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 #if 0
  41 #define QDEBUG(x, args ...)     printk(KERN_DEBUG "%s(%d):%s(): " x,       \
  42                                         __FILE__, __LINE__, __FUNCTION__,  \
  43                                         ## args)
  44 #else
  45 #define QDEBUG(x, ...)
  46 #endif
  47
  48 struct nfqnl_instance {
  49         struct hlist_node hlist;                /* global list of queues */
  50         struct rcu_head rcu;
  51
  52         int peer_pid;
  53         unsigned int queue_maxlen;
  54         unsigned int copy_range;
  55         unsigned int queue_total;
  56         unsigned int queue_dropped;
  57         unsigned int queue_user_dropped;
  58
  59         unsigned int id_sequence;               /* 'sequence' of pkt ids */
  60
  61         u_int16_t queue_num;                    /* number of this queue */
  62         u_int8_t copy_mode;
  63
  64         spinlock_t lock;
  65
  66         struct list_head queue_list;            /* packets in queue */
  67 };
  68
  69 typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
  70
  71 static DEFINE_SPINLOCK(instances_lock);
  72
  73 #define INSTANCE_BUCKETS        16
  74 static struct hlist_head instance_table[INSTANCE_BUCKETS];
  75
  76 static inline u_int8_t instance_hashfn(u_int16_t queue_num)
  77 {
  78         return ((queue_num >> 8) | queue_num) % INSTANCE_BUCKETS;
  79 }
  80
  81 static struct nfqnl_instance *
  82 instance_lookup(u_int16_t queue_num)
  83 {
  84         struct hlist_head *head;
  85         struct hlist_node *pos;
  86         struct nfqnl_instance *inst;
  87
  88         head = &instance_table[instance_hashfn(queue_num)];
  89         hlist_for_each_entry_rcu(inst, pos, head, hlist) {
  90                 if (inst->queue_num == queue_num)
  91                         return inst;
  92         }
  93         return NULL;
  94 }
  95
  96 static struct nfqnl_instance *
  97 instance_create(u_int16_t queue_num, int pid)
  98 {
  99         struct nfqnl_instance *inst;
 100         unsigned int h;
 101
 102         QDEBUG("entering for queue_num=%u, pid=%d\n", queue_num, pid);
 103
 104         spin_lock(&instances_lock);
 105         if (instance_lookup(queue_num)) {
 106                 inst = NULL;
 107                 QDEBUG("aborting, instance already exists\n");
 108                 goto out_unlock;
 109         }
 110
 111         inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
 112         if (!inst)
 113                 goto out_unlock;
 114
 115         inst->queue_num = queue_num;
 116         inst->peer_pid = pid;
 117         inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
 118         inst->copy_range = 0xfffff;
 119         inst->copy_mode = NFQNL_COPY_NONE;
 120         spin_lock_init(&inst->lock);
 121         INIT_LIST_HEAD(&inst->queue_list);
 122         INIT_RCU_HEAD(&inst->rcu);
 123
 124         if (!try_module_get(THIS_MODULE))
 125                 goto out_free;
 126
 127         h = instance_hashfn(queue_num);
 128         hlist_add_head_rcu(&inst->hlist, &instance_table[h]);
 129
 130         spin_unlock(&instances_lock);
 131
 132         QDEBUG("successfully created new instance\n");
 133
 134         return inst;
 135
 136 out_free:
 137         kfree(inst);
 138 out_unlock:
 139         spin_unlock(&instances_lock);
 140         return NULL;
 141 }
 142
 143 static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
 144                         unsigned long data);
 145
 146 static void
 147 instance_destroy_rcu(struct rcu_head *head)
 148 {
 149         struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
 150                                                    rcu);
 151
 152         nfqnl_flush(inst, NULL, 0);
 153         kfree(inst);
 154         module_put(THIS_MODULE);
 155 }
 156
 157 static void
 158 __instance_destroy(struct nfqnl_instance *inst)
 159 {
 160         hlist_del_rcu(&inst->hlist);
 161         call_rcu(&inst->rcu, instance_destroy_rcu);
 162 }
 163
 164 static void
 165 instance_destroy(struct nfqnl_instance *inst)
 166 {
 167         spin_lock(&instances_lock);
 168         __instance_destroy(inst);
 169         spin_unlock(&instances_lock);
 170 }
 171
 172 static inline void
 173 __enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
 174 {
 175        list_add_tail(&entry->list, &queue->queue_list);
 176        queue->queue_total++;
 177 }
 178
 179 static struct nf_queue_entry *
 180 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
 181 {
 182         struct nf_queue_entry *entry = NULL, *i;
 183
 184         spin_lock_bh(&queue->lock);
 185
 186         list_for_each_entry(i, &queue->queue_list, list) {
 187                 if (i->id == id) {
 188                         entry = i;
 189                         break;
 190                 }
 191         }
 192
 193         if (entry) {
 194                 list_del(&entry->list);
 195                 queue->queue_total--;
 196         }
 197
 198         spin_unlock_bh(&queue->lock);
 199
 200         return entry;
 201 }
 202
 203 static void
 204 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
 205 {
 206         struct nf_queue_entry *entry, *next;
 207
 208         spin_lock_bh(&queue->lock);
 209         list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
 210                 if (!cmpfn || cmpfn(entry, data)) {
 211                         list_del(&entry->list);
 212                         queue->queue_total--;
 213                         nf_reinject(entry, NF_DROP);
 214                 }
 215         }
 216         spin_unlock_bh(&queue->lock);
 217 }
 218
 219 static struct sk_buff *
 220 nfqnl_build_packet_message(struct nfqnl_instance *queue,
 221                            struct nf_queue_entry *entry, int *errp)
 222 {
 223         sk_buff_data_t old_tail;
 224         size_t size;
 225         size_t data_len = 0;
 226         struct sk_buff *skb;
 227         struct nfqnl_msg_packet_hdr pmsg;
 228         struct nlmsghdr *nlh;
 229         struct nfgenmsg *nfmsg;
 230         struct sk_buff *entskb = entry->skb;
 231         struct net_device *indev;
 232         struct net_device *outdev;
 233         __be32 tmp_uint;
 234
 235         QDEBUG("entered\n");
 236
 237         size =    NLMSG_ALIGN(sizeof(struct nfgenmsg))
 238                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
 239                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 240                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 241 #ifdef CONFIG_BRIDGE_NETFILTER
 242                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 243                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 244 #endif
 245                 + nla_total_size(sizeof(u_int32_t))     /* mark */
 246                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
 247                 + nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
 248
 249         outdev = entry->outdev;
 250
 251         spin_lock_bh(&queue->lock);
 252
 253         switch (queue->copy_mode) {
 254         case NFQNL_COPY_META:
 255         case NFQNL_COPY_NONE:
 256                 data_len = 0;
 257                 break;
 258
 259         case NFQNL_COPY_PACKET:
 260                 if ((entskb->ip_summed == CHECKSUM_PARTIAL ||
 261                      entskb->ip_summed == CHECKSUM_COMPLETE) &&
 262                     (*errp = skb_checksum_help(entskb))) {
 263                         spin_unlock_bh(&queue->lock);
 264                         return NULL;
 265                 }
 266                 if (queue->copy_range == 0
 267                     || queue->copy_range > entskb->len)
 268                         data_len = entskb->len;
 269                 else
 270                         data_len = queue->copy_range;
 271
 272                 size += nla_total_size(data_len);
 273                 break;
 274
 275         default:
 276                 *errp = -EINVAL;
 277                 spin_unlock_bh(&queue->lock);
 278                 return NULL;
 279         }
 280
 281         entry->id = queue->id_sequence++;
 282
 283         spin_unlock_bh(&queue->lock);
 284
 285         skb = alloc_skb(size, GFP_ATOMIC);
 286         if (!skb)
 287                 goto nlmsg_failure;
 288
 289         old_tail = skb->tail;
 290         nlh = NLMSG_PUT(skb, 0, 0,
 291                         NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
 292                         sizeof(struct nfgenmsg));
 293         nfmsg = NLMSG_DATA(nlh);
 294         nfmsg->nfgen_family = entry->pf;
 295         nfmsg->version = NFNETLINK_V0;
 296         nfmsg->res_id = htons(queue->queue_num);
 297
 298         pmsg.packet_id          = htonl(entry->id);
 299         pmsg.hw_protocol        = entskb->protocol;
 300         pmsg.hook               = entry->hook;
 301
 302         NLA_PUT(skb, NFQA_PACKET_HDR, sizeof(pmsg), &pmsg);
 303
 304         indev = entry->indev;
 305         if (indev) {
 306                 tmp_uint = htonl(indev->ifindex);
 307 #ifndef CONFIG_BRIDGE_NETFILTER
 308                 NLA_PUT(skb, NFQA_IFINDEX_INDEV, sizeof(tmp_uint), &tmp_uint);
 309 #else
 310                 if (entry->pf == PF_BRIDGE) {
 311                         /* Case 1: indev is physical input device, we need to
 312                          * look for bridge group (when called from
 313                          * netfilter_bridge) */
 314                         NLA_PUT(skb, NFQA_IFINDEX_PHYSINDEV, sizeof(tmp_uint),
 315                                 &tmp_uint);
 316                         /* this is the bridge group "brX" */
 317                         tmp_uint = htonl(indev->br_port->br->dev->ifindex);
 318                         NLA_PUT(skb, NFQA_IFINDEX_INDEV, sizeof(tmp_uint),
 319                                 &tmp_uint);
 320                 } else {
 321                         /* Case 2: indev is bridge group, we need to look for
 322                          * physical device (when called from ipv4) */
 323                         NLA_PUT(skb, NFQA_IFINDEX_INDEV, sizeof(tmp_uint),
 324                                 &tmp_uint);
 325                         if (entskb->nf_bridge
 326                             && entskb->nf_bridge->physindev) {
 327                                 tmp_uint = htonl(entskb->nf_bridge->physindev->ifindex);
 328                                 NLA_PUT(skb, NFQA_IFINDEX_PHYSINDEV,
 329                                         sizeof(tmp_uint), &tmp_uint);
 330                         }
 331                 }
 332 #endif
 333         }
 334
 335         if (outdev) {
 336                 tmp_uint = htonl(outdev->ifindex);
 337 #ifndef CONFIG_BRIDGE_NETFILTER
 338                 NLA_PUT(skb, NFQA_IFINDEX_OUTDEV, sizeof(tmp_uint), &tmp_uint);
 339 #else
 340                 if (entry->pf == PF_BRIDGE) {
 341                         /* Case 1: outdev is physical output device, we need to
 342                          * look for bridge group (when called from
 343                          * netfilter_bridge) */
 344                         NLA_PUT(skb, NFQA_IFINDEX_PHYSOUTDEV, sizeof(tmp_uint),
 345                                 &tmp_uint);
 346                         /* this is the bridge group "brX" */
 347                         tmp_uint = htonl(outdev->br_port->br->dev->ifindex);
 348                         NLA_PUT(skb, NFQA_IFINDEX_OUTDEV, sizeof(tmp_uint),
 349                                 &tmp_uint);
 350                 } else {
 351                         /* Case 2: outdev is bridge group, we need to look for
 352                          * physical output device (when called from ipv4) */
 353                         NLA_PUT(skb, NFQA_IFINDEX_OUTDEV, sizeof(tmp_uint),
 354                                 &tmp_uint);
 355                         if (entskb->nf_bridge
 356                             && entskb->nf_bridge->physoutdev) {
 357                                 tmp_uint = htonl(entskb->nf_bridge->physoutdev->ifindex);
 358                                 NLA_PUT(skb, NFQA_IFINDEX_PHYSOUTDEV,
 359                                         sizeof(tmp_uint), &tmp_uint);
 360                         }
 361                 }
 362 #endif
 363         }
 364
 365         if (entskb->mark) {
 366                 tmp_uint = htonl(entskb->mark);
 367                 NLA_PUT(skb, NFQA_MARK, sizeof(u_int32_t), &tmp_uint);
 368         }
 369
 370         if (indev && entskb->dev) {
 371                 struct nfqnl_msg_packet_hw phw;
 372                 int len = dev_parse_header(entskb, phw.hw_addr);
 373                 if (len) {
 374                         phw.hw_addrlen = htons(len);
 375                         NLA_PUT(skb, NFQA_HWADDR, sizeof(phw), &phw);
 376                 }
 377         }
 378
 379         if (entskb->tstamp.tv64) {
 380                 struct nfqnl_msg_packet_timestamp ts;
 381                 struct timeval tv = ktime_to_timeval(entskb->tstamp);
 382                 ts.sec = cpu_to_be64(tv.tv_sec);
 383                 ts.usec = cpu_to_be64(tv.tv_usec);
 384
 385                 NLA_PUT(skb, NFQA_TIMESTAMP, sizeof(ts), &ts);
 386         }
 387
 388         if (data_len) {
 389                 struct nlattr *nla;
 390                 int size = nla_attr_size(data_len);
 391
 392                 if (skb_tailroom(skb) < nla_total_size(data_len)) {
 393                         printk(KERN_WARNING "nf_queue: no tailroom!\n");
 394                         goto nlmsg_failure;
 395                 }
 396
 397                 nla = (struct nlattr *)skb_put(skb, nla_total_size(data_len));
 398                 nla->nla_type = NFQA_PAYLOAD;
 399                 nla->nla_len = size;
 400
 401                 if (skb_copy_bits(entskb, 0, nla_data(nla), data_len))
 402                         BUG();
 403         }
 404
 405         nlh->nlmsg_len = skb->tail - old_tail;
 406         return skb;
 407
 408 nlmsg_failure:
 409 nla_put_failure:
 410         if (skb)
 411                 kfree_skb(skb);
 412         *errp = -EINVAL;
 413         if (net_ratelimit())
 414                 printk(KERN_ERR "nf_queue: error creating packet message\n");
 415         return NULL;
 416 }
 417
 418 static int
 419 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 420 {
 421         int status = -EINVAL;
 422         struct sk_buff *nskb;
 423         struct nfqnl_instance *queue;
 424
 425         QDEBUG("entered\n");
 426
 427         /* rcu_read_lock()ed by nf_hook_slow() */
 428         queue = instance_lookup(queuenum);
 429         if (!queue) {
 430                 QDEBUG("no queue instance matching\n");
 431                 return -EINVAL;
 432         }
 433
 434         if (queue->copy_mode == NFQNL_COPY_NONE) {
 435                 QDEBUG("mode COPY_NONE, aborting\n");
 436                 return -EAGAIN;
 437         }
 438
 439         nskb = nfqnl_build_packet_message(queue, entry, &status);
 440         if (nskb == NULL)
 441                 return status;
 442
 443         spin_lock_bh(&queue->lock);
 444
 445         if (!queue->peer_pid)
 446                 goto err_out_free_nskb;
 447
 448         if (queue->queue_total >= queue->queue_maxlen) {
 449                 queue->queue_dropped++;
 450                 status = -ENOSPC;
 451                 if (net_ratelimit())
 452                           printk(KERN_WARNING "nf_queue: full at %d entries, "
 453                                  "dropping packets(s). Dropped: %d\n",
 454                                  queue->queue_total, queue->queue_dropped);
 455                 goto err_out_free_nskb;
 456         }
 457
 458         /* nfnetlink_unicast will either free the nskb or add it to a socket */
 459         status = nfnetlink_unicast(nskb, queue->peer_pid, MSG_DONTWAIT);
 460         if (status < 0) {
 461                 queue->queue_user_dropped++;
 462                 goto err_out_unlock;
 463         }
 464
 465         __enqueue_entry(queue, entry);
 466
 467         spin_unlock_bh(&queue->lock);
 468         return status;
 469
 470 err_out_free_nskb:
 471         kfree_skb(nskb);
 472
 473 err_out_unlock:
 474         spin_unlock_bh(&queue->lock);
 475         return status;
 476 }
 477
 478 static int
 479 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e)
 480 {
 481         int diff;
 482         int err;
 483
 484         diff = data_len - e->skb->len;
 485         if (diff < 0) {
 486                 if (pskb_trim(e->skb, data_len))
 487                         return -ENOMEM;
 488         } else if (diff > 0) {
 489                 if (data_len > 0xFFFF)
 490                         return -EINVAL;
 491                 if (diff > skb_tailroom(e->skb)) {
 492                         err = pskb_expand_head(e->skb, 0,
 493                                                diff - skb_tailroom(e->skb),
 494                                                GFP_ATOMIC);
 495                         if (err) {
 496                                 printk(KERN_WARNING "nf_queue: OOM "
 497                                       "in mangle, dropping packet\n");
 498                                 return err;
 499                         }
 500                 }
 501                 skb_put(e->skb, diff);
 502         }
 503         if (!skb_make_writable(e->skb, data_len))
 504                 return -ENOMEM;
 505         skb_copy_to_linear_data(e->skb, data, data_len);
 506         e->skb->ip_summed = CHECKSUM_NONE;
 507         return 0;
 508 }
 509
 510 static int
 511 nfqnl_set_mode(struct nfqnl_instance *queue,
 512                unsigned char mode, unsigned int range)
 513 {
 514         int status = 0;
 515
 516         spin_lock_bh(&queue->lock);
 517         switch (mode) {
 518         case NFQNL_COPY_NONE:
 519         case NFQNL_COPY_META:
 520                 queue->copy_mode = mode;
 521                 queue->copy_range = 0;
 522                 break;
 523
 524         case NFQNL_COPY_PACKET:
 525                 queue->copy_mode = mode;
 526                 /* we're using struct nlattr which has 16bit nla_len */
 527                 if (range > 0xffff)
 528                         queue->copy_range = 0xffff;
 529                 else
 530                         queue->copy_range = range;
 531                 break;
 532
 533         default:
 534                 status = -EINVAL;
 535
 536         }
 537         spin_unlock_bh(&queue->lock);
 538
 539         return status;
 540 }
 541
 542 static int
 543 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
 544 {
 545         if (entry->indev)
 546                 if (entry->indev->ifindex == ifindex)
 547                         return 1;
 548         if (entry->outdev)
 549                 if (entry->outdev->ifindex == ifindex)
 550                         return 1;
 551 #ifdef CONFIG_BRIDGE_NETFILTER
 552         if (entry->skb->nf_bridge) {
 553                 if (entry->skb->nf_bridge->physindev &&
 554                     entry->skb->nf_bridge->physindev->ifindex == ifindex)
 555                         return 1;
 556                 if (entry->skb->nf_bridge->physoutdev &&
 557                     entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
 558                         return 1;
 559         }
 560 #endif
 561         return 0;
 562 }
 563
 564 /* drop all packets with either indev or outdev == ifindex from all queue
 565  * instances */
 566 static void
 567 nfqnl_dev_drop(int ifindex)
 568 {
 569         int i;
 570
 571         QDEBUG("entering for ifindex %u\n", ifindex);
 572
 573         rcu_read_lock();
 574
 575         for (i = 0; i < INSTANCE_BUCKETS; i++) {
 576                 struct hlist_node *tmp;
 577                 struct nfqnl_instance *inst;
 578                 struct hlist_head *head = &instance_table[i];
 579
 580                 hlist_for_each_entry_rcu(inst, tmp, head, hlist)
 581                         nfqnl_flush(inst, dev_cmp, ifindex);
 582         }
 583
 584         rcu_read_unlock();
 585 }
 586
 587 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
 588
 589 static int
 590 nfqnl_rcv_dev_event(struct notifier_block *this,
 591                     unsigned long event, void *ptr)
 592 {
 593         struct net_device *dev = ptr;
 594
 595         if (dev->nd_net != &init_net)
 596                 return NOTIFY_DONE;
 597
 598         /* Drop any packets associated with the downed device */
 599         if (event == NETDEV_DOWN)
 600                 nfqnl_dev_drop(dev->ifindex);
 601         return NOTIFY_DONE;
 602 }
 603
 604 static struct notifier_block nfqnl_dev_notifier = {
 605         .notifier_call  = nfqnl_rcv_dev_event,
 606 };
 607
 608 static int
 609 nfqnl_rcv_nl_event(struct notifier_block *this,
 610                    unsigned long event, void *ptr)
 611 {
 612         struct netlink_notify *n = ptr;
 613
 614         if (event == NETLINK_URELEASE &&
 615             n->protocol == NETLINK_NETFILTER && n->pid) {
 616                 int i;
 617
 618                 /* destroy all instances for this pid */
 619                 spin_lock(&instances_lock);
 620                 for (i = 0; i < INSTANCE_BUCKETS; i++) {
 621                         struct hlist_node *tmp, *t2;
 622                         struct nfqnl_instance *inst;
 623                         struct hlist_head *head = &instance_table[i];
 624
 625                         hlist_for_each_entry_safe(inst, tmp, t2, head, hlist) {
 626                                 if ((n->net == &init_net) &&
 627                                     (n->pid == inst->peer_pid))
 628                                         __instance_destroy(inst);
 629                         }
 630                 }
 631                 spin_unlock(&instances_lock);
 632         }
 633         return NOTIFY_DONE;
 634 }
 635
 636 static struct notifier_block nfqnl_rtnl_notifier = {
 637         .notifier_call  = nfqnl_rcv_nl_event,
 638 };
 639
 640 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
 641         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 642         [NFQA_MARK]             = { .type = NLA_U32 },
 643         [NFQA_PAYLOAD]          = { .type = NLA_UNSPEC },
 644 };
 645
 646 static int
 647 nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
 648                    struct nlmsghdr *nlh, struct nlattr *nfqa[])
 649 {
 650         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 651         u_int16_t queue_num = ntohs(nfmsg->res_id);
 652
 653         struct nfqnl_msg_verdict_hdr *vhdr;
 654         struct nfqnl_instance *queue;
 655         unsigned int verdict;
 656         struct nf_queue_entry *entry;
 657         int err;
 658
 659         rcu_read_lock();
 660         queue = instance_lookup(queue_num);
 661         if (!queue) {
 662                 err = -ENODEV;
 663                 goto err_out_unlock;
 664         }
 665
 666         if (queue->peer_pid != NETLINK_CB(skb).pid) {
 667                 err = -EPERM;
 668                 goto err_out_unlock;
 669         }
 670
 671         if (!nfqa[NFQA_VERDICT_HDR]) {
 672                 err = -EINVAL;
 673                 goto err_out_unlock;
 674         }
 675
 676         vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
 677         verdict = ntohl(vhdr->verdict);
 678
 679         if ((verdict & NF_VERDICT_MASK) > NF_MAX_VERDICT) {
 680                 err = -EINVAL;
 681                 goto err_out_unlock;
 682         }
 683
 684         entry = find_dequeue_entry(queue, ntohl(vhdr->id));
 685         if (entry == NULL) {
 686                 err = -ENOENT;
 687                 goto err_out_unlock;
 688         }
 689         rcu_read_unlock();
 690
 691         if (nfqa[NFQA_PAYLOAD]) {
 692                 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
 693                                  nla_len(nfqa[NFQA_PAYLOAD]), entry) < 0)
 694                         verdict = NF_DROP;
 695         }
 696
 697         if (nfqa[NFQA_MARK])
 698                 entry->skb->mark = ntohl(*(__be32 *)
 699                                          nla_data(nfqa[NFQA_MARK]));
 700
 701         nf_reinject(entry, verdict);
 702         return 0;
 703
 704 err_out_unlock:
 705         rcu_read_unlock();
 706         return err;
 707 }
 708
 709 static int
 710 nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
 711                   struct nlmsghdr *nlh, struct nlattr *nfqa[])
 712 {
 713         return -ENOTSUPP;
 714 }
 715
 716 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
 717         [NFQA_CFG_CMD]          = { .len = sizeof(struct nfqnl_msg_config_cmd) },
 718         [NFQA_CFG_PARAMS]       = { .len = sizeof(struct nfqnl_msg_config_params) },
 719 };
 720
 721 static const struct nf_queue_handler nfqh = {
 722         .name   = "nf_queue",
 723         .outfn  = &nfqnl_enqueue_packet,
 724 };
 725
 726 static int
 727 nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
 728                   struct nlmsghdr *nlh, struct nlattr *nfqa[])
 729 {
 730         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 731         u_int16_t queue_num = ntohs(nfmsg->res_id);
 732         struct nfqnl_instance *queue;
 733         struct nfqnl_msg_config_cmd *cmd = NULL;
 734         int ret = 0;
 735
 736         QDEBUG("entering for msg %u\n", NFNL_MSG_TYPE(nlh->nlmsg_type));
 737
 738         if (nfqa[NFQA_CFG_CMD]) {
 739                 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
 740
 741                 /* Commands without queue context - might sleep */
 742                 switch (cmd->command) {
 743                 case NFQNL_CFG_CMD_PF_BIND:
 744                         ret = nf_register_queue_handler(ntohs(cmd->pf),
 745                                                         &nfqh);
 746                         break;
 747                 case NFQNL_CFG_CMD_PF_UNBIND:
 748                         ret = nf_unregister_queue_handler(ntohs(cmd->pf),
 749                                                           &nfqh);
 750                         break;
 751                 default:
 752                         break;
 753                 }
 754
 755                 if (ret < 0)
 756                         return ret;
 757         }
 758
 759         rcu_read_lock();
 760         queue = instance_lookup(queue_num);
 761         if (queue && queue->peer_pid != NETLINK_CB(skb).pid) {
 762                 ret = -EPERM;
 763                 goto err_out_unlock;
 764         }
 765
 766         if (cmd != NULL) {
 767                 switch (cmd->command) {
 768                 case NFQNL_CFG_CMD_BIND:
 769                         if (queue) {
 770                                 ret = -EBUSY;
 771                                 goto err_out_unlock;
 772                         }
 773                         queue = instance_create(queue_num, NETLINK_CB(skb).pid);
 774                         if (!queue) {
 775                                 ret = -EINVAL;
 776                                 goto err_out_unlock;
 777                         }
 778                         break;
 779                 case NFQNL_CFG_CMD_UNBIND:
 780                         if (!queue) {
 781                                 ret = -ENODEV;
 782                                 goto err_out_unlock;
 783                         }
 784                         instance_destroy(queue);
 785                         break;
 786                 case NFQNL_CFG_CMD_PF_BIND:
 787                 case NFQNL_CFG_CMD_PF_UNBIND:
 788                         break;
 789                 default:
 790                         ret = -EINVAL;
 791                         break;
 792                 }
 793         }
 794
 795         if (nfqa[NFQA_CFG_PARAMS]) {
 796                 struct nfqnl_msg_config_params *params;
 797
 798                 if (!queue) {
 799                         ret = -ENODEV;
 800                         goto err_out_unlock;
 801                 }
 802                 params = nla_data(nfqa[NFQA_CFG_PARAMS]);
 803                 nfqnl_set_mode(queue, params->copy_mode,
 804                                 ntohl(params->copy_range));
 805         }
 806
 807         if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
 808                 __be32 *queue_maxlen;
 809
 810                 if (!queue) {
 811                         ret = -ENODEV;
 812                         goto err_out_unlock;
 813                 }
 814                 queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
 815                 spin_lock_bh(&queue->lock);
 816                 queue->queue_maxlen = ntohl(*queue_maxlen);
 817                 spin_unlock_bh(&queue->lock);
 818         }
 819
 820 err_out_unlock:
 821         rcu_read_unlock();
 822         return ret;
 823 }
 824
 825 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
 826         [NFQNL_MSG_PACKET]      = { .call = nfqnl_recv_unsupp,
 827                                     .attr_count = NFQA_MAX, },
 828         [NFQNL_MSG_VERDICT]     = { .call = nfqnl_recv_verdict,
 829                                     .attr_count = NFQA_MAX,
 830                                     .policy = nfqa_verdict_policy },
 831         [NFQNL_MSG_CONFIG]      = { .call = nfqnl_recv_config,
 832                                     .attr_count = NFQA_CFG_MAX,
 833                                     .policy = nfqa_cfg_policy },
 834 };
 835
 836 static const struct nfnetlink_subsystem nfqnl_subsys = {
 837         .name           = "nf_queue",
 838         .subsys_id      = NFNL_SUBSYS_QUEUE,
 839         .cb_count       = NFQNL_MSG_MAX,
 840         .cb             = nfqnl_cb,
 841 };
 842
 843 #ifdef CONFIG_PROC_FS
 844 struct iter_state {
 845         unsigned int bucket;
 846 };
 847
 848 static struct hlist_node *get_first(struct seq_file *seq)
 849 {
 850         struct iter_state *st = seq->private;
 851
 852         if (!st)
 853                 return NULL;
 854
 855         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
 856                 if (!hlist_empty(&instance_table[st->bucket]))
 857                         return instance_table[st->bucket].first;
 858         }
 859         return NULL;
 860 }
 861
 862 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
 863 {
 864         struct iter_state *st = seq->private;
 865
 866         h = h->next;
 867         while (!h) {
 868                 if (++st->bucket >= INSTANCE_BUCKETS)
 869                         return NULL;
 870
 871                 h = instance_table[st->bucket].first;
 872         }
 873         return h;
 874 }
 875
 876 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
 877 {
 878         struct hlist_node *head;
 879         head = get_first(seq);
 880
 881         if (head)
 882                 while (pos && (head = get_next(seq, head)))
 883                         pos--;
 884         return pos ? NULL : head;
 885 }
 886
 887 static void *seq_start(struct seq_file *seq, loff_t *pos)
 888 {
 889         spin_lock(&instances_lock);
 890         return get_idx(seq, *pos);
 891 }
 892
 893 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
 894 {
 895         (*pos)++;
 896         return get_next(s, v);
 897 }
 898
 899 static void seq_stop(struct seq_file *s, void *v)
 900 {
 901         spin_unlock(&instances_lock);
 902 }
 903
 904 static int seq_show(struct seq_file *s, void *v)
 905 {
 906         const struct nfqnl_instance *inst = v;
 907
 908         return seq_printf(s, "%5d %6d %5d %1d %5d %5d %5d %8d %2d\n",
 909                           inst->queue_num,
 910                           inst->peer_pid, inst->queue_total,
 911                           inst->copy_mode, inst->copy_range,
 912                           inst->queue_dropped, inst->queue_user_dropped,
 913                           inst->id_sequence, 1);
 914 }
 915
 916 static const struct seq_operations nfqnl_seq_ops = {
 917         .start  = seq_start,
 918         .next   = seq_next,
 919         .stop   = seq_stop,
 920         .show   = seq_show,
 921 };
 922
 923 static int nfqnl_open(struct inode *inode, struct file *file)
 924 {
 925         return seq_open_private(file, &nfqnl_seq_ops,
 926                         sizeof(struct iter_state));
 927 }
 928
 929 static const struct file_operations nfqnl_file_ops = {
 930         .owner   = THIS_MODULE,
 931         .open    = nfqnl_open,
 932         .read    = seq_read,
 933         .llseek  = seq_lseek,
 934         .release = seq_release_private,
 935 };
 936
 937 #endif /* PROC_FS */
 938
 939 static int __init nfnetlink_queue_init(void)
 940 {
 941         int i, status = -ENOMEM;
 942 #ifdef CONFIG_PROC_FS
 943         struct proc_dir_entry *proc_nfqueue;
 944 #endif
 945
 946         for (i = 0; i < INSTANCE_BUCKETS; i++)
 947                 INIT_HLIST_HEAD(&instance_table[i]);
 948
 949         netlink_register_notifier(&nfqnl_rtnl_notifier);
 950         status = nfnetlink_subsys_register(&nfqnl_subsys);
 951         if (status < 0) {
 952                 printk(KERN_ERR "nf_queue: failed to create netlink socket\n");
 953                 goto cleanup_netlink_notifier;
 954         }
 955
 956 #ifdef CONFIG_PROC_FS
 957         proc_nfqueue = create_proc_entry("nfnetlink_queue", 0440,
 958                                          proc_net_netfilter);
 959         if (!proc_nfqueue)
 960                 goto cleanup_subsys;
 961         proc_nfqueue->proc_fops = &nfqnl_file_ops;
 962 #endif
 963
 964         register_netdevice_notifier(&nfqnl_dev_notifier);
 965         return status;
 966
 967 #ifdef CONFIG_PROC_FS
 968 cleanup_subsys:
 969         nfnetlink_subsys_unregister(&nfqnl_subsys);
 970 #endif
 971 cleanup_netlink_notifier:
 972         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
 973         return status;
 974 }
 975
 976 static void __exit nfnetlink_queue_fini(void)
 977 {
 978         nf_unregister_queue_handlers(&nfqh);
 979         unregister_netdevice_notifier(&nfqnl_dev_notifier);
 980 #ifdef CONFIG_PROC_FS
 981         remove_proc_entry("nfnetlink_queue", proc_net_netfilter);
 982 #endif
 983         nfnetlink_subsys_unregister(&nfqnl_subsys);
 984         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
 985 }
 986
 987 MODULE_DESCRIPTION("netfilter packet queue handler");
 988 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
 989 MODULE_LICENSE("GPL");
 990 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
 991
 992 module_init(nfnetlink_queue_init);
 993 module_exit(nfnetlink_queue_fini);