net/netfilter/nfnetlink_log.c

   1 /*
   2  * This is a module which is used for logging packets to userspace via
   3  * nfetlink.
   4  *
   5  * (C) 2005 by Harald Welte <laforge@netfilter.org>
   6  * (C) 2006-2012 Patrick McHardy <kaber@trash.net>
   7  *
   8  * Based on the old ipv4-only ipt_ULOG.c:
   9  * (C) 2000-2004 by Harald Welte <laforge@netfilter.org>
  10  *
  11  * This program is free software; you can redistribute it and/or modify
  12  * it under the terms of the GNU General Public License version 2 as
  13  * published by the Free Software Foundation.
  14  */
  15
  16 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  17
  18 #include <linux/module.h>
  19 #include <linux/skbuff.h>
  20 #include <linux/if_arp.h>
  21 #include <linux/init.h>
  22 #include <linux/ip.h>
  23 #include <linux/ipv6.h>
  24 #include <linux/netdevice.h>
  25 #include <linux/netfilter.h>
  26 #include <linux/netfilter_bridge.h>
  27 #include <net/netlink.h>
  28 #include <linux/netfilter/nfnetlink.h>
  29 #include <linux/netfilter/nfnetlink_log.h>
  30 #include <linux/netfilter/nf_conntrack_common.h>
  31 #include <linux/spinlock.h>
  32 #include <linux/sysctl.h>
  33 #include <linux/proc_fs.h>
  34 #include <linux/security.h>
  35 #include <linux/list.h>
  36 #include <linux/slab.h>
  37 #include <net/sock.h>
  38 #include <net/netfilter/nf_log.h>
  39 #include <net/netns/generic.h>
  40 #include <net/netfilter/nfnetlink_log.h>
  41
  42 #include <linux/atomic.h>
  43
  44 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  45 #include "../bridge/br_private.h"
  46 #endif
  47
  48 #define NFULNL_NLBUFSIZ_DEFAULT NLMSG_GOODSIZE
  49 #define NFULNL_TIMEOUT_DEFAULT  100     /* every second */
  50 #define NFULNL_QTHRESH_DEFAULT  100     /* 100 packets */
  51 /* max packet size is limited by 16-bit struct nfattr nfa_len field */
  52 #define NFULNL_COPY_RANGE_MAX   (0xFFFF - NLA_HDRLEN)
  53
  54 #define PRINTR(x, args...)      do { if (net_ratelimit()) \
  55                                      printk(x, ## args); } while (0);
  56
  57 struct nfulnl_instance {
  58         struct hlist_node hlist;        /* global list of instances */
  59         spinlock_t lock;
  60         atomic_t use;                   /* use count */
  61
  62         unsigned int qlen;              /* number of nlmsgs in skb */
  63         struct sk_buff *skb;            /* pre-allocatd skb */
  64         struct timer_list timer;
  65         struct net *net;
  66         struct user_namespace *peer_user_ns;    /* User namespace of the peer process */
  67         u32 peer_portid;                /* PORTID of the peer process */
  68
  69         /* configurable parameters */
  70         unsigned int flushtimeout;      /* timeout until queue flush */
  71         unsigned int nlbufsiz;          /* netlink buffer allocation size */
  72         unsigned int qthreshold;        /* threshold of the queue */
  73         u_int32_t copy_range;
  74         u_int32_t seq;                  /* instance-local sequential counter */
  75         u_int16_t group_num;            /* number of this queue */
  76         u_int16_t flags;
  77         u_int8_t copy_mode;
  78         struct rcu_head rcu;
  79 };
  80
  81 #define INSTANCE_BUCKETS        16
  82
  83 static int nfnl_log_net_id __read_mostly;
  84
  85 struct nfnl_log_net {
  86         spinlock_t instances_lock;
  87         struct hlist_head instance_table[INSTANCE_BUCKETS];
  88         atomic_t global_seq;
  89 };
  90
  91 static struct nfnl_log_net *nfnl_log_pernet(struct net *net)
  92 {
  93         return net_generic(net, nfnl_log_net_id);
  94 }
  95
  96 static inline u_int8_t instance_hashfn(u_int16_t group_num)
  97 {
  98         return ((group_num & 0xff) % INSTANCE_BUCKETS);
  99 }
 100
 101 static struct nfulnl_instance *
 102 __instance_lookup(struct nfnl_log_net *log, u_int16_t group_num)
 103 {
 104         struct hlist_head *head;
 105         struct nfulnl_instance *inst;
 106
 107         head = &log->instance_table[instance_hashfn(group_num)];
 108         hlist_for_each_entry_rcu(inst, head, hlist) {
 109                 if (inst->group_num == group_num)
 110                         return inst;
 111         }
 112         return NULL;
 113 }
 114
 115 static inline void
 116 instance_get(struct nfulnl_instance *inst)
 117 {
 118         atomic_inc(&inst->use);
 119 }
 120
 121 static struct nfulnl_instance *
 122 instance_lookup_get(struct nfnl_log_net *log, u_int16_t group_num)
 123 {
 124         struct nfulnl_instance *inst;
 125
 126         rcu_read_lock_bh();
 127         inst = __instance_lookup(log, group_num);
 128         if (inst && !atomic_inc_not_zero(&inst->use))
 129                 inst = NULL;
 130         rcu_read_unlock_bh();
 131
 132         return inst;
 133 }
 134
 135 static void nfulnl_instance_free_rcu(struct rcu_head *head)
 136 {
 137         struct nfulnl_instance *inst =
 138                 container_of(head, struct nfulnl_instance, rcu);
 139
 140         put_net(inst->net);
 141         kfree(inst);
 142         module_put(THIS_MODULE);
 143 }
 144
 145 static void
 146 instance_put(struct nfulnl_instance *inst)
 147 {
 148         if (inst && atomic_dec_and_test(&inst->use))
 149                 call_rcu_bh(&inst->rcu, nfulnl_instance_free_rcu);
 150 }
 151
 152 static void nfulnl_timer(unsigned long data);
 153
 154 static struct nfulnl_instance *
 155 instance_create(struct net *net, u_int16_t group_num,
 156                 u32 portid, struct user_namespace *user_ns)
 157 {
 158         struct nfulnl_instance *inst;
 159         struct nfnl_log_net *log = nfnl_log_pernet(net);
 160         int err;
 161
 162         spin_lock_bh(&log->instances_lock);
 163         if (__instance_lookup(log, group_num)) {
 164                 err = -EEXIST;
 165                 goto out_unlock;
 166         }
 167
 168         inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
 169         if (!inst) {
 170                 err = -ENOMEM;
 171                 goto out_unlock;
 172         }
 173
 174         if (!try_module_get(THIS_MODULE)) {
 175                 kfree(inst);
 176                 err = -EAGAIN;
 177                 goto out_unlock;
 178         }
 179
 180         INIT_HLIST_NODE(&inst->hlist);
 181         spin_lock_init(&inst->lock);
 182         /* needs to be two, since we _put() after creation */
 183         atomic_set(&inst->use, 2);
 184
 185         setup_timer(&inst->timer, nfulnl_timer, (unsigned long)inst);
 186
 187         inst->net = get_net(net);
 188         inst->peer_user_ns = user_ns;
 189         inst->peer_portid = portid;
 190         inst->group_num = group_num;
 191
 192         inst->qthreshold        = NFULNL_QTHRESH_DEFAULT;
 193         inst->flushtimeout      = NFULNL_TIMEOUT_DEFAULT;
 194         inst->nlbufsiz          = NFULNL_NLBUFSIZ_DEFAULT;
 195         inst->copy_mode         = NFULNL_COPY_PACKET;
 196         inst->copy_range        = NFULNL_COPY_RANGE_MAX;
 197
 198         hlist_add_head_rcu(&inst->hlist,
 199                        &log->instance_table[instance_hashfn(group_num)]);
 200
 201
 202         spin_unlock_bh(&log->instances_lock);
 203
 204         return inst;
 205
 206 out_unlock:
 207         spin_unlock_bh(&log->instances_lock);
 208         return ERR_PTR(err);
 209 }
 210
 211 static void __nfulnl_flush(struct nfulnl_instance *inst);
 212
 213 /* called with BH disabled */
 214 static void
 215 __instance_destroy(struct nfulnl_instance *inst)
 216 {
 217         /* first pull it out of the global list */
 218         hlist_del_rcu(&inst->hlist);
 219
 220         /* then flush all pending packets from skb */
 221
 222         spin_lock(&inst->lock);
 223
 224         /* lockless readers wont be able to use us */
 225         inst->copy_mode = NFULNL_COPY_DISABLED;
 226
 227         if (inst->skb)
 228                 __nfulnl_flush(inst);
 229         spin_unlock(&inst->lock);
 230
 231         /* and finally put the refcount */
 232         instance_put(inst);
 233 }
 234
 235 static inline void
 236 instance_destroy(struct nfnl_log_net *log,
 237                  struct nfulnl_instance *inst)
 238 {
 239         spin_lock_bh(&log->instances_lock);
 240         __instance_destroy(inst);
 241         spin_unlock_bh(&log->instances_lock);
 242 }
 243
 244 static int
 245 nfulnl_set_mode(struct nfulnl_instance *inst, u_int8_t mode,
 246                   unsigned int range)
 247 {
 248         int status = 0;
 249
 250         spin_lock_bh(&inst->lock);
 251
 252         switch (mode) {
 253         case NFULNL_COPY_NONE:
 254         case NFULNL_COPY_META:
 255                 inst->copy_mode = mode;
 256                 inst->copy_range = 0;
 257                 break;
 258
 259         case NFULNL_COPY_PACKET:
 260                 inst->copy_mode = mode;
 261                 if (range == 0)
 262                         range = NFULNL_COPY_RANGE_MAX;
 263                 inst->copy_range = min_t(unsigned int,
 264                                          range, NFULNL_COPY_RANGE_MAX);
 265                 break;
 266
 267         default:
 268                 status = -EINVAL;
 269                 break;
 270         }
 271
 272         spin_unlock_bh(&inst->lock);
 273
 274         return status;
 275 }
 276
 277 static int
 278 nfulnl_set_nlbufsiz(struct nfulnl_instance *inst, u_int32_t nlbufsiz)
 279 {
 280         int status;
 281
 282         spin_lock_bh(&inst->lock);
 283         if (nlbufsiz < NFULNL_NLBUFSIZ_DEFAULT)
 284                 status = -ERANGE;
 285         else if (nlbufsiz > 131072)
 286                 status = -ERANGE;
 287         else {
 288                 inst->nlbufsiz = nlbufsiz;
 289                 status = 0;
 290         }
 291         spin_unlock_bh(&inst->lock);
 292
 293         return status;
 294 }
 295
 296 static void
 297 nfulnl_set_timeout(struct nfulnl_instance *inst, u_int32_t timeout)
 298 {
 299         spin_lock_bh(&inst->lock);
 300         inst->flushtimeout = timeout;
 301         spin_unlock_bh(&inst->lock);
 302 }
 303
 304 static void
 305 nfulnl_set_qthresh(struct nfulnl_instance *inst, u_int32_t qthresh)
 306 {
 307         spin_lock_bh(&inst->lock);
 308         inst->qthreshold = qthresh;
 309         spin_unlock_bh(&inst->lock);
 310 }
 311
 312 static int
 313 nfulnl_set_flags(struct nfulnl_instance *inst, u_int16_t flags)
 314 {
 315         spin_lock_bh(&inst->lock);
 316         inst->flags = flags;
 317         spin_unlock_bh(&inst->lock);
 318
 319         return 0;
 320 }
 321
 322 static struct sk_buff *
 323 nfulnl_alloc_skb(struct net *net, u32 peer_portid, unsigned int inst_size,
 324                  unsigned int pkt_size)
 325 {
 326         struct sk_buff *skb;
 327         unsigned int n;
 328
 329         /* alloc skb which should be big enough for a whole multipart
 330          * message.  WARNING: has to be <= 128k due to slab restrictions */
 331
 332         n = max(inst_size, pkt_size);
 333         skb = nfnetlink_alloc_skb(net, n, peer_portid, GFP_ATOMIC);
 334         if (!skb) {
 335                 if (n > pkt_size) {
 336                         /* try to allocate only as much as we need for current
 337                          * packet */
 338
 339                         skb = nfnetlink_alloc_skb(net, pkt_size,
 340                                                   peer_portid, GFP_ATOMIC);
 341                 }
 342         }
 343
 344         return skb;
 345 }
 346
 347 static void
 348 __nfulnl_send(struct nfulnl_instance *inst)
 349 {
 350         if (inst->qlen > 1) {
 351                 struct nlmsghdr *nlh = nlmsg_put(inst->skb, 0, 0,
 352                                                  NLMSG_DONE,
 353                                                  sizeof(struct nfgenmsg),
 354                                                  0);
 355                 if (WARN_ONCE(!nlh, "bad nlskb size: %u, tailroom %d\n",
 356                               inst->skb->len, skb_tailroom(inst->skb))) {
 357                         kfree_skb(inst->skb);
 358                         goto out;
 359                 }
 360         }
 361         nfnetlink_unicast(inst->skb, inst->net, inst->peer_portid,
 362                           MSG_DONTWAIT);
 363 out:
 364         inst->qlen = 0;
 365         inst->skb = NULL;
 366 }
 367
 368 static void
 369 __nfulnl_flush(struct nfulnl_instance *inst)
 370 {
 371         /* timer holds a reference */
 372         if (del_timer(&inst->timer))
 373                 instance_put(inst);
 374         if (inst->skb)
 375                 __nfulnl_send(inst);
 376 }
 377
 378 static void
 379 nfulnl_timer(unsigned long data)
 380 {
 381         struct nfulnl_instance *inst = (struct nfulnl_instance *)data;
 382
 383         spin_lock_bh(&inst->lock);
 384         if (inst->skb)
 385                 __nfulnl_send(inst);
 386         spin_unlock_bh(&inst->lock);
 387         instance_put(inst);
 388 }
 389
 390 /* This is an inline function, we don't really care about a long
 391  * list of arguments */
 392 static inline int
 393 __build_packet_message(struct nfnl_log_net *log,
 394                         struct nfulnl_instance *inst,
 395                         const struct sk_buff *skb,
 396                         unsigned int data_len,
 397                         u_int8_t pf,
 398                         unsigned int hooknum,
 399                         const struct net_device *indev,
 400                         const struct net_device *outdev,
 401                         const char *prefix, unsigned int plen,
 402                         const struct nfnl_ct_hook *nfnl_ct,
 403                         struct nf_conn *ct, enum ip_conntrack_info ctinfo)
 404 {
 405         struct nfulnl_msg_packet_hdr pmsg;
 406         struct nlmsghdr *nlh;
 407         struct nfgenmsg *nfmsg;
 408         sk_buff_data_t old_tail = inst->skb->tail;
 409         struct sock *sk;
 410         const unsigned char *hwhdrp;
 411
 412         nlh = nlmsg_put(inst->skb, 0, 0,
 413                         NFNL_SUBSYS_ULOG << 8 | NFULNL_MSG_PACKET,
 414                         sizeof(struct nfgenmsg), 0);
 415         if (!nlh)
 416                 return -1;
 417         nfmsg = nlmsg_data(nlh);
 418         nfmsg->nfgen_family = pf;
 419         nfmsg->version = NFNETLINK_V0;
 420         nfmsg->res_id = htons(inst->group_num);
 421
 422         memset(&pmsg, 0, sizeof(pmsg));
 423         pmsg.hw_protocol        = skb->protocol;
 424         pmsg.hook               = hooknum;
 425
 426         if (nla_put(inst->skb, NFULA_PACKET_HDR, sizeof(pmsg), &pmsg))
 427                 goto nla_put_failure;
 428
 429         if (prefix &&
 430             nla_put(inst->skb, NFULA_PREFIX, plen, prefix))
 431                 goto nla_put_failure;
 432
 433         if (indev) {
 434 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 435                 if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 436                                  htonl(indev->ifindex)))
 437                         goto nla_put_failure;
 438 #else
 439                 if (pf == PF_BRIDGE) {
 440                         /* Case 1: outdev is physical input device, we need to
 441                          * look for bridge group (when called from
 442                          * netfilter_bridge) */
 443                         if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
 444                                          htonl(indev->ifindex)) ||
 445                         /* this is the bridge group "brX" */
 446                         /* rcu_read_lock()ed by nf_hook_slow or nf_log_packet */
 447                             nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 448                                          htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
 449                                 goto nla_put_failure;
 450                 } else {
 451                         struct net_device *physindev;
 452
 453                         /* Case 2: indev is bridge group, we need to look for
 454                          * physical device (when called from ipv4) */
 455                         if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 456                                          htonl(indev->ifindex)))
 457                                 goto nla_put_failure;
 458
 459                         physindev = nf_bridge_get_physindev(skb);
 460                         if (physindev &&
 461                             nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
 462                                          htonl(physindev->ifindex)))
 463                                 goto nla_put_failure;
 464                 }
 465 #endif
 466         }
 467
 468         if (outdev) {
 469 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 470                 if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 471                                  htonl(outdev->ifindex)))
 472                         goto nla_put_failure;
 473 #else
 474                 if (pf == PF_BRIDGE) {
 475                         /* Case 1: outdev is physical output device, we need to
 476                          * look for bridge group (when called from
 477                          * netfilter_bridge) */
 478                         if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
 479                                          htonl(outdev->ifindex)) ||
 480                         /* this is the bridge group "brX" */
 481                         /* rcu_read_lock()ed by nf_hook_slow or nf_log_packet */
 482                             nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 483                                          htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
 484                                 goto nla_put_failure;
 485                 } else {
 486                         struct net_device *physoutdev;
 487
 488                         /* Case 2: indev is a bridge group, we need to look
 489                          * for physical device (when called from ipv4) */
 490                         if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 491                                          htonl(outdev->ifindex)))
 492                                 goto nla_put_failure;
 493
 494                         physoutdev = nf_bridge_get_physoutdev(skb);
 495                         if (physoutdev &&
 496                             nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
 497                                          htonl(physoutdev->ifindex)))
 498                                 goto nla_put_failure;
 499                 }
 500 #endif
 501         }
 502
 503         if (skb->mark &&
 504             nla_put_be32(inst->skb, NFULA_MARK, htonl(skb->mark)))
 505                 goto nla_put_failure;
 506
 507         if (indev && skb->dev &&
 508             skb->mac_header != skb->network_header) {
 509                 struct nfulnl_msg_packet_hw phw;
 510                 int len;
 511
 512                 memset(&phw, 0, sizeof(phw));
 513                 len = dev_parse_header(skb, phw.hw_addr);
 514                 if (len > 0) {
 515                         phw.hw_addrlen = htons(len);
 516                         if (nla_put(inst->skb, NFULA_HWADDR, sizeof(phw), &phw))
 517                                 goto nla_put_failure;
 518                 }
 519         }
 520
 521         if (indev && skb_mac_header_was_set(skb)) {
 522                 if (nla_put_be16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type)) ||
 523                     nla_put_be16(inst->skb, NFULA_HWLEN,
 524                                  htons(skb->dev->hard_header_len)))
 525                         goto nla_put_failure;
 526
 527                 hwhdrp = skb_mac_header(skb);
 528
 529                 if (skb->dev->type == ARPHRD_SIT)
 530                         hwhdrp -= ETH_HLEN;
 531
 532                 if (hwhdrp >= skb->head &&
 533                     nla_put(inst->skb, NFULA_HWHEADER,
 534                             skb->dev->hard_header_len, hwhdrp))
 535                         goto nla_put_failure;
 536         }
 537
 538         if (skb->tstamp.tv64) {
 539                 struct nfulnl_msg_packet_timestamp ts;
 540                 struct timespec64 kts = ktime_to_timespec64(skb->tstamp);
 541                 ts.sec = cpu_to_be64(kts.tv_sec);
 542                 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
 543
 544                 if (nla_put(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts))
 545                         goto nla_put_failure;
 546         }
 547
 548         /* UID */
 549         sk = skb->sk;
 550         if (sk && sk_fullsock(sk)) {
 551                 read_lock_bh(&sk->sk_callback_lock);
 552                 if (sk->sk_socket && sk->sk_socket->file) {
 553                         struct file *file = sk->sk_socket->file;
 554                         const struct cred *cred = file->f_cred;
 555                         struct user_namespace *user_ns = inst->peer_user_ns;
 556                         __be32 uid = htonl(from_kuid_munged(user_ns, cred->fsuid));
 557                         __be32 gid = htonl(from_kgid_munged(user_ns, cred->fsgid));
 558                         read_unlock_bh(&sk->sk_callback_lock);
 559                         if (nla_put_be32(inst->skb, NFULA_UID, uid) ||
 560                             nla_put_be32(inst->skb, NFULA_GID, gid))
 561                                 goto nla_put_failure;
 562                 } else
 563                         read_unlock_bh(&sk->sk_callback_lock);
 564         }
 565
 566         /* local sequence number */
 567         if ((inst->flags & NFULNL_CFG_F_SEQ) &&
 568             nla_put_be32(inst->skb, NFULA_SEQ, htonl(inst->seq++)))
 569                 goto nla_put_failure;
 570
 571         /* global sequence number */
 572         if ((inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) &&
 573             nla_put_be32(inst->skb, NFULA_SEQ_GLOBAL,
 574                          htonl(atomic_inc_return(&log->global_seq))))
 575                 goto nla_put_failure;
 576
 577         if (ct && nfnl_ct->build(inst->skb, ct, ctinfo,
 578                                  NFULA_CT, NFULA_CT_INFO) < 0)
 579                 goto nla_put_failure;
 580
 581         if (data_len) {
 582                 struct nlattr *nla;
 583                 int size = nla_attr_size(data_len);
 584
 585                 if (skb_tailroom(inst->skb) < nla_total_size(data_len))
 586                         goto nla_put_failure;
 587
 588                 nla = (struct nlattr *)skb_put(inst->skb, nla_total_size(data_len));
 589                 nla->nla_type = NFULA_PAYLOAD;
 590                 nla->nla_len = size;
 591
 592                 if (skb_copy_bits(skb, 0, nla_data(nla), data_len))
 593                         BUG();
 594         }
 595
 596         nlh->nlmsg_len = inst->skb->tail - old_tail;
 597         return 0;
 598
 599 nla_put_failure:
 600         PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n");
 601         return -1;
 602 }
 603
 604 static struct nf_loginfo default_loginfo = {
 605         .type =         NF_LOG_TYPE_ULOG,
 606         .u = {
 607                 .ulog = {
 608                         .copy_len       = 0xffff,
 609                         .group          = 0,
 610                         .qthreshold     = 1,
 611                 },
 612         },
 613 };
 614
 615 /* log handler for internal netfilter logging api */
 616 void
 617 nfulnl_log_packet(struct net *net,
 618                   u_int8_t pf,
 619                   unsigned int hooknum,
 620                   const struct sk_buff *skb,
 621                   const struct net_device *in,
 622                   const struct net_device *out,
 623                   const struct nf_loginfo *li_user,
 624                   const char *prefix)
 625 {
 626         size_t size;
 627         unsigned int data_len;
 628         struct nfulnl_instance *inst;
 629         const struct nf_loginfo *li;
 630         unsigned int qthreshold;
 631         unsigned int plen;
 632         struct nfnl_log_net *log = nfnl_log_pernet(net);
 633         const struct nfnl_ct_hook *nfnl_ct = NULL;
 634         struct nf_conn *ct = NULL;
 635         enum ip_conntrack_info uninitialized_var(ctinfo);
 636
 637         if (li_user && li_user->type == NF_LOG_TYPE_ULOG)
 638                 li = li_user;
 639         else
 640                 li = &default_loginfo;
 641
 642         inst = instance_lookup_get(log, li->u.ulog.group);
 643         if (!inst)
 644                 return;
 645
 646         plen = 0;
 647         if (prefix)
 648                 plen = strlen(prefix) + 1;
 649
 650         /* FIXME: do we want to make the size calculation conditional based on
 651          * what is actually present?  way more branches and checks, but more
 652          * memory efficient... */
 653         size =    nlmsg_total_size(sizeof(struct nfgenmsg))
 654                 + nla_total_size(sizeof(struct nfulnl_msg_packet_hdr))
 655                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 656                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 657 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 658                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 659                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 660 #endif
 661                 + nla_total_size(sizeof(u_int32_t))     /* mark */
 662                 + nla_total_size(sizeof(u_int32_t))     /* uid */
 663                 + nla_total_size(sizeof(u_int32_t))     /* gid */
 664                 + nla_total_size(plen)                  /* prefix */
 665                 + nla_total_size(sizeof(struct nfulnl_msg_packet_hw))
 666                 + nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp))
 667                 + nla_total_size(sizeof(struct nfgenmsg));      /* NLMSG_DONE */
 668
 669         if (in && skb_mac_header_was_set(skb)) {
 670                 size +=   nla_total_size(skb->dev->hard_header_len)
 671                         + nla_total_size(sizeof(u_int16_t))     /* hwtype */
 672                         + nla_total_size(sizeof(u_int16_t));    /* hwlen */
 673         }
 674
 675         spin_lock_bh(&inst->lock);
 676
 677         if (inst->flags & NFULNL_CFG_F_SEQ)
 678                 size += nla_total_size(sizeof(u_int32_t));
 679         if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
 680                 size += nla_total_size(sizeof(u_int32_t));
 681         if (inst->flags & NFULNL_CFG_F_CONNTRACK) {
 682                 nfnl_ct = rcu_dereference(nfnl_ct_hook);
 683                 if (nfnl_ct != NULL) {
 684                         ct = nfnl_ct->get_ct(skb, &ctinfo);
 685                         if (ct != NULL)
 686                                 size += nfnl_ct->build_size(ct);
 687                 }
 688         }
 689
 690         qthreshold = inst->qthreshold;
 691         /* per-rule qthreshold overrides per-instance */
 692         if (li->u.ulog.qthreshold)
 693                 if (qthreshold > li->u.ulog.qthreshold)
 694                         qthreshold = li->u.ulog.qthreshold;
 695
 696
 697         switch (inst->copy_mode) {
 698         case NFULNL_COPY_META:
 699         case NFULNL_COPY_NONE:
 700                 data_len = 0;
 701                 break;
 702
 703         case NFULNL_COPY_PACKET:
 704                 if (inst->copy_range > skb->len)
 705                         data_len = skb->len;
 706                 else
 707                         data_len = inst->copy_range;
 708
 709                 size += nla_total_size(data_len);
 710                 break;
 711
 712         case NFULNL_COPY_DISABLED:
 713         default:
 714                 goto unlock_and_release;
 715         }
 716
 717         if (inst->skb && size > skb_tailroom(inst->skb)) {
 718                 /* either the queue len is too high or we don't have
 719                  * enough room in the skb left. flush to userspace. */
 720                 __nfulnl_flush(inst);
 721         }
 722
 723         if (!inst->skb) {
 724                 inst->skb = nfulnl_alloc_skb(net, inst->peer_portid,
 725                                              inst->nlbufsiz, size);
 726                 if (!inst->skb)
 727                         goto alloc_failure;
 728         }
 729
 730         inst->qlen++;
 731
 732         __build_packet_message(log, inst, skb, data_len, pf,
 733                                 hooknum, in, out, prefix, plen,
 734                                 nfnl_ct, ct, ctinfo);
 735
 736         if (inst->qlen >= qthreshold)
 737                 __nfulnl_flush(inst);
 738         /* timer_pending always called within inst->lock, so there
 739          * is no chance of a race here */
 740         else if (!timer_pending(&inst->timer)) {
 741                 instance_get(inst);
 742                 inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100);
 743                 add_timer(&inst->timer);
 744         }
 745
 746 unlock_and_release:
 747         spin_unlock_bh(&inst->lock);
 748         instance_put(inst);
 749         return;
 750
 751 alloc_failure:
 752         /* FIXME: statistics */
 753         goto unlock_and_release;
 754 }
 755 EXPORT_SYMBOL_GPL(nfulnl_log_packet);
 756
 757 static int
 758 nfulnl_rcv_nl_event(struct notifier_block *this,
 759                    unsigned long event, void *ptr)
 760 {
 761         struct netlink_notify *n = ptr;
 762         struct nfnl_log_net *log = nfnl_log_pernet(n->net);
 763
 764         if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
 765                 int i;
 766
 767                 /* destroy all instances for this portid */
 768                 spin_lock_bh(&log->instances_lock);
 769                 for  (i = 0; i < INSTANCE_BUCKETS; i++) {
 770                         struct hlist_node *t2;
 771                         struct nfulnl_instance *inst;
 772                         struct hlist_head *head = &log->instance_table[i];
 773
 774                         hlist_for_each_entry_safe(inst, t2, head, hlist) {
 775                                 if (n->portid == inst->peer_portid)
 776                                         __instance_destroy(inst);
 777                         }
 778                 }
 779                 spin_unlock_bh(&log->instances_lock);
 780         }
 781         return NOTIFY_DONE;
 782 }
 783
 784 static struct notifier_block nfulnl_rtnl_notifier = {
 785         .notifier_call  = nfulnl_rcv_nl_event,
 786 };
 787
 788 static int nfulnl_recv_unsupp(struct net *net, struct sock *ctnl,
 789                               struct sk_buff *skb, const struct nlmsghdr *nlh,
 790                               const struct nlattr * const nfqa[])
 791 {
 792         return -ENOTSUPP;
 793 }
 794
 795 static struct nf_logger nfulnl_logger __read_mostly = {
 796         .name   = "nfnetlink_log",
 797         .type   = NF_LOG_TYPE_ULOG,
 798         .logfn  = &nfulnl_log_packet,
 799         .me     = THIS_MODULE,
 800 };
 801
 802 static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = {
 803         [NFULA_CFG_CMD]         = { .len = sizeof(struct nfulnl_msg_config_cmd) },
 804         [NFULA_CFG_MODE]        = { .len = sizeof(struct nfulnl_msg_config_mode) },
 805         [NFULA_CFG_TIMEOUT]     = { .type = NLA_U32 },
 806         [NFULA_CFG_QTHRESH]     = { .type = NLA_U32 },
 807         [NFULA_CFG_NLBUFSIZ]    = { .type = NLA_U32 },
 808         [NFULA_CFG_FLAGS]       = { .type = NLA_U16 },
 809 };
 810
 811 static int nfulnl_recv_config(struct net *net, struct sock *ctnl,
 812                               struct sk_buff *skb, const struct nlmsghdr *nlh,
 813                               const struct nlattr * const nfula[])
 814 {
 815         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
 816         u_int16_t group_num = ntohs(nfmsg->res_id);
 817         struct nfulnl_instance *inst;
 818         struct nfulnl_msg_config_cmd *cmd = NULL;
 819         struct nfnl_log_net *log = nfnl_log_pernet(net);
 820         int ret = 0;
 821         u16 flags = 0;
 822
 823         if (nfula[NFULA_CFG_CMD]) {
 824                 u_int8_t pf = nfmsg->nfgen_family;
 825                 cmd = nla_data(nfula[NFULA_CFG_CMD]);
 826
 827                 /* Commands without queue context */
 828                 switch (cmd->command) {
 829                 case NFULNL_CFG_CMD_PF_BIND:
 830                         return nf_log_bind_pf(net, pf, &nfulnl_logger);
 831                 case NFULNL_CFG_CMD_PF_UNBIND:
 832                         nf_log_unbind_pf(net, pf);
 833                         return 0;
 834                 }
 835         }
 836
 837         inst = instance_lookup_get(log, group_num);
 838         if (inst && inst->peer_portid != NETLINK_CB(skb).portid) {
 839                 ret = -EPERM;
 840                 goto out_put;
 841         }
 842
 843         /* Check if we support these flags in first place, dependencies should
 844          * be there too not to break atomicity.
 845          */
 846         if (nfula[NFULA_CFG_FLAGS]) {
 847                 flags = ntohs(nla_get_be16(nfula[NFULA_CFG_FLAGS]));
 848
 849                 if ((flags & NFULNL_CFG_F_CONNTRACK) &&
 850                     !rcu_access_pointer(nfnl_ct_hook)) {
 851 #ifdef CONFIG_MODULES
 852                         nfnl_unlock(NFNL_SUBSYS_ULOG);
 853                         request_module("ip_conntrack_netlink");
 854                         nfnl_lock(NFNL_SUBSYS_ULOG);
 855                         if (rcu_access_pointer(nfnl_ct_hook)) {
 856                                 ret = -EAGAIN;
 857                                 goto out_put;
 858                         }
 859 #endif
 860                         ret = -EOPNOTSUPP;
 861                         goto out_put;
 862                 }
 863         }
 864
 865         if (cmd != NULL) {
 866                 switch (cmd->command) {
 867                 case NFULNL_CFG_CMD_BIND:
 868                         if (inst) {
 869                                 ret = -EBUSY;
 870                                 goto out_put;
 871                         }
 872
 873                         inst = instance_create(net, group_num,
 874                                                NETLINK_CB(skb).portid,
 875                                                sk_user_ns(NETLINK_CB(skb).sk));
 876                         if (IS_ERR(inst)) {
 877                                 ret = PTR_ERR(inst);
 878                                 goto out;
 879                         }
 880                         break;
 881                 case NFULNL_CFG_CMD_UNBIND:
 882                         if (!inst) {
 883                                 ret = -ENODEV;
 884                                 goto out;
 885                         }
 886
 887                         instance_destroy(log, inst);
 888                         goto out_put;
 889                 default:
 890                         ret = -ENOTSUPP;
 891                         goto out_put;
 892                 }
 893         } else if (!inst) {
 894                 ret = -ENODEV;
 895                 goto out;
 896         }
 897
 898         if (nfula[NFULA_CFG_MODE]) {
 899                 struct nfulnl_msg_config_mode *params =
 900                         nla_data(nfula[NFULA_CFG_MODE]);
 901
 902                 nfulnl_set_mode(inst, params->copy_mode,
 903                                 ntohl(params->copy_range));
 904         }
 905
 906         if (nfula[NFULA_CFG_TIMEOUT]) {
 907                 __be32 timeout = nla_get_be32(nfula[NFULA_CFG_TIMEOUT]);
 908
 909                 nfulnl_set_timeout(inst, ntohl(timeout));
 910         }
 911
 912         if (nfula[NFULA_CFG_NLBUFSIZ]) {
 913                 __be32 nlbufsiz = nla_get_be32(nfula[NFULA_CFG_NLBUFSIZ]);
 914
 915                 nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz));
 916         }
 917
 918         if (nfula[NFULA_CFG_QTHRESH]) {
 919                 __be32 qthresh = nla_get_be32(nfula[NFULA_CFG_QTHRESH]);
 920
 921                 nfulnl_set_qthresh(inst, ntohl(qthresh));
 922         }
 923
 924         if (nfula[NFULA_CFG_FLAGS])
 925                 nfulnl_set_flags(inst, flags);
 926
 927 out_put:
 928         instance_put(inst);
 929 out:
 930         return ret;
 931 }
 932
 933 static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = {
 934         [NFULNL_MSG_PACKET]     = { .call = nfulnl_recv_unsupp,
 935                                     .attr_count = NFULA_MAX, },
 936         [NFULNL_MSG_CONFIG]     = { .call = nfulnl_recv_config,
 937                                     .attr_count = NFULA_CFG_MAX,
 938                                     .policy = nfula_cfg_policy },
 939 };
 940
 941 static const struct nfnetlink_subsystem nfulnl_subsys = {
 942         .name           = "log",
 943         .subsys_id      = NFNL_SUBSYS_ULOG,
 944         .cb_count       = NFULNL_MSG_MAX,
 945         .cb             = nfulnl_cb,
 946 };
 947
 948 #ifdef CONFIG_PROC_FS
 949 struct iter_state {
 950         struct seq_net_private p;
 951         unsigned int bucket;
 952 };
 953
 954 static struct hlist_node *get_first(struct net *net, struct iter_state *st)
 955 {
 956         struct nfnl_log_net *log;
 957         if (!st)
 958                 return NULL;
 959
 960         log = nfnl_log_pernet(net);
 961
 962         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
 963                 struct hlist_head *head = &log->instance_table[st->bucket];
 964
 965                 if (!hlist_empty(head))
 966                         return rcu_dereference_bh(hlist_first_rcu(head));
 967         }
 968         return NULL;
 969 }
 970
 971 static struct hlist_node *get_next(struct net *net, struct iter_state *st,
 972                                    struct hlist_node *h)
 973 {
 974         h = rcu_dereference_bh(hlist_next_rcu(h));
 975         while (!h) {
 976                 struct nfnl_log_net *log;
 977                 struct hlist_head *head;
 978
 979                 if (++st->bucket >= INSTANCE_BUCKETS)
 980                         return NULL;
 981
 982                 log = nfnl_log_pernet(net);
 983                 head = &log->instance_table[st->bucket];
 984                 h = rcu_dereference_bh(hlist_first_rcu(head));
 985         }
 986         return h;
 987 }
 988
 989 static struct hlist_node *get_idx(struct net *net, struct iter_state *st,
 990                                   loff_t pos)
 991 {
 992         struct hlist_node *head;
 993         head = get_first(net, st);
 994
 995         if (head)
 996                 while (pos && (head = get_next(net, st, head)))
 997                         pos--;
 998         return pos ? NULL : head;
 999 }
1000
1001 static void *seq_start(struct seq_file *s, loff_t *pos)
1002         __acquires(rcu_bh)
1003 {
1004         rcu_read_lock_bh();
1005         return get_idx(seq_file_net(s), s->private, *pos);
1006 }
1007
1008 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1009 {
1010         (*pos)++;
1011         return get_next(seq_file_net(s), s->private, v);
1012 }
1013
1014 static void seq_stop(struct seq_file *s, void *v)
1015         __releases(rcu_bh)
1016 {
1017         rcu_read_unlock_bh();
1018 }
1019
1020 static int seq_show(struct seq_file *s, void *v)
1021 {
1022         const struct nfulnl_instance *inst = v;
1023
1024         seq_printf(s, "%5u %6u %5u %1u %5u %6u %2u\n",
1025                    inst->group_num,
1026                    inst->peer_portid, inst->qlen,
1027                    inst->copy_mode, inst->copy_range,
1028                    inst->flushtimeout, atomic_read(&inst->use));
1029
1030         return 0;
1031 }
1032
1033 static const struct seq_operations nful_seq_ops = {
1034         .start  = seq_start,
1035         .next   = seq_next,
1036         .stop   = seq_stop,
1037         .show   = seq_show,
1038 };
1039
1040 static int nful_open(struct inode *inode, struct file *file)
1041 {
1042         return seq_open_net(inode, file, &nful_seq_ops,
1043                             sizeof(struct iter_state));
1044 }
1045
1046 static const struct file_operations nful_file_ops = {
1047         .owner   = THIS_MODULE,
1048         .open    = nful_open,
1049         .read    = seq_read,
1050         .llseek  = seq_lseek,
1051         .release = seq_release_net,
1052 };
1053
1054 #endif /* PROC_FS */
1055
1056 static int __net_init nfnl_log_net_init(struct net *net)
1057 {
1058         unsigned int i;
1059         struct nfnl_log_net *log = nfnl_log_pernet(net);
1060 #ifdef CONFIG_PROC_FS
1061         struct proc_dir_entry *proc;
1062         kuid_t root_uid;
1063         kgid_t root_gid;
1064 #endif
1065
1066         for (i = 0; i < INSTANCE_BUCKETS; i++)
1067                 INIT_HLIST_HEAD(&log->instance_table[i]);
1068         spin_lock_init(&log->instances_lock);
1069
1070 #ifdef CONFIG_PROC_FS
1071         proc = proc_create("nfnetlink_log", 0440,
1072                            net->nf.proc_netfilter, &nful_file_ops);
1073         if (!proc)
1074                 return -ENOMEM;
1075
1076         root_uid = make_kuid(net->user_ns, 0);
1077         root_gid = make_kgid(net->user_ns, 0);
1078         if (uid_valid(root_uid) && gid_valid(root_gid))
1079                 proc_set_user(proc, root_uid, root_gid);
1080 #endif
1081         return 0;
1082 }
1083
1084 static void __net_exit nfnl_log_net_exit(struct net *net)
1085 {
1086 #ifdef CONFIG_PROC_FS
1087         remove_proc_entry("nfnetlink_log", net->nf.proc_netfilter);
1088 #endif
1089         nf_log_unset(net, &nfulnl_logger);
1090 }
1091
1092 static struct pernet_operations nfnl_log_net_ops = {
1093         .init   = nfnl_log_net_init,
1094         .exit   = nfnl_log_net_exit,
1095         .id     = &nfnl_log_net_id,
1096         .size   = sizeof(struct nfnl_log_net),
1097 };
1098
1099 static int __init nfnetlink_log_init(void)
1100 {
1101         int status;
1102
1103         status = register_pernet_subsys(&nfnl_log_net_ops);
1104         if (status < 0) {
1105                 pr_err("failed to register pernet ops\n");
1106                 goto out;
1107         }
1108
1109         netlink_register_notifier(&nfulnl_rtnl_notifier);
1110         status = nfnetlink_subsys_register(&nfulnl_subsys);
1111         if (status < 0) {
1112                 pr_err("failed to create netlink socket\n");
1113                 goto cleanup_netlink_notifier;
1114         }
1115
1116         status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger);
1117         if (status < 0) {
1118                 pr_err("failed to register logger\n");
1119                 goto cleanup_subsys;
1120         }
1121
1122         return status;
1123
1124 cleanup_subsys:
1125         nfnetlink_subsys_unregister(&nfulnl_subsys);
1126 cleanup_netlink_notifier:
1127         netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1128         unregister_pernet_subsys(&nfnl_log_net_ops);
1129 out:
1130         return status;
1131 }
1132
1133 static void __exit nfnetlink_log_fini(void)
1134 {
1135         nf_log_unregister(&nfulnl_logger);
1136         nfnetlink_subsys_unregister(&nfulnl_subsys);
1137         netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1138         unregister_pernet_subsys(&nfnl_log_net_ops);
1139 }
1140
1141 MODULE_DESCRIPTION("netfilter userspace logging");
1142 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1143 MODULE_LICENSE("GPL");
1144 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG);
1145 MODULE_ALIAS_NF_LOGGER(AF_INET, 1);
1146 MODULE_ALIAS_NF_LOGGER(AF_INET6, 1);
1147 MODULE_ALIAS_NF_LOGGER(AF_BRIDGE, 1);
1148
1149 module_init(nfnetlink_log_init);
1150 module_exit(nfnetlink_log_fini);