Pileus Git - ~andy/linux/blob - 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/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
  34 #include <asm/atomic.h>
  35
  36 #ifdef CONFIG_BRIDGE_NETFILTER
  37 #include "../bridge/br_private.h"
  38 #endif
  39
  40 #define NFQNL_QMAX_DEFAULT 1024
  41
  42 struct nfqnl_instance {
  43         struct hlist_node hlist;                /* global list of queues */
  44         struct rcu_head rcu;
  45
  46         int peer_pid;
  47         unsigned int queue_maxlen;
  48         unsigned int copy_range;
  49         unsigned int queue_dropped;
  50         unsigned int queue_user_dropped;
  51
  52
  53         u_int16_t queue_num;                    /* number of this queue */
  54         u_int8_t copy_mode;
  55 /*
  56  * Following fields are dirtied for each queued packet,
  57  * keep them in same cache line if possible.
  58  */
  59         spinlock_t      lock;
  60         unsigned int    queue_total;
  61         unsigned int    id_sequence;            /* 'sequence' of pkt ids */
  62         struct list_head queue_list;            /* packets in queue */
  63 };
  64
  65 typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
  66
  67 static DEFINE_SPINLOCK(instances_lock);
  68
  69 #define INSTANCE_BUCKETS        16
  70 static struct hlist_head instance_table[INSTANCE_BUCKETS] __read_mostly;
  71
  72 static inline u_int8_t instance_hashfn(u_int16_t queue_num)
  73 {
  74         return ((queue_num >> 8) | queue_num) % INSTANCE_BUCKETS;
  75 }
  76
  77 static struct nfqnl_instance *
  78 instance_lookup(u_int16_t queue_num)
  79 {
  80         struct hlist_head *head;
  81         struct hlist_node *pos;
  82         struct nfqnl_instance *inst;
  83
  84         head = &instance_table[instance_hashfn(queue_num)];
  85         hlist_for_each_entry_rcu(inst, pos, head, hlist) {
  86                 if (inst->queue_num == queue_num)
  87                         return inst;
  88         }
  89         return NULL;
  90 }
  91
  92 static struct nfqnl_instance *
  93 instance_create(u_int16_t queue_num, int pid)
  94 {
  95         struct nfqnl_instance *inst;
  96         unsigned int h;
  97         int err;
  98
  99         spin_lock(&instances_lock);
 100         if (instance_lookup(queue_num)) {
 101                 err = -EEXIST;
 102                 goto out_unlock;
 103         }
 104
 105         inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
 106         if (!inst) {
 107                 err = -ENOMEM;
 108                 goto out_unlock;
 109         }
 110
 111         inst->queue_num = queue_num;
 112         inst->peer_pid = pid;
 113         inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
 114         inst->copy_range = 0xfffff;
 115         inst->copy_mode = NFQNL_COPY_NONE;
 116         spin_lock_init(&inst->lock);
 117         INIT_LIST_HEAD(&inst->queue_list);
 118
 119         if (!try_module_get(THIS_MODULE)) {
 120                 err = -EAGAIN;
 121                 goto out_free;
 122         }
 123
 124         h = instance_hashfn(queue_num);
 125         hlist_add_head_rcu(&inst->hlist, &instance_table[h]);
 126
 127         spin_unlock(&instances_lock);
 128
 129         return inst;
 130
 131 out_free:
 132         kfree(inst);
 133 out_unlock:
 134         spin_unlock(&instances_lock);
 135         return ERR_PTR(err);
 136 }
 137
 138 static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
 139                         unsigned long data);
 140
 141 static void
 142 instance_destroy_rcu(struct rcu_head *head)
 143 {
 144         struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
 145                                                    rcu);
 146
 147         nfqnl_flush(inst, NULL, 0);
 148         kfree(inst);
 149         module_put(THIS_MODULE);
 150 }
 151
 152 static void
 153 __instance_destroy(struct nfqnl_instance *inst)
 154 {
 155         hlist_del_rcu(&inst->hlist);
 156         call_rcu(&inst->rcu, instance_destroy_rcu);
 157 }
 158
 159 static void
 160 instance_destroy(struct nfqnl_instance *inst)
 161 {
 162         spin_lock(&instances_lock);
 163         __instance_destroy(inst);
 164         spin_unlock(&instances_lock);
 165 }
 166
 167 static inline void
 168 __enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
 169 {
 170        list_add_tail(&entry->list, &queue->queue_list);
 171        queue->queue_total++;
 172 }
 173
 174 static void
 175 __dequeue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
 176 {
 177         list_del(&entry->list);
 178         queue->queue_total--;
 179 }
 180
 181 static struct nf_queue_entry *
 182 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
 183 {
 184         struct nf_queue_entry *entry = NULL, *i;
 185
 186         spin_lock_bh(&queue->lock);
 187
 188         list_for_each_entry(i, &queue->queue_list, list) {
 189                 if (i->id == id) {
 190                         entry = i;
 191                         break;
 192                 }
 193         }
 194
 195         if (entry)
 196                 __dequeue_entry(queue, entry);
 197
 198         spin_unlock_bh(&queue->lock);
 199
 200         return entry;
 201 }
 202
 203 static void
 204 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
 205 {
 206         struct nf_queue_entry *entry, *next;
 207
 208         spin_lock_bh(&queue->lock);
 209         list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
 210                 if (!cmpfn || cmpfn(entry, data)) {
 211                         list_del(&entry->list);
 212                         queue->queue_total--;
 213                         nf_reinject(entry, NF_DROP);
 214                 }
 215         }
 216         spin_unlock_bh(&queue->lock);
 217 }
 218
 219 static struct sk_buff *
 220 nfqnl_build_packet_message(struct nfqnl_instance *queue,
 221                            struct nf_queue_entry *entry,
 222                            __be32 **packet_id_ptr)
 223 {
 224         sk_buff_data_t old_tail;
 225         size_t size;
 226         size_t data_len = 0;
 227         struct sk_buff *skb;
 228         struct nlattr *nla;
 229         struct nfqnl_msg_packet_hdr *pmsg;
 230         struct nlmsghdr *nlh;
 231         struct nfgenmsg *nfmsg;
 232         struct sk_buff *entskb = entry->skb;
 233         struct net_device *indev;
 234         struct net_device *outdev;
 235
 236         size =    NLMSG_SPACE(sizeof(struct nfgenmsg))
 237                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
 238                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 239                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 240 #ifdef CONFIG_BRIDGE_NETFILTER
 241                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 242                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 243 #endif
 244                 + nla_total_size(sizeof(u_int32_t))     /* mark */
 245                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
 246                 + nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
 247
 248         outdev = entry->outdev;
 249
 250         switch ((enum nfqnl_config_mode)ACCESS_ONCE(queue->copy_mode)) {
 251         case NFQNL_COPY_META:
 252         case NFQNL_COPY_NONE:
 253                 break;
 254
 255         case NFQNL_COPY_PACKET:
 256                 if (entskb->ip_summed == CHECKSUM_PARTIAL &&
 257                     skb_checksum_help(entskb))
 258                         return NULL;
 259
 260                 data_len = ACCESS_ONCE(queue->copy_range);
 261                 if (data_len == 0 || data_len > entskb->len)
 262                         data_len = entskb->len;
 263
 264                 size += nla_total_size(data_len);
 265                 break;
 266         }
 267
 268
 269         skb = alloc_skb(size, GFP_ATOMIC);
 270         if (!skb)
 271                 goto nlmsg_failure;
 272
 273         old_tail = skb->tail;
 274         nlh = NLMSG_PUT(skb, 0, 0,
 275                         NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
 276                         sizeof(struct nfgenmsg));
 277         nfmsg = NLMSG_DATA(nlh);
 278         nfmsg->nfgen_family = entry->pf;
 279         nfmsg->version = NFNETLINK_V0;
 280         nfmsg->res_id = htons(queue->queue_num);
 281
 282         nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
 283         pmsg = nla_data(nla);
 284         pmsg->hw_protocol       = entskb->protocol;
 285         pmsg->hook              = entry->hook;
 286         *packet_id_ptr          = &pmsg->packet_id;
 287
 288         indev = entry->indev;
 289         if (indev) {
 290 #ifndef CONFIG_BRIDGE_NETFILTER
 291                 NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex));
 292 #else
 293                 if (entry->pf == PF_BRIDGE) {
 294                         /* Case 1: indev is physical input device, we need to
 295                          * look for bridge group (when called from
 296                          * netfilter_bridge) */
 297                         NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSINDEV,
 298                                      htonl(indev->ifindex));
 299                         /* this is the bridge group "brX" */
 300                         /* rcu_read_lock()ed by __nf_queue */
 301                         NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV,
 302                                      htonl(br_port_get_rcu(indev)->br->dev->ifindex));
 303                 } else {
 304                         /* Case 2: indev is bridge group, we need to look for
 305                          * physical device (when called from ipv4) */
 306                         NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV,
 307                                      htonl(indev->ifindex));
 308                         if (entskb->nf_bridge && entskb->nf_bridge->physindev)
 309                                 NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSINDEV,
 310                                              htonl(entskb->nf_bridge->physindev->ifindex));
 311                 }
 312 #endif
 313         }
 314
 315         if (outdev) {
 316 #ifndef CONFIG_BRIDGE_NETFILTER
 317                 NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex));
 318 #else
 319                 if (entry->pf == PF_BRIDGE) {
 320                         /* Case 1: outdev is physical output device, we need to
 321                          * look for bridge group (when called from
 322                          * netfilter_bridge) */
 323                         NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 324                                      htonl(outdev->ifindex));
 325                         /* this is the bridge group "brX" */
 326                         /* rcu_read_lock()ed by __nf_queue */
 327                         NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV,
 328                                      htonl(br_port_get_rcu(outdev)->br->dev->ifindex));
 329                 } else {
 330                         /* Case 2: outdev is bridge group, we need to look for
 331                          * physical output device (when called from ipv4) */
 332                         NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV,
 333                                      htonl(outdev->ifindex));
 334                         if (entskb->nf_bridge && entskb->nf_bridge->physoutdev)
 335                                 NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 336                                              htonl(entskb->nf_bridge->physoutdev->ifindex));
 337                 }
 338 #endif
 339         }
 340
 341         if (entskb->mark)
 342                 NLA_PUT_BE32(skb, NFQA_MARK, htonl(entskb->mark));
 343
 344         if (indev && entskb->dev) {
 345                 struct nfqnl_msg_packet_hw phw;
 346                 int len = dev_parse_header(entskb, phw.hw_addr);
 347                 if (len) {
 348                         phw.hw_addrlen = htons(len);
 349                         NLA_PUT(skb, NFQA_HWADDR, sizeof(phw), &phw);
 350                 }
 351         }
 352
 353         if (entskb->tstamp.tv64) {
 354                 struct nfqnl_msg_packet_timestamp ts;
 355                 struct timeval tv = ktime_to_timeval(entskb->tstamp);
 356                 ts.sec = cpu_to_be64(tv.tv_sec);
 357                 ts.usec = cpu_to_be64(tv.tv_usec);
 358
 359                 NLA_PUT(skb, NFQA_TIMESTAMP, sizeof(ts), &ts);
 360         }
 361
 362         if (data_len) {
 363                 struct nlattr *nla;
 364                 int sz = nla_attr_size(data_len);
 365
 366                 if (skb_tailroom(skb) < nla_total_size(data_len)) {
 367                         printk(KERN_WARNING "nf_queue: no tailroom!\n");
 368                         goto nlmsg_failure;
 369                 }
 370
 371                 nla = (struct nlattr *)skb_put(skb, nla_total_size(data_len));
 372                 nla->nla_type = NFQA_PAYLOAD;
 373                 nla->nla_len = sz;
 374
 375                 if (skb_copy_bits(entskb, 0, nla_data(nla), data_len))
 376                         BUG();
 377         }
 378
 379         nlh->nlmsg_len = skb->tail - old_tail;
 380         return skb;
 381
 382 nlmsg_failure:
 383 nla_put_failure:
 384         if (skb)
 385                 kfree_skb(skb);
 386         if (net_ratelimit())
 387                 printk(KERN_ERR "nf_queue: error creating packet message\n");
 388         return NULL;
 389 }
 390
 391 static int
 392 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 393 {
 394         struct sk_buff *nskb;
 395         struct nfqnl_instance *queue;
 396         int err = -ENOBUFS;
 397         __be32 *packet_id_ptr;
 398
 399         /* rcu_read_lock()ed by nf_hook_slow() */
 400         queue = instance_lookup(queuenum);
 401         if (!queue) {
 402                 err = -ESRCH;
 403                 goto err_out;
 404         }
 405
 406         if (queue->copy_mode == NFQNL_COPY_NONE) {
 407                 err = -EINVAL;
 408                 goto err_out;
 409         }
 410
 411         nskb = nfqnl_build_packet_message(queue, entry, &packet_id_ptr);
 412         if (nskb == NULL) {
 413                 err = -ENOMEM;
 414                 goto err_out;
 415         }
 416         spin_lock_bh(&queue->lock);
 417
 418         if (!queue->peer_pid) {
 419                 err = -EINVAL;
 420                 goto err_out_free_nskb;
 421         }
 422         if (queue->queue_total >= queue->queue_maxlen) {
 423                 queue->queue_dropped++;
 424                 if (net_ratelimit())
 425                           printk(KERN_WARNING "nf_queue: full at %d entries, "
 426                                  "dropping packets(s).\n",
 427                                  queue->queue_total);
 428                 goto err_out_free_nskb;
 429         }
 430         entry->id = ++queue->id_sequence;
 431         *packet_id_ptr = htonl(entry->id);
 432
 433         /* nfnetlink_unicast will either free the nskb or add it to a socket */
 434         err = nfnetlink_unicast(nskb, &init_net, queue->peer_pid, MSG_DONTWAIT);
 435         if (err < 0) {
 436                 queue->queue_user_dropped++;
 437                 goto err_out_unlock;
 438         }
 439
 440         __enqueue_entry(queue, entry);
 441
 442         spin_unlock_bh(&queue->lock);
 443         return 0;
 444
 445 err_out_free_nskb:
 446         kfree_skb(nskb);
 447 err_out_unlock:
 448         spin_unlock_bh(&queue->lock);
 449 err_out:
 450         return err;
 451 }
 452
 453 static int
 454 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e)
 455 {
 456         struct sk_buff *nskb;
 457         int diff;
 458
 459         diff = data_len - e->skb->len;
 460         if (diff < 0) {
 461                 if (pskb_trim(e->skb, data_len))
 462                         return -ENOMEM;
 463         } else if (diff > 0) {
 464                 if (data_len > 0xFFFF)
 465                         return -EINVAL;
 466                 if (diff > skb_tailroom(e->skb)) {
 467                         nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
 468                                                diff, GFP_ATOMIC);
 469                         if (!nskb) {
 470                                 printk(KERN_WARNING "nf_queue: OOM "
 471                                       "in mangle, dropping packet\n");
 472                                 return -ENOMEM;
 473                         }
 474                         kfree_skb(e->skb);
 475                         e->skb = nskb;
 476                 }
 477                 skb_put(e->skb, diff);
 478         }
 479         if (!skb_make_writable(e->skb, data_len))
 480                 return -ENOMEM;
 481         skb_copy_to_linear_data(e->skb, data, data_len);
 482         e->skb->ip_summed = CHECKSUM_NONE;
 483         return 0;
 484 }
 485
 486 static int
 487 nfqnl_set_mode(struct nfqnl_instance *queue,
 488                unsigned char mode, unsigned int range)
 489 {
 490         int status = 0;
 491
 492         spin_lock_bh(&queue->lock);
 493         switch (mode) {
 494         case NFQNL_COPY_NONE:
 495         case NFQNL_COPY_META:
 496                 queue->copy_mode = mode;
 497                 queue->copy_range = 0;
 498                 break;
 499
 500         case NFQNL_COPY_PACKET:
 501                 queue->copy_mode = mode;
 502                 /* we're using struct nlattr which has 16bit nla_len */
 503                 if (range > 0xffff)
 504                         queue->copy_range = 0xffff;
 505                 else
 506                         queue->copy_range = range;
 507                 break;
 508
 509         default:
 510                 status = -EINVAL;
 511
 512         }
 513         spin_unlock_bh(&queue->lock);
 514
 515         return status;
 516 }
 517
 518 static int
 519 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
 520 {
 521         if (entry->indev)
 522                 if (entry->indev->ifindex == ifindex)
 523                         return 1;
 524         if (entry->outdev)
 525                 if (entry->outdev->ifindex == ifindex)
 526                         return 1;
 527 #ifdef CONFIG_BRIDGE_NETFILTER
 528         if (entry->skb->nf_bridge) {
 529                 if (entry->skb->nf_bridge->physindev &&
 530                     entry->skb->nf_bridge->physindev->ifindex == ifindex)
 531                         return 1;
 532                 if (entry->skb->nf_bridge->physoutdev &&
 533                     entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
 534                         return 1;
 535         }
 536 #endif
 537         return 0;
 538 }
 539
 540 /* drop all packets with either indev or outdev == ifindex from all queue
 541  * instances */
 542 static void
 543 nfqnl_dev_drop(int ifindex)
 544 {
 545         int i;
 546
 547         rcu_read_lock();
 548
 549         for (i = 0; i < INSTANCE_BUCKETS; i++) {
 550                 struct hlist_node *tmp;
 551                 struct nfqnl_instance *inst;
 552                 struct hlist_head *head = &instance_table[i];
 553
 554                 hlist_for_each_entry_rcu(inst, tmp, head, hlist)
 555                         nfqnl_flush(inst, dev_cmp, ifindex);
 556         }
 557
 558         rcu_read_unlock();
 559 }
 560
 561 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
 562
 563 static int
 564 nfqnl_rcv_dev_event(struct notifier_block *this,
 565                     unsigned long event, void *ptr)
 566 {
 567         struct net_device *dev = ptr;
 568
 569         if (!net_eq(dev_net(dev), &init_net))
 570                 return NOTIFY_DONE;
 571
 572         /* Drop any packets associated with the downed device */
 573         if (event == NETDEV_DOWN)
 574                 nfqnl_dev_drop(dev->ifindex);
 575         return NOTIFY_DONE;
 576 }
 577
 578 static struct notifier_block nfqnl_dev_notifier = {
 579         .notifier_call  = nfqnl_rcv_dev_event,
 580 };
 581
 582 static int
 583 nfqnl_rcv_nl_event(struct notifier_block *this,
 584                    unsigned long event, void *ptr)
 585 {
 586         struct netlink_notify *n = ptr;
 587
 588         if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
 589                 int i;
 590
 591                 /* destroy all instances for this pid */
 592                 spin_lock(&instances_lock);
 593                 for (i = 0; i < INSTANCE_BUCKETS; i++) {
 594                         struct hlist_node *tmp, *t2;
 595                         struct nfqnl_instance *inst;
 596                         struct hlist_head *head = &instance_table[i];
 597
 598                         hlist_for_each_entry_safe(inst, tmp, t2, head, hlist) {
 599                                 if ((n->net == &init_net) &&
 600                                     (n->pid == inst->peer_pid))
 601                                         __instance_destroy(inst);
 602                         }
 603                 }
 604                 spin_unlock(&instances_lock);
 605         }
 606         return NOTIFY_DONE;
 607 }
 608
 609 static struct notifier_block nfqnl_rtnl_notifier = {
 610         .notifier_call  = nfqnl_rcv_nl_event,
 611 };
 612
 613 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
 614         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 615         [NFQA_MARK]             = { .type = NLA_U32 },
 616         [NFQA_PAYLOAD]          = { .type = NLA_UNSPEC },
 617 };
 618
 619 static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
 620         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 621         [NFQA_MARK]             = { .type = NLA_U32 },
 622 };
 623
 624 static struct nfqnl_instance *verdict_instance_lookup(u16 queue_num, int nlpid)
 625 {
 626         struct nfqnl_instance *queue;
 627
 628         queue = instance_lookup(queue_num);
 629         if (!queue)
 630                 return ERR_PTR(-ENODEV);
 631
 632         if (queue->peer_pid != nlpid)
 633                 return ERR_PTR(-EPERM);
 634
 635         return queue;
 636 }
 637
 638 static struct nfqnl_msg_verdict_hdr*
 639 verdicthdr_get(const struct nlattr * const nfqa[])
 640 {
 641         struct nfqnl_msg_verdict_hdr *vhdr;
 642         unsigned int verdict;
 643
 644         if (!nfqa[NFQA_VERDICT_HDR])
 645                 return NULL;
 646
 647         vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
 648         verdict = ntohl(vhdr->verdict);
 649         if ((verdict & NF_VERDICT_MASK) > NF_MAX_VERDICT)
 650                 return NULL;
 651         return vhdr;
 652 }
 653
 654 static int nfq_id_after(unsigned int id, unsigned int max)
 655 {
 656         return (int)(id - max) > 0;
 657 }
 658
 659 static int
 660 nfqnl_recv_verdict_batch(struct sock *ctnl, struct sk_buff *skb,
 661                    const struct nlmsghdr *nlh,
 662                    const struct nlattr * const nfqa[])
 663 {
 664         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 665         struct nf_queue_entry *entry, *tmp;
 666         unsigned int verdict, maxid;
 667         struct nfqnl_msg_verdict_hdr *vhdr;
 668         struct nfqnl_instance *queue;
 669         LIST_HEAD(batch_list);
 670         u16 queue_num = ntohs(nfmsg->res_id);
 671
 672         queue = verdict_instance_lookup(queue_num, NETLINK_CB(skb).pid);
 673         if (IS_ERR(queue))
 674                 return PTR_ERR(queue);
 675
 676         vhdr = verdicthdr_get(nfqa);
 677         if (!vhdr)
 678                 return -EINVAL;
 679
 680         verdict = ntohl(vhdr->verdict);
 681         maxid = ntohl(vhdr->id);
 682
 683         spin_lock_bh(&queue->lock);
 684
 685         list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
 686                 if (nfq_id_after(entry->id, maxid))
 687                         break;
 688                 __dequeue_entry(queue, entry);
 689                 list_add_tail(&entry->list, &batch_list);
 690         }
 691
 692         spin_unlock_bh(&queue->lock);
 693
 694         if (list_empty(&batch_list))
 695                 return -ENOENT;
 696
 697         list_for_each_entry_safe(entry, tmp, &batch_list, list) {
 698                 if (nfqa[NFQA_MARK])
 699                         entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
 700                 nf_reinject(entry, verdict);
 701         }
 702         return 0;
 703 }
 704
 705 static int
 706 nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
 707                    const struct nlmsghdr *nlh,
 708                    const struct nlattr * const nfqa[])
 709 {
 710         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 711         u_int16_t queue_num = ntohs(nfmsg->res_id);
 712
 713         struct nfqnl_msg_verdict_hdr *vhdr;
 714         struct nfqnl_instance *queue;
 715         unsigned int verdict;
 716         struct nf_queue_entry *entry;
 717
 718         queue = instance_lookup(queue_num);
 719         if (!queue)
 720
 721         queue = verdict_instance_lookup(queue_num, NETLINK_CB(skb).pid);
 722         if (IS_ERR(queue))
 723                 return PTR_ERR(queue);
 724
 725         vhdr = verdicthdr_get(nfqa);
 726         if (!vhdr)
 727                 return -EINVAL;
 728
 729         verdict = ntohl(vhdr->verdict);
 730
 731         entry = find_dequeue_entry(queue, ntohl(vhdr->id));
 732         if (entry == NULL)
 733                 return -ENOENT;
 734
 735         if (nfqa[NFQA_PAYLOAD]) {
 736                 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
 737                                  nla_len(nfqa[NFQA_PAYLOAD]), entry) < 0)
 738                         verdict = NF_DROP;
 739         }
 740
 741         if (nfqa[NFQA_MARK])
 742                 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
 743
 744         nf_reinject(entry, verdict);
 745         return 0;
 746 }
 747
 748 static int
 749 nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
 750                   const struct nlmsghdr *nlh,
 751                   const struct nlattr * const nfqa[])
 752 {
 753         return -ENOTSUPP;
 754 }
 755
 756 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
 757         [NFQA_CFG_CMD]          = { .len = sizeof(struct nfqnl_msg_config_cmd) },
 758         [NFQA_CFG_PARAMS]       = { .len = sizeof(struct nfqnl_msg_config_params) },
 759 };
 760
 761 static const struct nf_queue_handler nfqh = {
 762         .name   = "nf_queue",
 763         .outfn  = &nfqnl_enqueue_packet,
 764 };
 765
 766 static int
 767 nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
 768                   const struct nlmsghdr *nlh,
 769                   const struct nlattr * const nfqa[])
 770 {
 771         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 772         u_int16_t queue_num = ntohs(nfmsg->res_id);
 773         struct nfqnl_instance *queue;
 774         struct nfqnl_msg_config_cmd *cmd = NULL;
 775         int ret = 0;
 776
 777         if (nfqa[NFQA_CFG_CMD]) {
 778                 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
 779
 780                 /* Commands without queue context - might sleep */
 781                 switch (cmd->command) {
 782                 case NFQNL_CFG_CMD_PF_BIND:
 783                         return nf_register_queue_handler(ntohs(cmd->pf),
 784                                                          &nfqh);
 785                 case NFQNL_CFG_CMD_PF_UNBIND:
 786                         return nf_unregister_queue_handler(ntohs(cmd->pf),
 787                                                            &nfqh);
 788                 }
 789         }
 790
 791         rcu_read_lock();
 792         queue = instance_lookup(queue_num);
 793         if (queue && queue->peer_pid != NETLINK_CB(skb).pid) {
 794                 ret = -EPERM;
 795                 goto err_out_unlock;
 796         }
 797
 798         if (cmd != NULL) {
 799                 switch (cmd->command) {
 800                 case NFQNL_CFG_CMD_BIND:
 801                         if (queue) {
 802                                 ret = -EBUSY;
 803                                 goto err_out_unlock;
 804                         }
 805                         queue = instance_create(queue_num, NETLINK_CB(skb).pid);
 806                         if (IS_ERR(queue)) {
 807                                 ret = PTR_ERR(queue);
 808                                 goto err_out_unlock;
 809                         }
 810                         break;
 811                 case NFQNL_CFG_CMD_UNBIND:
 812                         if (!queue) {
 813                                 ret = -ENODEV;
 814                                 goto err_out_unlock;
 815                         }
 816                         instance_destroy(queue);
 817                         break;
 818                 case NFQNL_CFG_CMD_PF_BIND:
 819                 case NFQNL_CFG_CMD_PF_UNBIND:
 820                         break;
 821                 default:
 822                         ret = -ENOTSUPP;
 823                         break;
 824                 }
 825         }
 826
 827         if (nfqa[NFQA_CFG_PARAMS]) {
 828                 struct nfqnl_msg_config_params *params;
 829
 830                 if (!queue) {
 831                         ret = -ENODEV;
 832                         goto err_out_unlock;
 833                 }
 834                 params = nla_data(nfqa[NFQA_CFG_PARAMS]);
 835                 nfqnl_set_mode(queue, params->copy_mode,
 836                                 ntohl(params->copy_range));
 837         }
 838
 839         if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
 840                 __be32 *queue_maxlen;
 841
 842                 if (!queue) {
 843                         ret = -ENODEV;
 844                         goto err_out_unlock;
 845                 }
 846                 queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
 847                 spin_lock_bh(&queue->lock);
 848                 queue->queue_maxlen = ntohl(*queue_maxlen);
 849                 spin_unlock_bh(&queue->lock);
 850         }
 851
 852 err_out_unlock:
 853         rcu_read_unlock();
 854         return ret;
 855 }
 856
 857 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
 858         [NFQNL_MSG_PACKET]      = { .call_rcu = nfqnl_recv_unsupp,
 859                                     .attr_count = NFQA_MAX, },
 860         [NFQNL_MSG_VERDICT]     = { .call_rcu = nfqnl_recv_verdict,
 861                                     .attr_count = NFQA_MAX,
 862                                     .policy = nfqa_verdict_policy },
 863         [NFQNL_MSG_CONFIG]      = { .call = nfqnl_recv_config,
 864                                     .attr_count = NFQA_CFG_MAX,
 865                                     .policy = nfqa_cfg_policy },
 866         [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
 867                                     .attr_count = NFQA_MAX,
 868                                     .policy = nfqa_verdict_batch_policy },
 869 };
 870
 871 static const struct nfnetlink_subsystem nfqnl_subsys = {
 872         .name           = "nf_queue",
 873         .subsys_id      = NFNL_SUBSYS_QUEUE,
 874         .cb_count       = NFQNL_MSG_MAX,
 875         .cb             = nfqnl_cb,
 876 };
 877
 878 #ifdef CONFIG_PROC_FS
 879 struct iter_state {
 880         unsigned int bucket;
 881 };
 882
 883 static struct hlist_node *get_first(struct seq_file *seq)
 884 {
 885         struct iter_state *st = seq->private;
 886
 887         if (!st)
 888                 return NULL;
 889
 890         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
 891                 if (!hlist_empty(&instance_table[st->bucket]))
 892                         return instance_table[st->bucket].first;
 893         }
 894         return NULL;
 895 }
 896
 897 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
 898 {
 899         struct iter_state *st = seq->private;
 900
 901         h = h->next;
 902         while (!h) {
 903                 if (++st->bucket >= INSTANCE_BUCKETS)
 904                         return NULL;
 905
 906                 h = instance_table[st->bucket].first;
 907         }
 908         return h;
 909 }
 910
 911 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
 912 {
 913         struct hlist_node *head;
 914         head = get_first(seq);
 915
 916         if (head)
 917                 while (pos && (head = get_next(seq, head)))
 918                         pos--;
 919         return pos ? NULL : head;
 920 }
 921
 922 static void *seq_start(struct seq_file *seq, loff_t *pos)
 923         __acquires(instances_lock)
 924 {
 925         spin_lock(&instances_lock);
 926         return get_idx(seq, *pos);
 927 }
 928
 929 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
 930 {
 931         (*pos)++;
 932         return get_next(s, v);
 933 }
 934
 935 static void seq_stop(struct seq_file *s, void *v)
 936         __releases(instances_lock)
 937 {
 938         spin_unlock(&instances_lock);
 939 }
 940
 941 static int seq_show(struct seq_file *s, void *v)
 942 {
 943         const struct nfqnl_instance *inst = v;
 944
 945         return seq_printf(s, "%5d %6d %5d %1d %5d %5d %5d %8d %2d\n",
 946                           inst->queue_num,
 947                           inst->peer_pid, inst->queue_total,
 948                           inst->copy_mode, inst->copy_range,
 949                           inst->queue_dropped, inst->queue_user_dropped,
 950                           inst->id_sequence, 1);
 951 }
 952
 953 static const struct seq_operations nfqnl_seq_ops = {
 954         .start  = seq_start,
 955         .next   = seq_next,
 956         .stop   = seq_stop,
 957         .show   = seq_show,
 958 };
 959
 960 static int nfqnl_open(struct inode *inode, struct file *file)
 961 {
 962         return seq_open_private(file, &nfqnl_seq_ops,
 963                         sizeof(struct iter_state));
 964 }
 965
 966 static const struct file_operations nfqnl_file_ops = {
 967         .owner   = THIS_MODULE,
 968         .open    = nfqnl_open,
 969         .read    = seq_read,
 970         .llseek  = seq_lseek,
 971         .release = seq_release_private,
 972 };
 973
 974 #endif /* PROC_FS */
 975
 976 static int __init nfnetlink_queue_init(void)
 977 {
 978         int i, status = -ENOMEM;
 979
 980         for (i = 0; i < INSTANCE_BUCKETS; i++)
 981                 INIT_HLIST_HEAD(&instance_table[i]);
 982
 983         netlink_register_notifier(&nfqnl_rtnl_notifier);
 984         status = nfnetlink_subsys_register(&nfqnl_subsys);
 985         if (status < 0) {
 986                 printk(KERN_ERR "nf_queue: failed to create netlink socket\n");
 987                 goto cleanup_netlink_notifier;
 988         }
 989
 990 #ifdef CONFIG_PROC_FS
 991         if (!proc_create("nfnetlink_queue", 0440,
 992                          proc_net_netfilter, &nfqnl_file_ops))
 993                 goto cleanup_subsys;
 994 #endif
 995
 996         register_netdevice_notifier(&nfqnl_dev_notifier);
 997         return status;
 998
 999 #ifdef CONFIG_PROC_FS
1000 cleanup_subsys:
1001         nfnetlink_subsys_unregister(&nfqnl_subsys);
1002 #endif
1003 cleanup_netlink_notifier:
1004         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1005         return status;
1006 }
1007
1008 static void __exit nfnetlink_queue_fini(void)
1009 {
1010         nf_unregister_queue_handlers(&nfqh);
1011         unregister_netdevice_notifier(&nfqnl_dev_notifier);
1012 #ifdef CONFIG_PROC_FS
1013         remove_proc_entry("nfnetlink_queue", proc_net_netfilter);
1014 #endif
1015         nfnetlink_subsys_unregister(&nfqnl_subsys);
1016         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1017
1018         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1019 }
1020
1021 MODULE_DESCRIPTION("netfilter packet queue handler");
1022 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1023 MODULE_LICENSE("GPL");
1024 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1025
1026 module_init(nfnetlink_queue_init);
1027 module_exit(nfnetlink_queue_fini);