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 <linux/slab.h>
  29 #include <net/net_namespace.h>
  30 #include <net/sock.h>
  31 #include <net/ipv6.h>
  32 #include <net/ip6_route.h>
  33 #include <net/netfilter/nf_queue.h>
  34 #include <linux/netfilter_ipv4/ip_queue.h>
  35 #include <linux/netfilter_ipv4/ip_tables.h>
  36 #include <linux/netfilter_ipv6/ip6_tables.h>
  37
  38 #define IPQ_QMAX_DEFAULT 1024
  39 #define IPQ_PROC_FS_NAME "ip6_queue"
  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_SPINLOCK(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                 if (range > 0xFFFF)
  77                         range = 0xFFFF;
  78                 copy_range = range;
  79                 copy_mode = mode;
  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         spin_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         spin_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         spin_lock_bh(&queue_lock);
 141         __ipq_flush(cmpfn, data);
 142         spin_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         switch (ACCESS_ONCE(copy_mode)) {
 157         case IPQ_COPY_META:
 158         case IPQ_COPY_NONE:
 159                 size = NLMSG_SPACE(sizeof(*pmsg));
 160                 break;
 161
 162         case IPQ_COPY_PACKET:
 163                 if (entry->skb->ip_summed == CHECKSUM_PARTIAL &&
 164                     (*errp = skb_checksum_help(entry->skb)))
 165                         return NULL;
 166
 167                 data_len = ACCESS_ONCE(copy_range);
 168                 if (data_len == 0 || data_len > entry->skb->len)
 169                         data_len = entry->skb->len;
 170
 171                 size = NLMSG_SPACE(sizeof(*pmsg) + data_len);
 172                 break;
 173
 174         default:
 175                 *errp = -EINVAL;
 176                 return NULL;
 177         }
 178
 179         skb = alloc_skb(size, GFP_ATOMIC);
 180         if (!skb)
 181                 goto nlmsg_failure;
 182
 183         old_tail = skb->tail;
 184         nlh = NLMSG_PUT(skb, 0, 0, IPQM_PACKET, size - sizeof(*nlh));
 185         pmsg = NLMSG_DATA(nlh);
 186         memset(pmsg, 0, sizeof(*pmsg));
 187
 188         pmsg->packet_id       = (unsigned long )entry;
 189         pmsg->data_len        = data_len;
 190         tv = ktime_to_timeval(entry->skb->tstamp);
 191         pmsg->timestamp_sec   = tv.tv_sec;
 192         pmsg->timestamp_usec  = tv.tv_usec;
 193         pmsg->mark            = entry->skb->mark;
 194         pmsg->hook            = entry->hook;
 195         pmsg->hw_protocol     = entry->skb->protocol;
 196
 197         if (entry->indev)
 198                 strcpy(pmsg->indev_name, entry->indev->name);
 199         else
 200                 pmsg->indev_name[0] = '\0';
 201
 202         if (entry->outdev)
 203                 strcpy(pmsg->outdev_name, entry->outdev->name);
 204         else
 205                 pmsg->outdev_name[0] = '\0';
 206
 207         if (entry->indev && entry->skb->dev &&
 208             entry->skb->mac_header != entry->skb->network_header) {
 209                 pmsg->hw_type = entry->skb->dev->type;
 210                 pmsg->hw_addrlen = dev_parse_header(entry->skb, pmsg->hw_addr);
 211         }
 212
 213         if (data_len)
 214                 if (skb_copy_bits(entry->skb, 0, pmsg->payload, data_len))
 215                         BUG();
 216
 217         nlh->nlmsg_len = skb->tail - old_tail;
 218         return skb;
 219
 220 nlmsg_failure:
 221         *errp = -EINVAL;
 222         printk(KERN_ERR "ip6_queue: error creating packet message\n");
 223         return NULL;
 224 }
 225
 226 static int
 227 ipq_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 228 {
 229         int status = -EINVAL;
 230         struct sk_buff *nskb;
 231
 232         if (copy_mode == IPQ_COPY_NONE)
 233                 return -EAGAIN;
 234
 235         nskb = ipq_build_packet_message(entry, &status);
 236         if (nskb == NULL)
 237                 return status;
 238
 239         spin_lock_bh(&queue_lock);
 240
 241         if (!peer_pid)
 242                 goto err_out_free_nskb;
 243
 244         if (queue_total >= queue_maxlen) {
 245                 queue_dropped++;
 246                 status = -ENOSPC;
 247                 if (net_ratelimit())
 248                         printk (KERN_WARNING "ip6_queue: fill at %d entries, "
 249                                 "dropping packet(s).  Dropped: %d\n", queue_total,
 250                                 queue_dropped);
 251                 goto err_out_free_nskb;
 252         }
 253
 254         /* netlink_unicast will either free the nskb or attach it to a socket */
 255         status = netlink_unicast(ipqnl, nskb, peer_pid, MSG_DONTWAIT);
 256         if (status < 0) {
 257                 queue_user_dropped++;
 258                 goto err_out_unlock;
 259         }
 260
 261         __ipq_enqueue_entry(entry);
 262
 263         spin_unlock_bh(&queue_lock);
 264         return status;
 265
 266 err_out_free_nskb:
 267         kfree_skb(nskb);
 268
 269 err_out_unlock:
 270         spin_unlock_bh(&queue_lock);
 271         return status;
 272 }
 273
 274 static int
 275 ipq_mangle_ipv6(ipq_verdict_msg_t *v, struct nf_queue_entry *e)
 276 {
 277         int diff;
 278         struct ipv6hdr *user_iph = (struct ipv6hdr *)v->payload;
 279         struct sk_buff *nskb;
 280
 281         if (v->data_len < sizeof(*user_iph))
 282                 return 0;
 283         diff = v->data_len - e->skb->len;
 284         if (diff < 0) {
 285                 if (pskb_trim(e->skb, v->data_len))
 286                         return -ENOMEM;
 287         } else if (diff > 0) {
 288                 if (v->data_len > 0xFFFF)
 289                         return -EINVAL;
 290                 if (diff > skb_tailroom(e->skb)) {
 291                         nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
 292                                                diff, GFP_ATOMIC);
 293                         if (!nskb) {
 294                                 printk(KERN_WARNING "ip6_queue: OOM "
 295                                       "in mangle, dropping packet\n");
 296                                 return -ENOMEM;
 297                         }
 298                         kfree_skb(e->skb);
 299                         e->skb = nskb;
 300                 }
 301                 skb_put(e->skb, diff);
 302         }
 303         if (!skb_make_writable(e->skb, v->data_len))
 304                 return -ENOMEM;
 305         skb_copy_to_linear_data(e->skb, v->payload, v->data_len);
 306         e->skb->ip_summed = CHECKSUM_NONE;
 307
 308         return 0;
 309 }
 310
 311 static int
 312 ipq_set_verdict(struct ipq_verdict_msg *vmsg, unsigned int len)
 313 {
 314         struct nf_queue_entry *entry;
 315
 316         if (vmsg->value > NF_MAX_VERDICT)
 317                 return -EINVAL;
 318
 319         entry = ipq_find_dequeue_entry(vmsg->id);
 320         if (entry == NULL)
 321                 return -ENOENT;
 322         else {
 323                 int verdict = vmsg->value;
 324
 325                 if (vmsg->data_len && vmsg->data_len == len)
 326                         if (ipq_mangle_ipv6(vmsg, entry) < 0)
 327                                 verdict = NF_DROP;
 328
 329                 nf_reinject(entry, verdict);
 330                 return 0;
 331         }
 332 }
 333
 334 static int
 335 ipq_set_mode(unsigned char mode, unsigned int range)
 336 {
 337         int status;
 338
 339         spin_lock_bh(&queue_lock);
 340         status = __ipq_set_mode(mode, range);
 341         spin_unlock_bh(&queue_lock);
 342         return status;
 343 }
 344
 345 static int
 346 ipq_receive_peer(struct ipq_peer_msg *pmsg,
 347                  unsigned char type, unsigned int len)
 348 {
 349         int status = 0;
 350
 351         if (len < sizeof(*pmsg))
 352                 return -EINVAL;
 353
 354         switch (type) {
 355         case IPQM_MODE:
 356                 status = ipq_set_mode(pmsg->msg.mode.value,
 357                                       pmsg->msg.mode.range);
 358                 break;
 359
 360         case IPQM_VERDICT:
 361                 if (pmsg->msg.verdict.value > NF_MAX_VERDICT)
 362                         status = -EINVAL;
 363                 else
 364                         status = ipq_set_verdict(&pmsg->msg.verdict,
 365                                                  len - sizeof(*pmsg));
 366                         break;
 367         default:
 368                 status = -EINVAL;
 369         }
 370         return status;
 371 }
 372
 373 static int
 374 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
 375 {
 376         if (entry->indev)
 377                 if (entry->indev->ifindex == ifindex)
 378                         return 1;
 379
 380         if (entry->outdev)
 381                 if (entry->outdev->ifindex == ifindex)
 382                         return 1;
 383 #ifdef CONFIG_BRIDGE_NETFILTER
 384         if (entry->skb->nf_bridge) {
 385                 if (entry->skb->nf_bridge->physindev &&
 386                     entry->skb->nf_bridge->physindev->ifindex == ifindex)
 387                         return 1;
 388                 if (entry->skb->nf_bridge->physoutdev &&
 389                     entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
 390                         return 1;
 391         }
 392 #endif
 393         return 0;
 394 }
 395
 396 static void
 397 ipq_dev_drop(int ifindex)
 398 {
 399         ipq_flush(dev_cmp, ifindex);
 400 }
 401
 402 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
 403
 404 static inline void
 405 __ipq_rcv_skb(struct sk_buff *skb)
 406 {
 407         int status, type, pid, flags;
 408         unsigned int nlmsglen, skblen;
 409         struct nlmsghdr *nlh;
 410
 411         skblen = skb->len;
 412         if (skblen < sizeof(*nlh))
 413                 return;
 414
 415         nlh = nlmsg_hdr(skb);
 416         nlmsglen = nlh->nlmsg_len;
 417         if (nlmsglen < sizeof(*nlh) || skblen < nlmsglen)
 418                 return;
 419
 420         pid = nlh->nlmsg_pid;
 421         flags = nlh->nlmsg_flags;
 422
 423         if(pid <= 0 || !(flags & NLM_F_REQUEST) || flags & NLM_F_MULTI)
 424                 RCV_SKB_FAIL(-EINVAL);
 425
 426         if (flags & MSG_TRUNC)
 427                 RCV_SKB_FAIL(-ECOMM);
 428
 429         type = nlh->nlmsg_type;
 430         if (type < NLMSG_NOOP || type >= IPQM_MAX)
 431                 RCV_SKB_FAIL(-EINVAL);
 432
 433         if (type <= IPQM_BASE)
 434                 return;
 435
 436         if (security_netlink_recv(skb, CAP_NET_ADMIN))
 437                 RCV_SKB_FAIL(-EPERM);
 438
 439         spin_lock_bh(&queue_lock);
 440
 441         if (peer_pid) {
 442                 if (peer_pid != pid) {
 443                         spin_unlock_bh(&queue_lock);
 444                         RCV_SKB_FAIL(-EBUSY);
 445                 }
 446         } else {
 447                 net_enable_timestamp();
 448                 peer_pid = pid;
 449         }
 450
 451         spin_unlock_bh(&queue_lock);
 452
 453         status = ipq_receive_peer(NLMSG_DATA(nlh), type,
 454                                   nlmsglen - NLMSG_LENGTH(0));
 455         if (status < 0)
 456                 RCV_SKB_FAIL(status);
 457
 458         if (flags & NLM_F_ACK)
 459                 netlink_ack(skb, nlh, 0);
 460 }
 461
 462 static void
 463 ipq_rcv_skb(struct sk_buff *skb)
 464 {
 465         mutex_lock(&ipqnl_mutex);
 466         __ipq_rcv_skb(skb);
 467         mutex_unlock(&ipqnl_mutex);
 468 }
 469
 470 static int
 471 ipq_rcv_dev_event(struct notifier_block *this,
 472                   unsigned long event, void *ptr)
 473 {
 474         struct net_device *dev = ptr;
 475
 476         if (!net_eq(dev_net(dev), &init_net))
 477                 return NOTIFY_DONE;
 478
 479         /* Drop any packets associated with the downed device */
 480         if (event == NETDEV_DOWN)
 481                 ipq_dev_drop(dev->ifindex);
 482         return NOTIFY_DONE;
 483 }
 484
 485 static struct notifier_block ipq_dev_notifier = {
 486         .notifier_call  = ipq_rcv_dev_event,
 487 };
 488
 489 static int
 490 ipq_rcv_nl_event(struct notifier_block *this,
 491                  unsigned long event, void *ptr)
 492 {
 493         struct netlink_notify *n = ptr;
 494
 495         if (event == NETLINK_URELEASE && n->protocol == NETLINK_IP6_FW) {
 496                 spin_lock_bh(&queue_lock);
 497                 if ((net_eq(n->net, &init_net)) && (n->pid == peer_pid))
 498                         __ipq_reset();
 499                 spin_unlock_bh(&queue_lock);
 500         }
 501         return NOTIFY_DONE;
 502 }
 503
 504 static struct notifier_block ipq_nl_notifier = {
 505         .notifier_call  = ipq_rcv_nl_event,
 506 };
 507
 508 #ifdef CONFIG_SYSCTL
 509 static struct ctl_table_header *ipq_sysctl_header;
 510
 511 static ctl_table ipq_table[] = {
 512         {
 513                 .procname       = NET_IPQ_QMAX_NAME,
 514                 .data           = &queue_maxlen,
 515                 .maxlen         = sizeof(queue_maxlen),
 516                 .mode           = 0644,
 517                 .proc_handler   = proc_dointvec
 518         },
 519         { }
 520 };
 521 #endif
 522
 523 #ifdef CONFIG_PROC_FS
 524 static int ip6_queue_show(struct seq_file *m, void *v)
 525 {
 526         spin_lock_bh(&queue_lock);
 527
 528         seq_printf(m,
 529                       "Peer PID          : %d\n"
 530                       "Copy mode         : %hu\n"
 531                       "Copy range        : %u\n"
 532                       "Queue length      : %u\n"
 533                       "Queue max. length : %u\n"
 534                       "Queue dropped     : %u\n"
 535                       "Netfilter dropped : %u\n",
 536                       peer_pid,
 537                       copy_mode,
 538                       copy_range,
 539                       queue_total,
 540                       queue_maxlen,
 541                       queue_dropped,
 542                       queue_user_dropped);
 543
 544         spin_unlock_bh(&queue_lock);
 545         return 0;
 546 }
 547
 548 static int ip6_queue_open(struct inode *inode, struct file *file)
 549 {
 550         return single_open(file, ip6_queue_show, NULL);
 551 }
 552
 553 static const struct file_operations ip6_queue_proc_fops = {
 554         .open           = ip6_queue_open,
 555         .read           = seq_read,
 556         .llseek         = seq_lseek,
 557         .release        = single_release,
 558         .owner          = THIS_MODULE,
 559 };
 560 #endif
 561
 562 static const struct nf_queue_handler nfqh = {
 563         .name   = "ip6_queue",
 564         .outfn  = &ipq_enqueue_packet,
 565 };
 566
 567 static int __init ip6_queue_init(void)
 568 {
 569         int status = -ENOMEM;
 570         struct proc_dir_entry *proc __maybe_unused;
 571
 572         netlink_register_notifier(&ipq_nl_notifier);
 573         ipqnl = netlink_kernel_create(&init_net, NETLINK_IP6_FW, 0,
 574                                       ipq_rcv_skb, NULL, THIS_MODULE);
 575         if (ipqnl == NULL) {
 576                 printk(KERN_ERR "ip6_queue: failed to create netlink socket\n");
 577                 goto cleanup_netlink_notifier;
 578         }
 579
 580 #ifdef CONFIG_PROC_FS
 581         proc = proc_create(IPQ_PROC_FS_NAME, 0, init_net.proc_net,
 582                            &ip6_queue_proc_fops);
 583         if (!proc) {
 584                 printk(KERN_ERR "ip6_queue: failed to create proc entry\n");
 585                 goto cleanup_ipqnl;
 586         }
 587 #endif
 588         register_netdevice_notifier(&ipq_dev_notifier);
 589 #ifdef CONFIG_SYSCTL
 590         ipq_sysctl_header = register_sysctl_paths(net_ipv6_ctl_path, ipq_table);
 591 #endif
 592         status = nf_register_queue_handler(NFPROTO_IPV6, &nfqh);
 593         if (status < 0) {
 594                 printk(KERN_ERR "ip6_queue: failed to register queue handler\n");
 595                 goto cleanup_sysctl;
 596         }
 597         return status;
 598
 599 cleanup_sysctl:
 600 #ifdef CONFIG_SYSCTL
 601         unregister_sysctl_table(ipq_sysctl_header);
 602 #endif
 603         unregister_netdevice_notifier(&ipq_dev_notifier);
 604         proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
 605
 606 cleanup_ipqnl: __maybe_unused
 607         netlink_kernel_release(ipqnl);
 608         mutex_lock(&ipqnl_mutex);
 609         mutex_unlock(&ipqnl_mutex);
 610
 611 cleanup_netlink_notifier:
 612         netlink_unregister_notifier(&ipq_nl_notifier);
 613         return status;
 614 }
 615
 616 static void __exit ip6_queue_fini(void)
 617 {
 618         nf_unregister_queue_handlers(&nfqh);
 619
 620         ipq_flush(NULL, 0);
 621
 622 #ifdef CONFIG_SYSCTL
 623         unregister_sysctl_table(ipq_sysctl_header);
 624 #endif
 625         unregister_netdevice_notifier(&ipq_dev_notifier);
 626         proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
 627
 628         netlink_kernel_release(ipqnl);
 629         mutex_lock(&ipqnl_mutex);
 630         mutex_unlock(&ipqnl_mutex);
 631
 632         netlink_unregister_notifier(&ipq_nl_notifier);
 633 }
 634
 635 MODULE_DESCRIPTION("IPv6 packet queue handler");
 636 MODULE_LICENSE("GPL");
 637 MODULE_ALIAS_NET_PF_PROTO(PF_NETLINK, NETLINK_IP6_FW);
 638
 639 module_init(ip6_queue_init);
 640 module_exit(ip6_queue_fini);