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  * (C) 2007 by Patrick McHardy <kaber@trash.net>
   7  *
   8  * Based on the old ipv4-only ip_queue.c:
   9  * (C) 2000-2002 James Morris <jmorris@intercode.com.au>
  10  * (C) 2003-2005 Netfilter Core Team <coreteam@netfilter.org>
  11  *
  12  * This program is free software; you can redistribute it and/or modify
  13  * it under the terms of the GNU General Public License version 2 as
  14  * published by the Free Software Foundation.
  15  *
  16  */
  17 #include <linux/module.h>
  18 #include <linux/skbuff.h>
  19 #include <linux/init.h>
  20 #include <linux/spinlock.h>
  21 #include <linux/notifier.h>
  22 #include <linux/netdevice.h>
  23 #include <linux/netfilter.h>
  24 #include <linux/proc_fs.h>
  25 #include <linux/netfilter_ipv4.h>
  26 #include <linux/netfilter_ipv6.h>
  27 #include <linux/netfilter/nfnetlink.h>
  28 #include <linux/netfilter/nfnetlink_queue.h>
  29 #include <linux/list.h>
  30 #include <net/sock.h>
  31 #include <net/netfilter/nf_queue.h>
  32
  33 #include <asm/atomic.h>
  34
  35 #ifdef CONFIG_BRIDGE_NETFILTER
  36 #include "../bridge/br_private.h"
  37 #endif
  38
  39 #define NFQNL_QMAX_DEFAULT 1024
  40
  41 struct nfqnl_instance {
  42         struct hlist_node hlist;                /* global list of queues */
  43         struct rcu_head rcu;
  44
  45         int peer_pid;
  46         unsigned int queue_maxlen;
  47         unsigned int copy_range;
  48         unsigned int queue_total;
  49         unsigned int queue_dropped;
  50         unsigned int queue_user_dropped;
  51
  52         unsigned int id_sequence;               /* 'sequence' of pkt ids */
  53
  54         u_int16_t queue_num;                    /* number of this queue */
  55         u_int8_t copy_mode;
  56
  57         spinlock_t lock;
  58
  59         struct list_head queue_list;            /* packets in queue */
  60 };
  61
  62 typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
  63
  64 static DEFINE_SPINLOCK(instances_lock);
  65
  66 #define INSTANCE_BUCKETS        16
  67 static struct hlist_head instance_table[INSTANCE_BUCKETS] __read_mostly;
  68
  69 static inline u_int8_t instance_hashfn(u_int16_t queue_num)
  70 {
  71         return ((queue_num >> 8) | queue_num) % INSTANCE_BUCKETS;
  72 }
  73
  74 static struct nfqnl_instance *
  75 instance_lookup(u_int16_t queue_num)
  76 {
  77         struct hlist_head *head;
  78         struct hlist_node *pos;
  79         struct nfqnl_instance *inst;
  80
  81         head = &instance_table[instance_hashfn(queue_num)];
  82         hlist_for_each_entry_rcu(inst, pos, head, hlist) {
  83                 if (inst->queue_num == queue_num)
  84                         return inst;
  85         }
  86         return NULL;
  87 }
  88
  89 static struct nfqnl_instance *
  90 instance_create(u_int16_t queue_num, int pid)
  91 {
  92         struct nfqnl_instance *inst;
  93         unsigned int h;
  94         int err;
  95
  96         spin_lock(&instances_lock);
  97         if (instance_lookup(queue_num)) {
  98                 err = -EEXIST;
  99                 goto out_unlock;
 100         }
 101
 102         inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
 103         if (!inst) {
 104                 err = -ENOMEM;
 105                 goto out_unlock;
 106         }
 107
 108         inst->queue_num = queue_num;
 109         inst->peer_pid = pid;
 110         inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
 111         inst->copy_range = 0xfffff;
 112         inst->copy_mode = NFQNL_COPY_NONE;
 113         spin_lock_init(&inst->lock);
 114         INIT_LIST_HEAD(&inst->queue_list);
 115         INIT_RCU_HEAD(&inst->rcu);
 116
 117         if (!try_module_get(THIS_MODULE)) {
 118                 err = -EAGAIN;
 119                 goto out_free;
 120         }
 121
 122         h = instance_hashfn(queue_num);
 123         hlist_add_head_rcu(&inst->hlist, &instance_table[h]);
 124
 125         spin_unlock(&instances_lock);
 126
 127         return inst;
 128
 129 out_free:
 130         kfree(inst);
 131 out_unlock:
 132         spin_unlock(&instances_lock);
 133         return ERR_PTR(err);
 134 }
 135
 136 static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
 137                         unsigned long data);
 138
 139 static void
 140 instance_destroy_rcu(struct rcu_head *head)
 141 {
 142         struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
 143                                                    rcu);
 144
 145         nfqnl_flush(inst, NULL, 0);
 146         kfree(inst);
 147         module_put(THIS_MODULE);
 148 }
 149
 150 static void
 151 __instance_destroy(struct nfqnl_instance *inst)
 152 {
 153         hlist_del_rcu(&inst->hlist);
 154         call_rcu(&inst->rcu, instance_destroy_rcu);
 155 }
 156
 157 static void
 158 instance_destroy(struct nfqnl_instance *inst)
 159 {
 160         spin_lock(&instances_lock);
 161         __instance_destroy(inst);
 162         spin_unlock(&instances_lock);
 163 }
 164
 165 static inline void
 166 __enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
 167 {
 168        list_add_tail(&entry->list, &queue->queue_list);
 169        queue->queue_total++;
 170 }
 171
 172 static struct nf_queue_entry *
 173 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
 174 {
 175         struct nf_queue_entry *entry = NULL, *i;
 176
 177         spin_lock_bh(&queue->lock);
 178
 179         list_for_each_entry(i, &queue->queue_list, list) {
 180                 if (i->id == id) {
 181                         entry = i;
 182                         break;
 183                 }
 184         }
 185
 186         if (entry) {
 187                 list_del(&entry->list);
 188                 queue->queue_total--;
 189         }
 190
 191         spin_unlock_bh(&queue->lock);
 192
 193         return entry;
 194 }
 195
 196 static void
 197 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
 198 {
 199         struct nf_queue_entry *entry, *next;
 200
 201         spin_lock_bh(&queue->lock);
 202         list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
 203                 if (!cmpfn || cmpfn(entry, data)) {
 204                         list_del(&entry->list);
 205                         queue->queue_total--;
 206                         nf_reinject(entry, NF_DROP);
 207                 }
 208         }
 209         spin_unlock_bh(&queue->lock);
 210 }
 211
 212 static struct sk_buff *
 213 nfqnl_build_packet_message(struct nfqnl_instance *queue,
 214                            struct nf_queue_entry *entry)
 215 {
 216         sk_buff_data_t old_tail;
 217         size_t size;
 218         size_t data_len = 0;
 219         struct sk_buff *skb;
 220         struct nfqnl_msg_packet_hdr pmsg;
 221         struct nlmsghdr *nlh;
 222         struct nfgenmsg *nfmsg;
 223         struct sk_buff *entskb = entry->skb;
 224         struct net_device *indev;
 225         struct net_device *outdev;
 226
 227         size =    NLMSG_SPACE(sizeof(struct nfgenmsg))
 228                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
 229                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 230                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 231 #ifdef CONFIG_BRIDGE_NETFILTER
 232                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 233                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 234 #endif
 235                 + nla_total_size(sizeof(u_int32_t))     /* mark */
 236                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
 237                 + nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
 238
 239         outdev = entry->outdev;
 240
 241         spin_lock_bh(&queue->lock);
 242
 243         switch ((enum nfqnl_config_mode)queue->copy_mode) {
 244         case NFQNL_COPY_META:
 245         case NFQNL_COPY_NONE:
 246                 data_len = 0;
 247                 break;
 248
 249         case NFQNL_COPY_PACKET:
 250                 if ((entskb->ip_summed == CHECKSUM_PARTIAL ||
 251                      entskb->ip_summed == CHECKSUM_COMPLETE) &&
 252                     skb_checksum_help(entskb)) {
 253                         spin_unlock_bh(&queue->lock);
 254                         return NULL;
 255                 }
 256                 if (queue->copy_range == 0
 257                     || queue->copy_range > entskb->len)
 258                         data_len = entskb->len;
 259                 else
 260                         data_len = queue->copy_range;
 261
 262                 size += nla_total_size(data_len);
 263                 break;
 264         }
 265
 266         entry->id = queue->id_sequence++;
 267
 268         spin_unlock_bh(&queue->lock);
 269
 270         skb = alloc_skb(size, GFP_ATOMIC);
 271         if (!skb)
 272                 goto nlmsg_failure;
 273
 274         old_tail = skb->tail;
 275         nlh = NLMSG_PUT(skb, 0, 0,
 276                         NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
 277                         sizeof(struct nfgenmsg));
 278         nfmsg = NLMSG_DATA(nlh);
 279         nfmsg->nfgen_family = entry->pf;
 280         nfmsg->version = NFNETLINK_V0;
 281         nfmsg->res_id = htons(queue->queue_num);
 282
 283         pmsg.packet_id          = htonl(entry->id);
 284         pmsg.hw_protocol        = entskb->protocol;
 285         pmsg.hook               = entry->hook;
 286
 287         NLA_PUT(skb, NFQA_PACKET_HDR, sizeof(pmsg), &pmsg);
 288
 289         indev = entry->indev;
 290         if (indev) {
 291 #ifndef CONFIG_BRIDGE_NETFILTER
 292                 NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex));
 293 #else
 294                 if (entry->pf == PF_BRIDGE) {
 295                         /* Case 1: indev is physical input device, we need to
 296                          * look for bridge group (when called from
 297                          * netfilter_bridge) */
 298                         NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSINDEV,
 299                                      htonl(indev->ifindex));
 300                         /* this is the bridge group "brX" */
 301                         NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV,
 302                                      htonl(indev->br_port->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                         NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV,
 327                                      htonl(outdev->br_port->br->dev->ifindex));
 328                 } else {
 329                         /* Case 2: outdev is bridge group, we need to look for
 330                          * physical output device (when called from ipv4) */
 331                         NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV,
 332                                      htonl(outdev->ifindex));
 333                         if (entskb->nf_bridge && entskb->nf_bridge->physoutdev)
 334                                 NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 335                                              htonl(entskb->nf_bridge->physoutdev->ifindex));
 336                 }
 337 #endif
 338         }
 339
 340         if (entskb->mark)
 341                 NLA_PUT_BE32(skb, NFQA_MARK, htonl(entskb->mark));
 342
 343         if (indev && entskb->dev) {
 344                 struct nfqnl_msg_packet_hw phw;
 345                 int len = dev_parse_header(entskb, phw.hw_addr);
 346                 if (len) {
 347                         phw.hw_addrlen = htons(len);
 348                         NLA_PUT(skb, NFQA_HWADDR, sizeof(phw), &phw);
 349                 }
 350         }
 351
 352         if (entskb->tstamp.tv64) {
 353                 struct nfqnl_msg_packet_timestamp ts;
 354                 struct timeval tv = ktime_to_timeval(entskb->tstamp);
 355                 ts.sec = cpu_to_be64(tv.tv_sec);
 356                 ts.usec = cpu_to_be64(tv.tv_usec);
 357
 358                 NLA_PUT(skb, NFQA_TIMESTAMP, sizeof(ts), &ts);
 359         }
 360
 361         if (data_len) {
 362                 struct nlattr *nla;
 363                 int sz = nla_attr_size(data_len);
 364
 365                 if (skb_tailroom(skb) < nla_total_size(data_len)) {
 366                         printk(KERN_WARNING "nf_queue: no tailroom!\n");
 367                         goto nlmsg_failure;
 368                 }
 369
 370                 nla = (struct nlattr *)skb_put(skb, nla_total_size(data_len));
 371                 nla->nla_type = NFQA_PAYLOAD;
 372                 nla->nla_len = sz;
 373
 374                 if (skb_copy_bits(entskb, 0, nla_data(nla), data_len))
 375                         BUG();
 376         }
 377
 378         nlh->nlmsg_len = skb->tail - old_tail;
 379         return skb;
 380
 381 nlmsg_failure:
 382 nla_put_failure:
 383         if (skb)
 384                 kfree_skb(skb);
 385         if (net_ratelimit())
 386                 printk(KERN_ERR "nf_queue: error creating packet message\n");
 387         return NULL;
 388 }
 389
 390 static int
 391 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 392 {
 393         struct sk_buff *nskb;
 394         struct nfqnl_instance *queue;
 395         int err;
 396
 397         /* rcu_read_lock()ed by nf_hook_slow() */
 398         queue = instance_lookup(queuenum);
 399         if (!queue)
 400                 goto err_out;
 401
 402         if (queue->copy_mode == NFQNL_COPY_NONE)
 403                 goto err_out;
 404
 405         nskb = nfqnl_build_packet_message(queue, entry);
 406         if (nskb == NULL)
 407                 goto err_out;
 408
 409         spin_lock_bh(&queue->lock);
 410
 411         if (!queue->peer_pid)
 412                 goto err_out_free_nskb;
 413
 414         if (queue->queue_total >= queue->queue_maxlen) {
 415                 queue->queue_dropped++;
 416                 if (net_ratelimit())
 417                           printk(KERN_WARNING "nf_queue: full at %d entries, "
 418                                  "dropping packets(s). Dropped: %d\n",
 419                                  queue->queue_total, queue->queue_dropped);
 420                 goto err_out_free_nskb;
 421         }
 422
 423         /* nfnetlink_unicast will either free the nskb or add it to a socket */
 424         err = nfnetlink_unicast(nskb, queue->peer_pid, MSG_DONTWAIT);
 425         if (err < 0) {
 426                 queue->queue_user_dropped++;
 427                 goto err_out_unlock;
 428         }
 429
 430         __enqueue_entry(queue, entry);
 431
 432         spin_unlock_bh(&queue->lock);
 433         return 0;
 434
 435 err_out_free_nskb:
 436         kfree_skb(nskb);
 437 err_out_unlock:
 438         spin_unlock_bh(&queue->lock);
 439 err_out:
 440         return -1;
 441 }
 442
 443 static int
 444 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e)
 445 {
 446         struct sk_buff *nskb;
 447         int diff;
 448
 449         diff = data_len - e->skb->len;
 450         if (diff < 0) {
 451                 if (pskb_trim(e->skb, data_len))
 452                         return -ENOMEM;
 453         } else if (diff > 0) {
 454                 if (data_len > 0xFFFF)
 455                         return -EINVAL;
 456                 if (diff > skb_tailroom(e->skb)) {
 457                         nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
 458                                                diff, GFP_ATOMIC);
 459                         if (!nskb) {
 460                                 printk(KERN_WARNING "nf_queue: OOM "
 461                                       "in mangle, dropping packet\n");
 462                                 return -ENOMEM;
 463                         }
 464                         kfree_skb(e->skb);
 465                         e->skb = nskb;
 466                 }
 467                 skb_put(e->skb, diff);
 468         }
 469         if (!skb_make_writable(e->skb, data_len))
 470                 return -ENOMEM;
 471         skb_copy_to_linear_data(e->skb, data, data_len);
 472         e->skb->ip_summed = CHECKSUM_NONE;
 473         return 0;
 474 }
 475
 476 static int
 477 nfqnl_set_mode(struct nfqnl_instance *queue,
 478                unsigned char mode, unsigned int range)
 479 {
 480         int status = 0;
 481
 482         spin_lock_bh(&queue->lock);
 483         switch (mode) {
 484         case NFQNL_COPY_NONE:
 485         case NFQNL_COPY_META:
 486                 queue->copy_mode = mode;
 487                 queue->copy_range = 0;
 488                 break;
 489
 490         case NFQNL_COPY_PACKET:
 491                 queue->copy_mode = mode;
 492                 /* we're using struct nlattr which has 16bit nla_len */
 493                 if (range > 0xffff)
 494                         queue->copy_range = 0xffff;
 495                 else
 496                         queue->copy_range = range;
 497                 break;
 498
 499         default:
 500                 status = -EINVAL;
 501
 502         }
 503         spin_unlock_bh(&queue->lock);
 504
 505         return status;
 506 }
 507
 508 static int
 509 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
 510 {
 511         if (entry->indev)
 512                 if (entry->indev->ifindex == ifindex)
 513                         return 1;
 514         if (entry->outdev)
 515                 if (entry->outdev->ifindex == ifindex)
 516                         return 1;
 517 #ifdef CONFIG_BRIDGE_NETFILTER
 518         if (entry->skb->nf_bridge) {
 519                 if (entry->skb->nf_bridge->physindev &&
 520                     entry->skb->nf_bridge->physindev->ifindex == ifindex)
 521                         return 1;
 522                 if (entry->skb->nf_bridge->physoutdev &&
 523                     entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
 524                         return 1;
 525         }
 526 #endif
 527         return 0;
 528 }
 529
 530 /* drop all packets with either indev or outdev == ifindex from all queue
 531  * instances */
 532 static void
 533 nfqnl_dev_drop(int ifindex)
 534 {
 535         int i;
 536
 537         rcu_read_lock();
 538
 539         for (i = 0; i < INSTANCE_BUCKETS; i++) {
 540                 struct hlist_node *tmp;
 541                 struct nfqnl_instance *inst;
 542                 struct hlist_head *head = &instance_table[i];
 543
 544                 hlist_for_each_entry_rcu(inst, tmp, head, hlist)
 545                         nfqnl_flush(inst, dev_cmp, ifindex);
 546         }
 547
 548         rcu_read_unlock();
 549 }
 550
 551 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
 552
 553 static int
 554 nfqnl_rcv_dev_event(struct notifier_block *this,
 555                     unsigned long event, void *ptr)
 556 {
 557         struct net_device *dev = ptr;
 558
 559         if (dev_net(dev) != &init_net)
 560                 return NOTIFY_DONE;
 561
 562         /* Drop any packets associated with the downed device */
 563         if (event == NETDEV_DOWN)
 564                 nfqnl_dev_drop(dev->ifindex);
 565         return NOTIFY_DONE;
 566 }
 567
 568 static struct notifier_block nfqnl_dev_notifier = {
 569         .notifier_call  = nfqnl_rcv_dev_event,
 570 };
 571
 572 static int
 573 nfqnl_rcv_nl_event(struct notifier_block *this,
 574                    unsigned long event, void *ptr)
 575 {
 576         struct netlink_notify *n = ptr;
 577
 578         if (event == NETLINK_URELEASE &&
 579             n->protocol == NETLINK_NETFILTER && n->pid) {
 580                 int i;
 581
 582                 /* destroy all instances for this pid */
 583                 spin_lock(&instances_lock);
 584                 for (i = 0; i < INSTANCE_BUCKETS; i++) {
 585                         struct hlist_node *tmp, *t2;
 586                         struct nfqnl_instance *inst;
 587                         struct hlist_head *head = &instance_table[i];
 588
 589                         hlist_for_each_entry_safe(inst, tmp, t2, head, hlist) {
 590                                 if ((n->net == &init_net) &&
 591                                     (n->pid == inst->peer_pid))
 592                                         __instance_destroy(inst);
 593                         }
 594                 }
 595                 spin_unlock(&instances_lock);
 596         }
 597         return NOTIFY_DONE;
 598 }
 599
 600 static struct notifier_block nfqnl_rtnl_notifier = {
 601         .notifier_call  = nfqnl_rcv_nl_event,
 602 };
 603
 604 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
 605         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 606         [NFQA_MARK]             = { .type = NLA_U32 },
 607         [NFQA_PAYLOAD]          = { .type = NLA_UNSPEC },
 608 };
 609
 610 static int
 611 nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
 612                    struct nlmsghdr *nlh, struct nlattr *nfqa[])
 613 {
 614         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 615         u_int16_t queue_num = ntohs(nfmsg->res_id);
 616
 617         struct nfqnl_msg_verdict_hdr *vhdr;
 618         struct nfqnl_instance *queue;
 619         unsigned int verdict;
 620         struct nf_queue_entry *entry;
 621         int err;
 622
 623         rcu_read_lock();
 624         queue = instance_lookup(queue_num);
 625         if (!queue) {
 626                 err = -ENODEV;
 627                 goto err_out_unlock;
 628         }
 629
 630         if (queue->peer_pid != NETLINK_CB(skb).pid) {
 631                 err = -EPERM;
 632                 goto err_out_unlock;
 633         }
 634
 635         if (!nfqa[NFQA_VERDICT_HDR]) {
 636                 err = -EINVAL;
 637                 goto err_out_unlock;
 638         }
 639
 640         vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
 641         verdict = ntohl(vhdr->verdict);
 642
 643         if ((verdict & NF_VERDICT_MASK) > NF_MAX_VERDICT) {
 644                 err = -EINVAL;
 645                 goto err_out_unlock;
 646         }
 647
 648         entry = find_dequeue_entry(queue, ntohl(vhdr->id));
 649         if (entry == NULL) {
 650                 err = -ENOENT;
 651                 goto err_out_unlock;
 652         }
 653         rcu_read_unlock();
 654
 655         if (nfqa[NFQA_PAYLOAD]) {
 656                 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
 657                                  nla_len(nfqa[NFQA_PAYLOAD]), entry) < 0)
 658                         verdict = NF_DROP;
 659         }
 660
 661         if (nfqa[NFQA_MARK])
 662                 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
 663
 664         nf_reinject(entry, verdict);
 665         return 0;
 666
 667 err_out_unlock:
 668         rcu_read_unlock();
 669         return err;
 670 }
 671
 672 static int
 673 nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
 674                   struct nlmsghdr *nlh, struct nlattr *nfqa[])
 675 {
 676         return -ENOTSUPP;
 677 }
 678
 679 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
 680         [NFQA_CFG_CMD]          = { .len = sizeof(struct nfqnl_msg_config_cmd) },
 681         [NFQA_CFG_PARAMS]       = { .len = sizeof(struct nfqnl_msg_config_params) },
 682 };
 683
 684 static const struct nf_queue_handler nfqh = {
 685         .name   = "nf_queue",
 686         .outfn  = &nfqnl_enqueue_packet,
 687 };
 688
 689 static int
 690 nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
 691                   struct nlmsghdr *nlh, struct nlattr *nfqa[])
 692 {
 693         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 694         u_int16_t queue_num = ntohs(nfmsg->res_id);
 695         struct nfqnl_instance *queue;
 696         struct nfqnl_msg_config_cmd *cmd = NULL;
 697         int ret = 0;
 698
 699         if (nfqa[NFQA_CFG_CMD]) {
 700                 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
 701
 702                 /* Commands without queue context - might sleep */
 703                 switch (cmd->command) {
 704                 case NFQNL_CFG_CMD_PF_BIND:
 705                         return nf_register_queue_handler(ntohs(cmd->pf),
 706                                                          &nfqh);
 707                 case NFQNL_CFG_CMD_PF_UNBIND:
 708                         return nf_unregister_queue_handler(ntohs(cmd->pf),
 709                                                            &nfqh);
 710                 }
 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 (IS_ERR(queue)) {
 729                                 ret = PTR_ERR(queue);
 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 = -ENOTSUPP;
 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         __acquires(instances_lock)
 843 {
 844         spin_lock(&instances_lock);
 845         return get_idx(seq, *pos);
 846 }
 847
 848 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
 849 {
 850         (*pos)++;
 851         return get_next(s, v);
 852 }
 853
 854 static void seq_stop(struct seq_file *s, void *v)
 855         __releases(instances_lock)
 856 {
 857         spin_unlock(&instances_lock);
 858 }
 859
 860 static int seq_show(struct seq_file *s, void *v)
 861 {
 862         const struct nfqnl_instance *inst = v;
 863
 864         return seq_printf(s, "%5d %6d %5d %1d %5d %5d %5d %8d %2d\n",
 865                           inst->queue_num,
 866                           inst->peer_pid, inst->queue_total,
 867                           inst->copy_mode, inst->copy_range,
 868                           inst->queue_dropped, inst->queue_user_dropped,
 869                           inst->id_sequence, 1);
 870 }
 871
 872 static const struct seq_operations nfqnl_seq_ops = {
 873         .start  = seq_start,
 874         .next   = seq_next,
 875         .stop   = seq_stop,
 876         .show   = seq_show,
 877 };
 878
 879 static int nfqnl_open(struct inode *inode, struct file *file)
 880 {
 881         return seq_open_private(file, &nfqnl_seq_ops,
 882                         sizeof(struct iter_state));
 883 }
 884
 885 static const struct file_operations nfqnl_file_ops = {
 886         .owner   = THIS_MODULE,
 887         .open    = nfqnl_open,
 888         .read    = seq_read,
 889         .llseek  = seq_lseek,
 890         .release = seq_release_private,
 891 };
 892
 893 #endif /* PROC_FS */
 894
 895 static int __init nfnetlink_queue_init(void)
 896 {
 897         int i, status = -ENOMEM;
 898
 899         for (i = 0; i < INSTANCE_BUCKETS; i++)
 900                 INIT_HLIST_HEAD(&instance_table[i]);
 901
 902         netlink_register_notifier(&nfqnl_rtnl_notifier);
 903         status = nfnetlink_subsys_register(&nfqnl_subsys);
 904         if (status < 0) {
 905                 printk(KERN_ERR "nf_queue: failed to create netlink socket\n");
 906                 goto cleanup_netlink_notifier;
 907         }
 908
 909 #ifdef CONFIG_PROC_FS
 910         if (!proc_create("nfnetlink_queue", 0440,
 911                          proc_net_netfilter, &nfqnl_file_ops))
 912                 goto cleanup_subsys;
 913 #endif
 914
 915         register_netdevice_notifier(&nfqnl_dev_notifier);
 916         return status;
 917
 918 #ifdef CONFIG_PROC_FS
 919 cleanup_subsys:
 920         nfnetlink_subsys_unregister(&nfqnl_subsys);
 921 #endif
 922 cleanup_netlink_notifier:
 923         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
 924         return status;
 925 }
 926
 927 static void __exit nfnetlink_queue_fini(void)
 928 {
 929         nf_unregister_queue_handlers(&nfqh);
 930         unregister_netdevice_notifier(&nfqnl_dev_notifier);
 931 #ifdef CONFIG_PROC_FS
 932         remove_proc_entry("nfnetlink_queue", proc_net_netfilter);
 933 #endif
 934         nfnetlink_subsys_unregister(&nfqnl_subsys);
 935         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
 936 }
 937
 938 MODULE_DESCRIPTION("netfilter packet queue handler");
 939 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
 940 MODULE_LICENSE("GPL");
 941 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
 942
 943 module_init(nfnetlink_queue_init);
 944 module_exit(nfnetlink_queue_fini);