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 nf_hook_cmp(struct nf_queue_entry *entry, unsigned long ops_ptr)
 854 {
 855         return entry->elem == (struct nf_hook_ops *)ops_ptr;
 856 }
 857
 858 static void nfqnl_nf_hook_drop(struct net *net, struct nf_hook_ops *hook)
 859 {
 860         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
 861         int i;
 862
 863         rcu_read_lock();
 864         for (i = 0; i < INSTANCE_BUCKETS; i++) {
 865                 struct nfqnl_instance *inst;
 866                 struct hlist_head *head = &q->instance_table[i];
 867
 868                 hlist_for_each_entry_rcu(inst, head, hlist)
 869                         nfqnl_flush(inst, nf_hook_cmp, (unsigned long)hook);
 870         }
 871         rcu_read_unlock();
 872 }
 873
 874 static int
 875 nfqnl_rcv_nl_event(struct notifier_block *this,
 876                    unsigned long event, void *ptr)
 877 {
 878         struct netlink_notify *n = ptr;
 879         struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
 880
 881         if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
 882                 int i;
 883
 884                 /* destroy all instances for this portid */
 885                 spin_lock(&q->instances_lock);
 886                 for (i = 0; i < INSTANCE_BUCKETS; i++) {
 887                         struct hlist_node *t2;
 888                         struct nfqnl_instance *inst;
 889                         struct hlist_head *head = &q->instance_table[i];
 890
 891                         hlist_for_each_entry_safe(inst, t2, head, hlist) {
 892                                 if (n->portid == inst->peer_portid)
 893                                         __instance_destroy(inst);
 894                         }
 895                 }
 896                 spin_unlock(&q->instances_lock);
 897         }
 898         return NOTIFY_DONE;
 899 }
 900
 901 static struct notifier_block nfqnl_rtnl_notifier = {
 902         .notifier_call  = nfqnl_rcv_nl_event,
 903 };
 904
 905 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
 906         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 907         [NFQA_MARK]             = { .type = NLA_U32 },
 908         [NFQA_PAYLOAD]          = { .type = NLA_UNSPEC },
 909         [NFQA_CT]               = { .type = NLA_UNSPEC },
 910         [NFQA_EXP]              = { .type = NLA_UNSPEC },
 911 };
 912
 913 static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
 914         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 915         [NFQA_MARK]             = { .type = NLA_U32 },
 916 };
 917
 918 static struct nfqnl_instance *
 919 verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, u32 nlportid)
 920 {
 921         struct nfqnl_instance *queue;
 922
 923         queue = instance_lookup(q, queue_num);
 924         if (!queue)
 925                 return ERR_PTR(-ENODEV);
 926
 927         if (queue->peer_portid != nlportid)
 928                 return ERR_PTR(-EPERM);
 929
 930         return queue;
 931 }
 932
 933 static struct nfqnl_msg_verdict_hdr*
 934 verdicthdr_get(const struct nlattr * const nfqa[])
 935 {
 936         struct nfqnl_msg_verdict_hdr *vhdr;
 937         unsigned int verdict;
 938
 939         if (!nfqa[NFQA_VERDICT_HDR])
 940                 return NULL;
 941
 942         vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
 943         verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
 944         if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
 945                 return NULL;
 946         return vhdr;
 947 }
 948
 949 static int nfq_id_after(unsigned int id, unsigned int max)
 950 {
 951         return (int)(id - max) > 0;
 952 }
 953
 954 static int
 955 nfqnl_recv_verdict_batch(struct sock *ctnl, struct sk_buff *skb,
 956                    const struct nlmsghdr *nlh,
 957                    const struct nlattr * const nfqa[])
 958 {
 959         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
 960         struct nf_queue_entry *entry, *tmp;
 961         unsigned int verdict, maxid;
 962         struct nfqnl_msg_verdict_hdr *vhdr;
 963         struct nfqnl_instance *queue;
 964         LIST_HEAD(batch_list);
 965         u16 queue_num = ntohs(nfmsg->res_id);
 966
 967         struct net *net = sock_net(ctnl);
 968         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
 969
 970         queue = verdict_instance_lookup(q, queue_num,
 971                                         NETLINK_CB(skb).portid);
 972         if (IS_ERR(queue))
 973                 return PTR_ERR(queue);
 974
 975         vhdr = verdicthdr_get(nfqa);
 976         if (!vhdr)
 977                 return -EINVAL;
 978
 979         verdict = ntohl(vhdr->verdict);
 980         maxid = ntohl(vhdr->id);
 981
 982         spin_lock_bh(&queue->lock);
 983
 984         list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
 985                 if (nfq_id_after(entry->id, maxid))
 986                         break;
 987                 __dequeue_entry(queue, entry);
 988                 list_add_tail(&entry->list, &batch_list);
 989         }
 990
 991         spin_unlock_bh(&queue->lock);
 992
 993         if (list_empty(&batch_list))
 994                 return -ENOENT;
 995
 996         list_for_each_entry_safe(entry, tmp, &batch_list, list) {
 997                 if (nfqa[NFQA_MARK])
 998                         entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
 999                 nf_reinject(entry, verdict);
1000         }
1001         return 0;
1002 }
1003
1004 static int
1005 nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
1006                    const struct nlmsghdr *nlh,
1007                    const struct nlattr * const nfqa[])
1008 {
1009         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1010         u_int16_t queue_num = ntohs(nfmsg->res_id);
1011
1012         struct nfqnl_msg_verdict_hdr *vhdr;
1013         struct nfqnl_instance *queue;
1014         unsigned int verdict;
1015         struct nf_queue_entry *entry;
1016         enum ip_conntrack_info uninitialized_var(ctinfo);
1017         struct nf_conn *ct = NULL;
1018
1019         struct net *net = sock_net(ctnl);
1020         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1021
1022         queue = instance_lookup(q, queue_num);
1023         if (!queue)
1024                 queue = verdict_instance_lookup(q, queue_num,
1025                                                 NETLINK_CB(skb).portid);
1026         if (IS_ERR(queue))
1027                 return PTR_ERR(queue);
1028
1029         vhdr = verdicthdr_get(nfqa);
1030         if (!vhdr)
1031                 return -EINVAL;
1032
1033         verdict = ntohl(vhdr->verdict);
1034
1035         entry = find_dequeue_entry(queue, ntohl(vhdr->id));
1036         if (entry == NULL)
1037                 return -ENOENT;
1038
1039         if (nfqa[NFQA_CT]) {
1040                 ct = nfqnl_ct_parse(entry->skb, nfqa[NFQA_CT], &ctinfo);
1041                 if (ct && nfqa[NFQA_EXP]) {
1042                         nfqnl_attach_expect(ct, nfqa[NFQA_EXP],
1043                                             NETLINK_CB(skb).portid,
1044                                             nlmsg_report(nlh));
1045                 }
1046         }
1047
1048         if (nfqa[NFQA_PAYLOAD]) {
1049                 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
1050                 int diff = payload_len - entry->skb->len;
1051
1052                 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
1053                                  payload_len, entry, diff) < 0)
1054                         verdict = NF_DROP;
1055
1056                 if (ct)
1057                         nfqnl_ct_seq_adjust(entry->skb, ct, ctinfo, diff);
1058         }
1059
1060         if (nfqa[NFQA_MARK])
1061                 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1062
1063         nf_reinject(entry, verdict);
1064         return 0;
1065 }
1066
1067 static int
1068 nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
1069                   const struct nlmsghdr *nlh,
1070                   const struct nlattr * const nfqa[])
1071 {
1072         return -ENOTSUPP;
1073 }
1074
1075 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1076         [NFQA_CFG_CMD]          = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1077         [NFQA_CFG_PARAMS]       = { .len = sizeof(struct nfqnl_msg_config_params) },
1078 };
1079
1080 static const struct nf_queue_handler nfqh = {
1081         .outfn          = &nfqnl_enqueue_packet,
1082         .nf_hook_drop   = &nfqnl_nf_hook_drop,
1083 };
1084
1085 static int
1086 nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
1087                   const struct nlmsghdr *nlh,
1088                   const struct nlattr * const nfqa[])
1089 {
1090         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1091         u_int16_t queue_num = ntohs(nfmsg->res_id);
1092         struct nfqnl_instance *queue;
1093         struct nfqnl_msg_config_cmd *cmd = NULL;
1094         struct net *net = sock_net(ctnl);
1095         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1096         int ret = 0;
1097
1098         if (nfqa[NFQA_CFG_CMD]) {
1099                 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1100
1101                 /* Obsolete commands without queue context */
1102                 switch (cmd->command) {
1103                 case NFQNL_CFG_CMD_PF_BIND: return 0;
1104                 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1105                 }
1106         }
1107
1108         rcu_read_lock();
1109         queue = instance_lookup(q, queue_num);
1110         if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1111                 ret = -EPERM;
1112                 goto err_out_unlock;
1113         }
1114
1115         if (cmd != NULL) {
1116                 switch (cmd->command) {
1117                 case NFQNL_CFG_CMD_BIND:
1118                         if (queue) {
1119                                 ret = -EBUSY;
1120                                 goto err_out_unlock;
1121                         }
1122                         queue = instance_create(q, queue_num,
1123                                                 NETLINK_CB(skb).portid);
1124                         if (IS_ERR(queue)) {
1125                                 ret = PTR_ERR(queue);
1126                                 goto err_out_unlock;
1127                         }
1128                         break;
1129                 case NFQNL_CFG_CMD_UNBIND:
1130                         if (!queue) {
1131                                 ret = -ENODEV;
1132                                 goto err_out_unlock;
1133                         }
1134                         instance_destroy(q, queue);
1135                         break;
1136                 case NFQNL_CFG_CMD_PF_BIND:
1137                 case NFQNL_CFG_CMD_PF_UNBIND:
1138                         break;
1139                 default:
1140                         ret = -ENOTSUPP;
1141                         break;
1142                 }
1143         }
1144
1145         if (nfqa[NFQA_CFG_PARAMS]) {
1146                 struct nfqnl_msg_config_params *params;
1147
1148                 if (!queue) {
1149                         ret = -ENODEV;
1150                         goto err_out_unlock;
1151                 }
1152                 params = nla_data(nfqa[NFQA_CFG_PARAMS]);
1153                 nfqnl_set_mode(queue, params->copy_mode,
1154                                 ntohl(params->copy_range));
1155         }
1156
1157         if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1158                 __be32 *queue_maxlen;
1159
1160                 if (!queue) {
1161                         ret = -ENODEV;
1162                         goto err_out_unlock;
1163                 }
1164                 queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1165                 spin_lock_bh(&queue->lock);
1166                 queue->queue_maxlen = ntohl(*queue_maxlen);
1167                 spin_unlock_bh(&queue->lock);
1168         }
1169
1170         if (nfqa[NFQA_CFG_FLAGS]) {
1171                 __u32 flags, mask;
1172
1173                 if (!queue) {
1174                         ret = -ENODEV;
1175                         goto err_out_unlock;
1176                 }
1177
1178                 if (!nfqa[NFQA_CFG_MASK]) {
1179                         /* A mask is needed to specify which flags are being
1180                          * changed.
1181                          */
1182                         ret = -EINVAL;
1183                         goto err_out_unlock;
1184                 }
1185
1186                 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1187                 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1188
1189                 if (flags >= NFQA_CFG_F_MAX) {
1190                         ret = -EOPNOTSUPP;
1191                         goto err_out_unlock;
1192                 }
1193 #if !IS_ENABLED(CONFIG_NETWORK_SECMARK)
1194                 if (flags & mask & NFQA_CFG_F_SECCTX) {
1195                         ret = -EOPNOTSUPP;
1196                         goto err_out_unlock;
1197                 }
1198 #endif
1199                 spin_lock_bh(&queue->lock);
1200                 queue->flags &= ~mask;
1201                 queue->flags |= flags & mask;
1202                 spin_unlock_bh(&queue->lock);
1203         }
1204
1205 err_out_unlock:
1206         rcu_read_unlock();
1207         return ret;
1208 }
1209
1210 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1211         [NFQNL_MSG_PACKET]      = { .call_rcu = nfqnl_recv_unsupp,
1212                                     .attr_count = NFQA_MAX, },
1213         [NFQNL_MSG_VERDICT]     = { .call_rcu = nfqnl_recv_verdict,
1214                                     .attr_count = NFQA_MAX,
1215                                     .policy = nfqa_verdict_policy },
1216         [NFQNL_MSG_CONFIG]      = { .call = nfqnl_recv_config,
1217                                     .attr_count = NFQA_CFG_MAX,
1218                                     .policy = nfqa_cfg_policy },
1219         [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
1220                                     .attr_count = NFQA_MAX,
1221                                     .policy = nfqa_verdict_batch_policy },
1222 };
1223
1224 static const struct nfnetlink_subsystem nfqnl_subsys = {
1225         .name           = "nf_queue",
1226         .subsys_id      = NFNL_SUBSYS_QUEUE,
1227         .cb_count       = NFQNL_MSG_MAX,
1228         .cb             = nfqnl_cb,
1229 };
1230
1231 #ifdef CONFIG_PROC_FS
1232 struct iter_state {
1233         struct seq_net_private p;
1234         unsigned int bucket;
1235 };
1236
1237 static struct hlist_node *get_first(struct seq_file *seq)
1238 {
1239         struct iter_state *st = seq->private;
1240         struct net *net;
1241         struct nfnl_queue_net *q;
1242
1243         if (!st)
1244                 return NULL;
1245
1246         net = seq_file_net(seq);
1247         q = nfnl_queue_pernet(net);
1248         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1249                 if (!hlist_empty(&q->instance_table[st->bucket]))
1250                         return q->instance_table[st->bucket].first;
1251         }
1252         return NULL;
1253 }
1254
1255 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1256 {
1257         struct iter_state *st = seq->private;
1258         struct net *net = seq_file_net(seq);
1259
1260         h = h->next;
1261         while (!h) {
1262                 struct nfnl_queue_net *q;
1263
1264                 if (++st->bucket >= INSTANCE_BUCKETS)
1265                         return NULL;
1266
1267                 q = nfnl_queue_pernet(net);
1268                 h = q->instance_table[st->bucket].first;
1269         }
1270         return h;
1271 }
1272
1273 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1274 {
1275         struct hlist_node *head;
1276         head = get_first(seq);
1277
1278         if (head)
1279                 while (pos && (head = get_next(seq, head)))
1280                         pos--;
1281         return pos ? NULL : head;
1282 }
1283
1284 static void *seq_start(struct seq_file *s, loff_t *pos)
1285         __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1286 {
1287         spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1288         return get_idx(s, *pos);
1289 }
1290
1291 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1292 {
1293         (*pos)++;
1294         return get_next(s, v);
1295 }
1296
1297 static void seq_stop(struct seq_file *s, void *v)
1298         __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1299 {
1300         spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1301 }
1302
1303 static int seq_show(struct seq_file *s, void *v)
1304 {
1305         const struct nfqnl_instance *inst = v;
1306
1307         seq_printf(s, "%5u %6u %5u %1u %5u %5u %5u %8u %2d\n",
1308                    inst->queue_num,
1309                    inst->peer_portid, inst->queue_total,
1310                    inst->copy_mode, inst->copy_range,
1311                    inst->queue_dropped, inst->queue_user_dropped,
1312                    inst->id_sequence, 1);
1313         return 0;
1314 }
1315
1316 static const struct seq_operations nfqnl_seq_ops = {
1317         .start  = seq_start,
1318         .next   = seq_next,
1319         .stop   = seq_stop,
1320         .show   = seq_show,
1321 };
1322
1323 static int nfqnl_open(struct inode *inode, struct file *file)
1324 {
1325         return seq_open_net(inode, file, &nfqnl_seq_ops,
1326                         sizeof(struct iter_state));
1327 }
1328
1329 static const struct file_operations nfqnl_file_ops = {
1330         .owner   = THIS_MODULE,
1331         .open    = nfqnl_open,
1332         .read    = seq_read,
1333         .llseek  = seq_lseek,
1334         .release = seq_release_net,
1335 };
1336
1337 #endif /* PROC_FS */
1338
1339 static int __net_init nfnl_queue_net_init(struct net *net)
1340 {
1341         unsigned int i;
1342         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1343
1344         for (i = 0; i < INSTANCE_BUCKETS; i++)
1345                 INIT_HLIST_HEAD(&q->instance_table[i]);
1346
1347         spin_lock_init(&q->instances_lock);
1348
1349 #ifdef CONFIG_PROC_FS
1350         if (!proc_create("nfnetlink_queue", 0440,
1351                          net->nf.proc_netfilter, &nfqnl_file_ops))
1352                 return -ENOMEM;
1353 #endif
1354         return 0;
1355 }
1356
1357 static void __net_exit nfnl_queue_net_exit(struct net *net)
1358 {
1359 #ifdef CONFIG_PROC_FS
1360         remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1361 #endif
1362 }
1363
1364 static struct pernet_operations nfnl_queue_net_ops = {
1365         .init   = nfnl_queue_net_init,
1366         .exit   = nfnl_queue_net_exit,
1367         .id     = &nfnl_queue_net_id,
1368         .size   = sizeof(struct nfnl_queue_net),
1369 };
1370
1371 static int __init nfnetlink_queue_init(void)
1372 {
1373         int status;
1374
1375         status = register_pernet_subsys(&nfnl_queue_net_ops);
1376         if (status < 0) {
1377                 pr_err("nf_queue: failed to register pernet ops\n");
1378                 goto out;
1379         }
1380
1381         netlink_register_notifier(&nfqnl_rtnl_notifier);
1382         status = nfnetlink_subsys_register(&nfqnl_subsys);
1383         if (status < 0) {
1384                 pr_err("nf_queue: failed to create netlink socket\n");
1385                 goto cleanup_netlink_notifier;
1386         }
1387
1388         register_netdevice_notifier(&nfqnl_dev_notifier);
1389         nf_register_queue_handler(&nfqh);
1390         return status;
1391
1392 cleanup_netlink_notifier:
1393         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1394 out:
1395         return status;
1396 }
1397
1398 static void __exit nfnetlink_queue_fini(void)
1399 {
1400         nf_unregister_queue_handler();
1401         unregister_netdevice_notifier(&nfqnl_dev_notifier);
1402         nfnetlink_subsys_unregister(&nfqnl_subsys);
1403         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1404         unregister_pernet_subsys(&nfnl_queue_net_ops);
1405
1406         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1407 }
1408
1409 MODULE_DESCRIPTION("netfilter packet queue handler");
1410 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1411 MODULE_LICENSE("GPL");
1412 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1413
1414 module_init(nfnetlink_queue_init);
1415 module_exit(nfnetlink_queue_fini);