net/netfilter/nfnetlink_queue.c

   1 /*
   2  * This is a module which is used for queueing packets and communicating with
   3  * userspace via nfetlink.
   4  *
   5  * (C) 2005 by Harald Welte <laforge@netfilter.org>
   6  *
   7  * Based on the old ipv4-only ip_queue.c:
   8  * (C) 2000-2002 James Morris <jmorris@intercode.com.au>
   9  * (C) 2003-2005 Netfilter Core Team <coreteam@netfilter.org>
  10  *
  11  * This program is free software; you can redistribute it and/or modify
  12  * it under the terms of the GNU General Public License version 2 as
  13  * published by the Free Software Foundation.
  14  *
  15  */
  16 #include <linux/module.h>
  17 #include <linux/skbuff.h>
  18 #include <linux/init.h>
  19 #include <linux/spinlock.h>
  20 #include <linux/notifier.h>
  21 #include <linux/netdevice.h>
  22 #include <linux/netfilter.h>
  23 #include <linux/proc_fs.h>
  24 #include <linux/netfilter_ipv4.h>
  25 #include <linux/netfilter_ipv6.h>
  26 #include <linux/netfilter/nfnetlink.h>
  27 #include <linux/netfilter/nfnetlink_queue.h>
  28 #include <linux/list.h>
  29 #include <net/sock.h>
  30 #include <net/netfilter/nf_queue.h>
  31
  32 #include <asm/atomic.h>
  33
  34 #ifdef CONFIG_BRIDGE_NETFILTER
  35 #include "../bridge/br_private.h"
  36 #endif
  37
  38 #define NFQNL_QMAX_DEFAULT 1024
  39
  40 struct nfqnl_instance {
  41         struct hlist_node hlist;                /* global list of queues */
  42         struct rcu_head rcu;
  43
  44         int peer_pid;
  45         unsigned int queue_maxlen;
  46         unsigned int copy_range;
  47         unsigned int queue_total;
  48         unsigned int queue_dropped;
  49         unsigned int queue_user_dropped;
  50
  51         unsigned int id_sequence;               /* 'sequence' of pkt ids */
  52
  53         u_int16_t queue_num;                    /* number of this queue */
  54         u_int8_t copy_mode;
  55
  56         spinlock_t lock;
  57
  58         struct list_head queue_list;            /* packets in queue */
  59 };
  60
  61 typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
  62
  63 static DEFINE_SPINLOCK(instances_lock);
  64
  65 #define INSTANCE_BUCKETS        16
  66 static struct hlist_head instance_table[INSTANCE_BUCKETS] __read_mostly;
  67
  68 static inline u_int8_t instance_hashfn(u_int16_t queue_num)
  69 {
  70         return ((queue_num >> 8) | queue_num) % INSTANCE_BUCKETS;
  71 }
  72
  73 static struct nfqnl_instance *
  74 instance_lookup(u_int16_t queue_num)
  75 {
  76         struct hlist_head *head;
  77         struct hlist_node *pos;
  78         struct nfqnl_instance *inst;
  79
  80         head = &instance_table[instance_hashfn(queue_num)];
  81         hlist_for_each_entry_rcu(inst, pos, head, hlist) {
  82                 if (inst->queue_num == queue_num)
  83                         return inst;
  84         }
  85         return NULL;
  86 }
  87
  88 static struct nfqnl_instance *
  89 instance_create(u_int16_t queue_num, int pid)
  90 {
  91         struct nfqnl_instance *inst = NULL;
  92         unsigned int h;
  93
  94         spin_lock(&instances_lock);
  95         if (instance_lookup(queue_num))
  96                 goto out_unlock;
  97
  98         inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
  99         if (!inst)
 100                 goto out_unlock;
 101
 102         inst->queue_num = queue_num;
 103         inst->peer_pid = pid;
 104         inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
 105         inst->copy_range = 0xfffff;
 106         inst->copy_mode = NFQNL_COPY_NONE;
 107         spin_lock_init(&inst->lock);
 108         INIT_LIST_HEAD(&inst->queue_list);
 109         INIT_RCU_HEAD(&inst->rcu);
 110
 111         if (!try_module_get(THIS_MODULE))
 112                 goto out_free;
 113
 114         h = instance_hashfn(queue_num);
 115         hlist_add_head_rcu(&inst->hlist, &instance_table[h]);
 116
 117         spin_unlock(&instances_lock);
 118
 119         return inst;
 120
 121 out_free:
 122         kfree(inst);
 123 out_unlock:
 124         spin_unlock(&instances_lock);
 125         return NULL;
 126 }
 127
 128 static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
 129                         unsigned long data);
 130
 131 static void
 132 instance_destroy_rcu(struct rcu_head *head)
 133 {
 134         struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
 135                                                    rcu);
 136
 137         nfqnl_flush(inst, NULL, 0);
 138         kfree(inst);
 139         module_put(THIS_MODULE);
 140 }
 141
 142 static void
 143 __instance_destroy(struct nfqnl_instance *inst)
 144 {
 145         hlist_del_rcu(&inst->hlist);
 146         call_rcu(&inst->rcu, instance_destroy_rcu);
 147 }
 148
 149 static void
 150 instance_destroy(struct nfqnl_instance *inst)
 151 {
 152         spin_lock(&instances_lock);
 153         __instance_destroy(inst);
 154         spin_unlock(&instances_lock);
 155 }
 156
 157 static inline void
 158 __enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
 159 {
 160        list_add_tail(&entry->list, &queue->queue_list);
 161        queue->queue_total++;
 162 }
 163
 164 static struct nf_queue_entry *
 165 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
 166 {
 167         struct nf_queue_entry *entry = NULL, *i;
 168
 169         spin_lock_bh(&queue->lock);
 170
 171         list_for_each_entry(i, &queue->queue_list, list) {
 172                 if (i->id == id) {
 173                         entry = i;
 174                         break;
 175                 }
 176         }
 177
 178         if (entry) {
 179                 list_del(&entry->list);
 180                 queue->queue_total--;
 181         }
 182
 183         spin_unlock_bh(&queue->lock);
 184
 185         return entry;
 186 }
 187
 188 static void
 189 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
 190 {
 191         struct nf_queue_entry *entry, *next;
 192
 193         spin_lock_bh(&queue->lock);
 194         list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
 195                 if (!cmpfn || cmpfn(entry, data)) {
 196                         list_del(&entry->list);
 197                         queue->queue_total--;
 198                         nf_reinject(entry, NF_DROP);
 199                 }
 200         }
 201         spin_unlock_bh(&queue->lock);
 202 }
 203
 204 static struct sk_buff *
 205 nfqnl_build_packet_message(struct nfqnl_instance *queue,
 206                            struct nf_queue_entry *entry, int *errp)
 207 {
 208         sk_buff_data_t old_tail;
 209         size_t size;
 210         size_t data_len = 0;
 211         struct sk_buff *skb;
 212         struct nfqnl_msg_packet_hdr pmsg;
 213         struct nlmsghdr *nlh;
 214         struct nfgenmsg *nfmsg;
 215         struct sk_buff *entskb = entry->skb;
 216         struct net_device *indev;
 217         struct net_device *outdev;
 218
 219         size =    NLMSG_ALIGN(sizeof(struct nfgenmsg))
 220                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
 221                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 222                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 223 #ifdef CONFIG_BRIDGE_NETFILTER
 224                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 225                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 226 #endif
 227                 + nla_total_size(sizeof(u_int32_t))     /* mark */
 228                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
 229                 + nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
 230
 231         outdev = entry->outdev;
 232
 233         spin_lock_bh(&queue->lock);
 234
 235         switch ((enum nfqnl_config_mode)queue->copy_mode) {
 236         case NFQNL_COPY_META:
 237         case NFQNL_COPY_NONE:
 238                 data_len = 0;
 239                 break;
 240
 241         case NFQNL_COPY_PACKET:
 242                 if ((entskb->ip_summed == CHECKSUM_PARTIAL ||
 243                      entskb->ip_summed == CHECKSUM_COMPLETE) &&
 244                     (*errp = skb_checksum_help(entskb))) {
 245                         spin_unlock_bh(&queue->lock);
 246                         return NULL;
 247                 }
 248                 if (queue->copy_range == 0
 249                     || queue->copy_range > entskb->len)
 250                         data_len = entskb->len;
 251                 else
 252                         data_len = queue->copy_range;
 253
 254                 size += nla_total_size(data_len);
 255                 break;
 256         }
 257
 258         entry->id = queue->id_sequence++;
 259
 260         spin_unlock_bh(&queue->lock);
 261
 262         skb = alloc_skb(size, GFP_ATOMIC);
 263         if (!skb)
 264                 goto nlmsg_failure;
 265
 266         old_tail = skb->tail;
 267         nlh = NLMSG_PUT(skb, 0, 0,
 268                         NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
 269                         sizeof(struct nfgenmsg));
 270         nfmsg = NLMSG_DATA(nlh);
 271         nfmsg->nfgen_family = entry->pf;
 272         nfmsg->version = NFNETLINK_V0;
 273         nfmsg->res_id = htons(queue->queue_num);
 274
 275         pmsg.packet_id          = htonl(entry->id);
 276         pmsg.hw_protocol        = entskb->protocol;
 277         pmsg.hook               = entry->hook;
 278
 279         NLA_PUT(skb, NFQA_PACKET_HDR, sizeof(pmsg), &pmsg);
 280
 281         indev = entry->indev;
 282         if (indev) {
 283 #ifndef CONFIG_BRIDGE_NETFILTER
 284                 NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex));
 285 #else
 286                 if (entry->pf == PF_BRIDGE) {
 287                         /* Case 1: indev is physical input device, we need to
 288                          * look for bridge group (when called from
 289                          * netfilter_bridge) */
 290                         NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSINDEV,
 291                                      htonl(indev->ifindex));
 292                         /* this is the bridge group "brX" */
 293                         NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV,
 294                                      htonl(indev->br_port->br->dev->ifindex));
 295                 } else {
 296                         /* Case 2: indev is bridge group, we need to look for
 297                          * physical device (when called from ipv4) */
 298                         NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV,
 299                                      htonl(indev->ifindex));
 300                         if (entskb->nf_bridge && entskb->nf_bridge->physindev)
 301                                 NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSINDEV,
 302                                              htonl(entskb->nf_bridge->physindev->ifindex));
 303                 }
 304 #endif
 305         }
 306
 307         if (outdev) {
 308 #ifndef CONFIG_BRIDGE_NETFILTER
 309                 NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex));
 310 #else
 311                 if (entry->pf == PF_BRIDGE) {
 312                         /* Case 1: outdev is physical output device, we need to
 313                          * look for bridge group (when called from
 314                          * netfilter_bridge) */
 315                         NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 316                                      htonl(outdev->ifindex));
 317                         /* this is the bridge group "brX" */
 318                         NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV,
 319                                      htonl(outdev->br_port->br->dev->ifindex));
 320                 } else {
 321                         /* Case 2: outdev is bridge group, we need to look for
 322                          * physical output device (when called from ipv4) */
 323                         NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV,
 324                                      htonl(outdev->ifindex));
 325                         if (entskb->nf_bridge && entskb->nf_bridge->physoutdev)
 326                                 NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 327                                              htonl(entskb->nf_bridge->physoutdev->ifindex));
 328                 }
 329 #endif
 330         }
 331
 332         if (entskb->mark)
 333                 NLA_PUT_BE32(skb, NFQA_MARK, htonl(entskb->mark));
 334
 335         if (indev && entskb->dev) {
 336                 struct nfqnl_msg_packet_hw phw;
 337                 int len = dev_parse_header(entskb, phw.hw_addr);
 338                 if (len) {
 339                         phw.hw_addrlen = htons(len);
 340                         NLA_PUT(skb, NFQA_HWADDR, sizeof(phw), &phw);
 341                 }
 342         }
 343
 344         if (entskb->tstamp.tv64) {
 345                 struct nfqnl_msg_packet_timestamp ts;
 346                 struct timeval tv = ktime_to_timeval(entskb->tstamp);
 347                 ts.sec = cpu_to_be64(tv.tv_sec);
 348                 ts.usec = cpu_to_be64(tv.tv_usec);
 349
 350                 NLA_PUT(skb, NFQA_TIMESTAMP, sizeof(ts), &ts);
 351         }
 352
 353         if (data_len) {
 354                 struct nlattr *nla;
 355                 int size = nla_attr_size(data_len);
 356
 357                 if (skb_tailroom(skb) < nla_total_size(data_len)) {
 358                         printk(KERN_WARNING "nf_queue: no tailroom!\n");
 359                         goto nlmsg_failure;
 360                 }
 361
 362                 nla = (struct nlattr *)skb_put(skb, nla_total_size(data_len));
 363                 nla->nla_type = NFQA_PAYLOAD;
 364                 nla->nla_len = size;
 365
 366                 if (skb_copy_bits(entskb, 0, nla_data(nla), data_len))
 367                         BUG();
 368         }
 369
 370         nlh->nlmsg_len = skb->tail - old_tail;
 371         return skb;
 372
 373 nlmsg_failure:
 374 nla_put_failure:
 375         if (skb)
 376                 kfree_skb(skb);
 377         *errp = -EINVAL;
 378         if (net_ratelimit())
 379                 printk(KERN_ERR "nf_queue: error creating packet message\n");
 380         return NULL;
 381 }
 382
 383 static int
 384 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 385 {
 386         int status = -EINVAL;
 387         struct sk_buff *nskb;
 388         struct nfqnl_instance *queue;
 389
 390         /* rcu_read_lock()ed by nf_hook_slow() */
 391         queue = instance_lookup(queuenum);
 392         if (!queue)
 393                 return -EINVAL;
 394
 395         if (queue->copy_mode == NFQNL_COPY_NONE)
 396                 return -EAGAIN;
 397
 398         nskb = nfqnl_build_packet_message(queue, entry, &status);
 399         if (nskb == NULL)
 400                 return status;
 401
 402         spin_lock_bh(&queue->lock);
 403
 404         if (!queue->peer_pid)
 405                 goto err_out_free_nskb;
 406
 407         if (queue->queue_total >= queue->queue_maxlen) {
 408                 queue->queue_dropped++;
 409                 status = -ENOSPC;
 410                 if (net_ratelimit())
 411                           printk(KERN_WARNING "nf_queue: full at %d entries, "
 412                                  "dropping packets(s). Dropped: %d\n",
 413                                  queue->queue_total, queue->queue_dropped);
 414                 goto err_out_free_nskb;
 415         }
 416
 417         /* nfnetlink_unicast will either free the nskb or add it to a socket */
 418         status = nfnetlink_unicast(nskb, queue->peer_pid, MSG_DONTWAIT);
 419         if (status < 0) {
 420                 queue->queue_user_dropped++;
 421                 goto err_out_unlock;
 422         }
 423
 424         __enqueue_entry(queue, entry);
 425
 426         spin_unlock_bh(&queue->lock);
 427         return status;
 428
 429 err_out_free_nskb:
 430         kfree_skb(nskb);
 431
 432 err_out_unlock:
 433         spin_unlock_bh(&queue->lock);
 434         return status;
 435 }
 436
 437 static int
 438 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e)
 439 {
 440         int diff;
 441         int err;
 442
 443         diff = data_len - e->skb->len;
 444         if (diff < 0) {
 445                 if (pskb_trim(e->skb, data_len))
 446                         return -ENOMEM;
 447         } else if (diff > 0) {
 448                 if (data_len > 0xFFFF)
 449                         return -EINVAL;
 450                 if (diff > skb_tailroom(e->skb)) {
 451                         err = pskb_expand_head(e->skb, 0,
 452                                                diff - skb_tailroom(e->skb),
 453                                                GFP_ATOMIC);
 454                         if (err) {
 455                                 printk(KERN_WARNING "nf_queue: OOM "
 456                                       "in mangle, dropping packet\n");
 457                                 return err;
 458                         }
 459                 }
 460                 skb_put(e->skb, diff);
 461         }
 462         if (!skb_make_writable(e->skb, data_len))
 463                 return -ENOMEM;
 464         skb_copy_to_linear_data(e->skb, data, data_len);
 465         e->skb->ip_summed = CHECKSUM_NONE;
 466         return 0;
 467 }
 468
 469 static int
 470 nfqnl_set_mode(struct nfqnl_instance *queue,
 471                unsigned char mode, unsigned int range)
 472 {
 473         int status = 0;
 474
 475         spin_lock_bh(&queue->lock);
 476         switch (mode) {
 477         case NFQNL_COPY_NONE:
 478         case NFQNL_COPY_META:
 479                 queue->copy_mode = mode;
 480                 queue->copy_range = 0;
 481                 break;
 482
 483         case NFQNL_COPY_PACKET:
 484                 queue->copy_mode = mode;
 485                 /* we're using struct nlattr which has 16bit nla_len */
 486                 if (range > 0xffff)
 487                         queue->copy_range = 0xffff;
 488                 else
 489                         queue->copy_range = range;
 490                 break;
 491
 492         default:
 493                 status = -EINVAL;
 494
 495         }
 496         spin_unlock_bh(&queue->lock);
 497
 498         return status;
 499 }
 500
 501 static int
 502 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
 503 {
 504         if (entry->indev)
 505                 if (entry->indev->ifindex == ifindex)
 506                         return 1;
 507         if (entry->outdev)
 508                 if (entry->outdev->ifindex == ifindex)
 509                         return 1;
 510 #ifdef CONFIG_BRIDGE_NETFILTER
 511         if (entry->skb->nf_bridge) {
 512                 if (entry->skb->nf_bridge->physindev &&
 513                     entry->skb->nf_bridge->physindev->ifindex == ifindex)
 514                         return 1;
 515                 if (entry->skb->nf_bridge->physoutdev &&
 516                     entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
 517                         return 1;
 518         }
 519 #endif
 520         return 0;
 521 }
 522
 523 /* drop all packets with either indev or outdev == ifindex from all queue
 524  * instances */
 525 static void
 526 nfqnl_dev_drop(int ifindex)
 527 {
 528         int i;
 529
 530         rcu_read_lock();
 531
 532         for (i = 0; i < INSTANCE_BUCKETS; i++) {
 533                 struct hlist_node *tmp;
 534                 struct nfqnl_instance *inst;
 535                 struct hlist_head *head = &instance_table[i];
 536
 537                 hlist_for_each_entry_rcu(inst, tmp, head, hlist)
 538                         nfqnl_flush(inst, dev_cmp, ifindex);
 539         }
 540
 541         rcu_read_unlock();
 542 }
 543
 544 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
 545
 546 static int
 547 nfqnl_rcv_dev_event(struct notifier_block *this,
 548                     unsigned long event, void *ptr)
 549 {
 550         struct net_device *dev = ptr;
 551
 552         if (dev->nd_net != &init_net)
 553                 return NOTIFY_DONE;
 554
 555         /* Drop any packets associated with the downed device */
 556         if (event == NETDEV_DOWN)
 557                 nfqnl_dev_drop(dev->ifindex);
 558         return NOTIFY_DONE;
 559 }
 560
 561 static struct notifier_block nfqnl_dev_notifier = {
 562         .notifier_call  = nfqnl_rcv_dev_event,
 563 };
 564
 565 static int
 566 nfqnl_rcv_nl_event(struct notifier_block *this,
 567                    unsigned long event, void *ptr)
 568 {
 569         struct netlink_notify *n = ptr;
 570
 571         if (event == NETLINK_URELEASE &&
 572             n->protocol == NETLINK_NETFILTER && n->pid) {
 573                 int i;
 574
 575                 /* destroy all instances for this pid */
 576                 spin_lock(&instances_lock);
 577                 for (i = 0; i < INSTANCE_BUCKETS; i++) {
 578                         struct hlist_node *tmp, *t2;
 579                         struct nfqnl_instance *inst;
 580                         struct hlist_head *head = &instance_table[i];
 581
 582                         hlist_for_each_entry_safe(inst, tmp, t2, head, hlist) {
 583                                 if ((n->net == &init_net) &&
 584                                     (n->pid == inst->peer_pid))
 585                                         __instance_destroy(inst);
 586                         }
 587                 }
 588                 spin_unlock(&instances_lock);
 589         }
 590         return NOTIFY_DONE;
 591 }
 592
 593 static struct notifier_block nfqnl_rtnl_notifier = {
 594         .notifier_call  = nfqnl_rcv_nl_event,
 595 };
 596
 597 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
 598         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 599         [NFQA_MARK]             = { .type = NLA_U32 },
 600         [NFQA_PAYLOAD]          = { .type = NLA_UNSPEC },
 601 };
 602
 603 static int
 604 nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
 605                    struct nlmsghdr *nlh, struct nlattr *nfqa[])
 606 {
 607         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 608         u_int16_t queue_num = ntohs(nfmsg->res_id);
 609
 610         struct nfqnl_msg_verdict_hdr *vhdr;
 611         struct nfqnl_instance *queue;
 612         unsigned int verdict;
 613         struct nf_queue_entry *entry;
 614         int err;
 615
 616         rcu_read_lock();
 617         queue = instance_lookup(queue_num);
 618         if (!queue) {
 619                 err = -ENODEV;
 620                 goto err_out_unlock;
 621         }
 622
 623         if (queue->peer_pid != NETLINK_CB(skb).pid) {
 624                 err = -EPERM;
 625                 goto err_out_unlock;
 626         }
 627
 628         if (!nfqa[NFQA_VERDICT_HDR]) {
 629                 err = -EINVAL;
 630                 goto err_out_unlock;
 631         }
 632
 633         vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
 634         verdict = ntohl(vhdr->verdict);
 635
 636         if ((verdict & NF_VERDICT_MASK) > NF_MAX_VERDICT) {
 637                 err = -EINVAL;
 638                 goto err_out_unlock;
 639         }
 640
 641         entry = find_dequeue_entry(queue, ntohl(vhdr->id));
 642         if (entry == NULL) {
 643                 err = -ENOENT;
 644                 goto err_out_unlock;
 645         }
 646         rcu_read_unlock();
 647
 648         if (nfqa[NFQA_PAYLOAD]) {
 649                 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
 650                                  nla_len(nfqa[NFQA_PAYLOAD]), entry) < 0)
 651                         verdict = NF_DROP;
 652         }
 653
 654         if (nfqa[NFQA_MARK])
 655                 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
 656
 657         nf_reinject(entry, verdict);
 658         return 0;
 659
 660 err_out_unlock:
 661         rcu_read_unlock();
 662         return err;
 663 }
 664
 665 static int
 666 nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
 667                   struct nlmsghdr *nlh, struct nlattr *nfqa[])
 668 {
 669         return -ENOTSUPP;
 670 }
 671
 672 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
 673         [NFQA_CFG_CMD]          = { .len = sizeof(struct nfqnl_msg_config_cmd) },
 674         [NFQA_CFG_PARAMS]       = { .len = sizeof(struct nfqnl_msg_config_params) },
 675 };
 676
 677 static const struct nf_queue_handler nfqh = {
 678         .name   = "nf_queue",
 679         .outfn  = &nfqnl_enqueue_packet,
 680 };
 681
 682 static int
 683 nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
 684                   struct nlmsghdr *nlh, struct nlattr *nfqa[])
 685 {
 686         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 687         u_int16_t queue_num = ntohs(nfmsg->res_id);
 688         struct nfqnl_instance *queue;
 689         struct nfqnl_msg_config_cmd *cmd = NULL;
 690         int ret = 0;
 691
 692         if (nfqa[NFQA_CFG_CMD]) {
 693                 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
 694
 695                 /* Commands without queue context - might sleep */
 696                 switch (cmd->command) {
 697                 case NFQNL_CFG_CMD_PF_BIND:
 698                         ret = nf_register_queue_handler(ntohs(cmd->pf),
 699                                                         &nfqh);
 700                         break;
 701                 case NFQNL_CFG_CMD_PF_UNBIND:
 702                         ret = nf_unregister_queue_handler(ntohs(cmd->pf),
 703                                                           &nfqh);
 704                         break;
 705                 default:
 706                         break;
 707                 }
 708
 709                 if (ret < 0)
 710                         return ret;
 711         }
 712
 713         rcu_read_lock();
 714         queue = instance_lookup(queue_num);
 715         if (queue && queue->peer_pid != NETLINK_CB(skb).pid) {
 716                 ret = -EPERM;
 717                 goto err_out_unlock;
 718         }
 719
 720         if (cmd != NULL) {
 721                 switch (cmd->command) {
 722                 case NFQNL_CFG_CMD_BIND:
 723                         if (queue) {
 724                                 ret = -EBUSY;
 725                                 goto err_out_unlock;
 726                         }
 727                         queue = instance_create(queue_num, NETLINK_CB(skb).pid);
 728                         if (!queue) {
 729                                 ret = -EINVAL;
 730                                 goto err_out_unlock;
 731                         }
 732                         break;
 733                 case NFQNL_CFG_CMD_UNBIND:
 734                         if (!queue) {
 735                                 ret = -ENODEV;
 736                                 goto err_out_unlock;
 737                         }
 738                         instance_destroy(queue);
 739                         break;
 740                 case NFQNL_CFG_CMD_PF_BIND:
 741                 case NFQNL_CFG_CMD_PF_UNBIND:
 742                         break;
 743                 default:
 744                         ret = -EINVAL;
 745                         break;
 746                 }
 747         }
 748
 749         if (nfqa[NFQA_CFG_PARAMS]) {
 750                 struct nfqnl_msg_config_params *params;
 751
 752                 if (!queue) {
 753                         ret = -ENODEV;
 754                         goto err_out_unlock;
 755                 }
 756                 params = nla_data(nfqa[NFQA_CFG_PARAMS]);
 757                 nfqnl_set_mode(queue, params->copy_mode,
 758                                 ntohl(params->copy_range));
 759         }
 760
 761         if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
 762                 __be32 *queue_maxlen;
 763
 764                 if (!queue) {
 765                         ret = -ENODEV;
 766                         goto err_out_unlock;
 767                 }
 768                 queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
 769                 spin_lock_bh(&queue->lock);
 770                 queue->queue_maxlen = ntohl(*queue_maxlen);
 771                 spin_unlock_bh(&queue->lock);
 772         }
 773
 774 err_out_unlock:
 775         rcu_read_unlock();
 776         return ret;
 777 }
 778
 779 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
 780         [NFQNL_MSG_PACKET]      = { .call = nfqnl_recv_unsupp,
 781                                     .attr_count = NFQA_MAX, },
 782         [NFQNL_MSG_VERDICT]     = { .call = nfqnl_recv_verdict,
 783                                     .attr_count = NFQA_MAX,
 784                                     .policy = nfqa_verdict_policy },
 785         [NFQNL_MSG_CONFIG]      = { .call = nfqnl_recv_config,
 786                                     .attr_count = NFQA_CFG_MAX,
 787                                     .policy = nfqa_cfg_policy },
 788 };
 789
 790 static const struct nfnetlink_subsystem nfqnl_subsys = {
 791         .name           = "nf_queue",
 792         .subsys_id      = NFNL_SUBSYS_QUEUE,
 793         .cb_count       = NFQNL_MSG_MAX,
 794         .cb             = nfqnl_cb,
 795 };
 796
 797 #ifdef CONFIG_PROC_FS
 798 struct iter_state {
 799         unsigned int bucket;
 800 };
 801
 802 static struct hlist_node *get_first(struct seq_file *seq)
 803 {
 804         struct iter_state *st = seq->private;
 805
 806         if (!st)
 807                 return NULL;
 808
 809         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
 810                 if (!hlist_empty(&instance_table[st->bucket]))
 811                         return instance_table[st->bucket].first;
 812         }
 813         return NULL;
 814 }
 815
 816 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
 817 {
 818         struct iter_state *st = seq->private;
 819
 820         h = h->next;
 821         while (!h) {
 822                 if (++st->bucket >= INSTANCE_BUCKETS)
 823                         return NULL;
 824
 825                 h = instance_table[st->bucket].first;
 826         }
 827         return h;
 828 }
 829
 830 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
 831 {
 832         struct hlist_node *head;
 833         head = get_first(seq);
 834
 835         if (head)
 836                 while (pos && (head = get_next(seq, head)))
 837                         pos--;
 838         return pos ? NULL : head;
 839 }
 840
 841 static void *seq_start(struct seq_file *seq, loff_t *pos)
 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 {
 855         spin_unlock(&instances_lock);
 856 }
 857
 858 static int seq_show(struct seq_file *s, void *v)
 859 {
 860         const struct nfqnl_instance *inst = v;
 861
 862         return seq_printf(s, "%5d %6d %5d %1d %5d %5d %5d %8d %2d\n",
 863                           inst->queue_num,
 864                           inst->peer_pid, inst->queue_total,
 865                           inst->copy_mode, inst->copy_range,
 866                           inst->queue_dropped, inst->queue_user_dropped,
 867                           inst->id_sequence, 1);
 868 }
 869
 870 static const struct seq_operations nfqnl_seq_ops = {
 871         .start  = seq_start,
 872         .next   = seq_next,
 873         .stop   = seq_stop,
 874         .show   = seq_show,
 875 };
 876
 877 static int nfqnl_open(struct inode *inode, struct file *file)
 878 {
 879         return seq_open_private(file, &nfqnl_seq_ops,
 880                         sizeof(struct iter_state));
 881 }
 882
 883 static const struct file_operations nfqnl_file_ops = {
 884         .owner   = THIS_MODULE,
 885         .open    = nfqnl_open,
 886         .read    = seq_read,
 887         .llseek  = seq_lseek,
 888         .release = seq_release_private,
 889 };
 890
 891 #endif /* PROC_FS */
 892
 893 static int __init nfnetlink_queue_init(void)
 894 {
 895         int i, status = -ENOMEM;
 896 #ifdef CONFIG_PROC_FS
 897         struct proc_dir_entry *proc_nfqueue;
 898 #endif
 899
 900         for (i = 0; i < INSTANCE_BUCKETS; i++)
 901                 INIT_HLIST_HEAD(&instance_table[i]);
 902
 903         netlink_register_notifier(&nfqnl_rtnl_notifier);
 904         status = nfnetlink_subsys_register(&nfqnl_subsys);
 905         if (status < 0) {
 906                 printk(KERN_ERR "nf_queue: failed to create netlink socket\n");
 907                 goto cleanup_netlink_notifier;
 908         }
 909
 910 #ifdef CONFIG_PROC_FS
 911         proc_nfqueue = create_proc_entry("nfnetlink_queue", 0440,
 912                                          proc_net_netfilter);
 913         if (!proc_nfqueue)
 914                 goto cleanup_subsys;
 915         proc_nfqueue->proc_fops = &nfqnl_file_ops;
 916 #endif
 917
 918         register_netdevice_notifier(&nfqnl_dev_notifier);
 919         return status;
 920
 921 #ifdef CONFIG_PROC_FS
 922 cleanup_subsys:
 923         nfnetlink_subsys_unregister(&nfqnl_subsys);
 924 #endif
 925 cleanup_netlink_notifier:
 926         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
 927         return status;
 928 }
 929
 930 static void __exit nfnetlink_queue_fini(void)
 931 {
 932         nf_unregister_queue_handlers(&nfqh);
 933         unregister_netdevice_notifier(&nfqnl_dev_notifier);
 934 #ifdef CONFIG_PROC_FS
 935         remove_proc_entry("nfnetlink_queue", proc_net_netfilter);
 936 #endif
 937         nfnetlink_subsys_unregister(&nfqnl_subsys);
 938         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
 939 }
 940
 941 MODULE_DESCRIPTION("netfilter packet queue handler");
 942 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
 943 MODULE_LICENSE("GPL");
 944 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
 945
 946 module_init(nfnetlink_queue_init);
 947 module_exit(nfnetlink_queue_fini);