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         atomic_t use;
  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_RWLOCK(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(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_lookup_get(u_int16_t queue_num)
  98 {
  99         struct nfqnl_instance *inst;
 100
 101         read_lock_bh(&instances_lock);
 102         inst = __instance_lookup(queue_num);
 103         if (inst)
 104                 atomic_inc(&inst->use);
 105         read_unlock_bh(&instances_lock);
 106
 107         return inst;
 108 }
 109
 110 static void
 111 instance_put(struct nfqnl_instance *inst)
 112 {
 113         if (inst && atomic_dec_and_test(&inst->use)) {
 114                 QDEBUG("kfree(inst=%p)\n", inst);
 115                 kfree(inst);
 116         }
 117 }
 118
 119 static struct nfqnl_instance *
 120 instance_create(u_int16_t queue_num, int pid)
 121 {
 122         struct nfqnl_instance *inst;
 123
 124         QDEBUG("entering for queue_num=%u, pid=%d\n", queue_num, pid);
 125
 126         write_lock_bh(&instances_lock);
 127         if (__instance_lookup(queue_num)) {
 128                 inst = NULL;
 129                 QDEBUG("aborting, instance already exists\n");
 130                 goto out_unlock;
 131         }
 132
 133         inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
 134         if (!inst)
 135                 goto out_unlock;
 136
 137         inst->queue_num = queue_num;
 138         inst->peer_pid = pid;
 139         inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
 140         inst->copy_range = 0xfffff;
 141         inst->copy_mode = NFQNL_COPY_NONE;
 142         /* needs to be two, since we _put() after creation */
 143         atomic_set(&inst->use, 2);
 144         spin_lock_init(&inst->lock);
 145         INIT_LIST_HEAD(&inst->queue_list);
 146
 147         if (!try_module_get(THIS_MODULE))
 148                 goto out_free;
 149
 150         hlist_add_head(&inst->hlist,
 151                        &instance_table[instance_hashfn(queue_num)]);
 152
 153         write_unlock_bh(&instances_lock);
 154
 155         QDEBUG("successfully created new instance\n");
 156
 157         return inst;
 158
 159 out_free:
 160         kfree(inst);
 161 out_unlock:
 162         write_unlock_bh(&instances_lock);
 163         return NULL;
 164 }
 165
 166 static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
 167                         unsigned long data);
 168
 169 static void
 170 _instance_destroy2(struct nfqnl_instance *inst, int lock)
 171 {
 172         /* first pull it out of the global list */
 173         if (lock)
 174                 write_lock_bh(&instances_lock);
 175
 176         QDEBUG("removing instance %p (queuenum=%u) from hash\n",
 177                 inst, inst->queue_num);
 178         hlist_del(&inst->hlist);
 179
 180         if (lock)
 181                 write_unlock_bh(&instances_lock);
 182
 183         /* then flush all pending skbs from the queue */
 184         nfqnl_flush(inst, NULL, 0);
 185
 186         /* and finally put the refcount */
 187         instance_put(inst);
 188
 189         module_put(THIS_MODULE);
 190 }
 191
 192 static inline void
 193 __instance_destroy(struct nfqnl_instance *inst)
 194 {
 195         _instance_destroy2(inst, 0);
 196 }
 197
 198 static inline void
 199 instance_destroy(struct nfqnl_instance *inst)
 200 {
 201         _instance_destroy2(inst, 1);
 202 }
 203
 204 static inline void
 205 __enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
 206 {
 207        list_add_tail(&entry->list, &queue->queue_list);
 208        queue->queue_total++;
 209 }
 210
 211 static inline int
 212 __nfqnl_set_mode(struct nfqnl_instance *queue,
 213                  unsigned char mode, unsigned int range)
 214 {
 215         int status = 0;
 216
 217         switch (mode) {
 218         case NFQNL_COPY_NONE:
 219         case NFQNL_COPY_META:
 220                 queue->copy_mode = mode;
 221                 queue->copy_range = 0;
 222                 break;
 223
 224         case NFQNL_COPY_PACKET:
 225                 queue->copy_mode = mode;
 226                 /* we're using struct nlattr which has 16bit nla_len */
 227                 if (range > 0xffff)
 228                         queue->copy_range = 0xffff;
 229                 else
 230                         queue->copy_range = range;
 231                 break;
 232
 233         default:
 234                 status = -EINVAL;
 235
 236         }
 237         return status;
 238 }
 239
 240 static struct nf_queue_entry *
 241 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
 242 {
 243         struct nf_queue_entry *entry = NULL, *i;
 244
 245         spin_lock_bh(&queue->lock);
 246
 247         list_for_each_entry(i, &queue->queue_list, list) {
 248                 if (i->id == id) {
 249                         entry = i;
 250                         break;
 251                 }
 252         }
 253
 254         if (entry) {
 255                 list_del(&entry->list);
 256                 queue->queue_total--;
 257         }
 258
 259         spin_unlock_bh(&queue->lock);
 260
 261         return entry;
 262 }
 263
 264 static void
 265 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
 266 {
 267         struct nf_queue_entry *entry, *next;
 268
 269         spin_lock_bh(&queue->lock);
 270         list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
 271                 if (!cmpfn || cmpfn(entry, data)) {
 272                         list_del(&entry->list);
 273                         queue->queue_total--;
 274                         nf_reinject(entry, NF_DROP);
 275                 }
 276         }
 277         spin_unlock_bh(&queue->lock);
 278 }
 279
 280 static struct sk_buff *
 281 nfqnl_build_packet_message(struct nfqnl_instance *queue,
 282                            struct nf_queue_entry *entry, int *errp)
 283 {
 284         sk_buff_data_t old_tail;
 285         size_t size;
 286         size_t data_len = 0;
 287         struct sk_buff *skb;
 288         struct nfqnl_msg_packet_hdr pmsg;
 289         struct nlmsghdr *nlh;
 290         struct nfgenmsg *nfmsg;
 291         struct sk_buff *entskb = entry->skb;
 292         struct net_device *indev;
 293         struct net_device *outdev;
 294         __be32 tmp_uint;
 295
 296         QDEBUG("entered\n");
 297
 298         size =    NLMSG_ALIGN(sizeof(struct nfgenmsg))
 299                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
 300                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 301                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 302 #ifdef CONFIG_BRIDGE_NETFILTER
 303                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 304                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 305 #endif
 306                 + nla_total_size(sizeof(u_int32_t))     /* mark */
 307                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
 308                 + nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
 309
 310         outdev = entry->outdev;
 311
 312         spin_lock_bh(&queue->lock);
 313
 314         switch (queue->copy_mode) {
 315         case NFQNL_COPY_META:
 316         case NFQNL_COPY_NONE:
 317                 data_len = 0;
 318                 break;
 319
 320         case NFQNL_COPY_PACKET:
 321                 if ((entskb->ip_summed == CHECKSUM_PARTIAL ||
 322                      entskb->ip_summed == CHECKSUM_COMPLETE) &&
 323                     (*errp = skb_checksum_help(entskb))) {
 324                         spin_unlock_bh(&queue->lock);
 325                         return NULL;
 326                 }
 327                 if (queue->copy_range == 0
 328                     || queue->copy_range > entskb->len)
 329                         data_len = entskb->len;
 330                 else
 331                         data_len = queue->copy_range;
 332
 333                 size += nla_total_size(data_len);
 334                 break;
 335
 336         default:
 337                 *errp = -EINVAL;
 338                 spin_unlock_bh(&queue->lock);
 339                 return NULL;
 340         }
 341
 342         entry->id = queue->id_sequence++;
 343
 344         spin_unlock_bh(&queue->lock);
 345
 346         skb = alloc_skb(size, GFP_ATOMIC);
 347         if (!skb)
 348                 goto nlmsg_failure;
 349
 350         old_tail = skb->tail;
 351         nlh = NLMSG_PUT(skb, 0, 0,
 352                         NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
 353                         sizeof(struct nfgenmsg));
 354         nfmsg = NLMSG_DATA(nlh);
 355         nfmsg->nfgen_family = entry->pf;
 356         nfmsg->version = NFNETLINK_V0;
 357         nfmsg->res_id = htons(queue->queue_num);
 358
 359         pmsg.packet_id          = htonl(entry->id);
 360         pmsg.hw_protocol        = entskb->protocol;
 361         pmsg.hook               = entry->hook;
 362
 363         NLA_PUT(skb, NFQA_PACKET_HDR, sizeof(pmsg), &pmsg);
 364
 365         indev = entry->indev;
 366         if (indev) {
 367                 tmp_uint = htonl(indev->ifindex);
 368 #ifndef CONFIG_BRIDGE_NETFILTER
 369                 NLA_PUT(skb, NFQA_IFINDEX_INDEV, sizeof(tmp_uint), &tmp_uint);
 370 #else
 371                 if (entry->pf == PF_BRIDGE) {
 372                         /* Case 1: indev is physical input device, we need to
 373                          * look for bridge group (when called from
 374                          * netfilter_bridge) */
 375                         NLA_PUT(skb, NFQA_IFINDEX_PHYSINDEV, sizeof(tmp_uint),
 376                                 &tmp_uint);
 377                         /* this is the bridge group "brX" */
 378                         tmp_uint = htonl(indev->br_port->br->dev->ifindex);
 379                         NLA_PUT(skb, NFQA_IFINDEX_INDEV, sizeof(tmp_uint),
 380                                 &tmp_uint);
 381                 } else {
 382                         /* Case 2: indev is bridge group, we need to look for
 383                          * physical device (when called from ipv4) */
 384                         NLA_PUT(skb, NFQA_IFINDEX_INDEV, sizeof(tmp_uint),
 385                                 &tmp_uint);
 386                         if (entskb->nf_bridge
 387                             && entskb->nf_bridge->physindev) {
 388                                 tmp_uint = htonl(entskb->nf_bridge->physindev->ifindex);
 389                                 NLA_PUT(skb, NFQA_IFINDEX_PHYSINDEV,
 390                                         sizeof(tmp_uint), &tmp_uint);
 391                         }
 392                 }
 393 #endif
 394         }
 395
 396         if (outdev) {
 397                 tmp_uint = htonl(outdev->ifindex);
 398 #ifndef CONFIG_BRIDGE_NETFILTER
 399                 NLA_PUT(skb, NFQA_IFINDEX_OUTDEV, sizeof(tmp_uint), &tmp_uint);
 400 #else
 401                 if (entry->pf == PF_BRIDGE) {
 402                         /* Case 1: outdev is physical output device, we need to
 403                          * look for bridge group (when called from
 404                          * netfilter_bridge) */
 405                         NLA_PUT(skb, NFQA_IFINDEX_PHYSOUTDEV, sizeof(tmp_uint),
 406                                 &tmp_uint);
 407                         /* this is the bridge group "brX" */
 408                         tmp_uint = htonl(outdev->br_port->br->dev->ifindex);
 409                         NLA_PUT(skb, NFQA_IFINDEX_OUTDEV, sizeof(tmp_uint),
 410                                 &tmp_uint);
 411                 } else {
 412                         /* Case 2: outdev is bridge group, we need to look for
 413                          * physical output device (when called from ipv4) */
 414                         NLA_PUT(skb, NFQA_IFINDEX_OUTDEV, sizeof(tmp_uint),
 415                                 &tmp_uint);
 416                         if (entskb->nf_bridge
 417                             && entskb->nf_bridge->physoutdev) {
 418                                 tmp_uint = htonl(entskb->nf_bridge->physoutdev->ifindex);
 419                                 NLA_PUT(skb, NFQA_IFINDEX_PHYSOUTDEV,
 420                                         sizeof(tmp_uint), &tmp_uint);
 421                         }
 422                 }
 423 #endif
 424         }
 425
 426         if (entskb->mark) {
 427                 tmp_uint = htonl(entskb->mark);
 428                 NLA_PUT(skb, NFQA_MARK, sizeof(u_int32_t), &tmp_uint);
 429         }
 430
 431         if (indev && entskb->dev) {
 432                 struct nfqnl_msg_packet_hw phw;
 433                 int len = dev_parse_header(entskb, phw.hw_addr);
 434                 if (len) {
 435                         phw.hw_addrlen = htons(len);
 436                         NLA_PUT(skb, NFQA_HWADDR, sizeof(phw), &phw);
 437                 }
 438         }
 439
 440         if (entskb->tstamp.tv64) {
 441                 struct nfqnl_msg_packet_timestamp ts;
 442                 struct timeval tv = ktime_to_timeval(entskb->tstamp);
 443                 ts.sec = cpu_to_be64(tv.tv_sec);
 444                 ts.usec = cpu_to_be64(tv.tv_usec);
 445
 446                 NLA_PUT(skb, NFQA_TIMESTAMP, sizeof(ts), &ts);
 447         }
 448
 449         if (data_len) {
 450                 struct nlattr *nla;
 451                 int size = nla_attr_size(data_len);
 452
 453                 if (skb_tailroom(skb) < nla_total_size(data_len)) {
 454                         printk(KERN_WARNING "nf_queue: no tailroom!\n");
 455                         goto nlmsg_failure;
 456                 }
 457
 458                 nla = (struct nlattr *)skb_put(skb, nla_total_size(data_len));
 459                 nla->nla_type = NFQA_PAYLOAD;
 460                 nla->nla_len = size;
 461
 462                 if (skb_copy_bits(entskb, 0, nla_data(nla), data_len))
 463                         BUG();
 464         }
 465
 466         nlh->nlmsg_len = skb->tail - old_tail;
 467         return skb;
 468
 469 nlmsg_failure:
 470 nla_put_failure:
 471         if (skb)
 472                 kfree_skb(skb);
 473         *errp = -EINVAL;
 474         if (net_ratelimit())
 475                 printk(KERN_ERR "nf_queue: error creating packet message\n");
 476         return NULL;
 477 }
 478
 479 static int
 480 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 481 {
 482         int status = -EINVAL;
 483         struct sk_buff *nskb;
 484         struct nfqnl_instance *queue;
 485
 486         QDEBUG("entered\n");
 487
 488         queue = instance_lookup_get(queuenum);
 489         if (!queue) {
 490                 QDEBUG("no queue instance matching\n");
 491                 return -EINVAL;
 492         }
 493
 494         if (queue->copy_mode == NFQNL_COPY_NONE) {
 495                 QDEBUG("mode COPY_NONE, aborting\n");
 496                 status = -EAGAIN;
 497                 goto err_out_put;
 498         }
 499
 500         nskb = nfqnl_build_packet_message(queue, entry, &status);
 501         if (nskb == NULL)
 502                 goto err_out_put;
 503
 504         spin_lock_bh(&queue->lock);
 505
 506         if (!queue->peer_pid)
 507                 goto err_out_free_nskb;
 508
 509         if (queue->queue_total >= queue->queue_maxlen) {
 510                 queue->queue_dropped++;
 511                 status = -ENOSPC;
 512                 if (net_ratelimit())
 513                           printk(KERN_WARNING "nf_queue: full at %d entries, "
 514                                  "dropping packets(s). Dropped: %d\n",
 515                                  queue->queue_total, queue->queue_dropped);
 516                 goto err_out_free_nskb;
 517         }
 518
 519         /* nfnetlink_unicast will either free the nskb or add it to a socket */
 520         status = nfnetlink_unicast(nskb, queue->peer_pid, MSG_DONTWAIT);
 521         if (status < 0) {
 522                 queue->queue_user_dropped++;
 523                 goto err_out_unlock;
 524         }
 525
 526         __enqueue_entry(queue, entry);
 527
 528         spin_unlock_bh(&queue->lock);
 529         instance_put(queue);
 530         return status;
 531
 532 err_out_free_nskb:
 533         kfree_skb(nskb);
 534
 535 err_out_unlock:
 536         spin_unlock_bh(&queue->lock);
 537
 538 err_out_put:
 539         instance_put(queue);
 540         return status;
 541 }
 542
 543 static int
 544 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e)
 545 {
 546         int diff;
 547         int err;
 548
 549         diff = data_len - e->skb->len;
 550         if (diff < 0) {
 551                 if (pskb_trim(e->skb, data_len))
 552                         return -ENOMEM;
 553         } else if (diff > 0) {
 554                 if (data_len > 0xFFFF)
 555                         return -EINVAL;
 556                 if (diff > skb_tailroom(e->skb)) {
 557                         err = pskb_expand_head(e->skb, 0,
 558                                                diff - skb_tailroom(e->skb),
 559                                                GFP_ATOMIC);
 560                         if (err) {
 561                                 printk(KERN_WARNING "nf_queue: OOM "
 562                                       "in mangle, dropping packet\n");
 563                                 return err;
 564                         }
 565                 }
 566                 skb_put(e->skb, diff);
 567         }
 568         if (!skb_make_writable(e->skb, data_len))
 569                 return -ENOMEM;
 570         skb_copy_to_linear_data(e->skb, data, data_len);
 571         e->skb->ip_summed = CHECKSUM_NONE;
 572         return 0;
 573 }
 574
 575 static int
 576 nfqnl_set_mode(struct nfqnl_instance *queue,
 577                unsigned char mode, unsigned int range)
 578 {
 579         int status;
 580
 581         spin_lock_bh(&queue->lock);
 582         status = __nfqnl_set_mode(queue, mode, range);
 583         spin_unlock_bh(&queue->lock);
 584
 585         return status;
 586 }
 587
 588 static int
 589 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
 590 {
 591         if (entry->indev)
 592                 if (entry->indev->ifindex == ifindex)
 593                         return 1;
 594         if (entry->outdev)
 595                 if (entry->outdev->ifindex == ifindex)
 596                         return 1;
 597 #ifdef CONFIG_BRIDGE_NETFILTER
 598         if (entry->skb->nf_bridge) {
 599                 if (entry->skb->nf_bridge->physindev &&
 600                     entry->skb->nf_bridge->physindev->ifindex == ifindex)
 601                         return 1;
 602                 if (entry->skb->nf_bridge->physoutdev &&
 603                     entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
 604                         return 1;
 605         }
 606 #endif
 607         return 0;
 608 }
 609
 610 /* drop all packets with either indev or outdev == ifindex from all queue
 611  * instances */
 612 static void
 613 nfqnl_dev_drop(int ifindex)
 614 {
 615         int i;
 616
 617         QDEBUG("entering for ifindex %u\n", ifindex);
 618
 619         /* this only looks like we have to hold the readlock for a way too long
 620          * time, issue_verdict(),  nf_reinject(), ... - but we always only
 621          * issue NF_DROP, which is processed directly in nf_reinject() */
 622         read_lock_bh(&instances_lock);
 623
 624         for  (i = 0; i < INSTANCE_BUCKETS; i++) {
 625                 struct hlist_node *tmp;
 626                 struct nfqnl_instance *inst;
 627                 struct hlist_head *head = &instance_table[i];
 628
 629                 hlist_for_each_entry(inst, tmp, head, hlist)
 630                         nfqnl_flush(inst, dev_cmp, ifindex);
 631         }
 632
 633         read_unlock_bh(&instances_lock);
 634 }
 635
 636 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
 637
 638 static int
 639 nfqnl_rcv_dev_event(struct notifier_block *this,
 640                     unsigned long event, void *ptr)
 641 {
 642         struct net_device *dev = ptr;
 643
 644         if (dev->nd_net != &init_net)
 645                 return NOTIFY_DONE;
 646
 647         /* Drop any packets associated with the downed device */
 648         if (event == NETDEV_DOWN)
 649                 nfqnl_dev_drop(dev->ifindex);
 650         return NOTIFY_DONE;
 651 }
 652
 653 static struct notifier_block nfqnl_dev_notifier = {
 654         .notifier_call  = nfqnl_rcv_dev_event,
 655 };
 656
 657 static int
 658 nfqnl_rcv_nl_event(struct notifier_block *this,
 659                    unsigned long event, void *ptr)
 660 {
 661         struct netlink_notify *n = ptr;
 662
 663         if (event == NETLINK_URELEASE &&
 664             n->protocol == NETLINK_NETFILTER && n->pid) {
 665                 int i;
 666
 667                 /* destroy all instances for this pid */
 668                 write_lock_bh(&instances_lock);
 669                 for  (i = 0; i < INSTANCE_BUCKETS; i++) {
 670                         struct hlist_node *tmp, *t2;
 671                         struct nfqnl_instance *inst;
 672                         struct hlist_head *head = &instance_table[i];
 673
 674                         hlist_for_each_entry_safe(inst, tmp, t2, head, hlist) {
 675                                 if ((n->net == &init_net) &&
 676                                     (n->pid == inst->peer_pid))
 677                                         __instance_destroy(inst);
 678                         }
 679                 }
 680                 write_unlock_bh(&instances_lock);
 681         }
 682         return NOTIFY_DONE;
 683 }
 684
 685 static struct notifier_block nfqnl_rtnl_notifier = {
 686         .notifier_call  = nfqnl_rcv_nl_event,
 687 };
 688
 689 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
 690         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 691         [NFQA_MARK]             = { .type = NLA_U32 },
 692         [NFQA_PAYLOAD]          = { .type = NLA_UNSPEC },
 693 };
 694
 695 static int
 696 nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
 697                    struct nlmsghdr *nlh, struct nlattr *nfqa[])
 698 {
 699         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 700         u_int16_t queue_num = ntohs(nfmsg->res_id);
 701
 702         struct nfqnl_msg_verdict_hdr *vhdr;
 703         struct nfqnl_instance *queue;
 704         unsigned int verdict;
 705         struct nf_queue_entry *entry;
 706         int err;
 707
 708         queue = instance_lookup_get(queue_num);
 709         if (!queue)
 710                 return -ENODEV;
 711
 712         if (queue->peer_pid != NETLINK_CB(skb).pid) {
 713                 err = -EPERM;
 714                 goto err_out_put;
 715         }
 716
 717         if (!nfqa[NFQA_VERDICT_HDR]) {
 718                 err = -EINVAL;
 719                 goto err_out_put;
 720         }
 721
 722         vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
 723         verdict = ntohl(vhdr->verdict);
 724
 725         if ((verdict & NF_VERDICT_MASK) > NF_MAX_VERDICT) {
 726                 err = -EINVAL;
 727                 goto err_out_put;
 728         }
 729
 730         entry = find_dequeue_entry(queue, ntohl(vhdr->id));
 731         if (entry == NULL) {
 732                 err = -ENOENT;
 733                 goto err_out_put;
 734         }
 735
 736         if (nfqa[NFQA_PAYLOAD]) {
 737                 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
 738                                  nla_len(nfqa[NFQA_PAYLOAD]), entry) < 0)
 739                         verdict = NF_DROP;
 740         }
 741
 742         if (nfqa[NFQA_MARK])
 743                 entry->skb->mark = ntohl(*(__be32 *)
 744                                          nla_data(nfqa[NFQA_MARK]));
 745
 746         nf_reinject(entry, verdict);
 747         instance_put(queue);
 748         return 0;
 749
 750 err_out_put:
 751         instance_put(queue);
 752         return err;
 753 }
 754
 755 static int
 756 nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
 757                   struct nlmsghdr *nlh, struct nlattr *nfqa[])
 758 {
 759         return -ENOTSUPP;
 760 }
 761
 762 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
 763         [NFQA_CFG_CMD]          = { .len = sizeof(struct nfqnl_msg_config_cmd) },
 764         [NFQA_CFG_PARAMS]       = { .len = sizeof(struct nfqnl_msg_config_params) },
 765 };
 766
 767 static const struct nf_queue_handler nfqh = {
 768         .name   = "nf_queue",
 769         .outfn  = &nfqnl_enqueue_packet,
 770 };
 771
 772 static int
 773 nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
 774                   struct nlmsghdr *nlh, struct nlattr *nfqa[])
 775 {
 776         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 777         u_int16_t queue_num = ntohs(nfmsg->res_id);
 778         struct nfqnl_instance *queue;
 779         int ret = 0;
 780
 781         QDEBUG("entering for msg %u\n", NFNL_MSG_TYPE(nlh->nlmsg_type));
 782
 783         queue = instance_lookup_get(queue_num);
 784         if (queue && queue->peer_pid != NETLINK_CB(skb).pid) {
 785                 ret = -EPERM;
 786                 goto out_put;
 787         }
 788
 789         if (nfqa[NFQA_CFG_CMD]) {
 790                 struct nfqnl_msg_config_cmd *cmd;
 791
 792                 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
 793                 QDEBUG("found CFG_CMD\n");
 794
 795                 switch (cmd->command) {
 796                 case NFQNL_CFG_CMD_BIND:
 797                         if (queue)
 798                                 return -EBUSY;
 799
 800                         queue = instance_create(queue_num, NETLINK_CB(skb).pid);
 801                         if (!queue)
 802                                 return -EINVAL;
 803                         break;
 804                 case NFQNL_CFG_CMD_UNBIND:
 805                         if (!queue)
 806                                 return -ENODEV;
 807                         instance_destroy(queue);
 808                         break;
 809                 case NFQNL_CFG_CMD_PF_BIND:
 810                         QDEBUG("registering queue handler for pf=%u\n",
 811                                 ntohs(cmd->pf));
 812                         ret = nf_register_queue_handler(ntohs(cmd->pf), &nfqh);
 813                         break;
 814                 case NFQNL_CFG_CMD_PF_UNBIND:
 815                         QDEBUG("unregistering queue handler for pf=%u\n",
 816                                 ntohs(cmd->pf));
 817                         ret = nf_unregister_queue_handler(ntohs(cmd->pf), &nfqh);
 818                         break;
 819                 default:
 820                         ret = -EINVAL;
 821                         break;
 822                 }
 823         }
 824
 825         if (nfqa[NFQA_CFG_PARAMS]) {
 826                 struct nfqnl_msg_config_params *params;
 827
 828                 if (!queue) {
 829                         ret = -ENODEV;
 830                         goto out_put;
 831                 }
 832                 params = nla_data(nfqa[NFQA_CFG_PARAMS]);
 833                 nfqnl_set_mode(queue, params->copy_mode,
 834                                 ntohl(params->copy_range));
 835         }
 836
 837         if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
 838                 __be32 *queue_maxlen;
 839
 840                 if (!queue) {
 841                         ret = -ENODEV;
 842                         goto out_put;
 843                 }
 844                 queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
 845                 spin_lock_bh(&queue->lock);
 846                 queue->queue_maxlen = ntohl(*queue_maxlen);
 847                 spin_unlock_bh(&queue->lock);
 848         }
 849
 850 out_put:
 851         instance_put(queue);
 852         return ret;
 853 }
 854
 855 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
 856         [NFQNL_MSG_PACKET]      = { .call = nfqnl_recv_unsupp,
 857                                     .attr_count = NFQA_MAX, },
 858         [NFQNL_MSG_VERDICT]     = { .call = nfqnl_recv_verdict,
 859                                     .attr_count = NFQA_MAX,
 860                                     .policy = nfqa_verdict_policy },
 861         [NFQNL_MSG_CONFIG]      = { .call = nfqnl_recv_config,
 862                                     .attr_count = NFQA_CFG_MAX,
 863                                     .policy = nfqa_cfg_policy },
 864 };
 865
 866 static const struct nfnetlink_subsystem nfqnl_subsys = {
 867         .name           = "nf_queue",
 868         .subsys_id      = NFNL_SUBSYS_QUEUE,
 869         .cb_count       = NFQNL_MSG_MAX,
 870         .cb             = nfqnl_cb,
 871 };
 872
 873 #ifdef CONFIG_PROC_FS
 874 struct iter_state {
 875         unsigned int bucket;
 876 };
 877
 878 static struct hlist_node *get_first(struct seq_file *seq)
 879 {
 880         struct iter_state *st = seq->private;
 881
 882         if (!st)
 883                 return NULL;
 884
 885         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
 886                 if (!hlist_empty(&instance_table[st->bucket]))
 887                         return instance_table[st->bucket].first;
 888         }
 889         return NULL;
 890 }
 891
 892 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
 893 {
 894         struct iter_state *st = seq->private;
 895
 896         h = h->next;
 897         while (!h) {
 898                 if (++st->bucket >= INSTANCE_BUCKETS)
 899                         return NULL;
 900
 901                 h = instance_table[st->bucket].first;
 902         }
 903         return h;
 904 }
 905
 906 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
 907 {
 908         struct hlist_node *head;
 909         head = get_first(seq);
 910
 911         if (head)
 912                 while (pos && (head = get_next(seq, head)))
 913                         pos--;
 914         return pos ? NULL : head;
 915 }
 916
 917 static void *seq_start(struct seq_file *seq, loff_t *pos)
 918 {
 919         read_lock_bh(&instances_lock);
 920         return get_idx(seq, *pos);
 921 }
 922
 923 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
 924 {
 925         (*pos)++;
 926         return get_next(s, v);
 927 }
 928
 929 static void seq_stop(struct seq_file *s, void *v)
 930 {
 931         read_unlock_bh(&instances_lock);
 932 }
 933
 934 static int seq_show(struct seq_file *s, void *v)
 935 {
 936         const struct nfqnl_instance *inst = v;
 937
 938         return seq_printf(s, "%5d %6d %5d %1d %5d %5d %5d %8d %2d\n",
 939                           inst->queue_num,
 940                           inst->peer_pid, inst->queue_total,
 941                           inst->copy_mode, inst->copy_range,
 942                           inst->queue_dropped, inst->queue_user_dropped,
 943                           inst->id_sequence,
 944                           atomic_read(&inst->use));
 945 }
 946
 947 static const struct seq_operations nfqnl_seq_ops = {
 948         .start  = seq_start,
 949         .next   = seq_next,
 950         .stop   = seq_stop,
 951         .show   = seq_show,
 952 };
 953
 954 static int nfqnl_open(struct inode *inode, struct file *file)
 955 {
 956         return seq_open_private(file, &nfqnl_seq_ops,
 957                         sizeof(struct iter_state));
 958 }
 959
 960 static const struct file_operations nfqnl_file_ops = {
 961         .owner   = THIS_MODULE,
 962         .open    = nfqnl_open,
 963         .read    = seq_read,
 964         .llseek  = seq_lseek,
 965         .release = seq_release_private,
 966 };
 967
 968 #endif /* PROC_FS */
 969
 970 static int __init nfnetlink_queue_init(void)
 971 {
 972         int i, status = -ENOMEM;
 973 #ifdef CONFIG_PROC_FS
 974         struct proc_dir_entry *proc_nfqueue;
 975 #endif
 976
 977         for (i = 0; i < INSTANCE_BUCKETS; i++)
 978                 INIT_HLIST_HEAD(&instance_table[i]);
 979
 980         netlink_register_notifier(&nfqnl_rtnl_notifier);
 981         status = nfnetlink_subsys_register(&nfqnl_subsys);
 982         if (status < 0) {
 983                 printk(KERN_ERR "nf_queue: failed to create netlink socket\n");
 984                 goto cleanup_netlink_notifier;
 985         }
 986
 987 #ifdef CONFIG_PROC_FS
 988         proc_nfqueue = create_proc_entry("nfnetlink_queue", 0440,
 989                                          proc_net_netfilter);
 990         if (!proc_nfqueue)
 991                 goto cleanup_subsys;
 992         proc_nfqueue->proc_fops = &nfqnl_file_ops;
 993 #endif
 994
 995         register_netdevice_notifier(&nfqnl_dev_notifier);
 996         return status;
 997
 998 #ifdef CONFIG_PROC_FS
 999 cleanup_subsys:
1000         nfnetlink_subsys_unregister(&nfqnl_subsys);
1001 #endif
1002 cleanup_netlink_notifier:
1003         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1004         return status;
1005 }
1006
1007 static void __exit nfnetlink_queue_fini(void)
1008 {
1009         nf_unregister_queue_handlers(&nfqh);
1010         unregister_netdevice_notifier(&nfqnl_dev_notifier);
1011 #ifdef CONFIG_PROC_FS
1012         remove_proc_entry("nfnetlink_queue", proc_net_netfilter);
1013 #endif
1014         nfnetlink_subsys_unregister(&nfqnl_subsys);
1015         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1016 }
1017
1018 MODULE_DESCRIPTION("netfilter packet queue handler");
1019 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1020 MODULE_LICENSE("GPL");
1021 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1022
1023 module_init(nfnetlink_queue_init);
1024 module_exit(nfnetlink_queue_fini);