net/netfilter/nfnetlink_queue.c

   1 /*
   2  * This is a module which is used for queueing packets and communicating with
   3  * userspace via nfnetlink.
   4  *
   5  * (C) 2005 by Harald Welte <laforge@netfilter.org>
   6  * (C) 2007 by Patrick McHardy <kaber@trash.net>
   7  *
   8  * Based on the old ipv4-only ip_queue.c:
   9  * (C) 2000-2002 James Morris <jmorris@intercode.com.au>
  10  * (C) 2003-2005 Netfilter Core Team <coreteam@netfilter.org>
  11  *
  12  * This program is free software; you can redistribute it and/or modify
  13  * it under the terms of the GNU General Public License version 2 as
  14  * published by the Free Software Foundation.
  15  *
  16  */
  17 #include <linux/module.h>
  18 #include <linux/skbuff.h>
  19 #include <linux/init.h>
  20 #include <linux/spinlock.h>
  21 #include <linux/slab.h>
  22 #include <linux/notifier.h>
  23 #include <linux/netdevice.h>
  24 #include <linux/netfilter.h>
  25 #include <linux/proc_fs.h>
  26 #include <linux/netfilter_ipv4.h>
  27 #include <linux/netfilter_ipv6.h>
  28 #include <linux/netfilter/nfnetlink.h>
  29 #include <linux/netfilter/nfnetlink_queue.h>
  30 #include <linux/list.h>
  31 #include <net/sock.h>
  32 #include <net/netfilter/nf_queue.h>
  33
  34 #include <asm/atomic.h>
  35
  36 #ifdef CONFIG_BRIDGE_NETFILTER
  37 #include "../bridge/br_private.h"
  38 #endif
  39
  40 #define NFQNL_QMAX_DEFAULT 1024
  41
  42 struct nfqnl_instance {
  43         struct hlist_node hlist;                /* global list of queues */
  44         struct rcu_head rcu;
  45
  46         int peer_pid;
  47         unsigned int queue_maxlen;
  48         unsigned int copy_range;
  49         unsigned int queue_dropped;
  50         unsigned int queue_user_dropped;
  51
  52
  53         u_int16_t queue_num;                    /* number of this queue */
  54         u_int8_t copy_mode;
  55 /*
  56  * Following fields are dirtied for each queued packet,
  57  * keep them in same cache line if possible.
  58  */
  59         spinlock_t      lock;
  60         unsigned int    queue_total;
  61         unsigned int    id_sequence;            /* 'sequence' of pkt ids */
  62         struct list_head queue_list;            /* packets in queue */
  63 };
  64
  65 typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
  66
  67 static DEFINE_SPINLOCK(instances_lock);
  68
  69 #define INSTANCE_BUCKETS        16
  70 static struct hlist_head instance_table[INSTANCE_BUCKETS] __read_mostly;
  71
  72 static inline u_int8_t instance_hashfn(u_int16_t queue_num)
  73 {
  74         return ((queue_num >> 8) | queue_num) % INSTANCE_BUCKETS;
  75 }
  76
  77 static struct nfqnl_instance *
  78 instance_lookup(u_int16_t queue_num)
  79 {
  80         struct hlist_head *head;
  81         struct hlist_node *pos;
  82         struct nfqnl_instance *inst;
  83
  84         head = &instance_table[instance_hashfn(queue_num)];
  85         hlist_for_each_entry_rcu(inst, pos, head, hlist) {
  86                 if (inst->queue_num == queue_num)
  87                         return inst;
  88         }
  89         return NULL;
  90 }
  91
  92 static struct nfqnl_instance *
  93 instance_create(u_int16_t queue_num, int pid)
  94 {
  95         struct nfqnl_instance *inst;
  96         unsigned int h;
  97         int err;
  98
  99         spin_lock(&instances_lock);
 100         if (instance_lookup(queue_num)) {
 101                 err = -EEXIST;
 102                 goto out_unlock;
 103         }
 104
 105         inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
 106         if (!inst) {
 107                 err = -ENOMEM;
 108                 goto out_unlock;
 109         }
 110
 111         inst->queue_num = queue_num;
 112         inst->peer_pid = pid;
 113         inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
 114         inst->copy_range = 0xfffff;
 115         inst->copy_mode = NFQNL_COPY_NONE;
 116         spin_lock_init(&inst->lock);
 117         INIT_LIST_HEAD(&inst->queue_list);
 118
 119         if (!try_module_get(THIS_MODULE)) {
 120                 err = -EAGAIN;
 121                 goto out_free;
 122         }
 123
 124         h = instance_hashfn(queue_num);
 125         hlist_add_head_rcu(&inst->hlist, &instance_table[h]);
 126
 127         spin_unlock(&instances_lock);
 128
 129         return inst;
 130
 131 out_free:
 132         kfree(inst);
 133 out_unlock:
 134         spin_unlock(&instances_lock);
 135         return ERR_PTR(err);
 136 }
 137
 138 static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
 139                         unsigned long data);
 140
 141 static void
 142 instance_destroy_rcu(struct rcu_head *head)
 143 {
 144         struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
 145                                                    rcu);
 146
 147         nfqnl_flush(inst, NULL, 0);
 148         kfree(inst);
 149         module_put(THIS_MODULE);
 150 }
 151
 152 static void
 153 __instance_destroy(struct nfqnl_instance *inst)
 154 {
 155         hlist_del_rcu(&inst->hlist);
 156         call_rcu(&inst->rcu, instance_destroy_rcu);
 157 }
 158
 159 static void
 160 instance_destroy(struct nfqnl_instance *inst)
 161 {
 162         spin_lock(&instances_lock);
 163         __instance_destroy(inst);
 164         spin_unlock(&instances_lock);
 165 }
 166
 167 static inline void
 168 __enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
 169 {
 170        list_add_tail(&entry->list, &queue->queue_list);
 171        queue->queue_total++;
 172 }
 173
 174 static struct nf_queue_entry *
 175 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
 176 {
 177         struct nf_queue_entry *entry = NULL, *i;
 178
 179         spin_lock_bh(&queue->lock);
 180
 181         list_for_each_entry(i, &queue->queue_list, list) {
 182                 if (i->id == id) {
 183                         entry = i;
 184                         break;
 185                 }
 186         }
 187
 188         if (entry) {
 189                 list_del(&entry->list);
 190                 queue->queue_total--;
 191         }
 192
 193         spin_unlock_bh(&queue->lock);
 194
 195         return entry;
 196 }
 197
 198 static void
 199 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
 200 {
 201         struct nf_queue_entry *entry, *next;
 202
 203         spin_lock_bh(&queue->lock);
 204         list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
 205                 if (!cmpfn || cmpfn(entry, data)) {
 206                         list_del(&entry->list);
 207                         queue->queue_total--;
 208                         nf_reinject(entry, NF_DROP);
 209                 }
 210         }
 211         spin_unlock_bh(&queue->lock);
 212 }
 213
 214 static struct sk_buff *
 215 nfqnl_build_packet_message(struct nfqnl_instance *queue,
 216                            struct nf_queue_entry *entry,
 217                            __be32 **packet_id_ptr)
 218 {
 219         sk_buff_data_t old_tail;
 220         size_t size;
 221         size_t data_len = 0;
 222         struct sk_buff *skb;
 223         struct nlattr *nla;
 224         struct nfqnl_msg_packet_hdr *pmsg;
 225         struct nlmsghdr *nlh;
 226         struct nfgenmsg *nfmsg;
 227         struct sk_buff *entskb = entry->skb;
 228         struct net_device *indev;
 229         struct net_device *outdev;
 230
 231         size =    NLMSG_SPACE(sizeof(struct nfgenmsg))
 232                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
 233                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 234                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 235 #ifdef CONFIG_BRIDGE_NETFILTER
 236                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 237                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 238 #endif
 239                 + nla_total_size(sizeof(u_int32_t))     /* mark */
 240                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
 241                 + nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
 242
 243         outdev = entry->outdev;
 244
 245         switch ((enum nfqnl_config_mode)ACCESS_ONCE(queue->copy_mode)) {
 246         case NFQNL_COPY_META:
 247         case NFQNL_COPY_NONE:
 248                 break;
 249
 250         case NFQNL_COPY_PACKET:
 251                 if (entskb->ip_summed == CHECKSUM_PARTIAL &&
 252                     skb_checksum_help(entskb))
 253                         return NULL;
 254
 255                 data_len = ACCESS_ONCE(queue->copy_range);
 256                 if (data_len == 0 || data_len > entskb->len)
 257                         data_len = entskb->len;
 258
 259                 size += nla_total_size(data_len);
 260                 break;
 261         }
 262
 263
 264         skb = alloc_skb(size, GFP_ATOMIC);
 265         if (!skb)
 266                 goto nlmsg_failure;
 267
 268         old_tail = skb->tail;
 269         nlh = NLMSG_PUT(skb, 0, 0,
 270                         NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
 271                         sizeof(struct nfgenmsg));
 272         nfmsg = NLMSG_DATA(nlh);
 273         nfmsg->nfgen_family = entry->pf;
 274         nfmsg->version = NFNETLINK_V0;
 275         nfmsg->res_id = htons(queue->queue_num);
 276
 277         nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
 278         pmsg = nla_data(nla);
 279         pmsg->hw_protocol       = entskb->protocol;
 280         pmsg->hook              = entry->hook;
 281         *packet_id_ptr          = &pmsg->packet_id;
 282
 283         indev = entry->indev;
 284         if (indev) {
 285 #ifndef CONFIG_BRIDGE_NETFILTER
 286                 NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex));
 287 #else
 288                 if (entry->pf == PF_BRIDGE) {
 289                         /* Case 1: indev is physical input device, we need to
 290                          * look for bridge group (when called from
 291                          * netfilter_bridge) */
 292                         NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSINDEV,
 293                                      htonl(indev->ifindex));
 294                         /* this is the bridge group "brX" */
 295                         /* rcu_read_lock()ed by __nf_queue */
 296                         NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV,
 297                                      htonl(br_port_get_rcu(indev)->br->dev->ifindex));
 298                 } else {
 299                         /* Case 2: indev is bridge group, we need to look for
 300                          * physical device (when called from ipv4) */
 301                         NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV,
 302                                      htonl(indev->ifindex));
 303                         if (entskb->nf_bridge && entskb->nf_bridge->physindev)
 304                                 NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSINDEV,
 305                                              htonl(entskb->nf_bridge->physindev->ifindex));
 306                 }
 307 #endif
 308         }
 309
 310         if (outdev) {
 311 #ifndef CONFIG_BRIDGE_NETFILTER
 312                 NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex));
 313 #else
 314                 if (entry->pf == PF_BRIDGE) {
 315                         /* Case 1: outdev is physical output device, we need to
 316                          * look for bridge group (when called from
 317                          * netfilter_bridge) */
 318                         NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 319                                      htonl(outdev->ifindex));
 320                         /* this is the bridge group "brX" */
 321                         /* rcu_read_lock()ed by __nf_queue */
 322                         NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV,
 323                                      htonl(br_port_get_rcu(outdev)->br->dev->ifindex));
 324                 } else {
 325                         /* Case 2: outdev is bridge group, we need to look for
 326                          * physical output device (when called from ipv4) */
 327                         NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV,
 328                                      htonl(outdev->ifindex));
 329                         if (entskb->nf_bridge && entskb->nf_bridge->physoutdev)
 330                                 NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 331                                              htonl(entskb->nf_bridge->physoutdev->ifindex));
 332                 }
 333 #endif
 334         }
 335
 336         if (entskb->mark)
 337                 NLA_PUT_BE32(skb, NFQA_MARK, htonl(entskb->mark));
 338
 339         if (indev && entskb->dev) {
 340                 struct nfqnl_msg_packet_hw phw;
 341                 int len = dev_parse_header(entskb, phw.hw_addr);
 342                 if (len) {
 343                         phw.hw_addrlen = htons(len);
 344                         NLA_PUT(skb, NFQA_HWADDR, sizeof(phw), &phw);
 345                 }
 346         }
 347
 348         if (entskb->tstamp.tv64) {
 349                 struct nfqnl_msg_packet_timestamp ts;
 350                 struct timeval tv = ktime_to_timeval(entskb->tstamp);
 351                 ts.sec = cpu_to_be64(tv.tv_sec);
 352                 ts.usec = cpu_to_be64(tv.tv_usec);
 353
 354                 NLA_PUT(skb, NFQA_TIMESTAMP, sizeof(ts), &ts);
 355         }
 356
 357         if (data_len) {
 358                 struct nlattr *nla;
 359                 int sz = nla_attr_size(data_len);
 360
 361                 if (skb_tailroom(skb) < nla_total_size(data_len)) {
 362                         printk(KERN_WARNING "nf_queue: no tailroom!\n");
 363                         goto nlmsg_failure;
 364                 }
 365
 366                 nla = (struct nlattr *)skb_put(skb, nla_total_size(data_len));
 367                 nla->nla_type = NFQA_PAYLOAD;
 368                 nla->nla_len = sz;
 369
 370                 if (skb_copy_bits(entskb, 0, nla_data(nla), data_len))
 371                         BUG();
 372         }
 373
 374         nlh->nlmsg_len = skb->tail - old_tail;
 375         return skb;
 376
 377 nlmsg_failure:
 378 nla_put_failure:
 379         if (skb)
 380                 kfree_skb(skb);
 381         if (net_ratelimit())
 382                 printk(KERN_ERR "nf_queue: error creating packet message\n");
 383         return NULL;
 384 }
 385
 386 static int
 387 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 388 {
 389         struct sk_buff *nskb;
 390         struct nfqnl_instance *queue;
 391         int err = -ENOBUFS;
 392         __be32 *packet_id_ptr;
 393
 394         /* rcu_read_lock()ed by nf_hook_slow() */
 395         queue = instance_lookup(queuenum);
 396         if (!queue) {
 397                 err = -ESRCH;
 398                 goto err_out;
 399         }
 400
 401         if (queue->copy_mode == NFQNL_COPY_NONE) {
 402                 err = -EINVAL;
 403                 goto err_out;
 404         }
 405
 406         nskb = nfqnl_build_packet_message(queue, entry, &packet_id_ptr);
 407         if (nskb == NULL) {
 408                 err = -ENOMEM;
 409                 goto err_out;
 410         }
 411         spin_lock_bh(&queue->lock);
 412
 413         if (!queue->peer_pid) {
 414                 err = -EINVAL;
 415                 goto err_out_free_nskb;
 416         }
 417         if (queue->queue_total >= queue->queue_maxlen) {
 418                 queue->queue_dropped++;
 419                 if (net_ratelimit())
 420                           printk(KERN_WARNING "nf_queue: full at %d entries, "
 421                                  "dropping packets(s).\n",
 422                                  queue->queue_total);
 423                 goto err_out_free_nskb;
 424         }
 425         entry->id = ++queue->id_sequence;
 426         *packet_id_ptr = htonl(entry->id);
 427
 428         /* nfnetlink_unicast will either free the nskb or add it to a socket */
 429         err = nfnetlink_unicast(nskb, &init_net, queue->peer_pid, MSG_DONTWAIT);
 430         if (err < 0) {
 431                 queue->queue_user_dropped++;
 432                 goto err_out_unlock;
 433         }
 434
 435         __enqueue_entry(queue, entry);
 436
 437         spin_unlock_bh(&queue->lock);
 438         return 0;
 439
 440 err_out_free_nskb:
 441         kfree_skb(nskb);
 442 err_out_unlock:
 443         spin_unlock_bh(&queue->lock);
 444 err_out:
 445         return err;
 446 }
 447
 448 static int
 449 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e)
 450 {
 451         struct sk_buff *nskb;
 452         int diff;
 453
 454         diff = data_len - e->skb->len;
 455         if (diff < 0) {
 456                 if (pskb_trim(e->skb, data_len))
 457                         return -ENOMEM;
 458         } else if (diff > 0) {
 459                 if (data_len > 0xFFFF)
 460                         return -EINVAL;
 461                 if (diff > skb_tailroom(e->skb)) {
 462                         nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
 463                                                diff, GFP_ATOMIC);
 464                         if (!nskb) {
 465                                 printk(KERN_WARNING "nf_queue: OOM "
 466                                       "in mangle, dropping packet\n");
 467                                 return -ENOMEM;
 468                         }
 469                         kfree_skb(e->skb);
 470                         e->skb = nskb;
 471                 }
 472                 skb_put(e->skb, diff);
 473         }
 474         if (!skb_make_writable(e->skb, data_len))
 475                 return -ENOMEM;
 476         skb_copy_to_linear_data(e->skb, data, data_len);
 477         e->skb->ip_summed = CHECKSUM_NONE;
 478         return 0;
 479 }
 480
 481 static int
 482 nfqnl_set_mode(struct nfqnl_instance *queue,
 483                unsigned char mode, unsigned int range)
 484 {
 485         int status = 0;
 486
 487         spin_lock_bh(&queue->lock);
 488         switch (mode) {
 489         case NFQNL_COPY_NONE:
 490         case NFQNL_COPY_META:
 491                 queue->copy_mode = mode;
 492                 queue->copy_range = 0;
 493                 break;
 494
 495         case NFQNL_COPY_PACKET:
 496                 queue->copy_mode = mode;
 497                 /* we're using struct nlattr which has 16bit nla_len */
 498                 if (range > 0xffff)
 499                         queue->copy_range = 0xffff;
 500                 else
 501                         queue->copy_range = range;
 502                 break;
 503
 504         default:
 505                 status = -EINVAL;
 506
 507         }
 508         spin_unlock_bh(&queue->lock);
 509
 510         return status;
 511 }
 512
 513 static int
 514 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
 515 {
 516         if (entry->indev)
 517                 if (entry->indev->ifindex == ifindex)
 518                         return 1;
 519         if (entry->outdev)
 520                 if (entry->outdev->ifindex == ifindex)
 521                         return 1;
 522 #ifdef CONFIG_BRIDGE_NETFILTER
 523         if (entry->skb->nf_bridge) {
 524                 if (entry->skb->nf_bridge->physindev &&
 525                     entry->skb->nf_bridge->physindev->ifindex == ifindex)
 526                         return 1;
 527                 if (entry->skb->nf_bridge->physoutdev &&
 528                     entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
 529                         return 1;
 530         }
 531 #endif
 532         return 0;
 533 }
 534
 535 /* drop all packets with either indev or outdev == ifindex from all queue
 536  * instances */
 537 static void
 538 nfqnl_dev_drop(int ifindex)
 539 {
 540         int i;
 541
 542         rcu_read_lock();
 543
 544         for (i = 0; i < INSTANCE_BUCKETS; i++) {
 545                 struct hlist_node *tmp;
 546                 struct nfqnl_instance *inst;
 547                 struct hlist_head *head = &instance_table[i];
 548
 549                 hlist_for_each_entry_rcu(inst, tmp, head, hlist)
 550                         nfqnl_flush(inst, dev_cmp, ifindex);
 551         }
 552
 553         rcu_read_unlock();
 554 }
 555
 556 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
 557
 558 static int
 559 nfqnl_rcv_dev_event(struct notifier_block *this,
 560                     unsigned long event, void *ptr)
 561 {
 562         struct net_device *dev = ptr;
 563
 564         if (!net_eq(dev_net(dev), &init_net))
 565                 return NOTIFY_DONE;
 566
 567         /* Drop any packets associated with the downed device */
 568         if (event == NETDEV_DOWN)
 569                 nfqnl_dev_drop(dev->ifindex);
 570         return NOTIFY_DONE;
 571 }
 572
 573 static struct notifier_block nfqnl_dev_notifier = {
 574         .notifier_call  = nfqnl_rcv_dev_event,
 575 };
 576
 577 static int
 578 nfqnl_rcv_nl_event(struct notifier_block *this,
 579                    unsigned long event, void *ptr)
 580 {
 581         struct netlink_notify *n = ptr;
 582
 583         if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
 584                 int i;
 585
 586                 /* destroy all instances for this pid */
 587                 spin_lock(&instances_lock);
 588                 for (i = 0; i < INSTANCE_BUCKETS; i++) {
 589                         struct hlist_node *tmp, *t2;
 590                         struct nfqnl_instance *inst;
 591                         struct hlist_head *head = &instance_table[i];
 592
 593                         hlist_for_each_entry_safe(inst, tmp, t2, head, hlist) {
 594                                 if ((n->net == &init_net) &&
 595                                     (n->pid == inst->peer_pid))
 596                                         __instance_destroy(inst);
 597                         }
 598                 }
 599                 spin_unlock(&instances_lock);
 600         }
 601         return NOTIFY_DONE;
 602 }
 603
 604 static struct notifier_block nfqnl_rtnl_notifier = {
 605         .notifier_call  = nfqnl_rcv_nl_event,
 606 };
 607
 608 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
 609         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 610         [NFQA_MARK]             = { .type = NLA_U32 },
 611         [NFQA_PAYLOAD]          = { .type = NLA_UNSPEC },
 612 };
 613
 614 static int
 615 nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
 616                    const struct nlmsghdr *nlh,
 617                    const struct nlattr * const nfqa[])
 618 {
 619         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 620         u_int16_t queue_num = ntohs(nfmsg->res_id);
 621
 622         struct nfqnl_msg_verdict_hdr *vhdr;
 623         struct nfqnl_instance *queue;
 624         unsigned int verdict;
 625         struct nf_queue_entry *entry;
 626
 627         queue = instance_lookup(queue_num);
 628         if (!queue)
 629                 return -ENODEV;
 630
 631         if (queue->peer_pid != NETLINK_CB(skb).pid)
 632                 return -EPERM;
 633
 634         if (!nfqa[NFQA_VERDICT_HDR])
 635                 return -EINVAL;
 636
 637         vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
 638         verdict = ntohl(vhdr->verdict);
 639
 640         if ((verdict & NF_VERDICT_MASK) > NF_MAX_VERDICT)
 641                 return -EINVAL;
 642
 643         entry = find_dequeue_entry(queue, ntohl(vhdr->id));
 644         if (entry == NULL)
 645                 return -ENOENT;
 646
 647         if (nfqa[NFQA_PAYLOAD]) {
 648                 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
 649                                  nla_len(nfqa[NFQA_PAYLOAD]), entry) < 0)
 650                         verdict = NF_DROP;
 651         }
 652
 653         if (nfqa[NFQA_MARK])
 654                 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
 655
 656         nf_reinject(entry, verdict);
 657         return 0;
 658 }
 659
 660 static int
 661 nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
 662                   const struct nlmsghdr *nlh,
 663                   const struct nlattr * const nfqa[])
 664 {
 665         return -ENOTSUPP;
 666 }
 667
 668 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
 669         [NFQA_CFG_CMD]          = { .len = sizeof(struct nfqnl_msg_config_cmd) },
 670         [NFQA_CFG_PARAMS]       = { .len = sizeof(struct nfqnl_msg_config_params) },
 671 };
 672
 673 static const struct nf_queue_handler nfqh = {
 674         .name   = "nf_queue",
 675         .outfn  = &nfqnl_enqueue_packet,
 676 };
 677
 678 static int
 679 nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
 680                   const struct nlmsghdr *nlh,
 681                   const struct nlattr * const nfqa[])
 682 {
 683         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 684         u_int16_t queue_num = ntohs(nfmsg->res_id);
 685         struct nfqnl_instance *queue;
 686         struct nfqnl_msg_config_cmd *cmd = NULL;
 687         int ret = 0;
 688
 689         if (nfqa[NFQA_CFG_CMD]) {
 690                 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
 691
 692                 /* Commands without queue context - might sleep */
 693                 switch (cmd->command) {
 694                 case NFQNL_CFG_CMD_PF_BIND:
 695                         return nf_register_queue_handler(ntohs(cmd->pf),
 696                                                          &nfqh);
 697                 case NFQNL_CFG_CMD_PF_UNBIND:
 698                         return nf_unregister_queue_handler(ntohs(cmd->pf),
 699                                                            &nfqh);
 700                 }
 701         }
 702
 703         rcu_read_lock();
 704         queue = instance_lookup(queue_num);
 705         if (queue && queue->peer_pid != NETLINK_CB(skb).pid) {
 706                 ret = -EPERM;
 707                 goto err_out_unlock;
 708         }
 709
 710         if (cmd != NULL) {
 711                 switch (cmd->command) {
 712                 case NFQNL_CFG_CMD_BIND:
 713                         if (queue) {
 714                                 ret = -EBUSY;
 715                                 goto err_out_unlock;
 716                         }
 717                         queue = instance_create(queue_num, NETLINK_CB(skb).pid);
 718                         if (IS_ERR(queue)) {
 719                                 ret = PTR_ERR(queue);
 720                                 goto err_out_unlock;
 721                         }
 722                         break;
 723                 case NFQNL_CFG_CMD_UNBIND:
 724                         if (!queue) {
 725                                 ret = -ENODEV;
 726                                 goto err_out_unlock;
 727                         }
 728                         instance_destroy(queue);
 729                         break;
 730                 case NFQNL_CFG_CMD_PF_BIND:
 731                 case NFQNL_CFG_CMD_PF_UNBIND:
 732                         break;
 733                 default:
 734                         ret = -ENOTSUPP;
 735                         break;
 736                 }
 737         }
 738
 739         if (nfqa[NFQA_CFG_PARAMS]) {
 740                 struct nfqnl_msg_config_params *params;
 741
 742                 if (!queue) {
 743                         ret = -ENODEV;
 744                         goto err_out_unlock;
 745                 }
 746                 params = nla_data(nfqa[NFQA_CFG_PARAMS]);
 747                 nfqnl_set_mode(queue, params->copy_mode,
 748                                 ntohl(params->copy_range));
 749         }
 750
 751         if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
 752                 __be32 *queue_maxlen;
 753
 754                 if (!queue) {
 755                         ret = -ENODEV;
 756                         goto err_out_unlock;
 757                 }
 758                 queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
 759                 spin_lock_bh(&queue->lock);
 760                 queue->queue_maxlen = ntohl(*queue_maxlen);
 761                 spin_unlock_bh(&queue->lock);
 762         }
 763
 764 err_out_unlock:
 765         rcu_read_unlock();
 766         return ret;
 767 }
 768
 769 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
 770         [NFQNL_MSG_PACKET]      = { .call_rcu = nfqnl_recv_unsupp,
 771                                     .attr_count = NFQA_MAX, },
 772         [NFQNL_MSG_VERDICT]     = { .call_rcu = nfqnl_recv_verdict,
 773                                     .attr_count = NFQA_MAX,
 774                                     .policy = nfqa_verdict_policy },
 775         [NFQNL_MSG_CONFIG]      = { .call = nfqnl_recv_config,
 776                                     .attr_count = NFQA_CFG_MAX,
 777                                     .policy = nfqa_cfg_policy },
 778 };
 779
 780 static const struct nfnetlink_subsystem nfqnl_subsys = {
 781         .name           = "nf_queue",
 782         .subsys_id      = NFNL_SUBSYS_QUEUE,
 783         .cb_count       = NFQNL_MSG_MAX,
 784         .cb             = nfqnl_cb,
 785 };
 786
 787 #ifdef CONFIG_PROC_FS
 788 struct iter_state {
 789         unsigned int bucket;
 790 };
 791
 792 static struct hlist_node *get_first(struct seq_file *seq)
 793 {
 794         struct iter_state *st = seq->private;
 795
 796         if (!st)
 797                 return NULL;
 798
 799         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
 800                 if (!hlist_empty(&instance_table[st->bucket]))
 801                         return instance_table[st->bucket].first;
 802         }
 803         return NULL;
 804 }
 805
 806 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
 807 {
 808         struct iter_state *st = seq->private;
 809
 810         h = h->next;
 811         while (!h) {
 812                 if (++st->bucket >= INSTANCE_BUCKETS)
 813                         return NULL;
 814
 815                 h = instance_table[st->bucket].first;
 816         }
 817         return h;
 818 }
 819
 820 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
 821 {
 822         struct hlist_node *head;
 823         head = get_first(seq);
 824
 825         if (head)
 826                 while (pos && (head = get_next(seq, head)))
 827                         pos--;
 828         return pos ? NULL : head;
 829 }
 830
 831 static void *seq_start(struct seq_file *seq, loff_t *pos)
 832         __acquires(instances_lock)
 833 {
 834         spin_lock(&instances_lock);
 835         return get_idx(seq, *pos);
 836 }
 837
 838 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
 839 {
 840         (*pos)++;
 841         return get_next(s, v);
 842 }
 843
 844 static void seq_stop(struct seq_file *s, void *v)
 845         __releases(instances_lock)
 846 {
 847         spin_unlock(&instances_lock);
 848 }
 849
 850 static int seq_show(struct seq_file *s, void *v)
 851 {
 852         const struct nfqnl_instance *inst = v;
 853
 854         return seq_printf(s, "%5d %6d %5d %1d %5d %5d %5d %8d %2d\n",
 855                           inst->queue_num,
 856                           inst->peer_pid, inst->queue_total,
 857                           inst->copy_mode, inst->copy_range,
 858                           inst->queue_dropped, inst->queue_user_dropped,
 859                           inst->id_sequence, 1);
 860 }
 861
 862 static const struct seq_operations nfqnl_seq_ops = {
 863         .start  = seq_start,
 864         .next   = seq_next,
 865         .stop   = seq_stop,
 866         .show   = seq_show,
 867 };
 868
 869 static int nfqnl_open(struct inode *inode, struct file *file)
 870 {
 871         return seq_open_private(file, &nfqnl_seq_ops,
 872                         sizeof(struct iter_state));
 873 }
 874
 875 static const struct file_operations nfqnl_file_ops = {
 876         .owner   = THIS_MODULE,
 877         .open    = nfqnl_open,
 878         .read    = seq_read,
 879         .llseek  = seq_lseek,
 880         .release = seq_release_private,
 881 };
 882
 883 #endif /* PROC_FS */
 884
 885 static int __init nfnetlink_queue_init(void)
 886 {
 887         int i, status = -ENOMEM;
 888
 889         for (i = 0; i < INSTANCE_BUCKETS; i++)
 890                 INIT_HLIST_HEAD(&instance_table[i]);
 891
 892         netlink_register_notifier(&nfqnl_rtnl_notifier);
 893         status = nfnetlink_subsys_register(&nfqnl_subsys);
 894         if (status < 0) {
 895                 printk(KERN_ERR "nf_queue: failed to create netlink socket\n");
 896                 goto cleanup_netlink_notifier;
 897         }
 898
 899 #ifdef CONFIG_PROC_FS
 900         if (!proc_create("nfnetlink_queue", 0440,
 901                          proc_net_netfilter, &nfqnl_file_ops))
 902                 goto cleanup_subsys;
 903 #endif
 904
 905         register_netdevice_notifier(&nfqnl_dev_notifier);
 906         return status;
 907
 908 #ifdef CONFIG_PROC_FS
 909 cleanup_subsys:
 910         nfnetlink_subsys_unregister(&nfqnl_subsys);
 911 #endif
 912 cleanup_netlink_notifier:
 913         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
 914         return status;
 915 }
 916
 917 static void __exit nfnetlink_queue_fini(void)
 918 {
 919         nf_unregister_queue_handlers(&nfqh);
 920         unregister_netdevice_notifier(&nfqnl_dev_notifier);
 921 #ifdef CONFIG_PROC_FS
 922         remove_proc_entry("nfnetlink_queue", proc_net_netfilter);
 923 #endif
 924         nfnetlink_subsys_unregister(&nfqnl_subsys);
 925         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
 926
 927         rcu_barrier(); /* Wait for completion of call_rcu()'s */
 928 }
 929
 930 MODULE_DESCRIPTION("netfilter packet queue handler");
 931 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
 932 MODULE_LICENSE("GPL");
 933 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
 934
 935 module_init(nfnetlink_queue_init);
 936 module_exit(nfnetlink_queue_fini);