net/netfilter/nfnetlink_queue.c

   1 /*
   2  * This is a module which is used for queueing packets and communicating with
   3  * userspace via nfnetlink.
   4  *
   5  * (C) 2005 by Harald Welte <[email protected]>
   6  * (C) 2007 by Patrick McHardy <[email protected]>
   7  *
   8  * Based on the old ipv4-only ip_queue.c:
   9  * (C) 2000-2002 James Morris <[email protected]>
  10  * (C) 2003-2005 Netfilter Core Team <[email protected]>
  11  *
  12  * This program is free software; you can redistribute it and/or modify
  13  * it under the terms of the GNU General Public License version 2 as
  14  * published by the Free Software Foundation.
  15  *
  16  */
  17 #include <linux/module.h>
  18 #include <linux/skbuff.h>
  19 #include <linux/init.h>
  20 #include <linux/spinlock.h>
  21 #include <linux/slab.h>
  22 #include <linux/notifier.h>
  23 #include <linux/netdevice.h>
  24 #include <linux/netfilter.h>
  25 #include <linux/proc_fs.h>
  26 #include <linux/netfilter_ipv4.h>
  27 #include <linux/netfilter_ipv6.h>
  28 #include <linux/netfilter_bridge.h>
  29 #include <linux/netfilter/nfnetlink.h>
  30 #include <linux/netfilter/nfnetlink_queue.h>
  31 #include <linux/netfilter/nf_conntrack_common.h>
  32 #include <linux/list.h>
  33 #include <net/sock.h>
  34 #include <net/tcp_states.h>
  35 #include <net/netfilter/nf_queue.h>
  36 #include <net/netns/generic.h>
  37
  38 #include <linux/atomic.h>
  39
  40 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  41 #include "../bridge/br_private.h"
  42 #endif
  43
  44 #define NFQNL_QMAX_DEFAULT 1024
  45
  46 /* We're using struct nlattr which has 16bit nla_len. Note that nla_len
  47  * includes the header length. Thus, the maximum packet length that we
  48  * support is 65531 bytes. We send truncated packets if the specified length
  49  * is larger than that.  Userspace can check for presence of NFQA_CAP_LEN
  50  * attribute to detect truncation.
  51  */
  52 #define NFQNL_MAX_COPY_RANGE (0xffff - NLA_HDRLEN)
  53
  54 struct nfqnl_instance {
  55         struct hlist_node hlist;                /* global list of queues */
  56         struct rcu_head rcu;
  57
  58         u32 peer_portid;
  59         unsigned int queue_maxlen;
  60         unsigned int copy_range;
  61         unsigned int queue_dropped;
  62         unsigned int queue_user_dropped;
  63
  64
  65         u_int16_t queue_num;                    /* number of this queue */
  66         u_int8_t copy_mode;
  67         u_int32_t flags;                        /* Set using NFQA_CFG_FLAGS */
  68 /*
  69  * Following fields are dirtied for each queued packet,
  70  * keep them in same cache line if possible.
  71  */
  72         spinlock_t      lock;
  73         unsigned int    queue_total;
  74         unsigned int    id_sequence;            /* 'sequence' of pkt ids */
  75         struct list_head queue_list;            /* packets in queue */
  76 };
  77
  78 typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
  79
  80 static int nfnl_queue_net_id __read_mostly;
  81
  82 #define INSTANCE_BUCKETS        16
  83 struct nfnl_queue_net {
  84         spinlock_t instances_lock;
  85         struct hlist_head instance_table[INSTANCE_BUCKETS];
  86 };
  87
  88 static struct nfnl_queue_net *nfnl_queue_pernet(struct net *net)
  89 {
  90         return net_generic(net, nfnl_queue_net_id);
  91 }
  92
  93 static inline u_int8_t instance_hashfn(u_int16_t queue_num)
  94 {
  95         return ((queue_num >> 8) ^ queue_num) % INSTANCE_BUCKETS;
  96 }
  97
  98 static struct nfqnl_instance *
  99 instance_lookup(struct nfnl_queue_net *q, u_int16_t queue_num)
 100 {
 101         struct hlist_head *head;
 102         struct nfqnl_instance *inst;
 103
 104         head = &q->instance_table[instance_hashfn(queue_num)];
 105         hlist_for_each_entry_rcu(inst, head, hlist) {
 106                 if (inst->queue_num == queue_num)
 107                         return inst;
 108         }
 109         return NULL;
 110 }
 111
 112 static struct nfqnl_instance *
 113 instance_create(struct nfnl_queue_net *q, u_int16_t queue_num, u32 portid)
 114 {
 115         struct nfqnl_instance *inst;
 116         unsigned int h;
 117         int err;
 118
 119         spin_lock(&q->instances_lock);
 120         if (instance_lookup(q, queue_num)) {
 121                 err = -EEXIST;
 122                 goto out_unlock;
 123         }
 124
 125         inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
 126         if (!inst) {
 127                 err = -ENOMEM;
 128                 goto out_unlock;
 129         }
 130
 131         inst->queue_num = queue_num;
 132         inst->peer_portid = portid;
 133         inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
 134         inst->copy_range = NFQNL_MAX_COPY_RANGE;
 135         inst->copy_mode = NFQNL_COPY_NONE;
 136         spin_lock_init(&inst->lock);
 137         INIT_LIST_HEAD(&inst->queue_list);
 138
 139         if (!try_module_get(THIS_MODULE)) {
 140                 err = -EAGAIN;
 141                 goto out_free;
 142         }
 143
 144         h = instance_hashfn(queue_num);
 145         hlist_add_head_rcu(&inst->hlist, &q->instance_table[h]);
 146
 147         spin_unlock(&q->instances_lock);
 148
 149         return inst;
 150
 151 out_free:
 152         kfree(inst);
 153 out_unlock:
 154         spin_unlock(&q->instances_lock);
 155         return ERR_PTR(err);
 156 }
 157
 158 static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
 159                         unsigned long data);
 160
 161 static void
 162 instance_destroy_rcu(struct rcu_head *head)
 163 {
 164         struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
 165                                                    rcu);
 166
 167         nfqnl_flush(inst, NULL, 0);
 168         kfree(inst);
 169         module_put(THIS_MODULE);
 170 }
 171
 172 static void
 173 __instance_destroy(struct nfqnl_instance *inst)
 174 {
 175         hlist_del_rcu(&inst->hlist);
 176         call_rcu(&inst->rcu, instance_destroy_rcu);
 177 }
 178
 179 static void
 180 instance_destroy(struct nfnl_queue_net *q, struct nfqnl_instance *inst)
 181 {
 182         spin_lock(&q->instances_lock);
 183         __instance_destroy(inst);
 184         spin_unlock(&q->instances_lock);
 185 }
 186
 187 static inline void
 188 __enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
 189 {
 190        list_add_tail(&entry->list, &queue->queue_list);
 191        queue->queue_total++;
 192 }
 193
 194 static void
 195 __dequeue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
 196 {
 197         list_del(&entry->list);
 198         queue->queue_total--;
 199 }
 200
 201 static struct nf_queue_entry *
 202 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
 203 {
 204         struct nf_queue_entry *entry = NULL, *i;
 205
 206         spin_lock_bh(&queue->lock);
 207
 208         list_for_each_entry(i, &queue->queue_list, list) {
 209                 if (i->id == id) {
 210                         entry = i;
 211                         break;
 212                 }
 213         }
 214
 215         if (entry)
 216                 __dequeue_entry(queue, entry);
 217
 218         spin_unlock_bh(&queue->lock);
 219
 220         return entry;
 221 }
 222
 223 static void
 224 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
 225 {
 226         struct nf_queue_entry *entry, *next;
 227
 228         spin_lock_bh(&queue->lock);
 229         list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
 230                 if (!cmpfn || cmpfn(entry, data)) {
 231                         list_del(&entry->list);
 232                         queue->queue_total--;
 233                         nf_reinject(entry, NF_DROP);
 234                 }
 235         }
 236         spin_unlock_bh(&queue->lock);
 237 }
 238
 239 static int
 240 nfqnl_put_packet_info(struct sk_buff *nlskb, struct sk_buff *packet,
 241                       bool csum_verify)
 242 {
 243         __u32 flags = 0;
 244
 245         if (packet->ip_summed == CHECKSUM_PARTIAL)
 246                 flags = NFQA_SKB_CSUMNOTREADY;
 247         else if (csum_verify)
 248                 flags = NFQA_SKB_CSUM_NOTVERIFIED;
 249
 250         if (skb_is_gso(packet))
 251                 flags |= NFQA_SKB_GSO;
 252
 253         return flags ? nla_put_be32(nlskb, NFQA_SKB_INFO, htonl(flags)) : 0;
 254 }
 255
 256 static int nfqnl_put_sk_uidgid(struct sk_buff *skb, struct sock *sk)
 257 {
 258         const struct cred *cred;
 259
 260         if (!sk_fullsock(sk))
 261                 return 0;
 262
 263         read_lock_bh(&sk->sk_callback_lock);
 264         if (sk->sk_socket && sk->sk_socket->file) {
 265                 cred = sk->sk_socket->file->f_cred;
 266                 if (nla_put_be32(skb, NFQA_UID,
 267                     htonl(from_kuid_munged(&init_user_ns, cred->fsuid))))
 268                         goto nla_put_failure;
 269                 if (nla_put_be32(skb, NFQA_GID,
 270                     htonl(from_kgid_munged(&init_user_ns, cred->fsgid))))
 271                         goto nla_put_failure;
 272         }
 273         read_unlock_bh(&sk->sk_callback_lock);
 274         return 0;
 275
 276 nla_put_failure:
 277         read_unlock_bh(&sk->sk_callback_lock);
 278         return -1;
 279 }
 280
 281 static u32 nfqnl_get_sk_secctx(struct sk_buff *skb, char **secdata)
 282 {
 283         u32 seclen = 0;
 284 #if IS_ENABLED(CONFIG_NETWORK_SECMARK)
 285         if (!skb || !sk_fullsock(skb->sk))
 286                 return 0;
 287
 288         read_lock_bh(&skb->sk->sk_callback_lock);
 289
 290         if (skb->secmark)
 291                 security_secid_to_secctx(skb->secmark, secdata, &seclen);
 292
 293         read_unlock_bh(&skb->sk->sk_callback_lock);
 294 #endif
 295         return seclen;
 296 }
 297
 298 static struct sk_buff *
 299 nfqnl_build_packet_message(struct net *net, struct nfqnl_instance *queue,
 300                            struct nf_queue_entry *entry,
 301                            __be32 **packet_id_ptr)
 302 {
 303         size_t size;
 304         size_t data_len = 0, cap_len = 0, rem_len = 0;
 305         unsigned int hlen = 0;
 306         struct sk_buff *skb;
 307         struct nlattr *nla;
 308         struct nfqnl_msg_packet_hdr *pmsg;
 309         struct nlmsghdr *nlh;
 310         struct nfgenmsg *nfmsg;
 311         struct sk_buff *entskb = entry->skb;
 312         struct net_device *indev;
 313         struct net_device *outdev;
 314         struct nf_conn *ct = NULL;
 315         enum ip_conntrack_info uninitialized_var(ctinfo);
 316         struct nfnl_ct_hook *nfnl_ct;
 317         bool csum_verify;
 318         char *secdata = NULL;
 319         u32 seclen = 0;
 320
 321         size =    nlmsg_total_size(sizeof(struct nfgenmsg))
 322                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
 323                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 324                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 325 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 326                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 327                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 328 #endif
 329                 + nla_total_size(sizeof(u_int32_t))     /* mark */
 330                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
 331                 + nla_total_size(sizeof(u_int32_t))     /* skbinfo */
 332                 + nla_total_size(sizeof(u_int32_t));    /* cap_len */
 333
 334         if (entskb->tstamp.tv64)
 335                 size += nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
 336
 337         if (entry->state.hook <= NF_INET_FORWARD ||
 338            (entry->state.hook == NF_INET_POST_ROUTING && entskb->sk == NULL))
 339                 csum_verify = !skb_csum_unnecessary(entskb);
 340         else
 341                 csum_verify = false;
 342
 343         outdev = entry->state.out;
 344
 345         switch ((enum nfqnl_config_mode)ACCESS_ONCE(queue->copy_mode)) {
 346         case NFQNL_COPY_META:
 347         case NFQNL_COPY_NONE:
 348                 break;
 349
 350         case NFQNL_COPY_PACKET:
 351                 if (!(queue->flags & NFQA_CFG_F_GSO) &&
 352                     entskb->ip_summed == CHECKSUM_PARTIAL &&
 353                     skb_checksum_help(entskb))
 354                         return NULL;
 355
 356                 data_len = ACCESS_ONCE(queue->copy_range);
 357                 if (data_len > entskb->len)
 358                         data_len = entskb->len;
 359
 360                 hlen = skb_zerocopy_headlen(entskb);
 361                 hlen = min_t(unsigned int, hlen, data_len);
 362                 size += sizeof(struct nlattr) + hlen;
 363                 cap_len = entskb->len;
 364                 rem_len = data_len - hlen;
 365                 break;
 366         }
 367
 368         if (queue->flags & NFQA_CFG_F_CONNTRACK) {
 369                 nfnl_ct = rcu_dereference(nfnl_ct_hook);
 370                 if (nfnl_ct != NULL) {
 371                         ct = nfnl_ct->get_ct(entskb, &ctinfo);
 372                         if (ct != NULL)
 373                                 size += nfnl_ct->build_size(ct);
 374                 }
 375         }
 376
 377         if (queue->flags & NFQA_CFG_F_UID_GID) {
 378                 size +=  (nla_total_size(sizeof(u_int32_t))     /* uid */
 379                         + nla_total_size(sizeof(u_int32_t)));   /* gid */
 380         }
 381
 382         if ((queue->flags & NFQA_CFG_F_SECCTX) && entskb->sk) {
 383                 seclen = nfqnl_get_sk_secctx(entskb, &secdata);
 384                 if (seclen)
 385                         size += nla_total_size(seclen);
 386         }
 387
 388         skb = __netlink_alloc_skb(net->nfnl, size, rem_len, queue->peer_portid,
 389                                   GFP_ATOMIC);
 390         if (!skb) {
 391                 skb_tx_error(entskb);
 392                 return NULL;
 393         }
 394
 395         nlh = nlmsg_put(skb, 0, 0,
 396                         NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
 397                         sizeof(struct nfgenmsg), 0);
 398         if (!nlh) {
 399                 skb_tx_error(entskb);
 400                 kfree_skb(skb);
 401                 return NULL;
 402         }
 403         nfmsg = nlmsg_data(nlh);
 404         nfmsg->nfgen_family = entry->state.pf;
 405         nfmsg->version = NFNETLINK_V0;
 406         nfmsg->res_id = htons(queue->queue_num);
 407
 408         nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
 409         pmsg = nla_data(nla);
 410         pmsg->hw_protocol       = entskb->protocol;
 411         pmsg->hook              = entry->state.hook;
 412         *packet_id_ptr          = &pmsg->packet_id;
 413
 414         indev = entry->state.in;
 415         if (indev) {
 416 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 417                 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
 418                         goto nla_put_failure;
 419 #else
 420                 if (entry->state.pf == PF_BRIDGE) {
 421                         /* Case 1: indev is physical input device, we need to
 422                          * look for bridge group (when called from
 423                          * netfilter_bridge) */
 424                         if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
 425                                          htonl(indev->ifindex)) ||
 426                         /* this is the bridge group "brX" */
 427                         /* rcu_read_lock()ed by __nf_queue */
 428                             nla_put_be32(skb, NFQA_IFINDEX_INDEV,
 429                                          htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
 430                                 goto nla_put_failure;
 431                 } else {
 432                         int physinif;
 433
 434                         /* Case 2: indev is bridge group, we need to look for
 435                          * physical device (when called from ipv4) */
 436                         if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
 437                                          htonl(indev->ifindex)))
 438                                 goto nla_put_failure;
 439
 440                         physinif = nf_bridge_get_physinif(entskb);
 441                         if (physinif &&
 442                             nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
 443                                          htonl(physinif)))
 444                                 goto nla_put_failure;
 445                 }
 446 #endif
 447         }
 448
 449         if (outdev) {
 450 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 451                 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
 452                         goto nla_put_failure;
 453 #else
 454                 if (entry->state.pf == PF_BRIDGE) {
 455                         /* Case 1: outdev is physical output device, we need to
 456                          * look for bridge group (when called from
 457                          * netfilter_bridge) */
 458                         if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 459                                          htonl(outdev->ifindex)) ||
 460                         /* this is the bridge group "brX" */
 461                         /* rcu_read_lock()ed by __nf_queue */
 462                             nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
 463                                          htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
 464                                 goto nla_put_failure;
 465                 } else {
 466                         int physoutif;
 467
 468                         /* Case 2: outdev is bridge group, we need to look for
 469                          * physical output device (when called from ipv4) */
 470                         if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
 471                                          htonl(outdev->ifindex)))
 472                                 goto nla_put_failure;
 473
 474                         physoutif = nf_bridge_get_physoutif(entskb);
 475                         if (physoutif &&
 476                             nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 477                                          htonl(physoutif)))
 478                                 goto nla_put_failure;
 479                 }
 480 #endif
 481         }
 482
 483         if (entskb->mark &&
 484             nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
 485                 goto nla_put_failure;
 486
 487         if (indev && entskb->dev &&
 488             entskb->mac_header != entskb->network_header) {
 489                 struct nfqnl_msg_packet_hw phw;
 490                 int len;
 491
 492                 memset(&phw, 0, sizeof(phw));
 493                 len = dev_parse_header(entskb, phw.hw_addr);
 494                 if (len) {
 495                         phw.hw_addrlen = htons(len);
 496                         if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
 497                                 goto nla_put_failure;
 498                 }
 499         }
 500
 501         if (entskb->tstamp.tv64) {
 502                 struct nfqnl_msg_packet_timestamp ts;
 503                 struct timespec64 kts = ktime_to_timespec64(skb->tstamp);
 504
 505                 ts.sec = cpu_to_be64(kts.tv_sec);
 506                 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
 507
 508                 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
 509                         goto nla_put_failure;
 510         }
 511
 512         if ((queue->flags & NFQA_CFG_F_UID_GID) && entskb->sk &&
 513             nfqnl_put_sk_uidgid(skb, entskb->sk) < 0)
 514                 goto nla_put_failure;
 515
 516         if (seclen && nla_put(skb, NFQA_SECCTX, seclen, secdata))
 517                 goto nla_put_failure;
 518
 519         if (ct && nfnl_ct->build(skb, ct, ctinfo, NFQA_CT, NFQA_CT_INFO) < 0)
 520                 goto nla_put_failure;
 521
 522         if (cap_len > data_len &&
 523             nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
 524                 goto nla_put_failure;
 525
 526         if (nfqnl_put_packet_info(skb, entskb, csum_verify))
 527                 goto nla_put_failure;
 528
 529         if (data_len) {
 530                 struct nlattr *nla;
 531
 532                 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
 533                         goto nla_put_failure;
 534
 535                 nla = (struct nlattr *)skb_put(skb, sizeof(*nla));
 536                 nla->nla_type = NFQA_PAYLOAD;
 537                 nla->nla_len = nla_attr_size(data_len);
 538
 539                 if (skb_zerocopy(skb, entskb, data_len, hlen))
 540                         goto nla_put_failure;
 541         }
 542
 543         nlh->nlmsg_len = skb->len;
 544         return skb;
 545
 546 nla_put_failure:
 547         skb_tx_error(entskb);
 548         kfree_skb(skb);
 549         net_err_ratelimited("nf_queue: error creating packet message\n");
 550         return NULL;
 551 }
 552
 553 static int
 554 __nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
 555                         struct nf_queue_entry *entry)
 556 {
 557         struct sk_buff *nskb;
 558         int err = -ENOBUFS;
 559         __be32 *packet_id_ptr;
 560         int failopen = 0;
 561
 562         nskb = nfqnl_build_packet_message(net, queue, entry, &packet_id_ptr);
 563         if (nskb == NULL) {
 564                 err = -ENOMEM;
 565                 goto err_out;
 566         }
 567         spin_lock_bh(&queue->lock);
 568
 569         if (queue->queue_total >= queue->queue_maxlen) {
 570                 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
 571                         failopen = 1;
 572                         err = 0;
 573                 } else {
 574                         queue->queue_dropped++;
 575                         net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
 576                                              queue->queue_total);
 577                 }
 578                 goto err_out_free_nskb;
 579         }
 580         entry->id = ++queue->id_sequence;
 581         *packet_id_ptr = htonl(entry->id);
 582
 583         /* nfnetlink_unicast will either free the nskb or add it to a socket */
 584         err = nfnetlink_unicast(nskb, net, queue->peer_portid, MSG_DONTWAIT);
 585         if (err < 0) {
 586                 queue->queue_user_dropped++;
 587                 goto err_out_unlock;
 588         }
 589
 590         __enqueue_entry(queue, entry);
 591
 592         spin_unlock_bh(&queue->lock);
 593         return 0;
 594
 595 err_out_free_nskb:
 596         kfree_skb(nskb);
 597 err_out_unlock:
 598         spin_unlock_bh(&queue->lock);
 599         if (failopen)
 600                 nf_reinject(entry, NF_ACCEPT);
 601 err_out:
 602         return err;
 603 }
 604
 605 static struct nf_queue_entry *
 606 nf_queue_entry_dup(struct nf_queue_entry *e)
 607 {
 608         struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
 609         if (entry)
 610                 nf_queue_entry_get_refs(entry);
 611         return entry;
 612 }
 613
 614 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 615 /* When called from bridge netfilter, skb->data must point to MAC header
 616  * before calling skb_gso_segment(). Else, original MAC header is lost
 617  * and segmented skbs will be sent to wrong destination.
 618  */
 619 static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
 620 {
 621         if (skb->nf_bridge)
 622                 __skb_push(skb, skb->network_header - skb->mac_header);
 623 }
 624
 625 static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
 626 {
 627         if (skb->nf_bridge)
 628                 __skb_pull(skb, skb->network_header - skb->mac_header);
 629 }
 630 #else
 631 #define nf_bridge_adjust_skb_data(s) do {} while (0)
 632 #define nf_bridge_adjust_segmented_data(s) do {} while (0)
 633 #endif
 634
 635 static void free_entry(struct nf_queue_entry *entry)
 636 {
 637         nf_queue_entry_release_refs(entry);
 638         kfree(entry);
 639 }
 640
 641 static int
 642 __nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
 643                            struct sk_buff *skb, struct nf_queue_entry *entry)
 644 {
 645         int ret = -ENOMEM;
 646         struct nf_queue_entry *entry_seg;
 647
 648         nf_bridge_adjust_segmented_data(skb);
 649
 650         if (skb->next == NULL) { /* last packet, no need to copy entry */
 651                 struct sk_buff *gso_skb = entry->skb;
 652                 entry->skb = skb;
 653                 ret = __nfqnl_enqueue_packet(net, queue, entry);
 654                 if (ret)
 655                         entry->skb = gso_skb;
 656                 return ret;
 657         }
 658
 659         skb->next = NULL;
 660
 661         entry_seg = nf_queue_entry_dup(entry);
 662         if (entry_seg) {
 663                 entry_seg->skb = skb;
 664                 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
 665                 if (ret)
 666                         free_entry(entry_seg);
 667         }
 668         return ret;
 669 }
 670
 671 static int
 672 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 673 {
 674         unsigned int queued;
 675         struct nfqnl_instance *queue;
 676         struct sk_buff *skb, *segs;
 677         int err = -ENOBUFS;
 678         struct net *net = entry->state.net;
 679         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
 680
 681         /* rcu_read_lock()ed by nf_hook_slow() */
 682         queue = instance_lookup(q, queuenum);
 683         if (!queue)
 684                 return -ESRCH;
 685
 686         if (queue->copy_mode == NFQNL_COPY_NONE)
 687                 return -EINVAL;
 688
 689         skb = entry->skb;
 690
 691         switch (entry->state.pf) {
 692         case NFPROTO_IPV4:
 693                 skb->protocol = htons(ETH_P_IP);
 694                 break;
 695         case NFPROTO_IPV6:
 696                 skb->protocol = htons(ETH_P_IPV6);
 697                 break;
 698         }
 699
 700         if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(skb))
 701                 return __nfqnl_enqueue_packet(net, queue, entry);
 702
 703         nf_bridge_adjust_skb_data(skb);
 704         segs = skb_gso_segment(skb, 0);
 705         /* Does not use PTR_ERR to limit the number of error codes that can be
 706          * returned by nf_queue.  For instance, callers rely on -ESRCH to
 707          * mean 'ignore this hook'.
 708          */
 709         if (IS_ERR_OR_NULL(segs))
 710                 goto out_err;
 711         queued = 0;
 712         err = 0;
 713         do {
 714                 struct sk_buff *nskb = segs->next;
 715                 if (err == 0)
 716                         err = __nfqnl_enqueue_packet_gso(net, queue,
 717                                                         segs, entry);
 718                 if (err == 0)
 719                         queued++;
 720                 else
 721                         kfree_skb(segs);
 722                 segs = nskb;
 723         } while (segs);
 724
 725         if (queued) {
 726                 if (err) /* some segments are already queued */
 727                         free_entry(entry);
 728                 kfree_skb(skb);
 729                 return 0;
 730         }
 731  out_err:
 732         nf_bridge_adjust_segmented_data(skb);
 733         return err;
 734 }
 735
 736 static int
 737 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e, int diff)
 738 {
 739         struct sk_buff *nskb;
 740
 741         if (diff < 0) {
 742                 if (pskb_trim(e->skb, data_len))
 743                         return -ENOMEM;
 744         } else if (diff > 0) {
 745                 if (data_len > 0xFFFF)
 746                         return -EINVAL;
 747                 if (diff > skb_tailroom(e->skb)) {
 748                         nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
 749                                                diff, GFP_ATOMIC);
 750                         if (!nskb) {
 751                                 printk(KERN_WARNING "nf_queue: OOM "
 752                                       "in mangle, dropping packet\n");
 753                                 return -ENOMEM;
 754                         }
 755                         kfree_skb(e->skb);
 756                         e->skb = nskb;
 757                 }
 758                 skb_put(e->skb, diff);
 759         }
 760         if (!skb_make_writable(e->skb, data_len))
 761                 return -ENOMEM;
 762         skb_copy_to_linear_data(e->skb, data, data_len);
 763         e->skb->ip_summed = CHECKSUM_NONE;
 764         return 0;
 765 }
 766
 767 static int
 768 nfqnl_set_mode(struct nfqnl_instance *queue,
 769                unsigned char mode, unsigned int range)
 770 {
 771         int status = 0;
 772
 773         spin_lock_bh(&queue->lock);
 774         switch (mode) {
 775         case NFQNL_COPY_NONE:
 776         case NFQNL_COPY_META:
 777                 queue->copy_mode = mode;
 778                 queue->copy_range = 0;
 779                 break;
 780
 781         case NFQNL_COPY_PACKET:
 782                 queue->copy_mode = mode;
 783                 if (range == 0 || range > NFQNL_MAX_COPY_RANGE)
 784                         queue->copy_range = NFQNL_MAX_COPY_RANGE;
 785                 else
 786                         queue->copy_range = range;
 787                 break;
 788
 789         default:
 790                 status = -EINVAL;
 791
 792         }
 793         spin_unlock_bh(&queue->lock);
 794
 795         return status;
 796 }
 797
 798 static int
 799 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
 800 {
 801         if (entry->state.in)
 802                 if (entry->state.in->ifindex == ifindex)
 803                         return 1;
 804         if (entry->state.out)
 805                 if (entry->state.out->ifindex == ifindex)
 806                         return 1;
 807 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 808         if (entry->skb->nf_bridge) {
 809                 int physinif, physoutif;
 810
 811                 physinif = nf_bridge_get_physinif(entry->skb);
 812                 physoutif = nf_bridge_get_physoutif(entry->skb);
 813
 814                 if (physinif == ifindex || physoutif == ifindex)
 815                         return 1;
 816         }
 817 #endif
 818         return 0;
 819 }
 820
 821 /* drop all packets with either indev or outdev == ifindex from all queue
 822  * instances */
 823 static void
 824 nfqnl_dev_drop(struct net *net, int ifindex)
 825 {
 826         int i;
 827         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
 828
 829         rcu_read_lock();
 830
 831         for (i = 0; i < INSTANCE_BUCKETS; i++) {
 832                 struct nfqnl_instance *inst;
 833                 struct hlist_head *head = &q->instance_table[i];
 834
 835                 hlist_for_each_entry_rcu(inst, head, hlist)
 836                         nfqnl_flush(inst, dev_cmp, ifindex);
 837         }
 838
 839         rcu_read_unlock();
 840 }
 841
 842 static int
 843 nfqnl_rcv_dev_event(struct notifier_block *this,
 844                     unsigned long event, void *ptr)
 845 {
 846         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
 847
 848         /* Drop any packets associated with the downed device */
 849         if (event == NETDEV_DOWN)
 850                 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
 851         return NOTIFY_DONE;
 852 }
 853
 854 static struct notifier_block nfqnl_dev_notifier = {
 855         .notifier_call  = nfqnl_rcv_dev_event,
 856 };
 857
 858 static int nf_hook_cmp(struct nf_queue_entry *entry, unsigned long ops_ptr)
 859 {
 860         return entry->elem == (struct nf_hook_ops *)ops_ptr;
 861 }
 862
 863 static void nfqnl_nf_hook_drop(struct net *net, struct nf_hook_ops *hook)
 864 {
 865         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
 866         int i;
 867
 868         rcu_read_lock();
 869         for (i = 0; i < INSTANCE_BUCKETS; i++) {
 870                 struct nfqnl_instance *inst;
 871                 struct hlist_head *head = &q->instance_table[i];
 872
 873                 hlist_for_each_entry_rcu(inst, head, hlist)
 874                         nfqnl_flush(inst, nf_hook_cmp, (unsigned long)hook);
 875         }
 876         rcu_read_unlock();
 877 }
 878
 879 static int
 880 nfqnl_rcv_nl_event(struct notifier_block *this,
 881                    unsigned long event, void *ptr)
 882 {
 883         struct netlink_notify *n = ptr;
 884         struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
 885
 886         if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
 887                 int i;
 888
 889                 /* destroy all instances for this portid */
 890                 spin_lock(&q->instances_lock);
 891                 for (i = 0; i < INSTANCE_BUCKETS; i++) {
 892                         struct hlist_node *t2;
 893                         struct nfqnl_instance *inst;
 894                         struct hlist_head *head = &q->instance_table[i];
 895
 896                         hlist_for_each_entry_safe(inst, t2, head, hlist) {
 897                                 if (n->portid == inst->peer_portid)
 898                                         __instance_destroy(inst);
 899                         }
 900                 }
 901                 spin_unlock(&q->instances_lock);
 902         }
 903         return NOTIFY_DONE;
 904 }
 905
 906 static struct notifier_block nfqnl_rtnl_notifier = {
 907         .notifier_call  = nfqnl_rcv_nl_event,
 908 };
 909
 910 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
 911         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 912         [NFQA_MARK]             = { .type = NLA_U32 },
 913         [NFQA_PAYLOAD]          = { .type = NLA_UNSPEC },
 914         [NFQA_CT]               = { .type = NLA_UNSPEC },
 915         [NFQA_EXP]              = { .type = NLA_UNSPEC },
 916 };
 917
 918 static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
 919         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 920         [NFQA_MARK]             = { .type = NLA_U32 },
 921 };
 922
 923 static struct nfqnl_instance *
 924 verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, u32 nlportid)
 925 {
 926         struct nfqnl_instance *queue;
 927
 928         queue = instance_lookup(q, queue_num);
 929         if (!queue)
 930                 return ERR_PTR(-ENODEV);
 931
 932         if (queue->peer_portid != nlportid)
 933                 return ERR_PTR(-EPERM);
 934
 935         return queue;
 936 }
 937
 938 static struct nfqnl_msg_verdict_hdr*
 939 verdicthdr_get(const struct nlattr * const nfqa[])
 940 {
 941         struct nfqnl_msg_verdict_hdr *vhdr;
 942         unsigned int verdict;
 943
 944         if (!nfqa[NFQA_VERDICT_HDR])
 945                 return NULL;
 946
 947         vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
 948         verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
 949         if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
 950                 return NULL;
 951         return vhdr;
 952 }
 953
 954 static int nfq_id_after(unsigned int id, unsigned int max)
 955 {
 956         return (int)(id - max) > 0;
 957 }
 958
 959 static int
 960 nfqnl_recv_verdict_batch(struct sock *ctnl, struct sk_buff *skb,
 961                    const struct nlmsghdr *nlh,
 962                    const struct nlattr * const nfqa[])
 963 {
 964         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
 965         struct nf_queue_entry *entry, *tmp;
 966         unsigned int verdict, maxid;
 967         struct nfqnl_msg_verdict_hdr *vhdr;
 968         struct nfqnl_instance *queue;
 969         LIST_HEAD(batch_list);
 970         u16 queue_num = ntohs(nfmsg->res_id);
 971
 972         struct net *net = sock_net(ctnl);
 973         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
 974
 975         queue = verdict_instance_lookup(q, queue_num,
 976                                         NETLINK_CB(skb).portid);
 977         if (IS_ERR(queue))
 978                 return PTR_ERR(queue);
 979
 980         vhdr = verdicthdr_get(nfqa);
 981         if (!vhdr)
 982                 return -EINVAL;
 983
 984         verdict = ntohl(vhdr->verdict);
 985         maxid = ntohl(vhdr->id);
 986
 987         spin_lock_bh(&queue->lock);
 988
 989         list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
 990                 if (nfq_id_after(entry->id, maxid))
 991                         break;
 992                 __dequeue_entry(queue, entry);
 993                 list_add_tail(&entry->list, &batch_list);
 994         }
 995
 996         spin_unlock_bh(&queue->lock);
 997
 998         if (list_empty(&batch_list))
 999                 return -ENOENT;
1000
1001         list_for_each_entry_safe(entry, tmp, &batch_list, list) {
1002                 if (nfqa[NFQA_MARK])
1003                         entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1004                 nf_reinject(entry, verdict);
1005         }
1006         return 0;
1007 }
1008
1009 static struct nf_conn *nfqnl_ct_parse(struct nfnl_ct_hook *nfnl_ct,
1010                                       const struct nlmsghdr *nlh,
1011                                       const struct nlattr * const nfqa[],
1012                                       struct nf_queue_entry *entry,
1013                                       enum ip_conntrack_info *ctinfo)
1014 {
1015         struct nf_conn *ct;
1016
1017         ct = nfnl_ct->get_ct(entry->skb, ctinfo);
1018         if (ct == NULL)
1019                 return NULL;
1020
1021         if (nfnl_ct->parse(nfqa[NFQA_CT], ct) < 0)
1022                 return NULL;
1023
1024         if (nfqa[NFQA_EXP])
1025                 nfnl_ct->attach_expect(nfqa[NFQA_EXP], ct,
1026                                       NETLINK_CB(entry->skb).portid,
1027                                       nlmsg_report(nlh));
1028         return ct;
1029 }
1030
1031 static int
1032 nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
1033                    const struct nlmsghdr *nlh,
1034                    const struct nlattr * const nfqa[])
1035 {
1036         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1037         u_int16_t queue_num = ntohs(nfmsg->res_id);
1038
1039         struct nfqnl_msg_verdict_hdr *vhdr;
1040         struct nfqnl_instance *queue;
1041         unsigned int verdict;
1042         struct nf_queue_entry *entry;
1043         enum ip_conntrack_info uninitialized_var(ctinfo);
1044         struct nfnl_ct_hook *nfnl_ct;
1045         struct nf_conn *ct = NULL;
1046
1047         struct net *net = sock_net(ctnl);
1048         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1049
1050         queue = instance_lookup(q, queue_num);
1051         if (!queue)
1052                 queue = verdict_instance_lookup(q, queue_num,
1053                                                 NETLINK_CB(skb).portid);
1054         if (IS_ERR(queue))
1055                 return PTR_ERR(queue);
1056
1057         vhdr = verdicthdr_get(nfqa);
1058         if (!vhdr)
1059                 return -EINVAL;
1060
1061         verdict = ntohl(vhdr->verdict);
1062
1063         entry = find_dequeue_entry(queue, ntohl(vhdr->id));
1064         if (entry == NULL)
1065                 return -ENOENT;
1066
1067         if (nfqa[NFQA_CT]) {
1068                 /* rcu lock already held from nfnl->call_rcu. */
1069                 nfnl_ct = rcu_dereference(nfnl_ct_hook);
1070                 if (nfnl_ct != NULL)
1071                         ct = nfqnl_ct_parse(nfnl_ct, nlh, nfqa, entry, &ctinfo);
1072         }
1073
1074         if (nfqa[NFQA_PAYLOAD]) {
1075                 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
1076                 int diff = payload_len - entry->skb->len;
1077
1078                 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
1079                                  payload_len, entry, diff) < 0)
1080                         verdict = NF_DROP;
1081
1082                 if (ct && diff)
1083                         nfnl_ct->seq_adjust(entry->skb, ct, ctinfo, diff);
1084         }
1085
1086         if (nfqa[NFQA_MARK])
1087                 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1088
1089         nf_reinject(entry, verdict);
1090         return 0;
1091 }
1092
1093 static int
1094 nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
1095                   const struct nlmsghdr *nlh,
1096                   const struct nlattr * const nfqa[])
1097 {
1098         return -ENOTSUPP;
1099 }
1100
1101 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1102         [NFQA_CFG_CMD]          = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1103         [NFQA_CFG_PARAMS]       = { .len = sizeof(struct nfqnl_msg_config_params) },
1104 };
1105
1106 static const struct nf_queue_handler nfqh = {
1107         .outfn          = &nfqnl_enqueue_packet,
1108         .nf_hook_drop   = &nfqnl_nf_hook_drop,
1109 };
1110
1111 static int
1112 nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
1113                   const struct nlmsghdr *nlh,
1114                   const struct nlattr * const nfqa[])
1115 {
1116         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1117         u_int16_t queue_num = ntohs(nfmsg->res_id);
1118         struct nfqnl_instance *queue;
1119         struct nfqnl_msg_config_cmd *cmd = NULL;
1120         struct net *net = sock_net(ctnl);
1121         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1122         int ret = 0;
1123
1124         if (nfqa[NFQA_CFG_CMD]) {
1125                 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1126
1127                 /* Obsolete commands without queue context */
1128                 switch (cmd->command) {
1129                 case NFQNL_CFG_CMD_PF_BIND: return 0;
1130                 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1131                 }
1132         }
1133
1134         rcu_read_lock();
1135         queue = instance_lookup(q, queue_num);
1136         if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1137                 ret = -EPERM;
1138                 goto err_out_unlock;
1139         }
1140
1141         if (cmd != NULL) {
1142                 switch (cmd->command) {
1143                 case NFQNL_CFG_CMD_BIND:
1144                         if (queue) {
1145                                 ret = -EBUSY;
1146                                 goto err_out_unlock;
1147                         }
1148                         queue = instance_create(q, queue_num,
1149                                                 NETLINK_CB(skb).portid);
1150                         if (IS_ERR(queue)) {
1151                                 ret = PTR_ERR(queue);
1152                                 goto err_out_unlock;
1153                         }
1154                         break;
1155                 case NFQNL_CFG_CMD_UNBIND:
1156                         if (!queue) {
1157                                 ret = -ENODEV;
1158                                 goto err_out_unlock;
1159                         }
1160                         instance_destroy(q, queue);
1161                         break;
1162                 case NFQNL_CFG_CMD_PF_BIND:
1163                 case NFQNL_CFG_CMD_PF_UNBIND:
1164                         break;
1165                 default:
1166                         ret = -ENOTSUPP;
1167                         break;
1168                 }
1169         }
1170
1171         if (nfqa[NFQA_CFG_PARAMS]) {
1172                 struct nfqnl_msg_config_params *params;
1173
1174                 if (!queue) {
1175                         ret = -ENODEV;
1176                         goto err_out_unlock;
1177                 }
1178                 params = nla_data(nfqa[NFQA_CFG_PARAMS]);
1179                 nfqnl_set_mode(queue, params->copy_mode,
1180                                 ntohl(params->copy_range));
1181         }
1182
1183         if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1184                 __be32 *queue_maxlen;
1185
1186                 if (!queue) {
1187                         ret = -ENODEV;
1188                         goto err_out_unlock;
1189                 }
1190                 queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1191                 spin_lock_bh(&queue->lock);
1192                 queue->queue_maxlen = ntohl(*queue_maxlen);
1193                 spin_unlock_bh(&queue->lock);
1194         }
1195
1196         if (nfqa[NFQA_CFG_FLAGS]) {
1197                 __u32 flags, mask;
1198
1199                 if (!queue) {
1200                         ret = -ENODEV;
1201                         goto err_out_unlock;
1202                 }
1203
1204                 if (!nfqa[NFQA_CFG_MASK]) {
1205                         /* A mask is needed to specify which flags are being
1206                          * changed.
1207                          */
1208                         ret = -EINVAL;
1209                         goto err_out_unlock;
1210                 }
1211
1212                 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1213                 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1214
1215                 if (flags >= NFQA_CFG_F_MAX) {
1216                         ret = -EOPNOTSUPP;
1217                         goto err_out_unlock;
1218                 }
1219 #if !IS_ENABLED(CONFIG_NETWORK_SECMARK)
1220                 if (flags & mask & NFQA_CFG_F_SECCTX) {
1221                         ret = -EOPNOTSUPP;
1222                         goto err_out_unlock;
1223                 }
1224 #endif
1225                 spin_lock_bh(&queue->lock);
1226                 queue->flags &= ~mask;
1227                 queue->flags |= flags & mask;
1228                 spin_unlock_bh(&queue->lock);
1229         }
1230
1231 err_out_unlock:
1232         rcu_read_unlock();
1233         return ret;
1234 }
1235
1236 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1237         [NFQNL_MSG_PACKET]      = { .call_rcu = nfqnl_recv_unsupp,
1238                                     .attr_count = NFQA_MAX, },
1239         [NFQNL_MSG_VERDICT]     = { .call_rcu = nfqnl_recv_verdict,
1240                                     .attr_count = NFQA_MAX,
1241                                     .policy = nfqa_verdict_policy },
1242         [NFQNL_MSG_CONFIG]      = { .call = nfqnl_recv_config,
1243                                     .attr_count = NFQA_CFG_MAX,
1244                                     .policy = nfqa_cfg_policy },
1245         [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
1246                                     .attr_count = NFQA_MAX,
1247                                     .policy = nfqa_verdict_batch_policy },
1248 };
1249
1250 static const struct nfnetlink_subsystem nfqnl_subsys = {
1251         .name           = "nf_queue",
1252         .subsys_id      = NFNL_SUBSYS_QUEUE,
1253         .cb_count       = NFQNL_MSG_MAX,
1254         .cb             = nfqnl_cb,
1255 };
1256
1257 #ifdef CONFIG_PROC_FS
1258 struct iter_state {
1259         struct seq_net_private p;
1260         unsigned int bucket;
1261 };
1262
1263 static struct hlist_node *get_first(struct seq_file *seq)
1264 {
1265         struct iter_state *st = seq->private;
1266         struct net *net;
1267         struct nfnl_queue_net *q;
1268
1269         if (!st)
1270                 return NULL;
1271
1272         net = seq_file_net(seq);
1273         q = nfnl_queue_pernet(net);
1274         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1275                 if (!hlist_empty(&q->instance_table[st->bucket]))
1276                         return q->instance_table[st->bucket].first;
1277         }
1278         return NULL;
1279 }
1280
1281 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1282 {
1283         struct iter_state *st = seq->private;
1284         struct net *net = seq_file_net(seq);
1285
1286         h = h->next;
1287         while (!h) {
1288                 struct nfnl_queue_net *q;
1289
1290                 if (++st->bucket >= INSTANCE_BUCKETS)
1291                         return NULL;
1292
1293                 q = nfnl_queue_pernet(net);
1294                 h = q->instance_table[st->bucket].first;
1295         }
1296         return h;
1297 }
1298
1299 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1300 {
1301         struct hlist_node *head;
1302         head = get_first(seq);
1303
1304         if (head)
1305                 while (pos && (head = get_next(seq, head)))
1306                         pos--;
1307         return pos ? NULL : head;
1308 }
1309
1310 static void *seq_start(struct seq_file *s, loff_t *pos)
1311         __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1312 {
1313         spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1314         return get_idx(s, *pos);
1315 }
1316
1317 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1318 {
1319         (*pos)++;
1320         return get_next(s, v);
1321 }
1322
1323 static void seq_stop(struct seq_file *s, void *v)
1324         __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1325 {
1326         spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1327 }
1328
1329 static int seq_show(struct seq_file *s, void *v)
1330 {
1331         const struct nfqnl_instance *inst = v;
1332
1333         seq_printf(s, "%5u %6u %5u %1u %5u %5u %5u %8u %2d\n",
1334                    inst->queue_num,
1335                    inst->peer_portid, inst->queue_total,
1336                    inst->copy_mode, inst->copy_range,
1337                    inst->queue_dropped, inst->queue_user_dropped,
1338                    inst->id_sequence, 1);
1339         return 0;
1340 }
1341
1342 static const struct seq_operations nfqnl_seq_ops = {
1343         .start  = seq_start,
1344         .next   = seq_next,
1345         .stop   = seq_stop,
1346         .show   = seq_show,
1347 };
1348
1349 static int nfqnl_open(struct inode *inode, struct file *file)
1350 {
1351         return seq_open_net(inode, file, &nfqnl_seq_ops,
1352                         sizeof(struct iter_state));
1353 }
1354
1355 static const struct file_operations nfqnl_file_ops = {
1356         .owner   = THIS_MODULE,
1357         .open    = nfqnl_open,
1358         .read    = seq_read,
1359         .llseek  = seq_lseek,
1360         .release = seq_release_net,
1361 };
1362
1363 #endif /* PROC_FS */
1364
1365 static int __net_init nfnl_queue_net_init(struct net *net)
1366 {
1367         unsigned int i;
1368         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1369
1370         for (i = 0; i < INSTANCE_BUCKETS; i++)
1371                 INIT_HLIST_HEAD(&q->instance_table[i]);
1372
1373         spin_lock_init(&q->instances_lock);
1374
1375 #ifdef CONFIG_PROC_FS
1376         if (!proc_create("nfnetlink_queue", 0440,
1377                          net->nf.proc_netfilter, &nfqnl_file_ops))
1378                 return -ENOMEM;
1379 #endif
1380         return 0;
1381 }
1382
1383 static void __net_exit nfnl_queue_net_exit(struct net *net)
1384 {
1385 #ifdef CONFIG_PROC_FS
1386         remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1387 #endif
1388 }
1389
1390 static struct pernet_operations nfnl_queue_net_ops = {
1391         .init   = nfnl_queue_net_init,
1392         .exit   = nfnl_queue_net_exit,
1393         .id     = &nfnl_queue_net_id,
1394         .size   = sizeof(struct nfnl_queue_net),
1395 };
1396
1397 static int __init nfnetlink_queue_init(void)
1398 {
1399         int status;
1400
1401         status = register_pernet_subsys(&nfnl_queue_net_ops);
1402         if (status < 0) {
1403                 pr_err("nf_queue: failed to register pernet ops\n");
1404                 goto out;
1405         }
1406
1407         netlink_register_notifier(&nfqnl_rtnl_notifier);
1408         status = nfnetlink_subsys_register(&nfqnl_subsys);
1409         if (status < 0) {
1410                 pr_err("nf_queue: failed to create netlink socket\n");
1411                 goto cleanup_netlink_notifier;
1412         }
1413
1414         register_netdevice_notifier(&nfqnl_dev_notifier);
1415         nf_register_queue_handler(&nfqh);
1416         return status;
1417
1418 cleanup_netlink_notifier:
1419         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1420 out:
1421         return status;
1422 }
1423
1424 static void __exit nfnetlink_queue_fini(void)
1425 {
1426         nf_unregister_queue_handler();
1427         unregister_netdevice_notifier(&nfqnl_dev_notifier);
1428         nfnetlink_subsys_unregister(&nfqnl_subsys);
1429         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1430         unregister_pernet_subsys(&nfnl_queue_net_ops);
1431
1432         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1433 }
1434
1435 MODULE_DESCRIPTION("netfilter packet queue handler");
1436 MODULE_AUTHOR("Harald Welte <[email protected]>");
1437 MODULE_LICENSE("GPL");
1438 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1439
1440 module_init(nfnetlink_queue_init);
1441 module_exit(nfnetlink_queue_fini);