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
  18 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  19
  20 #include <linux/module.h>
  21 #include <linux/skbuff.h>
  22 #include <linux/init.h>
  23 #include <linux/spinlock.h>
  24 #include <linux/slab.h>
  25 #include <linux/notifier.h>
  26 #include <linux/netdevice.h>
  27 #include <linux/netfilter.h>
  28 #include <linux/proc_fs.h>
  29 #include <linux/netfilter_ipv4.h>
  30 #include <linux/netfilter_ipv6.h>
  31 #include <linux/netfilter_bridge.h>
  32 #include <linux/netfilter/nfnetlink.h>
  33 #include <linux/netfilter/nfnetlink_queue.h>
  34 #include <linux/netfilter/nf_conntrack_common.h>
  35 #include <linux/list.h>
  36 #include <net/sock.h>
  37 #include <net/tcp_states.h>
  38 #include <net/netfilter/nf_queue.h>
  39 #include <net/netns/generic.h>
  40
  41 #include <linux/atomic.h>
  42
  43 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  44 #include "../bridge/br_private.h"
  45 #endif
  46
  47 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
  48 #include <net/netfilter/nf_conntrack.h>
  49 #endif
  50
  51 #define NFQNL_QMAX_DEFAULT 1024
  52
  53 /* We're using struct nlattr which has 16bit nla_len. Note that nla_len
  54  * includes the header length. Thus, the maximum packet length that we
  55  * support is 65531 bytes. We send truncated packets if the specified length
  56  * is larger than that.  Userspace can check for presence of NFQA_CAP_LEN
  57  * attribute to detect truncation.
  58  */
  59 #define NFQNL_MAX_COPY_RANGE (0xffff - NLA_HDRLEN)
  60
  61 struct nfqnl_instance {
  62         struct hlist_node hlist;                /* global list of queues */
  63         struct rcu_head rcu;
  64
  65         u32 peer_portid;
  66         unsigned int queue_maxlen;
  67         unsigned int copy_range;
  68         unsigned int queue_dropped;
  69         unsigned int queue_user_dropped;
  70
  71
  72         u_int16_t queue_num;                    /* number of this queue */
  73         u_int8_t copy_mode;
  74         u_int32_t flags;                        /* Set using NFQA_CFG_FLAGS */
  75 /*
  76  * Following fields are dirtied for each queued packet,
  77  * keep them in same cache line if possible.
  78  */
  79         spinlock_t      lock    ____cacheline_aligned_in_smp;
  80         unsigned int    queue_total;
  81         unsigned int    id_sequence;            /* 'sequence' of pkt ids */
  82         struct list_head queue_list;            /* packets in queue */
  83 };
  84
  85 typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
  86
  87 static unsigned int nfnl_queue_net_id __read_mostly;
  88
  89 #define INSTANCE_BUCKETS        16
  90 struct nfnl_queue_net {
  91         spinlock_t instances_lock;
  92         struct hlist_head instance_table[INSTANCE_BUCKETS];
  93 };
  94
  95 static struct nfnl_queue_net *nfnl_queue_pernet(struct net *net)
  96 {
  97         return net_generic(net, nfnl_queue_net_id);
  98 }
  99
 100 static inline u_int8_t instance_hashfn(u_int16_t queue_num)
 101 {
 102         return ((queue_num >> 8) ^ queue_num) % INSTANCE_BUCKETS;
 103 }
 104
 105 static struct nfqnl_instance *
 106 instance_lookup(struct nfnl_queue_net *q, u_int16_t queue_num)
 107 {
 108         struct hlist_head *head;
 109         struct nfqnl_instance *inst;
 110
 111         head = &q->instance_table[instance_hashfn(queue_num)];
 112         hlist_for_each_entry_rcu(inst, head, hlist) {
 113                 if (inst->queue_num == queue_num)
 114                         return inst;
 115         }
 116         return NULL;
 117 }
 118
 119 static struct nfqnl_instance *
 120 instance_create(struct nfnl_queue_net *q, u_int16_t queue_num, u32 portid)
 121 {
 122         struct nfqnl_instance *inst;
 123         unsigned int h;
 124         int err;
 125
 126         spin_lock(&q->instances_lock);
 127         if (instance_lookup(q, queue_num)) {
 128                 err = -EEXIST;
 129                 goto out_unlock;
 130         }
 131
 132         inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
 133         if (!inst) {
 134                 err = -ENOMEM;
 135                 goto out_unlock;
 136         }
 137
 138         inst->queue_num = queue_num;
 139         inst->peer_portid = portid;
 140         inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
 141         inst->copy_range = NFQNL_MAX_COPY_RANGE;
 142         inst->copy_mode = NFQNL_COPY_NONE;
 143         spin_lock_init(&inst->lock);
 144         INIT_LIST_HEAD(&inst->queue_list);
 145
 146         if (!try_module_get(THIS_MODULE)) {
 147                 err = -EAGAIN;
 148                 goto out_free;
 149         }
 150
 151         h = instance_hashfn(queue_num);
 152         hlist_add_head_rcu(&inst->hlist, &q->instance_table[h]);
 153
 154         spin_unlock(&q->instances_lock);
 155
 156         return inst;
 157
 158 out_free:
 159         kfree(inst);
 160 out_unlock:
 161         spin_unlock(&q->instances_lock);
 162         return ERR_PTR(err);
 163 }
 164
 165 static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
 166                         unsigned long data);
 167
 168 static void
 169 instance_destroy_rcu(struct rcu_head *head)
 170 {
 171         struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
 172                                                    rcu);
 173
 174         nfqnl_flush(inst, NULL, 0);
 175         kfree(inst);
 176         module_put(THIS_MODULE);
 177 }
 178
 179 static void
 180 __instance_destroy(struct nfqnl_instance *inst)
 181 {
 182         hlist_del_rcu(&inst->hlist);
 183         call_rcu(&inst->rcu, instance_destroy_rcu);
 184 }
 185
 186 static void
 187 instance_destroy(struct nfnl_queue_net *q, struct nfqnl_instance *inst)
 188 {
 189         spin_lock(&q->instances_lock);
 190         __instance_destroy(inst);
 191         spin_unlock(&q->instances_lock);
 192 }
 193
 194 static inline void
 195 __enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
 196 {
 197        list_add_tail(&entry->list, &queue->queue_list);
 198        queue->queue_total++;
 199 }
 200
 201 static void
 202 __dequeue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
 203 {
 204         list_del(&entry->list);
 205         queue->queue_total--;
 206 }
 207
 208 static struct nf_queue_entry *
 209 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
 210 {
 211         struct nf_queue_entry *entry = NULL, *i;
 212
 213         spin_lock_bh(&queue->lock);
 214
 215         list_for_each_entry(i, &queue->queue_list, list) {
 216                 if (i->id == id) {
 217                         entry = i;
 218                         break;
 219                 }
 220         }
 221
 222         if (entry)
 223                 __dequeue_entry(queue, entry);
 224
 225         spin_unlock_bh(&queue->lock);
 226
 227         return entry;
 228 }
 229
 230 static void
 231 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
 232 {
 233         struct nf_queue_entry *entry, *next;
 234
 235         spin_lock_bh(&queue->lock);
 236         list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
 237                 if (!cmpfn || cmpfn(entry, data)) {
 238                         list_del(&entry->list);
 239                         queue->queue_total--;
 240                         nf_reinject(entry, NF_DROP);
 241                 }
 242         }
 243         spin_unlock_bh(&queue->lock);
 244 }
 245
 246 static int
 247 nfqnl_put_packet_info(struct sk_buff *nlskb, struct sk_buff *packet,
 248                       bool csum_verify)
 249 {
 250         __u32 flags = 0;
 251
 252         if (packet->ip_summed == CHECKSUM_PARTIAL)
 253                 flags = NFQA_SKB_CSUMNOTREADY;
 254         else if (csum_verify)
 255                 flags = NFQA_SKB_CSUM_NOTVERIFIED;
 256
 257         if (skb_is_gso(packet))
 258                 flags |= NFQA_SKB_GSO;
 259
 260         return flags ? nla_put_be32(nlskb, NFQA_SKB_INFO, htonl(flags)) : 0;
 261 }
 262
 263 static int nfqnl_put_sk_uidgid(struct sk_buff *skb, struct sock *sk)
 264 {
 265         const struct cred *cred;
 266
 267         if (!sk_fullsock(sk))
 268                 return 0;
 269
 270         read_lock_bh(&sk->sk_callback_lock);
 271         if (sk->sk_socket && sk->sk_socket->file) {
 272                 cred = sk->sk_socket->file->f_cred;
 273                 if (nla_put_be32(skb, NFQA_UID,
 274                     htonl(from_kuid_munged(&init_user_ns, cred->fsuid))))
 275                         goto nla_put_failure;
 276                 if (nla_put_be32(skb, NFQA_GID,
 277                     htonl(from_kgid_munged(&init_user_ns, cred->fsgid))))
 278                         goto nla_put_failure;
 279         }
 280         read_unlock_bh(&sk->sk_callback_lock);
 281         return 0;
 282
 283 nla_put_failure:
 284         read_unlock_bh(&sk->sk_callback_lock);
 285         return -1;
 286 }
 287
 288 static u32 nfqnl_get_sk_secctx(struct sk_buff *skb, char **secdata)
 289 {
 290         u32 seclen = 0;
 291 #if IS_ENABLED(CONFIG_NETWORK_SECMARK)
 292         if (!skb || !sk_fullsock(skb->sk))
 293                 return 0;
 294
 295         read_lock_bh(&skb->sk->sk_callback_lock);
 296
 297         if (skb->secmark)
 298                 security_secid_to_secctx(skb->secmark, secdata, &seclen);
 299
 300         read_unlock_bh(&skb->sk->sk_callback_lock);
 301 #endif
 302         return seclen;
 303 }
 304
 305 static u32 nfqnl_get_bridge_size(struct nf_queue_entry *entry)
 306 {
 307         struct sk_buff *entskb = entry->skb;
 308         u32 nlalen = 0;
 309
 310         if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb))
 311                 return 0;
 312
 313         if (skb_vlan_tag_present(entskb))
 314                 nlalen += nla_total_size(nla_total_size(sizeof(__be16)) +
 315                                          nla_total_size(sizeof(__be16)));
 316
 317         if (entskb->network_header > entskb->mac_header)
 318                 nlalen += nla_total_size((entskb->network_header -
 319                                           entskb->mac_header));
 320
 321         return nlalen;
 322 }
 323
 324 static int nfqnl_put_bridge(struct nf_queue_entry *entry, struct sk_buff *skb)
 325 {
 326         struct sk_buff *entskb = entry->skb;
 327
 328         if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb))
 329                 return 0;
 330
 331         if (skb_vlan_tag_present(entskb)) {
 332                 struct nlattr *nest;
 333
 334                 nest = nla_nest_start(skb, NFQA_VLAN | NLA_F_NESTED);
 335                 if (!nest)
 336                         goto nla_put_failure;
 337
 338                 if (nla_put_be16(skb, NFQA_VLAN_TCI, htons(entskb->vlan_tci)) ||
 339                     nla_put_be16(skb, NFQA_VLAN_PROTO, entskb->vlan_proto))
 340                         goto nla_put_failure;
 341
 342                 nla_nest_end(skb, nest);
 343         }
 344
 345         if (entskb->mac_header < entskb->network_header) {
 346                 int len = (int)(entskb->network_header - entskb->mac_header);
 347
 348                 if (nla_put(skb, NFQA_L2HDR, len, skb_mac_header(entskb)))
 349                         goto nla_put_failure;
 350         }
 351
 352         return 0;
 353
 354 nla_put_failure:
 355         return -1;
 356 }
 357
 358 static struct sk_buff *
 359 nfqnl_build_packet_message(struct net *net, struct nfqnl_instance *queue,
 360                            struct nf_queue_entry *entry,
 361                            __be32 **packet_id_ptr)
 362 {
 363         size_t size;
 364         size_t data_len = 0, cap_len = 0;
 365         unsigned int hlen = 0;
 366         struct sk_buff *skb;
 367         struct nlattr *nla;
 368         struct nfqnl_msg_packet_hdr *pmsg;
 369         struct nlmsghdr *nlh;
 370         struct nfgenmsg *nfmsg;
 371         struct sk_buff *entskb = entry->skb;
 372         struct net_device *indev;
 373         struct net_device *outdev;
 374         struct nf_conn *ct = NULL;
 375         enum ip_conntrack_info uninitialized_var(ctinfo);
 376         struct nfnl_ct_hook *nfnl_ct;
 377         bool csum_verify;
 378         char *secdata = NULL;
 379         u32 seclen = 0;
 380
 381         size =    nlmsg_total_size(sizeof(struct nfgenmsg))
 382                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
 383                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 384                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 385 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 386                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 387                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 388 #endif
 389                 + nla_total_size(sizeof(u_int32_t))     /* mark */
 390                 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
 391                 + nla_total_size(sizeof(u_int32_t))     /* skbinfo */
 392                 + nla_total_size(sizeof(u_int32_t));    /* cap_len */
 393
 394         if (entskb->tstamp)
 395                 size += nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
 396
 397         size += nfqnl_get_bridge_size(entry);
 398
 399         if (entry->state.hook <= NF_INET_FORWARD ||
 400            (entry->state.hook == NF_INET_POST_ROUTING && entskb->sk == NULL))
 401                 csum_verify = !skb_csum_unnecessary(entskb);
 402         else
 403                 csum_verify = false;
 404
 405         outdev = entry->state.out;
 406
 407         switch ((enum nfqnl_config_mode)READ_ONCE(queue->copy_mode)) {
 408         case NFQNL_COPY_META:
 409         case NFQNL_COPY_NONE:
 410                 break;
 411
 412         case NFQNL_COPY_PACKET:
 413                 if (!(queue->flags & NFQA_CFG_F_GSO) &&
 414                     entskb->ip_summed == CHECKSUM_PARTIAL &&
 415                     skb_checksum_help(entskb))
 416                         return NULL;
 417
 418                 data_len = READ_ONCE(queue->copy_range);
 419                 if (data_len > entskb->len)
 420                         data_len = entskb->len;
 421
 422                 hlen = skb_zerocopy_headlen(entskb);
 423                 hlen = min_t(unsigned int, hlen, data_len);
 424                 size += sizeof(struct nlattr) + hlen;
 425                 cap_len = entskb->len;
 426                 break;
 427         }
 428
 429         nfnl_ct = rcu_dereference(nfnl_ct_hook);
 430
 431         if (queue->flags & NFQA_CFG_F_CONNTRACK) {
 432                 if (nfnl_ct != NULL) {
 433                         ct = nfnl_ct->get_ct(entskb, &ctinfo);
 434                         if (ct != NULL)
 435                                 size += nfnl_ct->build_size(ct);
 436                 }
 437         }
 438
 439         if (queue->flags & NFQA_CFG_F_UID_GID) {
 440                 size +=  (nla_total_size(sizeof(u_int32_t))     /* uid */
 441                         + nla_total_size(sizeof(u_int32_t)));   /* gid */
 442         }
 443
 444         if ((queue->flags & NFQA_CFG_F_SECCTX) && entskb->sk) {
 445                 seclen = nfqnl_get_sk_secctx(entskb, &secdata);
 446                 if (seclen)
 447                         size += nla_total_size(seclen);
 448         }
 449
 450         skb = alloc_skb(size, GFP_ATOMIC);
 451         if (!skb) {
 452                 skb_tx_error(entskb);
 453                 goto nlmsg_failure;
 454         }
 455
 456         nlh = nlmsg_put(skb, 0, 0,
 457                         nfnl_msg_type(NFNL_SUBSYS_QUEUE, NFQNL_MSG_PACKET),
 458                         sizeof(struct nfgenmsg), 0);
 459         if (!nlh) {
 460                 skb_tx_error(entskb);
 461                 kfree_skb(skb);
 462                 goto nlmsg_failure;
 463         }
 464         nfmsg = nlmsg_data(nlh);
 465         nfmsg->nfgen_family = entry->state.pf;
 466         nfmsg->version = NFNETLINK_V0;
 467         nfmsg->res_id = htons(queue->queue_num);
 468
 469         nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
 470         pmsg = nla_data(nla);
 471         pmsg->hw_protocol       = entskb->protocol;
 472         pmsg->hook              = entry->state.hook;
 473         *packet_id_ptr          = &pmsg->packet_id;
 474
 475         indev = entry->state.in;
 476         if (indev) {
 477 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 478                 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
 479                         goto nla_put_failure;
 480 #else
 481                 if (entry->state.pf == PF_BRIDGE) {
 482                         /* Case 1: indev is physical input device, we need to
 483                          * look for bridge group (when called from
 484                          * netfilter_bridge) */
 485                         if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
 486                                          htonl(indev->ifindex)) ||
 487                         /* this is the bridge group "brX" */
 488                         /* rcu_read_lock()ed by __nf_queue */
 489                             nla_put_be32(skb, NFQA_IFINDEX_INDEV,
 490                                          htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
 491                                 goto nla_put_failure;
 492                 } else {
 493                         int physinif;
 494
 495                         /* Case 2: indev is bridge group, we need to look for
 496                          * physical device (when called from ipv4) */
 497                         if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
 498                                          htonl(indev->ifindex)))
 499                                 goto nla_put_failure;
 500
 501                         physinif = nf_bridge_get_physinif(entskb);
 502                         if (physinif &&
 503                             nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
 504                                          htonl(physinif)))
 505                                 goto nla_put_failure;
 506                 }
 507 #endif
 508         }
 509
 510         if (outdev) {
 511 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 512                 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
 513                         goto nla_put_failure;
 514 #else
 515                 if (entry->state.pf == PF_BRIDGE) {
 516                         /* Case 1: outdev is physical output device, we need to
 517                          * look for bridge group (when called from
 518                          * netfilter_bridge) */
 519                         if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 520                                          htonl(outdev->ifindex)) ||
 521                         /* this is the bridge group "brX" */
 522                         /* rcu_read_lock()ed by __nf_queue */
 523                             nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
 524                                          htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
 525                                 goto nla_put_failure;
 526                 } else {
 527                         int physoutif;
 528
 529                         /* Case 2: outdev is bridge group, we need to look for
 530                          * physical output device (when called from ipv4) */
 531                         if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
 532                                          htonl(outdev->ifindex)))
 533                                 goto nla_put_failure;
 534
 535                         physoutif = nf_bridge_get_physoutif(entskb);
 536                         if (physoutif &&
 537                             nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
 538                                          htonl(physoutif)))
 539                                 goto nla_put_failure;
 540                 }
 541 #endif
 542         }
 543
 544         if (entskb->mark &&
 545             nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
 546                 goto nla_put_failure;
 547
 548         if (indev && entskb->dev &&
 549             entskb->mac_header != entskb->network_header) {
 550                 struct nfqnl_msg_packet_hw phw;
 551                 int len;
 552
 553                 memset(&phw, 0, sizeof(phw));
 554                 len = dev_parse_header(entskb, phw.hw_addr);
 555                 if (len) {
 556                         phw.hw_addrlen = htons(len);
 557                         if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
 558                                 goto nla_put_failure;
 559                 }
 560         }
 561
 562         if (nfqnl_put_bridge(entry, skb) < 0)
 563                 goto nla_put_failure;
 564
 565         if (entskb->tstamp) {
 566                 struct nfqnl_msg_packet_timestamp ts;
 567                 struct timespec64 kts = ktime_to_timespec64(entskb->tstamp);
 568
 569                 ts.sec = cpu_to_be64(kts.tv_sec);
 570                 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
 571
 572                 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
 573                         goto nla_put_failure;
 574         }
 575
 576         if ((queue->flags & NFQA_CFG_F_UID_GID) && entskb->sk &&
 577             nfqnl_put_sk_uidgid(skb, entskb->sk) < 0)
 578                 goto nla_put_failure;
 579
 580         if (seclen && nla_put(skb, NFQA_SECCTX, seclen, secdata))
 581                 goto nla_put_failure;
 582
 583         if (ct && nfnl_ct->build(skb, ct, ctinfo, NFQA_CT, NFQA_CT_INFO) < 0)
 584                 goto nla_put_failure;
 585
 586         if (cap_len > data_len &&
 587             nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
 588                 goto nla_put_failure;
 589
 590         if (nfqnl_put_packet_info(skb, entskb, csum_verify))
 591                 goto nla_put_failure;
 592
 593         if (data_len) {
 594                 struct nlattr *nla;
 595
 596                 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
 597                         goto nla_put_failure;
 598
 599                 nla = skb_put(skb, sizeof(*nla));
 600                 nla->nla_type = NFQA_PAYLOAD;
 601                 nla->nla_len = nla_attr_size(data_len);
 602
 603                 if (skb_zerocopy(skb, entskb, data_len, hlen))
 604                         goto nla_put_failure;
 605         }
 606
 607         nlh->nlmsg_len = skb->len;
 608         if (seclen)
 609                 security_release_secctx(secdata, seclen);
 610         return skb;
 611
 612 nla_put_failure:
 613         skb_tx_error(entskb);
 614         kfree_skb(skb);
 615         net_err_ratelimited("nf_queue: error creating packet message\n");
 616 nlmsg_failure:
 617         if (seclen)
 618                 security_release_secctx(secdata, seclen);
 619         return NULL;
 620 }
 621
 622 static bool nf_ct_drop_unconfirmed(const struct nf_queue_entry *entry)
 623 {
 624 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
 625         static const unsigned long flags = IPS_CONFIRMED | IPS_DYING;
 626         const struct nf_conn *ct = (void *)skb_nfct(entry->skb);
 627
 628         if (ct && ((ct->status & flags) == IPS_DYING))
 629                 return true;
 630 #endif
 631         return false;
 632 }
 633
 634 static int
 635 __nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
 636                         struct nf_queue_entry *entry)
 637 {
 638         struct sk_buff *nskb;
 639         int err = -ENOBUFS;
 640         __be32 *packet_id_ptr;
 641         int failopen = 0;
 642
 643         nskb = nfqnl_build_packet_message(net, queue, entry, &packet_id_ptr);
 644         if (nskb == NULL) {
 645                 err = -ENOMEM;
 646                 goto err_out;
 647         }
 648         spin_lock_bh(&queue->lock);
 649
 650         if (nf_ct_drop_unconfirmed(entry))
 651                 goto err_out_free_nskb;
 652
 653         if (queue->queue_total >= queue->queue_maxlen) {
 654                 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
 655                         failopen = 1;
 656                         err = 0;
 657                 } else {
 658                         queue->queue_dropped++;
 659                         net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
 660                                              queue->queue_total);
 661                 }
 662                 goto err_out_free_nskb;
 663         }
 664         entry->id = ++queue->id_sequence;
 665         *packet_id_ptr = htonl(entry->id);
 666
 667         /* nfnetlink_unicast will either free the nskb or add it to a socket */
 668         err = nfnetlink_unicast(nskb, net, queue->peer_portid, MSG_DONTWAIT);
 669         if (err < 0) {
 670                 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
 671                         failopen = 1;
 672                         err = 0;
 673                 } else {
 674                         queue->queue_user_dropped++;
 675                 }
 676                 goto err_out_unlock;
 677         }
 678
 679         __enqueue_entry(queue, entry);
 680
 681         spin_unlock_bh(&queue->lock);
 682         return 0;
 683
 684 err_out_free_nskb:
 685         kfree_skb(nskb);
 686 err_out_unlock:
 687         spin_unlock_bh(&queue->lock);
 688         if (failopen)
 689                 nf_reinject(entry, NF_ACCEPT);
 690 err_out:
 691         return err;
 692 }
 693
 694 static struct nf_queue_entry *
 695 nf_queue_entry_dup(struct nf_queue_entry *e)
 696 {
 697         struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
 698         if (entry)
 699                 nf_queue_entry_get_refs(entry);
 700         return entry;
 701 }
 702
 703 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 704 /* When called from bridge netfilter, skb->data must point to MAC header
 705  * before calling skb_gso_segment(). Else, original MAC header is lost
 706  * and segmented skbs will be sent to wrong destination.
 707  */
 708 static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
 709 {
 710         if (skb->nf_bridge)
 711                 __skb_push(skb, skb->network_header - skb->mac_header);
 712 }
 713
 714 static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
 715 {
 716         if (skb->nf_bridge)
 717                 __skb_pull(skb, skb->network_header - skb->mac_header);
 718 }
 719 #else
 720 #define nf_bridge_adjust_skb_data(s) do {} while (0)
 721 #define nf_bridge_adjust_segmented_data(s) do {} while (0)
 722 #endif
 723
 724 static void free_entry(struct nf_queue_entry *entry)
 725 {
 726         nf_queue_entry_release_refs(entry);
 727         kfree(entry);
 728 }
 729
 730 static int
 731 __nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
 732                            struct sk_buff *skb, struct nf_queue_entry *entry)
 733 {
 734         int ret = -ENOMEM;
 735         struct nf_queue_entry *entry_seg;
 736
 737         nf_bridge_adjust_segmented_data(skb);
 738
 739         if (skb->next == NULL) { /* last packet, no need to copy entry */
 740                 struct sk_buff *gso_skb = entry->skb;
 741                 entry->skb = skb;
 742                 ret = __nfqnl_enqueue_packet(net, queue, entry);
 743                 if (ret)
 744                         entry->skb = gso_skb;
 745                 return ret;
 746         }
 747
 748         skb->next = NULL;
 749
 750         entry_seg = nf_queue_entry_dup(entry);
 751         if (entry_seg) {
 752                 entry_seg->skb = skb;
 753                 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
 754                 if (ret)
 755                         free_entry(entry_seg);
 756         }
 757         return ret;
 758 }
 759
 760 static int
 761 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 762 {
 763         unsigned int queued;
 764         struct nfqnl_instance *queue;
 765         struct sk_buff *skb, *segs;
 766         int err = -ENOBUFS;
 767         struct net *net = entry->state.net;
 768         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
 769
 770         /* rcu_read_lock()ed by nf_hook_thresh */
 771         queue = instance_lookup(q, queuenum);
 772         if (!queue)
 773                 return -ESRCH;
 774
 775         if (queue->copy_mode == NFQNL_COPY_NONE)
 776                 return -EINVAL;
 777
 778         skb = entry->skb;
 779
 780         switch (entry->state.pf) {
 781         case NFPROTO_IPV4:
 782                 skb->protocol = htons(ETH_P_IP);
 783                 break;
 784         case NFPROTO_IPV6:
 785                 skb->protocol = htons(ETH_P_IPV6);
 786                 break;
 787         }
 788
 789         if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(skb))
 790                 return __nfqnl_enqueue_packet(net, queue, entry);
 791
 792         nf_bridge_adjust_skb_data(skb);
 793         segs = skb_gso_segment(skb, 0);
 794         /* Does not use PTR_ERR to limit the number of error codes that can be
 795          * returned by nf_queue.  For instance, callers rely on -ESRCH to
 796          * mean 'ignore this hook'.
 797          */
 798         if (IS_ERR_OR_NULL(segs))
 799                 goto out_err;
 800         queued = 0;
 801         err = 0;
 802         do {
 803                 struct sk_buff *nskb = segs->next;
 804                 if (err == 0)
 805                         err = __nfqnl_enqueue_packet_gso(net, queue,
 806                                                         segs, entry);
 807                 if (err == 0)
 808                         queued++;
 809                 else
 810                         kfree_skb(segs);
 811                 segs = nskb;
 812         } while (segs);
 813
 814         if (queued) {
 815                 if (err) /* some segments are already queued */
 816                         free_entry(entry);
 817                 kfree_skb(skb);
 818                 return 0;
 819         }
 820  out_err:
 821         nf_bridge_adjust_segmented_data(skb);
 822         return err;
 823 }
 824
 825 static int
 826 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e, int diff)
 827 {
 828         struct sk_buff *nskb;
 829
 830         if (diff < 0) {
 831                 if (pskb_trim(e->skb, data_len))
 832                         return -ENOMEM;
 833         } else if (diff > 0) {
 834                 if (data_len > 0xFFFF)
 835                         return -EINVAL;
 836                 if (diff > skb_tailroom(e->skb)) {
 837                         nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
 838                                                diff, GFP_ATOMIC);
 839                         if (!nskb)
 840                                 return -ENOMEM;
 841                         kfree_skb(e->skb);
 842                         e->skb = nskb;
 843                 }
 844                 skb_put(e->skb, diff);
 845         }
 846         if (!skb_make_writable(e->skb, data_len))
 847                 return -ENOMEM;
 848         skb_copy_to_linear_data(e->skb, data, data_len);
 849         e->skb->ip_summed = CHECKSUM_NONE;
 850         return 0;
 851 }
 852
 853 static int
 854 nfqnl_set_mode(struct nfqnl_instance *queue,
 855                unsigned char mode, unsigned int range)
 856 {
 857         int status = 0;
 858
 859         spin_lock_bh(&queue->lock);
 860         switch (mode) {
 861         case NFQNL_COPY_NONE:
 862         case NFQNL_COPY_META:
 863                 queue->copy_mode = mode;
 864                 queue->copy_range = 0;
 865                 break;
 866
 867         case NFQNL_COPY_PACKET:
 868                 queue->copy_mode = mode;
 869                 if (range == 0 || range > NFQNL_MAX_COPY_RANGE)
 870                         queue->copy_range = NFQNL_MAX_COPY_RANGE;
 871                 else
 872                         queue->copy_range = range;
 873                 break;
 874
 875         default:
 876                 status = -EINVAL;
 877
 878         }
 879         spin_unlock_bh(&queue->lock);
 880
 881         return status;
 882 }
 883
 884 static int
 885 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
 886 {
 887         if (entry->state.in)
 888                 if (entry->state.in->ifindex == ifindex)
 889                         return 1;
 890         if (entry->state.out)
 891                 if (entry->state.out->ifindex == ifindex)
 892                         return 1;
 893 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 894         if (entry->skb->nf_bridge) {
 895                 int physinif, physoutif;
 896
 897                 physinif = nf_bridge_get_physinif(entry->skb);
 898                 physoutif = nf_bridge_get_physoutif(entry->skb);
 899
 900                 if (physinif == ifindex || physoutif == ifindex)
 901                         return 1;
 902         }
 903 #endif
 904         return 0;
 905 }
 906
 907 /* drop all packets with either indev or outdev == ifindex from all queue
 908  * instances */
 909 static void
 910 nfqnl_dev_drop(struct net *net, int ifindex)
 911 {
 912         int i;
 913         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
 914
 915         rcu_read_lock();
 916
 917         for (i = 0; i < INSTANCE_BUCKETS; i++) {
 918                 struct nfqnl_instance *inst;
 919                 struct hlist_head *head = &q->instance_table[i];
 920
 921                 hlist_for_each_entry_rcu(inst, head, hlist)
 922                         nfqnl_flush(inst, dev_cmp, ifindex);
 923         }
 924
 925         rcu_read_unlock();
 926 }
 927
 928 static int
 929 nfqnl_rcv_dev_event(struct notifier_block *this,
 930                     unsigned long event, void *ptr)
 931 {
 932         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
 933
 934         /* Drop any packets associated with the downed device */
 935         if (event == NETDEV_DOWN)
 936                 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
 937         return NOTIFY_DONE;
 938 }
 939
 940 static struct notifier_block nfqnl_dev_notifier = {
 941         .notifier_call  = nfqnl_rcv_dev_event,
 942 };
 943
 944 static void nfqnl_nf_hook_drop(struct net *net)
 945 {
 946         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
 947         int i;
 948
 949         for (i = 0; i < INSTANCE_BUCKETS; i++) {
 950                 struct nfqnl_instance *inst;
 951                 struct hlist_head *head = &q->instance_table[i];
 952
 953                 hlist_for_each_entry_rcu(inst, head, hlist)
 954                         nfqnl_flush(inst, NULL, 0);
 955         }
 956 }
 957
 958 static int
 959 nfqnl_rcv_nl_event(struct notifier_block *this,
 960                    unsigned long event, void *ptr)
 961 {
 962         struct netlink_notify *n = ptr;
 963         struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
 964
 965         if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
 966                 int i;
 967
 968                 /* destroy all instances for this portid */
 969                 spin_lock(&q->instances_lock);
 970                 for (i = 0; i < INSTANCE_BUCKETS; i++) {
 971                         struct hlist_node *t2;
 972                         struct nfqnl_instance *inst;
 973                         struct hlist_head *head = &q->instance_table[i];
 974
 975                         hlist_for_each_entry_safe(inst, t2, head, hlist) {
 976                                 if (n->portid == inst->peer_portid)
 977                                         __instance_destroy(inst);
 978                         }
 979                 }
 980                 spin_unlock(&q->instances_lock);
 981         }
 982         return NOTIFY_DONE;
 983 }
 984
 985 static struct notifier_block nfqnl_rtnl_notifier = {
 986         .notifier_call  = nfqnl_rcv_nl_event,
 987 };
 988
 989 static const struct nla_policy nfqa_vlan_policy[NFQA_VLAN_MAX + 1] = {
 990         [NFQA_VLAN_TCI]         = { .type = NLA_U16},
 991         [NFQA_VLAN_PROTO]       = { .type = NLA_U16},
 992 };
 993
 994 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
 995         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
 996         [NFQA_MARK]             = { .type = NLA_U32 },
 997         [NFQA_PAYLOAD]          = { .type = NLA_UNSPEC },
 998         [NFQA_CT]               = { .type = NLA_UNSPEC },
 999         [NFQA_EXP]              = { .type = NLA_UNSPEC },
1000         [NFQA_VLAN]             = { .type = NLA_NESTED },
1001 };
1002
1003 static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
1004         [NFQA_VERDICT_HDR]      = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
1005         [NFQA_MARK]             = { .type = NLA_U32 },
1006 };
1007
1008 static struct nfqnl_instance *
1009 verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, u32 nlportid)
1010 {
1011         struct nfqnl_instance *queue;
1012
1013         queue = instance_lookup(q, queue_num);
1014         if (!queue)
1015                 return ERR_PTR(-ENODEV);
1016
1017         if (queue->peer_portid != nlportid)
1018                 return ERR_PTR(-EPERM);
1019
1020         return queue;
1021 }
1022
1023 static struct nfqnl_msg_verdict_hdr*
1024 verdicthdr_get(const struct nlattr * const nfqa[])
1025 {
1026         struct nfqnl_msg_verdict_hdr *vhdr;
1027         unsigned int verdict;
1028
1029         if (!nfqa[NFQA_VERDICT_HDR])
1030                 return NULL;
1031
1032         vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
1033         verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
1034         if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
1035                 return NULL;
1036         return vhdr;
1037 }
1038
1039 static int nfq_id_after(unsigned int id, unsigned int max)
1040 {
1041         return (int)(id - max) > 0;
1042 }
1043
1044 static int nfqnl_recv_verdict_batch(struct net *net, struct sock *ctnl,
1045                                     struct sk_buff *skb,
1046                                     const struct nlmsghdr *nlh,
1047                                     const struct nlattr * const nfqa[],
1048                                     struct netlink_ext_ack *extack)
1049 {
1050         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1051         struct nf_queue_entry *entry, *tmp;
1052         unsigned int verdict, maxid;
1053         struct nfqnl_msg_verdict_hdr *vhdr;
1054         struct nfqnl_instance *queue;
1055         LIST_HEAD(batch_list);
1056         u16 queue_num = ntohs(nfmsg->res_id);
1057         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1058
1059         queue = verdict_instance_lookup(q, queue_num,
1060                                         NETLINK_CB(skb).portid);
1061         if (IS_ERR(queue))
1062                 return PTR_ERR(queue);
1063
1064         vhdr = verdicthdr_get(nfqa);
1065         if (!vhdr)
1066                 return -EINVAL;
1067
1068         verdict = ntohl(vhdr->verdict);
1069         maxid = ntohl(vhdr->id);
1070
1071         spin_lock_bh(&queue->lock);
1072
1073         list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
1074                 if (nfq_id_after(entry->id, maxid))
1075                         break;
1076                 __dequeue_entry(queue, entry);
1077                 list_add_tail(&entry->list, &batch_list);
1078         }
1079
1080         spin_unlock_bh(&queue->lock);
1081
1082         if (list_empty(&batch_list))
1083                 return -ENOENT;
1084
1085         list_for_each_entry_safe(entry, tmp, &batch_list, list) {
1086                 if (nfqa[NFQA_MARK])
1087                         entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1088                 nf_reinject(entry, verdict);
1089         }
1090         return 0;
1091 }
1092
1093 static struct nf_conn *nfqnl_ct_parse(struct nfnl_ct_hook *nfnl_ct,
1094                                       const struct nlmsghdr *nlh,
1095                                       const struct nlattr * const nfqa[],
1096                                       struct nf_queue_entry *entry,
1097                                       enum ip_conntrack_info *ctinfo)
1098 {
1099         struct nf_conn *ct;
1100
1101         ct = nfnl_ct->get_ct(entry->skb, ctinfo);
1102         if (ct == NULL)
1103                 return NULL;
1104
1105         if (nfnl_ct->parse(nfqa[NFQA_CT], ct) < 0)
1106                 return NULL;
1107
1108         if (nfqa[NFQA_EXP])
1109                 nfnl_ct->attach_expect(nfqa[NFQA_EXP], ct,
1110                                       NETLINK_CB(entry->skb).portid,
1111                                       nlmsg_report(nlh));
1112         return ct;
1113 }
1114
1115 static int nfqa_parse_bridge(struct nf_queue_entry *entry,
1116                              const struct nlattr * const nfqa[])
1117 {
1118         if (nfqa[NFQA_VLAN]) {
1119                 struct nlattr *tb[NFQA_VLAN_MAX + 1];
1120                 int err;
1121
1122                 err = nla_parse_nested(tb, NFQA_VLAN_MAX, nfqa[NFQA_VLAN],
1123                                        nfqa_vlan_policy, NULL);
1124                 if (err < 0)
1125                         return err;
1126
1127                 if (!tb[NFQA_VLAN_TCI] || !tb[NFQA_VLAN_PROTO])
1128                         return -EINVAL;
1129
1130                 entry->skb->vlan_tci = ntohs(nla_get_be16(tb[NFQA_VLAN_TCI]));
1131                 entry->skb->vlan_proto = nla_get_be16(tb[NFQA_VLAN_PROTO]);
1132         }
1133
1134         if (nfqa[NFQA_L2HDR]) {
1135                 int mac_header_len = entry->skb->network_header -
1136                         entry->skb->mac_header;
1137
1138                 if (mac_header_len != nla_len(nfqa[NFQA_L2HDR]))
1139                         return -EINVAL;
1140                 else if (mac_header_len > 0)
1141                         memcpy(skb_mac_header(entry->skb),
1142                                nla_data(nfqa[NFQA_L2HDR]),
1143                                mac_header_len);
1144         }
1145
1146         return 0;
1147 }
1148
1149 static int nfqnl_recv_verdict(struct net *net, struct sock *ctnl,
1150                               struct sk_buff *skb,
1151                               const struct nlmsghdr *nlh,
1152                               const struct nlattr * const nfqa[],
1153                               struct netlink_ext_ack *extack)
1154 {
1155         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1156         u_int16_t queue_num = ntohs(nfmsg->res_id);
1157         struct nfqnl_msg_verdict_hdr *vhdr;
1158         struct nfqnl_instance *queue;
1159         unsigned int verdict;
1160         struct nf_queue_entry *entry;
1161         enum ip_conntrack_info uninitialized_var(ctinfo);
1162         struct nfnl_ct_hook *nfnl_ct;
1163         struct nf_conn *ct = NULL;
1164         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1165         int err;
1166
1167         queue = verdict_instance_lookup(q, queue_num,
1168                                         NETLINK_CB(skb).portid);
1169         if (IS_ERR(queue))
1170                 return PTR_ERR(queue);
1171
1172         vhdr = verdicthdr_get(nfqa);
1173         if (!vhdr)
1174                 return -EINVAL;
1175
1176         verdict = ntohl(vhdr->verdict);
1177
1178         entry = find_dequeue_entry(queue, ntohl(vhdr->id));
1179         if (entry == NULL)
1180                 return -ENOENT;
1181
1182         /* rcu lock already held from nfnl->call_rcu. */
1183         nfnl_ct = rcu_dereference(nfnl_ct_hook);
1184
1185         if (nfqa[NFQA_CT]) {
1186                 if (nfnl_ct != NULL)
1187                         ct = nfqnl_ct_parse(nfnl_ct, nlh, nfqa, entry, &ctinfo);
1188         }
1189
1190         if (entry->state.pf == PF_BRIDGE) {
1191                 err = nfqa_parse_bridge(entry, nfqa);
1192                 if (err < 0)
1193                         return err;
1194         }
1195
1196         if (nfqa[NFQA_PAYLOAD]) {
1197                 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
1198                 int diff = payload_len - entry->skb->len;
1199
1200                 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
1201                                  payload_len, entry, diff) < 0)
1202                         verdict = NF_DROP;
1203
1204                 if (ct && diff)
1205                         nfnl_ct->seq_adjust(entry->skb, ct, ctinfo, diff);
1206         }
1207
1208         if (nfqa[NFQA_MARK])
1209                 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1210
1211         nf_reinject(entry, verdict);
1212         return 0;
1213 }
1214
1215 static int nfqnl_recv_unsupp(struct net *net, struct sock *ctnl,
1216                              struct sk_buff *skb, const struct nlmsghdr *nlh,
1217                              const struct nlattr * const nfqa[],
1218                              struct netlink_ext_ack *extack)
1219 {
1220         return -ENOTSUPP;
1221 }
1222
1223 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1224         [NFQA_CFG_CMD]          = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1225         [NFQA_CFG_PARAMS]       = { .len = sizeof(struct nfqnl_msg_config_params) },
1226 };
1227
1228 static const struct nf_queue_handler nfqh = {
1229         .outfn          = nfqnl_enqueue_packet,
1230         .nf_hook_drop   = nfqnl_nf_hook_drop,
1231 };
1232
1233 static int nfqnl_recv_config(struct net *net, struct sock *ctnl,
1234                              struct sk_buff *skb, const struct nlmsghdr *nlh,
1235                              const struct nlattr * const nfqa[],
1236                              struct netlink_ext_ack *extack)
1237 {
1238         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1239         u_int16_t queue_num = ntohs(nfmsg->res_id);
1240         struct nfqnl_instance *queue;
1241         struct nfqnl_msg_config_cmd *cmd = NULL;
1242         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1243         __u32 flags = 0, mask = 0;
1244         int ret = 0;
1245
1246         if (nfqa[NFQA_CFG_CMD]) {
1247                 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1248
1249                 /* Obsolete commands without queue context */
1250                 switch (cmd->command) {
1251                 case NFQNL_CFG_CMD_PF_BIND: return 0;
1252                 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1253                 }
1254         }
1255
1256         /* Check if we support these flags in first place, dependencies should
1257          * be there too not to break atomicity.
1258          */
1259         if (nfqa[NFQA_CFG_FLAGS]) {
1260                 if (!nfqa[NFQA_CFG_MASK]) {
1261                         /* A mask is needed to specify which flags are being
1262                          * changed.
1263                          */
1264                         return -EINVAL;
1265                 }
1266
1267                 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1268                 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1269
1270                 if (flags >= NFQA_CFG_F_MAX)
1271                         return -EOPNOTSUPP;
1272
1273 #if !IS_ENABLED(CONFIG_NETWORK_SECMARK)
1274                 if (flags & mask & NFQA_CFG_F_SECCTX)
1275                         return -EOPNOTSUPP;
1276 #endif
1277                 if ((flags & mask & NFQA_CFG_F_CONNTRACK) &&
1278                     !rcu_access_pointer(nfnl_ct_hook)) {
1279 #ifdef CONFIG_MODULES
1280                         nfnl_unlock(NFNL_SUBSYS_QUEUE);
1281                         request_module("ip_conntrack_netlink");
1282                         nfnl_lock(NFNL_SUBSYS_QUEUE);
1283                         if (rcu_access_pointer(nfnl_ct_hook))
1284                                 return -EAGAIN;
1285 #endif
1286                         return -EOPNOTSUPP;
1287                 }
1288         }
1289
1290         rcu_read_lock();
1291         queue = instance_lookup(q, queue_num);
1292         if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1293                 ret = -EPERM;
1294                 goto err_out_unlock;
1295         }
1296
1297         if (cmd != NULL) {
1298                 switch (cmd->command) {
1299                 case NFQNL_CFG_CMD_BIND:
1300                         if (queue) {
1301                                 ret = -EBUSY;
1302                                 goto err_out_unlock;
1303                         }
1304                         queue = instance_create(q, queue_num,
1305                                                 NETLINK_CB(skb).portid);
1306                         if (IS_ERR(queue)) {
1307                                 ret = PTR_ERR(queue);
1308                                 goto err_out_unlock;
1309                         }
1310                         break;
1311                 case NFQNL_CFG_CMD_UNBIND:
1312                         if (!queue) {
1313                                 ret = -ENODEV;
1314                                 goto err_out_unlock;
1315                         }
1316                         instance_destroy(q, queue);
1317                         goto err_out_unlock;
1318                 case NFQNL_CFG_CMD_PF_BIND:
1319                 case NFQNL_CFG_CMD_PF_UNBIND:
1320                         break;
1321                 default:
1322                         ret = -ENOTSUPP;
1323                         goto err_out_unlock;
1324                 }
1325         }
1326
1327         if (!queue) {
1328                 ret = -ENODEV;
1329                 goto err_out_unlock;
1330         }
1331
1332         if (nfqa[NFQA_CFG_PARAMS]) {
1333                 struct nfqnl_msg_config_params *params =
1334                         nla_data(nfqa[NFQA_CFG_PARAMS]);
1335
1336                 nfqnl_set_mode(queue, params->copy_mode,
1337                                 ntohl(params->copy_range));
1338         }
1339
1340         if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1341                 __be32 *queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1342
1343                 spin_lock_bh(&queue->lock);
1344                 queue->queue_maxlen = ntohl(*queue_maxlen);
1345                 spin_unlock_bh(&queue->lock);
1346         }
1347
1348         if (nfqa[NFQA_CFG_FLAGS]) {
1349                 spin_lock_bh(&queue->lock);
1350                 queue->flags &= ~mask;
1351                 queue->flags |= flags & mask;
1352                 spin_unlock_bh(&queue->lock);
1353         }
1354
1355 err_out_unlock:
1356         rcu_read_unlock();
1357         return ret;
1358 }
1359
1360 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1361         [NFQNL_MSG_PACKET]      = { .call_rcu = nfqnl_recv_unsupp,
1362                                     .attr_count = NFQA_MAX, },
1363         [NFQNL_MSG_VERDICT]     = { .call_rcu = nfqnl_recv_verdict,
1364                                     .attr_count = NFQA_MAX,
1365                                     .policy = nfqa_verdict_policy },
1366         [NFQNL_MSG_CONFIG]      = { .call = nfqnl_recv_config,
1367                                     .attr_count = NFQA_CFG_MAX,
1368                                     .policy = nfqa_cfg_policy },
1369         [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
1370                                     .attr_count = NFQA_MAX,
1371                                     .policy = nfqa_verdict_batch_policy },
1372 };
1373
1374 static const struct nfnetlink_subsystem nfqnl_subsys = {
1375         .name           = "nf_queue",
1376         .subsys_id      = NFNL_SUBSYS_QUEUE,
1377         .cb_count       = NFQNL_MSG_MAX,
1378         .cb             = nfqnl_cb,
1379 };
1380
1381 #ifdef CONFIG_PROC_FS
1382 struct iter_state {
1383         struct seq_net_private p;
1384         unsigned int bucket;
1385 };
1386
1387 static struct hlist_node *get_first(struct seq_file *seq)
1388 {
1389         struct iter_state *st = seq->private;
1390         struct net *net;
1391         struct nfnl_queue_net *q;
1392
1393         if (!st)
1394                 return NULL;
1395
1396         net = seq_file_net(seq);
1397         q = nfnl_queue_pernet(net);
1398         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1399                 if (!hlist_empty(&q->instance_table[st->bucket]))
1400                         return q->instance_table[st->bucket].first;
1401         }
1402         return NULL;
1403 }
1404
1405 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1406 {
1407         struct iter_state *st = seq->private;
1408         struct net *net = seq_file_net(seq);
1409
1410         h = h->next;
1411         while (!h) {
1412                 struct nfnl_queue_net *q;
1413
1414                 if (++st->bucket >= INSTANCE_BUCKETS)
1415                         return NULL;
1416
1417                 q = nfnl_queue_pernet(net);
1418                 h = q->instance_table[st->bucket].first;
1419         }
1420         return h;
1421 }
1422
1423 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1424 {
1425         struct hlist_node *head;
1426         head = get_first(seq);
1427
1428         if (head)
1429                 while (pos && (head = get_next(seq, head)))
1430                         pos--;
1431         return pos ? NULL : head;
1432 }
1433
1434 static void *seq_start(struct seq_file *s, loff_t *pos)
1435         __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1436 {
1437         spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1438         return get_idx(s, *pos);
1439 }
1440
1441 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1442 {
1443         (*pos)++;
1444         return get_next(s, v);
1445 }
1446
1447 static void seq_stop(struct seq_file *s, void *v)
1448         __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1449 {
1450         spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1451 }
1452
1453 static int seq_show(struct seq_file *s, void *v)
1454 {
1455         const struct nfqnl_instance *inst = v;
1456
1457         seq_printf(s, "%5u %6u %5u %1u %5u %5u %5u %8u %2d\n",
1458                    inst->queue_num,
1459                    inst->peer_portid, inst->queue_total,
1460                    inst->copy_mode, inst->copy_range,
1461                    inst->queue_dropped, inst->queue_user_dropped,
1462                    inst->id_sequence, 1);
1463         return 0;
1464 }
1465
1466 static const struct seq_operations nfqnl_seq_ops = {
1467         .start  = seq_start,
1468         .next   = seq_next,
1469         .stop   = seq_stop,
1470         .show   = seq_show,
1471 };
1472
1473 static int nfqnl_open(struct inode *inode, struct file *file)
1474 {
1475         return seq_open_net(inode, file, &nfqnl_seq_ops,
1476                         sizeof(struct iter_state));
1477 }
1478
1479 static const struct file_operations nfqnl_file_ops = {
1480         .open    = nfqnl_open,
1481         .read    = seq_read,
1482         .llseek  = seq_lseek,
1483         .release = seq_release_net,
1484 };
1485
1486 #endif /* PROC_FS */
1487
1488 static int __net_init nfnl_queue_net_init(struct net *net)
1489 {
1490         unsigned int i;
1491         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1492
1493         for (i = 0; i < INSTANCE_BUCKETS; i++)
1494                 INIT_HLIST_HEAD(&q->instance_table[i]);
1495
1496         spin_lock_init(&q->instances_lock);
1497
1498 #ifdef CONFIG_PROC_FS
1499         if (!proc_create("nfnetlink_queue", 0440,
1500                          net->nf.proc_netfilter, &nfqnl_file_ops))
1501                 return -ENOMEM;
1502 #endif
1503         nf_register_queue_handler(net, &nfqh);
1504         return 0;
1505 }
1506
1507 static void __net_exit nfnl_queue_net_exit(struct net *net)
1508 {
1509         struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1510         unsigned int i;
1511
1512         nf_unregister_queue_handler(net);
1513 #ifdef CONFIG_PROC_FS
1514         remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1515 #endif
1516         for (i = 0; i < INSTANCE_BUCKETS; i++)
1517                 WARN_ON_ONCE(!hlist_empty(&q->instance_table[i]));
1518 }
1519
1520 static void nfnl_queue_net_exit_batch(struct list_head *net_exit_list)
1521 {
1522         synchronize_rcu();
1523 }
1524
1525 static struct pernet_operations nfnl_queue_net_ops = {
1526         .init           = nfnl_queue_net_init,
1527         .exit           = nfnl_queue_net_exit,
1528         .exit_batch     = nfnl_queue_net_exit_batch,
1529         .id             = &nfnl_queue_net_id,
1530         .size           = sizeof(struct nfnl_queue_net),
1531 };
1532
1533 static int __init nfnetlink_queue_init(void)
1534 {
1535         int status;
1536
1537         status = register_pernet_subsys(&nfnl_queue_net_ops);
1538         if (status < 0) {
1539                 pr_err("failed to register pernet ops\n");
1540                 goto out;
1541         }
1542
1543         netlink_register_notifier(&nfqnl_rtnl_notifier);
1544         status = nfnetlink_subsys_register(&nfqnl_subsys);
1545         if (status < 0) {
1546                 pr_err("failed to create netlink socket\n");
1547                 goto cleanup_netlink_notifier;
1548         }
1549
1550         status = register_netdevice_notifier(&nfqnl_dev_notifier);
1551         if (status < 0) {
1552                 pr_err("failed to register netdevice notifier\n");
1553                 goto cleanup_netlink_subsys;
1554         }
1555
1556         return status;
1557
1558 cleanup_netlink_subsys:
1559         nfnetlink_subsys_unregister(&nfqnl_subsys);
1560 cleanup_netlink_notifier:
1561         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1562         unregister_pernet_subsys(&nfnl_queue_net_ops);
1563 out:
1564         return status;
1565 }
1566
1567 static void __exit nfnetlink_queue_fini(void)
1568 {
1569         unregister_netdevice_notifier(&nfqnl_dev_notifier);
1570         nfnetlink_subsys_unregister(&nfqnl_subsys);
1571         netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1572         unregister_pernet_subsys(&nfnl_queue_net_ops);
1573
1574         rcu_barrier(); /* Wait for completion of call_rcu()'s */
1575 }
1576
1577 MODULE_DESCRIPTION("netfilter packet queue handler");
1578 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1579 MODULE_LICENSE("GPL");
1580 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1581
1582 module_init(nfnetlink_queue_init);
1583 module_exit(nfnetlink_queue_fini);