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
 219         size =    NLMSG_ALIGN(sizeof(struct nfgenmsg))
 220                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
 221                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 222                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 223 #ifdef CONFIG_BRIDGE_NETFILTER
 224                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 225                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 226 #endif
 227                 + nla_total_size(sizeof(u_int32_t))     /* mark */
 228                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
 229                 + nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
 230
 231         outdev = entry->outdev;
 232
 233         spin_lock_bh(&queue->lock);
 234
 235         switch (queue->copy_mode) {
 236         case NFQNL_COPY_META:
 237         case NFQNL_COPY_NONE:
 238                 data_len = 0;
 239                 break;
 240
 241         case NFQNL_COPY_PACKET:
 242                 if ((entskb->ip_summed == CHECKSUM_PARTIAL ||
 243                      entskb->ip_summed == CHECKSUM_COMPLETE) &&
 244                     (*errp = skb_checksum_help(entskb))) {
 245                         spin_unlock_bh(&queue->lock);
 246                         return NULL;
 247                 }
 248                 if (queue->copy_range == 0
 249                     || queue->copy_range > entskb->len)
 250                         data_len = entskb->len;
 251                 else
 252                         data_len = queue->copy_range;
 253
 254                 size += nla_total_size(data_len);
 255                 break;
 256
 257         default:
 258                 *errp = -EINVAL;
 259                 spin_unlock_bh(&queue->lock);
 260                 return NULL;
 261         }
 262
 263         entry->id = queue->id_sequence++;
 264
 265         spin_unlock_bh(&queue->lock);
 266
 267         skb = alloc_skb(size, GFP_ATOMIC);
 268         if (!skb)
 269                 goto nlmsg_failure;
 270
 271         old_tail = skb->tail;
 272         nlh = NLMSG_PUT(skb, 0, 0,
 273                         NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
 274                         sizeof(struct nfgenmsg));
 275         nfmsg = NLMSG_DATA(nlh);
 276         nfmsg->nfgen_family = entry->pf;
 277         nfmsg->version = NFNETLINK_V0;
 278         nfmsg->res_id = htons(queue->queue_num);
 279
 280         pmsg.packet_id          = htonl(entry->id);
 281         pmsg.hw_protocol        = entskb->protocol;
 282         pmsg.hook               = entry->hook;
 283
 284         NLA_PUT(skb, NFQA_PACKET_HDR, sizeof(pmsg), &pmsg);
 285
 286         indev = entry->indev;
 287         if (indev) {
 288 #ifndef CONFIG_BRIDGE_NETFILTER
 289                 NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex));
 290 #else
 291                 if (entry->pf == PF_BRIDGE) {
 292                         /* Case 1: indev is physical input device, we need to
 293                          * look for bridge group (when called from
 294                          * netfilter_bridge) */
 295                         NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSINDEV,
 296                                      htonl(indev->ifindex));
 297                         /* this is the bridge group "brX" */
 298                         NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV,
 299                                      htonl(indev->br_port->br->dev->ifindex));
 300                 } else {
 301                         /* Case 2: indev is bridge group, we need to look for
 302                          * physical device (when called from ipv4) */
 303                         NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV,
 304                                      htonl(indev->ifindex));
 305                         if (entskb->nf_bridge && entskb->nf_bridge->physindev)
 306                                 NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSINDEV,
 307                                              htonl(entskb->nf_bridge->physindev->ifindex));
 308                 }
 309 #endif
 310         }
 311
 312         if (outdev) {
 313 #ifndef CONFIG_BRIDGE_NETFILTER
 314                 NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex));
 315 #else
 316                 if (entry->pf == PF_BRIDGE) {
 317                         /* Case 1: outdev is physical output device, we need to
 318                          * look for bridge group (when called from
 319                          * netfilter_bridge) */
 320                         NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 321                                      htonl(outdev->ifindex));
 322                         /* this is the bridge group "brX" */
 323                         NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV,
 324                                      htonl(outdev->br_port->br->dev->ifindex));
 325                 } else {
 326                         /* Case 2: outdev is bridge group, we need to look for
 327                          * physical output device (when called from ipv4) */
 328                         NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV,
 329                                      htonl(outdev->ifindex));
 330                         if (entskb->nf_bridge && entskb->nf_bridge->physoutdev)
 331                                 NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 332                                              htonl(entskb->nf_bridge->physoutdev->ifindex));
 333                 }
 334 #endif
 335         }
 336
 337         if (entskb->mark)
 338                 NLA_PUT_BE32(skb, NFQA_MARK, htonl(entskb->mark));
 339
 340         if (indev && entskb->dev) {
 341                 struct nfqnl_msg_packet_hw phw;
 342                 int len = dev_parse_header(entskb, phw.hw_addr);
 343                 if (len) {
 344                         phw.hw_addrlen = htons(len);
 345                         NLA_PUT(skb, NFQA_HWADDR, sizeof(phw), &phw);
 346                 }
 347         }
 348
 349         if (entskb->tstamp.tv64) {
 350                 struct nfqnl_msg_packet_timestamp ts;
 351                 struct timeval tv = ktime_to_timeval(entskb->tstamp);
 352                 ts.sec = cpu_to_be64(tv.tv_sec);
 353                 ts.usec = cpu_to_be64(tv.tv_usec);
 354
 355                 NLA_PUT(skb, NFQA_TIMESTAMP, sizeof(ts), &ts);
 356         }
 357
 358         if (data_len) {
 359                 struct nlattr *nla;
 360                 int size = nla_attr_size(data_len);
 361
 362                 if (skb_tailroom(skb) < nla_total_size(data_len)) {
 363                         printk(KERN_WARNING "nf_queue: no tailroom!\n");
 364                         goto nlmsg_failure;
 365                 }
 366
 367                 nla = (struct nlattr *)skb_put(skb, nla_total_size(data_len));
 368                 nla->nla_type = NFQA_PAYLOAD;
 369                 nla->nla_len = size;
 370
 371                 if (skb_copy_bits(entskb, 0, nla_data(nla), data_len))
 372                         BUG();
 373         }
 374
 375         nlh->nlmsg_len = skb->tail - old_tail;
 376         return skb;
 377
 378 nlmsg_failure:
 379 nla_put_failure:
 380         if (skb)
 381                 kfree_skb(skb);
 382         *errp = -EINVAL;
 383         if (net_ratelimit())
 384                 printk(KERN_ERR "nf_queue: error creating packet message\n");
 385         return NULL;
 386 }
 387
 388 static int
 389 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 390 {
 391         int status = -EINVAL;
 392         struct sk_buff *nskb;
 393         struct nfqnl_instance *queue;
 394
 395         /* rcu_read_lock()ed by nf_hook_slow() */
 396         queue = instance_lookup(queuenum);
 397         if (!queue)
 398                 return -EINVAL;
 399
 400         if (queue->copy_mode == NFQNL_COPY_NONE)
 401                 return -EAGAIN;
 402
 403         nskb = nfqnl_build_packet_message(queue, entry, &status);
 404         if (nskb == NULL)
 405                 return status;
 406
 407         spin_lock_bh(&queue->lock);
 408
 409         if (!queue->peer_pid)
 410                 goto err_out_free_nskb;
 411
 412         if (queue->queue_total >= queue->queue_maxlen) {
 413                 queue->queue_dropped++;
 414                 status = -ENOSPC;
 415                 if (net_ratelimit())
 416                           printk(KERN_WARNING "nf_queue: full at %d entries, "
 417                                  "dropping packets(s). Dropped: %d\n",
 418                                  queue->queue_total, queue->queue_dropped);
 419                 goto err_out_free_nskb;
 420         }
 421
 422         /* nfnetlink_unicast will either free the nskb or add it to a socket */
 423         status = nfnetlink_unicast(nskb, queue->peer_pid, MSG_DONTWAIT);
 424         if (status < 0) {
 425                 queue->queue_user_dropped++;
 426                 goto err_out_unlock;
 427         }
 428
 429         __enqueue_entry(queue, entry);
 430
 431         spin_unlock_bh(&queue->lock);
 432         return status;
 433
 434 err_out_free_nskb:
 435         kfree_skb(nskb);
 436
 437 err_out_unlock:
 438         spin_unlock_bh(&queue->lock);
 439         return status;
 440 }
 441
 442 static int
 443 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e)
 444 {
 445         int diff;
 446         int err;
 447
 448         diff = data_len - e->skb->len;
 449         if (diff < 0) {
 450                 if (pskb_trim(e->skb, data_len))
 451                         return -ENOMEM;
 452         } else if (diff > 0) {
 453                 if (data_len > 0xFFFF)
 454                         return -EINVAL;
 455                 if (diff > skb_tailroom(e->skb)) {
 456                         err = pskb_expand_head(e->skb, 0,
 457                                                diff - skb_tailroom(e->skb),
 458                                                GFP_ATOMIC);
 459                         if (err) {
 460                                 printk(KERN_WARNING "nf_queue: OOM "
 461                                       "in mangle, dropping packet\n");
 462                                 return err;
 463                         }
 464                 }
 465                 skb_put(e->skb, diff);
 466         }
 467         if (!skb_make_writable(e->skb, data_len))
 468                 return -ENOMEM;
 469         skb_copy_to_linear_data(e->skb, data, data_len);
 470         e->skb->ip_summed = CHECKSUM_NONE;
 471         return 0;
 472 }
 473
 474 static int
 475 nfqnl_set_mode(struct nfqnl_instance *queue,
 476                unsigned char mode, unsigned int range)
 477 {
 478         int status = 0;
 479
 480         spin_lock_bh(&queue->lock);
 481         switch (mode) {
 482         case NFQNL_COPY_NONE:
 483         case NFQNL_COPY_META:
 484                 queue->copy_mode = mode;
 485                 queue->copy_range = 0;
 486                 break;
 487
 488         case NFQNL_COPY_PACKET:
 489                 queue->copy_mode = mode;
 490                 /* we're using struct nlattr which has 16bit nla_len */
 491                 if (range > 0xffff)
 492                         queue->copy_range = 0xffff;
 493                 else
 494                         queue->copy_range = range;
 495                 break;
 496
 497         default:
 498                 status = -EINVAL;
 499
 500         }
 501         spin_unlock_bh(&queue->lock);
 502
 503         return status;
 504 }
 505
 506 static int
 507 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
 508 {
 509         if (entry->indev)
 510                 if (entry->indev->ifindex == ifindex)
 511                         return 1;
 512         if (entry->outdev)
 513                 if (entry->outdev->ifindex == ifindex)
 514                         return 1;
 515 #ifdef CONFIG_BRIDGE_NETFILTER
 516         if (entry->skb->nf_bridge) {
 517                 if (entry->skb->nf_bridge->physindev &&
 518                     entry->skb->nf_bridge->physindev->ifindex == ifindex)
 519                         return 1;
 520                 if (entry->skb->nf_bridge->physoutdev &&
 521                     entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
 522                         return 1;
 523         }
 524 #endif
 525         return 0;
 526 }
 527
 528 /* drop all packets with either indev or outdev == ifindex from all queue
 529  * instances */
 530 static void
 531 nfqnl_dev_drop(int ifindex)
 532 {
 533         int i;
 534
 535         rcu_read_lock();
 536
 537         for (i = 0; i < INSTANCE_BUCKETS; i++) {
 538                 struct hlist_node *tmp;
 539                 struct nfqnl_instance *inst;
 540                 struct hlist_head *head = &instance_table[i];
 541
 542                 hlist_for_each_entry_rcu(inst, tmp, head, hlist)
 543                         nfqnl_flush(inst, dev_cmp, ifindex);
 544         }
 545
 546         rcu_read_unlock();
 547 }
 548
 549 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
 550
 551 static int
 552 nfqnl_rcv_dev_event(struct notifier_block *this,
 553                     unsigned long event, void *ptr)
 554 {
 555         struct net_device *dev = ptr;
 556
 557         if (dev->nd_net != &init_net)
 558                 return NOTIFY_DONE;
 559
 560         /* Drop any packets associated with the downed device */
 561         if (event == NETDEV_DOWN)
 562                 nfqnl_dev_drop(dev->ifindex);
 563         return NOTIFY_DONE;
 564 }
 565
 566 static struct notifier_block nfqnl_dev_notifier = {
 567         .notifier_call  = nfqnl_rcv_dev_event,
 568 };
 569
 570 static int
 571 nfqnl_rcv_nl_event(struct notifier_block *this,
 572                    unsigned long event, void *ptr)
 573 {
 574         struct netlink_notify *n = ptr;
 575
 576         if (event == NETLINK_URELEASE &&
 577             n->protocol == NETLINK_NETFILTER && n->pid) {
 578                 int i;
 579
 580                 /* destroy all instances for this pid */
 581                 spin_lock(&instances_lock);
 582                 for (i = 0; i < INSTANCE_BUCKETS; i++) {
 583                         struct hlist_node *tmp, *t2;
 584                         struct nfqnl_instance *inst;
 585                         struct hlist_head *head = &instance_table[i];
 586
 587                         hlist_for_each_entry_safe(inst, tmp, t2, head, hlist) {
 588                                 if ((n->net == &init_net) &&
 589                                     (n->pid == inst->peer_pid))
 590                                         __instance_destroy(inst);
 591                         }
 592                 }
 593                 spin_unlock(&instances_lock);
 594         }
 595         return NOTIFY_DONE;
 596 }
 597
 598 static struct notifier_block nfqnl_rtnl_notifier = {
 599         .notifier_call  = nfqnl_rcv_nl_event,
 600 };
 601
 602 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
 603         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 604         [NFQA_MARK]             = { .type = NLA_U32 },
 605         [NFQA_PAYLOAD]          = { .type = NLA_UNSPEC },
 606 };
 607
 608 static int
 609 nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
 610                    struct nlmsghdr *nlh, struct nlattr *nfqa[])
 611 {
 612         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 613         u_int16_t queue_num = ntohs(nfmsg->res_id);
 614
 615         struct nfqnl_msg_verdict_hdr *vhdr;
 616         struct nfqnl_instance *queue;
 617         unsigned int verdict;
 618         struct nf_queue_entry *entry;
 619         int err;
 620
 621         rcu_read_lock();
 622         queue = instance_lookup(queue_num);
 623         if (!queue) {
 624                 err = -ENODEV;
 625                 goto err_out_unlock;
 626         }
 627
 628         if (queue->peer_pid != NETLINK_CB(skb).pid) {
 629                 err = -EPERM;
 630                 goto err_out_unlock;
 631         }
 632
 633         if (!nfqa[NFQA_VERDICT_HDR]) {
 634                 err = -EINVAL;
 635                 goto err_out_unlock;
 636         }
 637
 638         vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
 639         verdict = ntohl(vhdr->verdict);
 640
 641         if ((verdict & NF_VERDICT_MASK) > NF_MAX_VERDICT) {
 642                 err = -EINVAL;
 643                 goto err_out_unlock;
 644         }
 645
 646         entry = find_dequeue_entry(queue, ntohl(vhdr->id));
 647         if (entry == NULL) {
 648                 err = -ENOENT;
 649                 goto err_out_unlock;
 650         }
 651         rcu_read_unlock();
 652
 653         if (nfqa[NFQA_PAYLOAD]) {
 654                 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
 655                                  nla_len(nfqa[NFQA_PAYLOAD]), entry) < 0)
 656                         verdict = NF_DROP;
 657         }
 658
 659         if (nfqa[NFQA_MARK])
 660                 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
 661
 662         nf_reinject(entry, verdict);
 663         return 0;
 664
 665 err_out_unlock:
 666         rcu_read_unlock();
 667         return err;
 668 }
 669
 670 static int
 671 nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
 672                   struct nlmsghdr *nlh, struct nlattr *nfqa[])
 673 {
 674         return -ENOTSUPP;
 675 }
 676
 677 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
 678         [NFQA_CFG_CMD]          = { .len = sizeof(struct nfqnl_msg_config_cmd) },
 679         [NFQA_CFG_PARAMS]       = { .len = sizeof(struct nfqnl_msg_config_params) },
 680 };
 681
 682 static const struct nf_queue_handler nfqh = {
 683         .name   = "nf_queue",
 684         .outfn  = &nfqnl_enqueue_packet,
 685 };
 686
 687 static int
 688 nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
 689                   struct nlmsghdr *nlh, struct nlattr *nfqa[])
 690 {
 691         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 692         u_int16_t queue_num = ntohs(nfmsg->res_id);
 693         struct nfqnl_instance *queue;
 694         struct nfqnl_msg_config_cmd *cmd = NULL;
 695         int ret = 0;
 696
 697         if (nfqa[NFQA_CFG_CMD]) {
 698                 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
 699
 700                 /* Commands without queue context - might sleep */
 701                 switch (cmd->command) {
 702                 case NFQNL_CFG_CMD_PF_BIND:
 703                         ret = nf_register_queue_handler(ntohs(cmd->pf),
 704                                                         &nfqh);
 705                         break;
 706                 case NFQNL_CFG_CMD_PF_UNBIND:
 707                         ret = nf_unregister_queue_handler(ntohs(cmd->pf),
 708                                                           &nfqh);
 709                         break;
 710                 default:
 711                         break;
 712                 }
 713
 714                 if (ret < 0)
 715                         return ret;
 716         }
 717
 718         rcu_read_lock();
 719         queue = instance_lookup(queue_num);
 720         if (queue && queue->peer_pid != NETLINK_CB(skb).pid) {
 721                 ret = -EPERM;
 722                 goto err_out_unlock;
 723         }
 724
 725         if (cmd != NULL) {
 726                 switch (cmd->command) {
 727                 case NFQNL_CFG_CMD_BIND:
 728                         if (queue) {
 729                                 ret = -EBUSY;
 730                                 goto err_out_unlock;
 731                         }
 732                         queue = instance_create(queue_num, NETLINK_CB(skb).pid);
 733                         if (!queue) {
 734                                 ret = -EINVAL;
 735                                 goto err_out_unlock;
 736                         }
 737                         break;
 738                 case NFQNL_CFG_CMD_UNBIND:
 739                         if (!queue) {
 740                                 ret = -ENODEV;
 741                                 goto err_out_unlock;
 742                         }
 743                         instance_destroy(queue);
 744                         break;
 745                 case NFQNL_CFG_CMD_PF_BIND:
 746                 case NFQNL_CFG_CMD_PF_UNBIND:
 747                         break;
 748                 default:
 749                         ret = -EINVAL;
 750                         break;
 751                 }
 752         }
 753
 754         if (nfqa[NFQA_CFG_PARAMS]) {
 755                 struct nfqnl_msg_config_params *params;
 756
 757                 if (!queue) {
 758                         ret = -ENODEV;
 759                         goto err_out_unlock;
 760                 }
 761                 params = nla_data(nfqa[NFQA_CFG_PARAMS]);
 762                 nfqnl_set_mode(queue, params->copy_mode,
 763                                 ntohl(params->copy_range));
 764         }
 765
 766         if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
 767                 __be32 *queue_maxlen;
 768
 769                 if (!queue) {
 770                         ret = -ENODEV;
 771                         goto err_out_unlock;
 772                 }
 773                 queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
 774                 spin_lock_bh(&queue->lock);
 775                 queue->queue_maxlen = ntohl(*queue_maxlen);
 776                 spin_unlock_bh(&queue->lock);
 777         }
 778
 779 err_out_unlock:
 780         rcu_read_unlock();
 781         return ret;
 782 }
 783
 784 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
 785         [NFQNL_MSG_PACKET]      = { .call = nfqnl_recv_unsupp,
 786                                     .attr_count = NFQA_MAX, },
 787         [NFQNL_MSG_VERDICT]     = { .call = nfqnl_recv_verdict,
 788                                     .attr_count = NFQA_MAX,
 789                                     .policy = nfqa_verdict_policy },
 790         [NFQNL_MSG_CONFIG]      = { .call = nfqnl_recv_config,
 791                                     .attr_count = NFQA_CFG_MAX,
 792                                     .policy = nfqa_cfg_policy },
 793 };
 794
 795 static const struct nfnetlink_subsystem nfqnl_subsys = {
 796         .name           = "nf_queue",
 797         .subsys_id      = NFNL_SUBSYS_QUEUE,
 798         .cb_count       = NFQNL_MSG_MAX,
 799         .cb             = nfqnl_cb,
 800 };
 801
 802 #ifdef CONFIG_PROC_FS
 803 struct iter_state {
 804         unsigned int bucket;
 805 };
 806
 807 static struct hlist_node *get_first(struct seq_file *seq)
 808 {
 809         struct iter_state *st = seq->private;
 810
 811         if (!st)
 812                 return NULL;
 813
 814         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
 815                 if (!hlist_empty(&instance_table[st->bucket]))
 816                         return instance_table[st->bucket].first;
 817         }
 818         return NULL;
 819 }
 820
 821 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
 822 {
 823         struct iter_state *st = seq->private;
 824
 825         h = h->next;
 826         while (!h) {
 827                 if (++st->bucket >= INSTANCE_BUCKETS)
 828                         return NULL;
 829
 830                 h = instance_table[st->bucket].first;
 831         }
 832         return h;
 833 }
 834
 835 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
 836 {
 837         struct hlist_node *head;
 838         head = get_first(seq);
 839
 840         if (head)
 841                 while (pos && (head = get_next(seq, head)))
 842                         pos--;
 843         return pos ? NULL : head;
 844 }
 845
 846 static void *seq_start(struct seq_file *seq, loff_t *pos)
 847 {
 848         spin_lock(&instances_lock);
 849         return get_idx(seq, *pos);
 850 }
 851
 852 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
 853 {
 854         (*pos)++;
 855         return get_next(s, v);
 856 }
 857
 858 static void seq_stop(struct seq_file *s, void *v)
 859 {
 860         spin_unlock(&instances_lock);
 861 }
 862
 863 static int seq_show(struct seq_file *s, void *v)
 864 {
 865         const struct nfqnl_instance *inst = v;
 866
 867         return seq_printf(s, "%5d %6d %5d %1d %5d %5d %5d %8d %2d\n",
 868                           inst->queue_num,
 869                           inst->peer_pid, inst->queue_total,
 870                           inst->copy_mode, inst->copy_range,
 871                           inst->queue_dropped, inst->queue_user_dropped,
 872                           inst->id_sequence, 1);
 873 }
 874
 875 static const struct seq_operations nfqnl_seq_ops = {
 876         .start  = seq_start,
 877         .next   = seq_next,
 878         .stop   = seq_stop,
 879         .show   = seq_show,
 880 };
 881
 882 static int nfqnl_open(struct inode *inode, struct file *file)
 883 {
 884         return seq_open_private(file, &nfqnl_seq_ops,
 885                         sizeof(struct iter_state));
 886 }
 887
 888 static const struct file_operations nfqnl_file_ops = {
 889         .owner   = THIS_MODULE,
 890         .open    = nfqnl_open,
 891         .read    = seq_read,
 892         .llseek  = seq_lseek,
 893         .release = seq_release_private,
 894 };
 895
 896 #endif /* PROC_FS */
 897
 898 static int __init nfnetlink_queue_init(void)
 899 {
 900         int i, status = -ENOMEM;
 901 #ifdef CONFIG_PROC_FS
 902         struct proc_dir_entry *proc_nfqueue;
 903 #endif
 904
 905         for (i = 0; i < INSTANCE_BUCKETS; i++)
 906                 INIT_HLIST_HEAD(&instance_table[i]);
 907
 908         netlink_register_notifier(&nfqnl_rtnl_notifier);
 909         status = nfnetlink_subsys_register(&nfqnl_subsys);
 910         if (status < 0) {
 911                 printk(KERN_ERR "nf_queue: failed to create netlink socket\n");
 912                 goto cleanup_netlink_notifier;
 913         }
 914
 915 #ifdef CONFIG_PROC_FS
 916         proc_nfqueue = create_proc_entry("nfnetlink_queue", 0440,
 917                                          proc_net_netfilter);
 918         if (!proc_nfqueue)
 919                 goto cleanup_subsys;
 920         proc_nfqueue->proc_fops = &nfqnl_file_ops;
 921 #endif
 922
 923         register_netdevice_notifier(&nfqnl_dev_notifier);
 924         return status;
 925
 926 #ifdef CONFIG_PROC_FS
 927 cleanup_subsys:
 928         nfnetlink_subsys_unregister(&nfqnl_subsys);
 929 #endif
 930 cleanup_netlink_notifier:
 931         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
 932         return status;
 933 }
 934
 935 static void __exit nfnetlink_queue_fini(void)
 936 {
 937         nf_unregister_queue_handlers(&nfqh);
 938         unregister_netdevice_notifier(&nfqnl_dev_notifier);
 939 #ifdef CONFIG_PROC_FS
 940         remove_proc_entry("nfnetlink_queue", proc_net_netfilter);
 941 #endif
 942         nfnetlink_subsys_unregister(&nfqnl_subsys);
 943         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
 944 }
 945
 946 MODULE_DESCRIPTION("netfilter packet queue handler");
 947 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
 948 MODULE_LICENSE("GPL");
 949 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
 950
 951 module_init(nfnetlink_queue_init);
 952 module_exit(nfnetlink_queue_fini);