net/netfilter/nfnetlink_queue.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/notifier.h>
  22 #include <linux/netdevice.h>
  23 #include <linux/netfilter.h>
  24 #include <linux/proc_fs.h>
  25 #include <linux/netfilter_ipv4.h>
  26 #include <linux/netfilter_ipv6.h>
  27 #include <linux/netfilter/nfnetlink.h>
  28 #include <linux/netfilter/nfnetlink_queue.h>
  29 #include <linux/list.h>
  30 #include <net/sock.h>
  31 #include <net/netfilter/nf_queue.h>
  32
  33 #include <asm/atomic.h>
  34
  35 #ifdef CONFIG_BRIDGE_NETFILTER
  36 #include "../bridge/br_private.h"
  37 #endif
  38
  39 #define NFQNL_QMAX_DEFAULT 1024
  40
  41 struct nfqnl_instance {
  42         struct hlist_node hlist;                /* global list of queues */
  43         struct rcu_head rcu;
  44
  45         int peer_pid;
  46         unsigned int queue_maxlen;
  47         unsigned int copy_range;
  48         unsigned int queue_total;
  49         unsigned int queue_dropped;
  50         unsigned int queue_user_dropped;
  51
  52         unsigned int id_sequence;               /* 'sequence' of pkt ids */
  53
  54         u_int16_t queue_num;                    /* number of this queue */
  55         u_int8_t copy_mode;
  56
  57         spinlock_t lock;
  58
  59         struct list_head queue_list;            /* packets in queue */
  60 };
  61
  62 typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
  63
  64 static DEFINE_SPINLOCK(instances_lock);
  65
  66 #define INSTANCE_BUCKETS        16
  67 static struct hlist_head instance_table[INSTANCE_BUCKETS] __read_mostly;
  68
  69 static inline u_int8_t instance_hashfn(u_int16_t queue_num)
  70 {
  71         return ((queue_num >> 8) | queue_num) % INSTANCE_BUCKETS;
  72 }
  73
  74 static struct nfqnl_instance *
  75 instance_lookup(u_int16_t queue_num)
  76 {
  77         struct hlist_head *head;
  78         struct hlist_node *pos;
  79         struct nfqnl_instance *inst;
  80
  81         head = &instance_table[instance_hashfn(queue_num)];
  82         hlist_for_each_entry_rcu(inst, pos, head, hlist) {
  83                 if (inst->queue_num == queue_num)
  84                         return inst;
  85         }
  86         return NULL;
  87 }
  88
  89 static struct nfqnl_instance *
  90 instance_create(u_int16_t queue_num, int pid)
  91 {
  92         struct nfqnl_instance *inst;
  93         unsigned int h;
  94         int err;
  95
  96         spin_lock(&instances_lock);
  97         if (instance_lookup(queue_num)) {
  98                 err = -EEXIST;
  99                 goto out_unlock;
 100         }
 101
 102         inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
 103         if (!inst) {
 104                 err = -ENOMEM;
 105                 goto out_unlock;
 106         }
 107
 108         inst->queue_num = queue_num;
 109         inst->peer_pid = pid;
 110         inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
 111         inst->copy_range = 0xfffff;
 112         inst->copy_mode = NFQNL_COPY_NONE;
 113         spin_lock_init(&inst->lock);
 114         INIT_LIST_HEAD(&inst->queue_list);
 115
 116         if (!try_module_get(THIS_MODULE)) {
 117                 err = -EAGAIN;
 118                 goto out_free;
 119         }
 120
 121         h = instance_hashfn(queue_num);
 122         hlist_add_head_rcu(&inst->hlist, &instance_table[h]);
 123
 124         spin_unlock(&instances_lock);
 125
 126         return inst;
 127
 128 out_free:
 129         kfree(inst);
 130 out_unlock:
 131         spin_unlock(&instances_lock);
 132         return ERR_PTR(err);
 133 }
 134
 135 static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
 136                         unsigned long data);
 137
 138 static void
 139 instance_destroy_rcu(struct rcu_head *head)
 140 {
 141         struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
 142                                                    rcu);
 143
 144         nfqnl_flush(inst, NULL, 0);
 145         kfree(inst);
 146         module_put(THIS_MODULE);
 147 }
 148
 149 static void
 150 __instance_destroy(struct nfqnl_instance *inst)
 151 {
 152         hlist_del_rcu(&inst->hlist);
 153         call_rcu(&inst->rcu, instance_destroy_rcu);
 154 }
 155
 156 static void
 157 instance_destroy(struct nfqnl_instance *inst)
 158 {
 159         spin_lock(&instances_lock);
 160         __instance_destroy(inst);
 161         spin_unlock(&instances_lock);
 162 }
 163
 164 static inline void
 165 __enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
 166 {
 167        list_add_tail(&entry->list, &queue->queue_list);
 168        queue->queue_total++;
 169 }
 170
 171 static struct nf_queue_entry *
 172 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
 173 {
 174         struct nf_queue_entry *entry = NULL, *i;
 175
 176         spin_lock_bh(&queue->lock);
 177
 178         list_for_each_entry(i, &queue->queue_list, list) {
 179                 if (i->id == id) {
 180                         entry = i;
 181                         break;
 182                 }
 183         }
 184
 185         if (entry) {
 186                 list_del(&entry->list);
 187                 queue->queue_total--;
 188         }
 189
 190         spin_unlock_bh(&queue->lock);
 191
 192         return entry;
 193 }
 194
 195 static void
 196 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
 197 {
 198         struct nf_queue_entry *entry, *next;
 199
 200         spin_lock_bh(&queue->lock);
 201         list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
 202                 if (!cmpfn || cmpfn(entry, data)) {
 203                         list_del(&entry->list);
 204                         queue->queue_total--;
 205                         nf_reinject(entry, NF_DROP);
 206                 }
 207         }
 208         spin_unlock_bh(&queue->lock);
 209 }
 210
 211 static struct sk_buff *
 212 nfqnl_build_packet_message(struct nfqnl_instance *queue,
 213                            struct nf_queue_entry *entry)
 214 {
 215         sk_buff_data_t old_tail;
 216         size_t size;
 217         size_t data_len = 0;
 218         struct sk_buff *skb;
 219         struct nfqnl_msg_packet_hdr pmsg;
 220         struct nlmsghdr *nlh;
 221         struct nfgenmsg *nfmsg;
 222         struct sk_buff *entskb = entry->skb;
 223         struct net_device *indev;
 224         struct net_device *outdev;
 225
 226         size =    NLMSG_SPACE(sizeof(struct nfgenmsg))
 227                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
 228                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 229                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 230 #ifdef CONFIG_BRIDGE_NETFILTER
 231                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 232                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 233 #endif
 234                 + nla_total_size(sizeof(u_int32_t))     /* mark */
 235                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
 236                 + nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
 237
 238         outdev = entry->outdev;
 239
 240         spin_lock_bh(&queue->lock);
 241
 242         switch ((enum nfqnl_config_mode)queue->copy_mode) {
 243         case NFQNL_COPY_META:
 244         case NFQNL_COPY_NONE:
 245                 break;
 246
 247         case NFQNL_COPY_PACKET:
 248                 if (entskb->ip_summed == CHECKSUM_PARTIAL &&
 249                     skb_checksum_help(entskb)) {
 250                         spin_unlock_bh(&queue->lock);
 251                         return NULL;
 252                 }
 253                 if (queue->copy_range == 0
 254                     || queue->copy_range > entskb->len)
 255                         data_len = entskb->len;
 256                 else
 257                         data_len = queue->copy_range;
 258
 259                 size += nla_total_size(data_len);
 260                 break;
 261         }
 262
 263         entry->id = queue->id_sequence++;
 264
 265         spin_unlock_bh(&queue->lock);
 266
 267         skb = alloc_skb(size, GFP_ATOMIC);
 268         if (!skb)
 269                 goto nlmsg_failure;
 270
 271         old_tail = skb->tail;
 272         nlh = NLMSG_PUT(skb, 0, 0,
 273                         NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
 274                         sizeof(struct nfgenmsg));
 275         nfmsg = NLMSG_DATA(nlh);
 276         nfmsg->nfgen_family = entry->pf;
 277         nfmsg->version = NFNETLINK_V0;
 278         nfmsg->res_id = htons(queue->queue_num);
 279
 280         pmsg.packet_id          = htonl(entry->id);
 281         pmsg.hw_protocol        = entskb->protocol;
 282         pmsg.hook               = entry->hook;
 283
 284         NLA_PUT(skb, NFQA_PACKET_HDR, sizeof(pmsg), &pmsg);
 285
 286         indev = entry->indev;
 287         if (indev) {
 288 #ifndef CONFIG_BRIDGE_NETFILTER
 289                 NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex));
 290 #else
 291                 if (entry->pf == PF_BRIDGE) {
 292                         /* Case 1: indev is physical input device, we need to
 293                          * look for bridge group (when called from
 294                          * netfilter_bridge) */
 295                         NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSINDEV,
 296                                      htonl(indev->ifindex));
 297                         /* this is the bridge group "brX" */
 298                         NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV,
 299                                      htonl(indev->br_port->br->dev->ifindex));
 300                 } else {
 301                         /* Case 2: indev is bridge group, we need to look for
 302                          * physical device (when called from ipv4) */
 303                         NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV,
 304                                      htonl(indev->ifindex));
 305                         if (entskb->nf_bridge && entskb->nf_bridge->physindev)
 306                                 NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSINDEV,
 307                                              htonl(entskb->nf_bridge->physindev->ifindex));
 308                 }
 309 #endif
 310         }
 311
 312         if (outdev) {
 313 #ifndef CONFIG_BRIDGE_NETFILTER
 314                 NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex));
 315 #else
 316                 if (entry->pf == PF_BRIDGE) {
 317                         /* Case 1: outdev is physical output device, we need to
 318                          * look for bridge group (when called from
 319                          * netfilter_bridge) */
 320                         NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 321                                      htonl(outdev->ifindex));
 322                         /* this is the bridge group "brX" */
 323                         NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV,
 324                                      htonl(outdev->br_port->br->dev->ifindex));
 325                 } else {
 326                         /* Case 2: outdev is bridge group, we need to look for
 327                          * physical output device (when called from ipv4) */
 328                         NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV,
 329                                      htonl(outdev->ifindex));
 330                         if (entskb->nf_bridge && entskb->nf_bridge->physoutdev)
 331                                 NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 332                                              htonl(entskb->nf_bridge->physoutdev->ifindex));
 333                 }
 334 #endif
 335         }
 336
 337         if (entskb->mark)
 338                 NLA_PUT_BE32(skb, NFQA_MARK, htonl(entskb->mark));
 339
 340         if (indev && entskb->dev) {
 341                 struct nfqnl_msg_packet_hw phw;
 342                 int len = dev_parse_header(entskb, phw.hw_addr);
 343                 if (len) {
 344                         phw.hw_addrlen = htons(len);
 345                         NLA_PUT(skb, NFQA_HWADDR, sizeof(phw), &phw);
 346                 }
 347         }
 348
 349         if (entskb->tstamp.tv64) {
 350                 struct nfqnl_msg_packet_timestamp ts;
 351                 struct timeval tv = ktime_to_timeval(entskb->tstamp);
 352                 ts.sec = cpu_to_be64(tv.tv_sec);
 353                 ts.usec = cpu_to_be64(tv.tv_usec);
 354
 355                 NLA_PUT(skb, NFQA_TIMESTAMP, sizeof(ts), &ts);
 356         }
 357
 358         if (data_len) {
 359                 struct nlattr *nla;
 360                 int sz = nla_attr_size(data_len);
 361
 362                 if (skb_tailroom(skb) < nla_total_size(data_len)) {
 363                         printk(KERN_WARNING "nf_queue: no tailroom!\n");
 364                         goto nlmsg_failure;
 365                 }
 366
 367                 nla = (struct nlattr *)skb_put(skb, nla_total_size(data_len));
 368                 nla->nla_type = NFQA_PAYLOAD;
 369                 nla->nla_len = sz;
 370
 371                 if (skb_copy_bits(entskb, 0, nla_data(nla), data_len))
 372                         BUG();
 373         }
 374
 375         nlh->nlmsg_len = skb->tail - old_tail;
 376         return skb;
 377
 378 nlmsg_failure:
 379 nla_put_failure:
 380         if (skb)
 381                 kfree_skb(skb);
 382         if (net_ratelimit())
 383                 printk(KERN_ERR "nf_queue: error creating packet message\n");
 384         return NULL;
 385 }
 386
 387 static int
 388 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 389 {
 390         struct sk_buff *nskb;
 391         struct nfqnl_instance *queue;
 392         int err;
 393
 394         /* rcu_read_lock()ed by nf_hook_slow() */
 395         queue = instance_lookup(queuenum);
 396         if (!queue)
 397                 goto err_out;
 398
 399         if (queue->copy_mode == NFQNL_COPY_NONE)
 400                 goto err_out;
 401
 402         nskb = nfqnl_build_packet_message(queue, entry);
 403         if (nskb == NULL)
 404                 goto err_out;
 405
 406         spin_lock_bh(&queue->lock);
 407
 408         if (!queue->peer_pid)
 409                 goto err_out_free_nskb;
 410
 411         if (queue->queue_total >= queue->queue_maxlen) {
 412                 queue->queue_dropped++;
 413                 if (net_ratelimit())
 414                           printk(KERN_WARNING "nf_queue: full at %d entries, "
 415                                  "dropping packets(s).\n",
 416                                  queue->queue_total);
 417                 goto err_out_free_nskb;
 418         }
 419
 420         /* nfnetlink_unicast will either free the nskb or add it to a socket */
 421         err = nfnetlink_unicast(nskb, &init_net, queue->peer_pid, MSG_DONTWAIT);
 422         if (err < 0) {
 423                 queue->queue_user_dropped++;
 424                 goto err_out_unlock;
 425         }
 426
 427         __enqueue_entry(queue, entry);
 428
 429         spin_unlock_bh(&queue->lock);
 430         return 0;
 431
 432 err_out_free_nskb:
 433         kfree_skb(nskb);
 434 err_out_unlock:
 435         spin_unlock_bh(&queue->lock);
 436 err_out:
 437         return -1;
 438 }
 439
 440 static int
 441 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e)
 442 {
 443         struct sk_buff *nskb;
 444         int diff;
 445
 446         diff = data_len - e->skb->len;
 447         if (diff < 0) {
 448                 if (pskb_trim(e->skb, data_len))
 449                         return -ENOMEM;
 450         } else if (diff > 0) {
 451                 if (data_len > 0xFFFF)
 452                         return -EINVAL;
 453                 if (diff > skb_tailroom(e->skb)) {
 454                         nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
 455                                                diff, GFP_ATOMIC);
 456                         if (!nskb) {
 457                                 printk(KERN_WARNING "nf_queue: OOM "
 458                                       "in mangle, dropping packet\n");
 459                                 return -ENOMEM;
 460                         }
 461                         kfree_skb(e->skb);
 462                         e->skb = nskb;
 463                 }
 464                 skb_put(e->skb, diff);
 465         }
 466         if (!skb_make_writable(e->skb, data_len))
 467                 return -ENOMEM;
 468         skb_copy_to_linear_data(e->skb, data, data_len);
 469         e->skb->ip_summed = CHECKSUM_NONE;
 470         return 0;
 471 }
 472
 473 static int
 474 nfqnl_set_mode(struct nfqnl_instance *queue,
 475                unsigned char mode, unsigned int range)
 476 {
 477         int status = 0;
 478
 479         spin_lock_bh(&queue->lock);
 480         switch (mode) {
 481         case NFQNL_COPY_NONE:
 482         case NFQNL_COPY_META:
 483                 queue->copy_mode = mode;
 484                 queue->copy_range = 0;
 485                 break;
 486
 487         case NFQNL_COPY_PACKET:
 488                 queue->copy_mode = mode;
 489                 /* we're using struct nlattr which has 16bit nla_len */
 490                 if (range > 0xffff)
 491                         queue->copy_range = 0xffff;
 492                 else
 493                         queue->copy_range = range;
 494                 break;
 495
 496         default:
 497                 status = -EINVAL;
 498
 499         }
 500         spin_unlock_bh(&queue->lock);
 501
 502         return status;
 503 }
 504
 505 static int
 506 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
 507 {
 508         if (entry->indev)
 509                 if (entry->indev->ifindex == ifindex)
 510                         return 1;
 511         if (entry->outdev)
 512                 if (entry->outdev->ifindex == ifindex)
 513                         return 1;
 514 #ifdef CONFIG_BRIDGE_NETFILTER
 515         if (entry->skb->nf_bridge) {
 516                 if (entry->skb->nf_bridge->physindev &&
 517                     entry->skb->nf_bridge->physindev->ifindex == ifindex)
 518                         return 1;
 519                 if (entry->skb->nf_bridge->physoutdev &&
 520                     entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
 521                         return 1;
 522         }
 523 #endif
 524         return 0;
 525 }
 526
 527 /* drop all packets with either indev or outdev == ifindex from all queue
 528  * instances */
 529 static void
 530 nfqnl_dev_drop(int ifindex)
 531 {
 532         int i;
 533
 534         rcu_read_lock();
 535
 536         for (i = 0; i < INSTANCE_BUCKETS; i++) {
 537                 struct hlist_node *tmp;
 538                 struct nfqnl_instance *inst;
 539                 struct hlist_head *head = &instance_table[i];
 540
 541                 hlist_for_each_entry_rcu(inst, tmp, head, hlist)
 542                         nfqnl_flush(inst, dev_cmp, ifindex);
 543         }
 544
 545         rcu_read_unlock();
 546 }
 547
 548 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
 549
 550 static int
 551 nfqnl_rcv_dev_event(struct notifier_block *this,
 552                     unsigned long event, void *ptr)
 553 {
 554         struct net_device *dev = ptr;
 555
 556         if (!net_eq(dev_net(dev), &init_net))
 557                 return NOTIFY_DONE;
 558
 559         /* Drop any packets associated with the downed device */
 560         if (event == NETDEV_DOWN)
 561                 nfqnl_dev_drop(dev->ifindex);
 562         return NOTIFY_DONE;
 563 }
 564
 565 static struct notifier_block nfqnl_dev_notifier = {
 566         .notifier_call  = nfqnl_rcv_dev_event,
 567 };
 568
 569 static int
 570 nfqnl_rcv_nl_event(struct notifier_block *this,
 571                    unsigned long event, void *ptr)
 572 {
 573         struct netlink_notify *n = ptr;
 574
 575         if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
 576                 int i;
 577
 578                 /* destroy all instances for this pid */
 579                 spin_lock(&instances_lock);
 580                 for (i = 0; i < INSTANCE_BUCKETS; i++) {
 581                         struct hlist_node *tmp, *t2;
 582                         struct nfqnl_instance *inst;
 583                         struct hlist_head *head = &instance_table[i];
 584
 585                         hlist_for_each_entry_safe(inst, tmp, t2, head, hlist) {
 586                                 if ((n->net == &init_net) &&
 587                                     (n->pid == inst->peer_pid))
 588                                         __instance_destroy(inst);
 589                         }
 590                 }
 591                 spin_unlock(&instances_lock);
 592         }
 593         return NOTIFY_DONE;
 594 }
 595
 596 static struct notifier_block nfqnl_rtnl_notifier = {
 597         .notifier_call  = nfqnl_rcv_nl_event,
 598 };
 599
 600 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
 601         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 602         [NFQA_MARK]             = { .type = NLA_U32 },
 603         [NFQA_PAYLOAD]          = { .type = NLA_UNSPEC },
 604 };
 605
 606 static int
 607 nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
 608                    const struct nlmsghdr *nlh,
 609                    const struct nlattr * const nfqa[])
 610 {
 611         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 612         u_int16_t queue_num = ntohs(nfmsg->res_id);
 613
 614         struct nfqnl_msg_verdict_hdr *vhdr;
 615         struct nfqnl_instance *queue;
 616         unsigned int verdict;
 617         struct nf_queue_entry *entry;
 618         int err;
 619
 620         rcu_read_lock();
 621         queue = instance_lookup(queue_num);
 622         if (!queue) {
 623                 err = -ENODEV;
 624                 goto err_out_unlock;
 625         }
 626
 627         if (queue->peer_pid != NETLINK_CB(skb).pid) {
 628                 err = -EPERM;
 629                 goto err_out_unlock;
 630         }
 631
 632         if (!nfqa[NFQA_VERDICT_HDR]) {
 633                 err = -EINVAL;
 634                 goto err_out_unlock;
 635         }
 636
 637         vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
 638         verdict = ntohl(vhdr->verdict);
 639
 640         if ((verdict & NF_VERDICT_MASK) > NF_MAX_VERDICT) {
 641                 err = -EINVAL;
 642                 goto err_out_unlock;
 643         }
 644
 645         entry = find_dequeue_entry(queue, ntohl(vhdr->id));
 646         if (entry == NULL) {
 647                 err = -ENOENT;
 648                 goto err_out_unlock;
 649         }
 650         rcu_read_unlock();
 651
 652         if (nfqa[NFQA_PAYLOAD]) {
 653                 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
 654                                  nla_len(nfqa[NFQA_PAYLOAD]), entry) < 0)
 655                         verdict = NF_DROP;
 656         }
 657
 658         if (nfqa[NFQA_MARK])
 659                 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
 660
 661         nf_reinject(entry, verdict);
 662         return 0;
 663
 664 err_out_unlock:
 665         rcu_read_unlock();
 666         return err;
 667 }
 668
 669 static int
 670 nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
 671                   const struct nlmsghdr *nlh,
 672                   const struct nlattr * const nfqa[])
 673 {
 674         return -ENOTSUPP;
 675 }
 676
 677 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
 678         [NFQA_CFG_CMD]          = { .len = sizeof(struct nfqnl_msg_config_cmd) },
 679         [NFQA_CFG_PARAMS]       = { .len = sizeof(struct nfqnl_msg_config_params) },
 680 };
 681
 682 static const struct nf_queue_handler nfqh = {
 683         .name   = "nf_queue",
 684         .outfn  = &nfqnl_enqueue_packet,
 685 };
 686
 687 static int
 688 nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
 689                   const struct nlmsghdr *nlh,
 690                   const struct nlattr * const nfqa[])
 691 {
 692         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 693         u_int16_t queue_num = ntohs(nfmsg->res_id);
 694         struct nfqnl_instance *queue;
 695         struct nfqnl_msg_config_cmd *cmd = NULL;
 696         int ret = 0;
 697
 698         if (nfqa[NFQA_CFG_CMD]) {
 699                 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
 700
 701                 /* Commands without queue context - might sleep */
 702                 switch (cmd->command) {
 703                 case NFQNL_CFG_CMD_PF_BIND:
 704                         return nf_register_queue_handler(ntohs(cmd->pf),
 705                                                          &nfqh);
 706                 case NFQNL_CFG_CMD_PF_UNBIND:
 707                         return nf_unregister_queue_handler(ntohs(cmd->pf),
 708                                                            &nfqh);
 709                 }
 710         }
 711
 712         rcu_read_lock();
 713         queue = instance_lookup(queue_num);
 714         if (queue && queue->peer_pid != NETLINK_CB(skb).pid) {
 715                 ret = -EPERM;
 716                 goto err_out_unlock;
 717         }
 718
 719         if (cmd != NULL) {
 720                 switch (cmd->command) {
 721                 case NFQNL_CFG_CMD_BIND:
 722                         if (queue) {
 723                                 ret = -EBUSY;
 724                                 goto err_out_unlock;
 725                         }
 726                         queue = instance_create(queue_num, NETLINK_CB(skb).pid);
 727                         if (IS_ERR(queue)) {
 728                                 ret = PTR_ERR(queue);
 729                                 goto err_out_unlock;
 730                         }
 731                         break;
 732                 case NFQNL_CFG_CMD_UNBIND:
 733                         if (!queue) {
 734                                 ret = -ENODEV;
 735                                 goto err_out_unlock;
 736                         }
 737                         instance_destroy(queue);
 738                         break;
 739                 case NFQNL_CFG_CMD_PF_BIND:
 740                 case NFQNL_CFG_CMD_PF_UNBIND:
 741                         break;
 742                 default:
 743                         ret = -ENOTSUPP;
 744                         break;
 745                 }
 746         }
 747
 748         if (nfqa[NFQA_CFG_PARAMS]) {
 749                 struct nfqnl_msg_config_params *params;
 750
 751                 if (!queue) {
 752                         ret = -ENODEV;
 753                         goto err_out_unlock;
 754                 }
 755                 params = nla_data(nfqa[NFQA_CFG_PARAMS]);
 756                 nfqnl_set_mode(queue, params->copy_mode,
 757                                 ntohl(params->copy_range));
 758         }
 759
 760         if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
 761                 __be32 *queue_maxlen;
 762
 763                 if (!queue) {
 764                         ret = -ENODEV;
 765                         goto err_out_unlock;
 766                 }
 767                 queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
 768                 spin_lock_bh(&queue->lock);
 769                 queue->queue_maxlen = ntohl(*queue_maxlen);
 770                 spin_unlock_bh(&queue->lock);
 771         }
 772
 773 err_out_unlock:
 774         rcu_read_unlock();
 775         return ret;
 776 }
 777
 778 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
 779         [NFQNL_MSG_PACKET]      = { .call = nfqnl_recv_unsupp,
 780                                     .attr_count = NFQA_MAX, },
 781         [NFQNL_MSG_VERDICT]     = { .call = nfqnl_recv_verdict,
 782                                     .attr_count = NFQA_MAX,
 783                                     .policy = nfqa_verdict_policy },
 784         [NFQNL_MSG_CONFIG]      = { .call = nfqnl_recv_config,
 785                                     .attr_count = NFQA_CFG_MAX,
 786                                     .policy = nfqa_cfg_policy },
 787 };
 788
 789 static const struct nfnetlink_subsystem nfqnl_subsys = {
 790         .name           = "nf_queue",
 791         .subsys_id      = NFNL_SUBSYS_QUEUE,
 792         .cb_count       = NFQNL_MSG_MAX,
 793         .cb             = nfqnl_cb,
 794 };
 795
 796 #ifdef CONFIG_PROC_FS
 797 struct iter_state {
 798         unsigned int bucket;
 799 };
 800
 801 static struct hlist_node *get_first(struct seq_file *seq)
 802 {
 803         struct iter_state *st = seq->private;
 804
 805         if (!st)
 806                 return NULL;
 807
 808         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
 809                 if (!hlist_empty(&instance_table[st->bucket]))
 810                         return instance_table[st->bucket].first;
 811         }
 812         return NULL;
 813 }
 814
 815 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
 816 {
 817         struct iter_state *st = seq->private;
 818
 819         h = h->next;
 820         while (!h) {
 821                 if (++st->bucket >= INSTANCE_BUCKETS)
 822                         return NULL;
 823
 824                 h = instance_table[st->bucket].first;
 825         }
 826         return h;
 827 }
 828
 829 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
 830 {
 831         struct hlist_node *head;
 832         head = get_first(seq);
 833
 834         if (head)
 835                 while (pos && (head = get_next(seq, head)))
 836                         pos--;
 837         return pos ? NULL : head;
 838 }
 839
 840 static void *seq_start(struct seq_file *seq, loff_t *pos)
 841         __acquires(instances_lock)
 842 {
 843         spin_lock(&instances_lock);
 844         return get_idx(seq, *pos);
 845 }
 846
 847 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
 848 {
 849         (*pos)++;
 850         return get_next(s, v);
 851 }
 852
 853 static void seq_stop(struct seq_file *s, void *v)
 854         __releases(instances_lock)
 855 {
 856         spin_unlock(&instances_lock);
 857 }
 858
 859 static int seq_show(struct seq_file *s, void *v)
 860 {
 861         const struct nfqnl_instance *inst = v;
 862
 863         return seq_printf(s, "%5d %6d %5d %1d %5d %5d %5d %8d %2d\n",
 864                           inst->queue_num,
 865                           inst->peer_pid, inst->queue_total,
 866                           inst->copy_mode, inst->copy_range,
 867                           inst->queue_dropped, inst->queue_user_dropped,
 868                           inst->id_sequence, 1);
 869 }
 870
 871 static const struct seq_operations nfqnl_seq_ops = {
 872         .start  = seq_start,
 873         .next   = seq_next,
 874         .stop   = seq_stop,
 875         .show   = seq_show,
 876 };
 877
 878 static int nfqnl_open(struct inode *inode, struct file *file)
 879 {
 880         return seq_open_private(file, &nfqnl_seq_ops,
 881                         sizeof(struct iter_state));
 882 }
 883
 884 static const struct file_operations nfqnl_file_ops = {
 885         .owner   = THIS_MODULE,
 886         .open    = nfqnl_open,
 887         .read    = seq_read,
 888         .llseek  = seq_lseek,
 889         .release = seq_release_private,
 890 };
 891
 892 #endif /* PROC_FS */
 893
 894 static int __init nfnetlink_queue_init(void)
 895 {
 896         int i, status = -ENOMEM;
 897
 898         for (i = 0; i < INSTANCE_BUCKETS; i++)
 899                 INIT_HLIST_HEAD(&instance_table[i]);
 900
 901         netlink_register_notifier(&nfqnl_rtnl_notifier);
 902         status = nfnetlink_subsys_register(&nfqnl_subsys);
 903         if (status < 0) {
 904                 printk(KERN_ERR "nf_queue: failed to create netlink socket\n");
 905                 goto cleanup_netlink_notifier;
 906         }
 907
 908 #ifdef CONFIG_PROC_FS
 909         if (!proc_create("nfnetlink_queue", 0440,
 910                          proc_net_netfilter, &nfqnl_file_ops))
 911                 goto cleanup_subsys;
 912 #endif
 913
 914         register_netdevice_notifier(&nfqnl_dev_notifier);
 915         return status;
 916
 917 #ifdef CONFIG_PROC_FS
 918 cleanup_subsys:
 919         nfnetlink_subsys_unregister(&nfqnl_subsys);
 920 #endif
 921 cleanup_netlink_notifier:
 922         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
 923         return status;
 924 }
 925
 926 static void __exit nfnetlink_queue_fini(void)
 927 {
 928         nf_unregister_queue_handlers(&nfqh);
 929         unregister_netdevice_notifier(&nfqnl_dev_notifier);
 930 #ifdef CONFIG_PROC_FS
 931         remove_proc_entry("nfnetlink_queue", proc_net_netfilter);
 932 #endif
 933         nfnetlink_subsys_unregister(&nfqnl_subsys);
 934         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
 935
 936         rcu_barrier(); /* Wait for completion of call_rcu()'s */
 937 }
 938
 939 MODULE_DESCRIPTION("netfilter packet queue handler");
 940 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
 941 MODULE_LICENSE("GPL");
 942 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
 943
 944 module_init(nfnetlink_queue_init);
 945 module_exit(nfnetlink_queue_fini);