net/netfilter/nfnetlink_log.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /*
   3  * This is a module which is used for logging packets to userspace via
   4  * nfetlink.
   5  *
   6  * (C) 2005 by Harald Welte <laforge@netfilter.org>
   7  * (C) 2006-2012 Patrick McHardy <kaber@trash.net>
   8  *
   9  * Based on the old ipv4-only ipt_ULOG.c:
  10  * (C) 2000-2004 by Harald Welte <laforge@netfilter.org>
  11  */
  12
  13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  14
  15 #include <linux/module.h>
  16 #include <linux/skbuff.h>
  17 #include <linux/if_arp.h>
  18 #include <linux/init.h>
  19 #include <linux/ip.h>
  20 #include <linux/ipv6.h>
  21 #include <linux/netdevice.h>
  22 #include <linux/netfilter.h>
  23 #include <linux/netfilter_bridge.h>
  24 #include <net/netlink.h>
  25 #include <linux/netfilter/nfnetlink.h>
  26 #include <linux/netfilter/nfnetlink_log.h>
  27 #include <linux/netfilter/nf_conntrack_common.h>
  28 #include <linux/spinlock.h>
  29 #include <linux/sysctl.h>
  30 #include <linux/proc_fs.h>
  31 #include <linux/security.h>
  32 #include <linux/list.h>
  33 #include <linux/slab.h>
  34 #include <net/sock.h>
  35 #include <net/netfilter/nf_log.h>
  36 #include <net/netns/generic.h>
  37
  38 #include <linux/atomic.h>
  39 #include <linux/refcount.h>
  40
  41
  42 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  43 #include "../bridge/br_private.h"
  44 #endif
  45
  46 #define NFULNL_COPY_DISABLED    0xff
  47 #define NFULNL_NLBUFSIZ_DEFAULT NLMSG_GOODSIZE
  48 #define NFULNL_TIMEOUT_DEFAULT  100     /* every second */
  49 #define NFULNL_QTHRESH_DEFAULT  100     /* 100 packets */
  50 /* max packet size is limited by 16-bit struct nfattr nfa_len field */
  51 #define NFULNL_COPY_RANGE_MAX   (0xFFFF - NLA_HDRLEN)
  52
  53 #define PRINTR(x, args...)      do { if (net_ratelimit()) \
  54                                      printk(x, ## args); } while (0);
  55
  56 struct nfulnl_instance {
  57         struct hlist_node hlist;        /* global list of instances */
  58         spinlock_t lock;
  59         refcount_t use;                 /* use count */
  60
  61         unsigned int qlen;              /* number of nlmsgs in skb */
  62         struct sk_buff *skb;            /* pre-allocatd skb */
  63         struct timer_list timer;
  64         struct net *net;
  65         struct user_namespace *peer_user_ns;    /* User namespace of the peer process */
  66         u32 peer_portid;                /* PORTID of the peer process */
  67
  68         /* configurable parameters */
  69         unsigned int flushtimeout;      /* timeout until queue flush */
  70         unsigned int nlbufsiz;          /* netlink buffer allocation size */
  71         unsigned int qthreshold;        /* threshold of the queue */
  72         u_int32_t copy_range;
  73         u_int32_t seq;                  /* instance-local sequential counter */
  74         u_int16_t group_num;            /* number of this queue */
  75         u_int16_t flags;
  76         u_int8_t copy_mode;
  77         struct rcu_head rcu;
  78 };
  79
  80 #define INSTANCE_BUCKETS        16
  81
  82 static unsigned int nfnl_log_net_id __read_mostly;
  83
  84 struct nfnl_log_net {
  85         spinlock_t instances_lock;
  86         struct hlist_head instance_table[INSTANCE_BUCKETS];
  87         atomic_t global_seq;
  88 };
  89
  90 static struct nfnl_log_net *nfnl_log_pernet(struct net *net)
  91 {
  92         return net_generic(net, nfnl_log_net_id);
  93 }
  94
  95 static inline u_int8_t instance_hashfn(u_int16_t group_num)
  96 {
  97         return ((group_num & 0xff) % INSTANCE_BUCKETS);
  98 }
  99
 100 static struct nfulnl_instance *
 101 __instance_lookup(struct nfnl_log_net *log, u_int16_t group_num)
 102 {
 103         struct hlist_head *head;
 104         struct nfulnl_instance *inst;
 105
 106         head = &log->instance_table[instance_hashfn(group_num)];
 107         hlist_for_each_entry_rcu(inst, head, hlist) {
 108                 if (inst->group_num == group_num)
 109                         return inst;
 110         }
 111         return NULL;
 112 }
 113
 114 static inline void
 115 instance_get(struct nfulnl_instance *inst)
 116 {
 117         refcount_inc(&inst->use);
 118 }
 119
 120 static struct nfulnl_instance *
 121 instance_lookup_get(struct nfnl_log_net *log, u_int16_t group_num)
 122 {
 123         struct nfulnl_instance *inst;
 124
 125         rcu_read_lock_bh();
 126         inst = __instance_lookup(log, group_num);
 127         if (inst && !refcount_inc_not_zero(&inst->use))
 128                 inst = NULL;
 129         rcu_read_unlock_bh();
 130
 131         return inst;
 132 }
 133
 134 static void nfulnl_instance_free_rcu(struct rcu_head *head)
 135 {
 136         struct nfulnl_instance *inst =
 137                 container_of(head, struct nfulnl_instance, rcu);
 138
 139         put_net(inst->net);
 140         kfree(inst);
 141         module_put(THIS_MODULE);
 142 }
 143
 144 static void
 145 instance_put(struct nfulnl_instance *inst)
 146 {
 147         if (inst && refcount_dec_and_test(&inst->use))
 148                 call_rcu(&inst->rcu, nfulnl_instance_free_rcu);
 149 }
 150
 151 static void nfulnl_timer(struct timer_list *t);
 152
 153 static struct nfulnl_instance *
 154 instance_create(struct net *net, u_int16_t group_num,
 155                 u32 portid, struct user_namespace *user_ns)
 156 {
 157         struct nfulnl_instance *inst;
 158         struct nfnl_log_net *log = nfnl_log_pernet(net);
 159         int err;
 160
 161         spin_lock_bh(&log->instances_lock);
 162         if (__instance_lookup(log, group_num)) {
 163                 err = -EEXIST;
 164                 goto out_unlock;
 165         }
 166
 167         inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
 168         if (!inst) {
 169                 err = -ENOMEM;
 170                 goto out_unlock;
 171         }
 172
 173         if (!try_module_get(THIS_MODULE)) {
 174                 kfree(inst);
 175                 err = -EAGAIN;
 176                 goto out_unlock;
 177         }
 178
 179         INIT_HLIST_NODE(&inst->hlist);
 180         spin_lock_init(&inst->lock);
 181         /* needs to be two, since we _put() after creation */
 182         refcount_set(&inst->use, 2);
 183
 184         timer_setup(&inst->timer, nfulnl_timer, 0);
 185
 186         inst->net = get_net(net);
 187         inst->peer_user_ns = user_ns;
 188         inst->peer_portid = portid;
 189         inst->group_num = group_num;
 190
 191         inst->qthreshold        = NFULNL_QTHRESH_DEFAULT;
 192         inst->flushtimeout      = NFULNL_TIMEOUT_DEFAULT;
 193         inst->nlbufsiz          = NFULNL_NLBUFSIZ_DEFAULT;
 194         inst->copy_mode         = NFULNL_COPY_PACKET;
 195         inst->copy_range        = NFULNL_COPY_RANGE_MAX;
 196
 197         hlist_add_head_rcu(&inst->hlist,
 198                        &log->instance_table[instance_hashfn(group_num)]);
 199
 200
 201         spin_unlock_bh(&log->instances_lock);
 202
 203         return inst;
 204
 205 out_unlock:
 206         spin_unlock_bh(&log->instances_lock);
 207         return ERR_PTR(err);
 208 }
 209
 210 static void __nfulnl_flush(struct nfulnl_instance *inst);
 211
 212 /* called with BH disabled */
 213 static void
 214 __instance_destroy(struct nfulnl_instance *inst)
 215 {
 216         /* first pull it out of the global list */
 217         hlist_del_rcu(&inst->hlist);
 218
 219         /* then flush all pending packets from skb */
 220
 221         spin_lock(&inst->lock);
 222
 223         /* lockless readers wont be able to use us */
 224         inst->copy_mode = NFULNL_COPY_DISABLED;
 225
 226         if (inst->skb)
 227                 __nfulnl_flush(inst);
 228         spin_unlock(&inst->lock);
 229
 230         /* and finally put the refcount */
 231         instance_put(inst);
 232 }
 233
 234 static inline void
 235 instance_destroy(struct nfnl_log_net *log,
 236                  struct nfulnl_instance *inst)
 237 {
 238         spin_lock_bh(&log->instances_lock);
 239         __instance_destroy(inst);
 240         spin_unlock_bh(&log->instances_lock);
 241 }
 242
 243 static int
 244 nfulnl_set_mode(struct nfulnl_instance *inst, u_int8_t mode,
 245                   unsigned int range)
 246 {
 247         int status = 0;
 248
 249         spin_lock_bh(&inst->lock);
 250
 251         switch (mode) {
 252         case NFULNL_COPY_NONE:
 253         case NFULNL_COPY_META:
 254                 inst->copy_mode = mode;
 255                 inst->copy_range = 0;
 256                 break;
 257
 258         case NFULNL_COPY_PACKET:
 259                 inst->copy_mode = mode;
 260                 if (range == 0)
 261                         range = NFULNL_COPY_RANGE_MAX;
 262                 inst->copy_range = min_t(unsigned int,
 263                                          range, NFULNL_COPY_RANGE_MAX);
 264                 break;
 265
 266         default:
 267                 status = -EINVAL;
 268                 break;
 269         }
 270
 271         spin_unlock_bh(&inst->lock);
 272
 273         return status;
 274 }
 275
 276 static int
 277 nfulnl_set_nlbufsiz(struct nfulnl_instance *inst, u_int32_t nlbufsiz)
 278 {
 279         int status;
 280
 281         spin_lock_bh(&inst->lock);
 282         if (nlbufsiz < NFULNL_NLBUFSIZ_DEFAULT)
 283                 status = -ERANGE;
 284         else if (nlbufsiz > 131072)
 285                 status = -ERANGE;
 286         else {
 287                 inst->nlbufsiz = nlbufsiz;
 288                 status = 0;
 289         }
 290         spin_unlock_bh(&inst->lock);
 291
 292         return status;
 293 }
 294
 295 static void
 296 nfulnl_set_timeout(struct nfulnl_instance *inst, u_int32_t timeout)
 297 {
 298         spin_lock_bh(&inst->lock);
 299         inst->flushtimeout = timeout;
 300         spin_unlock_bh(&inst->lock);
 301 }
 302
 303 static void
 304 nfulnl_set_qthresh(struct nfulnl_instance *inst, u_int32_t qthresh)
 305 {
 306         spin_lock_bh(&inst->lock);
 307         inst->qthreshold = qthresh;
 308         spin_unlock_bh(&inst->lock);
 309 }
 310
 311 static int
 312 nfulnl_set_flags(struct nfulnl_instance *inst, u_int16_t flags)
 313 {
 314         spin_lock_bh(&inst->lock);
 315         inst->flags = flags;
 316         spin_unlock_bh(&inst->lock);
 317
 318         return 0;
 319 }
 320
 321 static struct sk_buff *
 322 nfulnl_alloc_skb(struct net *net, u32 peer_portid, unsigned int inst_size,
 323                  unsigned int pkt_size)
 324 {
 325         struct sk_buff *skb;
 326         unsigned int n;
 327
 328         /* alloc skb which should be big enough for a whole multipart
 329          * message.  WARNING: has to be <= 128k due to slab restrictions */
 330
 331         n = max(inst_size, pkt_size);
 332         skb = alloc_skb(n, GFP_ATOMIC | __GFP_NOWARN);
 333         if (!skb) {
 334                 if (n > pkt_size) {
 335                         /* try to allocate only as much as we need for current
 336                          * packet */
 337
 338                         skb = alloc_skb(pkt_size, GFP_ATOMIC);
 339                 }
 340         }
 341
 342         return skb;
 343 }
 344
 345 static void
 346 __nfulnl_send(struct nfulnl_instance *inst)
 347 {
 348         if (inst->qlen > 1) {
 349                 struct nlmsghdr *nlh = nlmsg_put(inst->skb, 0, 0,
 350                                                  NLMSG_DONE,
 351                                                  sizeof(struct nfgenmsg),
 352                                                  0);
 353                 if (WARN_ONCE(!nlh, "bad nlskb size: %u, tailroom %d\n",
 354                               inst->skb->len, skb_tailroom(inst->skb))) {
 355                         kfree_skb(inst->skb);
 356                         goto out;
 357                 }
 358         }
 359         nfnetlink_unicast(inst->skb, inst->net, inst->peer_portid);
 360 out:
 361         inst->qlen = 0;
 362         inst->skb = NULL;
 363 }
 364
 365 static void
 366 __nfulnl_flush(struct nfulnl_instance *inst)
 367 {
 368         /* timer holds a reference */
 369         if (del_timer(&inst->timer))
 370                 instance_put(inst);
 371         if (inst->skb)
 372                 __nfulnl_send(inst);
 373 }
 374
 375 static void
 376 nfulnl_timer(struct timer_list *t)
 377 {
 378         struct nfulnl_instance *inst = from_timer(inst, t, timer);
 379
 380         spin_lock_bh(&inst->lock);
 381         if (inst->skb)
 382                 __nfulnl_send(inst);
 383         spin_unlock_bh(&inst->lock);
 384         instance_put(inst);
 385 }
 386
 387 static u32 nfulnl_get_bridge_size(const struct sk_buff *skb)
 388 {
 389         u32 size = 0;
 390
 391         if (!skb_mac_header_was_set(skb))
 392                 return 0;
 393
 394         if (skb_vlan_tag_present(skb)) {
 395                 size += nla_total_size(0); /* nested */
 396                 size += nla_total_size(sizeof(u16)); /* id */
 397                 size += nla_total_size(sizeof(u16)); /* tag */
 398         }
 399
 400         if (skb->network_header > skb->mac_header)
 401                 size += nla_total_size(skb->network_header - skb->mac_header);
 402
 403         return size;
 404 }
 405
 406 static int nfulnl_put_bridge(struct nfulnl_instance *inst, const struct sk_buff *skb)
 407 {
 408         if (!skb_mac_header_was_set(skb))
 409                 return 0;
 410
 411         if (skb_vlan_tag_present(skb)) {
 412                 struct nlattr *nest;
 413
 414                 nest = nla_nest_start(inst->skb, NFULA_VLAN);
 415                 if (!nest)
 416                         goto nla_put_failure;
 417
 418                 if (nla_put_be16(inst->skb, NFULA_VLAN_TCI, htons(skb->vlan_tci)) ||
 419                     nla_put_be16(inst->skb, NFULA_VLAN_PROTO, skb->vlan_proto))
 420                         goto nla_put_failure;
 421
 422                 nla_nest_end(inst->skb, nest);
 423         }
 424
 425         if (skb->mac_header < skb->network_header) {
 426                 int len = (int)(skb->network_header - skb->mac_header);
 427
 428                 if (nla_put(inst->skb, NFULA_L2HDR, len, skb_mac_header(skb)))
 429                         goto nla_put_failure;
 430         }
 431
 432         return 0;
 433
 434 nla_put_failure:
 435         return -1;
 436 }
 437
 438 /* This is an inline function, we don't really care about a long
 439  * list of arguments */
 440 static inline int
 441 __build_packet_message(struct nfnl_log_net *log,
 442                         struct nfulnl_instance *inst,
 443                         const struct sk_buff *skb,
 444                         unsigned int data_len,
 445                         u_int8_t pf,
 446                         unsigned int hooknum,
 447                         const struct net_device *indev,
 448                         const struct net_device *outdev,
 449                         const char *prefix, unsigned int plen,
 450                         const struct nfnl_ct_hook *nfnl_ct,
 451                         struct nf_conn *ct, enum ip_conntrack_info ctinfo)
 452 {
 453         struct nfulnl_msg_packet_hdr pmsg;
 454         struct nlmsghdr *nlh;
 455         sk_buff_data_t old_tail = inst->skb->tail;
 456         struct sock *sk;
 457         const unsigned char *hwhdrp;
 458
 459         nlh = nfnl_msg_put(inst->skb, 0, 0,
 460                            nfnl_msg_type(NFNL_SUBSYS_ULOG, NFULNL_MSG_PACKET),
 461                            0, pf, NFNETLINK_V0, htons(inst->group_num));
 462         if (!nlh)
 463                 return -1;
 464
 465         memset(&pmsg, 0, sizeof(pmsg));
 466         pmsg.hw_protocol        = skb->protocol;
 467         pmsg.hook               = hooknum;
 468
 469         if (nla_put(inst->skb, NFULA_PACKET_HDR, sizeof(pmsg), &pmsg))
 470                 goto nla_put_failure;
 471
 472         if (prefix &&
 473             nla_put(inst->skb, NFULA_PREFIX, plen, prefix))
 474                 goto nla_put_failure;
 475
 476         if (indev) {
 477 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 478                 if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 479                                  htonl(indev->ifindex)))
 480                         goto nla_put_failure;
 481 #else
 482                 if (pf == PF_BRIDGE) {
 483                         /* Case 1: outdev is physical input device, we need to
 484                          * look for bridge group (when called from
 485                          * netfilter_bridge) */
 486                         if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
 487                                          htonl(indev->ifindex)) ||
 488                         /* this is the bridge group "brX" */
 489                         /* rcu_read_lock()ed by nf_hook_thresh or
 490                          * nf_log_packet.
 491                          */
 492                             nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 493                                          htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
 494                                 goto nla_put_failure;
 495                 } else {
 496                         struct net_device *physindev;
 497
 498                         /* Case 2: indev is bridge group, we need to look for
 499                          * physical device (when called from ipv4) */
 500                         if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 501                                          htonl(indev->ifindex)))
 502                                 goto nla_put_failure;
 503
 504                         physindev = nf_bridge_get_physindev(skb);
 505                         if (physindev &&
 506                             nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
 507                                          htonl(physindev->ifindex)))
 508                                 goto nla_put_failure;
 509                 }
 510 #endif
 511         }
 512
 513         if (outdev) {
 514 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 515                 if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 516                                  htonl(outdev->ifindex)))
 517                         goto nla_put_failure;
 518 #else
 519                 if (pf == PF_BRIDGE) {
 520                         /* Case 1: outdev is physical output device, we need to
 521                          * look for bridge group (when called from
 522                          * netfilter_bridge) */
 523                         if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
 524                                          htonl(outdev->ifindex)) ||
 525                         /* this is the bridge group "brX" */
 526                         /* rcu_read_lock()ed by nf_hook_thresh or
 527                          * nf_log_packet.
 528                          */
 529                             nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 530                                          htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
 531                                 goto nla_put_failure;
 532                 } else {
 533                         struct net_device *physoutdev;
 534
 535                         /* Case 2: indev is a bridge group, we need to look
 536                          * for physical device (when called from ipv4) */
 537                         if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 538                                          htonl(outdev->ifindex)))
 539                                 goto nla_put_failure;
 540
 541                         physoutdev = nf_bridge_get_physoutdev(skb);
 542                         if (physoutdev &&
 543                             nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
 544                                          htonl(physoutdev->ifindex)))
 545                                 goto nla_put_failure;
 546                 }
 547 #endif
 548         }
 549
 550         if (skb->mark &&
 551             nla_put_be32(inst->skb, NFULA_MARK, htonl(skb->mark)))
 552                 goto nla_put_failure;
 553
 554         if (indev && skb->dev &&
 555             skb_mac_header_was_set(skb) &&
 556             skb_mac_header_len(skb) != 0) {
 557                 struct nfulnl_msg_packet_hw phw;
 558                 int len;
 559
 560                 memset(&phw, 0, sizeof(phw));
 561                 len = dev_parse_header(skb, phw.hw_addr);
 562                 if (len > 0) {
 563                         phw.hw_addrlen = htons(len);
 564                         if (nla_put(inst->skb, NFULA_HWADDR, sizeof(phw), &phw))
 565                                 goto nla_put_failure;
 566                 }
 567         }
 568
 569         if (indev && skb_mac_header_was_set(skb)) {
 570                 if (nla_put_be16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type)) ||
 571                     nla_put_be16(inst->skb, NFULA_HWLEN,
 572                                  htons(skb->dev->hard_header_len)))
 573                         goto nla_put_failure;
 574
 575                 hwhdrp = skb_mac_header(skb);
 576
 577                 if (skb->dev->type == ARPHRD_SIT)
 578                         hwhdrp -= ETH_HLEN;
 579
 580                 if (hwhdrp >= skb->head &&
 581                     nla_put(inst->skb, NFULA_HWHEADER,
 582                             skb->dev->hard_header_len, hwhdrp))
 583                         goto nla_put_failure;
 584         }
 585
 586         if (hooknum <= NF_INET_FORWARD && skb->tstamp) {
 587                 struct nfulnl_msg_packet_timestamp ts;
 588                 struct timespec64 kts = ktime_to_timespec64(skb->tstamp);
 589                 ts.sec = cpu_to_be64(kts.tv_sec);
 590                 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
 591
 592                 if (nla_put(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts))
 593                         goto nla_put_failure;
 594         }
 595
 596         /* UID */
 597         sk = skb->sk;
 598         if (sk && sk_fullsock(sk)) {
 599                 read_lock_bh(&sk->sk_callback_lock);
 600                 if (sk->sk_socket && sk->sk_socket->file) {
 601                         struct file *file = sk->sk_socket->file;
 602                         const struct cred *cred = file->f_cred;
 603                         struct user_namespace *user_ns = inst->peer_user_ns;
 604                         __be32 uid = htonl(from_kuid_munged(user_ns, cred->fsuid));
 605                         __be32 gid = htonl(from_kgid_munged(user_ns, cred->fsgid));
 606                         read_unlock_bh(&sk->sk_callback_lock);
 607                         if (nla_put_be32(inst->skb, NFULA_UID, uid) ||
 608                             nla_put_be32(inst->skb, NFULA_GID, gid))
 609                                 goto nla_put_failure;
 610                 } else
 611                         read_unlock_bh(&sk->sk_callback_lock);
 612         }
 613
 614         /* local sequence number */
 615         if ((inst->flags & NFULNL_CFG_F_SEQ) &&
 616             nla_put_be32(inst->skb, NFULA_SEQ, htonl(inst->seq++)))
 617                 goto nla_put_failure;
 618
 619         /* global sequence number */
 620         if ((inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) &&
 621             nla_put_be32(inst->skb, NFULA_SEQ_GLOBAL,
 622                          htonl(atomic_inc_return(&log->global_seq))))
 623                 goto nla_put_failure;
 624
 625         if (ct && nfnl_ct->build(inst->skb, ct, ctinfo,
 626                                  NFULA_CT, NFULA_CT_INFO) < 0)
 627                 goto nla_put_failure;
 628
 629         if ((pf == NFPROTO_NETDEV || pf == NFPROTO_BRIDGE) &&
 630             nfulnl_put_bridge(inst, skb) < 0)
 631                 goto nla_put_failure;
 632
 633         if (data_len) {
 634                 struct nlattr *nla;
 635                 int size = nla_attr_size(data_len);
 636
 637                 if (skb_tailroom(inst->skb) < nla_total_size(data_len))
 638                         goto nla_put_failure;
 639
 640                 nla = skb_put(inst->skb, nla_total_size(data_len));
 641                 nla->nla_type = NFULA_PAYLOAD;
 642                 nla->nla_len = size;
 643
 644                 if (skb_copy_bits(skb, 0, nla_data(nla), data_len))
 645                         BUG();
 646         }
 647
 648         nlh->nlmsg_len = inst->skb->tail - old_tail;
 649         return 0;
 650
 651 nla_put_failure:
 652         PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n");
 653         return -1;
 654 }
 655
 656 static const struct nf_loginfo default_loginfo = {
 657         .type =         NF_LOG_TYPE_ULOG,
 658         .u = {
 659                 .ulog = {
 660                         .copy_len       = 0xffff,
 661                         .group          = 0,
 662                         .qthreshold     = 1,
 663                 },
 664         },
 665 };
 666
 667 /* log handler for internal netfilter logging api */
 668 static void
 669 nfulnl_log_packet(struct net *net,
 670                   u_int8_t pf,
 671                   unsigned int hooknum,
 672                   const struct sk_buff *skb,
 673                   const struct net_device *in,
 674                   const struct net_device *out,
 675                   const struct nf_loginfo *li_user,
 676                   const char *prefix)
 677 {
 678         size_t size;
 679         unsigned int data_len;
 680         struct nfulnl_instance *inst;
 681         const struct nf_loginfo *li;
 682         unsigned int qthreshold;
 683         unsigned int plen = 0;
 684         struct nfnl_log_net *log = nfnl_log_pernet(net);
 685         const struct nfnl_ct_hook *nfnl_ct = NULL;
 686         struct nf_conn *ct = NULL;
 687         enum ip_conntrack_info ctinfo;
 688
 689         if (li_user && li_user->type == NF_LOG_TYPE_ULOG)
 690                 li = li_user;
 691         else
 692                 li = &default_loginfo;
 693
 694         inst = instance_lookup_get(log, li->u.ulog.group);
 695         if (!inst)
 696                 return;
 697
 698         if (prefix)
 699                 plen = strlen(prefix) + 1;
 700
 701         /* FIXME: do we want to make the size calculation conditional based on
 702          * what is actually present?  way more branches and checks, but more
 703          * memory efficient... */
 704         size = nlmsg_total_size(sizeof(struct nfgenmsg))
 705                 + nla_total_size(sizeof(struct nfulnl_msg_packet_hdr))
 706                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 707                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 708 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 709                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 710                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 711 #endif
 712                 + nla_total_size(sizeof(u_int32_t))     /* mark */
 713                 + nla_total_size(sizeof(u_int32_t))     /* uid */
 714                 + nla_total_size(sizeof(u_int32_t))     /* gid */
 715                 + nla_total_size(plen)                  /* prefix */
 716                 + nla_total_size(sizeof(struct nfulnl_msg_packet_hw))
 717                 + nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp))
 718                 + nla_total_size(sizeof(struct nfgenmsg));      /* NLMSG_DONE */
 719
 720         if (in && skb_mac_header_was_set(skb)) {
 721                 size += nla_total_size(skb->dev->hard_header_len)
 722                         + nla_total_size(sizeof(u_int16_t))     /* hwtype */
 723                         + nla_total_size(sizeof(u_int16_t));    /* hwlen */
 724         }
 725
 726         spin_lock_bh(&inst->lock);
 727
 728         if (inst->flags & NFULNL_CFG_F_SEQ)
 729                 size += nla_total_size(sizeof(u_int32_t));
 730         if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
 731                 size += nla_total_size(sizeof(u_int32_t));
 732         if (inst->flags & NFULNL_CFG_F_CONNTRACK) {
 733                 nfnl_ct = rcu_dereference(nfnl_ct_hook);
 734                 if (nfnl_ct != NULL) {
 735                         ct = nfnl_ct->get_ct(skb, &ctinfo);
 736                         if (ct != NULL)
 737                                 size += nfnl_ct->build_size(ct);
 738                 }
 739         }
 740         if (pf == NFPROTO_NETDEV || pf == NFPROTO_BRIDGE)
 741                 size += nfulnl_get_bridge_size(skb);
 742
 743         qthreshold = inst->qthreshold;
 744         /* per-rule qthreshold overrides per-instance */
 745         if (li->u.ulog.qthreshold)
 746                 if (qthreshold > li->u.ulog.qthreshold)
 747                         qthreshold = li->u.ulog.qthreshold;
 748
 749
 750         switch (inst->copy_mode) {
 751         case NFULNL_COPY_META:
 752         case NFULNL_COPY_NONE:
 753                 data_len = 0;
 754                 break;
 755
 756         case NFULNL_COPY_PACKET:
 757                 data_len = inst->copy_range;
 758                 if ((li->u.ulog.flags & NF_LOG_F_COPY_LEN) &&
 759                     (li->u.ulog.copy_len < data_len))
 760                         data_len = li->u.ulog.copy_len;
 761
 762                 if (data_len > skb->len)
 763                         data_len = skb->len;
 764
 765                 size += nla_total_size(data_len);
 766                 break;
 767
 768         case NFULNL_COPY_DISABLED:
 769         default:
 770                 goto unlock_and_release;
 771         }
 772
 773         if (inst->skb && size > skb_tailroom(inst->skb)) {
 774                 /* either the queue len is too high or we don't have
 775                  * enough room in the skb left. flush to userspace. */
 776                 __nfulnl_flush(inst);
 777         }
 778
 779         if (!inst->skb) {
 780                 inst->skb = nfulnl_alloc_skb(net, inst->peer_portid,
 781                                              inst->nlbufsiz, size);
 782                 if (!inst->skb)
 783                         goto alloc_failure;
 784         }
 785
 786         inst->qlen++;
 787
 788         __build_packet_message(log, inst, skb, data_len, pf,
 789                                 hooknum, in, out, prefix, plen,
 790                                 nfnl_ct, ct, ctinfo);
 791
 792         if (inst->qlen >= qthreshold)
 793                 __nfulnl_flush(inst);
 794         /* timer_pending always called within inst->lock, so there
 795          * is no chance of a race here */
 796         else if (!timer_pending(&inst->timer)) {
 797                 instance_get(inst);
 798                 inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100);
 799                 add_timer(&inst->timer);
 800         }
 801
 802 unlock_and_release:
 803         spin_unlock_bh(&inst->lock);
 804         instance_put(inst);
 805         return;
 806
 807 alloc_failure:
 808         /* FIXME: statistics */
 809         goto unlock_and_release;
 810 }
 811
 812 static int
 813 nfulnl_rcv_nl_event(struct notifier_block *this,
 814                    unsigned long event, void *ptr)
 815 {
 816         struct netlink_notify *n = ptr;
 817         struct nfnl_log_net *log = nfnl_log_pernet(n->net);
 818
 819         if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
 820                 int i;
 821
 822                 /* destroy all instances for this portid */
 823                 spin_lock_bh(&log->instances_lock);
 824                 for  (i = 0; i < INSTANCE_BUCKETS; i++) {
 825                         struct hlist_node *t2;
 826                         struct nfulnl_instance *inst;
 827                         struct hlist_head *head = &log->instance_table[i];
 828
 829                         hlist_for_each_entry_safe(inst, t2, head, hlist) {
 830                                 if (n->portid == inst->peer_portid)
 831                                         __instance_destroy(inst);
 832                         }
 833                 }
 834                 spin_unlock_bh(&log->instances_lock);
 835         }
 836         return NOTIFY_DONE;
 837 }
 838
 839 static struct notifier_block nfulnl_rtnl_notifier = {
 840         .notifier_call  = nfulnl_rcv_nl_event,
 841 };
 842
 843 static int nfulnl_recv_unsupp(struct net *net, struct sock *ctnl,
 844                               struct sk_buff *skb, const struct nlmsghdr *nlh,
 845                               const struct nlattr * const nfqa[],
 846                               struct netlink_ext_ack *extack)
 847 {
 848         return -ENOTSUPP;
 849 }
 850
 851 static struct nf_logger nfulnl_logger __read_mostly = {
 852         .name   = "nfnetlink_log",
 853         .type   = NF_LOG_TYPE_ULOG,
 854         .logfn  = nfulnl_log_packet,
 855         .me     = THIS_MODULE,
 856 };
 857
 858 static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = {
 859         [NFULA_CFG_CMD]         = { .len = sizeof(struct nfulnl_msg_config_cmd) },
 860         [NFULA_CFG_MODE]        = { .len = sizeof(struct nfulnl_msg_config_mode) },
 861         [NFULA_CFG_TIMEOUT]     = { .type = NLA_U32 },
 862         [NFULA_CFG_QTHRESH]     = { .type = NLA_U32 },
 863         [NFULA_CFG_NLBUFSIZ]    = { .type = NLA_U32 },
 864         [NFULA_CFG_FLAGS]       = { .type = NLA_U16 },
 865 };
 866
 867 static int nfulnl_recv_config(struct net *net, struct sock *ctnl,
 868                               struct sk_buff *skb, const struct nlmsghdr *nlh,
 869                               const struct nlattr * const nfula[],
 870                               struct netlink_ext_ack *extack)
 871 {
 872         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
 873         u_int16_t group_num = ntohs(nfmsg->res_id);
 874         struct nfulnl_instance *inst;
 875         struct nfulnl_msg_config_cmd *cmd = NULL;
 876         struct nfnl_log_net *log = nfnl_log_pernet(net);
 877         int ret = 0;
 878         u16 flags = 0;
 879
 880         if (nfula[NFULA_CFG_CMD]) {
 881                 u_int8_t pf = nfmsg->nfgen_family;
 882                 cmd = nla_data(nfula[NFULA_CFG_CMD]);
 883
 884                 /* Commands without queue context */
 885                 switch (cmd->command) {
 886                 case NFULNL_CFG_CMD_PF_BIND:
 887                         return nf_log_bind_pf(net, pf, &nfulnl_logger);
 888                 case NFULNL_CFG_CMD_PF_UNBIND:
 889                         nf_log_unbind_pf(net, pf);
 890                         return 0;
 891                 }
 892         }
 893
 894         inst = instance_lookup_get(log, group_num);
 895         if (inst && inst->peer_portid != NETLINK_CB(skb).portid) {
 896                 ret = -EPERM;
 897                 goto out_put;
 898         }
 899
 900         /* Check if we support these flags in first place, dependencies should
 901          * be there too not to break atomicity.
 902          */
 903         if (nfula[NFULA_CFG_FLAGS]) {
 904                 flags = ntohs(nla_get_be16(nfula[NFULA_CFG_FLAGS]));
 905
 906                 if ((flags & NFULNL_CFG_F_CONNTRACK) &&
 907                     !rcu_access_pointer(nfnl_ct_hook)) {
 908 #ifdef CONFIG_MODULES
 909                         nfnl_unlock(NFNL_SUBSYS_ULOG);
 910                         request_module("ip_conntrack_netlink");
 911                         nfnl_lock(NFNL_SUBSYS_ULOG);
 912                         if (rcu_access_pointer(nfnl_ct_hook)) {
 913                                 ret = -EAGAIN;
 914                                 goto out_put;
 915                         }
 916 #endif
 917                         ret = -EOPNOTSUPP;
 918                         goto out_put;
 919                 }
 920         }
 921
 922         if (cmd != NULL) {
 923                 switch (cmd->command) {
 924                 case NFULNL_CFG_CMD_BIND:
 925                         if (inst) {
 926                                 ret = -EBUSY;
 927                                 goto out_put;
 928                         }
 929
 930                         inst = instance_create(net, group_num,
 931                                                NETLINK_CB(skb).portid,
 932                                                sk_user_ns(NETLINK_CB(skb).sk));
 933                         if (IS_ERR(inst)) {
 934                                 ret = PTR_ERR(inst);
 935                                 goto out;
 936                         }
 937                         break;
 938                 case NFULNL_CFG_CMD_UNBIND:
 939                         if (!inst) {
 940                                 ret = -ENODEV;
 941                                 goto out;
 942                         }
 943
 944                         instance_destroy(log, inst);
 945                         goto out_put;
 946                 default:
 947                         ret = -ENOTSUPP;
 948                         goto out_put;
 949                 }
 950         } else if (!inst) {
 951                 ret = -ENODEV;
 952                 goto out;
 953         }
 954
 955         if (nfula[NFULA_CFG_MODE]) {
 956                 struct nfulnl_msg_config_mode *params =
 957                         nla_data(nfula[NFULA_CFG_MODE]);
 958
 959                 nfulnl_set_mode(inst, params->copy_mode,
 960                                 ntohl(params->copy_range));
 961         }
 962
 963         if (nfula[NFULA_CFG_TIMEOUT]) {
 964                 __be32 timeout = nla_get_be32(nfula[NFULA_CFG_TIMEOUT]);
 965
 966                 nfulnl_set_timeout(inst, ntohl(timeout));
 967         }
 968
 969         if (nfula[NFULA_CFG_NLBUFSIZ]) {
 970                 __be32 nlbufsiz = nla_get_be32(nfula[NFULA_CFG_NLBUFSIZ]);
 971
 972                 nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz));
 973         }
 974
 975         if (nfula[NFULA_CFG_QTHRESH]) {
 976                 __be32 qthresh = nla_get_be32(nfula[NFULA_CFG_QTHRESH]);
 977
 978                 nfulnl_set_qthresh(inst, ntohl(qthresh));
 979         }
 980
 981         if (nfula[NFULA_CFG_FLAGS])
 982                 nfulnl_set_flags(inst, flags);
 983
 984 out_put:
 985         instance_put(inst);
 986 out:
 987         return ret;
 988 }
 989
 990 static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = {
 991         [NFULNL_MSG_PACKET]     = { .call = nfulnl_recv_unsupp,
 992                                     .attr_count = NFULA_MAX, },
 993         [NFULNL_MSG_CONFIG]     = { .call = nfulnl_recv_config,
 994                                     .attr_count = NFULA_CFG_MAX,
 995                                     .policy = nfula_cfg_policy },
 996 };
 997
 998 static const struct nfnetlink_subsystem nfulnl_subsys = {
 999         .name           = "log",
1000         .subsys_id      = NFNL_SUBSYS_ULOG,
1001         .cb_count       = NFULNL_MSG_MAX,
1002         .cb             = nfulnl_cb,
1003 };
1004
1005 #ifdef CONFIG_PROC_FS
1006 struct iter_state {
1007         struct seq_net_private p;
1008         unsigned int bucket;
1009 };
1010
1011 static struct hlist_node *get_first(struct net *net, struct iter_state *st)
1012 {
1013         struct nfnl_log_net *log;
1014         if (!st)
1015                 return NULL;
1016
1017         log = nfnl_log_pernet(net);
1018
1019         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1020                 struct hlist_head *head = &log->instance_table[st->bucket];
1021
1022                 if (!hlist_empty(head))
1023                         return rcu_dereference_bh(hlist_first_rcu(head));
1024         }
1025         return NULL;
1026 }
1027
1028 static struct hlist_node *get_next(struct net *net, struct iter_state *st,
1029                                    struct hlist_node *h)
1030 {
1031         h = rcu_dereference_bh(hlist_next_rcu(h));
1032         while (!h) {
1033                 struct nfnl_log_net *log;
1034                 struct hlist_head *head;
1035
1036                 if (++st->bucket >= INSTANCE_BUCKETS)
1037                         return NULL;
1038
1039                 log = nfnl_log_pernet(net);
1040                 head = &log->instance_table[st->bucket];
1041                 h = rcu_dereference_bh(hlist_first_rcu(head));
1042         }
1043         return h;
1044 }
1045
1046 static struct hlist_node *get_idx(struct net *net, struct iter_state *st,
1047                                   loff_t pos)
1048 {
1049         struct hlist_node *head;
1050         head = get_first(net, st);
1051
1052         if (head)
1053                 while (pos && (head = get_next(net, st, head)))
1054                         pos--;
1055         return pos ? NULL : head;
1056 }
1057
1058 static void *seq_start(struct seq_file *s, loff_t *pos)
1059         __acquires(rcu_bh)
1060 {
1061         rcu_read_lock_bh();
1062         return get_idx(seq_file_net(s), s->private, *pos);
1063 }
1064
1065 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1066 {
1067         (*pos)++;
1068         return get_next(seq_file_net(s), s->private, v);
1069 }
1070
1071 static void seq_stop(struct seq_file *s, void *v)
1072         __releases(rcu_bh)
1073 {
1074         rcu_read_unlock_bh();
1075 }
1076
1077 static int seq_show(struct seq_file *s, void *v)
1078 {
1079         const struct nfulnl_instance *inst = v;
1080
1081         seq_printf(s, "%5u %6u %5u %1u %5u %6u %2u\n",
1082                    inst->group_num,
1083                    inst->peer_portid, inst->qlen,
1084                    inst->copy_mode, inst->copy_range,
1085                    inst->flushtimeout, refcount_read(&inst->use));
1086
1087         return 0;
1088 }
1089
1090 static const struct seq_operations nful_seq_ops = {
1091         .start  = seq_start,
1092         .next   = seq_next,
1093         .stop   = seq_stop,
1094         .show   = seq_show,
1095 };
1096 #endif /* PROC_FS */
1097
1098 static int __net_init nfnl_log_net_init(struct net *net)
1099 {
1100         unsigned int i;
1101         struct nfnl_log_net *log = nfnl_log_pernet(net);
1102 #ifdef CONFIG_PROC_FS
1103         struct proc_dir_entry *proc;
1104         kuid_t root_uid;
1105         kgid_t root_gid;
1106 #endif
1107
1108         for (i = 0; i < INSTANCE_BUCKETS; i++)
1109                 INIT_HLIST_HEAD(&log->instance_table[i]);
1110         spin_lock_init(&log->instances_lock);
1111
1112 #ifdef CONFIG_PROC_FS
1113         proc = proc_create_net("nfnetlink_log", 0440, net->nf.proc_netfilter,
1114                         &nful_seq_ops, sizeof(struct iter_state));
1115         if (!proc)
1116                 return -ENOMEM;
1117
1118         root_uid = make_kuid(net->user_ns, 0);
1119         root_gid = make_kgid(net->user_ns, 0);
1120         if (uid_valid(root_uid) && gid_valid(root_gid))
1121                 proc_set_user(proc, root_uid, root_gid);
1122 #endif
1123         return 0;
1124 }
1125
1126 static void __net_exit nfnl_log_net_exit(struct net *net)
1127 {
1128         struct nfnl_log_net *log = nfnl_log_pernet(net);
1129         unsigned int i;
1130
1131 #ifdef CONFIG_PROC_FS
1132         remove_proc_entry("nfnetlink_log", net->nf.proc_netfilter);
1133 #endif
1134         nf_log_unset(net, &nfulnl_logger);
1135         for (i = 0; i < INSTANCE_BUCKETS; i++)
1136                 WARN_ON_ONCE(!hlist_empty(&log->instance_table[i]));
1137 }
1138
1139 static struct pernet_operations nfnl_log_net_ops = {
1140         .init   = nfnl_log_net_init,
1141         .exit   = nfnl_log_net_exit,
1142         .id     = &nfnl_log_net_id,
1143         .size   = sizeof(struct nfnl_log_net),
1144 };
1145
1146 static int __init nfnetlink_log_init(void)
1147 {
1148         int status;
1149
1150         status = register_pernet_subsys(&nfnl_log_net_ops);
1151         if (status < 0) {
1152                 pr_err("failed to register pernet ops\n");
1153                 goto out;
1154         }
1155
1156         netlink_register_notifier(&nfulnl_rtnl_notifier);
1157         status = nfnetlink_subsys_register(&nfulnl_subsys);
1158         if (status < 0) {
1159                 pr_err("failed to create netlink socket\n");
1160                 goto cleanup_netlink_notifier;
1161         }
1162
1163         status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger);
1164         if (status < 0) {
1165                 pr_err("failed to register logger\n");
1166                 goto cleanup_subsys;
1167         }
1168
1169         return status;
1170
1171 cleanup_subsys:
1172         nfnetlink_subsys_unregister(&nfulnl_subsys);
1173 cleanup_netlink_notifier:
1174         netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1175         unregister_pernet_subsys(&nfnl_log_net_ops);
1176 out:
1177         return status;
1178 }
1179
1180 static void __exit nfnetlink_log_fini(void)
1181 {
1182         nfnetlink_subsys_unregister(&nfulnl_subsys);
1183         netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1184         unregister_pernet_subsys(&nfnl_log_net_ops);
1185         nf_log_unregister(&nfulnl_logger);
1186 }
1187
1188 MODULE_DESCRIPTION("netfilter userspace logging");
1189 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1190 MODULE_LICENSE("GPL");
1191 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG);
1192 MODULE_ALIAS_NF_LOGGER(AF_INET, 1);
1193 MODULE_ALIAS_NF_LOGGER(AF_INET6, 1);
1194 MODULE_ALIAS_NF_LOGGER(AF_BRIDGE, 1);
1195 MODULE_ALIAS_NF_LOGGER(3, 1); /* NFPROTO_ARP */
1196 MODULE_ALIAS_NF_LOGGER(5, 1); /* NFPROTO_NETDEV */
1197
1198 module_init(nfnetlink_log_init);
1199 module_exit(nfnetlink_log_fini);