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         struct ipv6hdr *user_iph = (struct ipv6hdr *)v->payload;
 289         struct sk_buff *nskb;
 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                         nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
 302                                                diff, GFP_ATOMIC);
 303                         if (!nskb) {
 304                                 printk(KERN_WARNING "ip6_queue: OOM "
 305                                       "in mangle, dropping packet\n");
 306                                 return -ENOMEM;
 307                         }
 308                         kfree_skb(e->skb);
 309                         e->skb = nskb;
 310                 }
 311                 skb_put(e->skb, diff);
 312         }
 313         if (!skb_make_writable(e->skb, v->data_len))
 314                 return -ENOMEM;
 315         skb_copy_to_linear_data(e->skb, v->payload, v->data_len);
 316         e->skb->ip_summed = CHECKSUM_NONE;
 317
 318         return 0;
 319 }
 320
 321 static int
 322 ipq_set_verdict(struct ipq_verdict_msg *vmsg, unsigned int len)
 323 {
 324         struct nf_queue_entry *entry;
 325
 326         if (vmsg->value > NF_MAX_VERDICT)
 327                 return -EINVAL;
 328
 329         entry = ipq_find_dequeue_entry(vmsg->id);
 330         if (entry == NULL)
 331                 return -ENOENT;
 332         else {
 333                 int verdict = vmsg->value;
 334
 335                 if (vmsg->data_len && vmsg->data_len == len)
 336                         if (ipq_mangle_ipv6(vmsg, entry) < 0)
 337                                 verdict = NF_DROP;
 338
 339                 nf_reinject(entry, verdict);
 340                 return 0;
 341         }
 342 }
 343
 344 static int
 345 ipq_set_mode(unsigned char mode, unsigned int range)
 346 {
 347         int status;
 348
 349         write_lock_bh(&queue_lock);
 350         status = __ipq_set_mode(mode, range);
 351         write_unlock_bh(&queue_lock);
 352         return status;
 353 }
 354
 355 static int
 356 ipq_receive_peer(struct ipq_peer_msg *pmsg,
 357                  unsigned char type, unsigned int len)
 358 {
 359         int status = 0;
 360
 361         if (len < sizeof(*pmsg))
 362                 return -EINVAL;
 363
 364         switch (type) {
 365         case IPQM_MODE:
 366                 status = ipq_set_mode(pmsg->msg.mode.value,
 367                                       pmsg->msg.mode.range);
 368                 break;
 369
 370         case IPQM_VERDICT:
 371                 if (pmsg->msg.verdict.value > NF_MAX_VERDICT)
 372                         status = -EINVAL;
 373                 else
 374                         status = ipq_set_verdict(&pmsg->msg.verdict,
 375                                                  len - sizeof(*pmsg));
 376                         break;
 377         default:
 378                 status = -EINVAL;
 379         }
 380         return status;
 381 }
 382
 383 static int
 384 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
 385 {
 386         if (entry->indev)
 387                 if (entry->indev->ifindex == ifindex)
 388                         return 1;
 389
 390         if (entry->outdev)
 391                 if (entry->outdev->ifindex == ifindex)
 392                         return 1;
 393 #ifdef CONFIG_BRIDGE_NETFILTER
 394         if (entry->skb->nf_bridge) {
 395                 if (entry->skb->nf_bridge->physindev &&
 396                     entry->skb->nf_bridge->physindev->ifindex == ifindex)
 397                         return 1;
 398                 if (entry->skb->nf_bridge->physoutdev &&
 399                     entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
 400                         return 1;
 401         }
 402 #endif
 403         return 0;
 404 }
 405
 406 static void
 407 ipq_dev_drop(int ifindex)
 408 {
 409         ipq_flush(dev_cmp, ifindex);
 410 }
 411
 412 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
 413
 414 static inline void
 415 __ipq_rcv_skb(struct sk_buff *skb)
 416 {
 417         int status, type, pid, flags, nlmsglen, skblen;
 418         struct nlmsghdr *nlh;
 419
 420         skblen = skb->len;
 421         if (skblen < sizeof(*nlh))
 422                 return;
 423
 424         nlh = nlmsg_hdr(skb);
 425         nlmsglen = nlh->nlmsg_len;
 426         if (nlmsglen < sizeof(*nlh) || skblen < nlmsglen)
 427                 return;
 428
 429         pid = nlh->nlmsg_pid;
 430         flags = nlh->nlmsg_flags;
 431
 432         if(pid <= 0 || !(flags & NLM_F_REQUEST) || flags & NLM_F_MULTI)
 433                 RCV_SKB_FAIL(-EINVAL);
 434
 435         if (flags & MSG_TRUNC)
 436                 RCV_SKB_FAIL(-ECOMM);
 437
 438         type = nlh->nlmsg_type;
 439         if (type < NLMSG_NOOP || type >= IPQM_MAX)
 440                 RCV_SKB_FAIL(-EINVAL);
 441
 442         if (type <= IPQM_BASE)
 443                 return;
 444
 445         if (security_netlink_recv(skb, CAP_NET_ADMIN))
 446                 RCV_SKB_FAIL(-EPERM);
 447
 448         write_lock_bh(&queue_lock);
 449
 450         if (peer_pid) {
 451                 if (peer_pid != pid) {
 452                         write_unlock_bh(&queue_lock);
 453                         RCV_SKB_FAIL(-EBUSY);
 454                 }
 455         } else {
 456                 net_enable_timestamp();
 457                 peer_pid = pid;
 458         }
 459
 460         write_unlock_bh(&queue_lock);
 461
 462         status = ipq_receive_peer(NLMSG_DATA(nlh), type,
 463                                   nlmsglen - NLMSG_LENGTH(0));
 464         if (status < 0)
 465                 RCV_SKB_FAIL(status);
 466
 467         if (flags & NLM_F_ACK)
 468                 netlink_ack(skb, nlh, 0);
 469         return;
 470 }
 471
 472 static void
 473 ipq_rcv_skb(struct sk_buff *skb)
 474 {
 475         mutex_lock(&ipqnl_mutex);
 476         __ipq_rcv_skb(skb);
 477         mutex_unlock(&ipqnl_mutex);
 478 }
 479
 480 static int
 481 ipq_rcv_dev_event(struct notifier_block *this,
 482                   unsigned long event, void *ptr)
 483 {
 484         struct net_device *dev = ptr;
 485
 486         if (dev->nd_net != &init_net)
 487                 return NOTIFY_DONE;
 488
 489         /* Drop any packets associated with the downed device */
 490         if (event == NETDEV_DOWN)
 491                 ipq_dev_drop(dev->ifindex);
 492         return NOTIFY_DONE;
 493 }
 494
 495 static struct notifier_block ipq_dev_notifier = {
 496         .notifier_call  = ipq_rcv_dev_event,
 497 };
 498
 499 static int
 500 ipq_rcv_nl_event(struct notifier_block *this,
 501                  unsigned long event, void *ptr)
 502 {
 503         struct netlink_notify *n = ptr;
 504
 505         if (event == NETLINK_URELEASE &&
 506             n->protocol == NETLINK_IP6_FW && n->pid) {
 507                 write_lock_bh(&queue_lock);
 508                 if ((n->net == &init_net) && (n->pid == peer_pid))
 509                         __ipq_reset();
 510                 write_unlock_bh(&queue_lock);
 511         }
 512         return NOTIFY_DONE;
 513 }
 514
 515 static struct notifier_block ipq_nl_notifier = {
 516         .notifier_call  = ipq_rcv_nl_event,
 517 };
 518
 519 #ifdef CONFIG_SYSCTL
 520 static struct ctl_table_header *ipq_sysctl_header;
 521
 522 static ctl_table ipq_table[] = {
 523         {
 524                 .ctl_name       = NET_IPQ_QMAX,
 525                 .procname       = NET_IPQ_QMAX_NAME,
 526                 .data           = &queue_maxlen,
 527                 .maxlen         = sizeof(queue_maxlen),
 528                 .mode           = 0644,
 529                 .proc_handler   = proc_dointvec
 530         },
 531         { .ctl_name = 0 }
 532 };
 533 #endif
 534
 535 #ifdef CONFIG_PROC_FS
 536 static int ip6_queue_show(struct seq_file *m, void *v)
 537 {
 538         read_lock_bh(&queue_lock);
 539
 540         seq_printf(m,
 541                       "Peer PID          : %d\n"
 542                       "Copy mode         : %hu\n"
 543                       "Copy range        : %u\n"
 544                       "Queue length      : %u\n"
 545                       "Queue max. length : %u\n"
 546                       "Queue dropped     : %u\n"
 547                       "Netfilter dropped : %u\n",
 548                       peer_pid,
 549                       copy_mode,
 550                       copy_range,
 551                       queue_total,
 552                       queue_maxlen,
 553                       queue_dropped,
 554                       queue_user_dropped);
 555
 556         read_unlock_bh(&queue_lock);
 557         return 0;
 558 }
 559
 560 static int ip6_queue_open(struct inode *inode, struct file *file)
 561 {
 562         return single_open(file, ip6_queue_show, NULL);
 563 }
 564
 565 static const struct file_operations ip6_queue_proc_fops = {
 566         .open           = ip6_queue_open,
 567         .read           = seq_read,
 568         .llseek         = seq_lseek,
 569         .release        = single_release,
 570         .owner          = THIS_MODULE,
 571 };
 572 #endif
 573
 574 static const struct nf_queue_handler nfqh = {
 575         .name   = "ip6_queue",
 576         .outfn  = &ipq_enqueue_packet,
 577 };
 578
 579 static int __init ip6_queue_init(void)
 580 {
 581         int status = -ENOMEM;
 582         struct proc_dir_entry *proc __maybe_unused;
 583
 584         netlink_register_notifier(&ipq_nl_notifier);
 585         ipqnl = netlink_kernel_create(&init_net, NETLINK_IP6_FW, 0,
 586                                       ipq_rcv_skb, NULL, THIS_MODULE);
 587         if (ipqnl == NULL) {
 588                 printk(KERN_ERR "ip6_queue: failed to create netlink socket\n");
 589                 goto cleanup_netlink_notifier;
 590         }
 591
 592 #ifdef CONFIG_PROC_FS
 593         proc = proc_create(IPQ_PROC_FS_NAME, 0, init_net.proc_net,
 594                            &ip6_queue_proc_fops);
 595         if (!proc) {
 596                 printk(KERN_ERR "ip6_queue: failed to create proc entry\n");
 597                 goto cleanup_ipqnl;
 598         }
 599 #endif
 600         register_netdevice_notifier(&ipq_dev_notifier);
 601 #ifdef CONFIG_SYSCTL
 602         ipq_sysctl_header = register_sysctl_paths(net_ipv6_ctl_path, ipq_table);
 603 #endif
 604         status = nf_register_queue_handler(PF_INET6, &nfqh);
 605         if (status < 0) {
 606                 printk(KERN_ERR "ip6_queue: failed to register queue handler\n");
 607                 goto cleanup_sysctl;
 608         }
 609         return status;
 610
 611 cleanup_sysctl:
 612 #ifdef CONFIG_SYSCTL
 613         unregister_sysctl_table(ipq_sysctl_header);
 614 #endif
 615         unregister_netdevice_notifier(&ipq_dev_notifier);
 616         proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
 617
 618 cleanup_ipqnl: __maybe_unused
 619         netlink_kernel_release(ipqnl);
 620         mutex_lock(&ipqnl_mutex);
 621         mutex_unlock(&ipqnl_mutex);
 622
 623 cleanup_netlink_notifier:
 624         netlink_unregister_notifier(&ipq_nl_notifier);
 625         return status;
 626 }
 627
 628 static void __exit ip6_queue_fini(void)
 629 {
 630         nf_unregister_queue_handlers(&nfqh);
 631         synchronize_net();
 632         ipq_flush(NULL, 0);
 633
 634 #ifdef CONFIG_SYSCTL
 635         unregister_sysctl_table(ipq_sysctl_header);
 636 #endif
 637         unregister_netdevice_notifier(&ipq_dev_notifier);
 638         proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
 639
 640         netlink_kernel_release(ipqnl);
 641         mutex_lock(&ipqnl_mutex);
 642         mutex_unlock(&ipqnl_mutex);
 643
 644         netlink_unregister_notifier(&ipq_nl_notifier);
 645 }
 646
 647 MODULE_DESCRIPTION("IPv6 packet queue handler");
 648 MODULE_LICENSE("GPL");
 649
 650 module_init(ip6_queue_init);
 651 module_exit(ip6_queue_fini);