git.oblomov.eu Git - linux-2.6/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  * (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_ALIGN(sizeof(struct nfgenmsg))
 228                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
 229                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 230                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 231 #ifdef CONFIG_BRIDGE_NETFILTER
 232                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 233                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 234 #endif
 235                 + nla_total_size(sizeof(u_int32_t))     /* mark */
 236                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
 237                 + nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
 238
 239         outdev = entry->outdev;
 240
 241         spin_lock_bh(&queue->lock);
 242
 243         switch ((enum nfqnl_config_mode)queue->copy_mode) {
 244         case NFQNL_COPY_META:
 245         case NFQNL_COPY_NONE:
 246                 data_len = 0;
 247                 break;
 248
 249         case NFQNL_COPY_PACKET:
 250                 if ((entskb->ip_summed == CHECKSUM_PARTIAL ||
 251                      entskb->ip_summed == CHECKSUM_COMPLETE) &&
 252                     skb_checksum_help(entskb)) {
 253                         spin_unlock_bh(&queue->lock);
 254                         return NULL;
 255                 }
 256                 if (queue->copy_range == 0
 257                     || queue->copy_range > entskb->len)
 258                         data_len = entskb->len;
 259                 else
 260                         data_len = queue->copy_range;
 261
 262                 size += nla_total_size(data_len);
 263                 break;
 264         }
 265
 266         entry->id = queue->id_sequence++;
 267
 268         spin_unlock_bh(&queue->lock);
 269
 270         skb = alloc_skb(size, GFP_ATOMIC);
 271         if (!skb)
 272                 goto nlmsg_failure;
 273
 274         old_tail = skb->tail;
 275         nlh = NLMSG_PUT(skb, 0, 0,
 276                         NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
 277                         sizeof(struct nfgenmsg));
 278         nfmsg = NLMSG_DATA(nlh);
 279         nfmsg->nfgen_family = entry->pf;
 280         nfmsg->version = NFNETLINK_V0;
 281         nfmsg->res_id = htons(queue->queue_num);
 282
 283         pmsg.packet_id          = htonl(entry->id);
 284         pmsg.hw_protocol        = entskb->protocol;
 285         pmsg.hook               = entry->hook;
 286
 287         NLA_PUT(skb, NFQA_PACKET_HDR, sizeof(pmsg), &pmsg);
 288
 289         indev = entry->indev;
 290         if (indev) {
 291 #ifndef CONFIG_BRIDGE_NETFILTER
 292                 NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex));
 293 #else
 294                 if (entry->pf == PF_BRIDGE) {
 295                         /* Case 1: indev is physical input device, we need to
 296                          * look for bridge group (when called from
 297                          * netfilter_bridge) */
 298                         NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSINDEV,
 299                                      htonl(indev->ifindex));
 300                         /* this is the bridge group "brX" */
 301                         NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV,
 302                                      htonl(indev->br_port->br->dev->ifindex));
 303                 } else {
 304                         /* Case 2: indev is bridge group, we need to look for
 305                          * physical device (when called from ipv4) */
 306                         NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV,
 307                                      htonl(indev->ifindex));
 308                         if (entskb->nf_bridge && entskb->nf_bridge->physindev)
 309                                 NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSINDEV,
 310                                              htonl(entskb->nf_bridge->physindev->ifindex));
 311                 }
 312 #endif
 313         }
 314
 315         if (outdev) {
 316 #ifndef CONFIG_BRIDGE_NETFILTER
 317                 NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex));
 318 #else
 319                 if (entry->pf == PF_BRIDGE) {
 320                         /* Case 1: outdev is physical output device, we need to
 321                          * look for bridge group (when called from
 322                          * netfilter_bridge) */
 323                         NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 324                                      htonl(outdev->ifindex));
 325                         /* this is the bridge group "brX" */
 326                         NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV,
 327                                      htonl(outdev->br_port->br->dev->ifindex));
 328                 } else {
 329                         /* Case 2: outdev is bridge group, we need to look for
 330                          * physical output device (when called from ipv4) */
 331                         NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV,
 332                                      htonl(outdev->ifindex));
 333                         if (entskb->nf_bridge && entskb->nf_bridge->physoutdev)
 334                                 NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 335                                              htonl(entskb->nf_bridge->physoutdev->ifindex));
 336                 }
 337 #endif
 338         }
 339
 340         if (entskb->mark)
 341                 NLA_PUT_BE32(skb, NFQA_MARK, htonl(entskb->mark));
 342
 343         if (indev && entskb->dev) {
 344                 struct nfqnl_msg_packet_hw phw;
 345                 int len = dev_parse_header(entskb, phw.hw_addr);
 346                 if (len) {
 347                         phw.hw_addrlen = htons(len);
 348                         NLA_PUT(skb, NFQA_HWADDR, sizeof(phw), &phw);
 349                 }
 350         }
 351
 352         if (entskb->tstamp.tv64) {
 353                 struct nfqnl_msg_packet_timestamp ts;
 354                 struct timeval tv = ktime_to_timeval(entskb->tstamp);
 355                 ts.sec = cpu_to_be64(tv.tv_sec);
 356                 ts.usec = cpu_to_be64(tv.tv_usec);
 357
 358                 NLA_PUT(skb, NFQA_TIMESTAMP, sizeof(ts), &ts);
 359         }
 360
 361         if (data_len) {
 362                 struct nlattr *nla;
 363                 int size = nla_attr_size(data_len);
 364
 365                 if (skb_tailroom(skb) < nla_total_size(data_len)) {
 366                         printk(KERN_WARNING "nf_queue: no tailroom!\n");
 367                         goto nlmsg_failure;
 368                 }
 369
 370                 nla = (struct nlattr *)skb_put(skb, nla_total_size(data_len));
 371                 nla->nla_type = NFQA_PAYLOAD;
 372                 nla->nla_len = size;
 373
 374                 if (skb_copy_bits(entskb, 0, nla_data(nla), data_len))
 375                         BUG();
 376         }
 377
 378         nlh->nlmsg_len = skb->tail - old_tail;
 379         return skb;
 380
 381 nlmsg_failure:
 382 nla_put_failure:
 383         if (skb)
 384                 kfree_skb(skb);
 385         if (net_ratelimit())
 386                 printk(KERN_ERR "nf_queue: error creating packet message\n");
 387         return NULL;
 388 }
 389
 390 static int
 391 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 392 {
 393         struct sk_buff *nskb;
 394         struct nfqnl_instance *queue;
 395         int err;
 396
 397         /* rcu_read_lock()ed by nf_hook_slow() */
 398         queue = instance_lookup(queuenum);
 399         if (!queue)
 400                 goto err_out;
 401
 402         if (queue->copy_mode == NFQNL_COPY_NONE)
 403                 goto err_out;
 404
 405         nskb = nfqnl_build_packet_message(queue, entry);
 406         if (nskb == NULL)
 407                 goto err_out;
 408
 409         spin_lock_bh(&queue->lock);
 410
 411         if (!queue->peer_pid)
 412                 goto err_out_free_nskb;
 413
 414         if (queue->queue_total >= queue->queue_maxlen) {
 415                 queue->queue_dropped++;
 416                 if (net_ratelimit())
 417                           printk(KERN_WARNING "nf_queue: full at %d entries, "
 418                                  "dropping packets(s). Dropped: %d\n",
 419                                  queue->queue_total, queue->queue_dropped);
 420                 goto err_out_free_nskb;
 421         }
 422
 423         /* nfnetlink_unicast will either free the nskb or add it to a socket */
 424         err = nfnetlink_unicast(nskb, queue->peer_pid, MSG_DONTWAIT);
 425         if (err < 0) {
 426                 queue->queue_user_dropped++;
 427                 goto err_out_unlock;
 428         }
 429
 430         __enqueue_entry(queue, entry);
 431
 432         spin_unlock_bh(&queue->lock);
 433         return 0;
 434
 435 err_out_free_nskb:
 436         kfree_skb(nskb);
 437 err_out_unlock:
 438         spin_unlock_bh(&queue->lock);
 439 err_out:
 440         return -1;
 441 }
 442
 443 static int
 444 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e)
 445 {
 446         int diff;
 447         int err;
 448
 449         diff = data_len - e->skb->len;
 450         if (diff < 0) {
 451                 if (pskb_trim(e->skb, data_len))
 452                         return -ENOMEM;
 453         } else if (diff > 0) {
 454                 if (data_len > 0xFFFF)
 455                         return -EINVAL;
 456                 if (diff > skb_tailroom(e->skb)) {
 457                         err = pskb_expand_head(e->skb, 0,
 458                                                diff - skb_tailroom(e->skb),
 459                                                GFP_ATOMIC);
 460                         if (err) {
 461                                 printk(KERN_WARNING "nf_queue: OOM "
 462                                       "in mangle, dropping packet\n");
 463                                 return err;
 464                         }
 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 (dev->nd_net != &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                         ret = nf_register_queue_handler(ntohs(cmd->pf),
 705                                                         &nfqh);
 706                         break;
 707                 case NFQNL_CFG_CMD_PF_UNBIND:
 708                         ret = nf_unregister_queue_handler(ntohs(cmd->pf),
 709                                                           &nfqh);
 710                         break;
 711                 default:
 712                         break;
 713                 }
 714
 715                 if (ret < 0)
 716                         return ret;
 717         }
 718
 719         rcu_read_lock();
 720         queue = instance_lookup(queue_num);
 721         if (queue && queue->peer_pid != NETLINK_CB(skb).pid) {
 722                 ret = -EPERM;
 723                 goto err_out_unlock;
 724         }
 725
 726         if (cmd != NULL) {
 727                 switch (cmd->command) {
 728                 case NFQNL_CFG_CMD_BIND:
 729                         if (queue) {
 730                                 ret = -EBUSY;
 731                                 goto err_out_unlock;
 732                         }
 733                         queue = instance_create(queue_num, NETLINK_CB(skb).pid);
 734                         if (IS_ERR(queue)) {
 735                                 ret = PTR_ERR(queue);
 736                                 goto err_out_unlock;
 737                         }
 738                         break;
 739                 case NFQNL_CFG_CMD_UNBIND:
 740                         if (!queue) {
 741                                 ret = -ENODEV;
 742                                 goto err_out_unlock;
 743                         }
 744                         instance_destroy(queue);
 745                         break;
 746                 case NFQNL_CFG_CMD_PF_BIND:
 747                 case NFQNL_CFG_CMD_PF_UNBIND:
 748                         break;
 749                 default:
 750                         ret = -ENOTSUPP;
 751                         break;
 752                 }
 753         }
 754
 755         if (nfqa[NFQA_CFG_PARAMS]) {
 756                 struct nfqnl_msg_config_params *params;
 757
 758                 if (!queue) {
 759                         ret = -ENODEV;
 760                         goto err_out_unlock;
 761                 }
 762                 params = nla_data(nfqa[NFQA_CFG_PARAMS]);
 763                 nfqnl_set_mode(queue, params->copy_mode,
 764                                 ntohl(params->copy_range));
 765         }
 766
 767         if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
 768                 __be32 *queue_maxlen;
 769
 770                 if (!queue) {
 771                         ret = -ENODEV;
 772                         goto err_out_unlock;
 773                 }
 774                 queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
 775                 spin_lock_bh(&queue->lock);
 776                 queue->queue_maxlen = ntohl(*queue_maxlen);
 777                 spin_unlock_bh(&queue->lock);
 778         }
 779
 780 err_out_unlock:
 781         rcu_read_unlock();
 782         return ret;
 783 }
 784
 785 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
 786         [NFQNL_MSG_PACKET]      = { .call = nfqnl_recv_unsupp,
 787                                     .attr_count = NFQA_MAX, },
 788         [NFQNL_MSG_VERDICT]     = { .call = nfqnl_recv_verdict,
 789                                     .attr_count = NFQA_MAX,
 790                                     .policy = nfqa_verdict_policy },
 791         [NFQNL_MSG_CONFIG]      = { .call = nfqnl_recv_config,
 792                                     .attr_count = NFQA_CFG_MAX,
 793                                     .policy = nfqa_cfg_policy },
 794 };
 795
 796 static const struct nfnetlink_subsystem nfqnl_subsys = {
 797         .name           = "nf_queue",
 798         .subsys_id      = NFNL_SUBSYS_QUEUE,
 799         .cb_count       = NFQNL_MSG_MAX,
 800         .cb             = nfqnl_cb,
 801 };
 802
 803 #ifdef CONFIG_PROC_FS
 804 struct iter_state {
 805         unsigned int bucket;
 806 };
 807
 808 static struct hlist_node *get_first(struct seq_file *seq)
 809 {
 810         struct iter_state *st = seq->private;
 811
 812         if (!st)
 813                 return NULL;
 814
 815         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
 816                 if (!hlist_empty(&instance_table[st->bucket]))
 817                         return instance_table[st->bucket].first;
 818         }
 819         return NULL;
 820 }
 821
 822 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
 823 {
 824         struct iter_state *st = seq->private;
 825
 826         h = h->next;
 827         while (!h) {
 828                 if (++st->bucket >= INSTANCE_BUCKETS)
 829                         return NULL;
 830
 831                 h = instance_table[st->bucket].first;
 832         }
 833         return h;
 834 }
 835
 836 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
 837 {
 838         struct hlist_node *head;
 839         head = get_first(seq);
 840
 841         if (head)
 842                 while (pos && (head = get_next(seq, head)))
 843                         pos--;
 844         return pos ? NULL : head;
 845 }
 846
 847 static void *seq_start(struct seq_file *seq, loff_t *pos)
 848 {
 849         spin_lock(&instances_lock);
 850         return get_idx(seq, *pos);
 851 }
 852
 853 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
 854 {
 855         (*pos)++;
 856         return get_next(s, v);
 857 }
 858
 859 static void seq_stop(struct seq_file *s, void *v)
 860 {
 861         spin_unlock(&instances_lock);
 862 }
 863
 864 static int seq_show(struct seq_file *s, void *v)
 865 {
 866         const struct nfqnl_instance *inst = v;
 867
 868         return seq_printf(s, "%5d %6d %5d %1d %5d %5d %5d %8d %2d\n",
 869                           inst->queue_num,
 870                           inst->peer_pid, inst->queue_total,
 871                           inst->copy_mode, inst->copy_range,
 872                           inst->queue_dropped, inst->queue_user_dropped,
 873                           inst->id_sequence, 1);
 874 }
 875
 876 static const struct seq_operations nfqnl_seq_ops = {
 877         .start  = seq_start,
 878         .next   = seq_next,
 879         .stop   = seq_stop,
 880         .show   = seq_show,
 881 };
 882
 883 static int nfqnl_open(struct inode *inode, struct file *file)
 884 {
 885         return seq_open_private(file, &nfqnl_seq_ops,
 886                         sizeof(struct iter_state));
 887 }
 888
 889 static const struct file_operations nfqnl_file_ops = {
 890         .owner   = THIS_MODULE,
 891         .open    = nfqnl_open,
 892         .read    = seq_read,
 893         .llseek  = seq_lseek,
 894         .release = seq_release_private,
 895 };
 896
 897 #endif /* PROC_FS */
 898
 899 static int __init nfnetlink_queue_init(void)
 900 {
 901         int i, status = -ENOMEM;
 902 #ifdef CONFIG_PROC_FS
 903         struct proc_dir_entry *proc_nfqueue;
 904 #endif
 905
 906         for (i = 0; i < INSTANCE_BUCKETS; i++)
 907                 INIT_HLIST_HEAD(&instance_table[i]);
 908
 909         netlink_register_notifier(&nfqnl_rtnl_notifier);
 910         status = nfnetlink_subsys_register(&nfqnl_subsys);
 911         if (status < 0) {
 912                 printk(KERN_ERR "nf_queue: failed to create netlink socket\n");
 913                 goto cleanup_netlink_notifier;
 914         }
 915
 916 #ifdef CONFIG_PROC_FS
 917         proc_nfqueue = create_proc_entry("nfnetlink_queue", 0440,
 918                                          proc_net_netfilter);
 919         if (!proc_nfqueue)
 920                 goto cleanup_subsys;
 921         proc_nfqueue->proc_fops = &nfqnl_file_ops;
 922 #endif
 923
 924         register_netdevice_notifier(&nfqnl_dev_notifier);
 925         return status;
 926
 927 #ifdef CONFIG_PROC_FS
 928 cleanup_subsys:
 929         nfnetlink_subsys_unregister(&nfqnl_subsys);
 930 #endif
 931 cleanup_netlink_notifier:
 932         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
 933         return status;
 934 }
 935
 936 static void __exit nfnetlink_queue_fini(void)
 937 {
 938         nf_unregister_queue_handlers(&nfqh);
 939         unregister_netdevice_notifier(&nfqnl_dev_notifier);
 940 #ifdef CONFIG_PROC_FS
 941         remove_proc_entry("nfnetlink_queue", proc_net_netfilter);
 942 #endif
 943         nfnetlink_subsys_unregister(&nfqnl_subsys);
 944         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
 945 }
 946
 947 MODULE_DESCRIPTION("netfilter packet queue handler");
 948 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
 949 MODULE_LICENSE("GPL");
 950 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
 951
 952 module_init(nfnetlink_queue_init);
 953 module_exit(nfnetlink_queue_fini);