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         struct nfgenmsg *nfmsg;
 456         sk_buff_data_t old_tail = inst->skb->tail;
 457         struct sock *sk;
 458         const unsigned char *hwhdrp;
 459
 460         nlh = nlmsg_put(inst->skb, 0, 0,
 461                         nfnl_msg_type(NFNL_SUBSYS_ULOG, NFULNL_MSG_PACKET),
 462                         sizeof(struct nfgenmsg), 0);
 463         if (!nlh)
 464                 return -1;
 465         nfmsg = nlmsg_data(nlh);
 466         nfmsg->nfgen_family = pf;
 467         nfmsg->version = NFNETLINK_V0;
 468         nfmsg->res_id = htons(inst->group_num);
 469
 470         memset(&pmsg, 0, sizeof(pmsg));
 471         pmsg.hw_protocol        = skb->protocol;
 472         pmsg.hook               = hooknum;
 473
 474         if (nla_put(inst->skb, NFULA_PACKET_HDR, sizeof(pmsg), &pmsg))
 475                 goto nla_put_failure;
 476
 477         if (prefix &&
 478             nla_put(inst->skb, NFULA_PREFIX, plen, prefix))
 479                 goto nla_put_failure;
 480
 481         if (indev) {
 482 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 483                 if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 484                                  htonl(indev->ifindex)))
 485                         goto nla_put_failure;
 486 #else
 487                 if (pf == PF_BRIDGE) {
 488                         /* Case 1: outdev is physical input device, we need to
 489                          * look for bridge group (when called from
 490                          * netfilter_bridge) */
 491                         if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
 492                                          htonl(indev->ifindex)) ||
 493                         /* this is the bridge group "brX" */
 494                         /* rcu_read_lock()ed by nf_hook_thresh or
 495                          * nf_log_packet.
 496                          */
 497                             nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 498                                          htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
 499                                 goto nla_put_failure;
 500                 } else {
 501                         struct net_device *physindev;
 502
 503                         /* Case 2: indev is bridge group, we need to look for
 504                          * physical device (when called from ipv4) */
 505                         if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 506                                          htonl(indev->ifindex)))
 507                                 goto nla_put_failure;
 508
 509                         physindev = nf_bridge_get_physindev(skb);
 510                         if (physindev &&
 511                             nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
 512                                          htonl(physindev->ifindex)))
 513                                 goto nla_put_failure;
 514                 }
 515 #endif
 516         }
 517
 518         if (outdev) {
 519 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 520                 if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 521                                  htonl(outdev->ifindex)))
 522                         goto nla_put_failure;
 523 #else
 524                 if (pf == PF_BRIDGE) {
 525                         /* Case 1: outdev is physical output device, we need to
 526                          * look for bridge group (when called from
 527                          * netfilter_bridge) */
 528                         if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
 529                                          htonl(outdev->ifindex)) ||
 530                         /* this is the bridge group "brX" */
 531                         /* rcu_read_lock()ed by nf_hook_thresh or
 532                          * nf_log_packet.
 533                          */
 534                             nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 535                                          htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
 536                                 goto nla_put_failure;
 537                 } else {
 538                         struct net_device *physoutdev;
 539
 540                         /* Case 2: indev is a bridge group, we need to look
 541                          * for physical device (when called from ipv4) */
 542                         if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 543                                          htonl(outdev->ifindex)))
 544                                 goto nla_put_failure;
 545
 546                         physoutdev = nf_bridge_get_physoutdev(skb);
 547                         if (physoutdev &&
 548                             nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
 549                                          htonl(physoutdev->ifindex)))
 550                                 goto nla_put_failure;
 551                 }
 552 #endif
 553         }
 554
 555         if (skb->mark &&
 556             nla_put_be32(inst->skb, NFULA_MARK, htonl(skb->mark)))
 557                 goto nla_put_failure;
 558
 559         if (indev && skb->dev &&
 560             skb->mac_header != skb->network_header) {
 561                 struct nfulnl_msg_packet_hw phw;
 562                 int len;
 563
 564                 memset(&phw, 0, sizeof(phw));
 565                 len = dev_parse_header(skb, phw.hw_addr);
 566                 if (len > 0) {
 567                         phw.hw_addrlen = htons(len);
 568                         if (nla_put(inst->skb, NFULA_HWADDR, sizeof(phw), &phw))
 569                                 goto nla_put_failure;
 570                 }
 571         }
 572
 573         if (indev && skb_mac_header_was_set(skb)) {
 574                 if (nla_put_be16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type)) ||
 575                     nla_put_be16(inst->skb, NFULA_HWLEN,
 576                                  htons(skb->dev->hard_header_len)))
 577                         goto nla_put_failure;
 578
 579                 hwhdrp = skb_mac_header(skb);
 580
 581                 if (skb->dev->type == ARPHRD_SIT)
 582                         hwhdrp -= ETH_HLEN;
 583
 584                 if (hwhdrp >= skb->head &&
 585                     nla_put(inst->skb, NFULA_HWHEADER,
 586                             skb->dev->hard_header_len, hwhdrp))
 587                         goto nla_put_failure;
 588         }
 589
 590         if (hooknum <= NF_INET_FORWARD && skb->tstamp) {
 591                 struct nfulnl_msg_packet_timestamp ts;
 592                 struct timespec64 kts = ktime_to_timespec64(skb->tstamp);
 593                 ts.sec = cpu_to_be64(kts.tv_sec);
 594                 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
 595
 596                 if (nla_put(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts))
 597                         goto nla_put_failure;
 598         }
 599
 600         /* UID */
 601         sk = skb->sk;
 602         if (sk && sk_fullsock(sk)) {
 603                 read_lock_bh(&sk->sk_callback_lock);
 604                 if (sk->sk_socket && sk->sk_socket->file) {
 605                         struct file *file = sk->sk_socket->file;
 606                         const struct cred *cred = file->f_cred;
 607                         struct user_namespace *user_ns = inst->peer_user_ns;
 608                         __be32 uid = htonl(from_kuid_munged(user_ns, cred->fsuid));
 609                         __be32 gid = htonl(from_kgid_munged(user_ns, cred->fsgid));
 610                         read_unlock_bh(&sk->sk_callback_lock);
 611                         if (nla_put_be32(inst->skb, NFULA_UID, uid) ||
 612                             nla_put_be32(inst->skb, NFULA_GID, gid))
 613                                 goto nla_put_failure;
 614                 } else
 615                         read_unlock_bh(&sk->sk_callback_lock);
 616         }
 617
 618         /* local sequence number */
 619         if ((inst->flags & NFULNL_CFG_F_SEQ) &&
 620             nla_put_be32(inst->skb, NFULA_SEQ, htonl(inst->seq++)))
 621                 goto nla_put_failure;
 622
 623         /* global sequence number */
 624         if ((inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) &&
 625             nla_put_be32(inst->skb, NFULA_SEQ_GLOBAL,
 626                          htonl(atomic_inc_return(&log->global_seq))))
 627                 goto nla_put_failure;
 628
 629         if (ct && nfnl_ct->build(inst->skb, ct, ctinfo,
 630                                  NFULA_CT, NFULA_CT_INFO) < 0)
 631                 goto nla_put_failure;
 632
 633         if ((pf == NFPROTO_NETDEV || pf == NFPROTO_BRIDGE) &&
 634             nfulnl_put_bridge(inst, skb) < 0)
 635                 goto nla_put_failure;
 636
 637         if (data_len) {
 638                 struct nlattr *nla;
 639                 int size = nla_attr_size(data_len);
 640
 641                 if (skb_tailroom(inst->skb) < nla_total_size(data_len))
 642                         goto nla_put_failure;
 643
 644                 nla = skb_put(inst->skb, nla_total_size(data_len));
 645                 nla->nla_type = NFULA_PAYLOAD;
 646                 nla->nla_len = size;
 647
 648                 if (skb_copy_bits(skb, 0, nla_data(nla), data_len))
 649                         BUG();
 650         }
 651
 652         nlh->nlmsg_len = inst->skb->tail - old_tail;
 653         return 0;
 654
 655 nla_put_failure:
 656         PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n");
 657         return -1;
 658 }
 659
 660 static const struct nf_loginfo default_loginfo = {
 661         .type =         NF_LOG_TYPE_ULOG,
 662         .u = {
 663                 .ulog = {
 664                         .copy_len       = 0xffff,
 665                         .group          = 0,
 666                         .qthreshold     = 1,
 667                 },
 668         },
 669 };
 670
 671 /* log handler for internal netfilter logging api */
 672 static void
 673 nfulnl_log_packet(struct net *net,
 674                   u_int8_t pf,
 675                   unsigned int hooknum,
 676                   const struct sk_buff *skb,
 677                   const struct net_device *in,
 678                   const struct net_device *out,
 679                   const struct nf_loginfo *li_user,
 680                   const char *prefix)
 681 {
 682         size_t size;
 683         unsigned int data_len;
 684         struct nfulnl_instance *inst;
 685         const struct nf_loginfo *li;
 686         unsigned int qthreshold;
 687         unsigned int plen = 0;
 688         struct nfnl_log_net *log = nfnl_log_pernet(net);
 689         const struct nfnl_ct_hook *nfnl_ct = NULL;
 690         struct nf_conn *ct = NULL;
 691         enum ip_conntrack_info ctinfo;
 692
 693         if (li_user && li_user->type == NF_LOG_TYPE_ULOG)
 694                 li = li_user;
 695         else
 696                 li = &default_loginfo;
 697
 698         inst = instance_lookup_get(log, li->u.ulog.group);
 699         if (!inst)
 700                 return;
 701
 702         if (prefix)
 703                 plen = strlen(prefix) + 1;
 704
 705         /* FIXME: do we want to make the size calculation conditional based on
 706          * what is actually present?  way more branches and checks, but more
 707          * memory efficient... */
 708         size = nlmsg_total_size(sizeof(struct nfgenmsg))
 709                 + nla_total_size(sizeof(struct nfulnl_msg_packet_hdr))
 710                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 711                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 712 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 713                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 714                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 715 #endif
 716                 + nla_total_size(sizeof(u_int32_t))     /* mark */
 717                 + nla_total_size(sizeof(u_int32_t))     /* uid */
 718                 + nla_total_size(sizeof(u_int32_t))     /* gid */
 719                 + nla_total_size(plen)                  /* prefix */
 720                 + nla_total_size(sizeof(struct nfulnl_msg_packet_hw))
 721                 + nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp))
 722                 + nla_total_size(sizeof(struct nfgenmsg));      /* NLMSG_DONE */
 723
 724         if (in && skb_mac_header_was_set(skb)) {
 725                 size += nla_total_size(skb->dev->hard_header_len)
 726                         + nla_total_size(sizeof(u_int16_t))     /* hwtype */
 727                         + nla_total_size(sizeof(u_int16_t));    /* hwlen */
 728         }
 729
 730         spin_lock_bh(&inst->lock);
 731
 732         if (inst->flags & NFULNL_CFG_F_SEQ)
 733                 size += nla_total_size(sizeof(u_int32_t));
 734         if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
 735                 size += nla_total_size(sizeof(u_int32_t));
 736         if (inst->flags & NFULNL_CFG_F_CONNTRACK) {
 737                 nfnl_ct = rcu_dereference(nfnl_ct_hook);
 738                 if (nfnl_ct != NULL) {
 739                         ct = nfnl_ct->get_ct(skb, &ctinfo);
 740                         if (ct != NULL)
 741                                 size += nfnl_ct->build_size(ct);
 742                 }
 743         }
 744         if (pf == NFPROTO_NETDEV || pf == NFPROTO_BRIDGE)
 745                 size += nfulnl_get_bridge_size(skb);
 746
 747         qthreshold = inst->qthreshold;
 748         /* per-rule qthreshold overrides per-instance */
 749         if (li->u.ulog.qthreshold)
 750                 if (qthreshold > li->u.ulog.qthreshold)
 751                         qthreshold = li->u.ulog.qthreshold;
 752
 753
 754         switch (inst->copy_mode) {
 755         case NFULNL_COPY_META:
 756         case NFULNL_COPY_NONE:
 757                 data_len = 0;
 758                 break;
 759
 760         case NFULNL_COPY_PACKET:
 761                 data_len = inst->copy_range;
 762                 if ((li->u.ulog.flags & NF_LOG_F_COPY_LEN) &&
 763                     (li->u.ulog.copy_len < data_len))
 764                         data_len = li->u.ulog.copy_len;
 765
 766                 if (data_len > skb->len)
 767                         data_len = skb->len;
 768
 769                 size += nla_total_size(data_len);
 770                 break;
 771
 772         case NFULNL_COPY_DISABLED:
 773         default:
 774                 goto unlock_and_release;
 775         }
 776
 777         if (inst->skb && size > skb_tailroom(inst->skb)) {
 778                 /* either the queue len is too high or we don't have
 779                  * enough room in the skb left. flush to userspace. */
 780                 __nfulnl_flush(inst);
 781         }
 782
 783         if (!inst->skb) {
 784                 inst->skb = nfulnl_alloc_skb(net, inst->peer_portid,
 785                                              inst->nlbufsiz, size);
 786                 if (!inst->skb)
 787                         goto alloc_failure;
 788         }
 789
 790         inst->qlen++;
 791
 792         __build_packet_message(log, inst, skb, data_len, pf,
 793                                 hooknum, in, out, prefix, plen,
 794                                 nfnl_ct, ct, ctinfo);
 795
 796         if (inst->qlen >= qthreshold)
 797                 __nfulnl_flush(inst);
 798         /* timer_pending always called within inst->lock, so there
 799          * is no chance of a race here */
 800         else if (!timer_pending(&inst->timer)) {
 801                 instance_get(inst);
 802                 inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100);
 803                 add_timer(&inst->timer);
 804         }
 805
 806 unlock_and_release:
 807         spin_unlock_bh(&inst->lock);
 808         instance_put(inst);
 809         return;
 810
 811 alloc_failure:
 812         /* FIXME: statistics */
 813         goto unlock_and_release;
 814 }
 815
 816 static int
 817 nfulnl_rcv_nl_event(struct notifier_block *this,
 818                    unsigned long event, void *ptr)
 819 {
 820         struct netlink_notify *n = ptr;
 821         struct nfnl_log_net *log = nfnl_log_pernet(n->net);
 822
 823         if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
 824                 int i;
 825
 826                 /* destroy all instances for this portid */
 827                 spin_lock_bh(&log->instances_lock);
 828                 for  (i = 0; i < INSTANCE_BUCKETS; i++) {
 829                         struct hlist_node *t2;
 830                         struct nfulnl_instance *inst;
 831                         struct hlist_head *head = &log->instance_table[i];
 832
 833                         hlist_for_each_entry_safe(inst, t2, head, hlist) {
 834                                 if (n->portid == inst->peer_portid)
 835                                         __instance_destroy(inst);
 836                         }
 837                 }
 838                 spin_unlock_bh(&log->instances_lock);
 839         }
 840         return NOTIFY_DONE;
 841 }
 842
 843 static struct notifier_block nfulnl_rtnl_notifier = {
 844         .notifier_call  = nfulnl_rcv_nl_event,
 845 };
 846
 847 static int nfulnl_recv_unsupp(struct net *net, struct sock *ctnl,
 848                               struct sk_buff *skb, const struct nlmsghdr *nlh,
 849                               const struct nlattr * const nfqa[],
 850                               struct netlink_ext_ack *extack)
 851 {
 852         return -ENOTSUPP;
 853 }
 854
 855 static struct nf_logger nfulnl_logger __read_mostly = {
 856         .name   = "nfnetlink_log",
 857         .type   = NF_LOG_TYPE_ULOG,
 858         .logfn  = nfulnl_log_packet,
 859         .me     = THIS_MODULE,
 860 };
 861
 862 static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = {
 863         [NFULA_CFG_CMD]         = { .len = sizeof(struct nfulnl_msg_config_cmd) },
 864         [NFULA_CFG_MODE]        = { .len = sizeof(struct nfulnl_msg_config_mode) },
 865         [NFULA_CFG_TIMEOUT]     = { .type = NLA_U32 },
 866         [NFULA_CFG_QTHRESH]     = { .type = NLA_U32 },
 867         [NFULA_CFG_NLBUFSIZ]    = { .type = NLA_U32 },
 868         [NFULA_CFG_FLAGS]       = { .type = NLA_U16 },
 869 };
 870
 871 static int nfulnl_recv_config(struct net *net, struct sock *ctnl,
 872                               struct sk_buff *skb, const struct nlmsghdr *nlh,
 873                               const struct nlattr * const nfula[],
 874                               struct netlink_ext_ack *extack)
 875 {
 876         struct nfgenmsg *nfmsg = nlmsg_data(nlh);
 877         u_int16_t group_num = ntohs(nfmsg->res_id);
 878         struct nfulnl_instance *inst;
 879         struct nfulnl_msg_config_cmd *cmd = NULL;
 880         struct nfnl_log_net *log = nfnl_log_pernet(net);
 881         int ret = 0;
 882         u16 flags = 0;
 883
 884         if (nfula[NFULA_CFG_CMD]) {
 885                 u_int8_t pf = nfmsg->nfgen_family;
 886                 cmd = nla_data(nfula[NFULA_CFG_CMD]);
 887
 888                 /* Commands without queue context */
 889                 switch (cmd->command) {
 890                 case NFULNL_CFG_CMD_PF_BIND:
 891                         return nf_log_bind_pf(net, pf, &nfulnl_logger);
 892                 case NFULNL_CFG_CMD_PF_UNBIND:
 893                         nf_log_unbind_pf(net, pf);
 894                         return 0;
 895                 }
 896         }
 897
 898         inst = instance_lookup_get(log, group_num);
 899         if (inst && inst->peer_portid != NETLINK_CB(skb).portid) {
 900                 ret = -EPERM;
 901                 goto out_put;
 902         }
 903
 904         /* Check if we support these flags in first place, dependencies should
 905          * be there too not to break atomicity.
 906          */
 907         if (nfula[NFULA_CFG_FLAGS]) {
 908                 flags = ntohs(nla_get_be16(nfula[NFULA_CFG_FLAGS]));
 909
 910                 if ((flags & NFULNL_CFG_F_CONNTRACK) &&
 911                     !rcu_access_pointer(nfnl_ct_hook)) {
 912 #ifdef CONFIG_MODULES
 913                         nfnl_unlock(NFNL_SUBSYS_ULOG);
 914                         request_module("ip_conntrack_netlink");
 915                         nfnl_lock(NFNL_SUBSYS_ULOG);
 916                         if (rcu_access_pointer(nfnl_ct_hook)) {
 917                                 ret = -EAGAIN;
 918                                 goto out_put;
 919                         }
 920 #endif
 921                         ret = -EOPNOTSUPP;
 922                         goto out_put;
 923                 }
 924         }
 925
 926         if (cmd != NULL) {
 927                 switch (cmd->command) {
 928                 case NFULNL_CFG_CMD_BIND:
 929                         if (inst) {
 930                                 ret = -EBUSY;
 931                                 goto out_put;
 932                         }
 933
 934                         inst = instance_create(net, group_num,
 935                                                NETLINK_CB(skb).portid,
 936                                                sk_user_ns(NETLINK_CB(skb).sk));
 937                         if (IS_ERR(inst)) {
 938                                 ret = PTR_ERR(inst);
 939                                 goto out;
 940                         }
 941                         break;
 942                 case NFULNL_CFG_CMD_UNBIND:
 943                         if (!inst) {
 944                                 ret = -ENODEV;
 945                                 goto out;
 946                         }
 947
 948                         instance_destroy(log, inst);
 949                         goto out_put;
 950                 default:
 951                         ret = -ENOTSUPP;
 952                         goto out_put;
 953                 }
 954         } else if (!inst) {
 955                 ret = -ENODEV;
 956                 goto out;
 957         }
 958
 959         if (nfula[NFULA_CFG_MODE]) {
 960                 struct nfulnl_msg_config_mode *params =
 961                         nla_data(nfula[NFULA_CFG_MODE]);
 962
 963                 nfulnl_set_mode(inst, params->copy_mode,
 964                                 ntohl(params->copy_range));
 965         }
 966
 967         if (nfula[NFULA_CFG_TIMEOUT]) {
 968                 __be32 timeout = nla_get_be32(nfula[NFULA_CFG_TIMEOUT]);
 969
 970                 nfulnl_set_timeout(inst, ntohl(timeout));
 971         }
 972
 973         if (nfula[NFULA_CFG_NLBUFSIZ]) {
 974                 __be32 nlbufsiz = nla_get_be32(nfula[NFULA_CFG_NLBUFSIZ]);
 975
 976                 nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz));
 977         }
 978
 979         if (nfula[NFULA_CFG_QTHRESH]) {
 980                 __be32 qthresh = nla_get_be32(nfula[NFULA_CFG_QTHRESH]);
 981
 982                 nfulnl_set_qthresh(inst, ntohl(qthresh));
 983         }
 984
 985         if (nfula[NFULA_CFG_FLAGS])
 986                 nfulnl_set_flags(inst, flags);
 987
 988 out_put:
 989         instance_put(inst);
 990 out:
 991         return ret;
 992 }
 993
 994 static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = {
 995         [NFULNL_MSG_PACKET]     = { .call = nfulnl_recv_unsupp,
 996                                     .attr_count = NFULA_MAX, },
 997         [NFULNL_MSG_CONFIG]     = { .call = nfulnl_recv_config,
 998                                     .attr_count = NFULA_CFG_MAX,
 999                                     .policy = nfula_cfg_policy },
1000 };
1001
1002 static const struct nfnetlink_subsystem nfulnl_subsys = {
1003         .name           = "log",
1004         .subsys_id      = NFNL_SUBSYS_ULOG,
1005         .cb_count       = NFULNL_MSG_MAX,
1006         .cb             = nfulnl_cb,
1007 };
1008
1009 #ifdef CONFIG_PROC_FS
1010 struct iter_state {
1011         struct seq_net_private p;
1012         unsigned int bucket;
1013 };
1014
1015 static struct hlist_node *get_first(struct net *net, struct iter_state *st)
1016 {
1017         struct nfnl_log_net *log;
1018         if (!st)
1019                 return NULL;
1020
1021         log = nfnl_log_pernet(net);
1022
1023         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1024                 struct hlist_head *head = &log->instance_table[st->bucket];
1025
1026                 if (!hlist_empty(head))
1027                         return rcu_dereference_bh(hlist_first_rcu(head));
1028         }
1029         return NULL;
1030 }
1031
1032 static struct hlist_node *get_next(struct net *net, struct iter_state *st,
1033                                    struct hlist_node *h)
1034 {
1035         h = rcu_dereference_bh(hlist_next_rcu(h));
1036         while (!h) {
1037                 struct nfnl_log_net *log;
1038                 struct hlist_head *head;
1039
1040                 if (++st->bucket >= INSTANCE_BUCKETS)
1041                         return NULL;
1042
1043                 log = nfnl_log_pernet(net);
1044                 head = &log->instance_table[st->bucket];
1045                 h = rcu_dereference_bh(hlist_first_rcu(head));
1046         }
1047         return h;
1048 }
1049
1050 static struct hlist_node *get_idx(struct net *net, struct iter_state *st,
1051                                   loff_t pos)
1052 {
1053         struct hlist_node *head;
1054         head = get_first(net, st);
1055
1056         if (head)
1057                 while (pos && (head = get_next(net, st, head)))
1058                         pos--;
1059         return pos ? NULL : head;
1060 }
1061
1062 static void *seq_start(struct seq_file *s, loff_t *pos)
1063         __acquires(rcu_bh)
1064 {
1065         rcu_read_lock_bh();
1066         return get_idx(seq_file_net(s), s->private, *pos);
1067 }
1068
1069 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1070 {
1071         (*pos)++;
1072         return get_next(seq_file_net(s), s->private, v);
1073 }
1074
1075 static void seq_stop(struct seq_file *s, void *v)
1076         __releases(rcu_bh)
1077 {
1078         rcu_read_unlock_bh();
1079 }
1080
1081 static int seq_show(struct seq_file *s, void *v)
1082 {
1083         const struct nfulnl_instance *inst = v;
1084
1085         seq_printf(s, "%5u %6u %5u %1u %5u %6u %2u\n",
1086                    inst->group_num,
1087                    inst->peer_portid, inst->qlen,
1088                    inst->copy_mode, inst->copy_range,
1089                    inst->flushtimeout, refcount_read(&inst->use));
1090
1091         return 0;
1092 }
1093
1094 static const struct seq_operations nful_seq_ops = {
1095         .start  = seq_start,
1096         .next   = seq_next,
1097         .stop   = seq_stop,
1098         .show   = seq_show,
1099 };
1100 #endif /* PROC_FS */
1101
1102 static int __net_init nfnl_log_net_init(struct net *net)
1103 {
1104         unsigned int i;
1105         struct nfnl_log_net *log = nfnl_log_pernet(net);
1106 #ifdef CONFIG_PROC_FS
1107         struct proc_dir_entry *proc;
1108         kuid_t root_uid;
1109         kgid_t root_gid;
1110 #endif
1111
1112         for (i = 0; i < INSTANCE_BUCKETS; i++)
1113                 INIT_HLIST_HEAD(&log->instance_table[i]);
1114         spin_lock_init(&log->instances_lock);
1115
1116 #ifdef CONFIG_PROC_FS
1117         proc = proc_create_net("nfnetlink_log", 0440, net->nf.proc_netfilter,
1118                         &nful_seq_ops, sizeof(struct iter_state));
1119         if (!proc)
1120                 return -ENOMEM;
1121
1122         root_uid = make_kuid(net->user_ns, 0);
1123         root_gid = make_kgid(net->user_ns, 0);
1124         if (uid_valid(root_uid) && gid_valid(root_gid))
1125                 proc_set_user(proc, root_uid, root_gid);
1126 #endif
1127         return 0;
1128 }
1129
1130 static void __net_exit nfnl_log_net_exit(struct net *net)
1131 {
1132         struct nfnl_log_net *log = nfnl_log_pernet(net);
1133         unsigned int i;
1134
1135 #ifdef CONFIG_PROC_FS
1136         remove_proc_entry("nfnetlink_log", net->nf.proc_netfilter);
1137 #endif
1138         nf_log_unset(net, &nfulnl_logger);
1139         for (i = 0; i < INSTANCE_BUCKETS; i++)
1140                 WARN_ON_ONCE(!hlist_empty(&log->instance_table[i]));
1141 }
1142
1143 static struct pernet_operations nfnl_log_net_ops = {
1144         .init   = nfnl_log_net_init,
1145         .exit   = nfnl_log_net_exit,
1146         .id     = &nfnl_log_net_id,
1147         .size   = sizeof(struct nfnl_log_net),
1148 };
1149
1150 static int __init nfnetlink_log_init(void)
1151 {
1152         int status;
1153
1154         status = register_pernet_subsys(&nfnl_log_net_ops);
1155         if (status < 0) {
1156                 pr_err("failed to register pernet ops\n");
1157                 goto out;
1158         }
1159
1160         netlink_register_notifier(&nfulnl_rtnl_notifier);
1161         status = nfnetlink_subsys_register(&nfulnl_subsys);
1162         if (status < 0) {
1163                 pr_err("failed to create netlink socket\n");
1164                 goto cleanup_netlink_notifier;
1165         }
1166
1167         status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger);
1168         if (status < 0) {
1169                 pr_err("failed to register logger\n");
1170                 goto cleanup_subsys;
1171         }
1172
1173         return status;
1174
1175 cleanup_subsys:
1176         nfnetlink_subsys_unregister(&nfulnl_subsys);
1177 cleanup_netlink_notifier:
1178         netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1179         unregister_pernet_subsys(&nfnl_log_net_ops);
1180 out:
1181         return status;
1182 }
1183
1184 static void __exit nfnetlink_log_fini(void)
1185 {
1186         nfnetlink_subsys_unregister(&nfulnl_subsys);
1187         netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1188         unregister_pernet_subsys(&nfnl_log_net_ops);
1189         nf_log_unregister(&nfulnl_logger);
1190 }
1191
1192 MODULE_DESCRIPTION("netfilter userspace logging");
1193 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1194 MODULE_LICENSE("GPL");
1195 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG);
1196 MODULE_ALIAS_NF_LOGGER(AF_INET, 1);
1197 MODULE_ALIAS_NF_LOGGER(AF_INET6, 1);
1198 MODULE_ALIAS_NF_LOGGER(AF_BRIDGE, 1);
1199 MODULE_ALIAS_NF_LOGGER(3, 1); /* NFPROTO_ARP */
1200 MODULE_ALIAS_NF_LOGGER(5, 1); /* NFPROTO_NETDEV */
1201
1202 module_init(nfnetlink_log_init);
1203 module_exit(nfnetlink_log_fini);