net/netfilter/nfnetlink_queue_core.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/list.h>
  32 #include <net/sock.h>
  33 #include <net/tcp_states.h>
  34 #include <net/netfilter/nf_queue.h>
  35 #include <net/netns/generic.h>
  36 #include <net/netfilter/nfnetlink_queue.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, rem_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         bool csum_verify;
 317         char *secdata = NULL;
 318         u32 seclen = 0;
 319
 320         size =    nlmsg_total_size(sizeof(struct nfgenmsg))
 321                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
 322                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 323                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 324 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 325                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 326                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 327 #endif
 328                 + nla_total_size(sizeof(u_int32_t))     /* mark */
 329                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
 330                 + nla_total_size(sizeof(u_int32_t))     /* skbinfo */
 331                 + nla_total_size(sizeof(u_int32_t));    /* cap_len */
 332
 333         if (entskb->tstamp.tv64)
 334                 size += nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
 335
 336         if (entry->state.hook <= NF_INET_FORWARD ||
 337            (entry->state.hook == NF_INET_POST_ROUTING && entskb->sk == NULL))
 338                 csum_verify = !skb_csum_unnecessary(entskb);
 339         else
 340                 csum_verify = false;
 341
 342         outdev = entry->state.out;
 343
 344         switch ((enum nfqnl_config_mode)ACCESS_ONCE(queue->copy_mode)) {
 345         case NFQNL_COPY_META:
 346         case NFQNL_COPY_NONE:
 347                 break;
 348
 349         case NFQNL_COPY_PACKET:
 350                 if (!(queue->flags & NFQA_CFG_F_GSO) &&
 351                     entskb->ip_summed == CHECKSUM_PARTIAL &&
 352                     skb_checksum_help(entskb))
 353                         return NULL;
 354
 355                 data_len = ACCESS_ONCE(queue->copy_range);
 356                 if (data_len > entskb->len)
 357                         data_len = entskb->len;
 358
 359                 hlen = skb_zerocopy_headlen(entskb);
 360                 hlen = min_t(unsigned int, hlen, data_len);
 361                 size += sizeof(struct nlattr) + hlen;
 362                 cap_len = entskb->len;
 363                 rem_len = data_len - hlen;
 364                 break;
 365         }
 366
 367         if (queue->flags & NFQA_CFG_F_CONNTRACK)
 368                 ct = nfqnl_ct_get(entskb, &size, &ctinfo);
 369
 370         if (queue->flags & NFQA_CFG_F_UID_GID) {
 371                 size +=  (nla_total_size(sizeof(u_int32_t))     /* uid */
 372                         + nla_total_size(sizeof(u_int32_t)));   /* gid */
 373         }
 374
 375         if ((queue->flags & NFQA_CFG_F_SECCTX) && entskb->sk) {
 376                 seclen = nfqnl_get_sk_secctx(entskb, &secdata);
 377                 if (seclen)
 378                         size += nla_total_size(seclen);
 379         }
 380
 381         skb = __netlink_alloc_skb(net->nfnl, size, rem_len, queue->peer_portid,
 382                                   GFP_ATOMIC);
 383         if (!skb) {
 384                 skb_tx_error(entskb);
 385                 return NULL;
 386         }
 387
 388         nlh = nlmsg_put(skb, 0, 0,
 389                         NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
 390                         sizeof(struct nfgenmsg), 0);
 391         if (!nlh) {
 392                 skb_tx_error(entskb);
 393                 kfree_skb(skb);
 394                 return NULL;
 395         }
 396         nfmsg = nlmsg_data(nlh);
 397         nfmsg->nfgen_family = entry->state.pf;
 398         nfmsg->version = NFNETLINK_V0;
 399         nfmsg->res_id = htons(queue->queue_num);
 400
 401         nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
 402         pmsg = nla_data(nla);
 403         pmsg->hw_protocol       = entskb->protocol;
 404         pmsg->hook              = entry->state.hook;
 405         *packet_id_ptr          = &pmsg->packet_id;
 406
 407         indev = entry->state.in;
 408         if (indev) {
 409 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 410                 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
 411                         goto nla_put_failure;
 412 #else
 413                 if (entry->state.pf == PF_BRIDGE) {
 414                         /* Case 1: indev is physical input device, we need to
 415                          * look for bridge group (when called from
 416                          * netfilter_bridge) */
 417                         if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
 418                                          htonl(indev->ifindex)) ||
 419                         /* this is the bridge group "brX" */
 420                         /* rcu_read_lock()ed by __nf_queue */
 421                             nla_put_be32(skb, NFQA_IFINDEX_INDEV,
 422                                          htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
 423                                 goto nla_put_failure;
 424                 } else {
 425                         int physinif;
 426
 427                         /* Case 2: indev is bridge group, we need to look for
 428                          * physical device (when called from ipv4) */
 429                         if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
 430                                          htonl(indev->ifindex)))
 431                                 goto nla_put_failure;
 432
 433                         physinif = nf_bridge_get_physinif(entskb);
 434                         if (physinif &&
 435                             nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
 436                                          htonl(physinif)))
 437                                 goto nla_put_failure;
 438                 }
 439 #endif
 440         }
 441
 442         if (outdev) {
 443 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 444                 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
 445                         goto nla_put_failure;
 446 #else
 447                 if (entry->state.pf == PF_BRIDGE) {
 448                         /* Case 1: outdev is physical output device, we need to
 449                          * look for bridge group (when called from
 450                          * netfilter_bridge) */
 451                         if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 452                                          htonl(outdev->ifindex)) ||
 453                         /* this is the bridge group "brX" */
 454                         /* rcu_read_lock()ed by __nf_queue */
 455                             nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
 456                                          htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
 457                                 goto nla_put_failure;
 458                 } else {
 459                         int physoutif;
 460
 461                         /* Case 2: outdev is bridge group, we need to look for
 462                          * physical output device (when called from ipv4) */
 463                         if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
 464                                          htonl(outdev->ifindex)))
 465                                 goto nla_put_failure;
 466
 467                         physoutif = nf_bridge_get_physoutif(entskb);
 468                         if (physoutif &&
 469                             nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 470                                          htonl(physoutif)))
 471                                 goto nla_put_failure;
 472                 }
 473 #endif
 474         }
 475
 476         if (entskb->mark &&
 477             nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
 478                 goto nla_put_failure;
 479
 480         if (indev && entskb->dev &&
 481             entskb->mac_header != entskb->network_header) {
 482                 struct nfqnl_msg_packet_hw phw;
 483                 int len;
 484
 485                 memset(&phw, 0, sizeof(phw));
 486                 len = dev_parse_header(entskb, phw.hw_addr);
 487                 if (len) {
 488                         phw.hw_addrlen = htons(len);
 489                         if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
 490                                 goto nla_put_failure;
 491                 }
 492         }
 493
 494         if (entskb->tstamp.tv64) {
 495                 struct nfqnl_msg_packet_timestamp ts;
 496                 struct timeval tv = ktime_to_timeval(entskb->tstamp);
 497                 ts.sec = cpu_to_be64(tv.tv_sec);
 498                 ts.usec = cpu_to_be64(tv.tv_usec);
 499
 500                 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
 501                         goto nla_put_failure;
 502         }
 503
 504         if ((queue->flags & NFQA_CFG_F_UID_GID) && entskb->sk &&
 505             nfqnl_put_sk_uidgid(skb, entskb->sk) < 0)
 506                 goto nla_put_failure;
 507
 508         if (seclen && nla_put(skb, NFQA_SECCTX, seclen, secdata))
 509                 goto nla_put_failure;
 510
 511         if (ct && nfqnl_ct_put(skb, ct, ctinfo) < 0)
 512                 goto nla_put_failure;
 513
 514         if (cap_len > data_len &&
 515             nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
 516                 goto nla_put_failure;
 517
 518         if (nfqnl_put_packet_info(skb, entskb, csum_verify))
 519                 goto nla_put_failure;
 520
 521         if (data_len) {
 522                 struct nlattr *nla;
 523
 524                 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
 525                         goto nla_put_failure;
 526
 527                 nla = (struct nlattr *)skb_put(skb, sizeof(*nla));
 528                 nla->nla_type = NFQA_PAYLOAD;
 529                 nla->nla_len = nla_attr_size(data_len);
 530
 531                 if (skb_zerocopy(skb, entskb, data_len, hlen))
 532                         goto nla_put_failure;
 533         }
 534
 535         nlh->nlmsg_len = skb->len;
 536         return skb;
 537
 538 nla_put_failure:
 539         skb_tx_error(entskb);
 540         kfree_skb(skb);
 541         net_err_ratelimited("nf_queue: error creating packet message\n");
 542         return NULL;
 543 }
 544
 545 static int
 546 __nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
 547                         struct nf_queue_entry *entry)
 548 {
 549         struct sk_buff *nskb;
 550         int err = -ENOBUFS;
 551         __be32 *packet_id_ptr;
 552         int failopen = 0;
 553
 554         nskb = nfqnl_build_packet_message(net, queue, entry, &packet_id_ptr);
 555         if (nskb == NULL) {
 556                 err = -ENOMEM;
 557                 goto err_out;
 558         }
 559         spin_lock_bh(&queue->lock);
 560
 561         if (queue->queue_total >= queue->queue_maxlen) {
 562                 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
 563                         failopen = 1;
 564                         err = 0;
 565                 } else {
 566                         queue->queue_dropped++;
 567                         net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
 568                                              queue->queue_total);
 569                 }
 570                 goto err_out_free_nskb;
 571         }
 572         entry->id = ++queue->id_sequence;
 573         *packet_id_ptr = htonl(entry->id);
 574
 575         /* nfnetlink_unicast will either free the nskb or add it to a socket */
 576         err = nfnetlink_unicast(nskb, net, queue->peer_portid, MSG_DONTWAIT);
 577         if (err < 0) {
 578                 queue->queue_user_dropped++;
 579                 goto err_out_unlock;
 580         }
 581
 582         __enqueue_entry(queue, entry);
 583
 584         spin_unlock_bh(&queue->lock);
 585         return 0;
 586
 587 err_out_free_nskb:
 588         kfree_skb(nskb);
 589 err_out_unlock:
 590         spin_unlock_bh(&queue->lock);
 591         if (failopen)
 592                 nf_reinject(entry, NF_ACCEPT);
 593 err_out:
 594         return err;
 595 }
 596
 597 static struct nf_queue_entry *
 598 nf_queue_entry_dup(struct nf_queue_entry *e)
 599 {
 600         struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
 601         if (entry) {
 602                 if (nf_queue_entry_get_refs(entry))
 603                         return entry;
 604                 kfree(entry);
 605         }
 606         return NULL;
 607 }
 608
 609 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 610 /* When called from bridge netfilter, skb->data must point to MAC header
 611  * before calling skb_gso_segment(). Else, original MAC header is lost
 612  * and segmented skbs will be sent to wrong destination.
 613  */
 614 static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
 615 {
 616         if (skb->nf_bridge)
 617                 __skb_push(skb, skb->network_header - skb->mac_header);
 618 }
 619
 620 static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
 621 {
 622         if (skb->nf_bridge)
 623                 __skb_pull(skb, skb->network_header - skb->mac_header);
 624 }
 625 #else
 626 #define nf_bridge_adjust_skb_data(s) do {} while (0)
 627 #define nf_bridge_adjust_segmented_data(s) do {} while (0)
 628 #endif
 629
 630 static void free_entry(struct nf_queue_entry *entry)
 631 {
 632         nf_queue_entry_release_refs(entry);
 633         kfree(entry);
 634 }
 635
 636 static int
 637 __nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
 638                            struct sk_buff *skb, struct nf_queue_entry *entry)
 639 {
 640         int ret = -ENOMEM;
 641         struct nf_queue_entry *entry_seg;
 642
 643         nf_bridge_adjust_segmented_data(skb);
 644
 645         if (skb->next == NULL) { /* last packet, no need to copy entry */
 646                 struct sk_buff *gso_skb = entry->skb;
 647                 entry->skb = skb;
 648                 ret = __nfqnl_enqueue_packet(net, queue, entry);
 649                 if (ret)
 650                         entry->skb = gso_skb;
 651                 return ret;
 652         }
 653
 654         skb->next = NULL;
 655
 656         entry_seg = nf_queue_entry_dup(entry);
 657         if (entry_seg) {
 658                 entry_seg->skb = skb;
 659                 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
 660                 if (ret)
 661                         free_entry(entry_seg);
 662         }
 663         return ret;
 664 }
 665
 666 static int
 667 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 668 {
 669         unsigned int queued;
 670         struct nfqnl_instance *queue;
 671         struct sk_buff *skb, *segs;
 672         int err = -ENOBUFS;
 673         struct net *net = dev_net(entry->state.in ?
 674                                   entry->state.in : entry->state.out);
 675         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
 676
 677         /* rcu_read_lock()ed by nf_hook_slow() */
 678         queue = instance_lookup(q, queuenum);
 679         if (!queue)
 680                 return -ESRCH;
 681
 682         if (queue->copy_mode == NFQNL_COPY_NONE)
 683                 return -EINVAL;
 684
 685         skb = entry->skb;
 686
 687         switch (entry->state.pf) {
 688         case NFPROTO_IPV4:
 689                 skb->protocol = htons(ETH_P_IP);
 690                 break;
 691         case NFPROTO_IPV6:
 692                 skb->protocol = htons(ETH_P_IPV6);
 693                 break;
 694         }
 695
 696         if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(skb))
 697                 return __nfqnl_enqueue_packet(net, queue, entry);
 698
 699         nf_bridge_adjust_skb_data(skb);
 700         segs = skb_gso_segment(skb, 0);
 701         /* Does not use PTR_ERR to limit the number of error codes that can be
 702          * returned by nf_queue.  For instance, callers rely on -ECANCELED to
 703          * mean 'ignore this hook'.
 704          */
 705         if (IS_ERR_OR_NULL(segs))
 706                 goto out_err;
 707         queued = 0;
 708         err = 0;
 709         do {
 710                 struct sk_buff *nskb = segs->next;
 711                 if (err == 0)
 712                         err = __nfqnl_enqueue_packet_gso(net, queue,
 713                                                         segs, entry);
 714                 if (err == 0)
 715                         queued++;
 716                 else
 717                         kfree_skb(segs);
 718                 segs = nskb;
 719         } while (segs);
 720
 721         if (queued) {
 722                 if (err) /* some segments are already queued */
 723                         free_entry(entry);
 724                 kfree_skb(skb);
 725                 return 0;
 726         }
 727  out_err:
 728         nf_bridge_adjust_segmented_data(skb);
 729         return err;
 730 }
 731
 732 static int
 733 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e, int diff)
 734 {
 735         struct sk_buff *nskb;
 736
 737         if (diff < 0) {
 738                 if (pskb_trim(e->skb, data_len))
 739                         return -ENOMEM;
 740         } else if (diff > 0) {
 741                 if (data_len > 0xFFFF)
 742                         return -EINVAL;
 743                 if (diff > skb_tailroom(e->skb)) {
 744                         nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
 745                                                diff, GFP_ATOMIC);
 746                         if (!nskb) {
 747                                 printk(KERN_WARNING "nf_queue: OOM "
 748                                       "in mangle, dropping packet\n");
 749                                 return -ENOMEM;
 750                         }
 751                         kfree_skb(e->skb);
 752                         e->skb = nskb;
 753                 }
 754                 skb_put(e->skb, diff);
 755         }
 756         if (!skb_make_writable(e->skb, data_len))
 757                 return -ENOMEM;
 758         skb_copy_to_linear_data(e->skb, data, data_len);
 759         e->skb->ip_summed = CHECKSUM_NONE;
 760         return 0;
 761 }
 762
 763 static int
 764 nfqnl_set_mode(struct nfqnl_instance *queue,
 765                unsigned char mode, unsigned int range)
 766 {
 767         int status = 0;
 768
 769         spin_lock_bh(&queue->lock);
 770         switch (mode) {
 771         case NFQNL_COPY_NONE:
 772         case NFQNL_COPY_META:
 773                 queue->copy_mode = mode;
 774                 queue->copy_range = 0;
 775                 break;
 776
 777         case NFQNL_COPY_PACKET:
 778                 queue->copy_mode = mode;
 779                 if (range == 0 || range > NFQNL_MAX_COPY_RANGE)
 780                         queue->copy_range = NFQNL_MAX_COPY_RANGE;
 781                 else
 782                         queue->copy_range = range;
 783                 break;
 784
 785         default:
 786                 status = -EINVAL;
 787
 788         }
 789         spin_unlock_bh(&queue->lock);
 790
 791         return status;
 792 }
 793
 794 static int
 795 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
 796 {
 797         if (entry->state.in)
 798                 if (entry->state.in->ifindex == ifindex)
 799                         return 1;
 800         if (entry->state.out)
 801                 if (entry->state.out->ifindex == ifindex)
 802                         return 1;
 803 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 804         if (entry->skb->nf_bridge) {
 805                 int physinif, physoutif;
 806
 807                 physinif = nf_bridge_get_physinif(entry->skb);
 808                 physoutif = nf_bridge_get_physoutif(entry->skb);
 809
 810                 if (physinif == ifindex || physoutif == ifindex)
 811                         return 1;
 812         }
 813 #endif
 814         return 0;
 815 }
 816
 817 /* drop all packets with either indev or outdev == ifindex from all queue
 818  * instances */
 819 static void
 820 nfqnl_dev_drop(struct net *net, int ifindex)
 821 {
 822         int i;
 823         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
 824
 825         rcu_read_lock();
 826
 827         for (i = 0; i < INSTANCE_BUCKETS; i++) {
 828                 struct nfqnl_instance *inst;
 829                 struct hlist_head *head = &q->instance_table[i];
 830
 831                 hlist_for_each_entry_rcu(inst, head, hlist)
 832                         nfqnl_flush(inst, dev_cmp, ifindex);
 833         }
 834
 835         rcu_read_unlock();
 836 }
 837
 838 static int
 839 nfqnl_rcv_dev_event(struct notifier_block *this,
 840                     unsigned long event, void *ptr)
 841 {
 842         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
 843
 844         /* Drop any packets associated with the downed device */
 845         if (event == NETDEV_DOWN)
 846                 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
 847         return NOTIFY_DONE;
 848 }
 849
 850 static struct notifier_block nfqnl_dev_notifier = {
 851         .notifier_call  = nfqnl_rcv_dev_event,
 852 };
 853
 854 static int nf_hook_cmp(struct nf_queue_entry *entry, unsigned long ops_ptr)
 855 {
 856         return entry->elem == (struct nf_hook_ops *)ops_ptr;
 857 }
 858
 859 static void nfqnl_nf_hook_drop(struct net *net, struct nf_hook_ops *hook)
 860 {
 861         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
 862         int i;
 863
 864         rcu_read_lock();
 865         for (i = 0; i < INSTANCE_BUCKETS; i++) {
 866                 struct nfqnl_instance *inst;
 867                 struct hlist_head *head = &q->instance_table[i];
 868
 869                 hlist_for_each_entry_rcu(inst, head, hlist)
 870                         nfqnl_flush(inst, nf_hook_cmp, (unsigned long)hook);
 871         }
 872         rcu_read_unlock();
 873 }
 874
 875 static int
 876 nfqnl_rcv_nl_event(struct notifier_block *this,
 877                    unsigned long event, void *ptr)
 878 {
 879         struct netlink_notify *n = ptr;
 880         struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
 881
 882         if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
 883                 int i;
 884
 885                 /* destroy all instances for this portid */
 886                 spin_lock(&q->instances_lock);
 887                 for (i = 0; i < INSTANCE_BUCKETS; i++) {
 888                         struct hlist_node *t2;
 889                         struct nfqnl_instance *inst;
 890                         struct hlist_head *head = &q->instance_table[i];
 891
 892                         hlist_for_each_entry_safe(inst, t2, head, hlist) {
 893                                 if (n->portid == inst->peer_portid)
 894                                         __instance_destroy(inst);
 895                         }
 896                 }
 897                 spin_unlock(&q->instances_lock);
 898         }
 899         return NOTIFY_DONE;
 900 }
 901
 902 static struct notifier_block nfqnl_rtnl_notifier = {
 903         .notifier_call  = nfqnl_rcv_nl_event,
 904 };
 905
 906 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
 907         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 908         [NFQA_MARK]             = { .type = NLA_U32 },
 909         [NFQA_PAYLOAD]          = { .type = NLA_UNSPEC },
 910         [NFQA_CT]               = { .type = NLA_UNSPEC },
 911         [NFQA_EXP]              = { .type = NLA_UNSPEC },
 912 };
 913
 914 static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
 915         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 916         [NFQA_MARK]             = { .type = NLA_U32 },
 917 };
 918
 919 static struct nfqnl_instance *
 920 verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, u32 nlportid)
 921 {
 922         struct nfqnl_instance *queue;
 923
 924         queue = instance_lookup(q, queue_num);
 925         if (!queue)
 926                 return ERR_PTR(-ENODEV);
 927
 928         if (queue->peer_portid != nlportid)
 929                 return ERR_PTR(-EPERM);
 930
 931         return queue;
 932 }
 933
 934 static struct nfqnl_msg_verdict_hdr*
 935 verdicthdr_get(const struct nlattr * const nfqa[])
 936 {
 937         struct nfqnl_msg_verdict_hdr *vhdr;
 938         unsigned int verdict;
 939
 940         if (!nfqa[NFQA_VERDICT_HDR])
 941                 return NULL;
 942
 943         vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
 944         verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
 945         if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
 946                 return NULL;
 947         return vhdr;
 948 }
 949
 950 static int nfq_id_after(unsigned int id, unsigned int max)
 951 {
 952         return (int)(id - max) > 0;
 953 }
 954
 955 static int
 956 nfqnl_recv_verdict_batch(struct sock *ctnl, struct sk_buff *skb,
 957                    const struct nlmsghdr *nlh,
 958                    const struct nlattr * const nfqa[])
 959 {
 960         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
 961         struct nf_queue_entry *entry, *tmp;
 962         unsigned int verdict, maxid;
 963         struct nfqnl_msg_verdict_hdr *vhdr;
 964         struct nfqnl_instance *queue;
 965         LIST_HEAD(batch_list);
 966         u16 queue_num = ntohs(nfmsg->res_id);
 967
 968         struct net *net = sock_net(ctnl);
 969         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
 970
 971         queue = verdict_instance_lookup(q, queue_num,
 972                                         NETLINK_CB(skb).portid);
 973         if (IS_ERR(queue))
 974                 return PTR_ERR(queue);
 975
 976         vhdr = verdicthdr_get(nfqa);
 977         if (!vhdr)
 978                 return -EINVAL;
 979
 980         verdict = ntohl(vhdr->verdict);
 981         maxid = ntohl(vhdr->id);
 982
 983         spin_lock_bh(&queue->lock);
 984
 985         list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
 986                 if (nfq_id_after(entry->id, maxid))
 987                         break;
 988                 __dequeue_entry(queue, entry);
 989                 list_add_tail(&entry->list, &batch_list);
 990         }
 991
 992         spin_unlock_bh(&queue->lock);
 993
 994         if (list_empty(&batch_list))
 995                 return -ENOENT;
 996
 997         list_for_each_entry_safe(entry, tmp, &batch_list, list) {
 998                 if (nfqa[NFQA_MARK])
 999                         entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1000                 nf_reinject(entry, verdict);
1001         }
1002         return 0;
1003 }
1004
1005 static int
1006 nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
1007                    const struct nlmsghdr *nlh,
1008                    const struct nlattr * const nfqa[])
1009 {
1010         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1011         u_int16_t queue_num = ntohs(nfmsg->res_id);
1012
1013         struct nfqnl_msg_verdict_hdr *vhdr;
1014         struct nfqnl_instance *queue;
1015         unsigned int verdict;
1016         struct nf_queue_entry *entry;
1017         enum ip_conntrack_info uninitialized_var(ctinfo);
1018         struct nf_conn *ct = NULL;
1019
1020         struct net *net = sock_net(ctnl);
1021         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1022
1023         queue = instance_lookup(q, queue_num);
1024         if (!queue)
1025                 queue = verdict_instance_lookup(q, queue_num,
1026                                                 NETLINK_CB(skb).portid);
1027         if (IS_ERR(queue))
1028                 return PTR_ERR(queue);
1029
1030         vhdr = verdicthdr_get(nfqa);
1031         if (!vhdr)
1032                 return -EINVAL;
1033
1034         verdict = ntohl(vhdr->verdict);
1035
1036         entry = find_dequeue_entry(queue, ntohl(vhdr->id));
1037         if (entry == NULL)
1038                 return -ENOENT;
1039
1040         if (nfqa[NFQA_CT]) {
1041                 ct = nfqnl_ct_parse(entry->skb, nfqa[NFQA_CT], &ctinfo);
1042                 if (ct && nfqa[NFQA_EXP]) {
1043                         nfqnl_attach_expect(ct, nfqa[NFQA_EXP],
1044                                             NETLINK_CB(skb).portid,
1045                                             nlmsg_report(nlh));
1046                 }
1047         }
1048
1049         if (nfqa[NFQA_PAYLOAD]) {
1050                 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
1051                 int diff = payload_len - entry->skb->len;
1052
1053                 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
1054                                  payload_len, entry, diff) < 0)
1055                         verdict = NF_DROP;
1056
1057                 if (ct)
1058                         nfqnl_ct_seq_adjust(entry->skb, ct, ctinfo, diff);
1059         }
1060
1061         if (nfqa[NFQA_MARK])
1062                 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1063
1064         nf_reinject(entry, verdict);
1065         return 0;
1066 }
1067
1068 static int
1069 nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
1070                   const struct nlmsghdr *nlh,
1071                   const struct nlattr * const nfqa[])
1072 {
1073         return -ENOTSUPP;
1074 }
1075
1076 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1077         [NFQA_CFG_CMD]          = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1078         [NFQA_CFG_PARAMS]       = { .len = sizeof(struct nfqnl_msg_config_params) },
1079 };
1080
1081 static const struct nf_queue_handler nfqh = {
1082         .outfn          = &nfqnl_enqueue_packet,
1083         .nf_hook_drop   = &nfqnl_nf_hook_drop,
1084 };
1085
1086 static int
1087 nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
1088                   const struct nlmsghdr *nlh,
1089                   const struct nlattr * const nfqa[])
1090 {
1091         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1092         u_int16_t queue_num = ntohs(nfmsg->res_id);
1093         struct nfqnl_instance *queue;
1094         struct nfqnl_msg_config_cmd *cmd = NULL;
1095         struct net *net = sock_net(ctnl);
1096         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1097         int ret = 0;
1098
1099         if (nfqa[NFQA_CFG_CMD]) {
1100                 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1101
1102                 /* Obsolete commands without queue context */
1103                 switch (cmd->command) {
1104                 case NFQNL_CFG_CMD_PF_BIND: return 0;
1105                 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1106                 }
1107         }
1108
1109         rcu_read_lock();
1110         queue = instance_lookup(q, queue_num);
1111         if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1112                 ret = -EPERM;
1113                 goto err_out_unlock;
1114         }
1115
1116         if (cmd != NULL) {
1117                 switch (cmd->command) {
1118                 case NFQNL_CFG_CMD_BIND:
1119                         if (queue) {
1120                                 ret = -EBUSY;
1121                                 goto err_out_unlock;
1122                         }
1123                         queue = instance_create(q, queue_num,
1124                                                 NETLINK_CB(skb).portid);
1125                         if (IS_ERR(queue)) {
1126                                 ret = PTR_ERR(queue);
1127                                 goto err_out_unlock;
1128                         }
1129                         break;
1130                 case NFQNL_CFG_CMD_UNBIND:
1131                         if (!queue) {
1132                                 ret = -ENODEV;
1133                                 goto err_out_unlock;
1134                         }
1135                         instance_destroy(q, queue);
1136                         break;
1137                 case NFQNL_CFG_CMD_PF_BIND:
1138                 case NFQNL_CFG_CMD_PF_UNBIND:
1139                         break;
1140                 default:
1141                         ret = -ENOTSUPP;
1142                         break;
1143                 }
1144         }
1145
1146         if (nfqa[NFQA_CFG_PARAMS]) {
1147                 struct nfqnl_msg_config_params *params;
1148
1149                 if (!queue) {
1150                         ret = -ENODEV;
1151                         goto err_out_unlock;
1152                 }
1153                 params = nla_data(nfqa[NFQA_CFG_PARAMS]);
1154                 nfqnl_set_mode(queue, params->copy_mode,
1155                                 ntohl(params->copy_range));
1156         }
1157
1158         if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1159                 __be32 *queue_maxlen;
1160
1161                 if (!queue) {
1162                         ret = -ENODEV;
1163                         goto err_out_unlock;
1164                 }
1165                 queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1166                 spin_lock_bh(&queue->lock);
1167                 queue->queue_maxlen = ntohl(*queue_maxlen);
1168                 spin_unlock_bh(&queue->lock);
1169         }
1170
1171         if (nfqa[NFQA_CFG_FLAGS]) {
1172                 __u32 flags, mask;
1173
1174                 if (!queue) {
1175                         ret = -ENODEV;
1176                         goto err_out_unlock;
1177                 }
1178
1179                 if (!nfqa[NFQA_CFG_MASK]) {
1180                         /* A mask is needed to specify which flags are being
1181                          * changed.
1182                          */
1183                         ret = -EINVAL;
1184                         goto err_out_unlock;
1185                 }
1186
1187                 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1188                 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1189
1190                 if (flags >= NFQA_CFG_F_MAX) {
1191                         ret = -EOPNOTSUPP;
1192                         goto err_out_unlock;
1193                 }
1194 #if !IS_ENABLED(CONFIG_NETWORK_SECMARK)
1195                 if (flags & mask & NFQA_CFG_F_SECCTX) {
1196                         ret = -EOPNOTSUPP;
1197                         goto err_out_unlock;
1198                 }
1199 #endif
1200                 spin_lock_bh(&queue->lock);
1201                 queue->flags &= ~mask;
1202                 queue->flags |= flags & mask;
1203                 spin_unlock_bh(&queue->lock);
1204         }
1205
1206 err_out_unlock:
1207         rcu_read_unlock();
1208         return ret;
1209 }
1210
1211 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1212         [NFQNL_MSG_PACKET]      = { .call_rcu = nfqnl_recv_unsupp,
1213                                     .attr_count = NFQA_MAX, },
1214         [NFQNL_MSG_VERDICT]     = { .call_rcu = nfqnl_recv_verdict,
1215                                     .attr_count = NFQA_MAX,
1216                                     .policy = nfqa_verdict_policy },
1217         [NFQNL_MSG_CONFIG]      = { .call = nfqnl_recv_config,
1218                                     .attr_count = NFQA_CFG_MAX,
1219                                     .policy = nfqa_cfg_policy },
1220         [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
1221                                     .attr_count = NFQA_MAX,
1222                                     .policy = nfqa_verdict_batch_policy },
1223 };
1224
1225 static const struct nfnetlink_subsystem nfqnl_subsys = {
1226         .name           = "nf_queue",
1227         .subsys_id      = NFNL_SUBSYS_QUEUE,
1228         .cb_count       = NFQNL_MSG_MAX,
1229         .cb             = nfqnl_cb,
1230 };
1231
1232 #ifdef CONFIG_PROC_FS
1233 struct iter_state {
1234         struct seq_net_private p;
1235         unsigned int bucket;
1236 };
1237
1238 static struct hlist_node *get_first(struct seq_file *seq)
1239 {
1240         struct iter_state *st = seq->private;
1241         struct net *net;
1242         struct nfnl_queue_net *q;
1243
1244         if (!st)
1245                 return NULL;
1246
1247         net = seq_file_net(seq);
1248         q = nfnl_queue_pernet(net);
1249         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1250                 if (!hlist_empty(&q->instance_table[st->bucket]))
1251                         return q->instance_table[st->bucket].first;
1252         }
1253         return NULL;
1254 }
1255
1256 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1257 {
1258         struct iter_state *st = seq->private;
1259         struct net *net = seq_file_net(seq);
1260
1261         h = h->next;
1262         while (!h) {
1263                 struct nfnl_queue_net *q;
1264
1265                 if (++st->bucket >= INSTANCE_BUCKETS)
1266                         return NULL;
1267
1268                 q = nfnl_queue_pernet(net);
1269                 h = q->instance_table[st->bucket].first;
1270         }
1271         return h;
1272 }
1273
1274 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1275 {
1276         struct hlist_node *head;
1277         head = get_first(seq);
1278
1279         if (head)
1280                 while (pos && (head = get_next(seq, head)))
1281                         pos--;
1282         return pos ? NULL : head;
1283 }
1284
1285 static void *seq_start(struct seq_file *s, loff_t *pos)
1286         __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1287 {
1288         spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1289         return get_idx(s, *pos);
1290 }
1291
1292 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1293 {
1294         (*pos)++;
1295         return get_next(s, v);
1296 }
1297
1298 static void seq_stop(struct seq_file *s, void *v)
1299         __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1300 {
1301         spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1302 }
1303
1304 static int seq_show(struct seq_file *s, void *v)
1305 {
1306         const struct nfqnl_instance *inst = v;
1307
1308         seq_printf(s, "%5u %6u %5u %1u %5u %5u %5u %8u %2d\n",
1309                    inst->queue_num,
1310                    inst->peer_portid, inst->queue_total,
1311                    inst->copy_mode, inst->copy_range,
1312                    inst->queue_dropped, inst->queue_user_dropped,
1313                    inst->id_sequence, 1);
1314         return 0;
1315 }
1316
1317 static const struct seq_operations nfqnl_seq_ops = {
1318         .start  = seq_start,
1319         .next   = seq_next,
1320         .stop   = seq_stop,
1321         .show   = seq_show,
1322 };
1323
1324 static int nfqnl_open(struct inode *inode, struct file *file)
1325 {
1326         return seq_open_net(inode, file, &nfqnl_seq_ops,
1327                         sizeof(struct iter_state));
1328 }
1329
1330 static const struct file_operations nfqnl_file_ops = {
1331         .owner   = THIS_MODULE,
1332         .open    = nfqnl_open,
1333         .read    = seq_read,
1334         .llseek  = seq_lseek,
1335         .release = seq_release_net,
1336 };
1337
1338 #endif /* PROC_FS */
1339
1340 static int __net_init nfnl_queue_net_init(struct net *net)
1341 {
1342         unsigned int i;
1343         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1344
1345         for (i = 0; i < INSTANCE_BUCKETS; i++)
1346                 INIT_HLIST_HEAD(&q->instance_table[i]);
1347
1348         spin_lock_init(&q->instances_lock);
1349
1350 #ifdef CONFIG_PROC_FS
1351         if (!proc_create("nfnetlink_queue", 0440,
1352                          net->nf.proc_netfilter, &nfqnl_file_ops))
1353                 return -ENOMEM;
1354 #endif
1355         return 0;
1356 }
1357
1358 static void __net_exit nfnl_queue_net_exit(struct net *net)
1359 {
1360 #ifdef CONFIG_PROC_FS
1361         remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1362 #endif
1363 }
1364
1365 static struct pernet_operations nfnl_queue_net_ops = {
1366         .init   = nfnl_queue_net_init,
1367         .exit   = nfnl_queue_net_exit,
1368         .id     = &nfnl_queue_net_id,
1369         .size   = sizeof(struct nfnl_queue_net),
1370 };
1371
1372 static int __init nfnetlink_queue_init(void)
1373 {
1374         int status;
1375
1376         status = register_pernet_subsys(&nfnl_queue_net_ops);
1377         if (status < 0) {
1378                 pr_err("nf_queue: failed to register pernet ops\n");
1379                 goto out;
1380         }
1381
1382         netlink_register_notifier(&nfqnl_rtnl_notifier);
1383         status = nfnetlink_subsys_register(&nfqnl_subsys);
1384         if (status < 0) {
1385                 pr_err("nf_queue: failed to create netlink socket\n");
1386                 goto cleanup_netlink_notifier;
1387         }
1388
1389         register_netdevice_notifier(&nfqnl_dev_notifier);
1390         nf_register_queue_handler(&nfqh);
1391         return status;
1392
1393 cleanup_netlink_notifier:
1394         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1395 out:
1396         return status;
1397 }
1398
1399 static void __exit nfnetlink_queue_fini(void)
1400 {
1401         nf_unregister_queue_handler();
1402         unregister_netdevice_notifier(&nfqnl_dev_notifier);
1403         nfnetlink_subsys_unregister(&nfqnl_subsys);
1404         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1405         unregister_pernet_subsys(&nfnl_queue_net_ops);
1406
1407         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1408 }
1409
1410 MODULE_DESCRIPTION("netfilter packet queue handler");
1411 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1412 MODULE_LICENSE("GPL");
1413 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1414
1415 module_init(nfnetlink_queue_init);
1416 module_exit(nfnetlink_queue_fini);