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