net/ipv4/netfilter/ip_queue.c

   1 /*
   2  * This is a module which is used for queueing IPv4 packets and
   3  * communicating with userspace via netlink.
   4  *
   5  * (C) 2000-2002 James Morris <jmorris@intercode.com.au>
   6  * (C) 2003-2005 Netfilter Core Team <coreteam@netfilter.org>
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License version 2 as
  10  * published by the Free Software Foundation.
  11  */
  12 #include <linux/module.h>
  13 #include <linux/skbuff.h>
  14 #include <linux/init.h>
  15 #include <linux/ip.h>
  16 #include <linux/notifier.h>
  17 #include <linux/netdevice.h>
  18 #include <linux/netfilter.h>
  19 #include <linux/netfilter_ipv4/ip_queue.h>
  20 #include <linux/netfilter_ipv4/ip_tables.h>
  21 #include <linux/netlink.h>
  22 #include <linux/spinlock.h>
  23 #include <linux/sysctl.h>
  24 #include <linux/proc_fs.h>
  25 #include <linux/seq_file.h>
  26 #include <linux/security.h>
  27 #include <linux/mutex.h>
  28 #include <net/net_namespace.h>
  29 #include <net/sock.h>
  30 #include <net/route.h>
  31 #include <net/netfilter/nf_queue.h>
  32 #include <net/ip.h>
  33
  34 #define IPQ_QMAX_DEFAULT 1024
  35 #define IPQ_PROC_FS_NAME "ip_queue"
  36 #define NET_IPQ_QMAX 2088
  37 #define NET_IPQ_QMAX_NAME "ip_queue_maxlen"
  38
  39 typedef int (*ipq_cmpfn)(struct nf_queue_entry *, unsigned long);
  40
  41 static unsigned char copy_mode __read_mostly = IPQ_COPY_NONE;
  42 static unsigned int queue_maxlen __read_mostly = IPQ_QMAX_DEFAULT;
  43 static DEFINE_RWLOCK(queue_lock);
  44 static int peer_pid __read_mostly;
  45 static unsigned int copy_range __read_mostly;
  46 static unsigned int queue_total;
  47 static unsigned int queue_dropped = 0;
  48 static unsigned int queue_user_dropped = 0;
  49 static struct sock *ipqnl __read_mostly;
  50 static LIST_HEAD(queue_list);
  51 static DEFINE_MUTEX(ipqnl_mutex);
  52
  53 static inline void
  54 __ipq_enqueue_entry(struct nf_queue_entry *entry)
  55 {
  56        list_add_tail(&entry->list, &queue_list);
  57        queue_total++;
  58 }
  59
  60 static inline int
  61 __ipq_set_mode(unsigned char mode, unsigned int range)
  62 {
  63         int status = 0;
  64
  65         switch(mode) {
  66         case IPQ_COPY_NONE:
  67         case IPQ_COPY_META:
  68                 copy_mode = mode;
  69                 copy_range = 0;
  70                 break;
  71
  72         case IPQ_COPY_PACKET:
  73                 copy_mode = mode;
  74                 copy_range = range;
  75                 if (copy_range > 0xFFFF)
  76                         copy_range = 0xFFFF;
  77                 break;
  78
  79         default:
  80                 status = -EINVAL;
  81
  82         }
  83         return status;
  84 }
  85
  86 static void __ipq_flush(ipq_cmpfn cmpfn, unsigned long data);
  87
  88 static inline void
  89 __ipq_reset(void)
  90 {
  91         peer_pid = 0;
  92         net_disable_timestamp();
  93         __ipq_set_mode(IPQ_COPY_NONE, 0);
  94         __ipq_flush(NULL, 0);
  95 }
  96
  97 static struct nf_queue_entry *
  98 ipq_find_dequeue_entry(unsigned long id)
  99 {
 100         struct nf_queue_entry *entry = NULL, *i;
 101
 102         write_lock_bh(&queue_lock);
 103
 104         list_for_each_entry(i, &queue_list, list) {
 105                 if ((unsigned long)i == id) {
 106                         entry = i;
 107                         break;
 108                 }
 109         }
 110
 111         if (entry) {
 112                 list_del(&entry->list);
 113                 queue_total--;
 114         }
 115
 116         write_unlock_bh(&queue_lock);
 117         return entry;
 118 }
 119
 120 static void
 121 __ipq_flush(ipq_cmpfn cmpfn, unsigned long data)
 122 {
 123         struct nf_queue_entry *entry, *next;
 124
 125         list_for_each_entry_safe(entry, next, &queue_list, list) {
 126                 if (!cmpfn || cmpfn(entry, data)) {
 127                         list_del(&entry->list);
 128                         queue_total--;
 129                         nf_reinject(entry, NF_DROP);
 130                 }
 131         }
 132 }
 133
 134 static void
 135 ipq_flush(ipq_cmpfn cmpfn, unsigned long data)
 136 {
 137         write_lock_bh(&queue_lock);
 138         __ipq_flush(cmpfn, data);
 139         write_unlock_bh(&queue_lock);
 140 }
 141
 142 static struct sk_buff *
 143 ipq_build_packet_message(struct nf_queue_entry *entry, int *errp)
 144 {
 145         sk_buff_data_t old_tail;
 146         size_t size = 0;
 147         size_t data_len = 0;
 148         struct sk_buff *skb;
 149         struct ipq_packet_msg *pmsg;
 150         struct nlmsghdr *nlh;
 151         struct timeval tv;
 152
 153         read_lock_bh(&queue_lock);
 154
 155         switch (copy_mode) {
 156         case IPQ_COPY_META:
 157         case IPQ_COPY_NONE:
 158                 size = NLMSG_SPACE(sizeof(*pmsg));
 159                 data_len = 0;
 160                 break;
 161
 162         case IPQ_COPY_PACKET:
 163                 if ((entry->skb->ip_summed == CHECKSUM_PARTIAL ||
 164                      entry->skb->ip_summed == CHECKSUM_COMPLETE) &&
 165                     (*errp = skb_checksum_help(entry->skb))) {
 166                         read_unlock_bh(&queue_lock);
 167                         return NULL;
 168                 }
 169                 if (copy_range == 0 || copy_range > entry->skb->len)
 170                         data_len = entry->skb->len;
 171                 else
 172                         data_len = copy_range;
 173
 174                 size = NLMSG_SPACE(sizeof(*pmsg) + data_len);
 175                 break;
 176
 177         default:
 178                 *errp = -EINVAL;
 179                 read_unlock_bh(&queue_lock);
 180                 return NULL;
 181         }
 182
 183         read_unlock_bh(&queue_lock);
 184
 185         skb = alloc_skb(size, GFP_ATOMIC);
 186         if (!skb)
 187                 goto nlmsg_failure;
 188
 189         old_tail = skb->tail;
 190         nlh = NLMSG_PUT(skb, 0, 0, IPQM_PACKET, size - sizeof(*nlh));
 191         pmsg = NLMSG_DATA(nlh);
 192         memset(pmsg, 0, sizeof(*pmsg));
 193
 194         pmsg->packet_id       = (unsigned long )entry;
 195         pmsg->data_len        = data_len;
 196         tv = ktime_to_timeval(entry->skb->tstamp);
 197         pmsg->timestamp_sec   = tv.tv_sec;
 198         pmsg->timestamp_usec  = tv.tv_usec;
 199         pmsg->mark            = entry->skb->mark;
 200         pmsg->hook            = entry->hook;
 201         pmsg->hw_protocol     = entry->skb->protocol;
 202
 203         if (entry->indev)
 204                 strcpy(pmsg->indev_name, entry->indev->name);
 205         else
 206                 pmsg->indev_name[0] = '\0';
 207
 208         if (entry->outdev)
 209                 strcpy(pmsg->outdev_name, entry->outdev->name);
 210         else
 211                 pmsg->outdev_name[0] = '\0';
 212
 213         if (entry->indev && entry->skb->dev) {
 214                 pmsg->hw_type = entry->skb->dev->type;
 215                 pmsg->hw_addrlen = dev_parse_header(entry->skb,
 216                                                     pmsg->hw_addr);
 217         }
 218
 219         if (data_len)
 220                 if (skb_copy_bits(entry->skb, 0, pmsg->payload, data_len))
 221                         BUG();
 222
 223         nlh->nlmsg_len = skb->tail - old_tail;
 224         return skb;
 225
 226 nlmsg_failure:
 227         if (skb)
 228                 kfree_skb(skb);
 229         *errp = -EINVAL;
 230         printk(KERN_ERR "ip_queue: error creating packet message\n");
 231         return NULL;
 232 }
 233
 234 static int
 235 ipq_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 236 {
 237         int status = -EINVAL;
 238         struct sk_buff *nskb;
 239
 240         if (copy_mode == IPQ_COPY_NONE)
 241                 return -EAGAIN;
 242
 243         nskb = ipq_build_packet_message(entry, &status);
 244         if (nskb == NULL)
 245                 return status;
 246
 247         write_lock_bh(&queue_lock);
 248
 249         if (!peer_pid)
 250                 goto err_out_free_nskb;
 251
 252         if (queue_total >= queue_maxlen) {
 253                 queue_dropped++;
 254                 status = -ENOSPC;
 255                 if (net_ratelimit())
 256                           printk (KERN_WARNING "ip_queue: full at %d entries, "
 257                                   "dropping packets(s). Dropped: %d\n", queue_total,
 258                                   queue_dropped);
 259                 goto err_out_free_nskb;
 260         }
 261
 262         /* netlink_unicast will either free the nskb or attach it to a socket */
 263         status = netlink_unicast(ipqnl, nskb, peer_pid, MSG_DONTWAIT);
 264         if (status < 0) {
 265                 queue_user_dropped++;
 266                 goto err_out_unlock;
 267         }
 268
 269         __ipq_enqueue_entry(entry);
 270
 271         write_unlock_bh(&queue_lock);
 272         return status;
 273
 274 err_out_free_nskb:
 275         kfree_skb(nskb);
 276
 277 err_out_unlock:
 278         write_unlock_bh(&queue_lock);
 279         return status;
 280 }
 281
 282 static int
 283 ipq_mangle_ipv4(ipq_verdict_msg_t *v, struct nf_queue_entry *e)
 284 {
 285         int diff;
 286         int err;
 287         struct iphdr *user_iph = (struct iphdr *)v->payload;
 288
 289         if (v->data_len < sizeof(*user_iph))
 290                 return 0;
 291         diff = v->data_len - e->skb->len;
 292         if (diff < 0) {
 293                 if (pskb_trim(e->skb, v->data_len))
 294                         return -ENOMEM;
 295         } else if (diff > 0) {
 296                 if (v->data_len > 0xFFFF)
 297                         return -EINVAL;
 298                 if (diff > skb_tailroom(e->skb)) {
 299                         err = pskb_expand_head(e->skb, 0,
 300                                                diff - skb_tailroom(e->skb),
 301                                                GFP_ATOMIC);
 302                         if (err) {
 303                                 printk(KERN_WARNING "ip_queue: error "
 304                                       "in mangle, dropping packet: %d\n", -err);
 305                                 return err;
 306                         }
 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_ipv4(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         if (entry->outdev)
 387                 if (entry->outdev->ifindex == ifindex)
 388                         return 1;
 389 #ifdef CONFIG_BRIDGE_NETFILTER
 390         if (entry->skb->nf_bridge) {
 391                 if (entry->skb->nf_bridge->physindev &&
 392                     entry->skb->nf_bridge->physindev->ifindex == ifindex)
 393                         return 1;
 394                 if (entry->skb->nf_bridge->physoutdev &&
 395                     entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
 396                         return 1;
 397         }
 398 #endif
 399         return 0;
 400 }
 401
 402 static void
 403 ipq_dev_drop(int ifindex)
 404 {
 405         ipq_flush(dev_cmp, ifindex);
 406 }
 407
 408 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
 409
 410 static inline void
 411 __ipq_rcv_skb(struct sk_buff *skb)
 412 {
 413         int status, type, pid, flags, nlmsglen, skblen;
 414         struct nlmsghdr *nlh;
 415
 416         skblen = skb->len;
 417         if (skblen < sizeof(*nlh))
 418                 return;
 419
 420         nlh = nlmsg_hdr(skb);
 421         nlmsglen = nlh->nlmsg_len;
 422         if (nlmsglen < sizeof(*nlh) || skblen < nlmsglen)
 423                 return;
 424
 425         pid = nlh->nlmsg_pid;
 426         flags = nlh->nlmsg_flags;
 427
 428         if(pid <= 0 || !(flags & NLM_F_REQUEST) || flags & NLM_F_MULTI)
 429                 RCV_SKB_FAIL(-EINVAL);
 430
 431         if (flags & MSG_TRUNC)
 432                 RCV_SKB_FAIL(-ECOMM);
 433
 434         type = nlh->nlmsg_type;
 435         if (type < NLMSG_NOOP || type >= IPQM_MAX)
 436                 RCV_SKB_FAIL(-EINVAL);
 437
 438         if (type <= IPQM_BASE)
 439                 return;
 440
 441         if (security_netlink_recv(skb, CAP_NET_ADMIN))
 442                 RCV_SKB_FAIL(-EPERM);
 443
 444         write_lock_bh(&queue_lock);
 445
 446         if (peer_pid) {
 447                 if (peer_pid != pid) {
 448                         write_unlock_bh(&queue_lock);
 449                         RCV_SKB_FAIL(-EBUSY);
 450                 }
 451         } else {
 452                 net_enable_timestamp();
 453                 peer_pid = pid;
 454         }
 455
 456         write_unlock_bh(&queue_lock);
 457
 458         status = ipq_receive_peer(NLMSG_DATA(nlh), type,
 459                                   nlmsglen - NLMSG_LENGTH(0));
 460         if (status < 0)
 461                 RCV_SKB_FAIL(status);
 462
 463         if (flags & NLM_F_ACK)
 464                 netlink_ack(skb, nlh, 0);
 465         return;
 466 }
 467
 468 static void
 469 ipq_rcv_skb(struct sk_buff *skb)
 470 {
 471         mutex_lock(&ipqnl_mutex);
 472         __ipq_rcv_skb(skb);
 473         mutex_unlock(&ipqnl_mutex);
 474 }
 475
 476 static int
 477 ipq_rcv_dev_event(struct notifier_block *this,
 478                   unsigned long event, void *ptr)
 479 {
 480         struct net_device *dev = ptr;
 481
 482         if (dev->nd_net != &init_net)
 483                 return NOTIFY_DONE;
 484
 485         /* Drop any packets associated with the downed device */
 486         if (event == NETDEV_DOWN)
 487                 ipq_dev_drop(dev->ifindex);
 488         return NOTIFY_DONE;
 489 }
 490
 491 static struct notifier_block ipq_dev_notifier = {
 492         .notifier_call  = ipq_rcv_dev_event,
 493 };
 494
 495 static int
 496 ipq_rcv_nl_event(struct notifier_block *this,
 497                  unsigned long event, void *ptr)
 498 {
 499         struct netlink_notify *n = ptr;
 500
 501         if (event == NETLINK_URELEASE &&
 502             n->protocol == NETLINK_FIREWALL && n->pid) {
 503                 write_lock_bh(&queue_lock);
 504                 if ((n->net == &init_net) && (n->pid == peer_pid))
 505                         __ipq_reset();
 506                 write_unlock_bh(&queue_lock);
 507         }
 508         return NOTIFY_DONE;
 509 }
 510
 511 static struct notifier_block ipq_nl_notifier = {
 512         .notifier_call  = ipq_rcv_nl_event,
 513 };
 514
 515 #ifdef CONFIG_SYSCTL
 516 static struct ctl_table_header *ipq_sysctl_header;
 517
 518 static ctl_table ipq_table[] = {
 519         {
 520                 .ctl_name       = NET_IPQ_QMAX,
 521                 .procname       = NET_IPQ_QMAX_NAME,
 522                 .data           = &queue_maxlen,
 523                 .maxlen         = sizeof(queue_maxlen),
 524                 .mode           = 0644,
 525                 .proc_handler   = proc_dointvec
 526         },
 527         { .ctl_name = 0 }
 528 };
 529 #endif
 530
 531 #ifdef CONFIG_PROC_FS
 532 static int ip_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                       "Netlink 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 ip_queue_open(struct inode *inode, struct file *file)
 557 {
 558         return single_open(file, ip_queue_show, NULL);
 559 }
 560
 561 static const struct file_operations ip_queue_proc_fops = {
 562         .open           = ip_queue_open,
 563         .read           = seq_read,
 564         .llseek         = seq_lseek,
 565         .release        = single_release,
 566         .owner          = THIS_MODULE,
 567 };
 568 #endif
 569
 570 static const struct nf_queue_handler nfqh = {
 571         .name   = "ip_queue",
 572         .outfn  = &ipq_enqueue_packet,
 573 };
 574
 575 static int __init ip_queue_init(void)
 576 {
 577         int status = -ENOMEM;
 578         struct proc_dir_entry *proc __maybe_unused;
 579
 580         netlink_register_notifier(&ipq_nl_notifier);
 581         ipqnl = netlink_kernel_create(&init_net, NETLINK_FIREWALL, 0,
 582                                       ipq_rcv_skb, NULL, THIS_MODULE);
 583         if (ipqnl == NULL) {
 584                 printk(KERN_ERR "ip_queue: failed to create netlink socket\n");
 585                 goto cleanup_netlink_notifier;
 586         }
 587
 588 #ifdef CONFIG_PROC_FS
 589         proc = create_proc_entry(IPQ_PROC_FS_NAME, 0, init_net.proc_net);
 590         if (proc) {
 591                 proc->owner = THIS_MODULE;
 592                 proc->proc_fops = &ip_queue_proc_fops;
 593         } else {
 594                 printk(KERN_ERR "ip_queue: failed to create proc entry\n");
 595                 goto cleanup_ipqnl;
 596         }
 597 #endif
 598         register_netdevice_notifier(&ipq_dev_notifier);
 599 #ifdef CONFIG_SYSCTL
 600         ipq_sysctl_header = register_sysctl_paths(net_ipv4_ctl_path, ipq_table);
 601 #endif
 602         status = nf_register_queue_handler(PF_INET, &nfqh);
 603         if (status < 0) {
 604                 printk(KERN_ERR "ip_queue: failed to register queue handler\n");
 605                 goto cleanup_sysctl;
 606         }
 607         return status;
 608
 609 cleanup_sysctl:
 610 #ifdef CONFIG_SYSCTL
 611         unregister_sysctl_table(ipq_sysctl_header);
 612 #endif
 613         unregister_netdevice_notifier(&ipq_dev_notifier);
 614         proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
 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 ip_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("IPv4 packet queue handler");
 645 MODULE_AUTHOR("James Morris <jmorris@intercode.com.au>");
 646 MODULE_LICENSE("GPL");
 647
 648 module_init(ip_queue_init);
 649 module_exit(ip_queue_fini);