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