYLOAD;
 505                 nla->nla_len = nla_attr_size(data_len);
 506
 507                 nfqnl_zcopy(skb, entskb, data_len, hlen);
 508         }
 509
 510         nlh->nlmsg_len = skb->len;
 511         return skb;
 512
 513 nla_put_failure:
 514         kfree_skb(skb);
 515         net_err_ratelimited("nf_queue: error creating packet message\n");
 516         return NULL;
 517 }
 518
 519 static int
 520 __nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
 521                         struct nf_queue_entry *entry)
 522 {
 523         struct sk_buff *nskb;
 524         int err = -ENOBUFS;
 525         __be32 *packet_id_ptr;
 526         int failopen = 0;
 527
 528         nskb = nfqnl_build_packet_message(queue, entry, &packet_id_ptr);
 529         if (nskb == NULL) {
 530                 err = -ENOMEM;
 531                 goto err_out;
 532         }
 533         spin_lock_bh(&queue->lock);
 534
 535         if (queue->queue_total >= queue->queue_maxlen) {
 536                 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
 537                         failopen = 1;
 538                         err = 0;
 539                 } else {
 540                         queue->queue_dropped++;
 541                         net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
 542                                              queue->queue_total);
 543                 }
 544                 goto err_out_free_nskb;
 545         }
 546         entry->id = ++queue->id_sequence;
 547         *packet_id_ptr = htonl(entry->id);
 548
 549         /* nfnetlink_unicast will either free the nskb or add it to a socket */
 550         err = nfnetlink_unicast(nskb, net, queue->peer_portid, MSG_DONTWAIT);
 551         if (err < 0) {
 552                 queue->queue_user_dropped++;
 553                 goto err_out_unlock;
 554         }
 555
 556         __enqueue_entry(queue, entry);
 557
 558         spin_unlock_bh(&queue->lock);
 559         return 0;
 560
 561 err_out_free_nskb:
 562         kfree_skb(nskb);
 563 err_out_unlock:
 564         spin_unlock_bh(&queue->lock);
 565         if (failopen)
 566                 nf_reinject(entry, NF_ACCEPT);
 567 err_out:
 568         return err;
 569 }
 570
 571 static struct nf_queue_entry *
 572 nf_queue_entry_dup(struct nf_queue_entry *e)
 573 {
 574         struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
 575         if (entry) {
 576                 if (nf_queue_entry_get_refs(entry))
 577                         return entry;
 578                 kfree(entry);
 579         }
 580         return NULL;
 581 }
 582
 583 #ifdef CONFIG_BRIDGE_NETFILTER
 584 /* When called from bridge netfilter, skb->data must point to MAC header
 585  * before calling skb_gso_segment(). Else, original MAC header is lost
 586  * and segmented skbs will be sent to wrong destination.
 587  */
 588 static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
 589 {
 590         if (skb->nf_bridge)
 591                 __skb_push(skb, skb->network_header - skb->mac_header);
 592 }
 593
 594 static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
 595 {
 596         if (skb->nf_bridge)
 597                 __skb_pull(skb, skb->network_header - skb->mac_header);
 598 }
 599 #else
 600 #define nf_bridge_adjust_skb_data(s) do {} while (0)
 601 #define nf_bridge_adjust_segmented_data(s) do {} while (0)
 602 #endif
 603
 604 static void free_entry(struct nf_queue_entry *entry)
 605 {
 606         nf_queue_entry_release_refs(entry);
 607         kfree(entry);
 608 }
 609
 610 static int
 611 __nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
 612                            struct sk_buff *skb, struct nf_queue_entry *entry)
 613 {
 614         int ret = -ENOMEM;
 615         struct nf_queue_entry *entry_seg;
 616
 617         nf_bridge_adjust_segmented_data(skb);
 618
 619         if (skb->next == NULL) { /* last packet, no need to copy entry */
 620                 struct sk_buff *gso_skb = entry->skb;
 621                 entry->skb = skb;
 622                 ret = __nfqnl_enqueue_packet(net, queue, entry);
 623                 if (ret)
 624                         entry->skb = gso_skb;
 625                 return ret;
 626         }
 627
 628         skb->next = NULL;
 629
 630         entry_seg = nf_queue_entry_dup(entry);
 631         if (entry_seg) {
 632                 entry_seg->skb = skb;
 633                 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
 634                 if (ret)
 635                         free_entry(entry_seg);
 636         }
 637         return ret;
 638 }
 639
 640 static int
 641 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 642 {
 643         unsigned int queued;
 644         struct nfqnl_instance *queue;
 645         struct sk_buff *skb, *segs;
 646         int err = -ENOBUFS;
 647         struct net *net = dev_net(entry->indev ?
 648                                   entry->indev : entry->outdev);
 649         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
 650
 651         /* rcu_read_lock()ed by nf_hook_slow() */
 652         queue = instance_lookup(q, queuenum);
 653         if (!queue)
 654                 return -ESRCH;
 655
 656         if (queue->copy_mode == NFQNL_COPY_NONE)
 657                 return -EINVAL;
 658
 659         skb = entry->skb;
 660
 661         switch (entry->pf) {
 662         case NFPROTO_IPV4:
 663                 skb->protocol = htons(ETH_P_IP);
 664                 break;
 665         case NFPROTO_IPV6:
 666                 skb->protocol = htons(ETH_P_IPV6);
 667                 break;
 668         }
 669
 670         if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(skb))
 671                 return __nfqnl_enqueue_packet(net, queue, entry);
 672
 673         nf_bridge_adjust_skb_data(skb);
 674         segs = skb_gso_segment(skb, 0);
 675         /* Does not use PTR_ERR to limit the number of error codes that can be
 676          * returned by nf_queue.  For instance, callers rely on -ECANCELED to
 677          * mean 'ignore this hook'.
 678          */
 679         if (IS_ERR(segs))
 680                 goto out_err;
 681         queued = 0;
 682         err = 0;
 683         do {
 684                 struct sk_buff *nskb = segs->next;
 685                 if (err == 0)
 686                         err = __nfqnl_enqueue_packet_gso(net, queue,
 687                                                         segs, entry);
 688                 if (err == 0)
 689                         queued++;
 690                 else
 691                         kfree_skb(segs);
 692                 segs = nskb;
 693         } while (segs);
 694
 695         if (queued) {
 696                 if (err) /* some segments are already queued */
 697                         free_entry(entry);
 698                 kfree_skb(skb);
 699                 return 0;
 700         }
 701  out_err:
 702         nf_bridge_adjust_segmented_data(skb);
 703         return err;
 704 }
 705
 706 static int
 707 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e, int diff)
 708 {
 709         struct sk_buff *nskb;
 710
 711         if (diff < 0) {
 712                 if (pskb_trim(e->skb, data_len))
 713                         return -ENOMEM;
 714         } else if (diff > 0) {
 715                 if (data_len > 0xFFFF)
 716                         return -EINVAL;
 717                 if (diff > skb_tailroom(e->skb)) {
 718                         nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
 719                                                diff, GFP_ATOMIC);
 720                         if (!nskb) {
 721                                 printk(KERN_WARNING "nf_queue: OOM "
 722                                       "in mangle, dropping packet\n");
 723                                 return -ENOMEM;
 724                         }
 725                         kfree_skb(e->skb);
 726                         e->skb = nskb;
 727                 }
 728                 skb_put(e->skb, diff);
 729         }
 730         if (!skb_make_writable(e->skb, data_len))
 731                 return -ENOMEM;
 732         skb_copy_to_linear_data(e->skb, data, data_len);
 733         e->skb->ip_summed = CHECKSUM_NONE;
 734         return 0;
 735 }
 736
 737 static int
 738 nfqnl_set_mode(struct nfqnl_instance *queue,
 739                unsigned char mode, unsigned int range)
 740 {
 741         int status = 0;
 742
 743         spin_lock_bh(&queue->lock);
 744         switch (mode) {
 745         case NFQNL_COPY_NONE:
 746         case NFQNL_COPY_META:
 747                 queue->copy_mode = mode;
 748                 queue->copy_range = 0;
 749                 break;
 750
 751         case NFQNL_COPY_PACKET:
 752                 queue->copy_mode = mode;
 753                 if (range == 0 || range > NFQNL_MAX_COPY_RANGE)
 754                         queue->copy_range = NFQNL_MAX_COPY_RANGE;
 755                 else
 756                         queue->copy_range = range;
 757                 break;
 758
 759         default:
 760                 status = -EINVAL;
 761
 762         }
 763         spin_unlock_bh(&queue->lock);
 764
 765         return status;
 766 }
 767
 768 static int
 769 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
 770 {
 771         if (entry->indev)
 772                 if (entry->indev->ifindex == ifindex)
 773                         return 1;
 774         if (entry->outdev)
 775                 if (entry->outdev->ifindex == ifindex)
 776                         return 1;
 777 #ifdef CONFIG_BRIDGE_NETFILTER
 778         if (entry->skb->nf_bridge) {
 779                 if (entry->skb->nf_bridge->physindev &&
 780                     entry->skb->nf_bridge->physindev->ifindex == ifindex)
 781                         return 1;
 782                 if (entry->skb->nf_bridge->physoutdev &&
 783                     entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
 784                         return 1;
 785         }
 786 #endif
 787         return 0;
 788 }
 789
 790 /* drop all packets with either indev or outdev == ifindex from all queue
 791  * instances */
 792 static void
 793 nfqnl_dev_drop(struct net *net, int ifindex)
 794 {
 795         int i;
 796         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
 797
 798         rcu_read_lock();
 799
 800         for (i = 0; i < INSTANCE_BUCKETS; i++) {
 801                 struct nfqnl_instance *inst;
 802                 struct hlist_head *head = &q->instance_table[i];
 803
 804                 hlist_for_each_entry_rcu(inst, head, hlist)
 805                         nfqnl_flush(inst, dev_cmp, ifindex);
 806         }
 807
 808         rcu_read_unlock();
 809 }
 810
 811 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
 812
 813 static int
 814 nfqnl_rcv_dev_event(struct notifier_block *this,
 815                     unsigned long event, void *ptr)
 816 {
 817         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
 818
 819         /* Drop any packets associated with the downed device */
 820         if (event == NETDEV_DOWN)
 821                 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
 822         return NOTIFY_DONE;
 823 }
 824
 825 static struct notifier_block nfqnl_dev_notifier = {
 826         .notifier_call  = nfqnl_rcv_dev_event,
 827 };
 828
 829 static int
 830 nfqnl_rcv_nl_event(struct notifier_block *this,
 831                    unsigned long event, void *ptr)
 832 {
 833         struct netlink_notify *n = ptr;
 834         struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
 835
 836         if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
 837                 int i;
 838
 839                 /* destroy all instances for this portid */
 840                 spin_lock(&q->instances_lock);
 841                 for (i = 0; i < INSTANCE_BUCKETS; i++) {
 842                         struct hlist_node *t2;
 843                         struct nfqnl_instance *inst;
 844                         struct hlist_head *head = &q->instance_table[i];
 845
 846                         hlist_for_each_entry_safe(inst, t2, head, hlist) {
 847                                 if (n->portid == inst->peer_portid)
 848                                         __instance_destroy(inst);
 849                         }
 850                 }
 851                 spin_unlock(&q->instances_lock);
 852         }
 853         return NOTIFY_DONE;
 854 }
 855
 856 static struct notifier_block nfqnl_rtnl_notifier = {
 857         .notifier_call  = nfqnl_rcv_nl_event,
 858 };
 859
 860 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
 861         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 862         [NFQA_MARK]             = { .type = NLA_U32 },
 863         [NFQA_PAYLOAD]          = { .type = NLA_UNSPEC },
 864         [NFQA_CT]               = { .type = NLA_UNSPEC },
 865 };
 866
 867 static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
 868         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 869         [NFQA_MARK]             = { .type = NLA_U32 },
 870 };
 871
 872 static struct nfqnl_instance *
 873 verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, int nlportid)
 874 {
 875         struct nfqnl_instance *queue;
 876
 877         queue = instance_lookup(q, queue_num);
 878         if (!queue)
 879                 return ERR_PTR(-ENODEV);
 880
 881         if (queue->peer_portid != nlportid)
 882                 return ERR_PTR(-EPERM);
 883
 884         return queue;
 885 }
 886
 887 static struct nfqnl_msg_verdict_hdr*
 888 verdicthdr_get(const struct nlattr * const nfqa[])
 889 {
 890         struct nfqnl_msg_verdict_hdr *vhdr;
 891         unsigned int verdict;
 892
 893         if (!nfqa[NFQA_VERDICT_HDR])
 894                 return NULL;
 895
 896         vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
 897         verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
 898         if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
 899                 return NULL;
 900         return vhdr;
 901 }
 902
 903 static int nfq_id_after(unsigned int id, unsigned int max)
 904 {
 905         return (int)(id - max) > 0;
 906 }
 907
 908 static int
 909 nfqnl_recv_verdict_batch(struct sock *ctnl, struct sk_buff *skb,
 910                    const struct nlmsghdr *nlh,
 911                    const struct nlattr * const nfqa[])
 912 {
 913         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
 914         struct nf_queue_entry *entry, *tmp;
 915         unsigned int verdict, maxid;
 916         struct nfqnl_msg_verdict_hdr *vhdr;
 917         struct nfqnl_instance *queue;
 918         LIST_HEAD(batch_list);
 919         u16 queue_num = ntohs(nfmsg->res_id);
 920
 921         struct net *net = sock_net(ctnl);
 922         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
 923
 924         queue = verdict_instance_lookup(q, queue_num,
 925                                         NETLINK_CB(skb).portid);
 926         if (IS_ERR(queue))
 927                 return PTR_ERR(queue);
 928
 929         vhdr = verdicthdr_get(nfqa);
 930         if (!vhdr)
 931                 return -EINVAL;
 932
 933         verdict = ntohl(vhdr->verdict);
 934         maxid = ntohl(vhdr->id);
 935
 936         spin_lock_bh(&queue->lock);
 937
 938         list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
 939                 if (nfq_id_after(entry->id, maxid))
 940                         break;
 941                 __dequeue_entry(queue, entry);
 942                 list_add_tail(&entry->list, &batch_list);
 943         }
 944
 945         spin_unlock_bh(&queue->lock);
 946
 947         if (list_empty(&batch_list))
 948                 return -ENOENT;
 949
 950         list_for_each_entry_safe(entry, tmp, &batch_list, list) {
 951                 if (nfqa[NFQA_MARK])
 952                         entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
 953                 nf_reinject(entry, verdict);
 954         }
 955         return 0;
 956 }
 957
 958 static int
 959 nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
 960                    const struct nlmsghdr *nlh,
 961                    const struct nlattr * const nfqa[])
 962 {
 963         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
 964         u_int16_t queue_num = ntohs(nfmsg->res_id);
 965
 966         struct nfqnl_msg_verdict_hdr *vhdr;
 967         struct nfqnl_instance *queue;
 968         unsigned int verdict;
 969         struct nf_queue_entry *entry;
 970         enum ip_conntrack_info uninitialized_var(ctinfo);
 971         struct nf_conn *ct = NULL;
 972
 973         struct net *net = sock_net(ctnl);
 974         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
 975
 976         queue = instance_lookup(q, queue_num);
 977         if (!queue)
 978                 queue = verdict_instance_lookup(q, queue_num,
 979                                                 NETLINK_CB(skb).portid);
 980         if (IS_ERR(queue))
 981                 return PTR_ERR(queue);
 982
 983         vhdr = verdicthdr_get(nfqa);
 984         if (!vhdr)
 985                 return -EINVAL;
 986
 987         verdict = ntohl(vhdr->verdict);
 988
 989         entry = find_dequeue_entry(queue, ntohl(vhdr->id));
 990         if (entry == NULL)
 991                 return -ENOENT;
 992
 993         rcu_read_lock();
 994         if (nfqa[NFQA_CT] && (queue->flags & NFQA_CFG_F_CONNTRACK))
 995                 ct = nfqnl_ct_parse(entry->skb, nfqa[NFQA_CT], &ctinfo);
 996
 997         if (nfqa[NFQA_PAYLOAD]) {
 998                 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
 999                 int diff = payload_len - entry->skb->len;
1000
1001                 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
1002                                  payload_len, entry, diff) < 0)
1003                         verdict = NF_DROP;
1004
1005                 if (ct)
1006                         nfqnl_ct_seq_adjust(skb, ct, ctinfo, diff);
1007         }
1008         rcu_read_unlock();
1009
1010         if (nfqa[NFQA_MARK])
1011                 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1012
1013         nf_reinject(entry, verdict);
1014         return 0;
1015 }
1016
1017 static int
1018 nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
1019                   const struct nlmsghdr *nlh,
1020                   const struct nlattr * const nfqa[])
1021 {
1022         return -ENOTSUPP;
1023 }
1024
1025 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1026         [NFQA_CFG_CMD]          = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1027         [NFQA_CFG_PARAMS]       = { .len = sizeof(struct nfqnl_msg_config_params) },
1028 };
1029
1030 static const struct nf_queue_handler nfqh = {
1031         .outfn  = &nfqnl_enqueue_packet,
1032 };
1033
1034 static int
1035 nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
1036                   const struct nlmsghdr *nlh,
1037                   const struct nlattr * const nfqa[])
1038 {
1039         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1040         u_int16_t queue_num = ntohs(nfmsg->res_id);
1041         struct nfqnl_instance *queue;
1042         struct nfqnl_msg_config_cmd *cmd = NULL;
1043         struct net *net = sock_net(ctnl);
1044         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1045         int ret = 0;
1046
1047         if (nfqa[NFQA_CFG_CMD]) {
1048                 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1049
1050                 /* Obsolete commands without queue context */
1051                 switch (cmd->command) {
1052                 case NFQNL_CFG_CMD_PF_BIND: return 0;
1053                 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1054                 }
1055         }
1056
1057         rcu_read_lock();
1058         queue = instance_lookup(q, queue_num);
1059         if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1060                 ret = -EPERM;
1061                 goto err_out_unlock;
1062         }
1063
1064         if (cmd != NULL) {
1065                 switch (cmd->command) {
1066                 case NFQNL_CFG_CMD_BIND:
1067                         if (queue) {
1068                                 ret = -EBUSY;
1069                                 goto err_out_unlock;
1070                         }
1071                         queue = instance_create(q, queue_num,
1072                                                 NETLINK_CB(skb).portid);
1073                         if (IS_ERR(queue)) {
1074                                 ret = PTR_ERR(queue);
1075                                 goto err_out_unlock;
1076                         }
1077                         break;
1078                 case NFQNL_CFG_CMD_UNBIND:
1079                         if (!queue) {
1080                                 ret = -ENODEV;
1081                                 goto err_out_unlock;
1082                         }
1083                         instance_destroy(q, queue);
1084                         break;
1085                 case NFQNL_CFG_CMD_PF_BIND:
1086                 case NFQNL_CFG_CMD_PF_UNBIND:
1087                         break;
1088                 default:
1089                         ret = -ENOTSUPP;
1090                         break;
1091                 }
1092         }
1093
1094         if (nfqa[NFQA_CFG_PARAMS]) {
1095                 struct nfqnl_msg_config_params *params;
1096
1097                 if (!queue) {
1098                         ret = -ENODEV;
1099                         goto err_out_unlock;
1100                 }
1101                 params = nla_data(nfqa[NFQA_CFG_PARAMS]);
1102                 nfqnl_set_mode(queue, params->copy_mode,
1103                                 ntohl(params->copy_range));
1104         }
1105
1106         if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1107                 __be32 *queue_maxlen;
1108
1109                 if (!queue) {
1110                         ret = -ENODEV;
1111                         goto err_out_unlock;
1112                 }
1113                 queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1114                 spin_lock_bh(&queue->lock);
1115                 queue->queue_maxlen = ntohl(*queue_maxlen);
1116                 spin_unlock_bh(&queue->lock);
1117         }
1118
1119         if (nfqa[NFQA_CFG_FLAGS]) {
1120                 __u32 flags, mask;
1121
1122                 if (!queue) {
1123                         ret = -ENODEV;
1124                         goto err_out_unlock;
1125                 }
1126
1127                 if (!nfqa[NFQA_CFG_MASK]) {
1128                         /* A mask is needed to specify which flags are being
1129                          * changed.
1130                          */
1131                         ret = -EINVAL;
1132                         goto err_out_unlock;
1133                 }
1134
1135                 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1136                 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1137
1138                 if (flags >= NFQA_CFG_F_MAX) {
1139                         ret = -EOPNOTSUPP;
1140                         goto err_out_unlock;
1141                 }
1142
1143                 spin_lock_bh(&queue->lock);
1144                 queue->flags &= ~mask;
1145                 queue->flags |= flags & mask;
1146                 spin_unlock_bh(&queue->lock);
1147         }
1148
1149 err_out_unlock:
1150         rcu_read_unlock();
1151         return ret;
1152 }
1153
1154 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1155         [NFQNL_MSG_PACKET]      = { .call_rcu = nfqnl_recv_unsupp,
1156                                     .attr_count = NFQA_MAX, },
1157         [NFQNL_MSG_VERDICT]     = { .call_rcu = nfqnl_recv_verdict,
1158                                     .attr_count = NFQA_MAX,
1159                                     .policy = nfqa_verdict_policy },
1160         [NFQNL_MSG_CONFIG]      = { .call = nfqnl_recv_config,
1161                                     .attr_count = NFQA_CFG_MAX,
1162                                     .policy = nfqa_cfg_policy },
1163         [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
1164                                     .attr_count = NFQA_MAX,
1165                                     .policy = nfqa_verdict_batch_policy },
1166 };
1167
1168 static const struct nfnetlink_subsystem nfqnl_subsys = {
1169         .name           = "nf_queue",
1170         .subsys_id      = NFNL_SUBSYS_QUEUE,
1171         .cb_count       = NFQNL_MSG_MAX,
1172         .cb             = nfqnl_cb,
1173 };
1174
1175 #ifdef CONFIG_PROC_FS
1176 struct iter_state {
1177         struct seq_net_private p;
1178         unsigned int bucket;
1179 };
1180
1181 static struct hlist_node *get_first(struct seq_file *seq)
1182 {
1183         struct iter_state *st = seq->private;
1184         struct net *net;
1185         struct nfnl_queue_net *q;
1186
1187         if (!st)
1188                 return NULL;
1189
1190         net = seq_file_net(seq);
1191         q = nfnl_queue_pernet(net);
1192         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1193                 if (!hlist_empty(&q->instance_table[st->bucket]))
1194                         return q->instance_table[st->bucket].first;
1195         }
1196         return NULL;
1197 }
1198
1199 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1200 {
1201         struct iter_state *st = seq->private;
1202         struct net *net = seq_file_net(seq);
1203
1204         h = h->next;
1205         while (!h) {
1206                 struct nfnl_queue_net *q;
1207
1208                 if (++st->bucket >= INSTANCE_BUCKETS)
1209                         return NULL;
1210
1211                 q = nfnl_queue_pernet(net);
1212                 h = q->instance_table[st->bucket].first;
1213         }
1214         return h;
1215 }
1216
1217 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1218 {
1219         struct hlist_node *head;
1220         head = get_first(seq);
1221
1222         if (head)
1223                 while (pos && (head = get_next(seq, head)))
1224                         pos--;
1225         return pos ? NULL : head;
1226 }
1227
1228 static void *seq_start(struct seq_file *s, loff_t *pos)
1229         __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1230 {
1231         spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1232         return get_idx(s, *pos);
1233 }
1234
1235 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1236 {
1237         (*pos)++;
1238         return get_next(s, v);
1239 }
1240
1241 static void seq_stop(struct seq_file *s, void *v)
1242         __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1243 {
1244         spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1245 }
1246
1247 static int seq_show(struct seq_file *s, void *v)
1248 {
1249         const struct nfqnl_instance *inst = v;
1250
1251         return seq_printf(s, "%5d %6d %5d %1d %5d %5d %5d %8d %2d\n",
1252                           inst->queue_num,
1253                           inst->peer_portid, inst->queue_total,
1254                           inst->copy_mode, inst->copy_range,
1255                           inst->queue_dropped, inst->queue_user_dropped,
1256                           inst->id_sequence, 1);
1257 }
1258
1259 static const struct seq_operations nfqnl_seq_ops = {
1260         .start  = seq_start,
1261         .next   = seq_next,
1262         .stop   = seq_stop,
1263         .show   = seq_show,
1264 };
1265
1266 static int nfqnl_open(struct inode *inode, struct file *file)
1267 {
1268         return seq_open_net(inode, file, &nfqnl_seq_ops,
1269                         sizeof(struct iter_state));
1270 }
1271
1272 static const struct file_operations nfqnl_file_ops = {
1273         .owner   = THIS_MODULE,
1274         .open    = nfqnl_open,
1275         .read    = seq_read,
1276         .llseek  = seq_lseek,
1277         .release = seq_release_net,
1278 };
1279
1280 #endif /* PROC_FS */
1281
1282 static int __net_init nfnl_queue_net_init(struct net *net)
1283 {
1284         unsigned int i;
1285         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1286
1287         for (i = 0; i < INSTANCE_BUCKETS; i++)
1288                 INIT_HLIST_HEAD(&q->instance_table[i]);
1289
1290         spin_lock_init(&q->instances_lock);
1291
1292 #ifdef CONFIG_PROC_FS
1293         if (!proc_create("nfnetlink_queue", 0440,
1294                          net->nf.proc_netfilter, &nfqnl_file_ops))
1295                 return -ENOMEM;
1296 #endif
1297         return 0;
1298 }
1299
1300 static void __net_exit nfnl_queue_net_exit(struct net *net)
1301 {
1302 #ifdef CONFIG_PROC_FS
1303         remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1304 #endif
1305 }
1306
1307 static struct pernet_operations nfnl_queue_net_ops = {
1308         .init   = nfnl_queue_net_init,
1309         .exit   = nfnl_queue_net_exit,
1310         .id     = &nfnl_queue_net_id,
1311         .size   = sizeof(struct nfnl_queue_net),
1312 };
1313
1314 static int __init nfnetlink_queue_init(void)
1315 {
1316         int status = -ENOMEM;
1317
1318         netlink_register_notifier(&nfqnl_rtnl_notifier);
1319         status = nfnetlink_subsys_register(&nfqnl_subsys);
1320         if (status < 0) {
1321                 pr_err("nf_queue: failed to create netlink socket\n");
1322                 goto cleanup_netlink_notifier;
1323         }
1324
1325         status = register_pernet_subsys(&nfnl_queue_net_ops);
1326         if (status < 0) {
1327                 pr_err("nf_queue: failed to register pernet ops\n");
1328                 goto cleanup_subsys;
1329         }
1330         register_netdevice_notifier(&nfqnl_dev_notifier);
1331         nf_register_queue_handler(&nfqh);
1332         return status;
1333
1334 cleanup_subsys:
1335         nfnetlink_subsys_unregister(&nfqnl_subsys);
1336 cleanup_netlink_notifier:
1337         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1338         return status;
1339 }
1340
1341 static void __exit nfnetlink_queue_fini(void)
1342 {
1343         nf_unregister_queue_handler();
1344         unregister_netdevice_notifier(&nfqnl_dev_notifier);
1345         unregister_pernet_subsys(&nfnl_queue_net_ops);
1346         nfnetlink_subsys_unregister(&nfqnl_subsys);
1347         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1348
1349         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1350 }
1351
1352 MODULE_DESCRIPTION("netfilter packet queue handler");
1353 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1354 MODULE_LICENSE("GPL");
1355 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1356
1357 module_init(nfnetlink_queue_init);
1358 module_exit(nfnetlink_queue_fini);