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                 break;
 247
 248         case NFQNL_COPY_PACKET:
 249                 if ((entskb->ip_summed == CHECKSUM_PARTIAL ||
 250                      entskb->ip_summed == CHECKSUM_COMPLETE) &&
 251                     skb_checksum_help(entskb)) {
 252                         spin_unlock_bh(&queue->lock);
 253                         return NULL;
 254                 }
 255                 if (queue->copy_range == 0
 256                     || queue->copy_range > entskb->len)
 257                         data_len = entskb->len;
 258                 else
 259                         data_len = queue->copy_range;
 260
 261                 size += nla_total_size(data_len);
 262                 break;
 263         }
 264
 265         entry->id = queue->id_sequence++;
 266
 267         spin_unlock_bh(&queue->lock);
 268
 269         skb = alloc_skb(size, GFP_ATOMIC);
 270         if (!skb)
 271                 goto nlmsg_failure;
 272
 273         old_tail = skb->tail;
 274         nlh = NLMSG_PUT(skb, 0, 0,
 275                         NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
 276                         sizeof(struct nfgenmsg));
 277         nfmsg = NLMSG_DATA(nlh);
 278         nfmsg->nfgen_family = entry->pf;
 279         nfmsg->version = NFNETLINK_V0;
 280         nfmsg->res_id = htons(queue->queue_num);
 281
 282         pmsg.packet_id          = htonl(entry->id);
 283         pmsg.hw_protocol        = entskb->protocol;
 284         pmsg.hook               = entry->hook;
 285
 286         NLA_PUT(skb, NFQA_PACKET_HDR, sizeof(pmsg), &pmsg);
 287
 288         indev = entry->indev;
 289         if (indev) {
 290 #ifndef CONFIG_BRIDGE_NETFILTER
 291                 NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex));
 292 #else
 293                 if (entry->pf == PF_BRIDGE) {
 294                         /* Case 1: indev is physical input device, we need to
 295                          * look for bridge group (when called from
 296                          * netfilter_bridge) */
 297                         NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSINDEV,
 298                                      htonl(indev->ifindex));
 299                         /* this is the bridge group "brX" */
 300                         NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV,
 301                                      htonl(indev->br_port->br->dev->ifindex));
 302                 } else {
 303                         /* Case 2: indev is bridge group, we need to look for
 304                          * physical device (when called from ipv4) */
 305                         NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV,
 306                                      htonl(indev->ifindex));
 307                         if (entskb->nf_bridge && entskb->nf_bridge->physindev)
 308                                 NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSINDEV,
 309                                              htonl(entskb->nf_bridge->physindev->ifindex));
 310                 }
 311 #endif
 312         }
 313
 314         if (outdev) {
 315 #ifndef CONFIG_BRIDGE_NETFILTER
 316                 NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex));
 317 #else
 318                 if (entry->pf == PF_BRIDGE) {
 319                         /* Case 1: outdev is physical output device, we need to
 320                          * look for bridge group (when called from
 321                          * netfilter_bridge) */
 322                         NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 323                                      htonl(outdev->ifindex));
 324                         /* this is the bridge group "brX" */
 325                         NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV,
 326                                      htonl(outdev->br_port->br->dev->ifindex));
 327                 } else {
 328                         /* Case 2: outdev is bridge group, we need to look for
 329                          * physical output device (when called from ipv4) */
 330                         NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV,
 331                                      htonl(outdev->ifindex));
 332                         if (entskb->nf_bridge && entskb->nf_bridge->physoutdev)
 333                                 NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 334                                              htonl(entskb->nf_bridge->physoutdev->ifindex));
 335                 }
 336 #endif
 337         }
 338
 339         if (entskb->mark)
 340                 NLA_PUT_BE32(skb, NFQA_MARK, htonl(entskb->mark));
 341
 342         if (indev && entskb->dev) {
 343                 struct nfqnl_msg_packet_hw phw;
 344                 int len = dev_parse_header(entskb, phw.hw_addr);
 345                 if (len) {
 346                         phw.hw_addrlen = htons(len);
 347                         NLA_PUT(skb, NFQA_HWADDR, sizeof(phw), &phw);
 348                 }
 349         }
 350
 351         if (entskb->tstamp.tv64) {
 352                 struct nfqnl_msg_packet_timestamp ts;
 353                 struct timeval tv = ktime_to_timeval(entskb->tstamp);
 354                 ts.sec = cpu_to_be64(tv.tv_sec);
 355                 ts.usec = cpu_to_be64(tv.tv_usec);
 356
 357                 NLA_PUT(skb, NFQA_TIMESTAMP, sizeof(ts), &ts);
 358         }
 359
 360         if (data_len) {
 361                 struct nlattr *nla;
 362                 int sz = nla_attr_size(data_len);
 363
 364                 if (skb_tailroom(skb) < nla_total_size(data_len)) {
 365                         printk(KERN_WARNING "nf_queue: no tailroom!\n");
 366                         goto nlmsg_failure;
 367                 }
 368
 369                 nla = (struct nlattr *)skb_put(skb, nla_total_size(data_len));
 370                 nla->nla_type = NFQA_PAYLOAD;
 371                 nla->nla_len = sz;
 372
 373                 if (skb_copy_bits(entskb, 0, nla_data(nla), data_len))
 374                         BUG();
 375         }
 376
 377         nlh->nlmsg_len = skb->tail - old_tail;
 378         return skb;
 379
 380 nlmsg_failure:
 381 nla_put_failure:
 382         if (skb)
 383                 kfree_skb(skb);
 384         if (net_ratelimit())
 385                 printk(KERN_ERR "nf_queue: error creating packet message\n");
 386         return NULL;
 387 }
 388
 389 static int
 390 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 391 {
 392         struct sk_buff *nskb;
 393         struct nfqnl_instance *queue;
 394         int err;
 395
 396         /* rcu_read_lock()ed by nf_hook_slow() */
 397         queue = instance_lookup(queuenum);
 398         if (!queue)
 399                 goto err_out;
 400
 401         if (queue->copy_mode == NFQNL_COPY_NONE)
 402                 goto err_out;
 403
 404         nskb = nfqnl_build_packet_message(queue, entry);
 405         if (nskb == NULL)
 406                 goto err_out;
 407
 408         spin_lock_bh(&queue->lock);
 409
 410         if (!queue->peer_pid)
 411                 goto err_out_free_nskb;
 412
 413         if (queue->queue_total >= queue->queue_maxlen) {
 414                 queue->queue_dropped++;
 415                 if (net_ratelimit())
 416                           printk(KERN_WARNING "nf_queue: full at %d entries, "
 417                                  "dropping packets(s). Dropped: %d\n",
 418                                  queue->queue_total, queue->queue_dropped);
 419                 goto err_out_free_nskb;
 420         }
 421
 422         /* nfnetlink_unicast will either free the nskb or add it to a socket */
 423         err = nfnetlink_unicast(nskb, queue->peer_pid, MSG_DONTWAIT);
 424         if (err < 0) {
 425                 queue->queue_user_dropped++;
 426                 goto err_out_unlock;
 427         }
 428
 429         __enqueue_entry(queue, entry);
 430
 431         spin_unlock_bh(&queue->lock);
 432         return 0;
 433
 434 err_out_free_nskb:
 435         kfree_skb(nskb);
 436 err_out_unlock:
 437         spin_unlock_bh(&queue->lock);
 438 err_out:
 439         return -1;
 440 }
 441
 442 static int
 443 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e)
 444 {
 445         struct sk_buff *nskb;
 446         int diff;
 447
 448         diff = data_len - e->skb->len;
 449         if (diff < 0) {
 450                 if (pskb_trim(e->skb, data_len))
 451                         return -ENOMEM;
 452         } else if (diff > 0) {
 453                 if (data_len > 0xFFFF)
 454                         return -EINVAL;
 455                 if (diff > skb_tailroom(e->skb)) {
 456                         nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
 457                                                diff, GFP_ATOMIC);
 458                         if (!nskb) {
 459                                 printk(KERN_WARNING "nf_queue: OOM "
 460                                       "in mangle, dropping packet\n");
 461                                 return -ENOMEM;
 462                         }
 463                         kfree_skb(e->skb);
 464                         e->skb = nskb;
 465                 }
 466                 skb_put(e->skb, diff);
 467         }
 468         if (!skb_make_writable(e->skb, data_len))
 469                 return -ENOMEM;
 470         skb_copy_to_linear_data(e->skb, data, data_len);
 471         e->skb->ip_summed = CHECKSUM_NONE;
 472         return 0;
 473 }
 474
 475 static int
 476 nfqnl_set_mode(struct nfqnl_instance *queue,
 477                unsigned char mode, unsigned int range)
 478 {
 479         int status = 0;
 480
 481         spin_lock_bh(&queue->lock);
 482         switch (mode) {
 483         case NFQNL_COPY_NONE:
 484         case NFQNL_COPY_META:
 485                 queue->copy_mode = mode;
 486                 queue->copy_range = 0;
 487                 break;
 488
 489         case NFQNL_COPY_PACKET:
 490                 queue->copy_mode = mode;
 491                 /* we're using struct nlattr which has 16bit nla_len */
 492                 if (range > 0xffff)
 493                         queue->copy_range = 0xffff;
 494                 else
 495                         queue->copy_range = range;
 496                 break;
 497
 498         default:
 499                 status = -EINVAL;
 500
 501         }
 502         spin_unlock_bh(&queue->lock);
 503
 504         return status;
 505 }
 506
 507 static int
 508 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
 509 {
 510         if (entry->indev)
 511                 if (entry->indev->ifindex == ifindex)
 512                         return 1;
 513         if (entry->outdev)
 514                 if (entry->outdev->ifindex == ifindex)
 515                         return 1;
 516 #ifdef CONFIG_BRIDGE_NETFILTER
 517         if (entry->skb->nf_bridge) {
 518                 if (entry->skb->nf_bridge->physindev &&
 519                     entry->skb->nf_bridge->physindev->ifindex == ifindex)
 520                         return 1;
 521                 if (entry->skb->nf_bridge->physoutdev &&
 522                     entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
 523                         return 1;
 524         }
 525 #endif
 526         return 0;
 527 }
 528
 529 /* drop all packets with either indev or outdev == ifindex from all queue
 530  * instances */
 531 static void
 532 nfqnl_dev_drop(int ifindex)
 533 {
 534         int i;
 535
 536         rcu_read_lock();
 537
 538         for (i = 0; i < INSTANCE_BUCKETS; i++) {
 539                 struct hlist_node *tmp;
 540                 struct nfqnl_instance *inst;
 541                 struct hlist_head *head = &instance_table[i];
 542
 543                 hlist_for_each_entry_rcu(inst, tmp, head, hlist)
 544                         nfqnl_flush(inst, dev_cmp, ifindex);
 545         }
 546
 547         rcu_read_unlock();
 548 }
 549
 550 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
 551
 552 static int
 553 nfqnl_rcv_dev_event(struct notifier_block *this,
 554                     unsigned long event, void *ptr)
 555 {
 556         struct net_device *dev = ptr;
 557
 558         if (!net_eq(dev_net(dev), &init_net))
 559                 return NOTIFY_DONE;
 560
 561         /* Drop any packets associated with the downed device */
 562         if (event == NETDEV_DOWN)
 563                 nfqnl_dev_drop(dev->ifindex);
 564         return NOTIFY_DONE;
 565 }
 566
 567 static struct notifier_block nfqnl_dev_notifier = {
 568         .notifier_call  = nfqnl_rcv_dev_event,
 569 };
 570
 571 static int
 572 nfqnl_rcv_nl_event(struct notifier_block *this,
 573                    unsigned long event, void *ptr)
 574 {
 575         struct netlink_notify *n = ptr;
 576
 577         if (event == NETLINK_URELEASE &&
 578             n->protocol == NETLINK_NETFILTER && n->pid) {
 579                 int i;
 580
 581                 /* destroy all instances for this pid */
 582                 spin_lock(&instances_lock);
 583                 for (i = 0; i < INSTANCE_BUCKETS; i++) {
 584                         struct hlist_node *tmp, *t2;
 585                         struct nfqnl_instance *inst;
 586                         struct hlist_head *head = &instance_table[i];
 587
 588                         hlist_for_each_entry_safe(inst, tmp, t2, head, hlist) {
 589                                 if ((n->net == &init_net) &&
 590                                     (n->pid == inst->peer_pid))
 591                                         __instance_destroy(inst);
 592                         }
 593                 }
 594                 spin_unlock(&instances_lock);
 595         }
 596         return NOTIFY_DONE;
 597 }
 598
 599 static struct notifier_block nfqnl_rtnl_notifier = {
 600         .notifier_call  = nfqnl_rcv_nl_event,
 601 };
 602
 603 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
 604         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 605         [NFQA_MARK]             = { .type = NLA_U32 },
 606         [NFQA_PAYLOAD]          = { .type = NLA_UNSPEC },
 607 };
 608
 609 static int
 610 nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
 611                    struct nlmsghdr *nlh, struct nlattr *nfqa[])
 612 {
 613         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 614         u_int16_t queue_num = ntohs(nfmsg->res_id);
 615
 616         struct nfqnl_msg_verdict_hdr *vhdr;
 617         struct nfqnl_instance *queue;
 618         unsigned int verdict;
 619         struct nf_queue_entry *entry;
 620         int err;
 621
 622         rcu_read_lock();
 623         queue = instance_lookup(queue_num);
 624         if (!queue) {
 625                 err = -ENODEV;
 626                 goto err_out_unlock;
 627         }
 628
 629         if (queue->peer_pid != NETLINK_CB(skb).pid) {
 630                 err = -EPERM;
 631                 goto err_out_unlock;
 632         }
 633
 634         if (!nfqa[NFQA_VERDICT_HDR]) {
 635                 err = -EINVAL;
 636                 goto err_out_unlock;
 637         }
 638
 639         vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
 640         verdict = ntohl(vhdr->verdict);
 641
 642         if ((verdict & NF_VERDICT_MASK) > NF_MAX_VERDICT) {
 643                 err = -EINVAL;
 644                 goto err_out_unlock;
 645         }
 646
 647         entry = find_dequeue_entry(queue, ntohl(vhdr->id));
 648         if (entry == NULL) {
 649                 err = -ENOENT;
 650                 goto err_out_unlock;
 651         }
 652         rcu_read_unlock();
 653
 654         if (nfqa[NFQA_PAYLOAD]) {
 655                 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
 656                                  nla_len(nfqa[NFQA_PAYLOAD]), entry) < 0)
 657                         verdict = NF_DROP;
 658         }
 659
 660         if (nfqa[NFQA_MARK])
 661                 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
 662
 663         nf_reinject(entry, verdict);
 664         return 0;
 665
 666 err_out_unlock:
 667         rcu_read_unlock();
 668         return err;
 669 }
 670
 671 static int
 672 nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
 673                   struct nlmsghdr *nlh, struct nlattr *nfqa[])
 674 {
 675         return -ENOTSUPP;
 676 }
 677
 678 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
 679         [NFQA_CFG_CMD]          = { .len = sizeof(struct nfqnl_msg_config_cmd) },
 680         [NFQA_CFG_PARAMS]       = { .len = sizeof(struct nfqnl_msg_config_params) },
 681 };
 682
 683 static const struct nf_queue_handler nfqh = {
 684         .name   = "nf_queue",
 685         .outfn  = &nfqnl_enqueue_packet,
 686 };
 687
 688 static int
 689 nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
 690                   struct nlmsghdr *nlh, struct nlattr *nfqa[])
 691 {
 692         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 693         u_int16_t queue_num = ntohs(nfmsg->res_id);
 694         struct nfqnl_instance *queue;
 695         struct nfqnl_msg_config_cmd *cmd = NULL;
 696         int ret = 0;
 697
 698         if (nfqa[NFQA_CFG_CMD]) {
 699                 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
 700
 701                 /* Commands without queue context - might sleep */
 702                 switch (cmd->command) {
 703                 case NFQNL_CFG_CMD_PF_BIND:
 704                         return nf_register_queue_handler(ntohs(cmd->pf),
 705                                                          &nfqh);
 706                 case NFQNL_CFG_CMD_PF_UNBIND:
 707                         return nf_unregister_queue_handler(ntohs(cmd->pf),
 708                                                            &nfqh);
 709                 }
 710         }
 711
 712         rcu_read_lock();
 713         queue = instance_lookup(queue_num);
 714         if (queue && queue->peer_pid != NETLINK_CB(skb).pid) {
 715                 ret = -EPERM;
 716                 goto err_out_unlock;
 717         }
 718
 719         if (cmd != NULL) {
 720                 switch (cmd->command) {
 721                 case NFQNL_CFG_CMD_BIND:
 722                         if (queue) {
 723                                 ret = -EBUSY;
 724                                 goto err_out_unlock;
 725                         }
 726                         queue = instance_create(queue_num, NETLINK_CB(skb).pid);
 727                         if (IS_ERR(queue)) {
 728                                 ret = PTR_ERR(queue);
 729                                 goto err_out_unlock;
 730                         }
 731                         break;
 732                 case NFQNL_CFG_CMD_UNBIND:
 733                         if (!queue) {
 734                                 ret = -ENODEV;
 735                                 goto err_out_unlock;
 736                         }
 737                         instance_destroy(queue);
 738                         break;
 739                 case NFQNL_CFG_CMD_PF_BIND:
 740                 case NFQNL_CFG_CMD_PF_UNBIND:
 741                         break;
 742                 default:
 743                         ret = -ENOTSUPP;
 744                         break;
 745                 }
 746         }
 747
 748         if (nfqa[NFQA_CFG_PARAMS]) {
 749                 struct nfqnl_msg_config_params *params;
 750
 751                 if (!queue) {
 752                         ret = -ENODEV;
 753                         goto err_out_unlock;
 754                 }
 755                 params = nla_data(nfqa[NFQA_CFG_PARAMS]);
 756                 nfqnl_set_mode(queue, params->copy_mode,
 757                                 ntohl(params->copy_range));
 758         }
 759
 760         if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
 761                 __be32 *queue_maxlen;
 762
 763                 if (!queue) {
 764                         ret = -ENODEV;
 765                         goto err_out_unlock;
 766                 }
 767                 queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
 768                 spin_lock_bh(&queue->lock);
 769                 queue->queue_maxlen = ntohl(*queue_maxlen);
 770                 spin_unlock_bh(&queue->lock);
 771         }
 772
 773 err_out_unlock:
 774         rcu_read_unlock();
 775         return ret;
 776 }
 777
 778 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
 779         [NFQNL_MSG_PACKET]      = { .call = nfqnl_recv_unsupp,
 780                                     .attr_count = NFQA_MAX, },
 781         [NFQNL_MSG_VERDICT]     = { .call = nfqnl_recv_verdict,
 782                                     .attr_count = NFQA_MAX,
 783                                     .policy = nfqa_verdict_policy },
 784         [NFQNL_MSG_CONFIG]      = { .call = nfqnl_recv_config,
 785                                     .attr_count = NFQA_CFG_MAX,
 786                                     .policy = nfqa_cfg_policy },
 787 };
 788
 789 static const struct nfnetlink_subsystem nfqnl_subsys = {
 790         .name           = "nf_queue",
 791         .subsys_id      = NFNL_SUBSYS_QUEUE,
 792         .cb_count       = NFQNL_MSG_MAX,
 793         .cb             = nfqnl_cb,
 794 };
 795
 796 #ifdef CONFIG_PROC_FS
 797 struct iter_state {
 798         unsigned int bucket;
 799 };
 800
 801 static struct hlist_node *get_first(struct seq_file *seq)
 802 {
 803         struct iter_state *st = seq->private;
 804
 805         if (!st)
 806                 return NULL;
 807
 808         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
 809                 if (!hlist_empty(&instance_table[st->bucket]))
 810                         return instance_table[st->bucket].first;
 811         }
 812         return NULL;
 813 }
 814
 815 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
 816 {
 817         struct iter_state *st = seq->private;
 818
 819         h = h->next;
 820         while (!h) {
 821                 if (++st->bucket >= INSTANCE_BUCKETS)
 822                         return NULL;
 823
 824                 h = instance_table[st->bucket].first;
 825         }
 826         return h;
 827 }
 828
 829 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
 830 {
 831         struct hlist_node *head;
 832         head = get_first(seq);
 833
 834         if (head)
 835                 while (pos && (head = get_next(seq, head)))
 836                         pos--;
 837         return pos ? NULL : head;
 838 }
 839
 840 static void *seq_start(struct seq_file *seq, loff_t *pos)
 841         __acquires(instances_lock)
 842 {
 843         spin_lock(&instances_lock);
 844         return get_idx(seq, *pos);
 845 }
 846
 847 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
 848 {
 849         (*pos)++;
 850         return get_next(s, v);
 851 }
 852
 853 static void seq_stop(struct seq_file *s, void *v)
 854         __releases(instances_lock)
 855 {
 856         spin_unlock(&instances_lock);
 857 }
 858
 859 static int seq_show(struct seq_file *s, void *v)
 860 {
 861         const struct nfqnl_instance *inst = v;
 862
 863         return seq_printf(s, "%5d %6d %5d %1d %5d %5d %5d %8d %2d\n",
 864                           inst->queue_num,
 865                           inst->peer_pid, inst->queue_total,
 866                           inst->copy_mode, inst->copy_range,
 867                           inst->queue_dropped, inst->queue_user_dropped,
 868                           inst->id_sequence, 1);
 869 }
 870
 871 static const struct seq_operations nfqnl_seq_ops = {
 872         .start  = seq_start,
 873         .next   = seq_next,
 874         .stop   = seq_stop,
 875         .show   = seq_show,
 876 };
 877
 878 static int nfqnl_open(struct inode *inode, struct file *file)
 879 {
 880         return seq_open_private(file, &nfqnl_seq_ops,
 881                         sizeof(struct iter_state));
 882 }
 883
 884 static const struct file_operations nfqnl_file_ops = {
 885         .owner   = THIS_MODULE,
 886         .open    = nfqnl_open,
 887         .read    = seq_read,
 888         .llseek  = seq_lseek,
 889         .release = seq_release_private,
 890 };
 891
 892 #endif /* PROC_FS */
 893
 894 static int __init nfnetlink_queue_init(void)
 895 {
 896         int i, status = -ENOMEM;
 897
 898         for (i = 0; i < INSTANCE_BUCKETS; i++)
 899                 INIT_HLIST_HEAD(&instance_table[i]);
 900
 901         netlink_register_notifier(&nfqnl_rtnl_notifier);
 902         status = nfnetlink_subsys_register(&nfqnl_subsys);
 903         if (status < 0) {
 904                 printk(KERN_ERR "nf_queue: failed to create netlink socket\n");
 905                 goto cleanup_netlink_notifier;
 906         }
 907
 908 #ifdef CONFIG_PROC_FS
 909         if (!proc_create("nfnetlink_queue", 0440,
 910                          proc_net_netfilter, &nfqnl_file_ops))
 911                 goto cleanup_subsys;
 912 #endif
 913
 914         register_netdevice_notifier(&nfqnl_dev_notifier);
 915         return status;
 916
 917 #ifdef CONFIG_PROC_FS
 918 cleanup_subsys:
 919         nfnetlink_subsys_unregister(&nfqnl_subsys);
 920 #endif
 921 cleanup_netlink_notifier:
 922         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
 923         return status;
 924 }
 925
 926 static void __exit nfnetlink_queue_fini(void)
 927 {
 928         nf_unregister_queue_handlers(&nfqh);
 929         unregister_netdevice_notifier(&nfqnl_dev_notifier);
 930 #ifdef CONFIG_PROC_FS
 931         remove_proc_entry("nfnetlink_queue", proc_net_netfilter);
 932 #endif
 933         nfnetlink_subsys_unregister(&nfqnl_subsys);
 934         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
 935 }
 936
 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);