Pileus Git - ~andy/linux/blob - net/netfilter/nfnetlink_queue.c

   1 /*
   2  * This is a module which is used for queueing packets and communicating with
   3  * userspace via nfetlink.
   4  *
   5  * (C) 2005 by Harald Welte <laforge@netfilter.org>
   6  *
   7  * Based on the old ipv4-only ip_queue.c:
   8  * (C) 2000-2002 James Morris <jmorris@intercode.com.au>
   9  * (C) 2003-2005 Netfilter Core Team <coreteam@netfilter.org>
  10  *
  11  * This program is free software; you can redistribute it and/or modify
  12  * it under the terms of the GNU General Public License version 2 as
  13  * published by the Free Software Foundation.
  14  *
  15  */
  16 #include <linux/module.h>
  17 #include <linux/skbuff.h>
  18 #include <linux/init.h>
  19 #include <linux/spinlock.h>
  20 #include <linux/notifier.h>
  21 #include <linux/netdevice.h>
  22 #include <linux/netfilter.h>
  23 #include <linux/proc_fs.h>
  24 #include <linux/netfilter_ipv4.h>
  25 #include <linux/netfilter_ipv6.h>
  26 #include <linux/netfilter/nfnetlink.h>
  27 #include <linux/netfilter/nfnetlink_queue.h>
  28 #include <linux/list.h>
  29 #include <net/sock.h>
  30 #include <net/netfilter/nf_queue.h>
  31
  32 #include <asm/atomic.h>
  33
  34 #ifdef CONFIG_BRIDGE_NETFILTER
  35 #include "../bridge/br_private.h"
  36 #endif
  37
  38 #define NFQNL_QMAX_DEFAULT 1024
  39
  40 struct nfqnl_instance {
  41         struct hlist_node hlist;                /* global list of queues */
  42         struct rcu_head rcu;
  43
  44         int peer_pid;
  45         unsigned int queue_maxlen;
  46         unsigned int copy_range;
  47         unsigned int queue_total;
  48         unsigned int queue_dropped;
  49         unsigned int queue_user_dropped;
  50
  51         unsigned int id_sequence;               /* 'sequence' of pkt ids */
  52
  53         u_int16_t queue_num;                    /* number of this queue */
  54         u_int8_t copy_mode;
  55
  56         spinlock_t lock;
  57
  58         struct list_head queue_list;            /* packets in queue */
  59 };
  60
  61 typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
  62
  63 static DEFINE_SPINLOCK(instances_lock);
  64
  65 #define INSTANCE_BUCKETS        16
  66 static struct hlist_head instance_table[INSTANCE_BUCKETS] __read_mostly;
  67
  68 static inline u_int8_t instance_hashfn(u_int16_t queue_num)
  69 {
  70         return ((queue_num >> 8) | queue_num) % INSTANCE_BUCKETS;
  71 }
  72
  73 static struct nfqnl_instance *
  74 instance_lookup(u_int16_t queue_num)
  75 {
  76         struct hlist_head *head;
  77         struct hlist_node *pos;
  78         struct nfqnl_instance *inst;
  79
  80         head = &instance_table[instance_hashfn(queue_num)];
  81         hlist_for_each_entry_rcu(inst, pos, head, hlist) {
  82                 if (inst->queue_num == queue_num)
  83                         return inst;
  84         }
  85         return NULL;
  86 }
  87
  88 static struct nfqnl_instance *
  89 instance_create(u_int16_t queue_num, int pid)
  90 {
  91         struct nfqnl_instance *inst = NULL;
  92         unsigned int h;
  93
  94         spin_lock(&instances_lock);
  95         if (instance_lookup(queue_num))
  96                 goto out_unlock;
  97
  98         inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
  99         if (!inst)
 100                 goto out_unlock;
 101
 102         inst->queue_num = queue_num;
 103         inst->peer_pid = pid;
 104         inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
 105         inst->copy_range = 0xfffff;
 106         inst->copy_mode = NFQNL_COPY_NONE;
 107         spin_lock_init(&inst->lock);
 108         INIT_LIST_HEAD(&inst->queue_list);
 109         INIT_RCU_HEAD(&inst->rcu);
 110
 111         if (!try_module_get(THIS_MODULE))
 112                 goto out_free;
 113
 114         h = instance_hashfn(queue_num);
 115         hlist_add_head_rcu(&inst->hlist, &instance_table[h]);
 116
 117         spin_unlock(&instances_lock);
 118
 119         return inst;
 120
 121 out_free:
 122         kfree(inst);
 123 out_unlock:
 124         spin_unlock(&instances_lock);
 125         return NULL;
 126 }
 127
 128 static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
 129                         unsigned long data);
 130
 131 static void
 132 instance_destroy_rcu(struct rcu_head *head)
 133 {
 134         struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
 135                                                    rcu);
 136
 137         nfqnl_flush(inst, NULL, 0);
 138         kfree(inst);
 139         module_put(THIS_MODULE);
 140 }
 141
 142 static void
 143 __instance_destroy(struct nfqnl_instance *inst)
 144 {
 145         hlist_del_rcu(&inst->hlist);
 146         call_rcu(&inst->rcu, instance_destroy_rcu);
 147 }
 148
 149 static void
 150 instance_destroy(struct nfqnl_instance *inst)
 151 {
 152         spin_lock(&instances_lock);
 153         __instance_destroy(inst);
 154         spin_unlock(&instances_lock);
 155 }
 156
 157 static inline void
 158 __enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
 159 {
 160        list_add_tail(&entry->list, &queue->queue_list);
 161        queue->queue_total++;
 162 }
 163
 164 static struct nf_queue_entry *
 165 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
 166 {
 167         struct nf_queue_entry *entry = NULL, *i;
 168
 169         spin_lock_bh(&queue->lock);
 170
 171         list_for_each_entry(i, &queue->queue_list, list) {
 172                 if (i->id == id) {
 173                         entry = i;
 174                         break;
 175                 }
 176         }
 177
 178         if (entry) {
 179                 list_del(&entry->list);
 180                 queue->queue_total--;
 181         }
 182
 183         spin_unlock_bh(&queue->lock);
 184
 185         return entry;
 186 }
 187
 188 static void
 189 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
 190 {
 191         struct nf_queue_entry *entry, *next;
 192
 193         spin_lock_bh(&queue->lock);
 194         list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
 195                 if (!cmpfn || cmpfn(entry, data)) {
 196                         list_del(&entry->list);
 197                         queue->queue_total--;
 198                         nf_reinject(entry, NF_DROP);
 199                 }
 200         }
 201         spin_unlock_bh(&queue->lock);
 202 }
 203
 204 static struct sk_buff *
 205 nfqnl_build_packet_message(struct nfqnl_instance *queue,
 206                            struct nf_queue_entry *entry, int *errp)
 207 {
 208         sk_buff_data_t old_tail;
 209         size_t size;
 210         size_t data_len = 0;
 211         struct sk_buff *skb;
 212         struct nfqnl_msg_packet_hdr pmsg;
 213         struct nlmsghdr *nlh;
 214         struct nfgenmsg *nfmsg;
 215         struct sk_buff *entskb = entry->skb;
 216         struct net_device *indev;
 217         struct net_device *outdev;
 218         __be32 tmp_uint;
 219
 220         size =    NLMSG_ALIGN(sizeof(struct nfgenmsg))
 221                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
 222                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 223                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 224 #ifdef CONFIG_BRIDGE_NETFILTER
 225                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 226                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 227 #endif
 228                 + nla_total_size(sizeof(u_int32_t))     /* mark */
 229                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
 230                 + nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
 231
 232         outdev = entry->outdev;
 233
 234         spin_lock_bh(&queue->lock);
 235
 236         switch (queue->copy_mode) {
 237         case NFQNL_COPY_META:
 238         case NFQNL_COPY_NONE:
 239                 data_len = 0;
 240                 break;
 241
 242         case NFQNL_COPY_PACKET:
 243                 if ((entskb->ip_summed == CHECKSUM_PARTIAL ||
 244                      entskb->ip_summed == CHECKSUM_COMPLETE) &&
 245                     (*errp = skb_checksum_help(entskb))) {
 246                         spin_unlock_bh(&queue->lock);
 247                         return NULL;
 248                 }
 249                 if (queue->copy_range == 0
 250                     || queue->copy_range > entskb->len)
 251                         data_len = entskb->len;
 252                 else
 253                         data_len = queue->copy_range;
 254
 255                 size += nla_total_size(data_len);
 256                 break;
 257
 258         default:
 259                 *errp = -EINVAL;
 260                 spin_unlock_bh(&queue->lock);
 261                 return NULL;
 262         }
 263
 264         entry->id = queue->id_sequence++;
 265
 266         spin_unlock_bh(&queue->lock);
 267
 268         skb = alloc_skb(size, GFP_ATOMIC);
 269         if (!skb)
 270                 goto nlmsg_failure;
 271
 272         old_tail = skb->tail;
 273         nlh = NLMSG_PUT(skb, 0, 0,
 274                         NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
 275                         sizeof(struct nfgenmsg));
 276         nfmsg = NLMSG_DATA(nlh);
 277         nfmsg->nfgen_family = entry->pf;
 278         nfmsg->version = NFNETLINK_V0;
 279         nfmsg->res_id = htons(queue->queue_num);
 280
 281         pmsg.packet_id          = htonl(entry->id);
 282         pmsg.hw_protocol        = entskb->protocol;
 283         pmsg.hook               = entry->hook;
 284
 285         NLA_PUT(skb, NFQA_PACKET_HDR, sizeof(pmsg), &pmsg);
 286
 287         indev = entry->indev;
 288         if (indev) {
 289                 tmp_uint = htonl(indev->ifindex);
 290 #ifndef CONFIG_BRIDGE_NETFILTER
 291                 NLA_PUT(skb, NFQA_IFINDEX_INDEV, sizeof(tmp_uint), &tmp_uint);
 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(skb, NFQA_IFINDEX_PHYSINDEV, sizeof(tmp_uint),
 298                                 &tmp_uint);
 299                         /* this is the bridge group "brX" */
 300                         tmp_uint = htonl(indev->br_port->br->dev->ifindex);
 301                         NLA_PUT(skb, NFQA_IFINDEX_INDEV, sizeof(tmp_uint),
 302                                 &tmp_uint);
 303                 } else {
 304                         /* Case 2: indev is bridge group, we need to look for
 305                          * physical device (when called from ipv4) */
 306                         NLA_PUT(skb, NFQA_IFINDEX_INDEV, sizeof(tmp_uint),
 307                                 &tmp_uint);
 308                         if (entskb->nf_bridge
 309                             && entskb->nf_bridge->physindev) {
 310                                 tmp_uint = htonl(entskb->nf_bridge->physindev->ifindex);
 311                                 NLA_PUT(skb, NFQA_IFINDEX_PHYSINDEV,
 312                                         sizeof(tmp_uint), &tmp_uint);
 313                         }
 314                 }
 315 #endif
 316         }
 317
 318         if (outdev) {
 319                 tmp_uint = htonl(outdev->ifindex);
 320 #ifndef CONFIG_BRIDGE_NETFILTER
 321                 NLA_PUT(skb, NFQA_IFINDEX_OUTDEV, sizeof(tmp_uint), &tmp_uint);
 322 #else
 323                 if (entry->pf == PF_BRIDGE) {
 324                         /* Case 1: outdev is physical output device, we need to
 325                          * look for bridge group (when called from
 326                          * netfilter_bridge) */
 327                         NLA_PUT(skb, NFQA_IFINDEX_PHYSOUTDEV, sizeof(tmp_uint),
 328                                 &tmp_uint);
 329                         /* this is the bridge group "brX" */
 330                         tmp_uint = htonl(outdev->br_port->br->dev->ifindex);
 331                         NLA_PUT(skb, NFQA_IFINDEX_OUTDEV, sizeof(tmp_uint),
 332                                 &tmp_uint);
 333                 } else {
 334                         /* Case 2: outdev is bridge group, we need to look for
 335                          * physical output device (when called from ipv4) */
 336                         NLA_PUT(skb, NFQA_IFINDEX_OUTDEV, sizeof(tmp_uint),
 337                                 &tmp_uint);
 338                         if (entskb->nf_bridge
 339                             && entskb->nf_bridge->physoutdev) {
 340                                 tmp_uint = htonl(entskb->nf_bridge->physoutdev->ifindex);
 341                                 NLA_PUT(skb, NFQA_IFINDEX_PHYSOUTDEV,
 342                                         sizeof(tmp_uint), &tmp_uint);
 343                         }
 344                 }
 345 #endif
 346         }
 347
 348         if (entskb->mark) {
 349                 tmp_uint = htonl(entskb->mark);
 350                 NLA_PUT(skb, NFQA_MARK, sizeof(u_int32_t), &tmp_uint);
 351         }
 352
 353         if (indev && entskb->dev) {
 354                 struct nfqnl_msg_packet_hw phw;
 355                 int len = dev_parse_header(entskb, phw.hw_addr);
 356                 if (len) {
 357                         phw.hw_addrlen = htons(len);
 358                         NLA_PUT(skb, NFQA_HWADDR, sizeof(phw), &phw);
 359                 }
 360         }
 361
 362         if (entskb->tstamp.tv64) {
 363                 struct nfqnl_msg_packet_timestamp ts;
 364                 struct timeval tv = ktime_to_timeval(entskb->tstamp);
 365                 ts.sec = cpu_to_be64(tv.tv_sec);
 366                 ts.usec = cpu_to_be64(tv.tv_usec);
 367
 368                 NLA_PUT(skb, NFQA_TIMESTAMP, sizeof(ts), &ts);
 369         }
 370
 371         if (data_len) {
 372                 struct nlattr *nla;
 373                 int size = nla_attr_size(data_len);
 374
 375                 if (skb_tailroom(skb) < nla_total_size(data_len)) {
 376                         printk(KERN_WARNING "nf_queue: no tailroom!\n");
 377                         goto nlmsg_failure;
 378                 }
 379
 380                 nla = (struct nlattr *)skb_put(skb, nla_total_size(data_len));
 381                 nla->nla_type = NFQA_PAYLOAD;
 382                 nla->nla_len = size;
 383
 384                 if (skb_copy_bits(entskb, 0, nla_data(nla), data_len))
 385                         BUG();
 386         }
 387
 388         nlh->nlmsg_len = skb->tail - old_tail;
 389         return skb;
 390
 391 nlmsg_failure:
 392 nla_put_failure:
 393         if (skb)
 394                 kfree_skb(skb);
 395         *errp = -EINVAL;
 396         if (net_ratelimit())
 397                 printk(KERN_ERR "nf_queue: error creating packet message\n");
 398         return NULL;
 399 }
 400
 401 static int
 402 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 403 {
 404         int status = -EINVAL;
 405         struct sk_buff *nskb;
 406         struct nfqnl_instance *queue;
 407
 408         /* rcu_read_lock()ed by nf_hook_slow() */
 409         queue = instance_lookup(queuenum);
 410         if (!queue)
 411                 return -EINVAL;
 412
 413         if (queue->copy_mode == NFQNL_COPY_NONE)
 414                 return -EAGAIN;
 415
 416         nskb = nfqnl_build_packet_message(queue, entry, &status);
 417         if (nskb == NULL)
 418                 return status;
 419
 420         spin_lock_bh(&queue->lock);
 421
 422         if (!queue->peer_pid)
 423                 goto err_out_free_nskb;
 424
 425         if (queue->queue_total >= queue->queue_maxlen) {
 426                 queue->queue_dropped++;
 427                 status = -ENOSPC;
 428                 if (net_ratelimit())
 429                           printk(KERN_WARNING "nf_queue: full at %d entries, "
 430                                  "dropping packets(s). Dropped: %d\n",
 431                                  queue->queue_total, queue->queue_dropped);
 432                 goto err_out_free_nskb;
 433         }
 434
 435         /* nfnetlink_unicast will either free the nskb or add it to a socket */
 436         status = nfnetlink_unicast(nskb, queue->peer_pid, MSG_DONTWAIT);
 437         if (status < 0) {
 438                 queue->queue_user_dropped++;
 439                 goto err_out_unlock;
 440         }
 441
 442         __enqueue_entry(queue, entry);
 443
 444         spin_unlock_bh(&queue->lock);
 445         return status;
 446
 447 err_out_free_nskb:
 448         kfree_skb(nskb);
 449
 450 err_out_unlock:
 451         spin_unlock_bh(&queue->lock);
 452         return status;
 453 }
 454
 455 static int
 456 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e)
 457 {
 458         int diff;
 459         int err;
 460
 461         diff = data_len - e->skb->len;
 462         if (diff < 0) {
 463                 if (pskb_trim(e->skb, data_len))
 464                         return -ENOMEM;
 465         } else if (diff > 0) {
 466                 if (data_len > 0xFFFF)
 467                         return -EINVAL;
 468                 if (diff > skb_tailroom(e->skb)) {
 469                         err = pskb_expand_head(e->skb, 0,
 470                                                diff - skb_tailroom(e->skb),
 471                                                GFP_ATOMIC);
 472                         if (err) {
 473                                 printk(KERN_WARNING "nf_queue: OOM "
 474                                       "in mangle, dropping packet\n");
 475                                 return err;
 476                         }
 477                 }
 478                 skb_put(e->skb, diff);
 479         }
 480         if (!skb_make_writable(e->skb, data_len))
 481                 return -ENOMEM;
 482         skb_copy_to_linear_data(e->skb, data, data_len);
 483         e->skb->ip_summed = CHECKSUM_NONE;
 484         return 0;
 485 }
 486
 487 static int
 488 nfqnl_set_mode(struct nfqnl_instance *queue,
 489                unsigned char mode, unsigned int range)
 490 {
 491         int status = 0;
 492
 493         spin_lock_bh(&queue->lock);
 494         switch (mode) {
 495         case NFQNL_COPY_NONE:
 496         case NFQNL_COPY_META:
 497                 queue->copy_mode = mode;
 498                 queue->copy_range = 0;
 499                 break;
 500
 501         case NFQNL_COPY_PACKET:
 502                 queue->copy_mode = mode;
 503                 /* we're using struct nlattr which has 16bit nla_len */
 504                 if (range > 0xffff)
 505                         queue->copy_range = 0xffff;
 506                 else
 507                         queue->copy_range = range;
 508                 break;
 509
 510         default:
 511                 status = -EINVAL;
 512
 513         }
 514         spin_unlock_bh(&queue->lock);
 515
 516         return status;
 517 }
 518
 519 static int
 520 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
 521 {
 522         if (entry->indev)
 523                 if (entry->indev->ifindex == ifindex)
 524                         return 1;
 525         if (entry->outdev)
 526                 if (entry->outdev->ifindex == ifindex)
 527                         return 1;
 528 #ifdef CONFIG_BRIDGE_NETFILTER
 529         if (entry->skb->nf_bridge) {
 530                 if (entry->skb->nf_bridge->physindev &&
 531                     entry->skb->nf_bridge->physindev->ifindex == ifindex)
 532                         return 1;
 533                 if (entry->skb->nf_bridge->physoutdev &&
 534                     entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
 535                         return 1;
 536         }
 537 #endif
 538         return 0;
 539 }
 540
 541 /* drop all packets with either indev or outdev == ifindex from all queue
 542  * instances */
 543 static void
 544 nfqnl_dev_drop(int ifindex)
 545 {
 546         int i;
 547
 548         rcu_read_lock();
 549
 550         for (i = 0; i < INSTANCE_BUCKETS; i++) {
 551                 struct hlist_node *tmp;
 552                 struct nfqnl_instance *inst;
 553                 struct hlist_head *head = &instance_table[i];
 554
 555                 hlist_for_each_entry_rcu(inst, tmp, head, hlist)
 556                         nfqnl_flush(inst, dev_cmp, ifindex);
 557         }
 558
 559         rcu_read_unlock();
 560 }
 561
 562 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
 563
 564 static int
 565 nfqnl_rcv_dev_event(struct notifier_block *this,
 566                     unsigned long event, void *ptr)
 567 {
 568         struct net_device *dev = ptr;
 569
 570         if (dev->nd_net != &init_net)
 571                 return NOTIFY_DONE;
 572
 573         /* Drop any packets associated with the downed device */
 574         if (event == NETDEV_DOWN)
 575                 nfqnl_dev_drop(dev->ifindex);
 576         return NOTIFY_DONE;
 577 }
 578
 579 static struct notifier_block nfqnl_dev_notifier = {
 580         .notifier_call  = nfqnl_rcv_dev_event,
 581 };
 582
 583 static int
 584 nfqnl_rcv_nl_event(struct notifier_block *this,
 585                    unsigned long event, void *ptr)
 586 {
 587         struct netlink_notify *n = ptr;
 588
 589         if (event == NETLINK_URELEASE &&
 590             n->protocol == NETLINK_NETFILTER && n->pid) {
 591                 int i;
 592
 593                 /* destroy all instances for this pid */
 594                 spin_lock(&instances_lock);
 595                 for (i = 0; i < INSTANCE_BUCKETS; i++) {
 596                         struct hlist_node *tmp, *t2;
 597                         struct nfqnl_instance *inst;
 598                         struct hlist_head *head = &instance_table[i];
 599
 600                         hlist_for_each_entry_safe(inst, tmp, t2, head, hlist) {
 601                                 if ((n->net == &init_net) &&
 602                                     (n->pid == inst->peer_pid))
 603                                         __instance_destroy(inst);
 604                         }
 605                 }
 606                 spin_unlock(&instances_lock);
 607         }
 608         return NOTIFY_DONE;
 609 }
 610
 611 static struct notifier_block nfqnl_rtnl_notifier = {
 612         .notifier_call  = nfqnl_rcv_nl_event,
 613 };
 614
 615 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
 616         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 617         [NFQA_MARK]             = { .type = NLA_U32 },
 618         [NFQA_PAYLOAD]          = { .type = NLA_UNSPEC },
 619 };
 620
 621 static int
 622 nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
 623                    struct nlmsghdr *nlh, struct nlattr *nfqa[])
 624 {
 625         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 626         u_int16_t queue_num = ntohs(nfmsg->res_id);
 627
 628         struct nfqnl_msg_verdict_hdr *vhdr;
 629         struct nfqnl_instance *queue;
 630         unsigned int verdict;
 631         struct nf_queue_entry *entry;
 632         int err;
 633
 634         rcu_read_lock();
 635         queue = instance_lookup(queue_num);
 636         if (!queue) {
 637                 err = -ENODEV;
 638                 goto err_out_unlock;
 639         }
 640
 641         if (queue->peer_pid != NETLINK_CB(skb).pid) {
 642                 err = -EPERM;
 643                 goto err_out_unlock;
 644         }
 645
 646         if (!nfqa[NFQA_VERDICT_HDR]) {
 647                 err = -EINVAL;
 648                 goto err_out_unlock;
 649         }
 650
 651         vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
 652         verdict = ntohl(vhdr->verdict);
 653
 654         if ((verdict & NF_VERDICT_MASK) > NF_MAX_VERDICT) {
 655                 err = -EINVAL;
 656                 goto err_out_unlock;
 657         }
 658
 659         entry = find_dequeue_entry(queue, ntohl(vhdr->id));
 660         if (entry == NULL) {
 661                 err = -ENOENT;
 662                 goto err_out_unlock;
 663         }
 664         rcu_read_unlock();
 665
 666         if (nfqa[NFQA_PAYLOAD]) {
 667                 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
 668                                  nla_len(nfqa[NFQA_PAYLOAD]), entry) < 0)
 669                         verdict = NF_DROP;
 670         }
 671
 672         if (nfqa[NFQA_MARK])
 673                 entry->skb->mark = ntohl(*(__be32 *)
 674                                          nla_data(nfqa[NFQA_MARK]));
 675
 676         nf_reinject(entry, verdict);
 677         return 0;
 678
 679 err_out_unlock:
 680         rcu_read_unlock();
 681         return err;
 682 }
 683
 684 static int
 685 nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
 686                   struct nlmsghdr *nlh, struct nlattr *nfqa[])
 687 {
 688         return -ENOTSUPP;
 689 }
 690
 691 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
 692         [NFQA_CFG_CMD]          = { .len = sizeof(struct nfqnl_msg_config_cmd) },
 693         [NFQA_CFG_PARAMS]       = { .len = sizeof(struct nfqnl_msg_config_params) },
 694 };
 695
 696 static const struct nf_queue_handler nfqh = {
 697         .name   = "nf_queue",
 698         .outfn  = &nfqnl_enqueue_packet,
 699 };
 700
 701 static int
 702 nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
 703                   struct nlmsghdr *nlh, struct nlattr *nfqa[])
 704 {
 705         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 706         u_int16_t queue_num = ntohs(nfmsg->res_id);
 707         struct nfqnl_instance *queue;
 708         struct nfqnl_msg_config_cmd *cmd = NULL;
 709         int ret = 0;
 710
 711         if (nfqa[NFQA_CFG_CMD]) {
 712                 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
 713
 714                 /* Commands without queue context - might sleep */
 715                 switch (cmd->command) {
 716                 case NFQNL_CFG_CMD_PF_BIND:
 717                         ret = nf_register_queue_handler(ntohs(cmd->pf),
 718                                                         &nfqh);
 719                         break;
 720                 case NFQNL_CFG_CMD_PF_UNBIND:
 721                         ret = nf_unregister_queue_handler(ntohs(cmd->pf),
 722                                                           &nfqh);
 723                         break;
 724                 default:
 725                         break;
 726                 }
 727
 728                 if (ret < 0)
 729                         return ret;
 730         }
 731
 732         rcu_read_lock();
 733         queue = instance_lookup(queue_num);
 734         if (queue && queue->peer_pid != NETLINK_CB(skb).pid) {
 735                 ret = -EPERM;
 736                 goto err_out_unlock;
 737         }
 738
 739         if (cmd != NULL) {
 740                 switch (cmd->command) {
 741                 case NFQNL_CFG_CMD_BIND:
 742                         if (queue) {
 743                                 ret = -EBUSY;
 744                                 goto err_out_unlock;
 745                         }
 746                         queue = instance_create(queue_num, NETLINK_CB(skb).pid);
 747                         if (!queue) {
 748                                 ret = -EINVAL;
 749                                 goto err_out_unlock;
 750                         }
 751                         break;
 752                 case NFQNL_CFG_CMD_UNBIND:
 753                         if (!queue) {
 754                                 ret = -ENODEV;
 755                                 goto err_out_unlock;
 756                         }
 757                         instance_destroy(queue);
 758                         break;
 759                 case NFQNL_CFG_CMD_PF_BIND:
 760                 case NFQNL_CFG_CMD_PF_UNBIND:
 761                         break;
 762                 default:
 763                         ret = -EINVAL;
 764                         break;
 765                 }
 766         }
 767
 768         if (nfqa[NFQA_CFG_PARAMS]) {
 769                 struct nfqnl_msg_config_params *params;
 770
 771                 if (!queue) {
 772                         ret = -ENODEV;
 773                         goto err_out_unlock;
 774                 }
 775                 params = nla_data(nfqa[NFQA_CFG_PARAMS]);
 776                 nfqnl_set_mode(queue, params->copy_mode,
 777                                 ntohl(params->copy_range));
 778         }
 779
 780         if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
 781                 __be32 *queue_maxlen;
 782
 783                 if (!queue) {
 784                         ret = -ENODEV;
 785                         goto err_out_unlock;
 786                 }
 787                 queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
 788                 spin_lock_bh(&queue->lock);
 789                 queue->queue_maxlen = ntohl(*queue_maxlen);
 790                 spin_unlock_bh(&queue->lock);
 791         }
 792
 793 err_out_unlock:
 794         rcu_read_unlock();
 795         return ret;
 796 }
 797
 798 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
 799         [NFQNL_MSG_PACKET]      = { .call = nfqnl_recv_unsupp,
 800                                     .attr_count = NFQA_MAX, },
 801         [NFQNL_MSG_VERDICT]     = { .call = nfqnl_recv_verdict,
 802                                     .attr_count = NFQA_MAX,
 803                                     .policy = nfqa_verdict_policy },
 804         [NFQNL_MSG_CONFIG]      = { .call = nfqnl_recv_config,
 805                                     .attr_count = NFQA_CFG_MAX,
 806                                     .policy = nfqa_cfg_policy },
 807 };
 808
 809 static const struct nfnetlink_subsystem nfqnl_subsys = {
 810         .name           = "nf_queue",
 811         .subsys_id      = NFNL_SUBSYS_QUEUE,
 812         .cb_count       = NFQNL_MSG_MAX,
 813         .cb             = nfqnl_cb,
 814 };
 815
 816 #ifdef CONFIG_PROC_FS
 817 struct iter_state {
 818         unsigned int bucket;
 819 };
 820
 821 static struct hlist_node *get_first(struct seq_file *seq)
 822 {
 823         struct iter_state *st = seq->private;
 824
 825         if (!st)
 826                 return NULL;
 827
 828         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
 829                 if (!hlist_empty(&instance_table[st->bucket]))
 830                         return instance_table[st->bucket].first;
 831         }
 832         return NULL;
 833 }
 834
 835 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
 836 {
 837         struct iter_state *st = seq->private;
 838
 839         h = h->next;
 840         while (!h) {
 841                 if (++st->bucket >= INSTANCE_BUCKETS)
 842                         return NULL;
 843
 844                 h = instance_table[st->bucket].first;
 845         }
 846         return h;
 847 }
 848
 849 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
 850 {
 851         struct hlist_node *head;
 852         head = get_first(seq);
 853
 854         if (head)
 855                 while (pos && (head = get_next(seq, head)))
 856                         pos--;
 857         return pos ? NULL : head;
 858 }
 859
 860 static void *seq_start(struct seq_file *seq, loff_t *pos)
 861 {
 862         spin_lock(&instances_lock);
 863         return get_idx(seq, *pos);
 864 }
 865
 866 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
 867 {
 868         (*pos)++;
 869         return get_next(s, v);
 870 }
 871
 872 static void seq_stop(struct seq_file *s, void *v)
 873 {
 874         spin_unlock(&instances_lock);
 875 }
 876
 877 static int seq_show(struct seq_file *s, void *v)
 878 {
 879         const struct nfqnl_instance *inst = v;
 880
 881         return seq_printf(s, "%5d %6d %5d %1d %5d %5d %5d %8d %2d\n",
 882                           inst->queue_num,
 883                           inst->peer_pid, inst->queue_total,
 884                           inst->copy_mode, inst->copy_range,
 885                           inst->queue_dropped, inst->queue_user_dropped,
 886                           inst->id_sequence, 1);
 887 }
 888
 889 static const struct seq_operations nfqnl_seq_ops = {
 890         .start  = seq_start,
 891         .next   = seq_next,
 892         .stop   = seq_stop,
 893         .show   = seq_show,
 894 };
 895
 896 static int nfqnl_open(struct inode *inode, struct file *file)
 897 {
 898         return seq_open_private(file, &nfqnl_seq_ops,
 899                         sizeof(struct iter_state));
 900 }
 901
 902 static const struct file_operations nfqnl_file_ops = {
 903         .owner   = THIS_MODULE,
 904         .open    = nfqnl_open,
 905         .read    = seq_read,
 906         .llseek  = seq_lseek,
 907         .release = seq_release_private,
 908 };
 909
 910 #endif /* PROC_FS */
 911
 912 static int __init nfnetlink_queue_init(void)
 913 {
 914         int i, status = -ENOMEM;
 915 #ifdef CONFIG_PROC_FS
 916         struct proc_dir_entry *proc_nfqueue;
 917 #endif
 918
 919         for (i = 0; i < INSTANCE_BUCKETS; i++)
 920                 INIT_HLIST_HEAD(&instance_table[i]);
 921
 922         netlink_register_notifier(&nfqnl_rtnl_notifier);
 923         status = nfnetlink_subsys_register(&nfqnl_subsys);
 924         if (status < 0) {
 925                 printk(KERN_ERR "nf_queue: failed to create netlink socket\n");
 926                 goto cleanup_netlink_notifier;
 927         }
 928
 929 #ifdef CONFIG_PROC_FS
 930         proc_nfqueue = create_proc_entry("nfnetlink_queue", 0440,
 931                                          proc_net_netfilter);
 932         if (!proc_nfqueue)
 933                 goto cleanup_subsys;
 934         proc_nfqueue->proc_fops = &nfqnl_file_ops;
 935 #endif
 936
 937         register_netdevice_notifier(&nfqnl_dev_notifier);
 938         return status;
 939
 940 #ifdef CONFIG_PROC_FS
 941 cleanup_subsys:
 942         nfnetlink_subsys_unregister(&nfqnl_subsys);
 943 #endif
 944 cleanup_netlink_notifier:
 945         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
 946         return status;
 947 }
 948
 949 static void __exit nfnetlink_queue_fini(void)
 950 {
 951         nf_unregister_queue_handlers(&nfqh);
 952         unregister_netdevice_notifier(&nfqnl_dev_notifier);
 953 #ifdef CONFIG_PROC_FS
 954         remove_proc_entry("nfnetlink_queue", proc_net_netfilter);
 955 #endif
 956         nfnetlink_subsys_unregister(&nfqnl_subsys);
 957         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
 958 }
 959
 960 MODULE_DESCRIPTION("netfilter packet queue handler");
 961 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
 962 MODULE_LICENSE("GPL");
 963 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
 964
 965 module_init(nfnetlink_queue_init);
 966 module_exit(nfnetlink_queue_fini);