net/netfilter/nfnetlink_queue.c

   1 /*
   2  * This is a module which is used for queueing packets and communicating with
   3  * userspace via nfnetlink.
   4  *
   5  * (C) 2005 by Harald Welte <laforge@netfilter.org>
   6  * (C) 2007 by Patrick McHardy <kaber@trash.net>
   7  *
   8  * Based on the old ipv4-only ip_queue.c:
   9  * (C) 2000-2002 James Morris <jmorris@intercode.com.au>
  10  * (C) 2003-2005 Netfilter Core Team <coreteam@netfilter.org>
  11  *
  12  * This program is free software; you can redistribute it and/or modify
  13  * it under the terms of the GNU General Public License version 2 as
  14  * published by the Free Software Foundation.
  15  *
  16  */
  17 #include <linux/module.h>
  18 #include <linux/skbuff.h>
  19 #include <linux/init.h>
  20 #include <linux/spinlock.h>
  21 #include <linux/slab.h>
  22 #include <linux/notifier.h>
  23 #include <linux/netdevice.h>
  24 #include <linux/netfilter.h>
  25 #include <linux/proc_fs.h>
  26 #include <linux/netfilter_ipv4.h>
  27 #include <linux/netfilter_ipv6.h>
  28 #include <linux/netfilter_bridge.h>
  29 #include <linux/netfilter/nfnetlink.h>
  30 #include <linux/netfilter/nfnetlink_queue.h>
  31 #include <linux/netfilter/nf_conntrack_common.h>
  32 #include <linux/list.h>
  33 #include <net/sock.h>
  34 #include <net/tcp_states.h>
  35 #include <net/netfilter/nf_queue.h>
  36 #include <net/netns/generic.h>
  37
  38 #include <linux/atomic.h>
  39
  40 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  41 #include "../bridge/br_private.h"
  42 #endif
  43
  44 #define NFQNL_QMAX_DEFAULT 1024
  45
  46 /* We're using struct nlattr which has 16bit nla_len. Note that nla_len
  47  * includes the header length. Thus, the maximum packet length that we
  48  * support is 65531 bytes. We send truncated packets if the specified length
  49  * is larger than that.  Userspace can check for presence of NFQA_CAP_LEN
  50  * attribute to detect truncation.
  51  */
  52 #define NFQNL_MAX_COPY_RANGE (0xffff - NLA_HDRLEN)
  53
  54 struct nfqnl_instance {
  55         struct hlist_node hlist;                /* global list of queues */
  56         struct rcu_head rcu;
  57
  58         u32 peer_portid;
  59         unsigned int queue_maxlen;
  60         unsigned int copy_range;
  61         unsigned int queue_dropped;
  62         unsigned int queue_user_dropped;
  63
  64
  65         u_int16_t queue_num;                    /* number of this queue */
  66         u_int8_t copy_mode;
  67         u_int32_t flags;                        /* Set using NFQA_CFG_FLAGS */
  68 /*
  69  * Following fields are dirtied for each queued packet,
  70  * keep them in same cache line if possible.
  71  */
  72         spinlock_t      lock;
  73         unsigned int    queue_total;
  74         unsigned int    id_sequence;            /* 'sequence' of pkt ids */
  75         struct list_head queue_list;            /* packets in queue */
  76 };
  77
  78 typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
  79
  80 static int nfnl_queue_net_id __read_mostly;
  81
  82 #define INSTANCE_BUCKETS        16
  83 struct nfnl_queue_net {
  84         spinlock_t instances_lock;
  85         struct hlist_head instance_table[INSTANCE_BUCKETS];
  86 };
  87
  88 static struct nfnl_queue_net *nfnl_queue_pernet(struct net *net)
  89 {
  90         return net_generic(net, nfnl_queue_net_id);
  91 }
  92
  93 static inline u_int8_t instance_hashfn(u_int16_t queue_num)
  94 {
  95         return ((queue_num >> 8) ^ queue_num) % INSTANCE_BUCKETS;
  96 }
  97
  98 static struct nfqnl_instance *
  99 instance_lookup(struct nfnl_queue_net *q, u_int16_t queue_num)
 100 {
 101         struct hlist_head *head;
 102         struct nfqnl_instance *inst;
 103
 104         head = &q->instance_table[instance_hashfn(queue_num)];
 105         hlist_for_each_entry_rcu(inst, head, hlist) {
 106                 if (inst->queue_num == queue_num)
 107                         return inst;
 108         }
 109         return NULL;
 110 }
 111
 112 static struct nfqnl_instance *
 113 instance_create(struct nfnl_queue_net *q, u_int16_t queue_num, u32 portid)
 114 {
 115         struct nfqnl_instance *inst;
 116         unsigned int h;
 117         int err;
 118
 119         spin_lock(&q->instances_lock);
 120         if (instance_lookup(q, queue_num)) {
 121                 err = -EEXIST;
 122                 goto out_unlock;
 123         }
 124
 125         inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
 126         if (!inst) {
 127                 err = -ENOMEM;
 128                 goto out_unlock;
 129         }
 130
 131         inst->queue_num = queue_num;
 132         inst->peer_portid = portid;
 133         inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
 134         inst->copy_range = NFQNL_MAX_COPY_RANGE;
 135         inst->copy_mode = NFQNL_COPY_NONE;
 136         spin_lock_init(&inst->lock);
 137         INIT_LIST_HEAD(&inst->queue_list);
 138
 139         if (!try_module_get(THIS_MODULE)) {
 140                 err = -EAGAIN;
 141                 goto out_free;
 142         }
 143
 144         h = instance_hashfn(queue_num);
 145         hlist_add_head_rcu(&inst->hlist, &q->instance_table[h]);
 146
 147         spin_unlock(&q->instances_lock);
 148
 149         return inst;
 150
 151 out_free:
 152         kfree(inst);
 153 out_unlock:
 154         spin_unlock(&q->instances_lock);
 155         return ERR_PTR(err);
 156 }
 157
 158 static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
 159                         unsigned long data);
 160
 161 static void
 162 instance_destroy_rcu(struct rcu_head *head)
 163 {
 164         struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
 165                                                    rcu);
 166
 167         nfqnl_flush(inst, NULL, 0);
 168         kfree(inst);
 169         module_put(THIS_MODULE);
 170 }
 171
 172 static void
 173 __instance_destroy(struct nfqnl_instance *inst)
 174 {
 175         hlist_del_rcu(&inst->hlist);
 176         call_rcu(&inst->rcu, instance_destroy_rcu);
 177 }
 178
 179 static void
 180 instance_destroy(struct nfnl_queue_net *q, struct nfqnl_instance *inst)
 181 {
 182         spin_lock(&q->instances_lock);
 183         __instance_destroy(inst);
 184         spin_unlock(&q->instances_lock);
 185 }
 186
 187 static inline void
 188 __enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
 189 {
 190        list_add_tail(&entry->list, &queue->queue_list);
 191        queue->queue_total++;
 192 }
 193
 194 static void
 195 __dequeue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
 196 {
 197         list_del(&entry->list);
 198         queue->queue_total--;
 199 }
 200
 201 static struct nf_queue_entry *
 202 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
 203 {
 204         struct nf_queue_entry *entry = NULL, *i;
 205
 206         spin_lock_bh(&queue->lock);
 207
 208         list_for_each_entry(i, &queue->queue_list, list) {
 209                 if (i->id == id) {
 210                         entry = i;
 211                         break;
 212                 }
 213         }
 214
 215         if (entry)
 216                 __dequeue_entry(queue, entry);
 217
 218         spin_unlock_bh(&queue->lock);
 219
 220         return entry;
 221 }
 222
 223 static void
 224 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
 225 {
 226         struct nf_queue_entry *entry, *next;
 227
 228         spin_lock_bh(&queue->lock);
 229         list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
 230                 if (!cmpfn || cmpfn(entry, data)) {
 231                         list_del(&entry->list);
 232                         queue->queue_total--;
 233                         nf_reinject(entry, NF_DROP);
 234                 }
 235         }
 236         spin_unlock_bh(&queue->lock);
 237 }
 238
 239 static int
 240 nfqnl_put_packet_info(struct sk_buff *nlskb, struct sk_buff *packet,
 241                       bool csum_verify)
 242 {
 243         __u32 flags = 0;
 244
 245         if (packet->ip_summed == CHECKSUM_PARTIAL)
 246                 flags = NFQA_SKB_CSUMNOTREADY;
 247         else if (csum_verify)
 248                 flags = NFQA_SKB_CSUM_NOTVERIFIED;
 249
 250         if (skb_is_gso(packet))
 251                 flags |= NFQA_SKB_GSO;
 252
 253         return flags ? nla_put_be32(nlskb, NFQA_SKB_INFO, htonl(flags)) : 0;
 254 }
 255
 256 static int nfqnl_put_sk_uidgid(struct sk_buff *skb, struct sock *sk)
 257 {
 258         const struct cred *cred;
 259
 260         if (!sk_fullsock(sk))
 261                 return 0;
 262
 263         read_lock_bh(&sk->sk_callback_lock);
 264         if (sk->sk_socket && sk->sk_socket->file) {
 265                 cred = sk->sk_socket->file->f_cred;
 266                 if (nla_put_be32(skb, NFQA_UID,
 267                     htonl(from_kuid_munged(&init_user_ns, cred->fsuid))))
 268                         goto nla_put_failure;
 269                 if (nla_put_be32(skb, NFQA_GID,
 270                     htonl(from_kgid_munged(&init_user_ns, cred->fsgid))))
 271                         goto nla_put_failure;
 272         }
 273         read_unlock_bh(&sk->sk_callback_lock);
 274         return 0;
 275
 276 nla_put_failure:
 277         read_unlock_bh(&sk->sk_callback_lock);
 278         return -1;
 279 }
 280
 281 static u32 nfqnl_get_sk_secctx(struct sk_buff *skb, char **secdata)
 282 {
 283         u32 seclen = 0;
 284 #if IS_ENABLED(CONFIG_NETWORK_SECMARK)
 285         if (!skb || !sk_fullsock(skb->sk))
 286                 return 0;
 287
 288         read_lock_bh(&skb->sk->sk_callback_lock);
 289
 290         if (skb->secmark)
 291                 security_secid_to_secctx(skb->secmark, secdata, &seclen);
 292
 293         read_unlock_bh(&skb->sk->sk_callback_lock);
 294 #endif
 295         return seclen;
 296 }
 297
 298 static u32 nfqnl_get_bridge_size(struct nf_queue_entry *entry)
 299 {
 300         struct sk_buff *entskb = entry->skb;
 301         u32 nlalen = 0;
 302
 303         if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb))
 304                 return 0;
 305
 306         if (skb_vlan_tag_present(entskb))
 307                 nlalen += nla_total_size(nla_total_size(sizeof(__be16)) +
 308                                          nla_total_size(sizeof(__be16)));
 309
 310         if (entskb->network_header > entskb->mac_header)
 311                 nlalen += nla_total_size((entskb->network_header -
 312                                           entskb->mac_header));
 313
 314         return nlalen;
 315 }
 316
 317 static int nfqnl_put_bridge(struct nf_queue_entry *entry, struct sk_buff *skb)
 318 {
 319         struct sk_buff *entskb = entry->skb;
 320
 321         if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb))
 322                 return 0;
 323
 324         if (skb_vlan_tag_present(entskb)) {
 325                 struct nlattr *nest;
 326
 327                 nest = nla_nest_start(skb, NFQA_VLAN | NLA_F_NESTED);
 328                 if (!nest)
 329                         goto nla_put_failure;
 330
 331                 if (nla_put_be16(skb, NFQA_VLAN_TCI, htons(entskb->vlan_tci)) ||
 332                     nla_put_be16(skb, NFQA_VLAN_PROTO, entskb->vlan_proto))
 333                         goto nla_put_failure;
 334
 335                 nla_nest_end(skb, nest);
 336         }
 337
 338         if (entskb->mac_header < entskb->network_header) {
 339                 int len = (int)(entskb->network_header - entskb->mac_header);
 340
 341                 if (nla_put(skb, NFQA_L2HDR, len, skb_mac_header(entskb)))
 342                         goto nla_put_failure;
 343         }
 344
 345         return 0;
 346
 347 nla_put_failure:
 348         return -1;
 349 }
 350
 351 static struct sk_buff *
 352 nfqnl_build_packet_message(struct net *net, struct nfqnl_instance *queue,
 353                            struct nf_queue_entry *entry,
 354                            __be32 **packet_id_ptr)
 355 {
 356         size_t size;
 357         size_t data_len = 0, cap_len = 0;
 358         unsigned int hlen = 0;
 359         struct sk_buff *skb;
 360         struct nlattr *nla;
 361         struct nfqnl_msg_packet_hdr *pmsg;
 362         struct nlmsghdr *nlh;
 363         struct nfgenmsg *nfmsg;
 364         struct sk_buff *entskb = entry->skb;
 365         struct net_device *indev;
 366         struct net_device *outdev;
 367         struct nf_conn *ct = NULL;
 368         enum ip_conntrack_info uninitialized_var(ctinfo);
 369         struct nfnl_ct_hook *nfnl_ct;
 370         bool csum_verify;
 371         char *secdata = NULL;
 372         u32 seclen = 0;
 373
 374         size =    nlmsg_total_size(sizeof(struct nfgenmsg))
 375                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
 376                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 377                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 378 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 379                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 380                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 381 #endif
 382                 + nla_total_size(sizeof(u_int32_t))     /* mark */
 383                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
 384                 + nla_total_size(sizeof(u_int32_t))     /* skbinfo */
 385                 + nla_total_size(sizeof(u_int32_t));    /* cap_len */
 386
 387         if (entskb->tstamp.tv64)
 388                 size += nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
 389
 390         size += nfqnl_get_bridge_size(entry);
 391
 392         if (entry->state.hook <= NF_INET_FORWARD ||
 393            (entry->state.hook == NF_INET_POST_ROUTING && entskb->sk == NULL))
 394                 csum_verify = !skb_csum_unnecessary(entskb);
 395         else
 396                 csum_verify = false;
 397
 398         outdev = entry->state.out;
 399
 400         switch ((enum nfqnl_config_mode)ACCESS_ONCE(queue->copy_mode)) {
 401         case NFQNL_COPY_META:
 402         case NFQNL_COPY_NONE:
 403                 break;
 404
 405         case NFQNL_COPY_PACKET:
 406                 if (!(queue->flags & NFQA_CFG_F_GSO) &&
 407                     entskb->ip_summed == CHECKSUM_PARTIAL &&
 408                     skb_checksum_help(entskb))
 409                         return NULL;
 410
 411                 data_len = ACCESS_ONCE(queue->copy_range);
 412                 if (data_len > entskb->len)
 413                         data_len = entskb->len;
 414
 415                 hlen = skb_zerocopy_headlen(entskb);
 416                 hlen = min_t(unsigned int, hlen, data_len);
 417                 size += sizeof(struct nlattr) + hlen;
 418                 cap_len = entskb->len;
 419                 break;
 420         }
 421
 422         nfnl_ct = rcu_dereference(nfnl_ct_hook);
 423
 424         if (queue->flags & NFQA_CFG_F_CONNTRACK) {
 425                 if (nfnl_ct != NULL) {
 426                         ct = nfnl_ct->get_ct(entskb, &ctinfo);
 427                         if (ct != NULL)
 428                                 size += nfnl_ct->build_size(ct);
 429                 }
 430         }
 431
 432         if (queue->flags & NFQA_CFG_F_UID_GID) {
 433                 size +=  (nla_total_size(sizeof(u_int32_t))     /* uid */
 434                         + nla_total_size(sizeof(u_int32_t)));   /* gid */
 435         }
 436
 437         if ((queue->flags & NFQA_CFG_F_SECCTX) && entskb->sk) {
 438                 seclen = nfqnl_get_sk_secctx(entskb, &secdata);
 439                 if (seclen)
 440                         size += nla_total_size(seclen);
 441         }
 442
 443         skb = alloc_skb(size, GFP_ATOMIC);
 444         if (!skb) {
 445                 skb_tx_error(entskb);
 446                 goto nlmsg_failure;
 447         }
 448
 449         nlh = nlmsg_put(skb, 0, 0,
 450                         NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
 451                         sizeof(struct nfgenmsg), 0);
 452         if (!nlh) {
 453                 skb_tx_error(entskb);
 454                 kfree_skb(skb);
 455                 goto nlmsg_failure;
 456         }
 457         nfmsg = nlmsg_data(nlh);
 458         nfmsg->nfgen_family = entry->state.pf;
 459         nfmsg->version = NFNETLINK_V0;
 460         nfmsg->res_id = htons(queue->queue_num);
 461
 462         nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
 463         pmsg = nla_data(nla);
 464         pmsg->hw_protocol       = entskb->protocol;
 465         pmsg->hook              = entry->state.hook;
 466         *packet_id_ptr          = &pmsg->packet_id;
 467
 468         indev = entry->state.in;
 469         if (indev) {
 470 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 471                 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
 472                         goto nla_put_failure;
 473 #else
 474                 if (entry->state.pf == PF_BRIDGE) {
 475                         /* Case 1: indev is physical input device, we need to
 476                          * look for bridge group (when called from
 477                          * netfilter_bridge) */
 478                         if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
 479                                          htonl(indev->ifindex)) ||
 480                         /* this is the bridge group "brX" */
 481                         /* rcu_read_lock()ed by __nf_queue */
 482                             nla_put_be32(skb, NFQA_IFINDEX_INDEV,
 483                                          htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
 484                                 goto nla_put_failure;
 485                 } else {
 486                         int physinif;
 487
 488                         /* Case 2: indev is bridge group, we need to look for
 489                          * physical device (when called from ipv4) */
 490                         if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
 491                                          htonl(indev->ifindex)))
 492                                 goto nla_put_failure;
 493
 494                         physinif = nf_bridge_get_physinif(entskb);
 495                         if (physinif &&
 496                             nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
 497                                          htonl(physinif)))
 498                                 goto nla_put_failure;
 499                 }
 500 #endif
 501         }
 502
 503         if (outdev) {
 504 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 505                 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
 506                         goto nla_put_failure;
 507 #else
 508                 if (entry->state.pf == PF_BRIDGE) {
 509                         /* Case 1: outdev is physical output device, we need to
 510                          * look for bridge group (when called from
 511                          * netfilter_bridge) */
 512                         if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 513                                          htonl(outdev->ifindex)) ||
 514                         /* this is the bridge group "brX" */
 515                         /* rcu_read_lock()ed by __nf_queue */
 516                             nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
 517                                          htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
 518                                 goto nla_put_failure;
 519                 } else {
 520                         int physoutif;
 521
 522                         /* Case 2: outdev is bridge group, we need to look for
 523                          * physical output device (when called from ipv4) */
 524                         if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
 525                                          htonl(outdev->ifindex)))
 526                                 goto nla_put_failure;
 527
 528                         physoutif = nf_bridge_get_physoutif(entskb);
 529                         if (physoutif &&
 530                             nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 531                                          htonl(physoutif)))
 532                                 goto nla_put_failure;
 533                 }
 534 #endif
 535         }
 536
 537         if (entskb->mark &&
 538             nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
 539                 goto nla_put_failure;
 540
 541         if (indev && entskb->dev &&
 542             skb_mac_header_was_set(entskb)) {
 543                 struct nfqnl_msg_packet_hw phw;
 544                 int len;
 545
 546                 memset(&phw, 0, sizeof(phw));
 547                 len = dev_parse_header(entskb, phw.hw_addr);
 548                 if (len) {
 549                         phw.hw_addrlen = htons(len);
 550                         if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
 551                                 goto nla_put_failure;
 552                 }
 553         }
 554
 555         if (nfqnl_put_bridge(entry, skb) < 0)
 556                 goto nla_put_failure;
 557
 558         if (entskb->tstamp.tv64) {
 559                 struct nfqnl_msg_packet_timestamp ts;
 560                 struct timespec64 kts = ktime_to_timespec64(entskb->tstamp);
 561
 562                 ts.sec = cpu_to_be64(kts.tv_sec);
 563                 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
 564
 565                 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
 566                         goto nla_put_failure;
 567         }
 568
 569         if ((queue->flags & NFQA_CFG_F_UID_GID) && entskb->sk &&
 570             nfqnl_put_sk_uidgid(skb, entskb->sk) < 0)
 571                 goto nla_put_failure;
 572
 573         if (seclen && nla_put(skb, NFQA_SECCTX, seclen, secdata))
 574                 goto nla_put_failure;
 575
 576         if (ct && nfnl_ct->build(skb, ct, ctinfo, NFQA_CT, NFQA_CT_INFO) < 0)
 577                 goto nla_put_failure;
 578
 579         if (cap_len > data_len &&
 580             nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
 581                 goto nla_put_failure;
 582
 583         if (nfqnl_put_packet_info(skb, entskb, csum_verify))
 584                 goto nla_put_failure;
 585
 586         if (data_len) {
 587                 struct nlattr *nla;
 588
 589                 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
 590                         goto nla_put_failure;
 591
 592                 nla = (struct nlattr *)skb_put(skb, sizeof(*nla));
 593                 nla->nla_type = NFQA_PAYLOAD;
 594                 nla->nla_len = nla_attr_size(data_len);
 595
 596                 if (skb_zerocopy(skb, entskb, data_len, hlen))
 597                         goto nla_put_failure;
 598         }
 599
 600         nlh->nlmsg_len = skb->len;
 601         if (seclen)
 602                 security_release_secctx(secdata, seclen);
 603         return skb;
 604
 605 nla_put_failure:
 606         skb_tx_error(entskb);
 607         kfree_skb(skb);
 608         net_err_ratelimited("nf_queue: error creating packet message\n");
 609 nlmsg_failure:
 610         if (seclen)
 611                 security_release_secctx(secdata, seclen);
 612         return NULL;
 613 }
 614
 615 static int
 616 __nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
 617                         struct nf_queue_entry *entry)
 618 {
 619         struct sk_buff *nskb;
 620         int err = -ENOBUFS;
 621         __be32 *packet_id_ptr;
 622         int failopen = 0;
 623
 624         nskb = nfqnl_build_packet_message(net, queue, entry, &packet_id_ptr);
 625         if (nskb == NULL) {
 626                 err = -ENOMEM;
 627                 goto err_out;
 628         }
 629         spin_lock_bh(&queue->lock);
 630
 631         if (queue->queue_total >= queue->queue_maxlen) {
 632                 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
 633                         failopen = 1;
 634                         err = 0;
 635                 } else {
 636                         queue->queue_dropped++;
 637                         net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
 638                                              queue->queue_total);
 639                 }
 640                 goto err_out_free_nskb;
 641         }
 642         entry->id = ++queue->id_sequence;
 643         *packet_id_ptr = htonl(entry->id);
 644
 645         /* nfnetlink_unicast will either free the nskb or add it to a socket */
 646         err = nfnetlink_unicast(nskb, net, queue->peer_portid, MSG_DONTWAIT);
 647         if (err < 0) {
 648                 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
 649                         failopen = 1;
 650                         err = 0;
 651                 } else {
 652                         queue->queue_user_dropped++;
 653                 }
 654                 goto err_out_unlock;
 655         }
 656
 657         __enqueue_entry(queue, entry);
 658
 659         spin_unlock_bh(&queue->lock);
 660         return 0;
 661
 662 err_out_free_nskb:
 663         kfree_skb(nskb);
 664 err_out_unlock:
 665         spin_unlock_bh(&queue->lock);
 666         if (failopen)
 667                 nf_reinject(entry, NF_ACCEPT);
 668 err_out:
 669         return err;
 670 }
 671
 672 static struct nf_queue_entry *
 673 nf_queue_entry_dup(struct nf_queue_entry *e)
 674 {
 675         struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
 676
 677         if (!entry)
 678                 return NULL;
 679
 680         if (nf_queue_entry_get_refs(entry))
 681                 return entry;
 682
 683         kfree(entry);
 684         return NULL;
 685 }
 686
 687 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 688 /* When called from bridge netfilter, skb->data must point to MAC header
 689  * before calling skb_gso_segment(). Else, original MAC header is lost
 690  * and segmented skbs will be sent to wrong destination.
 691  */
 692 static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
 693 {
 694         if (skb->nf_bridge)
 695                 __skb_push(skb, skb->network_header - skb->mac_header);
 696 }
 697
 698 static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
 699 {
 700         if (skb->nf_bridge)
 701                 __skb_pull(skb, skb->network_header - skb->mac_header);
 702 }
 703 #else
 704 #define nf_bridge_adjust_skb_data(s) do {} while (0)
 705 #define nf_bridge_adjust_segmented_data(s) do {} while (0)
 706 #endif
 707
 708 static void free_entry(struct nf_queue_entry *entry)
 709 {
 710         nf_queue_entry_release_refs(entry);
 711         kfree(entry);
 712 }
 713
 714 static int
 715 __nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
 716                            struct sk_buff *skb, struct nf_queue_entry *entry)
 717 {
 718         int ret = -ENOMEM;
 719         struct nf_queue_entry *entry_seg;
 720
 721         nf_bridge_adjust_segmented_data(skb);
 722
 723         if (skb->next == NULL) { /* last packet, no need to copy entry */
 724                 struct sk_buff *gso_skb = entry->skb;
 725                 entry->skb = skb;
 726                 ret = __nfqnl_enqueue_packet(net, queue, entry);
 727                 if (ret)
 728                         entry->skb = gso_skb;
 729                 return ret;
 730         }
 731
 732         skb->next = NULL;
 733
 734         entry_seg = nf_queue_entry_dup(entry);
 735         if (entry_seg) {
 736                 entry_seg->skb = skb;
 737                 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
 738                 if (ret)
 739                         free_entry(entry_seg);
 740         }
 741         return ret;
 742 }
 743
 744 static int
 745 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 746 {
 747         unsigned int queued;
 748         struct nfqnl_instance *queue;
 749         struct sk_buff *skb, *segs;
 750         int err = -ENOBUFS;
 751         struct net *net = entry->state.net;
 752         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
 753
 754         /* rcu_read_lock()ed by nf_hook_thresh */
 755         queue = instance_lookup(q, queuenum);
 756         if (!queue)
 757                 return -ESRCH;
 758
 759         if (queue->copy_mode == NFQNL_COPY_NONE)
 760                 return -EINVAL;
 761
 762         skb = entry->skb;
 763
 764         switch (entry->state.pf) {
 765         case NFPROTO_IPV4:
 766                 skb->protocol = htons(ETH_P_IP);
 767                 break;
 768         case NFPROTO_IPV6:
 769                 skb->protocol = htons(ETH_P_IPV6);
 770                 break;
 771         }
 772
 773         if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(skb))
 774                 return __nfqnl_enqueue_packet(net, queue, entry);
 775
 776         nf_bridge_adjust_skb_data(skb);
 777         segs = skb_gso_segment(skb, 0);
 778         /* Does not use PTR_ERR to limit the number of error codes that can be
 779          * returned by nf_queue.  For instance, callers rely on -ESRCH to
 780          * mean 'ignore this hook'.
 781          */
 782         if (IS_ERR_OR_NULL(segs))
 783                 goto out_err;
 784         queued = 0;
 785         err = 0;
 786         do {
 787                 struct sk_buff *nskb = segs->next;
 788                 if (err == 0)
 789                         err = __nfqnl_enqueue_packet_gso(net, queue,
 790                                                         segs, entry);
 791                 if (err == 0)
 792                         queued++;
 793                 else
 794                         kfree_skb(segs);
 795                 segs = nskb;
 796         } while (segs);
 797
 798         if (queued) {
 799                 if (err) /* some segments are already queued */
 800                         free_entry(entry);
 801                 kfree_skb(skb);
 802                 return 0;
 803         }
 804  out_err:
 805         nf_bridge_adjust_segmented_data(skb);
 806         return err;
 807 }
 808
 809 static int
 810 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e, int diff)
 811 {
 812         struct sk_buff *nskb;
 813
 814         if (diff < 0) {
 815                 if (pskb_trim(e->skb, data_len))
 816                         return -ENOMEM;
 817         } else if (diff > 0) {
 818                 if (data_len > 0xFFFF)
 819                         return -EINVAL;
 820                 if (diff > skb_tailroom(e->skb)) {
 821                         nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
 822                                                diff, GFP_ATOMIC);
 823                         if (!nskb) {
 824                                 printk(KERN_WARNING "nf_queue: OOM "
 825                                       "in mangle, dropping packet\n");
 826                                 return -ENOMEM;
 827                         }
 828                         kfree_skb(e->skb);
 829                         e->skb = nskb;
 830                 }
 831                 skb_put(e->skb, diff);
 832         }
 833         if (!skb_make_writable(e->skb, data_len))
 834                 return -ENOMEM;
 835         skb_copy_to_linear_data(e->skb, data, data_len);
 836         e->skb->ip_summed = CHECKSUM_NONE;
 837         return 0;
 838 }
 839
 840 static int
 841 nfqnl_set_mode(struct nfqnl_instance *queue,
 842                unsigned char mode, unsigned int range)
 843 {
 844         int status = 0;
 845
 846         spin_lock_bh(&queue->lock);
 847         switch (mode) {
 848         case NFQNL_COPY_NONE:
 849         case NFQNL_COPY_META:
 850                 queue->copy_mode = mode;
 851                 queue->copy_range = 0;
 852                 break;
 853
 854         case NFQNL_COPY_PACKET:
 855                 queue->copy_mode = mode;
 856                 if (range == 0 || range > NFQNL_MAX_COPY_RANGE)
 857                         queue->copy_range = NFQNL_MAX_COPY_RANGE;
 858                 else
 859                         queue->copy_range = range;
 860                 break;
 861
 862         default:
 863                 status = -EINVAL;
 864
 865         }
 866         spin_unlock_bh(&queue->lock);
 867
 868         return status;
 869 }
 870
 871 static int
 872 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
 873 {
 874         if (entry->state.in)
 875                 if (entry->state.in->ifindex == ifindex)
 876                         return 1;
 877         if (entry->state.out)
 878                 if (entry->state.out->ifindex == ifindex)
 879                         return 1;
 880 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 881         if (entry->skb->nf_bridge) {
 882                 int physinif, physoutif;
 883
 884                 physinif = nf_bridge_get_physinif(entry->skb);
 885                 physoutif = nf_bridge_get_physoutif(entry->skb);
 886
 887                 if (physinif == ifindex || physoutif == ifindex)
 888                         return 1;
 889         }
 890 #endif
 891         return 0;
 892 }
 893
 894 /* drop all packets with either indev or outdev == ifindex from all queue
 895  * instances */
 896 static void
 897 nfqnl_dev_drop(struct net *net, int ifindex)
 898 {
 899         int i;
 900         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
 901
 902         rcu_read_lock();
 903
 904         for (i = 0; i < INSTANCE_BUCKETS; i++) {
 905                 struct nfqnl_instance *inst;
 906                 struct hlist_head *head = &q->instance_table[i];
 907
 908                 hlist_for_each_entry_rcu(inst, head, hlist)
 909                         nfqnl_flush(inst, dev_cmp, ifindex);
 910         }
 911
 912         rcu_read_unlock();
 913 }
 914
 915 static int
 916 nfqnl_rcv_dev_event(struct notifier_block *this,
 917                     unsigned long event, void *ptr)
 918 {
 919         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
 920
 921         /* Drop any packets associated with the downed device */
 922         if (event == NETDEV_DOWN)
 923                 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
 924         return NOTIFY_DONE;
 925 }
 926
 927 static struct notifier_block nfqnl_dev_notifier = {
 928         .notifier_call  = nfqnl_rcv_dev_event,
 929 };
 930
 931 static int nf_hook_cmp(struct nf_queue_entry *entry, unsigned long entry_ptr)
 932 {
 933         return rcu_access_pointer(entry->state.hook_entries) ==
 934                 (struct nf_hook_entry *)entry_ptr;
 935 }
 936
 937 static void nfqnl_nf_hook_drop(struct net *net,
 938                                const struct nf_hook_entry *hook)
 939 {
 940         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
 941         int i;
 942
 943         rcu_read_lock();
 944         for (i = 0; i < INSTANCE_BUCKETS; i++) {
 945                 struct nfqnl_instance *inst;
 946                 struct hlist_head *head = &q->instance_table[i];
 947
 948                 hlist_for_each_entry_rcu(inst, head, hlist)
 949                         nfqnl_flush(inst, nf_hook_cmp, (unsigned long)hook);
 950         }
 951         rcu_read_unlock();
 952 }
 953
 954 static int
 955 nfqnl_rcv_nl_event(struct notifier_block *this,
 956                    unsigned long event, void *ptr)
 957 {
 958         struct netlink_notify *n = ptr;
 959         struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
 960
 961         if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
 962                 int i;
 963
 964                 /* destroy all instances for this portid */
 965                 spin_lock(&q->instances_lock);
 966                 for (i = 0; i < INSTANCE_BUCKETS; i++) {
 967                         struct hlist_node *t2;
 968                         struct nfqnl_instance *inst;
 969                         struct hlist_head *head = &q->instance_table[i];
 970
 971                         hlist_for_each_entry_safe(inst, t2, head, hlist) {
 972                                 if (n->portid == inst->peer_portid)
 973                                         __instance_destroy(inst);
 974                         }
 975                 }
 976                 spin_unlock(&q->instances_lock);
 977         }
 978         return NOTIFY_DONE;
 979 }
 980
 981 static struct notifier_block nfqnl_rtnl_notifier = {
 982         .notifier_call  = nfqnl_rcv_nl_event,
 983 };
 984
 985 static const struct nla_policy nfqa_vlan_policy[NFQA_VLAN_MAX + 1] = {
 986         [NFQA_VLAN_TCI]         = { .type = NLA_U16},
 987         [NFQA_VLAN_PROTO]       = { .type = NLA_U16},
 988 };
 989
 990 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
 991         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 992         [NFQA_MARK]             = { .type = NLA_U32 },
 993         [NFQA_PAYLOAD]          = { .type = NLA_UNSPEC },
 994         [NFQA_CT]               = { .type = NLA_UNSPEC },
 995         [NFQA_EXP]              = { .type = NLA_UNSPEC },
 996         [NFQA_VLAN]             = { .type = NLA_NESTED },
 997 };
 998
 999 static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
1000         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
1001         [NFQA_MARK]             = { .type = NLA_U32 },
1002 };
1003
1004 static struct nfqnl_instance *
1005 verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, u32 nlportid)
1006 {
1007         struct nfqnl_instance *queue;
1008
1009         queue = instance_lookup(q, queue_num);
1010         if (!queue)
1011                 return ERR_PTR(-ENODEV);
1012
1013         if (queue->peer_portid != nlportid)
1014                 return ERR_PTR(-EPERM);
1015
1016         return queue;
1017 }
1018
1019 static struct nfqnl_msg_verdict_hdr*
1020 verdicthdr_get(const struct nlattr * const nfqa[])
1021 {
1022         struct nfqnl_msg_verdict_hdr *vhdr;
1023         unsigned int verdict;
1024
1025         if (!nfqa[NFQA_VERDICT_HDR])
1026                 return NULL;
1027
1028         vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
1029         verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
1030         if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
1031                 return NULL;
1032         return vhdr;
1033 }
1034
1035 static int nfq_id_after(unsigned int id, unsigned int max)
1036 {
1037         return (int)(id - max) > 0;
1038 }
1039
1040 static int nfqnl_recv_verdict_batch(struct net *net, struct sock *ctnl,
1041                                     struct sk_buff *skb,
1042                                     const struct nlmsghdr *nlh,
1043                                     const struct nlattr * const nfqa[])
1044 {
1045         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1046         struct nf_queue_entry *entry, *tmp;
1047         unsigned int verdict, maxid;
1048         struct nfqnl_msg_verdict_hdr *vhdr;
1049         struct nfqnl_instance *queue;
1050         LIST_HEAD(batch_list);
1051         u16 queue_num = ntohs(nfmsg->res_id);
1052         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1053
1054         queue = verdict_instance_lookup(q, queue_num,
1055                                         NETLINK_CB(skb).portid);
1056         if (IS_ERR(queue))
1057                 return PTR_ERR(queue);
1058
1059         vhdr = verdicthdr_get(nfqa);
1060         if (!vhdr)
1061                 return -EINVAL;
1062
1063         verdict = ntohl(vhdr->verdict);
1064         maxid = ntohl(vhdr->id);
1065
1066         spin_lock_bh(&queue->lock);
1067
1068         list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
1069                 if (nfq_id_after(entry->id, maxid))
1070                         break;
1071                 __dequeue_entry(queue, entry);
1072                 list_add_tail(&entry->list, &batch_list);
1073         }
1074
1075         spin_unlock_bh(&queue->lock);
1076
1077         if (list_empty(&batch_list))
1078                 return -ENOENT;
1079
1080         list_for_each_entry_safe(entry, tmp, &batch_list, list) {
1081                 if (nfqa[NFQA_MARK])
1082                         entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1083                 nf_reinject(entry, verdict);
1084         }
1085         return 0;
1086 }
1087
1088 static struct nf_conn *nfqnl_ct_parse(struct nfnl_ct_hook *nfnl_ct,
1089                                       const struct nlmsghdr *nlh,
1090                                       const struct nlattr * const nfqa[],
1091                                       struct nf_queue_entry *entry,
1092                                       enum ip_conntrack_info *ctinfo)
1093 {
1094         struct nf_conn *ct;
1095
1096         ct = nfnl_ct->get_ct(entry->skb, ctinfo);
1097         if (ct == NULL)
1098                 return NULL;
1099
1100         if (nfnl_ct->parse(nfqa[NFQA_CT], ct) < 0)
1101                 return NULL;
1102
1103         if (nfqa[NFQA_EXP])
1104                 nfnl_ct->attach_expect(nfqa[NFQA_EXP], ct,
1105                                       NETLINK_CB(entry->skb).portid,
1106                                       nlmsg_report(nlh));
1107         return ct;
1108 }
1109
1110 static int nfqa_parse_bridge(struct nf_queue_entry *entry,
1111                              const struct nlattr * const nfqa[])
1112 {
1113         if (nfqa[NFQA_VLAN]) {
1114                 struct nlattr *tb[NFQA_VLAN_MAX + 1];
1115                 int err;
1116
1117                 err = nla_parse_nested(tb, NFQA_VLAN_MAX, nfqa[NFQA_VLAN],
1118                                        nfqa_vlan_policy);
1119                 if (err < 0)
1120                         return err;
1121
1122                 if (!tb[NFQA_VLAN_TCI] || !tb[NFQA_VLAN_PROTO])
1123                         return -EINVAL;
1124
1125                 entry->skb->vlan_tci = ntohs(nla_get_be16(tb[NFQA_VLAN_TCI]));
1126                 entry->skb->vlan_proto = nla_get_be16(tb[NFQA_VLAN_PROTO]);
1127         }
1128
1129         if (nfqa[NFQA_L2HDR]) {
1130                 int mac_header_len = entry->skb->network_header -
1131                         entry->skb->mac_header;
1132
1133                 if (mac_header_len != nla_len(nfqa[NFQA_L2HDR]))
1134                         return -EINVAL;
1135                 else if (mac_header_len > 0)
1136                         memcpy(skb_mac_header(entry->skb),
1137                                nla_data(nfqa[NFQA_L2HDR]),
1138                                mac_header_len);
1139         }
1140
1141         return 0;
1142 }
1143
1144 static int nfqnl_recv_verdict(struct net *net, struct sock *ctnl,
1145                               struct sk_buff *skb,
1146                               const struct nlmsghdr *nlh,
1147                               const struct nlattr * const nfqa[])
1148 {
1149         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1150         u_int16_t queue_num = ntohs(nfmsg->res_id);
1151         struct nfqnl_msg_verdict_hdr *vhdr;
1152         struct nfqnl_instance *queue;
1153         unsigned int verdict;
1154         struct nf_queue_entry *entry;
1155         enum ip_conntrack_info uninitialized_var(ctinfo);
1156         struct nfnl_ct_hook *nfnl_ct;
1157         struct nf_conn *ct = NULL;
1158         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1159         int err;
1160
1161         queue = verdict_instance_lookup(q, queue_num,
1162                                         NETLINK_CB(skb).portid);
1163         if (IS_ERR(queue))
1164                 return PTR_ERR(queue);
1165
1166         vhdr = verdicthdr_get(nfqa);
1167         if (!vhdr)
1168                 return -EINVAL;
1169
1170         verdict = ntohl(vhdr->verdict);
1171
1172         entry = find_dequeue_entry(queue, ntohl(vhdr->id));
1173         if (entry == NULL)
1174                 return -ENOENT;
1175
1176         /* rcu lock already held from nfnl->call_rcu. */
1177         nfnl_ct = rcu_dereference(nfnl_ct_hook);
1178
1179         if (nfqa[NFQA_CT]) {
1180                 if (nfnl_ct != NULL)
1181                         ct = nfqnl_ct_parse(nfnl_ct, nlh, nfqa, entry, &ctinfo);
1182         }
1183
1184         if (entry->state.pf == PF_BRIDGE) {
1185                 err = nfqa_parse_bridge(entry, nfqa);
1186                 if (err < 0)
1187                         return err;
1188         }
1189
1190         if (nfqa[NFQA_PAYLOAD]) {
1191                 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
1192                 int diff = payload_len - entry->skb->len;
1193
1194                 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
1195                                  payload_len, entry, diff) < 0)
1196                         verdict = NF_DROP;
1197
1198                 if (ct && diff)
1199                         nfnl_ct->seq_adjust(entry->skb, ct, ctinfo, diff);
1200         }
1201
1202         if (nfqa[NFQA_MARK])
1203                 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1204
1205         nf_reinject(entry, verdict);
1206         return 0;
1207 }
1208
1209 static int nfqnl_recv_unsupp(struct net *net, struct sock *ctnl,
1210                              struct sk_buff *skb, const struct nlmsghdr *nlh,
1211                              const struct nlattr * const nfqa[])
1212 {
1213         return -ENOTSUPP;
1214 }
1215
1216 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1217         [NFQA_CFG_CMD]          = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1218         [NFQA_CFG_PARAMS]       = { .len = sizeof(struct nfqnl_msg_config_params) },
1219         [NFQA_CFG_QUEUE_MAXLEN] = { .type = NLA_U32 },
1220         [NFQA_CFG_MASK]         = { .type = NLA_U32 },
1221         [NFQA_CFG_FLAGS]        = { .type = NLA_U32 },
1222 };
1223
1224 static const struct nf_queue_handler nfqh = {
1225         .outfn          = &nfqnl_enqueue_packet,
1226         .nf_hook_drop   = &nfqnl_nf_hook_drop,
1227 };
1228
1229 static int nfqnl_recv_config(struct net *net, struct sock *ctnl,
1230                              struct sk_buff *skb, const struct nlmsghdr *nlh,
1231                              const struct nlattr * const nfqa[])
1232 {
1233         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1234         u_int16_t queue_num = ntohs(nfmsg->res_id);
1235         struct nfqnl_instance *queue;
1236         struct nfqnl_msg_config_cmd *cmd = NULL;
1237         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1238         __u32 flags = 0, mask = 0;
1239         int ret = 0;
1240
1241         if (nfqa[NFQA_CFG_CMD]) {
1242                 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1243
1244                 /* Obsolete commands without queue context */
1245                 switch (cmd->command) {
1246                 case NFQNL_CFG_CMD_PF_BIND: return 0;
1247                 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1248                 }
1249         }
1250
1251         /* Check if we support these flags in first place, dependencies should
1252          * be there too not to break atomicity.
1253          */
1254         if (nfqa[NFQA_CFG_FLAGS]) {
1255                 if (!nfqa[NFQA_CFG_MASK]) {
1256                         /* A mask is needed to specify which flags are being
1257                          * changed.
1258                          */
1259                         return -EINVAL;
1260                 }
1261
1262                 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1263                 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1264
1265                 if (flags >= NFQA_CFG_F_MAX)
1266                         return -EOPNOTSUPP;
1267
1268 #if !IS_ENABLED(CONFIG_NETWORK_SECMARK)
1269                 if (flags & mask & NFQA_CFG_F_SECCTX)
1270                         return -EOPNOTSUPP;
1271 #endif
1272                 if ((flags & mask & NFQA_CFG_F_CONNTRACK) &&
1273                     !rcu_access_pointer(nfnl_ct_hook)) {
1274 #ifdef CONFIG_MODULES
1275                         nfnl_unlock(NFNL_SUBSYS_QUEUE);
1276                         request_module("ip_conntrack_netlink");
1277                         nfnl_lock(NFNL_SUBSYS_QUEUE);
1278                         if (rcu_access_pointer(nfnl_ct_hook))
1279                                 return -EAGAIN;
1280 #endif
1281                         return -EOPNOTSUPP;
1282                 }
1283         }
1284
1285         rcu_read_lock();
1286         queue = instance_lookup(q, queue_num);
1287         if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1288                 ret = -EPERM;
1289                 goto err_out_unlock;
1290         }
1291
1292         if (cmd != NULL) {
1293                 switch (cmd->command) {
1294                 case NFQNL_CFG_CMD_BIND:
1295                         if (queue) {
1296                                 ret = -EBUSY;
1297                                 goto err_out_unlock;
1298                         }
1299                         queue = instance_create(q, queue_num,
1300                                                 NETLINK_CB(skb).portid);
1301                         if (IS_ERR(queue)) {
1302                                 ret = PTR_ERR(queue);
1303                                 goto err_out_unlock;
1304                         }
1305                         break;
1306                 case NFQNL_CFG_CMD_UNBIND:
1307                         if (!queue) {
1308                                 ret = -ENODEV;
1309                                 goto err_out_unlock;
1310                         }
1311                         instance_destroy(q, queue);
1312                         goto err_out_unlock;
1313                 case NFQNL_CFG_CMD_PF_BIND:
1314                 case NFQNL_CFG_CMD_PF_UNBIND:
1315                         break;
1316                 default:
1317                         ret = -ENOTSUPP;
1318                         goto err_out_unlock;
1319                 }
1320         }
1321
1322         if (!queue) {
1323                 ret = -ENODEV;
1324                 goto err_out_unlock;
1325         }
1326
1327         if (nfqa[NFQA_CFG_PARAMS]) {
1328                 struct nfqnl_msg_config_params *params =
1329                         nla_data(nfqa[NFQA_CFG_PARAMS]);
1330
1331                 nfqnl_set_mode(queue, params->copy_mode,
1332                                 ntohl(params->copy_range));
1333         }
1334
1335         if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1336                 __be32 *queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1337
1338                 spin_lock_bh(&queue->lock);
1339                 queue->queue_maxlen = ntohl(*queue_maxlen);
1340                 spin_unlock_bh(&queue->lock);
1341         }
1342
1343         if (nfqa[NFQA_CFG_FLAGS]) {
1344                 spin_lock_bh(&queue->lock);
1345                 queue->flags &= ~mask;
1346                 queue->flags |= flags & mask;
1347                 spin_unlock_bh(&queue->lock);
1348         }
1349
1350 err_out_unlock:
1351         rcu_read_unlock();
1352         return ret;
1353 }
1354
1355 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1356         [NFQNL_MSG_PACKET]      = { .call_rcu = nfqnl_recv_unsupp,
1357                                     .attr_count = NFQA_MAX, },
1358         [NFQNL_MSG_VERDICT]     = { .call_rcu = nfqnl_recv_verdict,
1359                                     .attr_count = NFQA_MAX,
1360                                     .policy = nfqa_verdict_policy },
1361         [NFQNL_MSG_CONFIG]      = { .call = nfqnl_recv_config,
1362                                     .attr_count = NFQA_CFG_MAX,
1363                                     .policy = nfqa_cfg_policy },
1364         [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
1365                                     .attr_count = NFQA_MAX,
1366                                     .policy = nfqa_verdict_batch_policy },
1367 };
1368
1369 static const struct nfnetlink_subsystem nfqnl_subsys = {
1370         .name           = "nf_queue",
1371         .subsys_id      = NFNL_SUBSYS_QUEUE,
1372         .cb_count       = NFQNL_MSG_MAX,
1373         .cb             = nfqnl_cb,
1374 };
1375
1376 #ifdef CONFIG_PROC_FS
1377 struct iter_state {
1378         struct seq_net_private p;
1379         unsigned int bucket;
1380 };
1381
1382 static struct hlist_node *get_first(struct seq_file *seq)
1383 {
1384         struct iter_state *st = seq->private;
1385         struct net *net;
1386         struct nfnl_queue_net *q;
1387
1388         if (!st)
1389                 return NULL;
1390
1391         net = seq_file_net(seq);
1392         q = nfnl_queue_pernet(net);
1393         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1394                 if (!hlist_empty(&q->instance_table[st->bucket]))
1395                         return q->instance_table[st->bucket].first;
1396         }
1397         return NULL;
1398 }
1399
1400 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1401 {
1402         struct iter_state *st = seq->private;
1403         struct net *net = seq_file_net(seq);
1404
1405         h = h->next;
1406         while (!h) {
1407                 struct nfnl_queue_net *q;
1408
1409                 if (++st->bucket >= INSTANCE_BUCKETS)
1410                         return NULL;
1411
1412                 q = nfnl_queue_pernet(net);
1413                 h = q->instance_table[st->bucket].first;
1414         }
1415         return h;
1416 }
1417
1418 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1419 {
1420         struct hlist_node *head;
1421         head = get_first(seq);
1422
1423         if (head)
1424                 while (pos && (head = get_next(seq, head)))
1425                         pos--;
1426         return pos ? NULL : head;
1427 }
1428
1429 static void *seq_start(struct seq_file *s, loff_t *pos)
1430         __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1431 {
1432         spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1433         return get_idx(s, *pos);
1434 }
1435
1436 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1437 {
1438         (*pos)++;
1439         return get_next(s, v);
1440 }
1441
1442 static void seq_stop(struct seq_file *s, void *v)
1443         __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1444 {
1445         spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1446 }
1447
1448 static int seq_show(struct seq_file *s, void *v)
1449 {
1450         const struct nfqnl_instance *inst = v;
1451
1452         seq_printf(s, "%5u %6u %5u %1u %5u %5u %5u %8u %2d\n",
1453                    inst->queue_num,
1454                    inst->peer_portid, inst->queue_total,
1455                    inst->copy_mode, inst->copy_range,
1456                    inst->queue_dropped, inst->queue_user_dropped,
1457                    inst->id_sequence, 1);
1458         return 0;
1459 }
1460
1461 static const struct seq_operations nfqnl_seq_ops = {
1462         .start  = seq_start,
1463         .next   = seq_next,
1464         .stop   = seq_stop,
1465         .show   = seq_show,
1466 };
1467
1468 static int nfqnl_open(struct inode *inode, struct file *file)
1469 {
1470         return seq_open_net(inode, file, &nfqnl_seq_ops,
1471                         sizeof(struct iter_state));
1472 }
1473
1474 static const struct file_operations nfqnl_file_ops = {
1475         .owner   = THIS_MODULE,
1476         .open    = nfqnl_open,
1477         .read    = seq_read,
1478         .llseek  = seq_lseek,
1479         .release = seq_release_net,
1480 };
1481
1482 #endif /* PROC_FS */
1483
1484 static int __net_init nfnl_queue_net_init(struct net *net)
1485 {
1486         unsigned int i;
1487         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1488
1489         for (i = 0; i < INSTANCE_BUCKETS; i++)
1490                 INIT_HLIST_HEAD(&q->instance_table[i]);
1491
1492         spin_lock_init(&q->instances_lock);
1493
1494 #ifdef CONFIG_PROC_FS
1495         if (!proc_create("nfnetlink_queue", 0440,
1496                          net->nf.proc_netfilter, &nfqnl_file_ops))
1497                 return -ENOMEM;
1498 #endif
1499         nf_register_queue_handler(net, &nfqh);
1500         return 0;
1501 }
1502
1503 static void __net_exit nfnl_queue_net_exit(struct net *net)
1504 {
1505         nf_unregister_queue_handler(net);
1506 #ifdef CONFIG_PROC_FS
1507         remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1508 #endif
1509 }
1510
1511 static void nfnl_queue_net_exit_batch(struct list_head *net_exit_list)
1512 {
1513         synchronize_rcu();
1514 }
1515
1516 static struct pernet_operations nfnl_queue_net_ops = {
1517         .init           = nfnl_queue_net_init,
1518         .exit           = nfnl_queue_net_exit,
1519         .exit_batch     = nfnl_queue_net_exit_batch,
1520         .id             = &nfnl_queue_net_id,
1521         .size           = sizeof(struct nfnl_queue_net),
1522 };
1523
1524 static int __init nfnetlink_queue_init(void)
1525 {
1526         int status;
1527
1528         status = register_pernet_subsys(&nfnl_queue_net_ops);
1529         if (status < 0) {
1530                 pr_err("nf_queue: failed to register pernet ops\n");
1531                 goto out;
1532         }
1533
1534         netlink_register_notifier(&nfqnl_rtnl_notifier);
1535         status = nfnetlink_subsys_register(&nfqnl_subsys);
1536         if (status < 0) {
1537                 pr_err("nf_queue: failed to create netlink socket\n");
1538                 goto cleanup_netlink_notifier;
1539         }
1540
1541         status = register_netdevice_notifier(&nfqnl_dev_notifier);
1542         if (status < 0) {
1543                 pr_err("nf_queue: failed to register netdevice notifier\n");
1544                 goto cleanup_netlink_subsys;
1545         }
1546
1547         return status;
1548
1549 cleanup_netlink_subsys:
1550         nfnetlink_subsys_unregister(&nfqnl_subsys);
1551 cleanup_netlink_notifier:
1552         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1553         unregister_pernet_subsys(&nfnl_queue_net_ops);
1554 out:
1555         return status;
1556 }
1557
1558 static void __exit nfnetlink_queue_fini(void)
1559 {
1560         unregister_netdevice_notifier(&nfqnl_dev_notifier);
1561         nfnetlink_subsys_unregister(&nfqnl_subsys);
1562         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1563         unregister_pernet_subsys(&nfnl_queue_net_ops);
1564
1565         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1566 }
1567
1568 MODULE_DESCRIPTION("netfilter packet queue handler");
1569 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1570 MODULE_LICENSE("GPL");
1571 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1572
1573 module_init(nfnetlink_queue_init);
1574 module_exit(nfnetlink_queue_fini);