net/netfilter/nfnetlink_queue.c

   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_bridge.h>
  29 #include <linux/netfilter/nfnetlink.h>
  30 #include <linux/netfilter/nfnetlink_queue.h>
  31 #include <linux/netfilter/nf_conntrack_common.h>
  32 #include <linux/list.h>
  33 #include <net/sock.h>
  34 #include <net/tcp_states.h>
  35 #include <net/netfilter/nf_queue.h>
  36 #include <net/netns/generic.h>
  37
  38 #include <linux/atomic.h>
  39
  40 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  41 #include "../bridge/br_private.h"
  42 #endif
  43
  44 #define NFQNL_QMAX_DEFAULT 1024
  45
  46 /* We're using struct nlattr which has 16bit nla_len. Note that nla_len
  47  * includes the header length. Thus, the maximum packet length that we
  48  * support is 65531 bytes. We send truncated packets if the specified length
  49  * is larger than that.  Userspace can check for presence of NFQA_CAP_LEN
  50  * attribute to detect truncation.
  51  */
  52 #define NFQNL_MAX_COPY_RANGE (0xffff - NLA_HDRLEN)
  53
  54 struct nfqnl_instance {
  55         struct hlist_node hlist;                /* global list of queues */
  56         struct rcu_head rcu;
  57
  58         u32 peer_portid;
  59         unsigned int queue_maxlen;
  60         unsigned int copy_range;
  61         unsigned int queue_dropped;
  62         unsigned int queue_user_dropped;
  63
  64
  65         u_int16_t queue_num;                    /* number of this queue */
  66         u_int8_t copy_mode;
  67         u_int32_t flags;                        /* Set using NFQA_CFG_FLAGS */
  68 /*
  69  * Following fields are dirtied for each queued packet,
  70  * keep them in same cache line if possible.
  71  */
  72         spinlock_t      lock;
  73         unsigned int    queue_total;
  74         unsigned int    id_sequence;            /* 'sequence' of pkt ids */
  75         struct list_head queue_list;            /* packets in queue */
  76 };
  77
  78 typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
  79
  80 static int nfnl_queue_net_id __read_mostly;
  81
  82 #define INSTANCE_BUCKETS        16
  83 struct nfnl_queue_net {
  84         spinlock_t instances_lock;
  85         struct hlist_head instance_table[INSTANCE_BUCKETS];
  86 };
  87
  88 static struct nfnl_queue_net *nfnl_queue_pernet(struct net *net)
  89 {
  90         return net_generic(net, nfnl_queue_net_id);
  91 }
  92
  93 static inline u_int8_t instance_hashfn(u_int16_t queue_num)
  94 {
  95         return ((queue_num >> 8) ^ queue_num) % INSTANCE_BUCKETS;
  96 }
  97
  98 static struct nfqnl_instance *
  99 instance_lookup(struct nfnl_queue_net *q, u_int16_t queue_num)
 100 {
 101         struct hlist_head *head;
 102         struct nfqnl_instance *inst;
 103
 104         head = &q->instance_table[instance_hashfn(queue_num)];
 105         hlist_for_each_entry_rcu(inst, head, hlist) {
 106                 if (inst->queue_num == queue_num)
 107                         return inst;
 108         }
 109         return NULL;
 110 }
 111
 112 static struct nfqnl_instance *
 113 instance_create(struct nfnl_queue_net *q, u_int16_t queue_num, u32 portid)
 114 {
 115         struct nfqnl_instance *inst;
 116         unsigned int h;
 117         int err;
 118
 119         spin_lock(&q->instances_lock);
 120         if (instance_lookup(q, queue_num)) {
 121                 err = -EEXIST;
 122                 goto out_unlock;
 123         }
 124
 125         inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
 126         if (!inst) {
 127                 err = -ENOMEM;
 128                 goto out_unlock;
 129         }
 130
 131         inst->queue_num = queue_num;
 132         inst->peer_portid = portid;
 133         inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
 134         inst->copy_range = NFQNL_MAX_COPY_RANGE;
 135         inst->copy_mode = NFQNL_COPY_NONE;
 136         spin_lock_init(&inst->lock);
 137         INIT_LIST_HEAD(&inst->queue_list);
 138
 139         if (!try_module_get(THIS_MODULE)) {
 140                 err = -EAGAIN;
 141                 goto out_free;
 142         }
 143
 144         h = instance_hashfn(queue_num);
 145         hlist_add_head_rcu(&inst->hlist, &q->instance_table[h]);
 146
 147         spin_unlock(&q->instances_lock);
 148
 149         return inst;
 150
 151 out_free:
 152         kfree(inst);
 153 out_unlock:
 154         spin_unlock(&q->instances_lock);
 155         return ERR_PTR(err);
 156 }
 157
 158 static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
 159                         unsigned long data);
 160
 161 static void
 162 instance_destroy_rcu(struct rcu_head *head)
 163 {
 164         struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
 165                                                    rcu);
 166
 167         nfqnl_flush(inst, NULL, 0);
 168         kfree(inst);
 169         module_put(THIS_MODULE);
 170 }
 171
 172 static void
 173 __instance_destroy(struct nfqnl_instance *inst)
 174 {
 175         hlist_del_rcu(&inst->hlist);
 176         call_rcu(&inst->rcu, instance_destroy_rcu);
 177 }
 178
 179 static void
 180 instance_destroy(struct nfnl_queue_net *q, struct nfqnl_instance *inst)
 181 {
 182         spin_lock(&q->instances_lock);
 183         __instance_destroy(inst);
 184         spin_unlock(&q->instances_lock);
 185 }
 186
 187 static inline void
 188 __enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
 189 {
 190        list_add_tail(&entry->list, &queue->queue_list);
 191        queue->queue_total++;
 192 }
 193
 194 static void
 195 __dequeue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
 196 {
 197         list_del(&entry->list);
 198         queue->queue_total--;
 199 }
 200
 201 static struct nf_queue_entry *
 202 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
 203 {
 204         struct nf_queue_entry *entry = NULL, *i;
 205
 206         spin_lock_bh(&queue->lock);
 207
 208         list_for_each_entry(i, &queue->queue_list, list) {
 209                 if (i->id == id) {
 210                         entry = i;
 211                         break;
 212                 }
 213         }
 214
 215         if (entry)
 216                 __dequeue_entry(queue, entry);
 217
 218         spin_unlock_bh(&queue->lock);
 219
 220         return entry;
 221 }
 222
 223 static void
 224 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
 225 {
 226         struct nf_queue_entry *entry, *next;
 227
 228         spin_lock_bh(&queue->lock);
 229         list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
 230                 if (!cmpfn || cmpfn(entry, data)) {
 231                         list_del(&entry->list);
 232                         queue->queue_total--;
 233                         nf_reinject(entry, NF_DROP);
 234                 }
 235         }
 236         spin_unlock_bh(&queue->lock);
 237 }
 238
 239 static int
 240 nfqnl_put_packet_info(struct sk_buff *nlskb, struct sk_buff *packet,
 241                       bool csum_verify)
 242 {
 243         __u32 flags = 0;
 244
 245         if (packet->ip_summed == CHECKSUM_PARTIAL)
 246                 flags = NFQA_SKB_CSUMNOTREADY;
 247         else if (csum_verify)
 248                 flags = NFQA_SKB_CSUM_NOTVERIFIED;
 249
 250         if (skb_is_gso(packet))
 251                 flags |= NFQA_SKB_GSO;
 252
 253         return flags ? nla_put_be32(nlskb, NFQA_SKB_INFO, htonl(flags)) : 0;
 254 }
 255
 256 static int nfqnl_put_sk_uidgid(struct sk_buff *skb, struct sock *sk)
 257 {
 258         const struct cred *cred;
 259
 260         if (!sk_fullsock(sk))
 261                 return 0;
 262
 263         read_lock_bh(&sk->sk_callback_lock);
 264         if (sk->sk_socket && sk->sk_socket->file) {
 265                 cred = sk->sk_socket->file->f_cred;
 266                 if (nla_put_be32(skb, NFQA_UID,
 267                     htonl(from_kuid_munged(&init_user_ns, cred->fsuid))))
 268                         goto nla_put_failure;
 269                 if (nla_put_be32(skb, NFQA_GID,
 270                     htonl(from_kgid_munged(&init_user_ns, cred->fsgid))))
 271                         goto nla_put_failure;
 272         }
 273         read_unlock_bh(&sk->sk_callback_lock);
 274         return 0;
 275
 276 nla_put_failure:
 277         read_unlock_bh(&sk->sk_callback_lock);
 278         return -1;
 279 }
 280
 281 static u32 nfqnl_get_sk_secctx(struct sk_buff *skb, char **secdata)
 282 {
 283         u32 seclen = 0;
 284 #if IS_ENABLED(CONFIG_NETWORK_SECMARK)
 285         if (!skb || !sk_fullsock(skb->sk))
 286                 return 0;
 287
 288         read_lock_bh(&skb->sk->sk_callback_lock);
 289
 290         if (skb->secmark)
 291                 security_secid_to_secctx(skb->secmark, secdata, &seclen);
 292
 293         read_unlock_bh(&skb->sk->sk_callback_lock);
 294 #endif
 295         return seclen;
 296 }
 297
 298 static struct sk_buff *
 299 nfqnl_build_packet_message(struct net *net, struct nfqnl_instance *queue,
 300                            struct nf_queue_entry *entry,
 301                            __be32 **packet_id_ptr)
 302 {
 303         size_t size;
 304         size_t data_len = 0, cap_len = 0;
 305         unsigned int hlen = 0;
 306         struct sk_buff *skb;
 307         struct nlattr *nla;
 308         struct nfqnl_msg_packet_hdr *pmsg;
 309         struct nlmsghdr *nlh;
 310         struct nfgenmsg *nfmsg;
 311         struct sk_buff *entskb = entry->skb;
 312         struct net_device *indev;
 313         struct net_device *outdev;
 314         struct nf_conn *ct = NULL;
 315         enum ip_conntrack_info uninitialized_var(ctinfo);
 316         struct nfnl_ct_hook *nfnl_ct;
 317         bool csum_verify;
 318         char *secdata = NULL;
 319         u32 seclen = 0;
 320
 321         size =    nlmsg_total_size(sizeof(struct nfgenmsg))
 322                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
 323                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 324                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 325 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 326                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 327                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 328 #endif
 329                 + nla_total_size(sizeof(u_int32_t))     /* mark */
 330                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
 331                 + nla_total_size(sizeof(u_int32_t))     /* skbinfo */
 332                 + nla_total_size(sizeof(u_int32_t));    /* cap_len */
 333
 334         if (entskb->tstamp.tv64)
 335                 size += nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
 336
 337         if (entry->state.hook <= NF_INET_FORWARD ||
 338            (entry->state.hook == NF_INET_POST_ROUTING && entskb->sk == NULL))
 339                 csum_verify = !skb_csum_unnecessary(entskb);
 340         else
 341                 csum_verify = false;
 342
 343         outdev = entry->state.out;
 344
 345         switch ((enum nfqnl_config_mode)ACCESS_ONCE(queue->copy_mode)) {
 346         case NFQNL_COPY_META:
 347         case NFQNL_COPY_NONE:
 348                 break;
 349
 350         case NFQNL_COPY_PACKET:
 351                 if (!(queue->flags & NFQA_CFG_F_GSO) &&
 352                     entskb->ip_summed == CHECKSUM_PARTIAL &&
 353                     skb_checksum_help(entskb))
 354                         return NULL;
 355
 356                 data_len = ACCESS_ONCE(queue->copy_range);
 357                 if (data_len > entskb->len)
 358                         data_len = entskb->len;
 359
 360                 hlen = skb_zerocopy_headlen(entskb);
 361                 hlen = min_t(unsigned int, hlen, data_len);
 362                 size += sizeof(struct nlattr) + hlen;
 363                 cap_len = entskb->len;
 364                 break;
 365         }
 366
 367         nfnl_ct = rcu_dereference(nfnl_ct_hook);
 368
 369         if (queue->flags & NFQA_CFG_F_CONNTRACK) {
 370                 if (nfnl_ct != NULL) {
 371                         ct = nfnl_ct->get_ct(entskb, &ctinfo);
 372                         if (ct != NULL)
 373                                 size += nfnl_ct->build_size(ct);
 374                 }
 375         }
 376
 377         if (queue->flags & NFQA_CFG_F_UID_GID) {
 378                 size +=  (nla_total_size(sizeof(u_int32_t))     /* uid */
 379                         + nla_total_size(sizeof(u_int32_t)));   /* gid */
 380         }
 381
 382         if ((queue->flags & NFQA_CFG_F_SECCTX) && entskb->sk) {
 383                 seclen = nfqnl_get_sk_secctx(entskb, &secdata);
 384                 if (seclen)
 385                         size += nla_total_size(seclen);
 386         }
 387
 388         skb = alloc_skb(size, GFP_ATOMIC);
 389         if (!skb) {
 390                 skb_tx_error(entskb);
 391                 return NULL;
 392         }
 393
 394         nlh = nlmsg_put(skb, 0, 0,
 395                         NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
 396                         sizeof(struct nfgenmsg), 0);
 397         if (!nlh) {
 398                 skb_tx_error(entskb);
 399                 kfree_skb(skb);
 400                 return NULL;
 401         }
 402         nfmsg = nlmsg_data(nlh);
 403         nfmsg->nfgen_family = entry->state.pf;
 404         nfmsg->version = NFNETLINK_V0;
 405         nfmsg->res_id = htons(queue->queue_num);
 406
 407         nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
 408         pmsg = nla_data(nla);
 409         pmsg->hw_protocol       = entskb->protocol;
 410         pmsg->hook              = entry->state.hook;
 411         *packet_id_ptr          = &pmsg->packet_id;
 412
 413         indev = entry->state.in;
 414         if (indev) {
 415 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 416                 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
 417                         goto nla_put_failure;
 418 #else
 419                 if (entry->state.pf == PF_BRIDGE) {
 420                         /* Case 1: indev is physical input device, we need to
 421                          * look for bridge group (when called from
 422                          * netfilter_bridge) */
 423                         if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
 424                                          htonl(indev->ifindex)) ||
 425                         /* this is the bridge group "brX" */
 426                         /* rcu_read_lock()ed by __nf_queue */
 427                             nla_put_be32(skb, NFQA_IFINDEX_INDEV,
 428                                          htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
 429                                 goto nla_put_failure;
 430                 } else {
 431                         int physinif;
 432
 433                         /* Case 2: indev is bridge group, we need to look for
 434                          * physical device (when called from ipv4) */
 435                         if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
 436                                          htonl(indev->ifindex)))
 437                                 goto nla_put_failure;
 438
 439                         physinif = nf_bridge_get_physinif(entskb);
 440                         if (physinif &&
 441                             nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
 442                                          htonl(physinif)))
 443                                 goto nla_put_failure;
 444                 }
 445 #endif
 446         }
 447
 448         if (outdev) {
 449 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 450                 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
 451                         goto nla_put_failure;
 452 #else
 453                 if (entry->state.pf == PF_BRIDGE) {
 454                         /* Case 1: outdev is physical output device, we need to
 455                          * look for bridge group (when called from
 456                          * netfilter_bridge) */
 457                         if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 458                                          htonl(outdev->ifindex)) ||
 459                         /* this is the bridge group "brX" */
 460                         /* rcu_read_lock()ed by __nf_queue */
 461                             nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
 462                                          htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
 463                                 goto nla_put_failure;
 464                 } else {
 465                         int physoutif;
 466
 467                         /* Case 2: outdev is bridge group, we need to look for
 468                          * physical output device (when called from ipv4) */
 469                         if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
 470                                          htonl(outdev->ifindex)))
 471                                 goto nla_put_failure;
 472
 473                         physoutif = nf_bridge_get_physoutif(entskb);
 474                         if (physoutif &&
 475                             nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 476                                          htonl(physoutif)))
 477                                 goto nla_put_failure;
 478                 }
 479 #endif
 480         }
 481
 482         if (entskb->mark &&
 483             nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
 484                 goto nla_put_failure;
 485
 486         if (indev && entskb->dev &&
 487             entskb->mac_header != entskb->network_header) {
 488                 struct nfqnl_msg_packet_hw phw;
 489                 int len;
 490
 491                 memset(&phw, 0, sizeof(phw));
 492                 len = dev_parse_header(entskb, phw.hw_addr);
 493                 if (len) {
 494                         phw.hw_addrlen = htons(len);
 495                         if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
 496                                 goto nla_put_failure;
 497                 }
 498         }
 499
 500         if (entskb->tstamp.tv64) {
 501                 struct nfqnl_msg_packet_timestamp ts;
 502                 struct timespec64 kts = ktime_to_timespec64(skb->tstamp);
 503
 504                 ts.sec = cpu_to_be64(kts.tv_sec);
 505                 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
 506
 507                 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
 508                         goto nla_put_failure;
 509         }
 510
 511         if ((queue->flags & NFQA_CFG_F_UID_GID) && entskb->sk &&
 512             nfqnl_put_sk_uidgid(skb, entskb->sk) < 0)
 513                 goto nla_put_failure;
 514
 515         if (seclen && nla_put(skb, NFQA_SECCTX, seclen, secdata))
 516                 goto nla_put_failure;
 517
 518         if (ct && nfnl_ct->build(skb, ct, ctinfo, NFQA_CT, NFQA_CT_INFO) < 0)
 519                 goto nla_put_failure;
 520
 521         if (cap_len > data_len &&
 522             nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
 523                 goto nla_put_failure;
 524
 525         if (nfqnl_put_packet_info(skb, entskb, csum_verify))
 526                 goto nla_put_failure;
 527
 528         if (data_len) {
 529                 struct nlattr *nla;
 530
 531                 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
 532                         goto nla_put_failure;
 533
 534                 nla = (struct nlattr *)skb_put(skb, sizeof(*nla));
 535                 nla->nla_type = NFQA_PAYLOAD;
 536                 nla->nla_len = nla_attr_size(data_len);
 537
 538                 if (skb_zerocopy(skb, entskb, data_len, hlen))
 539                         goto nla_put_failure;
 540         }
 541
 542         nlh->nlmsg_len = skb->len;
 543         return skb;
 544
 545 nla_put_failure:
 546         skb_tx_error(entskb);
 547         kfree_skb(skb);
 548         net_err_ratelimited("nf_queue: error creating packet message\n");
 549         return NULL;
 550 }
 551
 552 static int
 553 __nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
 554                         struct nf_queue_entry *entry)
 555 {
 556         struct sk_buff *nskb;
 557         int err = -ENOBUFS;
 558         __be32 *packet_id_ptr;
 559         int failopen = 0;
 560
 561         nskb = nfqnl_build_packet_message(net, queue, entry, &packet_id_ptr);
 562         if (nskb == NULL) {
 563                 err = -ENOMEM;
 564                 goto err_out;
 565         }
 566         spin_lock_bh(&queue->lock);
 567
 568         if (queue->queue_total >= queue->queue_maxlen) {
 569                 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
 570                         failopen = 1;
 571                         err = 0;
 572                 } else {
 573                         queue->queue_dropped++;
 574                         net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
 575                                              queue->queue_total);
 576                 }
 577                 goto err_out_free_nskb;
 578         }
 579         entry->id = ++queue->id_sequence;
 580         *packet_id_ptr = htonl(entry->id);
 581
 582         /* nfnetlink_unicast will either free the nskb or add it to a socket */
 583         err = nfnetlink_unicast(nskb, net, queue->peer_portid, MSG_DONTWAIT);
 584         if (err < 0) {
 585                 queue->queue_user_dropped++;
 586                 goto err_out_unlock;
 587         }
 588
 589         __enqueue_entry(queue, entry);
 590
 591         spin_unlock_bh(&queue->lock);
 592         return 0;
 593
 594 err_out_free_nskb:
 595         kfree_skb(nskb);
 596 err_out_unlock:
 597         spin_unlock_bh(&queue->lock);
 598         if (failopen)
 599                 nf_reinject(entry, NF_ACCEPT);
 600 err_out:
 601         return err;
 602 }
 603
 604 static struct nf_queue_entry *
 605 nf_queue_entry_dup(struct nf_queue_entry *e)
 606 {
 607         struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
 608         if (entry)
 609                 nf_queue_entry_get_refs(entry);
 610         return entry;
 611 }
 612
 613 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 614 /* When called from bridge netfilter, skb->data must point to MAC header
 615  * before calling skb_gso_segment(). Else, original MAC header is lost
 616  * and segmented skbs will be sent to wrong destination.
 617  */
 618 static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
 619 {
 620         if (skb->nf_bridge)
 621                 __skb_push(skb, skb->network_header - skb->mac_header);
 622 }
 623
 624 static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
 625 {
 626         if (skb->nf_bridge)
 627                 __skb_pull(skb, skb->network_header - skb->mac_header);
 628 }
 629 #else
 630 #define nf_bridge_adjust_skb_data(s) do {} while (0)
 631 #define nf_bridge_adjust_segmented_data(s) do {} while (0)
 632 #endif
 633
 634 static void free_entry(struct nf_queue_entry *entry)
 635 {
 636         nf_queue_entry_release_refs(entry);
 637         kfree(entry);
 638 }
 639
 640 static int
 641 __nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
 642                            struct sk_buff *skb, struct nf_queue_entry *entry)
 643 {
 644         int ret = -ENOMEM;
 645         struct nf_queue_entry *entry_seg;
 646
 647         nf_bridge_adjust_segmented_data(skb);
 648
 649         if (skb->next == NULL) { /* last packet, no need to copy entry */
 650                 struct sk_buff *gso_skb = entry->skb;
 651                 entry->skb = skb;
 652                 ret = __nfqnl_enqueue_packet(net, queue, entry);
 653                 if (ret)
 654                         entry->skb = gso_skb;
 655                 return ret;
 656         }
 657
 658         skb->next = NULL;
 659
 660         entry_seg = nf_queue_entry_dup(entry);
 661         if (entry_seg) {
 662                 entry_seg->skb = skb;
 663                 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
 664                 if (ret)
 665                         free_entry(entry_seg);
 666         }
 667         return ret;
 668 }
 669
 670 static int
 671 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 672 {
 673         unsigned int queued;
 674         struct nfqnl_instance *queue;
 675         struct sk_buff *skb, *segs;
 676         int err = -ENOBUFS;
 677         struct net *net = entry->state.net;
 678         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
 679
 680         /* rcu_read_lock()ed by nf_hook_slow() */
 681         queue = instance_lookup(q, queuenum);
 682         if (!queue)
 683                 return -ESRCH;
 684
 685         if (queue->copy_mode == NFQNL_COPY_NONE)
 686                 return -EINVAL;
 687
 688         skb = entry->skb;
 689
 690         switch (entry->state.pf) {
 691         case NFPROTO_IPV4:
 692                 skb->protocol = htons(ETH_P_IP);
 693                 break;
 694         case NFPROTO_IPV6:
 695                 skb->protocol = htons(ETH_P_IPV6);
 696                 break;
 697         }
 698
 699         if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(skb))
 700                 return __nfqnl_enqueue_packet(net, queue, entry);
 701
 702         nf_bridge_adjust_skb_data(skb);
 703         segs = skb_gso_segment(skb, 0);
 704         /* Does not use PTR_ERR to limit the number of error codes that can be
 705          * returned by nf_queue.  For instance, callers rely on -ESRCH to
 706          * mean 'ignore this hook'.
 707          */
 708         if (IS_ERR_OR_NULL(segs))
 709                 goto out_err;
 710         queued = 0;
 711         err = 0;
 712         do {
 713                 struct sk_buff *nskb = segs->next;
 714                 if (err == 0)
 715                         err = __nfqnl_enqueue_packet_gso(net, queue,
 716                                                         segs, entry);
 717                 if (err == 0)
 718                         queued++;
 719                 else
 720                         kfree_skb(segs);
 721                 segs = nskb;
 722         } while (segs);
 723
 724         if (queued) {
 725                 if (err) /* some segments are already queued */
 726                         free_entry(entry);
 727                 kfree_skb(skb);
 728                 return 0;
 729         }
 730  out_err:
 731         nf_bridge_adjust_segmented_data(skb);
 732         return err;
 733 }
 734
 735 static int
 736 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e, int diff)
 737 {
 738         struct sk_buff *nskb;
 739
 740         if (diff < 0) {
 741                 if (pskb_trim(e->skb, data_len))
 742                         return -ENOMEM;
 743         } else if (diff > 0) {
 744                 if (data_len > 0xFFFF)
 745                         return -EINVAL;
 746                 if (diff > skb_tailroom(e->skb)) {
 747                         nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
 748                                                diff, GFP_ATOMIC);
 749                         if (!nskb) {
 750                                 printk(KERN_WARNING "nf_queue: OOM "
 751                                       "in mangle, dropping packet\n");
 752                                 return -ENOMEM;
 753                         }
 754                         kfree_skb(e->skb);
 755                         e->skb = nskb;
 756                 }
 757                 skb_put(e->skb, diff);
 758         }
 759         if (!skb_make_writable(e->skb, data_len))
 760                 return -ENOMEM;
 761         skb_copy_to_linear_data(e->skb, data, data_len);
 762         e->skb->ip_summed = CHECKSUM_NONE;
 763         return 0;
 764 }
 765
 766 static int
 767 nfqnl_set_mode(struct nfqnl_instance *queue,
 768                unsigned char mode, unsigned int range)
 769 {
 770         int status = 0;
 771
 772         spin_lock_bh(&queue->lock);
 773         switch (mode) {
 774         case NFQNL_COPY_NONE:
 775         case NFQNL_COPY_META:
 776                 queue->copy_mode = mode;
 777                 queue->copy_range = 0;
 778                 break;
 779
 780         case NFQNL_COPY_PACKET:
 781                 queue->copy_mode = mode;
 782                 if (range == 0 || range > NFQNL_MAX_COPY_RANGE)
 783                         queue->copy_range = NFQNL_MAX_COPY_RANGE;
 784                 else
 785                         queue->copy_range = range;
 786                 break;
 787
 788         default:
 789                 status = -EINVAL;
 790
 791         }
 792         spin_unlock_bh(&queue->lock);
 793
 794         return status;
 795 }
 796
 797 static int
 798 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
 799 {
 800         if (entry->state.in)
 801                 if (entry->state.in->ifindex == ifindex)
 802                         return 1;
 803         if (entry->state.out)
 804                 if (entry->state.out->ifindex == ifindex)
 805                         return 1;
 806 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 807         if (entry->skb->nf_bridge) {
 808                 int physinif, physoutif;
 809
 810                 physinif = nf_bridge_get_physinif(entry->skb);
 811                 physoutif = nf_bridge_get_physoutif(entry->skb);
 812
 813                 if (physinif == ifindex || physoutif == ifindex)
 814                         return 1;
 815         }
 816 #endif
 817         return 0;
 818 }
 819
 820 /* drop all packets with either indev or outdev == ifindex from all queue
 821  * instances */
 822 static void
 823 nfqnl_dev_drop(struct net *net, int ifindex)
 824 {
 825         int i;
 826         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
 827
 828         rcu_read_lock();
 829
 830         for (i = 0; i < INSTANCE_BUCKETS; i++) {
 831                 struct nfqnl_instance *inst;
 832                 struct hlist_head *head = &q->instance_table[i];
 833
 834                 hlist_for_each_entry_rcu(inst, head, hlist)
 835                         nfqnl_flush(inst, dev_cmp, ifindex);
 836         }
 837
 838         rcu_read_unlock();
 839 }
 840
 841 static int
 842 nfqnl_rcv_dev_event(struct notifier_block *this,
 843                     unsigned long event, void *ptr)
 844 {
 845         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
 846
 847         /* Drop any packets associated with the downed device */
 848         if (event == NETDEV_DOWN)
 849                 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
 850         return NOTIFY_DONE;
 851 }
 852
 853 static struct notifier_block nfqnl_dev_notifier = {
 854         .notifier_call  = nfqnl_rcv_dev_event,
 855 };
 856
 857 static int nf_hook_cmp(struct nf_queue_entry *entry, unsigned long ops_ptr)
 858 {
 859         return entry->elem == (struct nf_hook_ops *)ops_ptr;
 860 }
 861
 862 static void nfqnl_nf_hook_drop(struct net *net, struct nf_hook_ops *hook)
 863 {
 864         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
 865         int i;
 866
 867         rcu_read_lock();
 868         for (i = 0; i < INSTANCE_BUCKETS; i++) {
 869                 struct nfqnl_instance *inst;
 870                 struct hlist_head *head = &q->instance_table[i];
 871
 872                 hlist_for_each_entry_rcu(inst, head, hlist)
 873                         nfqnl_flush(inst, nf_hook_cmp, (unsigned long)hook);
 874         }
 875         rcu_read_unlock();
 876 }
 877
 878 static int
 879 nfqnl_rcv_nl_event(struct notifier_block *this,
 880                    unsigned long event, void *ptr)
 881 {
 882         struct netlink_notify *n = ptr;
 883         struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
 884
 885         if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
 886                 int i;
 887
 888                 /* destroy all instances for this portid */
 889                 spin_lock(&q->instances_lock);
 890                 for (i = 0; i < INSTANCE_BUCKETS; i++) {
 891                         struct hlist_node *t2;
 892                         struct nfqnl_instance *inst;
 893                         struct hlist_head *head = &q->instance_table[i];
 894
 895                         hlist_for_each_entry_safe(inst, t2, head, hlist) {
 896                                 if (n->portid == inst->peer_portid)
 897                                         __instance_destroy(inst);
 898                         }
 899                 }
 900                 spin_unlock(&q->instances_lock);
 901         }
 902         return NOTIFY_DONE;
 903 }
 904
 905 static struct notifier_block nfqnl_rtnl_notifier = {
 906         .notifier_call  = nfqnl_rcv_nl_event,
 907 };
 908
 909 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
 910         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 911         [NFQA_MARK]             = { .type = NLA_U32 },
 912         [NFQA_PAYLOAD]          = { .type = NLA_UNSPEC },
 913         [NFQA_CT]               = { .type = NLA_UNSPEC },
 914         [NFQA_EXP]              = { .type = NLA_UNSPEC },
 915 };
 916
 917 static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
 918         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 919         [NFQA_MARK]             = { .type = NLA_U32 },
 920 };
 921
 922 static struct nfqnl_instance *
 923 verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, u32 nlportid)
 924 {
 925         struct nfqnl_instance *queue;
 926
 927         queue = instance_lookup(q, queue_num);
 928         if (!queue)
 929                 return ERR_PTR(-ENODEV);
 930
 931         if (queue->peer_portid != nlportid)
 932                 return ERR_PTR(-EPERM);
 933
 934         return queue;
 935 }
 936
 937 static struct nfqnl_msg_verdict_hdr*
 938 verdicthdr_get(const struct nlattr * const nfqa[])
 939 {
 940         struct nfqnl_msg_verdict_hdr *vhdr;
 941         unsigned int verdict;
 942
 943         if (!nfqa[NFQA_VERDICT_HDR])
 944                 return NULL;
 945
 946         vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
 947         verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
 948         if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
 949                 return NULL;
 950         return vhdr;
 951 }
 952
 953 static int nfq_id_after(unsigned int id, unsigned int max)
 954 {
 955         return (int)(id - max) > 0;
 956 }
 957
 958 static int nfqnl_recv_verdict_batch(struct net *net, struct sock *ctnl,
 959                                     struct sk_buff *skb,
 960                                     const struct nlmsghdr *nlh,
 961                                     const struct nlattr * const nfqa[])
 962 {
 963         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
 964         struct nf_queue_entry *entry, *tmp;
 965         unsigned int verdict, maxid;
 966         struct nfqnl_msg_verdict_hdr *vhdr;
 967         struct nfqnl_instance *queue;
 968         LIST_HEAD(batch_list);
 969         u16 queue_num = ntohs(nfmsg->res_id);
 970         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
 971
 972         queue = verdict_instance_lookup(q, queue_num,
 973                                         NETLINK_CB(skb).portid);
 974         if (IS_ERR(queue))
 975                 return PTR_ERR(queue);
 976
 977         vhdr = verdicthdr_get(nfqa);
 978         if (!vhdr)
 979                 return -EINVAL;
 980
 981         verdict = ntohl(vhdr->verdict);
 982         maxid = ntohl(vhdr->id);
 983
 984         spin_lock_bh(&queue->lock);
 985
 986         list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
 987                 if (nfq_id_after(entry->id, maxid))
 988                         break;
 989                 __dequeue_entry(queue, entry);
 990                 list_add_tail(&entry->list, &batch_list);
 991         }
 992
 993         spin_unlock_bh(&queue->lock);
 994
 995         if (list_empty(&batch_list))
 996                 return -ENOENT;
 997
 998         list_for_each_entry_safe(entry, tmp, &batch_list, list) {
 999                 if (nfqa[NFQA_MARK])
1000                         entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1001                 nf_reinject(entry, verdict);
1002         }
1003         return 0;
1004 }
1005
1006 static struct nf_conn *nfqnl_ct_parse(struct nfnl_ct_hook *nfnl_ct,
1007                                       const struct nlmsghdr *nlh,
1008                                       const struct nlattr * const nfqa[],
1009                                       struct nf_queue_entry *entry,
1010                                       enum ip_conntrack_info *ctinfo)
1011 {
1012         struct nf_conn *ct;
1013
1014         ct = nfnl_ct->get_ct(entry->skb, ctinfo);
1015         if (ct == NULL)
1016                 return NULL;
1017
1018         if (nfnl_ct->parse(nfqa[NFQA_CT], ct) < 0)
1019                 return NULL;
1020
1021         if (nfqa[NFQA_EXP])
1022                 nfnl_ct->attach_expect(nfqa[NFQA_EXP], ct,
1023                                       NETLINK_CB(entry->skb).portid,
1024                                       nlmsg_report(nlh));
1025         return ct;
1026 }
1027
1028 static int nfqnl_recv_verdict(struct net *net, struct sock *ctnl,
1029                               struct sk_buff *skb,
1030                               const struct nlmsghdr *nlh,
1031                               const struct nlattr * const nfqa[])
1032 {
1033         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1034         u_int16_t queue_num = ntohs(nfmsg->res_id);
1035         struct nfqnl_msg_verdict_hdr *vhdr;
1036         struct nfqnl_instance *queue;
1037         unsigned int verdict;
1038         struct nf_queue_entry *entry;
1039         enum ip_conntrack_info uninitialized_var(ctinfo);
1040         struct nfnl_ct_hook *nfnl_ct;
1041         struct nf_conn *ct = NULL;
1042         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1043
1044         queue = instance_lookup(q, queue_num);
1045         if (!queue)
1046                 queue = verdict_instance_lookup(q, queue_num,
1047                                                 NETLINK_CB(skb).portid);
1048         if (IS_ERR(queue))
1049                 return PTR_ERR(queue);
1050
1051         vhdr = verdicthdr_get(nfqa);
1052         if (!vhdr)
1053                 return -EINVAL;
1054
1055         verdict = ntohl(vhdr->verdict);
1056
1057         entry = find_dequeue_entry(queue, ntohl(vhdr->id));
1058         if (entry == NULL)
1059                 return -ENOENT;
1060
1061         /* rcu lock already held from nfnl->call_rcu. */
1062         nfnl_ct = rcu_dereference(nfnl_ct_hook);
1063
1064         if (nfqa[NFQA_CT]) {
1065                 if (nfnl_ct != NULL)
1066                         ct = nfqnl_ct_parse(nfnl_ct, nlh, nfqa, entry, &ctinfo);
1067         }
1068
1069         if (nfqa[NFQA_PAYLOAD]) {
1070                 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
1071                 int diff = payload_len - entry->skb->len;
1072
1073                 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
1074                                  payload_len, entry, diff) < 0)
1075                         verdict = NF_DROP;
1076
1077                 if (ct && diff)
1078                         nfnl_ct->seq_adjust(entry->skb, ct, ctinfo, diff);
1079         }
1080
1081         if (nfqa[NFQA_MARK])
1082                 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1083
1084         nf_reinject(entry, verdict);
1085         return 0;
1086 }
1087
1088 static int nfqnl_recv_unsupp(struct net *net, struct sock *ctnl,
1089                              struct sk_buff *skb, const struct nlmsghdr *nlh,
1090                              const struct nlattr * const nfqa[])
1091 {
1092         return -ENOTSUPP;
1093 }
1094
1095 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1096         [NFQA_CFG_CMD]          = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1097         [NFQA_CFG_PARAMS]       = { .len = sizeof(struct nfqnl_msg_config_params) },
1098 };
1099
1100 static const struct nf_queue_handler nfqh = {
1101         .outfn          = &nfqnl_enqueue_packet,
1102         .nf_hook_drop   = &nfqnl_nf_hook_drop,
1103 };
1104
1105 static int nfqnl_recv_config(struct net *net, struct sock *ctnl,
1106                              struct sk_buff *skb, const struct nlmsghdr *nlh,
1107                              const struct nlattr * const nfqa[])
1108 {
1109         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1110         u_int16_t queue_num = ntohs(nfmsg->res_id);
1111         struct nfqnl_instance *queue;
1112         struct nfqnl_msg_config_cmd *cmd = NULL;
1113         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1114         __u32 flags = 0, mask = 0;
1115         int ret = 0;
1116
1117         if (nfqa[NFQA_CFG_CMD]) {
1118                 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1119
1120                 /* Obsolete commands without queue context */
1121                 switch (cmd->command) {
1122                 case NFQNL_CFG_CMD_PF_BIND: return 0;
1123                 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1124                 }
1125         }
1126
1127         /* Check if we support these flags in first place, dependencies should
1128          * be there too not to break atomicity.
1129          */
1130         if (nfqa[NFQA_CFG_FLAGS]) {
1131                 if (!nfqa[NFQA_CFG_MASK]) {
1132                         /* A mask is needed to specify which flags are being
1133                          * changed.
1134                          */
1135                         return -EINVAL;
1136                 }
1137
1138                 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1139                 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1140
1141                 if (flags >= NFQA_CFG_F_MAX)
1142                         return -EOPNOTSUPP;
1143
1144 #if !IS_ENABLED(CONFIG_NETWORK_SECMARK)
1145                 if (flags & mask & NFQA_CFG_F_SECCTX)
1146                         return -EOPNOTSUPP;
1147 #endif
1148                 if ((flags & mask & NFQA_CFG_F_CONNTRACK) &&
1149                     !rcu_access_pointer(nfnl_ct_hook)) {
1150 #ifdef CONFIG_MODULES
1151                         nfnl_unlock(NFNL_SUBSYS_QUEUE);
1152                         request_module("ip_conntrack_netlink");
1153                         nfnl_lock(NFNL_SUBSYS_QUEUE);
1154                         if (rcu_access_pointer(nfnl_ct_hook))
1155                                 return -EAGAIN;
1156 #endif
1157                         return -EOPNOTSUPP;
1158                 }
1159         }
1160
1161         rcu_read_lock();
1162         queue = instance_lookup(q, queue_num);
1163         if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1164                 ret = -EPERM;
1165                 goto err_out_unlock;
1166         }
1167
1168         if (cmd != NULL) {
1169                 switch (cmd->command) {
1170                 case NFQNL_CFG_CMD_BIND:
1171                         if (queue) {
1172                                 ret = -EBUSY;
1173                                 goto err_out_unlock;
1174                         }
1175                         queue = instance_create(q, queue_num,
1176                                                 NETLINK_CB(skb).portid);
1177                         if (IS_ERR(queue)) {
1178                                 ret = PTR_ERR(queue);
1179                                 goto err_out_unlock;
1180                         }
1181                         break;
1182                 case NFQNL_CFG_CMD_UNBIND:
1183                         if (!queue) {
1184                                 ret = -ENODEV;
1185                                 goto err_out_unlock;
1186                         }
1187                         instance_destroy(q, queue);
1188                         goto err_out_unlock;
1189                 case NFQNL_CFG_CMD_PF_BIND:
1190                 case NFQNL_CFG_CMD_PF_UNBIND:
1191                         break;
1192                 default:
1193                         ret = -ENOTSUPP;
1194                         goto err_out_unlock;
1195                 }
1196         }
1197
1198         if (!queue) {
1199                 ret = -ENODEV;
1200                 goto err_out_unlock;
1201         }
1202
1203         if (nfqa[NFQA_CFG_PARAMS]) {
1204                 struct nfqnl_msg_config_params *params =
1205                         nla_data(nfqa[NFQA_CFG_PARAMS]);
1206
1207                 nfqnl_set_mode(queue, params->copy_mode,
1208                                 ntohl(params->copy_range));
1209         }
1210
1211         if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1212                 __be32 *queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1213
1214                 spin_lock_bh(&queue->lock);
1215                 queue->queue_maxlen = ntohl(*queue_maxlen);
1216                 spin_unlock_bh(&queue->lock);
1217         }
1218
1219         if (nfqa[NFQA_CFG_FLAGS]) {
1220                 spin_lock_bh(&queue->lock);
1221                 queue->flags &= ~mask;
1222                 queue->flags |= flags & mask;
1223                 spin_unlock_bh(&queue->lock);
1224         }
1225
1226 err_out_unlock:
1227         rcu_read_unlock();
1228         return ret;
1229 }
1230
1231 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1232         [NFQNL_MSG_PACKET]      = { .call_rcu = nfqnl_recv_unsupp,
1233                                     .attr_count = NFQA_MAX, },
1234         [NFQNL_MSG_VERDICT]     = { .call_rcu = nfqnl_recv_verdict,
1235                                     .attr_count = NFQA_MAX,
1236                                     .policy = nfqa_verdict_policy },
1237         [NFQNL_MSG_CONFIG]      = { .call = nfqnl_recv_config,
1238                                     .attr_count = NFQA_CFG_MAX,
1239                                     .policy = nfqa_cfg_policy },
1240         [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
1241                                     .attr_count = NFQA_MAX,
1242                                     .policy = nfqa_verdict_batch_policy },
1243 };
1244
1245 static const struct nfnetlink_subsystem nfqnl_subsys = {
1246         .name           = "nf_queue",
1247         .subsys_id      = NFNL_SUBSYS_QUEUE,
1248         .cb_count       = NFQNL_MSG_MAX,
1249         .cb             = nfqnl_cb,
1250 };
1251
1252 #ifdef CONFIG_PROC_FS
1253 struct iter_state {
1254         struct seq_net_private p;
1255         unsigned int bucket;
1256 };
1257
1258 static struct hlist_node *get_first(struct seq_file *seq)
1259 {
1260         struct iter_state *st = seq->private;
1261         struct net *net;
1262         struct nfnl_queue_net *q;
1263
1264         if (!st)
1265                 return NULL;
1266
1267         net = seq_file_net(seq);
1268         q = nfnl_queue_pernet(net);
1269         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1270                 if (!hlist_empty(&q->instance_table[st->bucket]))
1271                         return q->instance_table[st->bucket].first;
1272         }
1273         return NULL;
1274 }
1275
1276 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1277 {
1278         struct iter_state *st = seq->private;
1279         struct net *net = seq_file_net(seq);
1280
1281         h = h->next;
1282         while (!h) {
1283                 struct nfnl_queue_net *q;
1284
1285                 if (++st->bucket >= INSTANCE_BUCKETS)
1286                         return NULL;
1287
1288                 q = nfnl_queue_pernet(net);
1289                 h = q->instance_table[st->bucket].first;
1290         }
1291         return h;
1292 }
1293
1294 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1295 {
1296         struct hlist_node *head;
1297         head = get_first(seq);
1298
1299         if (head)
1300                 while (pos && (head = get_next(seq, head)))
1301                         pos--;
1302         return pos ? NULL : head;
1303 }
1304
1305 static void *seq_start(struct seq_file *s, loff_t *pos)
1306         __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1307 {
1308         spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1309         return get_idx(s, *pos);
1310 }
1311
1312 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1313 {
1314         (*pos)++;
1315         return get_next(s, v);
1316 }
1317
1318 static void seq_stop(struct seq_file *s, void *v)
1319         __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1320 {
1321         spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1322 }
1323
1324 static int seq_show(struct seq_file *s, void *v)
1325 {
1326         const struct nfqnl_instance *inst = v;
1327
1328         seq_printf(s, "%5u %6u %5u %1u %5u %5u %5u %8u %2d\n",
1329                    inst->queue_num,
1330                    inst->peer_portid, inst->queue_total,
1331                    inst->copy_mode, inst->copy_range,
1332                    inst->queue_dropped, inst->queue_user_dropped,
1333                    inst->id_sequence, 1);
1334         return 0;
1335 }
1336
1337 static const struct seq_operations nfqnl_seq_ops = {
1338         .start  = seq_start,
1339         .next   = seq_next,
1340         .stop   = seq_stop,
1341         .show   = seq_show,
1342 };
1343
1344 static int nfqnl_open(struct inode *inode, struct file *file)
1345 {
1346         return seq_open_net(inode, file, &nfqnl_seq_ops,
1347                         sizeof(struct iter_state));
1348 }
1349
1350 static const struct file_operations nfqnl_file_ops = {
1351         .owner   = THIS_MODULE,
1352         .open    = nfqnl_open,
1353         .read    = seq_read,
1354         .llseek  = seq_lseek,
1355         .release = seq_release_net,
1356 };
1357
1358 #endif /* PROC_FS */
1359
1360 static int __net_init nfnl_queue_net_init(struct net *net)
1361 {
1362         unsigned int i;
1363         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1364
1365         for (i = 0; i < INSTANCE_BUCKETS; i++)
1366                 INIT_HLIST_HEAD(&q->instance_table[i]);
1367
1368         spin_lock_init(&q->instances_lock);
1369
1370 #ifdef CONFIG_PROC_FS
1371         if (!proc_create("nfnetlink_queue", 0440,
1372                          net->nf.proc_netfilter, &nfqnl_file_ops))
1373                 return -ENOMEM;
1374 #endif
1375         return 0;
1376 }
1377
1378 static void __net_exit nfnl_queue_net_exit(struct net *net)
1379 {
1380 #ifdef CONFIG_PROC_FS
1381         remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1382 #endif
1383 }
1384
1385 static struct pernet_operations nfnl_queue_net_ops = {
1386         .init   = nfnl_queue_net_init,
1387         .exit   = nfnl_queue_net_exit,
1388         .id     = &nfnl_queue_net_id,
1389         .size   = sizeof(struct nfnl_queue_net),
1390 };
1391
1392 static int __init nfnetlink_queue_init(void)
1393 {
1394         int status;
1395
1396         status = register_pernet_subsys(&nfnl_queue_net_ops);
1397         if (status < 0) {
1398                 pr_err("nf_queue: failed to register pernet ops\n");
1399                 goto out;
1400         }
1401
1402         netlink_register_notifier(&nfqnl_rtnl_notifier);
1403         status = nfnetlink_subsys_register(&nfqnl_subsys);
1404         if (status < 0) {
1405                 pr_err("nf_queue: failed to create netlink socket\n");
1406                 goto cleanup_netlink_notifier;
1407         }
1408
1409         register_netdevice_notifier(&nfqnl_dev_notifier);
1410         nf_register_queue_handler(&nfqh);
1411         return status;
1412
1413 cleanup_netlink_notifier:
1414         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1415         unregister_pernet_subsys(&nfnl_queue_net_ops);
1416 out:
1417         return status;
1418 }
1419
1420 static void __exit nfnetlink_queue_fini(void)
1421 {
1422         nf_unregister_queue_handler();
1423         unregister_netdevice_notifier(&nfqnl_dev_notifier);
1424         nfnetlink_subsys_unregister(&nfqnl_subsys);
1425         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1426         unregister_pernet_subsys(&nfnl_queue_net_ops);
1427
1428         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1429 }
1430
1431 MODULE_DESCRIPTION("netfilter packet queue handler");
1432 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1433 MODULE_LICENSE("GPL");
1434 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1435
1436 module_init(nfnetlink_queue_init);
1437 module_exit(nfnetlink_queue_fini);