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                 break;
 163
 164         case IPQ_COPY_PACKET:
 165                 if ((entry->skb->ip_summed == CHECKSUM_PARTIAL ||
 166                      entry->skb->ip_summed == CHECKSUM_COMPLETE) &&
 167                     (*errp = skb_checksum_help(entry->skb))) {
 168                         read_unlock_bh(&queue_lock);
 169                         return NULL;
 170                 }
 171                 if (copy_range == 0 || copy_range > entry->skb->len)
 172                         data_len = entry->skb->len;
 173                 else
 174                         data_len = copy_range;
 175
 176                 size = NLMSG_SPACE(sizeof(*pmsg) + data_len);
 177                 break;
 178
 179         default:
 180                 *errp = -EINVAL;
 181                 read_unlock_bh(&queue_lock);
 182                 return NULL;
 183         }
 184
 185         read_unlock_bh(&queue_lock);
 186
 187         skb = alloc_skb(size, GFP_ATOMIC);
 188         if (!skb)
 189                 goto nlmsg_failure;
 190
 191         old_tail = skb->tail;
 192         nlh = NLMSG_PUT(skb, 0, 0, IPQM_PACKET, size - sizeof(*nlh));
 193         pmsg = NLMSG_DATA(nlh);
 194         memset(pmsg, 0, sizeof(*pmsg));
 195
 196         pmsg->packet_id       = (unsigned long )entry;
 197         pmsg->data_len        = data_len;
 198         tv = ktime_to_timeval(entry->skb->tstamp);
 199         pmsg->timestamp_sec   = tv.tv_sec;
 200         pmsg->timestamp_usec  = tv.tv_usec;
 201         pmsg->mark            = entry->skb->mark;
 202         pmsg->hook            = entry->hook;
 203         pmsg->hw_protocol     = entry->skb->protocol;
 204
 205         if (entry->indev)
 206                 strcpy(pmsg->indev_name, entry->indev->name);
 207         else
 208                 pmsg->indev_name[0] = '\0';
 209
 210         if (entry->outdev)
 211                 strcpy(pmsg->outdev_name, entry->outdev->name);
 212         else
 213                 pmsg->outdev_name[0] = '\0';
 214
 215         if (entry->indev && entry->skb->dev) {
 216                 pmsg->hw_type = entry->skb->dev->type;
 217                 pmsg->hw_addrlen = dev_parse_header(entry->skb, pmsg->hw_addr);
 218         }
 219
 220         if (data_len)
 221                 if (skb_copy_bits(entry->skb, 0, pmsg->payload, data_len))
 222                         BUG();
 223
 224         nlh->nlmsg_len = skb->tail - old_tail;
 225         return skb;
 226
 227 nlmsg_failure:
 228         *errp = -EINVAL;
 229         printk(KERN_ERR "ip6_queue: error creating packet message\n");
 230         return NULL;
 231 }
 232
 233 static int
 234 ipq_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 235 {
 236         int status = -EINVAL;
 237         struct sk_buff *nskb;
 238
 239         if (copy_mode == IPQ_COPY_NONE)
 240                 return -EAGAIN;
 241
 242         nskb = ipq_build_packet_message(entry, &status);
 243         if (nskb == NULL)
 244                 return status;
 245
 246         write_lock_bh(&queue_lock);
 247
 248         if (!peer_pid)
 249                 goto err_out_free_nskb;
 250
 251         if (queue_total >= queue_maxlen) {
 252                 queue_dropped++;
 253                 status = -ENOSPC;
 254                 if (net_ratelimit())
 255                         printk (KERN_WARNING "ip6_queue: fill at %d entries, "
 256                                 "dropping packet(s).  Dropped: %d\n", queue_total,
 257                                 queue_dropped);
 258                 goto err_out_free_nskb;
 259         }
 260
 261         /* netlink_unicast will either free the nskb or attach it to a socket */
 262         status = netlink_unicast(ipqnl, nskb, peer_pid, MSG_DONTWAIT);
 263         if (status < 0) {
 264                 queue_user_dropped++;
 265                 goto err_out_unlock;
 266         }
 267
 268         __ipq_enqueue_entry(entry);
 269
 270         write_unlock_bh(&queue_lock);
 271         return status;
 272
 273 err_out_free_nskb:
 274         kfree_skb(nskb);
 275
 276 err_out_unlock:
 277         write_unlock_bh(&queue_lock);
 278         return status;
 279 }
 280
 281 static int
 282 ipq_mangle_ipv6(ipq_verdict_msg_t *v, struct nf_queue_entry *e)
 283 {
 284         int diff;
 285         struct ipv6hdr *user_iph = (struct ipv6hdr *)v->payload;
 286         struct sk_buff *nskb;
 287
 288         if (v->data_len < sizeof(*user_iph))
 289                 return 0;
 290         diff = v->data_len - e->skb->len;
 291         if (diff < 0) {
 292                 if (pskb_trim(e->skb, v->data_len))
 293                         return -ENOMEM;
 294         } else if (diff > 0) {
 295                 if (v->data_len > 0xFFFF)
 296                         return -EINVAL;
 297                 if (diff > skb_tailroom(e->skb)) {
 298                         nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
 299                                                diff, GFP_ATOMIC);
 300                         if (!nskb) {
 301                                 printk(KERN_WARNING "ip6_queue: OOM "
 302                                       "in mangle, dropping packet\n");
 303                                 return -ENOMEM;
 304                         }
 305                         kfree_skb(e->skb);
 306                         e->skb = nskb;
 307                 }
 308                 skb_put(e->skb, diff);
 309         }
 310         if (!skb_make_writable(e->skb, v->data_len))
 311                 return -ENOMEM;
 312         skb_copy_to_linear_data(e->skb, v->payload, v->data_len);
 313         e->skb->ip_summed = CHECKSUM_NONE;
 314
 315         return 0;
 316 }
 317
 318 static int
 319 ipq_set_verdict(struct ipq_verdict_msg *vmsg, unsigned int len)
 320 {
 321         struct nf_queue_entry *entry;
 322
 323         if (vmsg->value > NF_MAX_VERDICT)
 324                 return -EINVAL;
 325
 326         entry = ipq_find_dequeue_entry(vmsg->id);
 327         if (entry == NULL)
 328                 return -ENOENT;
 329         else {
 330                 int verdict = vmsg->value;
 331
 332                 if (vmsg->data_len && vmsg->data_len == len)
 333                         if (ipq_mangle_ipv6(vmsg, entry) < 0)
 334                                 verdict = NF_DROP;
 335
 336                 nf_reinject(entry, verdict);
 337                 return 0;
 338         }
 339 }
 340
 341 static int
 342 ipq_set_mode(unsigned char mode, unsigned int range)
 343 {
 344         int status;
 345
 346         write_lock_bh(&queue_lock);
 347         status = __ipq_set_mode(mode, range);
 348         write_unlock_bh(&queue_lock);
 349         return status;
 350 }
 351
 352 static int
 353 ipq_receive_peer(struct ipq_peer_msg *pmsg,
 354                  unsigned char type, unsigned int len)
 355 {
 356         int status = 0;
 357
 358         if (len < sizeof(*pmsg))
 359                 return -EINVAL;
 360
 361         switch (type) {
 362         case IPQM_MODE:
 363                 status = ipq_set_mode(pmsg->msg.mode.value,
 364                                       pmsg->msg.mode.range);
 365                 break;
 366
 367         case IPQM_VERDICT:
 368                 if (pmsg->msg.verdict.value > NF_MAX_VERDICT)
 369                         status = -EINVAL;
 370                 else
 371                         status = ipq_set_verdict(&pmsg->msg.verdict,
 372                                                  len - sizeof(*pmsg));
 373                         break;
 374         default:
 375                 status = -EINVAL;
 376         }
 377         return status;
 378 }
 379
 380 static int
 381 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
 382 {
 383         if (entry->indev)
 384                 if (entry->indev->ifindex == ifindex)
 385                         return 1;
 386
 387         if (entry->outdev)
 388                 if (entry->outdev->ifindex == ifindex)
 389                         return 1;
 390 #ifdef CONFIG_BRIDGE_NETFILTER
 391         if (entry->skb->nf_bridge) {
 392                 if (entry->skb->nf_bridge->physindev &&
 393                     entry->skb->nf_bridge->physindev->ifindex == ifindex)
 394                         return 1;
 395                 if (entry->skb->nf_bridge->physoutdev &&
 396                     entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
 397                         return 1;
 398         }
 399 #endif
 400         return 0;
 401 }
 402
 403 static void
 404 ipq_dev_drop(int ifindex)
 405 {
 406         ipq_flush(dev_cmp, ifindex);
 407 }
 408
 409 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
 410
 411 static inline void
 412 __ipq_rcv_skb(struct sk_buff *skb)
 413 {
 414         int status, type, pid, flags, nlmsglen, skblen;
 415         struct nlmsghdr *nlh;
 416
 417         skblen = skb->len;
 418         if (skblen < sizeof(*nlh))
 419                 return;
 420
 421         nlh = nlmsg_hdr(skb);
 422         nlmsglen = nlh->nlmsg_len;
 423         if (nlmsglen < sizeof(*nlh) || skblen < nlmsglen)
 424                 return;
 425
 426         pid = nlh->nlmsg_pid;
 427         flags = nlh->nlmsg_flags;
 428
 429         if(pid <= 0 || !(flags & NLM_F_REQUEST) || flags & NLM_F_MULTI)
 430                 RCV_SKB_FAIL(-EINVAL);
 431
 432         if (flags & MSG_TRUNC)
 433                 RCV_SKB_FAIL(-ECOMM);
 434
 435         type = nlh->nlmsg_type;
 436         if (type < NLMSG_NOOP || type >= IPQM_MAX)
 437                 RCV_SKB_FAIL(-EINVAL);
 438
 439         if (type <= IPQM_BASE)
 440                 return;
 441
 442         if (security_netlink_recv(skb, CAP_NET_ADMIN))
 443                 RCV_SKB_FAIL(-EPERM);
 444
 445         write_lock_bh(&queue_lock);
 446
 447         if (peer_pid) {
 448                 if (peer_pid != pid) {
 449                         write_unlock_bh(&queue_lock);
 450                         RCV_SKB_FAIL(-EBUSY);
 451                 }
 452         } else {
 453                 net_enable_timestamp();
 454                 peer_pid = pid;
 455         }
 456
 457         write_unlock_bh(&queue_lock);
 458
 459         status = ipq_receive_peer(NLMSG_DATA(nlh), type,
 460                                   nlmsglen - NLMSG_LENGTH(0));
 461         if (status < 0)
 462                 RCV_SKB_FAIL(status);
 463
 464         if (flags & NLM_F_ACK)
 465                 netlink_ack(skb, nlh, 0);
 466         return;
 467 }
 468
 469 static void
 470 ipq_rcv_skb(struct sk_buff *skb)
 471 {
 472         mutex_lock(&ipqnl_mutex);
 473         __ipq_rcv_skb(skb);
 474         mutex_unlock(&ipqnl_mutex);
 475 }
 476
 477 static int
 478 ipq_rcv_dev_event(struct notifier_block *this,
 479                   unsigned long event, void *ptr)
 480 {
 481         struct net_device *dev = ptr;
 482
 483         if (!net_eq(dev_net(dev), &init_net))
 484                 return NOTIFY_DONE;
 485
 486         /* Drop any packets associated with the downed device */
 487         if (event == NETDEV_DOWN)
 488                 ipq_dev_drop(dev->ifindex);
 489         return NOTIFY_DONE;
 490 }
 491
 492 static struct notifier_block ipq_dev_notifier = {
 493         .notifier_call  = ipq_rcv_dev_event,
 494 };
 495
 496 static int
 497 ipq_rcv_nl_event(struct notifier_block *this,
 498                  unsigned long event, void *ptr)
 499 {
 500         struct netlink_notify *n = ptr;
 501
 502         if (event == NETLINK_URELEASE &&
 503             n->protocol == NETLINK_IP6_FW && n->pid) {
 504                 write_lock_bh(&queue_lock);
 505                 if ((n->net == &init_net) && (n->pid == peer_pid))
 506                         __ipq_reset();
 507                 write_unlock_bh(&queue_lock);
 508         }
 509         return NOTIFY_DONE;
 510 }
 511
 512 static struct notifier_block ipq_nl_notifier = {
 513         .notifier_call  = ipq_rcv_nl_event,
 514 };
 515
 516 #ifdef CONFIG_SYSCTL
 517 static struct ctl_table_header *ipq_sysctl_header;
 518
 519 static ctl_table ipq_table[] = {
 520         {
 521                 .ctl_name       = NET_IPQ_QMAX,
 522                 .procname       = NET_IPQ_QMAX_NAME,
 523                 .data           = &queue_maxlen,
 524                 .maxlen         = sizeof(queue_maxlen),
 525                 .mode           = 0644,
 526                 .proc_handler   = proc_dointvec
 527         },
 528         { .ctl_name = 0 }
 529 };
 530 #endif
 531
 532 #ifdef CONFIG_PROC_FS
 533 static int ip6_queue_show(struct seq_file *m, void *v)
 534 {
 535         read_lock_bh(&queue_lock);
 536
 537         seq_printf(m,
 538                       "Peer PID          : %d\n"
 539                       "Copy mode         : %hu\n"
 540                       "Copy range        : %u\n"
 541                       "Queue length      : %u\n"
 542                       "Queue max. length : %u\n"
 543                       "Queue dropped     : %u\n"
 544                       "Netfilter dropped : %u\n",
 545                       peer_pid,
 546                       copy_mode,
 547                       copy_range,
 548                       queue_total,
 549                       queue_maxlen,
 550                       queue_dropped,
 551                       queue_user_dropped);
 552
 553         read_unlock_bh(&queue_lock);
 554         return 0;
 555 }
 556
 557 static int ip6_queue_open(struct inode *inode, struct file *file)
 558 {
 559         return single_open(file, ip6_queue_show, NULL);
 560 }
 561
 562 static const struct file_operations ip6_queue_proc_fops = {
 563         .open           = ip6_queue_open,
 564         .read           = seq_read,
 565         .llseek         = seq_lseek,
 566         .release        = single_release,
 567         .owner          = THIS_MODULE,
 568 };
 569 #endif
 570
 571 static const struct nf_queue_handler nfqh = {
 572         .name   = "ip6_queue",
 573         .outfn  = &ipq_enqueue_packet,
 574 };
 575
 576 static int __init ip6_queue_init(void)
 577 {
 578         int status = -ENOMEM;
 579         struct proc_dir_entry *proc __maybe_unused;
 580
 581         netlink_register_notifier(&ipq_nl_notifier);
 582         ipqnl = netlink_kernel_create(&init_net, NETLINK_IP6_FW, 0,
 583                                       ipq_rcv_skb, NULL, THIS_MODULE);
 584         if (ipqnl == NULL) {
 585                 printk(KERN_ERR "ip6_queue: failed to create netlink socket\n");
 586                 goto cleanup_netlink_notifier;
 587         }
 588
 589 #ifdef CONFIG_PROC_FS
 590         proc = proc_create(IPQ_PROC_FS_NAME, 0, init_net.proc_net,
 591                            &ip6_queue_proc_fops);
 592         if (!proc) {
 593                 printk(KERN_ERR "ip6_queue: failed to create proc entry\n");
 594                 goto cleanup_ipqnl;
 595         }
 596 #endif
 597         register_netdevice_notifier(&ipq_dev_notifier);
 598 #ifdef CONFIG_SYSCTL
 599         ipq_sysctl_header = register_sysctl_paths(net_ipv6_ctl_path, ipq_table);
 600 #endif
 601         status = nf_register_queue_handler(NFPROTO_IPV6, &nfqh);
 602         if (status < 0) {
 603                 printk(KERN_ERR "ip6_queue: failed to register queue handler\n");
 604                 goto cleanup_sysctl;
 605         }
 606         return status;
 607
 608 cleanup_sysctl:
 609 #ifdef CONFIG_SYSCTL
 610         unregister_sysctl_table(ipq_sysctl_header);
 611 #endif
 612         unregister_netdevice_notifier(&ipq_dev_notifier);
 613         proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
 614
 615 cleanup_ipqnl: __maybe_unused
 616         netlink_kernel_release(ipqnl);
 617         mutex_lock(&ipqnl_mutex);
 618         mutex_unlock(&ipqnl_mutex);
 619
 620 cleanup_netlink_notifier:
 621         netlink_unregister_notifier(&ipq_nl_notifier);
 622         return status;
 623 }
 624
 625 static void __exit ip6_queue_fini(void)
 626 {
 627         nf_unregister_queue_handlers(&nfqh);
 628         synchronize_net();
 629         ipq_flush(NULL, 0);
 630
 631 #ifdef CONFIG_SYSCTL
 632         unregister_sysctl_table(ipq_sysctl_header);
 633 #endif
 634         unregister_netdevice_notifier(&ipq_dev_notifier);
 635         proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
 636
 637         netlink_kernel_release(ipqnl);
 638         mutex_lock(&ipqnl_mutex);
 639         mutex_unlock(&ipqnl_mutex);
 640
 641         netlink_unregister_notifier(&ipq_nl_notifier);
 642 }
 643
 644 MODULE_DESCRIPTION("IPv6 packet queue handler");
 645 MODULE_LICENSE("GPL");
 646 MODULE_ALIAS_NET_PF_PROTO(PF_NETLINK, NETLINK_IP6_FW);
 647
 648 module_init(ip6_queue_init);
 649 module_exit(ip6_queue_fini);