linux/net/netfilter/nfnetlink_queue.c
<<
>>
Prefs
   1/*
   2 * This is a module which is used for queueing packets and communicating with
   3 * userspace via nfnetlink.
   4 *
   5 * (C) 2005 by Harald Welte <laforge@netfilter.org>
   6 * (C) 2007 by Patrick McHardy <kaber@trash.net>
   7 *
   8 * Based on the old ipv4-only ip_queue.c:
   9 * (C) 2000-2002 James Morris <jmorris@intercode.com.au>
  10 * (C) 2003-2005 Netfilter Core Team <coreteam@netfilter.org>
  11 *
  12 * This program is free software; you can redistribute it and/or modify
  13 * it under the terms of the GNU General Public License version 2 as
  14 * published by the Free Software Foundation.
  15 *
  16 */
  17#include <linux/module.h>
  18#include <linux/skbuff.h>
  19#include <linux/init.h>
  20#include <linux/spinlock.h>
  21#include <linux/slab.h>
  22#include <linux/notifier.h>
  23#include <linux/netdevice.h>
  24#include <linux/netfilter.h>
  25#include <linux/proc_fs.h>
  26#include <linux/netfilter_ipv4.h>
  27#include <linux/netfilter_ipv6.h>
  28#include <linux/netfilter/nfnetlink.h>
  29#include <linux/netfilter/nfnetlink_queue.h>
  30#include <linux/list.h>
  31#include <net/sock.h>
  32#include <net/netfilter/nf_queue.h>
  33
  34#include <linux/atomic.h>
  35
  36#ifdef CONFIG_BRIDGE_NETFILTER
  37#include "../bridge/br_private.h"
  38#endif
  39
  40#define NFQNL_QMAX_DEFAULT 1024
  41
  42struct nfqnl_instance {
  43        struct hlist_node hlist;                /* global list of queues */
  44        struct rcu_head rcu;
  45
  46        int peer_pid;
  47        unsigned int queue_maxlen;
  48        unsigned int copy_range;
  49        unsigned int queue_dropped;
  50        unsigned int queue_user_dropped;
  51
  52
  53        u_int16_t queue_num;                    /* number of this queue */
  54        u_int8_t copy_mode;
  55/*
  56 * Following fields are dirtied for each queued packet,
  57 * keep them in same cache line if possible.
  58 */
  59        spinlock_t      lock;
  60        unsigned int    queue_total;
  61        unsigned int    id_sequence;            /* 'sequence' of pkt ids */
  62        struct list_head queue_list;            /* packets in queue */
  63};
  64
  65typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
  66
  67static DEFINE_SPINLOCK(instances_lock);
  68
  69#define INSTANCE_BUCKETS        16
  70static struct hlist_head instance_table[INSTANCE_BUCKETS] __read_mostly;
  71
  72static inline u_int8_t instance_hashfn(u_int16_t queue_num)
  73{
  74        return ((queue_num >> 8) | queue_num) % INSTANCE_BUCKETS;
  75}
  76
  77static struct nfqnl_instance *
  78instance_lookup(u_int16_t queue_num)
  79{
  80        struct hlist_head *head;
  81        struct hlist_node *pos;
  82        struct nfqnl_instance *inst;
  83
  84        head = &instance_table[instance_hashfn(queue_num)];
  85        hlist_for_each_entry_rcu(inst, pos, head, hlist) {
  86                if (inst->queue_num == queue_num)
  87                        return inst;
  88        }
  89        return NULL;
  90}
  91
  92static struct nfqnl_instance *
  93instance_create(u_int16_t queue_num, int pid)
  94{
  95        struct nfqnl_instance *inst;
  96        unsigned int h;
  97        int err;
  98
  99        spin_lock(&instances_lock);
 100        if (instance_lookup(queue_num)) {
 101                err = -EEXIST;
 102                goto out_unlock;
 103        }
 104
 105        inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
 106        if (!inst) {
 107                err = -ENOMEM;
 108                goto out_unlock;
 109        }
 110
 111        inst->queue_num = queue_num;
 112        inst->peer_pid = pid;
 113        inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
 114        inst->copy_range = 0xfffff;
 115        inst->copy_mode = NFQNL_COPY_NONE;
 116        spin_lock_init(&inst->lock);
 117        INIT_LIST_HEAD(&inst->queue_list);
 118
 119        if (!try_module_get(THIS_MODULE)) {
 120                err = -EAGAIN;
 121                goto out_free;
 122        }
 123
 124        h = instance_hashfn(queue_num);
 125        hlist_add_head_rcu(&inst->hlist, &instance_table[h]);
 126
 127        spin_unlock(&instances_lock);
 128
 129        return inst;
 130
 131out_free:
 132        kfree(inst);
 133out_unlock:
 134        spin_unlock(&instances_lock);
 135        return ERR_PTR(err);
 136}
 137
 138static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
 139                        unsigned long data);
 140
 141static void
 142instance_destroy_rcu(struct rcu_head *head)
 143{
 144        struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
 145                                                   rcu);
 146
 147        nfqnl_flush(inst, NULL, 0);
 148        kfree(inst);
 149        module_put(THIS_MODULE);
 150}
 151
 152static void
 153__instance_destroy(struct nfqnl_instance *inst)
 154{
 155        hlist_del_rcu(&inst->hlist);
 156        call_rcu(&inst->rcu, instance_destroy_rcu);
 157}
 158
 159static void
 160instance_destroy(struct nfqnl_instance *inst)
 161{
 162        spin_lock(&instances_lock);
 163        __instance_destroy(inst);
 164        spin_unlock(&instances_lock);
 165}
 166
 167static inline void
 168__enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
 169{
 170       list_add_tail(&entry->list, &queue->queue_list);
 171       queue->queue_total++;
 172}
 173
 174static void
 175__dequeue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
 176{
 177        list_del(&entry->list);
 178        queue->queue_total--;
 179}
 180
 181static struct nf_queue_entry *
 182find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
 183{
 184        struct nf_queue_entry *entry = NULL, *i;
 185
 186        spin_lock_bh(&queue->lock);
 187
 188        list_for_each_entry(i, &queue->queue_list, list) {
 189                if (i->id == id) {
 190                        entry = i;
 191                        break;
 192                }
 193        }
 194
 195        if (entry)
 196                __dequeue_entry(queue, entry);
 197
 198        spin_unlock_bh(&queue->lock);
 199
 200        return entry;
 201}
 202
 203static void
 204nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
 205{
 206        struct nf_queue_entry *entry, *next;
 207
 208        spin_lock_bh(&queue->lock);
 209        list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
 210                if (!cmpfn || cmpfn(entry, data)) {
 211                        list_del(&entry->list);
 212                        queue->queue_total--;
 213                        nf_reinject(entry, NF_DROP);
 214                }
 215        }
 216        spin_unlock_bh(&queue->lock);
 217}
 218
 219static struct sk_buff *
 220nfqnl_build_packet_message(struct nfqnl_instance *queue,
 221                           struct nf_queue_entry *entry,
 222                           __be32 **packet_id_ptr)
 223{
 224        sk_buff_data_t old_tail;
 225        size_t size;
 226        size_t data_len = 0;
 227        struct sk_buff *skb;
 228        struct nlattr *nla;
 229        struct nfqnl_msg_packet_hdr *pmsg;
 230        struct nlmsghdr *nlh;
 231        struct nfgenmsg *nfmsg;
 232        struct sk_buff *entskb = entry->skb;
 233        struct net_device *indev;
 234        struct net_device *outdev;
 235
 236        size =    NLMSG_SPACE(sizeof(struct nfgenmsg))
 237                + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
 238                + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 239                + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 240#ifdef CONFIG_BRIDGE_NETFILTER
 241                + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 242                + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 243#endif
 244                + nla_total_size(sizeof(u_int32_t))     /* mark */
 245                + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
 246                + nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
 247
 248        outdev = entry->outdev;
 249
 250        switch ((enum nfqnl_config_mode)ACCESS_ONCE(queue->copy_mode)) {
 251        case NFQNL_COPY_META:
 252        case NFQNL_COPY_NONE:
 253                break;
 254
 255        case NFQNL_COPY_PACKET:
 256                if (entskb->ip_summed == CHECKSUM_PARTIAL &&
 257                    skb_checksum_help(entskb))
 258                        return NULL;
 259
 260                data_len = ACCESS_ONCE(queue->copy_range);
 261                if (data_len == 0 || data_len > entskb->len)
 262                        data_len = entskb->len;
 263
 264                size += nla_total_size(data_len);
 265                break;
 266        }
 267
 268
 269        skb = alloc_skb(size, GFP_ATOMIC);
 270        if (!skb)
 271                goto nlmsg_failure;
 272
 273        old_tail = skb->tail;
 274        nlh = NLMSG_PUT(skb, 0, 0,
 275                        NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
 276                        sizeof(struct nfgenmsg));
 277        nfmsg = NLMSG_DATA(nlh);
 278        nfmsg->nfgen_family = entry->pf;
 279        nfmsg->version = NFNETLINK_V0;
 280        nfmsg->res_id = htons(queue->queue_num);
 281
 282        nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
 283        pmsg = nla_data(nla);
 284        pmsg->hw_protocol       = entskb->protocol;
 285        pmsg->hook              = entry->hook;
 286        *packet_id_ptr          = &pmsg->packet_id;
 287
 288        indev = entry->indev;
 289        if (indev) {
 290#ifndef CONFIG_BRIDGE_NETFILTER
 291                NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex));
 292#else
 293                if (entry->pf == PF_BRIDGE) {
 294                        /* Case 1: indev is physical input device, we need to
 295                         * look for bridge group (when called from
 296                         * netfilter_bridge) */
 297                        NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSINDEV,
 298                                     htonl(indev->ifindex));
 299                        /* this is the bridge group "brX" */
 300                        /* rcu_read_lock()ed by __nf_queue */
 301                        NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV,
 302                                     htonl(br_port_get_rcu(indev)->br->dev->ifindex));
 303                } else {
 304                        /* Case 2: indev is bridge group, we need to look for
 305                         * physical device (when called from ipv4) */
 306                        NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV,
 307                                     htonl(indev->ifindex));
 308                        if (entskb->nf_bridge && entskb->nf_bridge->physindev)
 309                                NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSINDEV,
 310                                             htonl(entskb->nf_bridge->physindev->ifindex));
 311                }
 312#endif
 313        }
 314
 315        if (outdev) {
 316#ifndef CONFIG_BRIDGE_NETFILTER
 317                NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex));
 318#else
 319                if (entry->pf == PF_BRIDGE) {
 320                        /* Case 1: outdev is physical output device, we need to
 321                         * look for bridge group (when called from
 322                         * netfilter_bridge) */
 323                        NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 324                                     htonl(outdev->ifindex));
 325                        /* this is the bridge group "brX" */
 326                        /* rcu_read_lock()ed by __nf_queue */
 327                        NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV,
 328                                     htonl(br_port_get_rcu(outdev)->br->dev->ifindex));
 329                } else {
 330                        /* Case 2: outdev is bridge group, we need to look for
 331                         * physical output device (when called from ipv4) */
 332                        NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV,
 333                                     htonl(outdev->ifindex));
 334                        if (entskb->nf_bridge && entskb->nf_bridge->physoutdev)
 335                                NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 336                                             htonl(entskb->nf_bridge->physoutdev->ifindex));
 337                }
 338#endif
 339        }
 340
 341        if (entskb->mark)
 342                NLA_PUT_BE32(skb, NFQA_MARK, htonl(entskb->mark));
 343
 344        if (indev && entskb->dev &&
 345            entskb->mac_header != entskb->network_header) {
 346                struct nfqnl_msg_packet_hw phw;
 347                int len = dev_parse_header(entskb, phw.hw_addr);
 348                if (len) {
 349                        phw.hw_addrlen = htons(len);
 350                        NLA_PUT(skb, NFQA_HWADDR, sizeof(phw), &phw);
 351                }
 352        }
 353
 354        if (entskb->tstamp.tv64) {
 355                struct nfqnl_msg_packet_timestamp ts;
 356                struct timeval tv = ktime_to_timeval(entskb->tstamp);
 357                ts.sec = cpu_to_be64(tv.tv_sec);
 358                ts.usec = cpu_to_be64(tv.tv_usec);
 359
 360                NLA_PUT(skb, NFQA_TIMESTAMP, sizeof(ts), &ts);
 361        }
 362
 363        if (data_len) {
 364                struct nlattr *nla;
 365                int sz = nla_attr_size(data_len);
 366
 367                if (skb_tailroom(skb) < nla_total_size(data_len)) {
 368                        printk(KERN_WARNING "nf_queue: no tailroom!\n");
 369                        goto nlmsg_failure;
 370                }
 371
 372                nla = (struct nlattr *)skb_put(skb, nla_total_size(data_len));
 373                nla->nla_type = NFQA_PAYLOAD;
 374                nla->nla_len = sz;
 375
 376                if (skb_copy_bits(entskb, 0, nla_data(nla), data_len))
 377                        BUG();
 378        }
 379
 380        nlh->nlmsg_len = skb->tail - old_tail;
 381        return skb;
 382
 383nlmsg_failure:
 384nla_put_failure:
 385        if (skb)
 386                kfree_skb(skb);
 387        if (net_ratelimit())
 388                printk(KERN_ERR "nf_queue: error creating packet message\n");
 389        return NULL;
 390}
 391
 392static int
 393nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 394{
 395        struct sk_buff *nskb;
 396        struct nfqnl_instance *queue;
 397        int err = -ENOBUFS;
 398        __be32 *packet_id_ptr;
 399
 400        /* rcu_read_lock()ed by nf_hook_slow() */
 401        queue = instance_lookup(queuenum);
 402        if (!queue) {
 403                err = -ESRCH;
 404                goto err_out;
 405        }
 406
 407        if (queue->copy_mode == NFQNL_COPY_NONE) {
 408                err = -EINVAL;
 409                goto err_out;
 410        }
 411
 412        nskb = nfqnl_build_packet_message(queue, entry, &packet_id_ptr);
 413        if (nskb == NULL) {
 414                err = -ENOMEM;
 415                goto err_out;
 416        }
 417        spin_lock_bh(&queue->lock);
 418
 419        if (!queue->peer_pid) {
 420                err = -EINVAL;
 421                goto err_out_free_nskb;
 422        }
 423        if (queue->queue_total >= queue->queue_maxlen) {
 424                queue->queue_dropped++;
 425                if (net_ratelimit())
 426                          printk(KERN_WARNING "nf_queue: full at %d entries, "
 427                                 "dropping packets(s).\n",
 428                                 queue->queue_total);
 429                goto err_out_free_nskb;
 430        }
 431        entry->id = ++queue->id_sequence;
 432        *packet_id_ptr = htonl(entry->id);
 433
 434        /* nfnetlink_unicast will either free the nskb or add it to a socket */
 435        err = nfnetlink_unicast(nskb, &init_net, queue->peer_pid, MSG_DONTWAIT);
 436        if (err < 0) {
 437                queue->queue_user_dropped++;
 438                goto err_out_unlock;
 439        }
 440
 441        __enqueue_entry(queue, entry);
 442
 443        spin_unlock_bh(&queue->lock);
 444        return 0;
 445
 446err_out_free_nskb:
 447        kfree_skb(nskb);
 448err_out_unlock:
 449        spin_unlock_bh(&queue->lock);
 450err_out:
 451        return err;
 452}
 453
 454static int
 455nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e)
 456{
 457        struct sk_buff *nskb;
 458        int diff;
 459
 460        diff = data_len - e->skb->len;
 461        if (diff < 0) {
 462                if (pskb_trim(e->skb, data_len))
 463                        return -ENOMEM;
 464        } else if (diff > 0) {
 465                if (data_len > 0xFFFF)
 466                        return -EINVAL;
 467                if (diff > skb_tailroom(e->skb)) {
 468                        nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
 469                                               diff, GFP_ATOMIC);
 470                        if (!nskb) {
 471                                printk(KERN_WARNING "nf_queue: OOM "
 472                                      "in mangle, dropping packet\n");
 473                                return -ENOMEM;
 474                        }
 475                        kfree_skb(e->skb);
 476                        e->skb = nskb;
 477                }
 478                skb_put(e->skb, diff);
 479        }
 480        if (!skb_make_writable(e->skb, data_len))
 481                return -ENOMEM;
 482        skb_copy_to_linear_data(e->skb, data, data_len);
 483        e->skb->ip_summed = CHECKSUM_NONE;
 484        return 0;
 485}
 486
 487static int
 488nfqnl_set_mode(struct nfqnl_instance *queue,
 489               unsigned char mode, unsigned int range)
 490{
 491        int status = 0;
 492
 493        spin_lock_bh(&queue->lock);
 494        switch (mode) {
 495        case NFQNL_COPY_NONE:
 496        case NFQNL_COPY_META:
 497                queue->copy_mode = mode;
 498                queue->copy_range = 0;
 499                break;
 500
 501        case NFQNL_COPY_PACKET:
 502                queue->copy_mode = mode;
 503                /* we're using struct nlattr which has 16bit nla_len */
 504                if (range > 0xffff)
 505                        queue->copy_range = 0xffff;
 506                else
 507                        queue->copy_range = range;
 508                break;
 509
 510        default:
 511                status = -EINVAL;
 512
 513        }
 514        spin_unlock_bh(&queue->lock);
 515
 516        return status;
 517}
 518
 519static int
 520dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
 521{
 522        if (entry->indev)
 523                if (entry->indev->ifindex == ifindex)
 524                        return 1;
 525        if (entry->outdev)
 526                if (entry->outdev->ifindex == ifindex)
 527                        return 1;
 528#ifdef CONFIG_BRIDGE_NETFILTER
 529        if (entry->skb->nf_bridge) {
 530                if (entry->skb->nf_bridge->physindev &&
 531                    entry->skb->nf_bridge->physindev->ifindex == ifindex)
 532                        return 1;
 533                if (entry->skb->nf_bridge->physoutdev &&
 534                    entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
 535                        return 1;
 536        }
 537#endif
 538        return 0;
 539}
 540
 541/* drop all packets with either indev or outdev == ifindex from all queue
 542 * instances */
 543static void
 544nfqnl_dev_drop(int ifindex)
 545{
 546        int i;
 547
 548        rcu_read_lock();
 549
 550        for (i = 0; i < INSTANCE_BUCKETS; i++) {
 551                struct hlist_node *tmp;
 552                struct nfqnl_instance *inst;
 553                struct hlist_head *head = &instance_table[i];
 554
 555                hlist_for_each_entry_rcu(inst, tmp, head, hlist)
 556                        nfqnl_flush(inst, dev_cmp, ifindex);
 557        }
 558
 559        rcu_read_unlock();
 560}
 561
 562#define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
 563
 564static int
 565nfqnl_rcv_dev_event(struct notifier_block *this,
 566                    unsigned long event, void *ptr)
 567{
 568        struct net_device *dev = ptr;
 569
 570        if (!net_eq(dev_net(dev), &init_net))
 571                return NOTIFY_DONE;
 572
 573        /* Drop any packets associated with the downed device */
 574        if (event == NETDEV_DOWN)
 575                nfqnl_dev_drop(dev->ifindex);
 576        return NOTIFY_DONE;
 577}
 578
 579static struct notifier_block nfqnl_dev_notifier = {
 580        .notifier_call  = nfqnl_rcv_dev_event,
 581};
 582
 583static int
 584nfqnl_rcv_nl_event(struct notifier_block *this,
 585                   unsigned long event, void *ptr)
 586{
 587        struct netlink_notify *n = ptr;
 588
 589        if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
 590                int i;
 591
 592                /* destroy all instances for this pid */
 593                spin_lock(&instances_lock);
 594                for (i = 0; i < INSTANCE_BUCKETS; i++) {
 595                        struct hlist_node *tmp, *t2;
 596                        struct nfqnl_instance *inst;
 597                        struct hlist_head *head = &instance_table[i];
 598
 599                        hlist_for_each_entry_safe(inst, tmp, t2, head, hlist) {
 600                                if ((n->net == &init_net) &&
 601                                    (n->pid == inst->peer_pid))
 602                                        __instance_destroy(inst);
 603                        }
 604                }
 605                spin_unlock(&instances_lock);
 606        }
 607        return NOTIFY_DONE;
 608}
 609
 610static struct notifier_block nfqnl_rtnl_notifier = {
 611        .notifier_call  = nfqnl_rcv_nl_event,
 612};
 613
 614static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
 615        [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 616        [NFQA_MARK]             = { .type = NLA_U32 },
 617        [NFQA_PAYLOAD]          = { .type = NLA_UNSPEC },
 618};
 619
 620static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
 621        [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 622        [NFQA_MARK]             = { .type = NLA_U32 },
 623};
 624
 625static struct nfqnl_instance *verdict_instance_lookup(u16 queue_num, int nlpid)
 626{
 627        struct nfqnl_instance *queue;
 628
 629        queue = instance_lookup(queue_num);
 630        if (!queue)
 631                return ERR_PTR(-ENODEV);
 632
 633        if (queue->peer_pid != nlpid)
 634                return ERR_PTR(-EPERM);
 635
 636        return queue;
 637}
 638
 639static struct nfqnl_msg_verdict_hdr*
 640verdicthdr_get(const struct nlattr * const nfqa[])
 641{
 642        struct nfqnl_msg_verdict_hdr *vhdr;
 643        unsigned int verdict;
 644
 645        if (!nfqa[NFQA_VERDICT_HDR])
 646                return NULL;
 647
 648        vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
 649        verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
 650        if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
 651                return NULL;
 652        return vhdr;
 653}
 654
 655static int nfq_id_after(unsigned int id, unsigned int max)
 656{
 657        return (int)(id - max) > 0;
 658}
 659
 660static int
 661nfqnl_recv_verdict_batch(struct sock *ctnl, struct sk_buff *skb,
 662                   const struct nlmsghdr *nlh,
 663                   const struct nlattr * const nfqa[])
 664{
 665        struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 666        struct nf_queue_entry *entry, *tmp;
 667        unsigned int verdict, maxid;
 668        struct nfqnl_msg_verdict_hdr *vhdr;
 669        struct nfqnl_instance *queue;
 670        LIST_HEAD(batch_list);
 671        u16 queue_num = ntohs(nfmsg->res_id);
 672
 673        queue = verdict_instance_lookup(queue_num, NETLINK_CB(skb).pid);
 674        if (IS_ERR(queue))
 675                return PTR_ERR(queue);
 676
 677        vhdr = verdicthdr_get(nfqa);
 678        if (!vhdr)
 679                return -EINVAL;
 680
 681        verdict = ntohl(vhdr->verdict);
 682        maxid = ntohl(vhdr->id);
 683
 684        spin_lock_bh(&queue->lock);
 685
 686        list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
 687                if (nfq_id_after(entry->id, maxid))
 688                        break;
 689                __dequeue_entry(queue, entry);
 690                list_add_tail(&entry->list, &batch_list);
 691        }
 692
 693        spin_unlock_bh(&queue->lock);
 694
 695        if (list_empty(&batch_list))
 696                return -ENOENT;
 697
 698        list_for_each_entry_safe(entry, tmp, &batch_list, list) {
 699                if (nfqa[NFQA_MARK])
 700                        entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
 701                nf_reinject(entry, verdict);
 702        }
 703        return 0;
 704}
 705
 706static int
 707nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
 708                   const struct nlmsghdr *nlh,
 709                   const struct nlattr * const nfqa[])
 710{
 711        struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 712        u_int16_t queue_num = ntohs(nfmsg->res_id);
 713
 714        struct nfqnl_msg_verdict_hdr *vhdr;
 715        struct nfqnl_instance *queue;
 716        unsigned int verdict;
 717        struct nf_queue_entry *entry;
 718
 719        queue = instance_lookup(queue_num);
 720        if (!queue)
 721
 722        queue = verdict_instance_lookup(queue_num, NETLINK_CB(skb).pid);
 723        if (IS_ERR(queue))
 724                return PTR_ERR(queue);
 725
 726        vhdr = verdicthdr_get(nfqa);
 727        if (!vhdr)
 728                return -EINVAL;
 729
 730        verdict = ntohl(vhdr->verdict);
 731
 732        entry = find_dequeue_entry(queue, ntohl(vhdr->id));
 733        if (entry == NULL)
 734                return -ENOENT;
 735
 736        if (nfqa[NFQA_PAYLOAD]) {
 737                if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
 738                                 nla_len(nfqa[NFQA_PAYLOAD]), entry) < 0)
 739                        verdict = NF_DROP;
 740        }
 741
 742        if (nfqa[NFQA_MARK])
 743                entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
 744
 745        nf_reinject(entry, verdict);
 746        return 0;
 747}
 748
 749static int
 750nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
 751                  const struct nlmsghdr *nlh,
 752                  const struct nlattr * const nfqa[])
 753{
 754        return -ENOTSUPP;
 755}
 756
 757static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
 758        [NFQA_CFG_CMD]          = { .len = sizeof(struct nfqnl_msg_config_cmd) },
 759        [NFQA_CFG_PARAMS]       = { .len = sizeof(struct nfqnl_msg_config_params) },
 760};
 761
 762static const struct nf_queue_handler nfqh = {
 763        .name   = "nf_queue",
 764        .outfn  = &nfqnl_enqueue_packet,
 765};
 766
 767static int
 768nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
 769                  const struct nlmsghdr *nlh,
 770                  const struct nlattr * const nfqa[])
 771{
 772        struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
 773        u_int16_t queue_num = ntohs(nfmsg->res_id);
 774        struct nfqnl_instance *queue;
 775        struct nfqnl_msg_config_cmd *cmd = NULL;
 776        int ret = 0;
 777
 778        if (nfqa[NFQA_CFG_CMD]) {
 779                cmd = nla_data(nfqa[NFQA_CFG_CMD]);
 780
 781                /* Commands without queue context - might sleep */
 782                switch (cmd->command) {
 783                case NFQNL_CFG_CMD_PF_BIND:
 784                        return nf_register_queue_handler(ntohs(cmd->pf),
 785                                                         &nfqh);
 786                case NFQNL_CFG_CMD_PF_UNBIND:
 787                        return nf_unregister_queue_handler(ntohs(cmd->pf),
 788                                                           &nfqh);
 789                }
 790        }
 791
 792        rcu_read_lock();
 793        queue = instance_lookup(queue_num);
 794        if (queue && queue->peer_pid != NETLINK_CB(skb).pid) {
 795                ret = -EPERM;
 796                goto err_out_unlock;
 797        }
 798
 799        if (cmd != NULL) {
 800                switch (cmd->command) {
 801                case NFQNL_CFG_CMD_BIND:
 802                        if (queue) {
 803                                ret = -EBUSY;
 804                                goto err_out_unlock;
 805                        }
 806                        queue = instance_create(queue_num, NETLINK_CB(skb).pid);
 807                        if (IS_ERR(queue)) {
 808                                ret = PTR_ERR(queue);
 809                                goto err_out_unlock;
 810                        }
 811                        break;
 812                case NFQNL_CFG_CMD_UNBIND:
 813                        if (!queue) {
 814                                ret = -ENODEV;
 815                                goto err_out_unlock;
 816                        }
 817                        instance_destroy(queue);
 818                        break;
 819                case NFQNL_CFG_CMD_PF_BIND:
 820                case NFQNL_CFG_CMD_PF_UNBIND:
 821                        break;
 822                default:
 823                        ret = -ENOTSUPP;
 824                        break;
 825                }
 826        }
 827
 828        if (nfqa[NFQA_CFG_PARAMS]) {
 829                struct nfqnl_msg_config_params *params;
 830
 831                if (!queue) {
 832                        ret = -ENODEV;
 833                        goto err_out_unlock;
 834                }
 835                params = nla_data(nfqa[NFQA_CFG_PARAMS]);
 836                nfqnl_set_mode(queue, params->copy_mode,
 837                                ntohl(params->copy_range));
 838        }
 839
 840        if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
 841                __be32 *queue_maxlen;
 842
 843                if (!queue) {
 844                        ret = -ENODEV;
 845                        goto err_out_unlock;
 846                }
 847                queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
 848                spin_lock_bh(&queue->lock);
 849                queue->queue_maxlen = ntohl(*queue_maxlen);
 850                spin_unlock_bh(&queue->lock);
 851        }
 852
 853err_out_unlock:
 854        rcu_read_unlock();
 855        return ret;
 856}
 857
 858static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
 859        [NFQNL_MSG_PACKET]      = { .call_rcu = nfqnl_recv_unsupp,
 860                                    .attr_count = NFQA_MAX, },
 861        [NFQNL_MSG_VERDICT]     = { .call_rcu = nfqnl_recv_verdict,
 862                                    .attr_count = NFQA_MAX,
 863                                    .policy = nfqa_verdict_policy },
 864        [NFQNL_MSG_CONFIG]      = { .call = nfqnl_recv_config,
 865                                    .attr_count = NFQA_CFG_MAX,
 866                                    .policy = nfqa_cfg_policy },
 867        [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
 868                                    .attr_count = NFQA_MAX,
 869                                    .policy = nfqa_verdict_batch_policy },
 870};
 871
 872static const struct nfnetlink_subsystem nfqnl_subsys = {
 873        .name           = "nf_queue",
 874        .subsys_id      = NFNL_SUBSYS_QUEUE,
 875        .cb_count       = NFQNL_MSG_MAX,
 876        .cb             = nfqnl_cb,
 877};
 878
 879#ifdef CONFIG_PROC_FS
 880struct iter_state {
 881        unsigned int bucket;
 882};
 883
 884static struct hlist_node *get_first(struct seq_file *seq)
 885{
 886        struct iter_state *st = seq->private;
 887
 888        if (!st)
 889                return NULL;
 890
 891        for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
 892                if (!hlist_empty(&instance_table[st->bucket]))
 893                        return instance_table[st->bucket].first;
 894        }
 895        return NULL;
 896}
 897
 898static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
 899{
 900        struct iter_state *st = seq->private;
 901
 902        h = h->next;
 903        while (!h) {
 904                if (++st->bucket >= INSTANCE_BUCKETS)
 905                        return NULL;
 906
 907                h = instance_table[st->bucket].first;
 908        }
 909        return h;
 910}
 911
 912static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
 913{
 914        struct hlist_node *head;
 915        head = get_first(seq);
 916
 917        if (head)
 918                while (pos && (head = get_next(seq, head)))
 919                        pos--;
 920        return pos ? NULL : head;
 921}
 922
 923static void *seq_start(struct seq_file *seq, loff_t *pos)
 924        __acquires(instances_lock)
 925{
 926        spin_lock(&instances_lock);
 927        return get_idx(seq, *pos);
 928}
 929
 930static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
 931{
 932        (*pos)++;
 933        return get_next(s, v);
 934}
 935
 936static void seq_stop(struct seq_file *s, void *v)
 937        __releases(instances_lock)
 938{
 939        spin_unlock(&instances_lock);
 940}
 941
 942static int seq_show(struct seq_file *s, void *v)
 943{
 944        const struct nfqnl_instance *inst = v;
 945
 946        return seq_printf(s, "%5d %6d %5d %1d %5d %5d %5d %8d %2d\n",
 947                          inst->queue_num,
 948                          inst->peer_pid, inst->queue_total,
 949                          inst->copy_mode, inst->copy_range,
 950                          inst->queue_dropped, inst->queue_user_dropped,
 951                          inst->id_sequence, 1);
 952}
 953
 954static const struct seq_operations nfqnl_seq_ops = {
 955        .start  = seq_start,
 956        .next   = seq_next,
 957        .stop   = seq_stop,
 958        .show   = seq_show,
 959};
 960
 961static int nfqnl_open(struct inode *inode, struct file *file)
 962{
 963        return seq_open_private(file, &nfqnl_seq_ops,
 964                        sizeof(struct iter_state));
 965}
 966
 967static const struct file_operations nfqnl_file_ops = {
 968        .owner   = THIS_MODULE,
 969        .open    = nfqnl_open,
 970        .read    = seq_read,
 971        .llseek  = seq_lseek,
 972        .release = seq_release_private,
 973};
 974
 975#endif /* PROC_FS */
 976
 977static int __init nfnetlink_queue_init(void)
 978{
 979        int i, status = -ENOMEM;
 980
 981        for (i = 0; i < INSTANCE_BUCKETS; i++)
 982                INIT_HLIST_HEAD(&instance_table[i]);
 983
 984        netlink_register_notifier(&nfqnl_rtnl_notifier);
 985        status = nfnetlink_subsys_register(&nfqnl_subsys);
 986        if (status < 0) {
 987                printk(KERN_ERR "nf_queue: failed to create netlink socket\n");
 988                goto cleanup_netlink_notifier;
 989        }
 990
 991#ifdef CONFIG_PROC_FS
 992        if (!proc_create("nfnetlink_queue", 0440,
 993                         proc_net_netfilter, &nfqnl_file_ops))
 994                goto cleanup_subsys;
 995#endif
 996
 997        register_netdevice_notifier(&nfqnl_dev_notifier);
 998        return status;
 999
1000#ifdef CONFIG_PROC_FS
1001cleanup_subsys:
1002        nfnetlink_subsys_unregister(&nfqnl_subsys);
1003#endif
1004cleanup_netlink_notifier:
1005        netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1006        return status;
1007}
1008
1009static void __exit nfnetlink_queue_fini(void)
1010{
1011        nf_unregister_queue_handlers(&nfqh);
1012        unregister_netdevice_notifier(&nfqnl_dev_notifier);
1013#ifdef CONFIG_PROC_FS
1014        remove_proc_entry("nfnetlink_queue", proc_net_netfilter);
1015#endif
1016        nfnetlink_subsys_unregister(&nfqnl_subsys);
1017        netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1018
1019        rcu_barrier(); /* Wait for completion of call_rcu()'s */
1020}
1021
1022MODULE_DESCRIPTION("netfilter packet queue handler");
1023MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1024MODULE_LICENSE("GPL");
1025MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1026
1027module_init(nfnetlink_queue_init);
1028module_exit(nfnetlink_queue_fini);
1029