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