err.no Git - linux-2.6/blob - 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 void
  57 ipq_issue_verdict(struct nf_queue_entry *entry, int verdict)
  58 {
  59         local_bh_disable();
  60         nf_reinject(entry, verdict);
  61         local_bh_enable();
  62 }
  63
  64 static inline void
  65 __ipq_enqueue_entry(struct nf_queue_entry *entry)
  66 {
  67        list_add_tail(&entry->list, &queue_list);
  68        queue_total++;
  69 }
  70
  71 static inline int
  72 __ipq_set_mode(unsigned char mode, unsigned int range)
  73 {
  74         int status = 0;
  75
  76         switch(mode) {
  77         case IPQ_COPY_NONE:
  78         case IPQ_COPY_META:
  79                 copy_mode = mode;
  80                 copy_range = 0;
  81                 break;
  82
  83         case IPQ_COPY_PACKET:
  84                 copy_mode = mode;
  85                 copy_range = range;
  86                 if (copy_range > 0xFFFF)
  87                         copy_range = 0xFFFF;
  88                 break;
  89
  90         default:
  91                 status = -EINVAL;
  92
  93         }
  94         return status;
  95 }
  96
  97 static void __ipq_flush(ipq_cmpfn cmpfn, unsigned long data);
  98
  99 static inline void
 100 __ipq_reset(void)
 101 {
 102         peer_pid = 0;
 103         net_disable_timestamp();
 104         __ipq_set_mode(IPQ_COPY_NONE, 0);
 105         __ipq_flush(NULL, 0);
 106 }
 107
 108 static struct nf_queue_entry *
 109 ipq_find_dequeue_entry(unsigned long id)
 110 {
 111         struct nf_queue_entry *entry = NULL, *i;
 112
 113         write_lock_bh(&queue_lock);
 114
 115         list_for_each_entry(i, &queue_list, list) {
 116                 if ((unsigned long)i == id) {
 117                         entry = i;
 118                         break;
 119                 }
 120         }
 121
 122         if (entry) {
 123                 list_del(&entry->list);
 124                 queue_total--;
 125         }
 126
 127         write_unlock_bh(&queue_lock);
 128         return entry;
 129 }
 130
 131 static void
 132 __ipq_flush(ipq_cmpfn cmpfn, unsigned long data)
 133 {
 134         struct nf_queue_entry *entry, *next;
 135
 136         list_for_each_entry_safe(entry, next, &queue_list, list) {
 137                 if (!cmpfn || cmpfn(entry, data)) {
 138                         list_del(&entry->list);
 139                         queue_total--;
 140                         ipq_issue_verdict(entry, NF_DROP);
 141                 }
 142         }
 143 }
 144
 145 static void
 146 ipq_flush(ipq_cmpfn cmpfn, unsigned long data)
 147 {
 148         write_lock_bh(&queue_lock);
 149         __ipq_flush(cmpfn, data);
 150         write_unlock_bh(&queue_lock);
 151 }
 152
 153 static struct sk_buff *
 154 ipq_build_packet_message(struct nf_queue_entry *entry, int *errp)
 155 {
 156         sk_buff_data_t old_tail;
 157         size_t size = 0;
 158         size_t data_len = 0;
 159         struct sk_buff *skb;
 160         struct ipq_packet_msg *pmsg;
 161         struct nlmsghdr *nlh;
 162         struct timeval tv;
 163
 164         read_lock_bh(&queue_lock);
 165
 166         switch (copy_mode) {
 167         case IPQ_COPY_META:
 168         case IPQ_COPY_NONE:
 169                 size = NLMSG_SPACE(sizeof(*pmsg));
 170                 data_len = 0;
 171                 break;
 172
 173         case IPQ_COPY_PACKET:
 174                 if ((entry->skb->ip_summed == CHECKSUM_PARTIAL ||
 175                      entry->skb->ip_summed == CHECKSUM_COMPLETE) &&
 176                     (*errp = skb_checksum_help(entry->skb))) {
 177                         read_unlock_bh(&queue_lock);
 178                         return NULL;
 179                 }
 180                 if (copy_range == 0 || copy_range > entry->skb->len)
 181                         data_len = entry->skb->len;
 182                 else
 183                         data_len = copy_range;
 184
 185                 size = NLMSG_SPACE(sizeof(*pmsg) + data_len);
 186                 break;
 187
 188         default:
 189                 *errp = -EINVAL;
 190                 read_unlock_bh(&queue_lock);
 191                 return NULL;
 192         }
 193
 194         read_unlock_bh(&queue_lock);
 195
 196         skb = alloc_skb(size, GFP_ATOMIC);
 197         if (!skb)
 198                 goto nlmsg_failure;
 199
 200         old_tail = skb->tail;
 201         nlh = NLMSG_PUT(skb, 0, 0, IPQM_PACKET, size - sizeof(*nlh));
 202         pmsg = NLMSG_DATA(nlh);
 203         memset(pmsg, 0, sizeof(*pmsg));
 204
 205         pmsg->packet_id       = (unsigned long )entry;
 206         pmsg->data_len        = data_len;
 207         tv = ktime_to_timeval(entry->skb->tstamp);
 208         pmsg->timestamp_sec   = tv.tv_sec;
 209         pmsg->timestamp_usec  = tv.tv_usec;
 210         pmsg->mark            = entry->skb->mark;
 211         pmsg->hook            = entry->hook;
 212         pmsg->hw_protocol     = entry->skb->protocol;
 213
 214         if (entry->indev)
 215                 strcpy(pmsg->indev_name, entry->indev->name);
 216         else
 217                 pmsg->indev_name[0] = '\0';
 218
 219         if (entry->outdev)
 220                 strcpy(pmsg->outdev_name, entry->outdev->name);
 221         else
 222                 pmsg->outdev_name[0] = '\0';
 223
 224         if (entry->indev && entry->skb->dev) {
 225                 pmsg->hw_type = entry->skb->dev->type;
 226                 pmsg->hw_addrlen = dev_parse_header(entry->skb, pmsg->hw_addr);
 227         }
 228
 229         if (data_len)
 230                 if (skb_copy_bits(entry->skb, 0, pmsg->payload, data_len))
 231                         BUG();
 232
 233         nlh->nlmsg_len = skb->tail - old_tail;
 234         return skb;
 235
 236 nlmsg_failure:
 237         if (skb)
 238                 kfree_skb(skb);
 239         *errp = -EINVAL;
 240         printk(KERN_ERR "ip6_queue: error creating packet message\n");
 241         return NULL;
 242 }
 243
 244 static int
 245 ipq_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 246 {
 247         int status = -EINVAL;
 248         struct sk_buff *nskb;
 249
 250         if (copy_mode == IPQ_COPY_NONE)
 251                 return -EAGAIN;
 252
 253         nskb = ipq_build_packet_message(entry, &status);
 254         if (nskb == NULL)
 255                 return status;
 256
 257         write_lock_bh(&queue_lock);
 258
 259         if (!peer_pid)
 260                 goto err_out_free_nskb;
 261
 262         if (queue_total >= queue_maxlen) {
 263                 queue_dropped++;
 264                 status = -ENOSPC;
 265                 if (net_ratelimit())
 266                         printk (KERN_WARNING "ip6_queue: fill at %d entries, "
 267                                 "dropping packet(s).  Dropped: %d\n", queue_total,
 268                                 queue_dropped);
 269                 goto err_out_free_nskb;
 270         }
 271
 272         /* netlink_unicast will either free the nskb or attach it to a socket */
 273         status = netlink_unicast(ipqnl, nskb, peer_pid, MSG_DONTWAIT);
 274         if (status < 0) {
 275                 queue_user_dropped++;
 276                 goto err_out_unlock;
 277         }
 278
 279         __ipq_enqueue_entry(entry);
 280
 281         write_unlock_bh(&queue_lock);
 282         return status;
 283
 284 err_out_free_nskb:
 285         kfree_skb(nskb);
 286
 287 err_out_unlock:
 288         write_unlock_bh(&queue_lock);
 289         return status;
 290 }
 291
 292 static int
 293 ipq_mangle_ipv6(ipq_verdict_msg_t *v, struct nf_queue_entry *e)
 294 {
 295         int diff;
 296         int err;
 297         struct ipv6hdr *user_iph = (struct ipv6hdr *)v->payload;
 298
 299         if (v->data_len < sizeof(*user_iph))
 300                 return 0;
 301         diff = v->data_len - e->skb->len;
 302         if (diff < 0) {
 303                 if (pskb_trim(e->skb, v->data_len))
 304                         return -ENOMEM;
 305         } else if (diff > 0) {
 306                 if (v->data_len > 0xFFFF)
 307                         return -EINVAL;
 308                 if (diff > skb_tailroom(e->skb)) {
 309                         err = pskb_expand_head(e->skb, 0,
 310                                                diff - skb_tailroom(e->skb),
 311                                                GFP_ATOMIC);
 312                         if (err) {
 313                                 printk(KERN_WARNING "ip6_queue: OOM "
 314                                       "in mangle, dropping packet\n");
 315                                 return err;
 316                         }
 317                 }
 318                 skb_put(e->skb, diff);
 319         }
 320         if (!skb_make_writable(e->skb, v->data_len))
 321                 return -ENOMEM;
 322         skb_copy_to_linear_data(e->skb, v->payload, v->data_len);
 323         e->skb->ip_summed = CHECKSUM_NONE;
 324
 325         return 0;
 326 }
 327
 328 static int
 329 ipq_set_verdict(struct ipq_verdict_msg *vmsg, unsigned int len)
 330 {
 331         struct nf_queue_entry *entry;
 332
 333         if (vmsg->value > NF_MAX_VERDICT)
 334                 return -EINVAL;
 335
 336         entry = ipq_find_dequeue_entry(vmsg->id);
 337         if (entry == NULL)
 338                 return -ENOENT;
 339         else {
 340                 int verdict = vmsg->value;
 341
 342                 if (vmsg->data_len && vmsg->data_len == len)
 343                         if (ipq_mangle_ipv6(vmsg, entry) < 0)
 344                                 verdict = NF_DROP;
 345
 346                 ipq_issue_verdict(entry, verdict);
 347                 return 0;
 348         }
 349 }
 350
 351 static int
 352 ipq_set_mode(unsigned char mode, unsigned int range)
 353 {
 354         int status;
 355
 356         write_lock_bh(&queue_lock);
 357         status = __ipq_set_mode(mode, range);
 358         write_unlock_bh(&queue_lock);
 359         return status;
 360 }
 361
 362 static int
 363 ipq_receive_peer(struct ipq_peer_msg *pmsg,
 364                  unsigned char type, unsigned int len)
 365 {
 366         int status = 0;
 367
 368         if (len < sizeof(*pmsg))
 369                 return -EINVAL;
 370
 371         switch (type) {
 372         case IPQM_MODE:
 373                 status = ipq_set_mode(pmsg->msg.mode.value,
 374                                       pmsg->msg.mode.range);
 375                 break;
 376
 377         case IPQM_VERDICT:
 378                 if (pmsg->msg.verdict.value > NF_MAX_VERDICT)
 379                         status = -EINVAL;
 380                 else
 381                         status = ipq_set_verdict(&pmsg->msg.verdict,
 382                                                  len - sizeof(*pmsg));
 383                         break;
 384         default:
 385                 status = -EINVAL;
 386         }
 387         return status;
 388 }
 389
 390 static int
 391 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
 392 {
 393         if (entry->indev)
 394                 if (entry->indev->ifindex == ifindex)
 395                         return 1;
 396
 397         if (entry->outdev)
 398                 if (entry->outdev->ifindex == ifindex)
 399                         return 1;
 400 #ifdef CONFIG_BRIDGE_NETFILTER
 401         if (entry->skb->nf_bridge) {
 402                 if (entry->skb->nf_bridge->physindev &&
 403                     entry->skb->nf_bridge->physindev->ifindex == ifindex)
 404                         return 1;
 405                 if (entry->skb->nf_bridge->physoutdev &&
 406                     entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
 407                         return 1;
 408         }
 409 #endif
 410         return 0;
 411 }
 412
 413 static void
 414 ipq_dev_drop(int ifindex)
 415 {
 416         ipq_flush(dev_cmp, ifindex);
 417 }
 418
 419 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
 420
 421 static inline void
 422 __ipq_rcv_skb(struct sk_buff *skb)
 423 {
 424         int status, type, pid, flags, nlmsglen, skblen;
 425         struct nlmsghdr *nlh;
 426
 427         skblen = skb->len;
 428         if (skblen < sizeof(*nlh))
 429                 return;
 430
 431         nlh = nlmsg_hdr(skb);
 432         nlmsglen = nlh->nlmsg_len;
 433         if (nlmsglen < sizeof(*nlh) || skblen < nlmsglen)
 434                 return;
 435
 436         pid = nlh->nlmsg_pid;
 437         flags = nlh->nlmsg_flags;
 438
 439         if(pid <= 0 || !(flags & NLM_F_REQUEST) || flags & NLM_F_MULTI)
 440                 RCV_SKB_FAIL(-EINVAL);
 441
 442         if (flags & MSG_TRUNC)
 443                 RCV_SKB_FAIL(-ECOMM);
 444
 445         type = nlh->nlmsg_type;
 446         if (type < NLMSG_NOOP || type >= IPQM_MAX)
 447                 RCV_SKB_FAIL(-EINVAL);
 448
 449         if (type <= IPQM_BASE)
 450                 return;
 451
 452         if (security_netlink_recv(skb, CAP_NET_ADMIN))
 453                 RCV_SKB_FAIL(-EPERM);
 454
 455         write_lock_bh(&queue_lock);
 456
 457         if (peer_pid) {
 458                 if (peer_pid != pid) {
 459                         write_unlock_bh(&queue_lock);
 460                         RCV_SKB_FAIL(-EBUSY);
 461                 }
 462         } else {
 463                 net_enable_timestamp();
 464                 peer_pid = pid;
 465         }
 466
 467         write_unlock_bh(&queue_lock);
 468
 469         status = ipq_receive_peer(NLMSG_DATA(nlh), type,
 470                                   nlmsglen - NLMSG_LENGTH(0));
 471         if (status < 0)
 472                 RCV_SKB_FAIL(status);
 473
 474         if (flags & NLM_F_ACK)
 475                 netlink_ack(skb, nlh, 0);
 476         return;
 477 }
 478
 479 static void
 480 ipq_rcv_skb(struct sk_buff *skb)
 481 {
 482         mutex_lock(&ipqnl_mutex);
 483         __ipq_rcv_skb(skb);
 484         mutex_unlock(&ipqnl_mutex);
 485 }
 486
 487 static int
 488 ipq_rcv_dev_event(struct notifier_block *this,
 489                   unsigned long event, void *ptr)
 490 {
 491         struct net_device *dev = ptr;
 492
 493         if (dev->nd_net != &init_net)
 494                 return NOTIFY_DONE;
 495
 496         /* Drop any packets associated with the downed device */
 497         if (event == NETDEV_DOWN)
 498                 ipq_dev_drop(dev->ifindex);
 499         return NOTIFY_DONE;
 500 }
 501
 502 static struct notifier_block ipq_dev_notifier = {
 503         .notifier_call  = ipq_rcv_dev_event,
 504 };
 505
 506 static int
 507 ipq_rcv_nl_event(struct notifier_block *this,
 508                  unsigned long event, void *ptr)
 509 {
 510         struct netlink_notify *n = ptr;
 511
 512         if (event == NETLINK_URELEASE &&
 513             n->protocol == NETLINK_IP6_FW && n->pid) {
 514                 write_lock_bh(&queue_lock);
 515                 if ((n->net == &init_net) && (n->pid == peer_pid))
 516                         __ipq_reset();
 517                 write_unlock_bh(&queue_lock);
 518         }
 519         return NOTIFY_DONE;
 520 }
 521
 522 static struct notifier_block ipq_nl_notifier = {
 523         .notifier_call  = ipq_rcv_nl_event,
 524 };
 525
 526 static struct ctl_table_header *ipq_sysctl_header;
 527
 528 static ctl_table ipq_table[] = {
 529         {
 530                 .ctl_name       = NET_IPQ_QMAX,
 531                 .procname       = NET_IPQ_QMAX_NAME,
 532                 .data           = &queue_maxlen,
 533                 .maxlen         = sizeof(queue_maxlen),
 534                 .mode           = 0644,
 535                 .proc_handler   = proc_dointvec
 536         },
 537         { .ctl_name = 0 }
 538 };
 539
 540 static ctl_table ipq_dir_table[] = {
 541         {
 542                 .ctl_name       = NET_IPV6,
 543                 .procname       = "ipv6",
 544                 .mode           = 0555,
 545                 .child          = ipq_table
 546         },
 547         { .ctl_name = 0 }
 548 };
 549
 550 static ctl_table ipq_root_table[] = {
 551         {
 552                 .ctl_name       = CTL_NET,
 553                 .procname       = "net",
 554                 .mode           = 0555,
 555                 .child          = ipq_dir_table
 556         },
 557         { .ctl_name = 0 }
 558 };
 559
 560 static int ip6_queue_show(struct seq_file *m, void *v)
 561 {
 562         read_lock_bh(&queue_lock);
 563
 564         seq_printf(m,
 565                       "Peer PID          : %d\n"
 566                       "Copy mode         : %hu\n"
 567                       "Copy range        : %u\n"
 568                       "Queue length      : %u\n"
 569                       "Queue max. length : %u\n"
 570                       "Queue dropped     : %u\n"
 571                       "Netfilter dropped : %u\n",
 572                       peer_pid,
 573                       copy_mode,
 574                       copy_range,
 575                       queue_total,
 576                       queue_maxlen,
 577                       queue_dropped,
 578                       queue_user_dropped);
 579
 580         read_unlock_bh(&queue_lock);
 581         return 0;
 582 }
 583
 584 static int ip6_queue_open(struct inode *inode, struct file *file)
 585 {
 586         return single_open(file, ip6_queue_show, NULL);
 587 }
 588
 589 static const struct file_operations ip6_queue_proc_fops = {
 590         .open           = ip6_queue_open,
 591         .read           = seq_read,
 592         .llseek         = seq_lseek,
 593         .release        = single_release,
 594         .owner          = THIS_MODULE,
 595 };
 596
 597 static const struct nf_queue_handler nfqh = {
 598         .name   = "ip6_queue",
 599         .outfn  = &ipq_enqueue_packet,
 600 };
 601
 602 static int __init ip6_queue_init(void)
 603 {
 604         int status = -ENOMEM;
 605         struct proc_dir_entry *proc;
 606
 607         netlink_register_notifier(&ipq_nl_notifier);
 608         ipqnl = netlink_kernel_create(&init_net, NETLINK_IP6_FW, 0,
 609                                       ipq_rcv_skb, NULL, THIS_MODULE);
 610         if (ipqnl == NULL) {
 611                 printk(KERN_ERR "ip6_queue: failed to create netlink socket\n");
 612                 goto cleanup_netlink_notifier;
 613         }
 614
 615         proc = create_proc_entry(IPQ_PROC_FS_NAME, 0, init_net.proc_net);
 616         if (proc) {
 617                 proc->owner = THIS_MODULE;
 618                 proc->proc_fops = &ip6_queue_proc_fops;
 619         } else {
 620                 printk(KERN_ERR "ip6_queue: failed to create proc entry\n");
 621                 goto cleanup_ipqnl;
 622         }
 623
 624         register_netdevice_notifier(&ipq_dev_notifier);
 625         ipq_sysctl_header = register_sysctl_table(ipq_root_table);
 626
 627         status = nf_register_queue_handler(PF_INET6, &nfqh);
 628         if (status < 0) {
 629                 printk(KERN_ERR "ip6_queue: failed to register queue handler\n");
 630                 goto cleanup_sysctl;
 631         }
 632         return status;
 633
 634 cleanup_sysctl:
 635         unregister_sysctl_table(ipq_sysctl_header);
 636         unregister_netdevice_notifier(&ipq_dev_notifier);
 637         proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
 638
 639 cleanup_ipqnl:
 640         sock_release(ipqnl->sk_socket);
 641         mutex_lock(&ipqnl_mutex);
 642         mutex_unlock(&ipqnl_mutex);
 643
 644 cleanup_netlink_notifier:
 645         netlink_unregister_notifier(&ipq_nl_notifier);
 646         return status;
 647 }
 648
 649 static void __exit ip6_queue_fini(void)
 650 {
 651         nf_unregister_queue_handlers(&nfqh);
 652         synchronize_net();
 653         ipq_flush(NULL, 0);
 654
 655         unregister_sysctl_table(ipq_sysctl_header);
 656         unregister_netdevice_notifier(&ipq_dev_notifier);
 657         proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
 658
 659         sock_release(ipqnl->sk_socket);
 660         mutex_lock(&ipqnl_mutex);
 661         mutex_unlock(&ipqnl_mutex);
 662
 663         netlink_unregister_notifier(&ipq_nl_notifier);
 664 }
 665
 666 MODULE_DESCRIPTION("IPv6 packet queue handler");
 667 MODULE_LICENSE("GPL");
 668
 669 module_init(ip6_queue_init);
 670 module_exit(ip6_queue_fini);