SAFE public projects git trees. - safe/jmp/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  *
   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 inline int
 180 __nfqnl_set_mode(struct nfqnl_instance *queue,
 181                  unsigned char mode, unsigned int range)
 182 {
 183         int status = 0;
 184
 185         switch (mode) {
 186         case NFQNL_COPY_NONE:
 187         case NFQNL_COPY_META:
 188                 queue->copy_mode = mode;
 189                 queue->copy_range = 0;
 190                 break;
 191
 192         case NFQNL_COPY_PACKET:
 193                 queue->copy_mode = mode;
 194                 /* we're using struct nlattr which has 16bit nla_len */
 195                 if (range > 0xffff)
 196                         queue->copy_range = 0xffff;
 197                 else
 198                         queue->copy_range = range;
 199                 break;
 200
 201         default:
 202                 status = -EINVAL;
 203
 204         }
 205         return status;
 206 }
 207
 208 static struct nf_queue_entry *
 209 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
 210 {
 211         struct nf_queue_entry *entry = NULL, *i;
 212
 213         spin_lock_bh(&queue->lock);
 214
 215         list_for_each_entry(i, &queue->queue_list, list) {
 216                 if (i->id == id) {
 217                         entry = i;
 218                         break;
 219                 }
 220         }
 221
 222         if (entry) {
 223                 list_del(&entry->list);
 224                 queue->queue_total--;
 225         }
 226
 227         spin_unlock_bh(&queue->lock);
 228
 229         return entry;
 230 }
 231
 232 static void
 233 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
 234 {
 235         struct nf_queue_entry *entry, *next;
 236
 237         spin_lock_bh(&queue->lock);
 238         list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
 239                 if (!cmpfn || cmpfn(entry, data)) {
 240                         list_del(&entry->list);
 241                         queue->queue_total--;
 242                         nf_reinject(entry, NF_DROP);
 243                 }
 244         }
 245         spin_unlock_bh(&queue->lock);
 246 }
 247
 248 static struct sk_buff *
 249 nfqnl_build_packet_message(struct nfqnl_instance *queue,
 250                            struct nf_queue_entry *entry, int *errp)
 251 {
 252         sk_buff_data_t old_tail;
 253         size_t size;
 254         size_t data_len = 0;
 255         struct sk_buff *skb;
 256         struct nfqnl_msg_packet_hdr pmsg;
 257         struct nlmsghdr *nlh;
 258         struct nfgenmsg *nfmsg;
 259         struct sk_buff *entskb = entry->skb;
 260         struct net_device *indev;
 261         struct net_device *outdev;
 262         __be32 tmp_uint;
 263
 264         QDEBUG("entered\n");
 265
 266         size =    NLMSG_ALIGN(sizeof(struct nfgenmsg))
 267                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
 268                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 269                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 270 #ifdef CONFIG_BRIDGE_NETFILTER
 271                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 272                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 273 #endif
 274                 + nla_total_size(sizeof(u_int32_t))     /* mark */
 275                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
 276                 + nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
 277
 278         outdev = entry->outdev;
 279
 280         spin_lock_bh(&queue->lock);
 281
 282         switch (queue->copy_mode) {
 283         case NFQNL_COPY_META:
 284         case NFQNL_COPY_NONE:
 285                 data_len = 0;
 286                 break;
 287
 288         case NFQNL_COPY_PACKET:
 289                 if ((entskb->ip_summed == CHECKSUM_PARTIAL ||
 290                      entskb->ip_summed == CHECKSUM_COMPLETE) &&
 291                     (*errp = skb_checksum_help(entskb))) {
 292                         spin_unlock_bh(&queue->lock);
 293                         return NULL;
 294                 }
 295                 if (queue->copy_range == 0
 296                     || queue->copy_range > entskb->len)
 297                         data_len = entskb->len;
 298                 else
 299                         data_len = queue->copy_range;
 300
 301                 size += nla_total_size(data_len);
 302                 break;
 303
 304         default:
 305                 *errp = -EINVAL;
 306                 spin_unlock_bh(&queue->lock);
 307                 return NULL;
 308         }
 309
 310         entry->id = queue->id_sequence++;
 311
 312         spin_unlock_bh(&queue->lock);
 313
 314         skb = alloc_skb(size, GFP_ATOMIC);
 315         if (!skb)
 316                 goto nlmsg_failure;
 317
 318         old_tail = skb->tail;
 319         nlh = NLMSG_PUT(skb, 0, 0,
 320                         NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
 321                         sizeof(struct nfgenmsg));
 322         nfmsg = NLMSG_DATA(nlh);
 323         nfmsg->nfgen_family = entry->pf;
 324         nfmsg->version = NFNETLINK_V0;
 325         nfmsg->res_id = htons(queue->queue_num);
 326
 327         pmsg.packet_id          = htonl(entry->id);
 328         pmsg.hw_protocol        = entskb->protocol;
 329         pmsg.hook               = entry->hook;
 330
 331         NLA_PUT(skb, NFQA_PACKET_HDR, sizeof(pmsg), &pmsg);
 332
 333         indev = entry->indev;
 334         if (indev) {
 335                 tmp_uint = htonl(indev->ifindex);
 336 #ifndef CONFIG_BRIDGE_NETFILTER
 337                 NLA_PUT(skb, NFQA_IFINDEX_INDEV, sizeof(tmp_uint), &tmp_uint);
 338 #else
 339                 if (entry->pf == PF_BRIDGE) {
 340                         /* Case 1: indev is physical input device, we need to
 341                          * look for bridge group (when called from
 342                          * netfilter_bridge) */
 343                         NLA_PUT(skb, NFQA_IFINDEX_PHYSINDEV, sizeof(tmp_uint),
 344                                 &tmp_uint);
 345                         /* this is the bridge group "brX" */
 346                         tmp_uint = htonl(indev->br_port->br->dev->ifindex);
 347                         NLA_PUT(skb, NFQA_IFINDEX_INDEV, sizeof(tmp_uint),
 348                                 &tmp_uint);
 349                 } else {
 350                         /* Case 2: indev is bridge group, we need to look for
 351                          * physical device (when called from ipv4) */
 352                         NLA_PUT(skb, NFQA_IFINDEX_INDEV, sizeof(tmp_uint),
 353                                 &tmp_uint);
 354                         if (entskb->nf_bridge
 355                             && entskb->nf_bridge->physindev) {
 356                                 tmp_uint = htonl(entskb->nf_bridge->physindev->ifindex);
 357                                 NLA_PUT(skb, NFQA_IFINDEX_PHYSINDEV,
 358                                         sizeof(tmp_uint), &tmp_uint);
 359                         }
 360                 }
 361 #endif
 362         }
 363
 364         if (outdev) {
 365                 tmp_uint = htonl(outdev->ifindex);
 366 #ifndef CONFIG_BRIDGE_NETFILTER
 367                 NLA_PUT(skb, NFQA_IFINDEX_OUTDEV, sizeof(tmp_uint), &tmp_uint);
 368 #else
 369                 if (entry->pf == PF_BRIDGE) {
 370                         /* Case 1: outdev is physical output device, we need to
 371                          * look for bridge group (when called from
 372                          * netfilter_bridge) */
 373                         NLA_PUT(skb, NFQA_IFINDEX_PHYSOUTDEV, sizeof(tmp_uint),
 374                                 &tmp_uint);
 375                         /* this is the bridge group "brX" */
 376                         tmp_uint = htonl(outdev->br_port->br->dev->ifindex);
 377                         NLA_PUT(skb, NFQA_IFINDEX_OUTDEV, sizeof(tmp_uint),
 378                                 &tmp_uint);
 379                 } else {
 380                         /* Case 2: outdev is bridge group, we need to look for
 381                          * physical output device (when called from ipv4) */
 382                         NLA_PUT(skb, NFQA_IFINDEX_OUTDEV, sizeof(tmp_uint),
 383                                 &tmp_uint);
 384                         if (entskb->nf_bridge
 385                             && entskb->nf_bridge->physoutdev) {
 386                                 tmp_uint = htonl(entskb->nf_bridge->physoutdev->ifindex);
 387                                 NLA_PUT(skb, NFQA_IFINDEX_PHYSOUTDEV,
 388                                         sizeof(tmp_uint), &tmp_uint);
 389                         }
 390                 }
 391 #endif
 392         }
 393
 394         if (entskb->mark) {
 395                 tmp_uint = htonl(entskb->mark);
 396                 NLA_PUT(skb, NFQA_MARK, sizeof(u_int32_t), &tmp_uint);
 397         }
 398
 399         if (indev && entskb->dev) {
 400                 struct nfqnl_msg_packet_hw phw;
 401                 int len = dev_parse_header(entskb, phw.hw_addr);
 402                 if (len) {
 403                         phw.hw_addrlen = htons(len);
 404                         NLA_PUT(skb, NFQA_HWADDR, sizeof(phw), &phw);
 405                 }
 406         }
 407
 408         if (entskb->tstamp.tv64) {
 409                 struct nfqnl_msg_packet_timestamp ts;
 410                 struct timeval tv = ktime_to_timeval(entskb->tstamp);
 411                 ts.sec = cpu_to_be64(tv.tv_sec);
 412                 ts.usec = cpu_to_be64(tv.tv_usec);
 413
 414                 NLA_PUT(skb, NFQA_TIMESTAMP, sizeof(ts), &ts);
 415         }
 416
 417         if (data_len) {
 418                 struct nlattr *nla;
 419                 int size = nla_attr_size(data_len);
 420
 421                 if (skb_tailroom(skb) < nla_total_size(data_len)) {
 422                         printk(KERN_WARNING "nf_queue: no tailroom!\n");
 423                         goto nlmsg_failure;
 424                 }
 425
 426                 nla = (struct nlattr *)skb_put(skb, nla_total_size(data_len));
 427                 nla->nla_type = NFQA_PAYLOAD;
 428                 nla->nla_len = size;
 429
 430                 if (skb_copy_bits(entskb, 0, nla_data(nla), data_len))
 431                         BUG();
 432         }
 433
 434         nlh->nlmsg_len = skb->tail - old_tail;
 435         return skb;
 436
 437 nlmsg_failure:
 438 nla_put_failure:
 439         if (skb)
 440                 kfree_skb(skb);
 441         *errp = -EINVAL;
 442         if (net_ratelimit())
 443                 printk(KERN_ERR "nf_queue: error creating packet message\n");
 444         return NULL;
 445 }
 446
 447 static int
 448 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 449 {
 450         int status = -EINVAL;
 451         struct sk_buff *nskb;
 452         struct nfqnl_instance *queue;
 453
 454         QDEBUG("entered\n");
 455
 456         /* rcu_read_lock()ed by nf_hook_slow() */
 457         queue = instance_lookup(queuenum);
 458         if (!queue) {
 459                 QDEBUG("no queue instance matching\n");
 460                 return -EINVAL;
 461         }
 462
 463         if (queue->copy_mode == NFQNL_COPY_NONE) {
 464                 QDEBUG("mode COPY_NONE, aborting\n");
 465                 return -EAGAIN;
 466         }
 467
 468         nskb = nfqnl_build_packet_message(queue, entry, &status);
 469         if (nskb == NULL)
 470                 return status;
 471
 472         spin_lock_bh(&queue->lock);
 473
 474         if (!queue->peer_pid)
 475                 goto err_out_free_nskb;
 476
 477         if (queue->queue_total >= queue->queue_maxlen) {
 478                 queue->queue_dropped++;
 479                 status = -ENOSPC;
 480                 if (net_ratelimit())
 481                           printk(KERN_WARNING "nf_queue: full at %d entries, "
 482                                  "dropping packets(s). Dropped: %d\n",
 483                                  queue->queue_total, queue->queue_dropped);
 484                 goto err_out_free_nskb;
 485         }
 486
 487         /* nfnetlink_unicast will either free the nskb or add it to a socket */
 488         status = nfnetlink_unicast(nskb, queue->peer_pid, MSG_DONTWAIT);
 489         if (status < 0) {
 490                 queue->queue_user_dropped++;
 491                 goto err_out_unlock;
 492         }
 493
 494         __enqueue_entry(queue, entry);
 495
 496         spin_unlock_bh(&queue->lock);
 497         return status;
 498
 499 err_out_free_nskb:
 500         kfree_skb(nskb);
 501
 502 err_out_unlock:
 503         spin_unlock_bh(&queue->lock);
 504         return status;
 505 }
 506
 507 static int
 508 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e)
 509 {
 510         int diff;
 511         int err;
 512
 513         diff = data_len - e->skb->len;
 514         if (diff < 0) {
 515                 if (pskb_trim(e->skb, data_len))
 516                         return -ENOMEM;
 517         } else if (diff > 0) {
 518                 if (data_len > 0xFFFF)
 519                         return -EINVAL;
 520                 if (diff > skb_tailroom(e->skb)) {
 521                         err = pskb_expand_head(e->skb, 0,
 522                                                diff - skb_tailroom(e->skb),
 523                                                GFP_ATOMIC);
 524                         if (err) {
 525                                 printk(KERN_WARNING "nf_queue: OOM "
 526                                       "in mangle, dropping packet\n");
 527                                 return err;
 528                         }
 529                 }
 530                 skb_put(e->skb, diff);
 531         }
 532         if (!skb_make_writable(e->skb, data_len))
 533                 return -ENOMEM;
 534         skb_copy_to_linear_data(e->skb, data, data_len);
 535         e->skb->ip_summed = CHECKSUM_NONE;
 536         return 0;
 537 }
 538
 539 static int
 540 nfqnl_set_mode(struct nfqnl_instance *queue,
 541                unsigned char mode, unsigned int range)
 542 {
 543         int status;
 544
 545         spin_lock_bh(&queue->lock);
 546         status = __nfqnl_set_mode(queue, mode, range);
 547         spin_unlock_bh(&queue->lock);
 548
 549         return status;
 550 }
 551
 552 static int
 553 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
 554 {
 555         if (entry->indev)
 556                 if (entry->indev->ifindex == ifindex)
 557                         return 1;
 558         if (entry->outdev)
 559                 if (entry->outdev->ifindex == ifindex)
 560                         return 1;
 561 #ifdef CONFIG_BRIDGE_NETFILTER
 562         if (entry->skb->nf_bridge) {
 563                 if (entry->skb->nf_bridge->physindev &&
 564                     entry->skb->nf_bridge->physindev->ifindex == ifindex)
 565                         return 1;
 566                 if (entry->skb->nf_bridge->physoutdev &&
 567                     entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
 568                         return 1;
 569         }
 570 #endif
 571         return 0;
 572 }
 573
 574 /* drop all packets with either indev or outdev == ifindex from all queue
 575  * instances */
 576 static void
 577 nfqnl_dev_drop(int ifindex)
 578 {
 579         int i;
 580
 581         QDEBUG("entering for ifindex %u\n", ifindex);
 582
 583         rcu_read_lock();
 584
 585         for (i = 0; i < INSTANCE_BUCKETS; i++) {
 586                 struct hlist_node *tmp;
 587                 struct nfqnl_instance *inst;
 588                 struct hlist_head *head = &instance_table[i];
 589
 590                 hlist_for_each_entry_rcu(inst, tmp, head, hlist)
 591                         nfqnl_flush(inst, dev_cmp, ifindex);
 592         }
 593
 594         rcu_read_unlock();
 595 }
 596
 597 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
 598
 599 static int
 600 nfqnl_rcv_dev_event(struct notifier_block *this,
 601                     unsigned long event, void *ptr)
 602 {
 603         struct net_device *dev = ptr;
 604
 605         if (dev->nd_net != &init_net)
 606                 return NOTIFY_DONE;
 607
 608         /* Drop any packets associated with the downed device */
 609         if (event == NETDEV_DOWN)
 610                 nfqnl_dev_drop(dev->ifindex);
 611         return NOTIFY_DONE;
 612 }
 613
 614 static struct notifier_block nfqnl_dev_notifier = {
 615         .notifier_call  = nfqnl_rcv_dev_event,
 616 };
 617
 618 static int
 619 nfqnl_rcv_nl_event(struct notifier_block *this,
 620                    unsigned long event, void *ptr)
 621 {
 622         struct netlink_notify *n = ptr;
 623
 624         if (event == NETLINK_URELEASE &&
 625             n->protocol == NETLINK_NETFILTER && n->pid) {
 626                 int i;
 627
 628                 /* destroy all instances for this pid */
 629                 spin_lock(&instances_lock);
 630                 for (i = 0; i < INSTANCE_BUCKETS; i++) {
 631                         struct hlist_node *tmp, *t2;
 632                         struct nfqnl_instance *inst;
 633                         struct hlist_head *head = &instance_table[i];
 634
 635                         hlist_for_each_entry_safe(inst, tmp, t2, head, hlist) {
 636                                 if ((n->net == &init_net) &&
 637                                     (n->pid == inst->peer_pid))
 638                                         __instance_destroy(inst);
 639                         }
 640                 }
 641                 spin_unlock(&instances_lock);
 642         }
 643         return NOTIFY_DONE;
 644 }
 645
 646 static struct notifier_block nfqnl_rtnl_notifier = {
 647         .notifier_call  = nfqnl_rcv_nl_event,
 648 };
 649
 650 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
 651         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 652         [NFQA_MARK]             = { .type = NLA_U32 },
 653         [NFQA_PAYLOAD]          = { .type = NLA_UNSPEC },
 654 };
 655
 656 static int
 657 nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
 658                    struct nlmsghdr *nlh, struct nlattr *nfqa[])
 659 {
 660         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 661         u_int16_t queue_num = ntohs(nfmsg->res_id);
 662
 663         struct nfqnl_msg_verdict_hdr *vhdr;
 664         struct nfqnl_instance *queue;
 665         unsigned int verdict;
 666         struct nf_queue_entry *entry;
 667         int err;
 668
 669         rcu_read_lock();
 670         queue = instance_lookup(queue_num);
 671         if (!queue) {
 672                 err = -ENODEV;
 673                 goto err_out_unlock;
 674         }
 675
 676         if (queue->peer_pid != NETLINK_CB(skb).pid) {
 677                 err = -EPERM;
 678                 goto err_out_unlock;
 679         }
 680
 681         if (!nfqa[NFQA_VERDICT_HDR]) {
 682                 err = -EINVAL;
 683                 goto err_out_unlock;
 684         }
 685
 686         vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
 687         verdict = ntohl(vhdr->verdict);
 688
 689         if ((verdict & NF_VERDICT_MASK) > NF_MAX_VERDICT) {
 690                 err = -EINVAL;
 691                 goto err_out_unlock;
 692         }
 693
 694         entry = find_dequeue_entry(queue, ntohl(vhdr->id));
 695         if (entry == NULL) {
 696                 err = -ENOENT;
 697                 goto err_out_unlock;
 698         }
 699         rcu_read_unlock();
 700
 701         if (nfqa[NFQA_PAYLOAD]) {
 702                 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
 703                                  nla_len(nfqa[NFQA_PAYLOAD]), entry) < 0)
 704                         verdict = NF_DROP;
 705         }
 706
 707         if (nfqa[NFQA_MARK])
 708                 entry->skb->mark = ntohl(*(__be32 *)
 709                                          nla_data(nfqa[NFQA_MARK]));
 710
 711         nf_reinject(entry, verdict);
 712         return 0;
 713
 714 err_out_unlock:
 715         rcu_read_unlock();
 716         return err;
 717 }
 718
 719 static int
 720 nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
 721                   struct nlmsghdr *nlh, struct nlattr *nfqa[])
 722 {
 723         return -ENOTSUPP;
 724 }
 725
 726 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
 727         [NFQA_CFG_CMD]          = { .len = sizeof(struct nfqnl_msg_config_cmd) },
 728         [NFQA_CFG_PARAMS]       = { .len = sizeof(struct nfqnl_msg_config_params) },
 729 };
 730
 731 static const struct nf_queue_handler nfqh = {
 732         .name   = "nf_queue",
 733         .outfn  = &nfqnl_enqueue_packet,
 734 };
 735
 736 static int
 737 nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
 738                   struct nlmsghdr *nlh, struct nlattr *nfqa[])
 739 {
 740         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 741         u_int16_t queue_num = ntohs(nfmsg->res_id);
 742         struct nfqnl_instance *queue;
 743         struct nfqnl_msg_config_cmd *cmd = NULL;
 744         int ret = 0;
 745
 746         QDEBUG("entering for msg %u\n", NFNL_MSG_TYPE(nlh->nlmsg_type));
 747
 748         if (nfqa[NFQA_CFG_CMD]) {
 749                 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
 750
 751                 /* Commands without queue context - might sleep */
 752                 switch (cmd->command) {
 753                 case NFQNL_CFG_CMD_PF_BIND:
 754                         ret = nf_register_queue_handler(ntohs(cmd->pf),
 755                                                         &nfqh);
 756                         break;
 757                 case NFQNL_CFG_CMD_PF_UNBIND:
 758                         ret = nf_unregister_queue_handler(ntohs(cmd->pf),
 759                                                           &nfqh);
 760                         break;
 761                 default:
 762                         break;
 763                 }
 764
 765                 if (ret < 0)
 766                         return ret;
 767         }
 768
 769         rcu_read_lock();
 770         queue = instance_lookup(queue_num);
 771         if (queue && queue->peer_pid != NETLINK_CB(skb).pid) {
 772                 ret = -EPERM;
 773                 goto err_out_unlock;
 774         }
 775
 776         if (cmd != NULL) {
 777                 switch (cmd->command) {
 778                 case NFQNL_CFG_CMD_BIND:
 779                         if (queue) {
 780                                 ret = -EBUSY;
 781                                 goto err_out_unlock;
 782                         }
 783                         queue = instance_create(queue_num, NETLINK_CB(skb).pid);
 784                         if (!queue) {
 785                                 ret = -EINVAL;
 786                                 goto err_out_unlock;
 787                         }
 788                         break;
 789                 case NFQNL_CFG_CMD_UNBIND:
 790                         if (!queue) {
 791                                 ret = -ENODEV;
 792                                 goto err_out_unlock;
 793                         }
 794                         instance_destroy(queue);
 795                         break;
 796                 case NFQNL_CFG_CMD_PF_BIND:
 797                 case NFQNL_CFG_CMD_PF_UNBIND:
 798                         break;
 799                 default:
 800                         ret = -EINVAL;
 801                         break;
 802                 }
 803         }
 804
 805         if (nfqa[NFQA_CFG_PARAMS]) {
 806                 struct nfqnl_msg_config_params *params;
 807
 808                 if (!queue) {
 809                         ret = -ENODEV;
 810                         goto err_out_unlock;
 811                 }
 812                 params = nla_data(nfqa[NFQA_CFG_PARAMS]);
 813                 nfqnl_set_mode(queue, params->copy_mode,
 814                                 ntohl(params->copy_range));
 815         }
 816
 817         if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
 818                 __be32 *queue_maxlen;
 819
 820                 if (!queue) {
 821                         ret = -ENODEV;
 822                         goto err_out_unlock;
 823                 }
 824                 queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
 825                 spin_lock_bh(&queue->lock);
 826                 queue->queue_maxlen = ntohl(*queue_maxlen);
 827                 spin_unlock_bh(&queue->lock);
 828         }
 829
 830 err_out_unlock:
 831         rcu_read_unlock();
 832         return ret;
 833 }
 834
 835 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
 836         [NFQNL_MSG_PACKET]      = { .call = nfqnl_recv_unsupp,
 837                                     .attr_count = NFQA_MAX, },
 838         [NFQNL_MSG_VERDICT]     = { .call = nfqnl_recv_verdict,
 839                                     .attr_count = NFQA_MAX,
 840                                     .policy = nfqa_verdict_policy },
 841         [NFQNL_MSG_CONFIG]      = { .call = nfqnl_recv_config,
 842                                     .attr_count = NFQA_CFG_MAX,
 843                                     .policy = nfqa_cfg_policy },
 844 };
 845
 846 static const struct nfnetlink_subsystem nfqnl_subsys = {
 847         .name           = "nf_queue",
 848         .subsys_id      = NFNL_SUBSYS_QUEUE,
 849         .cb_count       = NFQNL_MSG_MAX,
 850         .cb             = nfqnl_cb,
 851 };
 852
 853 #ifdef CONFIG_PROC_FS
 854 struct iter_state {
 855         unsigned int bucket;
 856 };
 857
 858 static struct hlist_node *get_first(struct seq_file *seq)
 859 {
 860         struct iter_state *st = seq->private;
 861
 862         if (!st)
 863                 return NULL;
 864
 865         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
 866                 if (!hlist_empty(&instance_table[st->bucket]))
 867                         return instance_table[st->bucket].first;
 868         }
 869         return NULL;
 870 }
 871
 872 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
 873 {
 874         struct iter_state *st = seq->private;
 875
 876         h = h->next;
 877         while (!h) {
 878                 if (++st->bucket >= INSTANCE_BUCKETS)
 879                         return NULL;
 880
 881                 h = instance_table[st->bucket].first;
 882         }
 883         return h;
 884 }
 885
 886 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
 887 {
 888         struct hlist_node *head;
 889         head = get_first(seq);
 890
 891         if (head)
 892                 while (pos && (head = get_next(seq, head)))
 893                         pos--;
 894         return pos ? NULL : head;
 895 }
 896
 897 static void *seq_start(struct seq_file *seq, loff_t *pos)
 898 {
 899         spin_lock(&instances_lock);
 900         return get_idx(seq, *pos);
 901 }
 902
 903 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
 904 {
 905         (*pos)++;
 906         return get_next(s, v);
 907 }
 908
 909 static void seq_stop(struct seq_file *s, void *v)
 910 {
 911         spin_unlock(&instances_lock);
 912 }
 913
 914 static int seq_show(struct seq_file *s, void *v)
 915 {
 916         const struct nfqnl_instance *inst = v;
 917
 918         return seq_printf(s, "%5d %6d %5d %1d %5d %5d %5d %8d %2d\n",
 919                           inst->queue_num,
 920                           inst->peer_pid, inst->queue_total,
 921                           inst->copy_mode, inst->copy_range,
 922                           inst->queue_dropped, inst->queue_user_dropped,
 923                           inst->id_sequence, 1);
 924 }
 925
 926 static const struct seq_operations nfqnl_seq_ops = {
 927         .start  = seq_start,
 928         .next   = seq_next,
 929         .stop   = seq_stop,
 930         .show   = seq_show,
 931 };
 932
 933 static int nfqnl_open(struct inode *inode, struct file *file)
 934 {
 935         return seq_open_private(file, &nfqnl_seq_ops,
 936                         sizeof(struct iter_state));
 937 }
 938
 939 static const struct file_operations nfqnl_file_ops = {
 940         .owner   = THIS_MODULE,
 941         .open    = nfqnl_open,
 942         .read    = seq_read,
 943         .llseek  = seq_lseek,
 944         .release = seq_release_private,
 945 };
 946
 947 #endif /* PROC_FS */
 948
 949 static int __init nfnetlink_queue_init(void)
 950 {
 951         int i, status = -ENOMEM;
 952 #ifdef CONFIG_PROC_FS
 953         struct proc_dir_entry *proc_nfqueue;
 954 #endif
 955
 956         for (i = 0; i < INSTANCE_BUCKETS; i++)
 957                 INIT_HLIST_HEAD(&instance_table[i]);
 958
 959         netlink_register_notifier(&nfqnl_rtnl_notifier);
 960         status = nfnetlink_subsys_register(&nfqnl_subsys);
 961         if (status < 0) {
 962                 printk(KERN_ERR "nf_queue: failed to create netlink socket\n");
 963                 goto cleanup_netlink_notifier;
 964         }
 965
 966 #ifdef CONFIG_PROC_FS
 967         proc_nfqueue = create_proc_entry("nfnetlink_queue", 0440,
 968                                          proc_net_netfilter);
 969         if (!proc_nfqueue)
 970                 goto cleanup_subsys;
 971         proc_nfqueue->proc_fops = &nfqnl_file_ops;
 972 #endif
 973
 974         register_netdevice_notifier(&nfqnl_dev_notifier);
 975         return status;
 976
 977 #ifdef CONFIG_PROC_FS
 978 cleanup_subsys:
 979         nfnetlink_subsys_unregister(&nfqnl_subsys);
 980 #endif
 981 cleanup_netlink_notifier:
 982         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
 983         return status;
 984 }
 985
 986 static void __exit nfnetlink_queue_fini(void)
 987 {
 988         nf_unregister_queue_handlers(&nfqh);
 989         unregister_netdevice_notifier(&nfqnl_dev_notifier);
 990 #ifdef CONFIG_PROC_FS
 991         remove_proc_entry("nfnetlink_queue", proc_net_netfilter);
 992 #endif
 993         nfnetlink_subsys_unregister(&nfqnl_subsys);
 994         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
 995 }
 996
 997 MODULE_DESCRIPTION("netfilter packet queue handler");
 998 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
 999 MODULE_LICENSE("GPL");
1000 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1001
1002 module_init(nfnetlink_queue_init);
1003 module_exit(nfnetlink_queue_fini);