net/ipv6/netfilter/ip6_queue.c

   1 /*
   2  * This is a module which is used for queueing IPv6 packets and
   3  * communicating with userspace via netlink.
   4  *
   5  * (C) 2001 Fernando Anton, this code is GPL.
   6  *     IPv64 Project - Work based in IPv64 draft by Arturo Azcorra.
   7  *     Universidad Carlos III de Madrid - Leganes (Madrid) - Spain
   8  *     Universidad Politecnica de Alcala de Henares - Alcala de H. (Madrid) - Spain
   9  *     email: fanton@it.uc3m.es
  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 #include <linux/module.h>
  16 #include <linux/skbuff.h>
  17 #include <linux/init.h>
  18 #include <linux/ipv6.h>
  19 #include <linux/notifier.h>
  20 #include <linux/netdevice.h>
  21 #include <linux/netfilter.h>
  22 #include <linux/netlink.h>
  23 #include <linux/spinlock.h>
  24 #include <linux/sysctl.h>
  25 #include <linux/proc_fs.h>
  26 #include <linux/seq_file.h>
  27 #include <linux/mutex.h>
  28 #include <net/net_namespace.h>
  29 #include <net/sock.h>
  30 #include <net/ipv6.h>
  31 #include <net/ip6_route.h>
  32 #include <net/netfilter/nf_queue.h>
  33 #include <linux/netfilter_ipv4/ip_queue.h>
  34 #include <linux/netfilter_ipv4/ip_tables.h>
  35 #include <linux/netfilter_ipv6/ip6_tables.h>
  36
  37 #define IPQ_QMAX_DEFAULT 1024
  38 #define IPQ_PROC_FS_NAME "ip6_queue"
  39 #define NET_IPQ_QMAX 2088
  40 #define NET_IPQ_QMAX_NAME "ip6_queue_maxlen"
  41
  42 typedef int (*ipq_cmpfn)(struct nf_queue_entry *, unsigned long);
  43
  44 static unsigned char copy_mode __read_mostly = IPQ_COPY_NONE;
  45 static unsigned int queue_maxlen __read_mostly = IPQ_QMAX_DEFAULT;
  46 static DEFINE_RWLOCK(queue_lock);
  47 static int peer_pid __read_mostly;
  48 static unsigned int copy_range __read_mostly;
  49 static unsigned int queue_total;
  50 static unsigned int queue_dropped = 0;
  51 static unsigned int queue_user_dropped = 0;
  52 static struct sock *ipqnl __read_mostly;
  53 static LIST_HEAD(queue_list);
  54 static DEFINE_MUTEX(ipqnl_mutex);
  55
  56 static inline void
  57 __ipq_enqueue_entry(struct nf_queue_entry *entry)
  58 {
  59        list_add_tail(&entry->list, &queue_list);
  60        queue_total++;
  61 }
  62
  63 static inline int
  64 __ipq_set_mode(unsigned char mode, unsigned int range)
  65 {
  66         int status = 0;
  67
  68         switch(mode) {
  69         case IPQ_COPY_NONE:
  70         case IPQ_COPY_META:
  71                 copy_mode = mode;
  72                 copy_range = 0;
  73                 break;
  74
  75         case IPQ_COPY_PACKET:
  76                 copy_mode = mode;
  77                 copy_range = range;
  78                 if (copy_range > 0xFFFF)
  79                         copy_range = 0xFFFF;
  80                 break;
  81
  82         default:
  83                 status = -EINVAL;
  84
  85         }
  86         return status;
  87 }
  88
  89 static void __ipq_flush(ipq_cmpfn cmpfn, unsigned long data);
  90
  91 static inline void
  92 __ipq_reset(void)
  93 {
  94         peer_pid = 0;
  95         net_disable_timestamp();
  96         __ipq_set_mode(IPQ_COPY_NONE, 0);
  97         __ipq_flush(NULL, 0);
  98 }
  99
 100 static struct nf_queue_entry *
 101 ipq_find_dequeue_entry(unsigned long id)
 102 {
 103         struct nf_queue_entry *entry = NULL, *i;
 104
 105         write_lock_bh(&queue_lock);
 106
 107         list_for_each_entry(i, &queue_list, list) {
 108                 if ((unsigned long)i == id) {
 109                         entry = i;
 110                         break;
 111                 }
 112         }
 113
 114         if (entry) {
 115                 list_del(&entry->list);
 116                 queue_total--;
 117         }
 118
 119         write_unlock_bh(&queue_lock);
 120         return entry;
 121 }
 122
 123 static void
 124 __ipq_flush(ipq_cmpfn cmpfn, unsigned long data)
 125 {
 126         struct nf_queue_entry *entry, *next;
 127
 128         list_for_each_entry_safe(entry, next, &queue_list, list) {
 129                 if (!cmpfn || cmpfn(entry, data)) {
 130                         list_del(&entry->list);
 131                         queue_total--;
 132                         nf_reinject(entry, NF_DROP);
 133                 }
 134         }
 135 }
 136
 137 static void
 138 ipq_flush(ipq_cmpfn cmpfn, unsigned long data)
 139 {
 140         write_lock_bh(&queue_lock);
 141         __ipq_flush(cmpfn, data);
 142         write_unlock_bh(&queue_lock);
 143 }
 144
 145 static struct sk_buff *
 146 ipq_build_packet_message(struct nf_queue_entry *entry, int *errp)
 147 {
 148         sk_buff_data_t old_tail;
 149         size_t size = 0;
 150         size_t data_len = 0;
 151         struct sk_buff *skb;
 152         struct ipq_packet_msg *pmsg;
 153         struct nlmsghdr *nlh;
 154         struct timeval tv;
 155
 156         read_lock_bh(&queue_lock);
 157
 158         switch (copy_mode) {
 159         case IPQ_COPY_META:
 160         case IPQ_COPY_NONE:
 161                 size = NLMSG_SPACE(sizeof(*pmsg));
 162                 data_len = 0;
 163                 break;
 164
 165         case IPQ_COPY_PACKET:
 166                 if ((entry->skb->ip_summed == CHECKSUM_PARTIAL ||
 167                      entry->skb->ip_summed == CHECKSUM_COMPLETE) &&
 168                     (*errp = skb_checksum_help(entry->skb))) {
 169                         read_unlock_bh(&queue_lock);
 170                         return NULL;
 171                 }
 172                 if (copy_range == 0 || copy_range > entry->skb->len)
 173                         data_len = entry->skb->len;
 174                 else
 175                         data_len = copy_range;
 176
 177                 size = NLMSG_SPACE(sizeof(*pmsg) + data_len);
 178                 break;
 179
 180         default:
 181                 *errp = -EINVAL;
 182                 read_unlock_bh(&queue_lock);
 183                 return NULL;
 184         }
 185
 186         read_unlock_bh(&queue_lock);
 187
 188         skb = alloc_skb(size, GFP_ATOMIC);
 189         if (!skb)
 190                 goto nlmsg_failure;
 191
 192         old_tail = skb->tail;
 193         nlh = NLMSG_PUT(skb, 0, 0, IPQM_PACKET, size - sizeof(*nlh));
 194         pmsg = NLMSG_DATA(nlh);
 195         memset(pmsg, 0, sizeof(*pmsg));
 196
 197         pmsg->packet_id       = (unsigned long )entry;
 198         pmsg->data_len        = data_len;
 199         tv = ktime_to_timeval(entry->skb->tstamp);
 200         pmsg->timestamp_sec   = tv.tv_sec;
 201         pmsg->timestamp_usec  = tv.tv_usec;
 202         pmsg->mark            = entry->skb->mark;
 203         pmsg->hook            = entry->hook;
 204         pmsg->hw_protocol     = entry->skb->protocol;
 205
 206         if (entry->indev)
 207                 strcpy(pmsg->indev_name, entry->indev->name);
 208         else
 209                 pmsg->indev_name[0] = '\0';
 210
 211         if (entry->outdev)
 212                 strcpy(pmsg->outdev_name, entry->outdev->name);
 213         else
 214                 pmsg->outdev_name[0] = '\0';
 215
 216         if (entry->indev && entry->skb->dev) {
 217                 pmsg->hw_type = entry->skb->dev->type;
 218                 pmsg->hw_addrlen = dev_parse_header(entry->skb, pmsg->hw_addr);
 219         }
 220
 221         if (data_len)
 222                 if (skb_copy_bits(entry->skb, 0, pmsg->payload, data_len))
 223                         BUG();
 224
 225         nlh->nlmsg_len = skb->tail - old_tail;
 226         return skb;
 227
 228 nlmsg_failure:
 229         if (skb)
 230                 kfree_skb(skb);
 231         *errp = -EINVAL;
 232         printk(KERN_ERR "ip6_queue: error creating packet message\n");
 233         return NULL;
 234 }
 235
 236 static int
 237 ipq_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 238 {
 239         int status = -EINVAL;
 240         struct sk_buff *nskb;
 241
 242         if (copy_mode == IPQ_COPY_NONE)
 243                 return -EAGAIN;
 244
 245         nskb = ipq_build_packet_message(entry, &status);
 246         if (nskb == NULL)
 247                 return status;
 248
 249         write_lock_bh(&queue_lock);
 250
 251         if (!peer_pid)
 252                 goto err_out_free_nskb;
 253
 254         if (queue_total >= queue_maxlen) {
 255                 queue_dropped++;
 256                 status = -ENOSPC;
 257                 if (net_ratelimit())
 258                         printk (KERN_WARNING "ip6_queue: fill at %d entries, "
 259                                 "dropping packet(s).  Dropped: %d\n", queue_total,
 260                                 queue_dropped);
 261                 goto err_out_free_nskb;
 262         }
 263
 264         /* netlink_unicast will either free the nskb or attach it to a socket */
 265         status = netlink_unicast(ipqnl, nskb, peer_pid, MSG_DONTWAIT);
 266         if (status < 0) {
 267                 queue_user_dropped++;
 268                 goto err_out_unlock;
 269         }
 270
 271         __ipq_enqueue_entry(entry);
 272
 273         write_unlock_bh(&queue_lock);
 274         return status;
 275
 276 err_out_free_nskb:
 277         kfree_skb(nskb);
 278
 279 err_out_unlock:
 280         write_unlock_bh(&queue_lock);
 281         return status;
 282 }
 283
 284 static int
 285 ipq_mangle_ipv6(ipq_verdict_msg_t *v, struct nf_queue_entry *e)
 286 {
 287         int diff;
 288         int err;
 289         struct ipv6hdr *user_iph = (struct ipv6hdr *)v->payload;
 290
 291         if (v->data_len < sizeof(*user_iph))
 292                 return 0;
 293         diff = v->data_len - e->skb->len;
 294         if (diff < 0) {
 295                 if (pskb_trim(e->skb, v->data_len))
 296                         return -ENOMEM;
 297         } else if (diff > 0) {
 298                 if (v->data_len > 0xFFFF)
 299                         return -EINVAL;
 300                 if (diff > skb_tailroom(e->skb)) {
 301                         err = pskb_expand_head(e->skb, 0,
 302                                                diff - skb_tailroom(e->skb),
 303                                                GFP_ATOMIC);
 304                         if (err) {
 305                                 printk(KERN_WARNING "ip6_queue: OOM "
 306                                       "in mangle, dropping packet\n");
 307                                 return err;
 308                         }
 309                 }
 310                 skb_put(e->skb, diff);
 311         }
 312         if (!skb_make_writable(e->skb, v->data_len))
 313                 return -ENOMEM;
 314         skb_copy_to_linear_data(e->skb, v->payload, v->data_len);
 315         e->skb->ip_summed = CHECKSUM_NONE;
 316
 317         return 0;
 318 }
 319
 320 static int
 321 ipq_set_verdict(struct ipq_verdict_msg *vmsg, unsigned int len)
 322 {
 323         struct nf_queue_entry *entry;
 324
 325         if (vmsg->value > NF_MAX_VERDICT)
 326                 return -EINVAL;
 327
 328         entry = ipq_find_dequeue_entry(vmsg->id);
 329         if (entry == NULL)
 330                 return -ENOENT;
 331         else {
 332                 int verdict = vmsg->value;
 333
 334                 if (vmsg->data_len && vmsg->data_len == len)
 335                         if (ipq_mangle_ipv6(vmsg, entry) < 0)
 336                                 verdict = NF_DROP;
 337
 338                 nf_reinject(entry, verdict);
 339                 return 0;
 340         }
 341 }
 342
 343 static int
 344 ipq_set_mode(unsigned char mode, unsigned int range)
 345 {
 346         int status;
 347
 348         write_lock_bh(&queue_lock);
 349         status = __ipq_set_mode(mode, range);
 350         write_unlock_bh(&queue_lock);
 351         return status;
 352 }
 353
 354 static int
 355 ipq_receive_peer(struct ipq_peer_msg *pmsg,
 356                  unsigned char type, unsigned int len)
 357 {
 358         int status = 0;
 359
 360         if (len < sizeof(*pmsg))
 361                 return -EINVAL;
 362
 363         switch (type) {
 364         case IPQM_MODE:
 365                 status = ipq_set_mode(pmsg->msg.mode.value,
 366                                       pmsg->msg.mode.range);
 367                 break;
 368
 369         case IPQM_VERDICT:
 370                 if (pmsg->msg.verdict.value > NF_MAX_VERDICT)
 371                         status = -EINVAL;
 372                 else
 373                         status = ipq_set_verdict(&pmsg->msg.verdict,
 374                                                  len - sizeof(*pmsg));
 375                         break;
 376         default:
 377                 status = -EINVAL;
 378         }
 379         return status;
 380 }
 381
 382 static int
 383 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
 384 {
 385         if (entry->indev)
 386                 if (entry->indev->ifindex == ifindex)
 387                         return 1;
 388
 389         if (entry->outdev)
 390                 if (entry->outdev->ifindex == ifindex)
 391                         return 1;
 392 #ifdef CONFIG_BRIDGE_NETFILTER
 393         if (entry->skb->nf_bridge) {
 394                 if (entry->skb->nf_bridge->physindev &&
 395                     entry->skb->nf_bridge->physindev->ifindex == ifindex)
 396                         return 1;
 397                 if (entry->skb->nf_bridge->physoutdev &&
 398                     entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
 399                         return 1;
 400         }
 401 #endif
 402         return 0;
 403 }
 404
 405 static void
 406 ipq_dev_drop(int ifindex)
 407 {
 408         ipq_flush(dev_cmp, ifindex);
 409 }
 410
 411 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
 412
 413 static inline void
 414 __ipq_rcv_skb(struct sk_buff *skb)
 415 {
 416         int status, type, pid, flags, nlmsglen, skblen;
 417         struct nlmsghdr *nlh;
 418
 419         skblen = skb->len;
 420         if (skblen < sizeof(*nlh))
 421                 return;
 422
 423         nlh = nlmsg_hdr(skb);
 424         nlmsglen = nlh->nlmsg_len;
 425         if (nlmsglen < sizeof(*nlh) || skblen < nlmsglen)
 426                 return;
 427
 428         pid = nlh->nlmsg_pid;
 429         flags = nlh->nlmsg_flags;
 430
 431         if(pid <= 0 || !(flags & NLM_F_REQUEST) || flags & NLM_F_MULTI)
 432                 RCV_SKB_FAIL(-EINVAL);
 433
 434         if (flags & MSG_TRUNC)
 435                 RCV_SKB_FAIL(-ECOMM);
 436
 437         type = nlh->nlmsg_type;
 438         if (type < NLMSG_NOOP || type >= IPQM_MAX)
 439                 RCV_SKB_FAIL(-EINVAL);
 440
 441         if (type <= IPQM_BASE)
 442                 return;
 443
 444         if (security_netlink_recv(skb, CAP_NET_ADMIN))
 445                 RCV_SKB_FAIL(-EPERM);
 446
 447         write_lock_bh(&queue_lock);
 448
 449         if (peer_pid) {
 450                 if (peer_pid != pid) {
 451                         write_unlock_bh(&queue_lock);
 452                         RCV_SKB_FAIL(-EBUSY);
 453                 }
 454         } else {
 455                 net_enable_timestamp();
 456                 peer_pid = pid;
 457         }
 458
 459         write_unlock_bh(&queue_lock);
 460
 461         status = ipq_receive_peer(NLMSG_DATA(nlh), type,
 462                                   nlmsglen - NLMSG_LENGTH(0));
 463         if (status < 0)
 464                 RCV_SKB_FAIL(status);
 465
 466         if (flags & NLM_F_ACK)
 467                 netlink_ack(skb, nlh, 0);
 468         return;
 469 }
 470
 471 static void
 472 ipq_rcv_skb(struct sk_buff *skb)
 473 {
 474         mutex_lock(&ipqnl_mutex);
 475         __ipq_rcv_skb(skb);
 476         mutex_unlock(&ipqnl_mutex);
 477 }
 478
 479 static int
 480 ipq_rcv_dev_event(struct notifier_block *this,
 481                   unsigned long event, void *ptr)
 482 {
 483         struct net_device *dev = ptr;
 484
 485         if (dev->nd_net != &init_net)
 486                 return NOTIFY_DONE;
 487
 488         /* Drop any packets associated with the downed device */
 489         if (event == NETDEV_DOWN)
 490                 ipq_dev_drop(dev->ifindex);
 491         return NOTIFY_DONE;
 492 }
 493
 494 static struct notifier_block ipq_dev_notifier = {
 495         .notifier_call  = ipq_rcv_dev_event,
 496 };
 497
 498 static int
 499 ipq_rcv_nl_event(struct notifier_block *this,
 500                  unsigned long event, void *ptr)
 501 {
 502         struct netlink_notify *n = ptr;
 503
 504         if (event == NETLINK_URELEASE &&
 505             n->protocol == NETLINK_IP6_FW && n->pid) {
 506                 write_lock_bh(&queue_lock);
 507                 if ((n->net == &init_net) && (n->pid == peer_pid))
 508                         __ipq_reset();
 509                 write_unlock_bh(&queue_lock);
 510         }
 511         return NOTIFY_DONE;
 512 }
 513
 514 static struct notifier_block ipq_nl_notifier = {
 515         .notifier_call  = ipq_rcv_nl_event,
 516 };
 517
 518 static struct ctl_table_header *ipq_sysctl_header;
 519
 520 static ctl_table ipq_table[] = {
 521         {
 522                 .ctl_name       = NET_IPQ_QMAX,
 523                 .procname       = NET_IPQ_QMAX_NAME,
 524                 .data           = &queue_maxlen,
 525                 .maxlen         = sizeof(queue_maxlen),
 526                 .mode           = 0644,
 527                 .proc_handler   = proc_dointvec
 528         },
 529         { .ctl_name = 0 }
 530 };
 531
 532 static int ip6_queue_show(struct seq_file *m, void *v)
 533 {
 534         read_lock_bh(&queue_lock);
 535
 536         seq_printf(m,
 537                       "Peer PID          : %d\n"
 538                       "Copy mode         : %hu\n"
 539                       "Copy range        : %u\n"
 540                       "Queue length      : %u\n"
 541                       "Queue max. length : %u\n"
 542                       "Queue dropped     : %u\n"
 543                       "Netfilter dropped : %u\n",
 544                       peer_pid,
 545                       copy_mode,
 546                       copy_range,
 547                       queue_total,
 548                       queue_maxlen,
 549                       queue_dropped,
 550                       queue_user_dropped);
 551
 552         read_unlock_bh(&queue_lock);
 553         return 0;
 554 }
 555
 556 static int ip6_queue_open(struct inode *inode, struct file *file)
 557 {
 558         return single_open(file, ip6_queue_show, NULL);
 559 }
 560
 561 static const struct file_operations ip6_queue_proc_fops = {
 562         .open           = ip6_queue_open,
 563         .read           = seq_read,
 564         .llseek         = seq_lseek,
 565         .release        = single_release,
 566         .owner          = THIS_MODULE,
 567 };
 568
 569 static const struct nf_queue_handler nfqh = {
 570         .name   = "ip6_queue",
 571         .outfn  = &ipq_enqueue_packet,
 572 };
 573
 574 static int __init ip6_queue_init(void)
 575 {
 576         int status = -ENOMEM;
 577         struct proc_dir_entry *proc;
 578
 579         netlink_register_notifier(&ipq_nl_notifier);
 580         ipqnl = netlink_kernel_create(&init_net, NETLINK_IP6_FW, 0,
 581                                       ipq_rcv_skb, NULL, THIS_MODULE);
 582         if (ipqnl == NULL) {
 583                 printk(KERN_ERR "ip6_queue: failed to create netlink socket\n");
 584                 goto cleanup_netlink_notifier;
 585         }
 586
 587         proc = create_proc_entry(IPQ_PROC_FS_NAME, 0, init_net.proc_net);
 588         if (proc) {
 589                 proc->owner = THIS_MODULE;
 590                 proc->proc_fops = &ip6_queue_proc_fops;
 591         } else {
 592                 printk(KERN_ERR "ip6_queue: failed to create proc entry\n");
 593                 goto cleanup_ipqnl;
 594         }
 595
 596         register_netdevice_notifier(&ipq_dev_notifier);
 597         ipq_sysctl_header = register_sysctl_paths(net_ipv6_ctl_path, ipq_table);
 598
 599         status = nf_register_queue_handler(PF_INET6, &nfqh);
 600         if (status < 0) {
 601                 printk(KERN_ERR "ip6_queue: failed to register queue handler\n");
 602                 goto cleanup_sysctl;
 603         }
 604         return status;
 605
 606 cleanup_sysctl:
 607         unregister_sysctl_table(ipq_sysctl_header);
 608         unregister_netdevice_notifier(&ipq_dev_notifier);
 609         proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
 610
 611 cleanup_ipqnl:
 612         netlink_kernel_release(ipqnl);
 613         mutex_lock(&ipqnl_mutex);
 614         mutex_unlock(&ipqnl_mutex);
 615
 616 cleanup_netlink_notifier:
 617         netlink_unregister_notifier(&ipq_nl_notifier);
 618         return status;
 619 }
 620
 621 static void __exit ip6_queue_fini(void)
 622 {
 623         nf_unregister_queue_handlers(&nfqh);
 624         synchronize_net();
 625         ipq_flush(NULL, 0);
 626
 627         unregister_sysctl_table(ipq_sysctl_header);
 628         unregister_netdevice_notifier(&ipq_dev_notifier);
 629         proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
 630
 631         netlink_kernel_release(ipqnl);
 632         mutex_lock(&ipqnl_mutex);
 633         mutex_unlock(&ipqnl_mutex);
 634
 635         netlink_unregister_notifier(&ipq_nl_notifier);
 636 }
 637
 638 MODULE_DESCRIPTION("IPv6 packet queue handler");
 639 MODULE_LICENSE("GPL");
 640
 641 module_init(ip6_queue_init);
 642 module_exit(ip6_queue_fini);