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 #include <linux/refcount.h>
  44
  45
  46 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  47 #include "../bridge/br_private.h"
  48 #endif
  49
  50 #define NFULNL_NLBUFSIZ_DEFAULT NLMSG_GOODSIZE
  51 #define NFULNL_TIMEOUT_DEFAULT  100     /* every second */
  52 #define NFULNL_QTHRESH_DEFAULT  100     /* 100 packets */
  53 /* max packet size is limited by 16-bit struct nfattr nfa_len field */
  54 #define NFULNL_COPY_RANGE_MAX   (0xFFFF - NLA_HDRLEN)
  55
  56 #define PRINTR(x, args...)      do { if (net_ratelimit()) \
  57                                      printk(x, ## args); } while (0);
  58
  59 struct nfulnl_instance {
  60         struct hlist_node hlist;        /* global list of instances */
  61         spinlock_t lock;
  62         refcount_t use;                 /* use count */
  63
  64         unsigned int qlen;              /* number of nlmsgs in skb */
  65         struct sk_buff *skb;            /* pre-allocatd skb */
  66         struct timer_list timer;
  67         struct net *net;
  68         struct user_namespace *peer_user_ns;    /* User namespace of the peer process */
  69         u32 peer_portid;                /* PORTID of the peer process */
  70
  71         /* configurable parameters */
  72         unsigned int flushtimeout;      /* timeout until queue flush */
  73         unsigned int nlbufsiz;          /* netlink buffer allocation size */
  74         unsigned int qthreshold;        /* threshold of the queue */
  75         u_int32_t copy_range;
  76         u_int32_t seq;                  /* instance-local sequential counter */
  77         u_int16_t group_num;            /* number of this queue */
  78         u_int16_t flags;
  79         u_int8_t copy_mode;
  80         struct rcu_head rcu;
  81 };
  82
  83 #define INSTANCE_BUCKETS        16
  84
  85 static unsigned int nfnl_log_net_id __read_mostly;
  86
  87 struct nfnl_log_net {
  88         spinlock_t instances_lock;
  89         struct hlist_head instance_table[INSTANCE_BUCKETS];
  90         atomic_t global_seq;
  91 };
  92
  93 static struct nfnl_log_net *nfnl_log_pernet(struct net *net)
  94 {
  95         return net_generic(net, nfnl_log_net_id);
  96 }
  97
  98 static inline u_int8_t instance_hashfn(u_int16_t group_num)
  99 {
 100         return ((group_num & 0xff) % INSTANCE_BUCKETS);
 101 }
 102
 103 static struct nfulnl_instance *
 104 __instance_lookup(struct nfnl_log_net *log, u_int16_t group_num)
 105 {
 106         struct hlist_head *head;
 107         struct nfulnl_instance *inst;
 108
 109         head = &log->instance_table[instance_hashfn(group_num)];
 110         hlist_for_each_entry_rcu(inst, head, hlist) {
 111                 if (inst->group_num == group_num)
 112                         return inst;
 113         }
 114         return NULL;
 115 }
 116
 117 static inline void
 118 instance_get(struct nfulnl_instance *inst)
 119 {
 120         refcount_inc(&inst->use);
 121 }
 122
 123 static struct nfulnl_instance *
 124 instance_lookup_get(struct nfnl_log_net *log, u_int16_t group_num)
 125 {
 126         struct nfulnl_instance *inst;
 127
 128         rcu_read_lock_bh();
 129         inst = __instance_lookup(log, group_num);
 130         if (inst && !refcount_inc_not_zero(&inst->use))
 131                 inst = NULL;
 132         rcu_read_unlock_bh();
 133
 134         return inst;
 135 }
 136
 137 static void nfulnl_instance_free_rcu(struct rcu_head *head)
 138 {
 139         struct nfulnl_instance *inst =
 140                 container_of(head, struct nfulnl_instance, rcu);
 141
 142         put_net(inst->net);
 143         kfree(inst);
 144         module_put(THIS_MODULE);
 145 }
 146
 147 static void
 148 instance_put(struct nfulnl_instance *inst)
 149 {
 150         if (inst && refcount_dec_and_test(&inst->use))
 151                 call_rcu_bh(&inst->rcu, nfulnl_instance_free_rcu);
 152 }
 153
 154 static void nfulnl_timer(unsigned long data);
 155
 156 static struct nfulnl_instance *
 157 instance_create(struct net *net, u_int16_t group_num,
 158                 u32 portid, struct user_namespace *user_ns)
 159 {
 160         struct nfulnl_instance *inst;
 161         struct nfnl_log_net *log = nfnl_log_pernet(net);
 162         int err;
 163
 164         spin_lock_bh(&log->instances_lock);
 165         if (__instance_lookup(log, group_num)) {
 166                 err = -EEXIST;
 167                 goto out_unlock;
 168         }
 169
 170         inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
 171         if (!inst) {
 172                 err = -ENOMEM;
 173                 goto out_unlock;
 174         }
 175
 176         if (!try_module_get(THIS_MODULE)) {
 177                 kfree(inst);
 178                 err = -EAGAIN;
 179                 goto out_unlock;
 180         }
 181
 182         INIT_HLIST_NODE(&inst->hlist);
 183         spin_lock_init(&inst->lock);
 184         /* needs to be two, since we _put() after creation */
 185         refcount_set(&inst->use, 2);
 186
 187         setup_timer(&inst->timer, nfulnl_timer, (unsigned long)inst);
 188
 189         inst->net = get_net(net);
 190         inst->peer_user_ns = user_ns;
 191         inst->peer_portid = portid;
 192         inst->group_num = group_num;
 193
 194         inst->qthreshold        = NFULNL_QTHRESH_DEFAULT;
 195         inst->flushtimeout      = NFULNL_TIMEOUT_DEFAULT;
 196         inst->nlbufsiz          = NFULNL_NLBUFSIZ_DEFAULT;
 197         inst->copy_mode         = NFULNL_COPY_PACKET;
 198         inst->copy_range        = NFULNL_COPY_RANGE_MAX;
 199
 200         hlist_add_head_rcu(&inst->hlist,
 201                        &log->instance_table[instance_hashfn(group_num)]);
 202
 203
 204         spin_unlock_bh(&log->instances_lock);
 205
 206         return inst;
 207
 208 out_unlock:
 209         spin_unlock_bh(&log->instances_lock);
 210         return ERR_PTR(err);
 211 }
 212
 213 static void __nfulnl_flush(struct nfulnl_instance *inst);
 214
 215 /* called with BH disabled */
 216 static void
 217 __instance_destroy(struct nfulnl_instance *inst)
 218 {
 219         /* first pull it out of the global list */
 220         hlist_del_rcu(&inst->hlist);
 221
 222         /* then flush all pending packets from skb */
 223
 224         spin_lock(&inst->lock);
 225
 226         /* lockless readers wont be able to use us */
 227         inst->copy_mode = NFULNL_COPY_DISABLED;
 228
 229         if (inst->skb)
 230                 __nfulnl_flush(inst);
 231         spin_unlock(&inst->lock);
 232
 233         /* and finally put the refcount */
 234         instance_put(inst);
 235 }
 236
 237 static inline void
 238 instance_destroy(struct nfnl_log_net *log,
 239                  struct nfulnl_instance *inst)
 240 {
 241         spin_lock_bh(&log->instances_lock);
 242         __instance_destroy(inst);
 243         spin_unlock_bh(&log->instances_lock);
 244 }
 245
 246 static int
 247 nfulnl_set_mode(struct nfulnl_instance *inst, u_int8_t mode,
 248                   unsigned int range)
 249 {
 250         int status = 0;
 251
 252         spin_lock_bh(&inst->lock);
 253
 254         switch (mode) {
 255         case NFULNL_COPY_NONE:
 256         case NFULNL_COPY_META:
 257                 inst->copy_mode = mode;
 258                 inst->copy_range = 0;
 259                 break;
 260
 261         case NFULNL_COPY_PACKET:
 262                 inst->copy_mode = mode;
 263                 if (range == 0)
 264                         range = NFULNL_COPY_RANGE_MAX;
 265                 inst->copy_range = min_t(unsigned int,
 266                                          range, NFULNL_COPY_RANGE_MAX);
 267                 break;
 268
 269         default:
 270                 status = -EINVAL;
 271                 break;
 272         }
 273
 274         spin_unlock_bh(&inst->lock);
 275
 276         return status;
 277 }
 278
 279 static int
 280 nfulnl_set_nlbufsiz(struct nfulnl_instance *inst, u_int32_t nlbufsiz)
 281 {
 282         int status;
 283
 284         spin_lock_bh(&inst->lock);
 285         if (nlbufsiz < NFULNL_NLBUFSIZ_DEFAULT)
 286                 status = -ERANGE;
 287         else if (nlbufsiz > 131072)
 288                 status = -ERANGE;
 289         else {
 290                 inst->nlbufsiz = nlbufsiz;
 291                 status = 0;
 292         }
 293         spin_unlock_bh(&inst->lock);
 294
 295         return status;
 296 }
 297
 298 static void
 299 nfulnl_set_timeout(struct nfulnl_instance *inst, u_int32_t timeout)
 300 {
 301         spin_lock_bh(&inst->lock);
 302         inst->flushtimeout = timeout;
 303         spin_unlock_bh(&inst->lock);
 304 }
 305
 306 static void
 307 nfulnl_set_qthresh(struct nfulnl_instance *inst, u_int32_t qthresh)
 308 {
 309         spin_lock_bh(&inst->lock);
 310         inst->qthreshold = qthresh;
 311         spin_unlock_bh(&inst->lock);
 312 }
 313
 314 static int
 315 nfulnl_set_flags(struct nfulnl_instance *inst, u_int16_t flags)
 316 {
 317         spin_lock_bh(&inst->lock);
 318         inst->flags = flags;
 319         spin_unlock_bh(&inst->lock);
 320
 321         return 0;
 322 }
 323
 324 static struct sk_buff *
 325 nfulnl_alloc_skb(struct net *net, u32 peer_portid, unsigned int inst_size,
 326                  unsigned int pkt_size)
 327 {
 328         struct sk_buff *skb;
 329         unsigned int n;
 330
 331         /* alloc skb which should be big enough for a whole multipart
 332          * message.  WARNING: has to be <= 128k due to slab restrictions */
 333
 334         n = max(inst_size, pkt_size);
 335         skb = alloc_skb(n, GFP_ATOMIC | __GFP_NOWARN);
 336         if (!skb) {
 337                 if (n > pkt_size) {
 338                         /* try to allocate only as much as we need for current
 339                          * packet */
 340
 341                         skb = alloc_skb(pkt_size, GFP_ATOMIC);
 342                 }
 343         }
 344
 345         return skb;
 346 }
 347
 348 static void
 349 __nfulnl_send(struct nfulnl_instance *inst)
 350 {
 351         if (inst->qlen > 1) {
 352                 struct nlmsghdr *nlh = nlmsg_put(inst->skb, 0, 0,
 353                                                  NLMSG_DONE,
 354                                                  sizeof(struct nfgenmsg),
 355                                                  0);
 356                 if (WARN_ONCE(!nlh, "bad nlskb size: %u, tailroom %d\n",
 357                               inst->skb->len, skb_tailroom(inst->skb))) {
 358                         kfree_skb(inst->skb);
 359                         goto out;
 360                 }
 361         }
 362         nfnetlink_unicast(inst->skb, inst->net, inst->peer_portid,
 363                           MSG_DONTWAIT);
 364 out:
 365         inst->qlen = 0;
 366         inst->skb = NULL;
 367 }
 368
 369 static void
 370 __nfulnl_flush(struct nfulnl_instance *inst)
 371 {
 372         /* timer holds a reference */
 373         if (del_timer(&inst->timer))
 374                 instance_put(inst);
 375         if (inst->skb)
 376                 __nfulnl_send(inst);
 377 }
 378
 379 static void
 380 nfulnl_timer(unsigned long data)
 381 {
 382         struct nfulnl_instance *inst = (struct nfulnl_instance *)data;
 383
 384         spin_lock_bh(&inst->lock);
 385         if (inst->skb)
 386                 __nfulnl_send(inst);
 387         spin_unlock_bh(&inst->lock);
 388         instance_put(inst);
 389 }
 390
 391 /* This is an inline function, we don't really care about a long
 392  * list of arguments */
 393 static inline int
 394 __build_packet_message(struct nfnl_log_net *log,
 395                         struct nfulnl_instance *inst,
 396                         const struct sk_buff *skb,
 397                         unsigned int data_len,
 398                         u_int8_t pf,
 399                         unsigned int hooknum,
 400                         const struct net_device *indev,
 401                         const struct net_device *outdev,
 402                         const char *prefix, unsigned int plen,
 403                         const struct nfnl_ct_hook *nfnl_ct,
 404                         struct nf_conn *ct, enum ip_conntrack_info ctinfo)
 405 {
 406         struct nfulnl_msg_packet_hdr pmsg;
 407         struct nlmsghdr *nlh;
 408         struct nfgenmsg *nfmsg;
 409         sk_buff_data_t old_tail = inst->skb->tail;
 410         struct sock *sk;
 411         const unsigned char *hwhdrp;
 412
 413         nlh = nlmsg_put(inst->skb, 0, 0,
 414                         nfnl_msg_type(NFNL_SUBSYS_ULOG, NFULNL_MSG_PACKET),
 415                         sizeof(struct nfgenmsg), 0);
 416         if (!nlh)
 417                 return -1;
 418         nfmsg = nlmsg_data(nlh);
 419         nfmsg->nfgen_family = pf;
 420         nfmsg->version = NFNETLINK_V0;
 421         nfmsg->res_id = htons(inst->group_num);
 422
 423         memset(&pmsg, 0, sizeof(pmsg));
 424         pmsg.hw_protocol        = skb->protocol;
 425         pmsg.hook               = hooknum;
 426
 427         if (nla_put(inst->skb, NFULA_PACKET_HDR, sizeof(pmsg), &pmsg))
 428                 goto nla_put_failure;
 429
 430         if (prefix &&
 431             nla_put(inst->skb, NFULA_PREFIX, plen, prefix))
 432                 goto nla_put_failure;
 433
 434         if (indev) {
 435 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 436                 if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 437                                  htonl(indev->ifindex)))
 438                         goto nla_put_failure;
 439 #else
 440                 if (pf == PF_BRIDGE) {
 441                         /* Case 1: outdev is physical input device, we need to
 442                          * look for bridge group (when called from
 443                          * netfilter_bridge) */
 444                         if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
 445                                          htonl(indev->ifindex)) ||
 446                         /* this is the bridge group "brX" */
 447                         /* rcu_read_lock()ed by nf_hook_thresh or
 448                          * nf_log_packet.
 449                          */
 450                             nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 451                                          htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
 452                                 goto nla_put_failure;
 453                 } else {
 454                         struct net_device *physindev;
 455
 456                         /* Case 2: indev is bridge group, we need to look for
 457                          * physical device (when called from ipv4) */
 458                         if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 459                                          htonl(indev->ifindex)))
 460                                 goto nla_put_failure;
 461
 462                         physindev = nf_bridge_get_physindev(skb);
 463                         if (physindev &&
 464                             nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
 465                                          htonl(physindev->ifindex)))
 466                                 goto nla_put_failure;
 467                 }
 468 #endif
 469         }
 470
 471         if (outdev) {
 472 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 473                 if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 474                                  htonl(outdev->ifindex)))
 475                         goto nla_put_failure;
 476 #else
 477                 if (pf == PF_BRIDGE) {
 478                         /* Case 1: outdev is physical output device, we need to
 479                          * look for bridge group (when called from
 480                          * netfilter_bridge) */
 481                         if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
 482                                          htonl(outdev->ifindex)) ||
 483                         /* this is the bridge group "brX" */
 484                         /* rcu_read_lock()ed by nf_hook_thresh or
 485                          * nf_log_packet.
 486                          */
 487                             nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 488                                          htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
 489                                 goto nla_put_failure;
 490                 } else {
 491                         struct net_device *physoutdev;
 492
 493                         /* Case 2: indev is a bridge group, we need to look
 494                          * for physical device (when called from ipv4) */
 495                         if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 496                                          htonl(outdev->ifindex)))
 497                                 goto nla_put_failure;
 498
 499                         physoutdev = nf_bridge_get_physoutdev(skb);
 500                         if (physoutdev &&
 501                             nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
 502                                          htonl(physoutdev->ifindex)))
 503                                 goto nla_put_failure;
 504                 }
 505 #endif
 506         }
 507
 508         if (skb->mark &&
 509             nla_put_be32(inst->skb, NFULA_MARK, htonl(skb->mark)))
 510                 goto nla_put_failure;
 511
 512         if (indev && skb->dev &&
 513             skb_mac_header_was_set(skb) &&
 514             skb_mac_header_len(skb) != 0) {
 515                 struct nfulnl_msg_packet_hw phw;
 516                 int len;
 517
 518                 memset(&phw, 0, sizeof(phw));
 519                 len = dev_parse_header(skb, phw.hw_addr);
 520                 if (len > 0) {
 521                         phw.hw_addrlen = htons(len);
 522                         if (nla_put(inst->skb, NFULA_HWADDR, sizeof(phw), &phw))
 523                                 goto nla_put_failure;
 524                 }
 525         }
 526
 527         if (indev && skb_mac_header_was_set(skb)) {
 528                 if (nla_put_be16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type)) ||
 529                     nla_put_be16(inst->skb, NFULA_HWLEN,
 530                                  htons(skb->dev->hard_header_len)))
 531                         goto nla_put_failure;
 532
 533                 hwhdrp = skb_mac_header(skb);
 534
 535                 if (skb->dev->type == ARPHRD_SIT)
 536                         hwhdrp -= ETH_HLEN;
 537
 538                 if (hwhdrp >= skb->head &&
 539                     nla_put(inst->skb, NFULA_HWHEADER,
 540                             skb->dev->hard_header_len, hwhdrp))
 541                         goto nla_put_failure;
 542         }
 543
 544         if (skb->tstamp) {
 545                 struct nfulnl_msg_packet_timestamp ts;
 546                 struct timespec64 kts = ktime_to_timespec64(skb->tstamp);
 547                 ts.sec = cpu_to_be64(kts.tv_sec);
 548                 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
 549
 550                 if (nla_put(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts))
 551                         goto nla_put_failure;
 552         }
 553
 554         /* UID */
 555         sk = skb->sk;
 556         if (sk && sk_fullsock(sk)) {
 557                 read_lock_bh(&sk->sk_callback_lock);
 558                 if (sk->sk_socket && sk->sk_socket->file) {
 559                         struct file *file = sk->sk_socket->file;
 560                         const struct cred *cred = file->f_cred;
 561                         struct user_namespace *user_ns = inst->peer_user_ns;
 562                         __be32 uid = htonl(from_kuid_munged(user_ns, cred->fsuid));
 563                         __be32 gid = htonl(from_kgid_munged(user_ns, cred->fsgid));
 564                         read_unlock_bh(&sk->sk_callback_lock);
 565                         if (nla_put_be32(inst->skb, NFULA_UID, uid) ||
 566                             nla_put_be32(inst->skb, NFULA_GID, gid))
 567                                 goto nla_put_failure;
 568                 } else
 569                         read_unlock_bh(&sk->sk_callback_lock);
 570         }
 571
 572         /* local sequence number */
 573         if ((inst->flags & NFULNL_CFG_F_SEQ) &&
 574             nla_put_be32(inst->skb, NFULA_SEQ, htonl(inst->seq++)))
 575                 goto nla_put_failure;
 576
 577         /* global sequence number */
 578         if ((inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) &&
 579             nla_put_be32(inst->skb, NFULA_SEQ_GLOBAL,
 580                          htonl(atomic_inc_return(&log->global_seq))))
 581                 goto nla_put_failure;
 582
 583         if (ct && nfnl_ct->build(inst->skb, ct, ctinfo,
 584                                  NFULA_CT, NFULA_CT_INFO) < 0)
 585                 goto nla_put_failure;
 586
 587         if (data_len) {
 588                 struct nlattr *nla;
 589                 int size = nla_attr_size(data_len);
 590
 591                 if (skb_tailroom(inst->skb) < nla_total_size(data_len))
 592                         goto nla_put_failure;
 593
 594                 nla = skb_put(inst->skb, nla_total_size(data_len));
 595                 nla->nla_type = NFULA_PAYLOAD;
 596                 nla->nla_len = size;
 597
 598                 if (skb_copy_bits(skb, 0, nla_data(nla), data_len))
 599                         BUG();
 600         }
 601
 602         nlh->nlmsg_len = inst->skb->tail - old_tail;
 603         return 0;
 604
 605 nla_put_failure:
 606         PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n");
 607         return -1;
 608 }
 609
 610 static const struct nf_loginfo default_loginfo = {
 611         .type =         NF_LOG_TYPE_ULOG,
 612         .u = {
 613                 .ulog = {
 614                         .copy_len       = 0xffff,
 615                         .group          = 0,
 616                         .qthreshold     = 1,
 617                 },
 618         },
 619 };
 620
 621 /* log handler for internal netfilter logging api */
 622 void
 623 nfulnl_log_packet(struct net *net,
 624                   u_int8_t pf,
 625                   unsigned int hooknum,
 626                   const struct sk_buff *skb,
 627                   const struct net_device *in,
 628                   const struct net_device *out,
 629                   const struct nf_loginfo *li_user,
 630                   const char *prefix)
 631 {
 632         size_t size;
 633         unsigned int data_len;
 634         struct nfulnl_instance *inst;
 635         const struct nf_loginfo *li;
 636         unsigned int qthreshold;
 637         unsigned int plen;
 638         struct nfnl_log_net *log = nfnl_log_pernet(net);
 639         const struct nfnl_ct_hook *nfnl_ct = NULL;
 640         struct nf_conn *ct = NULL;
 641         enum ip_conntrack_info uninitialized_var(ctinfo);
 642
 643         if (li_user && li_user->type == NF_LOG_TYPE_ULOG)
 644                 li = li_user;
 645         else
 646                 li = &default_loginfo;
 647
 648         inst = instance_lookup_get(log, li->u.ulog.group);
 649         if (!inst)
 650                 return;
 651
 652         plen = 0;
 653         if (prefix)
 654                 plen = strlen(prefix) + 1;
 655
 656         /* FIXME: do we want to make the size calculation conditional based on
 657          * what is actually present?  way more branches and checks, but more
 658          * memory efficient... */
 659         size =    nlmsg_total_size(sizeof(struct nfgenmsg))
 660                 + nla_total_size(sizeof(struct nfulnl_msg_packet_hdr))
 661                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 662                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 663 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 664                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 665                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 666 #endif
 667                 + nla_total_size(sizeof(u_int32_t))     /* mark */
 668                 + nla_total_size(sizeof(u_int32_t))     /* uid */
 669                 + nla_total_size(sizeof(u_int32_t))     /* gid */
 670                 + nla_total_size(plen)                  /* prefix */
 671                 + nla_total_size(sizeof(struct nfulnl_msg_packet_hw))
 672                 + nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp))
 673                 + nla_total_size(sizeof(struct nfgenmsg));      /* NLMSG_DONE */
 674
 675         if (in && skb_mac_header_was_set(skb)) {
 676                 size +=   nla_total_size(skb->dev->hard_header_len)
 677                         + nla_total_size(sizeof(u_int16_t))     /* hwtype */
 678                         + nla_total_size(sizeof(u_int16_t));    /* hwlen */
 679         }
 680
 681         spin_lock_bh(&inst->lock);
 682
 683         if (inst->flags & NFULNL_CFG_F_SEQ)
 684                 size += nla_total_size(sizeof(u_int32_t));
 685         if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
 686                 size += nla_total_size(sizeof(u_int32_t));
 687         if (inst->flags & NFULNL_CFG_F_CONNTRACK) {
 688                 nfnl_ct = rcu_dereference(nfnl_ct_hook);
 689                 if (nfnl_ct != NULL) {
 690                         ct = nfnl_ct->get_ct(skb, &ctinfo);
 691                         if (ct != NULL)
 692                                 size += nfnl_ct->build_size(ct);
 693                 }
 694         }
 695
 696         qthreshold = inst->qthreshold;
 697         /* per-rule qthreshold overrides per-instance */
 698         if (li->u.ulog.qthreshold)
 699                 if (qthreshold > li->u.ulog.qthreshold)
 700                         qthreshold = li->u.ulog.qthreshold;
 701
 702
 703         switch (inst->copy_mode) {
 704         case NFULNL_COPY_META:
 705         case NFULNL_COPY_NONE:
 706                 data_len = 0;
 707                 break;
 708
 709         case NFULNL_COPY_PACKET:
 710                 data_len = inst->copy_range;
 711                 if ((li->u.ulog.flags & NF_LOG_F_COPY_LEN) &&
 712                     (li->u.ulog.copy_len < data_len))
 713                         data_len = li->u.ulog.copy_len;
 714
 715                 if (data_len > skb->len)
 716                         data_len = skb->len;
 717
 718                 size += nla_total_size(data_len);
 719                 break;
 720
 721         case NFULNL_COPY_DISABLED:
 722         default:
 723                 goto unlock_and_release;
 724         }
 725
 726         if (inst->skb && size > skb_tailroom(inst->skb)) {
 727                 /* either the queue len is too high or we don't have
 728                  * enough room in the skb left. flush to userspace. */
 729                 __nfulnl_flush(inst);
 730         }
 731
 732         if (!inst->skb) {
 733                 inst->skb = nfulnl_alloc_skb(net, inst->peer_portid,
 734                                              inst->nlbufsiz, size);
 735                 if (!inst->skb)
 736                         goto alloc_failure;
 737         }
 738
 739         inst->qlen++;
 740
 741         __build_packet_message(log, inst, skb, data_len, pf,
 742                                 hooknum, in, out, prefix, plen,
 743                                 nfnl_ct, ct, ctinfo);
 744
 745         if (inst->qlen >= qthreshold)
 746                 __nfulnl_flush(inst);
 747         /* timer_pending always called within inst->lock, so there
 748          * is no chance of a race here */
 749         else if (!timer_pending(&inst->timer)) {
 750                 instance_get(inst);
 751                 inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100);
 752                 add_timer(&inst->timer);
 753         }
 754
 755 unlock_and_release:
 756         spin_unlock_bh(&inst->lock);
 757         instance_put(inst);
 758         return;
 759
 760 alloc_failure:
 761         /* FIXME: statistics */
 762         goto unlock_and_release;
 763 }
 764 EXPORT_SYMBOL_GPL(nfulnl_log_packet);
 765
 766 static int
 767 nfulnl_rcv_nl_event(struct notifier_block *this,
 768                    unsigned long event, void *ptr)
 769 {
 770         struct netlink_notify *n = ptr;
 771         struct nfnl_log_net *log = nfnl_log_pernet(n->net);
 772
 773         if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
 774                 int i;
 775
 776                 /* destroy all instances for this portid */
 777                 spin_lock_bh(&log->instances_lock);
 778                 for  (i = 0; i < INSTANCE_BUCKETS; i++) {
 779                         struct hlist_node *t2;
 780                         struct nfulnl_instance *inst;
 781                         struct hlist_head *head = &log->instance_table[i];
 782
 783                         hlist_for_each_entry_safe(inst, t2, head, hlist) {
 784                                 if (n->portid == inst->peer_portid)
 785                                         __instance_destroy(inst);
 786                         }
 787                 }
 788                 spin_unlock_bh(&log->instances_lock);
 789         }
 790         return NOTIFY_DONE;
 791 }
 792
 793 static struct notifier_block nfulnl_rtnl_notifier = {
 794         .notifier_call  = nfulnl_rcv_nl_event,
 795 };
 796
 797 static int nfulnl_recv_unsupp(struct net *net, struct sock *ctnl,
 798                               struct sk_buff *skb, const struct nlmsghdr *nlh,
 799                               const struct nlattr * const nfqa[],
 800                               struct netlink_ext_ack *extack)
 801 {
 802         return -ENOTSUPP;
 803 }
 804
 805 static struct nf_logger nfulnl_logger __read_mostly = {
 806         .name   = "nfnetlink_log",
 807         .type   = NF_LOG_TYPE_ULOG,
 808         .logfn  = nfulnl_log_packet,
 809         .me     = THIS_MODULE,
 810 };
 811
 812 static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = {
 813         [NFULA_CFG_CMD]         = { .len = sizeof(struct nfulnl_msg_config_cmd) },
 814         [NFULA_CFG_MODE]        = { .len = sizeof(struct nfulnl_msg_config_mode) },
 815         [NFULA_CFG_TIMEOUT]     = { .type = NLA_U32 },
 816         [NFULA_CFG_QTHRESH]     = { .type = NLA_U32 },
 817         [NFULA_CFG_NLBUFSIZ]    = { .type = NLA_U32 },
 818         [NFULA_CFG_FLAGS]       = { .type = NLA_U16 },
 819 };
 820
 821 static int nfulnl_recv_config(struct net *net, struct sock *ctnl,
 822                               struct sk_buff *skb, const struct nlmsghdr *nlh,
 823                               const struct nlattr * const nfula[],
 824                               struct netlink_ext_ack *extack)
 825 {
 826         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
 827         u_int16_t group_num = ntohs(nfmsg->res_id);
 828         struct nfulnl_instance *inst;
 829         struct nfulnl_msg_config_cmd *cmd = NULL;
 830         struct nfnl_log_net *log = nfnl_log_pernet(net);
 831         int ret = 0;
 832         u16 flags = 0;
 833
 834         if (nfula[NFULA_CFG_CMD]) {
 835                 u_int8_t pf = nfmsg->nfgen_family;
 836                 cmd = nla_data(nfula[NFULA_CFG_CMD]);
 837
 838                 /* Commands without queue context */
 839                 switch (cmd->command) {
 840                 case NFULNL_CFG_CMD_PF_BIND:
 841                         return nf_log_bind_pf(net, pf, &nfulnl_logger);
 842                 case NFULNL_CFG_CMD_PF_UNBIND:
 843                         nf_log_unbind_pf(net, pf);
 844                         return 0;
 845                 }
 846         }
 847
 848         inst = instance_lookup_get(log, group_num);
 849         if (inst && inst->peer_portid != NETLINK_CB(skb).portid) {
 850                 ret = -EPERM;
 851                 goto out_put;
 852         }
 853
 854         /* Check if we support these flags in first place, dependencies should
 855          * be there too not to break atomicity.
 856          */
 857         if (nfula[NFULA_CFG_FLAGS]) {
 858                 flags = ntohs(nla_get_be16(nfula[NFULA_CFG_FLAGS]));
 859
 860                 if ((flags & NFULNL_CFG_F_CONNTRACK) &&
 861                     !rcu_access_pointer(nfnl_ct_hook)) {
 862 #ifdef CONFIG_MODULES
 863                         nfnl_unlock(NFNL_SUBSYS_ULOG);
 864                         request_module("ip_conntrack_netlink");
 865                         nfnl_lock(NFNL_SUBSYS_ULOG);
 866                         if (rcu_access_pointer(nfnl_ct_hook)) {
 867                                 ret = -EAGAIN;
 868                                 goto out_put;
 869                         }
 870 #endif
 871                         ret = -EOPNOTSUPP;
 872                         goto out_put;
 873                 }
 874         }
 875
 876         if (cmd != NULL) {
 877                 switch (cmd->command) {
 878                 case NFULNL_CFG_CMD_BIND:
 879                         if (inst) {
 880                                 ret = -EBUSY;
 881                                 goto out_put;
 882                         }
 883
 884                         inst = instance_create(net, group_num,
 885                                                NETLINK_CB(skb).portid,
 886                                                sk_user_ns(NETLINK_CB(skb).sk));
 887                         if (IS_ERR(inst)) {
 888                                 ret = PTR_ERR(inst);
 889                                 goto out;
 890                         }
 891                         break;
 892                 case NFULNL_CFG_CMD_UNBIND:
 893                         if (!inst) {
 894                                 ret = -ENODEV;
 895                                 goto out;
 896                         }
 897
 898                         instance_destroy(log, inst);
 899                         goto out_put;
 900                 default:
 901                         ret = -ENOTSUPP;
 902                         goto out_put;
 903                 }
 904         } else if (!inst) {
 905                 ret = -ENODEV;
 906                 goto out;
 907         }
 908
 909         if (nfula[NFULA_CFG_MODE]) {
 910                 struct nfulnl_msg_config_mode *params =
 911                         nla_data(nfula[NFULA_CFG_MODE]);
 912
 913                 nfulnl_set_mode(inst, params->copy_mode,
 914                                 ntohl(params->copy_range));
 915         }
 916
 917         if (nfula[NFULA_CFG_TIMEOUT]) {
 918                 __be32 timeout = nla_get_be32(nfula[NFULA_CFG_TIMEOUT]);
 919
 920                 nfulnl_set_timeout(inst, ntohl(timeout));
 921         }
 922
 923         if (nfula[NFULA_CFG_NLBUFSIZ]) {
 924                 __be32 nlbufsiz = nla_get_be32(nfula[NFULA_CFG_NLBUFSIZ]);
 925
 926                 nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz));
 927         }
 928
 929         if (nfula[NFULA_CFG_QTHRESH]) {
 930                 __be32 qthresh = nla_get_be32(nfula[NFULA_CFG_QTHRESH]);
 931
 932                 nfulnl_set_qthresh(inst, ntohl(qthresh));
 933         }
 934
 935         if (nfula[NFULA_CFG_FLAGS])
 936                 nfulnl_set_flags(inst, flags);
 937
 938 out_put:
 939         instance_put(inst);
 940 out:
 941         return ret;
 942 }
 943
 944 static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = {
 945         [NFULNL_MSG_PACKET]     = { .call = nfulnl_recv_unsupp,
 946                                     .attr_count = NFULA_MAX, },
 947         [NFULNL_MSG_CONFIG]     = { .call = nfulnl_recv_config,
 948                                     .attr_count = NFULA_CFG_MAX,
 949                                     .policy = nfula_cfg_policy },
 950 };
 951
 952 static const struct nfnetlink_subsystem nfulnl_subsys = {
 953         .name           = "log",
 954         .subsys_id      = NFNL_SUBSYS_ULOG,
 955         .cb_count       = NFULNL_MSG_MAX,
 956         .cb             = nfulnl_cb,
 957 };
 958
 959 #ifdef CONFIG_PROC_FS
 960 struct iter_state {
 961         struct seq_net_private p;
 962         unsigned int bucket;
 963 };
 964
 965 static struct hlist_node *get_first(struct net *net, struct iter_state *st)
 966 {
 967         struct nfnl_log_net *log;
 968         if (!st)
 969                 return NULL;
 970
 971         log = nfnl_log_pernet(net);
 972
 973         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
 974                 struct hlist_head *head = &log->instance_table[st->bucket];
 975
 976                 if (!hlist_empty(head))
 977                         return rcu_dereference_bh(hlist_first_rcu(head));
 978         }
 979         return NULL;
 980 }
 981
 982 static struct hlist_node *get_next(struct net *net, struct iter_state *st,
 983                                    struct hlist_node *h)
 984 {
 985         h = rcu_dereference_bh(hlist_next_rcu(h));
 986         while (!h) {
 987                 struct nfnl_log_net *log;
 988                 struct hlist_head *head;
 989
 990                 if (++st->bucket >= INSTANCE_BUCKETS)
 991                         return NULL;
 992
 993                 log = nfnl_log_pernet(net);
 994                 head = &log->instance_table[st->bucket];
 995                 h = rcu_dereference_bh(hlist_first_rcu(head));
 996         }
 997         return h;
 998 }
 999
1000 static struct hlist_node *get_idx(struct net *net, struct iter_state *st,
1001                                   loff_t pos)
1002 {
1003         struct hlist_node *head;
1004         head = get_first(net, st);
1005
1006         if (head)
1007                 while (pos && (head = get_next(net, st, head)))
1008                         pos--;
1009         return pos ? NULL : head;
1010 }
1011
1012 static void *seq_start(struct seq_file *s, loff_t *pos)
1013         __acquires(rcu_bh)
1014 {
1015         rcu_read_lock_bh();
1016         return get_idx(seq_file_net(s), s->private, *pos);
1017 }
1018
1019 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1020 {
1021         (*pos)++;
1022         return get_next(seq_file_net(s), s->private, v);
1023 }
1024
1025 static void seq_stop(struct seq_file *s, void *v)
1026         __releases(rcu_bh)
1027 {
1028         rcu_read_unlock_bh();
1029 }
1030
1031 static int seq_show(struct seq_file *s, void *v)
1032 {
1033         const struct nfulnl_instance *inst = v;
1034
1035         seq_printf(s, "%5u %6u %5u %1u %5u %6u %2u\n",
1036                    inst->group_num,
1037                    inst->peer_portid, inst->qlen,
1038                    inst->copy_mode, inst->copy_range,
1039                    inst->flushtimeout, refcount_read(&inst->use));
1040
1041         return 0;
1042 }
1043
1044 static const struct seq_operations nful_seq_ops = {
1045         .start  = seq_start,
1046         .next   = seq_next,
1047         .stop   = seq_stop,
1048         .show   = seq_show,
1049 };
1050
1051 static int nful_open(struct inode *inode, struct file *file)
1052 {
1053         return seq_open_net(inode, file, &nful_seq_ops,
1054                             sizeof(struct iter_state));
1055 }
1056
1057 static const struct file_operations nful_file_ops = {
1058         .owner   = THIS_MODULE,
1059         .open    = nful_open,
1060         .read    = seq_read,
1061         .llseek  = seq_lseek,
1062         .release = seq_release_net,
1063 };
1064
1065 #endif /* PROC_FS */
1066
1067 static int __net_init nfnl_log_net_init(struct net *net)
1068 {
1069         unsigned int i;
1070         struct nfnl_log_net *log = nfnl_log_pernet(net);
1071 #ifdef CONFIG_PROC_FS
1072         struct proc_dir_entry *proc;
1073         kuid_t root_uid;
1074         kgid_t root_gid;
1075 #endif
1076
1077         for (i = 0; i < INSTANCE_BUCKETS; i++)
1078                 INIT_HLIST_HEAD(&log->instance_table[i]);
1079         spin_lock_init(&log->instances_lock);
1080
1081 #ifdef CONFIG_PROC_FS
1082         proc = proc_create("nfnetlink_log", 0440,
1083                            net->nf.proc_netfilter, &nful_file_ops);
1084         if (!proc)
1085                 return -ENOMEM;
1086
1087         root_uid = make_kuid(net->user_ns, 0);
1088         root_gid = make_kgid(net->user_ns, 0);
1089         if (uid_valid(root_uid) && gid_valid(root_gid))
1090                 proc_set_user(proc, root_uid, root_gid);
1091 #endif
1092         return 0;
1093 }
1094
1095 static void __net_exit nfnl_log_net_exit(struct net *net)
1096 {
1097 #ifdef CONFIG_PROC_FS
1098         remove_proc_entry("nfnetlink_log", net->nf.proc_netfilter);
1099 #endif
1100         nf_log_unset(net, &nfulnl_logger);
1101 }
1102
1103 static struct pernet_operations nfnl_log_net_ops = {
1104         .init   = nfnl_log_net_init,
1105         .exit   = nfnl_log_net_exit,
1106         .id     = &nfnl_log_net_id,
1107         .size   = sizeof(struct nfnl_log_net),
1108 };
1109
1110 static int __init nfnetlink_log_init(void)
1111 {
1112         int status;
1113
1114         status = register_pernet_subsys(&nfnl_log_net_ops);
1115         if (status < 0) {
1116                 pr_err("failed to register pernet ops\n");
1117                 goto out;
1118         }
1119
1120         netlink_register_notifier(&nfulnl_rtnl_notifier);
1121         status = nfnetlink_subsys_register(&nfulnl_subsys);
1122         if (status < 0) {
1123                 pr_err("failed to create netlink socket\n");
1124                 goto cleanup_netlink_notifier;
1125         }
1126
1127         status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger);
1128         if (status < 0) {
1129                 pr_err("failed to register logger\n");
1130                 goto cleanup_subsys;
1131         }
1132
1133         return status;
1134
1135 cleanup_subsys:
1136         nfnetlink_subsys_unregister(&nfulnl_subsys);
1137 cleanup_netlink_notifier:
1138         netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1139         unregister_pernet_subsys(&nfnl_log_net_ops);
1140 out:
1141         return status;
1142 }
1143
1144 static void __exit nfnetlink_log_fini(void)
1145 {
1146         nfnetlink_subsys_unregister(&nfulnl_subsys);
1147         netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1148         unregister_pernet_subsys(&nfnl_log_net_ops);
1149         nf_log_unregister(&nfulnl_logger);
1150 }
1151
1152 MODULE_DESCRIPTION("netfilter userspace logging");
1153 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1154 MODULE_LICENSE("GPL");
1155 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG);
1156 MODULE_ALIAS_NF_LOGGER(AF_INET, 1);
1157 MODULE_ALIAS_NF_LOGGER(AF_INET6, 1);
1158 MODULE_ALIAS_NF_LOGGER(AF_BRIDGE, 1);
1159 MODULE_ALIAS_NF_LOGGER(3, 1); /* NFPROTO_ARP */
1160 MODULE_ALIAS_NF_LOGGER(5, 1); /* NFPROTO_NETDEV */
1161
1162 module_init(nfnetlink_log_init);
1163 module_exit(nfnetlink_log_fini);