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