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#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 <net/netlink.h>
  24#include <linux/netfilter/nfnetlink.h>
  25#include <linux/netfilter/nfnetlink_log.h>
  26#include <linux/spinlock.h>
  27#include <linux/sysctl.h>
  28#include <linux/proc_fs.h>
  29#include <linux/security.h>
  30#include <linux/list.h>
  31#include <linux/jhash.h>
  32#include <linux/random.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#include <net/netfilter/nfnetlink_log.h>
  38
  39#include <linux/atomic.h>
  40
  41#ifdef CONFIG_BRIDGE_NETFILTER
  42#include "../bridge/br_private.h"
  43#endif
  44
  45#define NFULNL_NLBUFSIZ_DEFAULT NLMSG_GOODSIZE
  46#define NFULNL_TIMEOUT_DEFAULT  100     /* every second */
  47#define NFULNL_QTHRESH_DEFAULT  100     /* 100 packets */
  48#define NFULNL_COPY_RANGE_MAX   0xFFFF  /* max packet size is limited by 16-bit struct nfattr nfa_len field */
  49
  50#define PRINTR(x, args...)      do { if (net_ratelimit()) \
  51                                     printk(x, ## args); } while (0);
  52
  53struct nfulnl_instance {
  54        struct hlist_node hlist;        /* global list of instances */
  55        spinlock_t lock;
  56        atomic_t use;                   /* use count */
  57
  58        unsigned int qlen;              /* number of nlmsgs in skb */
  59        struct sk_buff *skb;            /* pre-allocatd skb */
  60        struct timer_list timer;
  61        struct net *net;
  62        struct user_namespace *peer_user_ns;    /* User namespace of the peer process */
  63        int peer_portid;                        /* PORTID of the peer process */
  64
  65        /* configurable parameters */
  66        unsigned int flushtimeout;      /* timeout until queue flush */
  67        unsigned int nlbufsiz;          /* netlink buffer allocation size */
  68        unsigned int qthreshold;        /* threshold of the queue */
  69        u_int32_t copy_range;
  70        u_int32_t seq;                  /* instance-local sequential counter */
  71        u_int16_t group_num;            /* number of this queue */
  72        u_int16_t flags;
  73        u_int8_t copy_mode;
  74        struct rcu_head rcu;
  75};
  76
  77#define INSTANCE_BUCKETS        16
  78static unsigned int hash_init;
  79
  80static int nfnl_log_net_id __read_mostly;
  81
  82struct nfnl_log_net {
  83        spinlock_t instances_lock;
  84        struct hlist_head instance_table[INSTANCE_BUCKETS];
  85        atomic_t global_seq;
  86};
  87
  88static struct nfnl_log_net *nfnl_log_pernet(struct net *net)
  89{
  90        return net_generic(net, nfnl_log_net_id);
  91}
  92
  93static inline u_int8_t instance_hashfn(u_int16_t group_num)
  94{
  95        return ((group_num & 0xff) % INSTANCE_BUCKETS);
  96}
  97
  98static struct nfulnl_instance *
  99__instance_lookup(struct nfnl_log_net *log, u_int16_t group_num)
 100{
 101        struct hlist_head *head;
 102        struct nfulnl_instance *inst;
 103
 104        head = &log->instance_table[instance_hashfn(group_num)];
 105        hlist_for_each_entry_rcu(inst, head, hlist) {
 106                if (inst->group_num == group_num)
 107                        return inst;
 108        }
 109        return NULL;
 110}
 111
 112static inline void
 113instance_get(struct nfulnl_instance *inst)
 114{
 115        atomic_inc(&inst->use);
 116}
 117
 118static struct nfulnl_instance *
 119instance_lookup_get(struct nfnl_log_net *log, u_int16_t group_num)
 120{
 121        struct nfulnl_instance *inst;
 122
 123        rcu_read_lock_bh();
 124        inst = __instance_lookup(log, group_num);
 125        if (inst && !atomic_inc_not_zero(&inst->use))
 126                inst = NULL;
 127        rcu_read_unlock_bh();
 128
 129        return inst;
 130}
 131
 132static void nfulnl_instance_free_rcu(struct rcu_head *head)
 133{
 134        struct nfulnl_instance *inst =
 135                container_of(head, struct nfulnl_instance, rcu);
 136
 137        put_net(inst->net);
 138        kfree(inst);
 139        module_put(THIS_MODULE);
 140}
 141
 142static void
 143instance_put(struct nfulnl_instance *inst)
 144{
 145        if (inst && atomic_dec_and_test(&inst->use))
 146                call_rcu_bh(&inst->rcu, nfulnl_instance_free_rcu);
 147}
 148
 149static void nfulnl_timer(unsigned long data);
 150
 151static struct nfulnl_instance *
 152instance_create(struct net *net, u_int16_t group_num,
 153                int portid, struct user_namespace *user_ns)
 154{
 155        struct nfulnl_instance *inst;
 156        struct nfnl_log_net *log = nfnl_log_pernet(net);
 157        int err;
 158
 159        spin_lock_bh(&log->instances_lock);
 160        if (__instance_lookup(log, group_num)) {
 161                err = -EEXIST;
 162                goto out_unlock;
 163        }
 164
 165        inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
 166        if (!inst) {
 167                err = -ENOMEM;
 168                goto out_unlock;
 169        }
 170
 171        if (!try_module_get(THIS_MODULE)) {
 172                kfree(inst);
 173                err = -EAGAIN;
 174                goto out_unlock;
 175        }
 176
 177        INIT_HLIST_NODE(&inst->hlist);
 178        spin_lock_init(&inst->lock);
 179        /* needs to be two, since we _put() after creation */
 180        atomic_set(&inst->use, 2);
 181
 182        setup_timer(&inst->timer, nfulnl_timer, (unsigned long)inst);
 183
 184        inst->net = get_net(net);
 185        inst->peer_user_ns = user_ns;
 186        inst->peer_portid = portid;
 187        inst->group_num = group_num;
 188
 189        inst->qthreshold        = NFULNL_QTHRESH_DEFAULT;
 190        inst->flushtimeout      = NFULNL_TIMEOUT_DEFAULT;
 191        inst->nlbufsiz          = NFULNL_NLBUFSIZ_DEFAULT;
 192        inst->copy_mode         = NFULNL_COPY_PACKET;
 193        inst->copy_range        = NFULNL_COPY_RANGE_MAX;
 194
 195        hlist_add_head_rcu(&inst->hlist,
 196                       &log->instance_table[instance_hashfn(group_num)]);
 197
 198
 199        spin_unlock_bh(&log->instances_lock);
 200
 201        return inst;
 202
 203out_unlock:
 204        spin_unlock_bh(&log->instances_lock);
 205        return ERR_PTR(err);
 206}
 207
 208static void __nfulnl_flush(struct nfulnl_instance *inst);
 209
 210/* called with BH disabled */
 211static void
 212__instance_destroy(struct nfulnl_instance *inst)
 213{
 214        /* first pull it out of the global list */
 215        hlist_del_rcu(&inst->hlist);
 216
 217        /* then flush all pending packets from skb */
 218
 219        spin_lock(&inst->lock);
 220
 221        /* lockless readers wont be able to use us */
 222        inst->copy_mode = NFULNL_COPY_DISABLED;
 223
 224        if (inst->skb)
 225                __nfulnl_flush(inst);
 226        spin_unlock(&inst->lock);
 227
 228        /* and finally put the refcount */
 229        instance_put(inst);
 230}
 231
 232static inline void
 233instance_destroy(struct nfnl_log_net *log,
 234                 struct nfulnl_instance *inst)
 235{
 236        spin_lock_bh(&log->instances_lock);
 237        __instance_destroy(inst);
 238        spin_unlock_bh(&log->instances_lock);
 239}
 240
 241static int
 242nfulnl_set_mode(struct nfulnl_instance *inst, u_int8_t mode,
 243                  unsigned int range)
 244{
 245        int status = 0;
 246
 247        spin_lock_bh(&inst->lock);
 248
 249        switch (mode) {
 250        case NFULNL_COPY_NONE:
 251        case NFULNL_COPY_META:
 252                inst->copy_mode = mode;
 253                inst->copy_range = 0;
 254                break;
 255
 256        case NFULNL_COPY_PACKET:
 257                inst->copy_mode = mode;
 258                inst->copy_range = min_t(unsigned int,
 259                                         range, NFULNL_COPY_RANGE_MAX);
 260                break;
 261
 262        default:
 263                status = -EINVAL;
 264                break;
 265        }
 266
 267        spin_unlock_bh(&inst->lock);
 268
 269        return status;
 270}
 271
 272static int
 273nfulnl_set_nlbufsiz(struct nfulnl_instance *inst, u_int32_t nlbufsiz)
 274{
 275        int status;
 276
 277        spin_lock_bh(&inst->lock);
 278        if (nlbufsiz < NFULNL_NLBUFSIZ_DEFAULT)
 279                status = -ERANGE;
 280        else if (nlbufsiz > 131072)
 281                status = -ERANGE;
 282        else {
 283                inst->nlbufsiz = nlbufsiz;
 284                status = 0;
 285        }
 286        spin_unlock_bh(&inst->lock);
 287
 288        return status;
 289}
 290
 291static int
 292nfulnl_set_timeout(struct nfulnl_instance *inst, u_int32_t timeout)
 293{
 294        spin_lock_bh(&inst->lock);
 295        inst->flushtimeout = timeout;
 296        spin_unlock_bh(&inst->lock);
 297
 298        return 0;
 299}
 300
 301static int
 302nfulnl_set_qthresh(struct nfulnl_instance *inst, u_int32_t qthresh)
 303{
 304        spin_lock_bh(&inst->lock);
 305        inst->qthreshold = qthresh;
 306        spin_unlock_bh(&inst->lock);
 307
 308        return 0;
 309}
 310
 311static int
 312nfulnl_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
 321static struct sk_buff *
 322nfulnl_alloc_skb(u32 peer_portid, unsigned int inst_size, unsigned int pkt_size)
 323{
 324        struct sk_buff *skb;
 325        unsigned int n;
 326
 327        /* alloc skb which should be big enough for a whole multipart
 328         * message.  WARNING: has to be <= 128k due to slab restrictions */
 329
 330        n = max(inst_size, pkt_size);
 331        skb = nfnetlink_alloc_skb(&init_net, n, peer_portid, GFP_ATOMIC);
 332        if (!skb) {
 333                if (n > pkt_size) {
 334                        /* try to allocate only as much as we need for current
 335                         * packet */
 336
 337                        skb = nfnetlink_alloc_skb(&init_net, pkt_size,
 338                                                  peer_portid, GFP_ATOMIC);
 339                        if (!skb)
 340                                pr_err("nfnetlink_log: can't even alloc %u bytes\n",
 341                                       pkt_size);
 342                }
 343        }
 344
 345        return skb;
 346}
 347
 348static int
 349__nfulnl_send(struct nfulnl_instance *inst)
 350{
 351        int status = -1;
 352
 353        if (inst->qlen > 1) {
 354                struct nlmsghdr *nlh = nlmsg_put(inst->skb, 0, 0,
 355                                                 NLMSG_DONE,
 356                                                 sizeof(struct nfgenmsg),
 357                                                 0);
 358                if (!nlh)
 359                        goto out;
 360        }
 361        status = nfnetlink_unicast(inst->skb, inst->net, inst->peer_portid,
 362                                   MSG_DONTWAIT);
 363
 364        inst->qlen = 0;
 365        inst->skb = NULL;
 366out:
 367        return status;
 368}
 369
 370static void
 371__nfulnl_flush(struct nfulnl_instance *inst)
 372{
 373        /* timer holds a reference */
 374        if (del_timer(&inst->timer))
 375                instance_put(inst);
 376        if (inst->skb)
 377                __nfulnl_send(inst);
 378}
 379
 380static void
 381nfulnl_timer(unsigned long data)
 382{
 383        struct nfulnl_instance *inst = (struct nfulnl_instance *)data;
 384
 385        spin_lock_bh(&inst->lock);
 386        if (inst->skb)
 387                __nfulnl_send(inst);
 388        spin_unlock_bh(&inst->lock);
 389        instance_put(inst);
 390}
 391
 392/* This is an inline function, we don't really care about a long
 393 * list of arguments */
 394static inline int
 395__build_packet_message(struct nfnl_log_net *log,
 396                        struct nfulnl_instance *inst,
 397                        const struct sk_buff *skb,
 398                        unsigned int data_len,
 399                        u_int8_t pf,
 400                        unsigned int hooknum,
 401                        const struct net_device *indev,
 402                        const struct net_device *outdev,
 403                        const char *prefix, unsigned int plen)
 404{
 405        struct nfulnl_msg_packet_hdr pmsg;
 406        struct nlmsghdr *nlh;
 407        struct nfgenmsg *nfmsg;
 408        sk_buff_data_t old_tail = inst->skb->tail;
 409        struct sock *sk;
 410        const unsigned char *hwhdrp;
 411
 412        nlh = nlmsg_put(inst->skb, 0, 0,
 413                        NFNL_SUBSYS_ULOG << 8 | NFULNL_MSG_PACKET,
 414                        sizeof(struct nfgenmsg), 0);
 415        if (!nlh)
 416                return -1;
 417        nfmsg = nlmsg_data(nlh);
 418        nfmsg->nfgen_family = pf;
 419        nfmsg->version = NFNETLINK_V0;
 420        nfmsg->res_id = htons(inst->group_num);
 421
 422        memset(&pmsg, 0, sizeof(pmsg));
 423        pmsg.hw_protocol        = skb->protocol;
 424        pmsg.hook               = hooknum;
 425
 426        if (nla_put(inst->skb, NFULA_PACKET_HDR, sizeof(pmsg), &pmsg))
 427                goto nla_put_failure;
 428
 429        if (prefix &&
 430            nla_put(inst->skb, NFULA_PREFIX, plen, prefix))
 431                goto nla_put_failure;
 432
 433        if (indev) {
 434#ifndef CONFIG_BRIDGE_NETFILTER
 435                if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 436                                 htonl(indev->ifindex)))
 437                        goto nla_put_failure;
 438#else
 439                if (pf == PF_BRIDGE) {
 440                        /* Case 1: outdev is physical input device, we need to
 441                         * look for bridge group (when called from
 442                         * netfilter_bridge) */
 443                        if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
 444                                         htonl(indev->ifindex)) ||
 445                        /* this is the bridge group "brX" */
 446                        /* rcu_read_lock()ed by nf_hook_slow or nf_log_packet */
 447                            nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 448                                         htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
 449                                goto nla_put_failure;
 450                } else {
 451                        /* Case 2: indev is bridge group, we need to look for
 452                         * physical device (when called from ipv4) */
 453                        if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 454                                         htonl(indev->ifindex)))
 455                                goto nla_put_failure;
 456                        if (skb->nf_bridge && skb->nf_bridge->physindev &&
 457                            nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
 458                                         htonl(skb->nf_bridge->physindev->ifindex)))
 459                                goto nla_put_failure;
 460                }
 461#endif
 462        }
 463
 464        if (outdev) {
 465#ifndef CONFIG_BRIDGE_NETFILTER
 466                if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 467                                 htonl(outdev->ifindex)))
 468                        goto nla_put_failure;
 469#else
 470                if (pf == PF_BRIDGE) {
 471                        /* Case 1: outdev is physical output device, we need to
 472                         * look for bridge group (when called from
 473                         * netfilter_bridge) */
 474                        if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
 475                                         htonl(outdev->ifindex)) ||
 476                        /* this is the bridge group "brX" */
 477                        /* rcu_read_lock()ed by nf_hook_slow or nf_log_packet */
 478                            nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 479                                         htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
 480                                goto nla_put_failure;
 481                } else {
 482                        /* Case 2: indev is a bridge group, we need to look
 483                         * for physical device (when called from ipv4) */
 484                        if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 485                                         htonl(outdev->ifindex)))
 486                                goto nla_put_failure;
 487                        if (skb->nf_bridge && skb->nf_bridge->physoutdev &&
 488                            nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
 489                                         htonl(skb->nf_bridge->physoutdev->ifindex)))
 490                                goto nla_put_failure;
 491                }
 492#endif
 493        }
 494
 495        if (skb->mark &&
 496            nla_put_be32(inst->skb, NFULA_MARK, htonl(skb->mark)))
 497                goto nla_put_failure;
 498
 499        if (indev && skb->dev &&
 500            skb->mac_header != skb->network_header) {
 501                struct nfulnl_msg_packet_hw phw;
 502                int len;
 503
 504                memset(&phw, 0, sizeof(phw));
 505                len = dev_parse_header(skb, phw.hw_addr);
 506                if (len > 0) {
 507                        phw.hw_addrlen = htons(len);
 508                        if (nla_put(inst->skb, NFULA_HWADDR, sizeof(phw), &phw))
 509                                goto nla_put_failure;
 510                }
 511        }
 512
 513        if (indev && skb_mac_header_was_set(skb)) {
 514                if (nla_put_be16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type)) ||
 515                    nla_put_be16(inst->skb, NFULA_HWLEN,
 516                                 htons(skb->dev->hard_header_len)))
 517                        goto nla_put_failure;
 518
 519                hwhdrp = skb_mac_header(skb);
 520
 521                if (skb->dev->type == ARPHRD_SIT)
 522                        hwhdrp -= ETH_HLEN;
 523
 524                if (hwhdrp >= skb->head &&
 525                    nla_put(inst->skb, NFULA_HWHEADER,
 526                            skb->dev->hard_header_len, hwhdrp))
 527                        goto nla_put_failure;
 528        }
 529
 530        if (skb->tstamp.tv64) {
 531                struct nfulnl_msg_packet_timestamp ts;
 532                struct timeval tv = ktime_to_timeval(skb->tstamp);
 533                ts.sec = cpu_to_be64(tv.tv_sec);
 534                ts.usec = cpu_to_be64(tv.tv_usec);
 535
 536                if (nla_put(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts))
 537                        goto nla_put_failure;
 538        }
 539
 540        /* UID */
 541        sk = skb->sk;
 542        if (sk && sk->sk_state != TCP_TIME_WAIT) {
 543                read_lock_bh(&sk->sk_callback_lock);
 544                if (sk->sk_socket && sk->sk_socket->file) {
 545                        struct file *file = sk->sk_socket->file;
 546                        const struct cred *cred = file->f_cred;
 547                        struct user_namespace *user_ns = inst->peer_user_ns;
 548                        __be32 uid = htonl(from_kuid_munged(user_ns, cred->fsuid));
 549                        __be32 gid = htonl(from_kgid_munged(user_ns, cred->fsgid));
 550                        read_unlock_bh(&sk->sk_callback_lock);
 551                        if (nla_put_be32(inst->skb, NFULA_UID, uid) ||
 552                            nla_put_be32(inst->skb, NFULA_GID, gid))
 553                                goto nla_put_failure;
 554                } else
 555                        read_unlock_bh(&sk->sk_callback_lock);
 556        }
 557
 558        /* local sequence number */
 559        if ((inst->flags & NFULNL_CFG_F_SEQ) &&
 560            nla_put_be32(inst->skb, NFULA_SEQ, htonl(inst->seq++)))
 561                goto nla_put_failure;
 562
 563        /* global sequence number */
 564        if ((inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) &&
 565            nla_put_be32(inst->skb, NFULA_SEQ_GLOBAL,
 566                         htonl(atomic_inc_return(&log->global_seq))))
 567                goto nla_put_failure;
 568
 569        if (data_len) {
 570                struct nlattr *nla;
 571                int size = nla_attr_size(data_len);
 572
 573                if (skb_tailroom(inst->skb) < nla_total_size(data_len)) {
 574                        printk(KERN_WARNING "nfnetlink_log: no tailroom!\n");
 575                        return -1;
 576                }
 577
 578                nla = (struct nlattr *)skb_put(inst->skb, nla_total_size(data_len));
 579                nla->nla_type = NFULA_PAYLOAD;
 580                nla->nla_len = size;
 581
 582                if (skb_copy_bits(skb, 0, nla_data(nla), data_len))
 583                        BUG();
 584        }
 585
 586        nlh->nlmsg_len = inst->skb->tail - old_tail;
 587        return 0;
 588
 589nla_put_failure:
 590        PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n");
 591        return -1;
 592}
 593
 594#define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
 595
 596static struct nf_loginfo default_loginfo = {
 597        .type =         NF_LOG_TYPE_ULOG,
 598        .u = {
 599                .ulog = {
 600                        .copy_len       = 0xffff,
 601                        .group          = 0,
 602                        .qthreshold     = 1,
 603                },
 604        },
 605};
 606
 607/* log handler for internal netfilter logging api */
 608void
 609nfulnl_log_packet(struct net *net,
 610                  u_int8_t pf,
 611                  unsigned int hooknum,
 612                  const struct sk_buff *skb,
 613                  const struct net_device *in,
 614                  const struct net_device *out,
 615                  const struct nf_loginfo *li_user,
 616                  const char *prefix)
 617{
 618        unsigned int size, data_len;
 619        struct nfulnl_instance *inst;
 620        const struct nf_loginfo *li;
 621        unsigned int qthreshold;
 622        unsigned int plen;
 623        struct nfnl_log_net *log = nfnl_log_pernet(net);
 624
 625        if (li_user && li_user->type == NF_LOG_TYPE_ULOG)
 626                li = li_user;
 627        else
 628                li = &default_loginfo;
 629
 630        inst = instance_lookup_get(log, li->u.ulog.group);
 631        if (!inst)
 632                return;
 633
 634        plen = 0;
 635        if (prefix)
 636                plen = strlen(prefix) + 1;
 637
 638        /* FIXME: do we want to make the size calculation conditional based on
 639         * what is actually present?  way more branches and checks, but more
 640         * memory efficient... */
 641        size =    nlmsg_total_size(sizeof(struct nfgenmsg))
 642                + nla_total_size(sizeof(struct nfulnl_msg_packet_hdr))
 643                + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 644                + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 645#ifdef CONFIG_BRIDGE_NETFILTER
 646                + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 647                + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 648#endif
 649                + nla_total_size(sizeof(u_int32_t))     /* mark */
 650                + nla_total_size(sizeof(u_int32_t))     /* uid */
 651                + nla_total_size(sizeof(u_int32_t))     /* gid */
 652                + nla_total_size(plen)                  /* prefix */
 653                + nla_total_size(sizeof(struct nfulnl_msg_packet_hw))
 654                + nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp));
 655
 656        if (in && skb_mac_header_was_set(skb)) {
 657                size +=   nla_total_size(skb->dev->hard_header_len)
 658                        + nla_total_size(sizeof(u_int16_t))     /* hwtype */
 659                        + nla_total_size(sizeof(u_int16_t));    /* hwlen */
 660        }
 661
 662        spin_lock_bh(&inst->lock);
 663
 664        if (inst->flags & NFULNL_CFG_F_SEQ)
 665                size += nla_total_size(sizeof(u_int32_t));
 666        if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
 667                size += nla_total_size(sizeof(u_int32_t));
 668
 669        qthreshold = inst->qthreshold;
 670        /* per-rule qthreshold overrides per-instance */
 671        if (li->u.ulog.qthreshold)
 672                if (qthreshold > li->u.ulog.qthreshold)
 673                        qthreshold = li->u.ulog.qthreshold;
 674
 675
 676        switch (inst->copy_mode) {
 677        case NFULNL_COPY_META:
 678        case NFULNL_COPY_NONE:
 679                data_len = 0;
 680                break;
 681
 682        case NFULNL_COPY_PACKET:
 683                if (inst->copy_range == 0
 684                    || inst->copy_range > skb->len)
 685                        data_len = skb->len;
 686                else
 687                        data_len = inst->copy_range;
 688
 689                size += nla_total_size(data_len);
 690                break;
 691
 692        case NFULNL_COPY_DISABLED:
 693        default:
 694                goto unlock_and_release;
 695        }
 696
 697        if (inst->skb &&
 698            size > skb_tailroom(inst->skb) - sizeof(struct nfgenmsg)) {
 699                /* either the queue len is too high or we don't have
 700                 * enough room in the skb left. flush to userspace. */
 701                __nfulnl_flush(inst);
 702        }
 703
 704        if (!inst->skb) {
 705                inst->skb = nfulnl_alloc_skb(inst->peer_portid, inst->nlbufsiz,
 706                                             size);
 707                if (!inst->skb)
 708                        goto alloc_failure;
 709        }
 710
 711        inst->qlen++;
 712
 713        __build_packet_message(log, inst, skb, data_len, pf,
 714                                hooknum, in, out, prefix, plen);
 715
 716        if (inst->qlen >= qthreshold)
 717                __nfulnl_flush(inst);
 718        /* timer_pending always called within inst->lock, so there
 719         * is no chance of a race here */
 720        else if (!timer_pending(&inst->timer)) {
 721                instance_get(inst);
 722                inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100);
 723                add_timer(&inst->timer);
 724        }
 725
 726unlock_and_release:
 727        spin_unlock_bh(&inst->lock);
 728        instance_put(inst);
 729        return;
 730
 731alloc_failure:
 732        /* FIXME: statistics */
 733        goto unlock_and_release;
 734}
 735EXPORT_SYMBOL_GPL(nfulnl_log_packet);
 736
 737static int
 738nfulnl_rcv_nl_event(struct notifier_block *this,
 739                   unsigned long event, void *ptr)
 740{
 741        struct netlink_notify *n = ptr;
 742        struct nfnl_log_net *log = nfnl_log_pernet(n->net);
 743
 744        if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
 745                int i;
 746
 747                /* destroy all instances for this portid */
 748                spin_lock_bh(&log->instances_lock);
 749                for  (i = 0; i < INSTANCE_BUCKETS; i++) {
 750                        struct hlist_node *t2;
 751                        struct nfulnl_instance *inst;
 752                        struct hlist_head *head = &log->instance_table[i];
 753
 754                        hlist_for_each_entry_safe(inst, t2, head, hlist) {
 755                                if (n->portid == inst->peer_portid)
 756                                        __instance_destroy(inst);
 757                        }
 758                }
 759                spin_unlock_bh(&log->instances_lock);
 760        }
 761        return NOTIFY_DONE;
 762}
 763
 764static struct notifier_block nfulnl_rtnl_notifier = {
 765        .notifier_call  = nfulnl_rcv_nl_event,
 766};
 767
 768static int
 769nfulnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
 770                   const struct nlmsghdr *nlh,
 771                   const struct nlattr * const nfqa[])
 772{
 773        return -ENOTSUPP;
 774}
 775
 776static struct nf_logger nfulnl_logger __read_mostly = {
 777        .name   = "nfnetlink_log",
 778        .logfn  = &nfulnl_log_packet,
 779        .me     = THIS_MODULE,
 780};
 781
 782static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = {
 783        [NFULA_CFG_CMD]         = { .len = sizeof(struct nfulnl_msg_config_cmd) },
 784        [NFULA_CFG_MODE]        = { .len = sizeof(struct nfulnl_msg_config_mode) },
 785        [NFULA_CFG_TIMEOUT]     = { .type = NLA_U32 },
 786        [NFULA_CFG_QTHRESH]     = { .type = NLA_U32 },
 787        [NFULA_CFG_NLBUFSIZ]    = { .type = NLA_U32 },
 788        [NFULA_CFG_FLAGS]       = { .type = NLA_U16 },
 789};
 790
 791static int
 792nfulnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
 793                   const struct nlmsghdr *nlh,
 794                   const struct nlattr * const nfula[])
 795{
 796        struct nfgenmsg *nfmsg = nlmsg_data(nlh);
 797        u_int16_t group_num = ntohs(nfmsg->res_id);
 798        struct nfulnl_instance *inst;
 799        struct nfulnl_msg_config_cmd *cmd = NULL;
 800        struct net *net = sock_net(ctnl);
 801        struct nfnl_log_net *log = nfnl_log_pernet(net);
 802        int ret = 0;
 803
 804        if (nfula[NFULA_CFG_CMD]) {
 805                u_int8_t pf = nfmsg->nfgen_family;
 806                cmd = nla_data(nfula[NFULA_CFG_CMD]);
 807
 808                /* Commands without queue context */
 809                switch (cmd->command) {
 810                case NFULNL_CFG_CMD_PF_BIND:
 811                        return nf_log_bind_pf(net, pf, &nfulnl_logger);
 812                case NFULNL_CFG_CMD_PF_UNBIND:
 813                        nf_log_unbind_pf(net, pf);
 814                        return 0;
 815                }
 816        }
 817
 818        inst = instance_lookup_get(log, group_num);
 819        if (inst && inst->peer_portid != NETLINK_CB(skb).portid) {
 820                ret = -EPERM;
 821                goto out_put;
 822        }
 823
 824        if (cmd != NULL) {
 825                switch (cmd->command) {
 826                case NFULNL_CFG_CMD_BIND:
 827                        if (inst) {
 828                                ret = -EBUSY;
 829                                goto out_put;
 830                        }
 831
 832                        inst = instance_create(net, group_num,
 833                                               NETLINK_CB(skb).portid,
 834                                               sk_user_ns(NETLINK_CB(skb).sk));
 835                        if (IS_ERR(inst)) {
 836                                ret = PTR_ERR(inst);
 837                                goto out;
 838                        }
 839                        break;
 840                case NFULNL_CFG_CMD_UNBIND:
 841                        if (!inst) {
 842                                ret = -ENODEV;
 843                                goto out;
 844                        }
 845
 846                        instance_destroy(log, inst);
 847                        goto out_put;
 848                default:
 849                        ret = -ENOTSUPP;
 850                        break;
 851                }
 852        }
 853
 854        if (nfula[NFULA_CFG_MODE]) {
 855                struct nfulnl_msg_config_mode *params;
 856                params = nla_data(nfula[NFULA_CFG_MODE]);
 857
 858                if (!inst) {
 859                        ret = -ENODEV;
 860                        goto out;
 861                }
 862                nfulnl_set_mode(inst, params->copy_mode,
 863                                ntohl(params->copy_range));
 864        }
 865
 866        if (nfula[NFULA_CFG_TIMEOUT]) {
 867                __be32 timeout = nla_get_be32(nfula[NFULA_CFG_TIMEOUT]);
 868
 869                if (!inst) {
 870                        ret = -ENODEV;
 871                        goto out;
 872                }
 873                nfulnl_set_timeout(inst, ntohl(timeout));
 874        }
 875
 876        if (nfula[NFULA_CFG_NLBUFSIZ]) {
 877                __be32 nlbufsiz = nla_get_be32(nfula[NFULA_CFG_NLBUFSIZ]);
 878
 879                if (!inst) {
 880                        ret = -ENODEV;
 881                        goto out;
 882                }
 883                nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz));
 884        }
 885
 886        if (nfula[NFULA_CFG_QTHRESH]) {
 887                __be32 qthresh = nla_get_be32(nfula[NFULA_CFG_QTHRESH]);
 888
 889                if (!inst) {
 890                        ret = -ENODEV;
 891                        goto out;
 892                }
 893                nfulnl_set_qthresh(inst, ntohl(qthresh));
 894        }
 895
 896        if (nfula[NFULA_CFG_FLAGS]) {
 897                __be16 flags = nla_get_be16(nfula[NFULA_CFG_FLAGS]);
 898
 899                if (!inst) {
 900                        ret = -ENODEV;
 901                        goto out;
 902                }
 903                nfulnl_set_flags(inst, ntohs(flags));
 904        }
 905
 906out_put:
 907        instance_put(inst);
 908out:
 909        return ret;
 910}
 911
 912static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = {
 913        [NFULNL_MSG_PACKET]     = { .call = nfulnl_recv_unsupp,
 914                                    .attr_count = NFULA_MAX, },
 915        [NFULNL_MSG_CONFIG]     = { .call = nfulnl_recv_config,
 916                                    .attr_count = NFULA_CFG_MAX,
 917                                    .policy = nfula_cfg_policy },
 918};
 919
 920static const struct nfnetlink_subsystem nfulnl_subsys = {
 921        .name           = "log",
 922        .subsys_id      = NFNL_SUBSYS_ULOG,
 923        .cb_count       = NFULNL_MSG_MAX,
 924        .cb             = nfulnl_cb,
 925};
 926
 927#ifdef CONFIG_PROC_FS
 928struct iter_state {
 929        struct seq_net_private p;
 930        unsigned int bucket;
 931};
 932
 933static struct hlist_node *get_first(struct net *net, struct iter_state *st)
 934{
 935        struct nfnl_log_net *log;
 936        if (!st)
 937                return NULL;
 938
 939        log = nfnl_log_pernet(net);
 940
 941        for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
 942                struct hlist_head *head = &log->instance_table[st->bucket];
 943
 944                if (!hlist_empty(head))
 945                        return rcu_dereference_bh(hlist_first_rcu(head));
 946        }
 947        return NULL;
 948}
 949
 950static struct hlist_node *get_next(struct net *net, struct iter_state *st,
 951                                   struct hlist_node *h)
 952{
 953        h = rcu_dereference_bh(hlist_next_rcu(h));
 954        while (!h) {
 955                struct nfnl_log_net *log;
 956                struct hlist_head *head;
 957
 958                if (++st->bucket >= INSTANCE_BUCKETS)
 959                        return NULL;
 960
 961                log = nfnl_log_pernet(net);
 962                head = &log->instance_table[st->bucket];
 963                h = rcu_dereference_bh(hlist_first_rcu(head));
 964        }
 965        return h;
 966}
 967
 968static struct hlist_node *get_idx(struct net *net, struct iter_state *st,
 969                                  loff_t pos)
 970{
 971        struct hlist_node *head;
 972        head = get_first(net, st);
 973
 974        if (head)
 975                while (pos && (head = get_next(net, st, head)))
 976                        pos--;
 977        return pos ? NULL : head;
 978}
 979
 980static void *seq_start(struct seq_file *s, loff_t *pos)
 981        __acquires(rcu_bh)
 982{
 983        rcu_read_lock_bh();
 984        return get_idx(seq_file_net(s), s->private, *pos);
 985}
 986
 987static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
 988{
 989        (*pos)++;
 990        return get_next(seq_file_net(s), s->private, v);
 991}
 992
 993static void seq_stop(struct seq_file *s, void *v)
 994        __releases(rcu_bh)
 995{
 996        rcu_read_unlock_bh();
 997}
 998
 999static int seq_show(struct seq_file *s, void *v)
1000{
1001        const struct nfulnl_instance *inst = v;
1002
1003        return seq_printf(s, "%5d %6d %5d %1d %5d %6d %2d\n",
1004                          inst->group_num,
1005                          inst->peer_portid, inst->qlen,
1006                          inst->copy_mode, inst->copy_range,
1007                          inst->flushtimeout, atomic_read(&inst->use));
1008}
1009
1010static const struct seq_operations nful_seq_ops = {
1011        .start  = seq_start,
1012        .next   = seq_next,
1013        .stop   = seq_stop,
1014        .show   = seq_show,
1015};
1016
1017static int nful_open(struct inode *inode, struct file *file)
1018{
1019        return seq_open_net(inode, file, &nful_seq_ops,
1020                            sizeof(struct iter_state));
1021}
1022
1023static const struct file_operations nful_file_ops = {
1024        .owner   = THIS_MODULE,
1025        .open    = nful_open,
1026        .read    = seq_read,
1027        .llseek  = seq_lseek,
1028        .release = seq_release_net,
1029};
1030
1031#endif /* PROC_FS */
1032
1033static int __net_init nfnl_log_net_init(struct net *net)
1034{
1035        unsigned int i;
1036        struct nfnl_log_net *log = nfnl_log_pernet(net);
1037
1038        for (i = 0; i < INSTANCE_BUCKETS; i++)
1039                INIT_HLIST_HEAD(&log->instance_table[i]);
1040        spin_lock_init(&log->instances_lock);
1041
1042#ifdef CONFIG_PROC_FS
1043        if (!proc_create("nfnetlink_log", 0440,
1044                         net->nf.proc_netfilter, &nful_file_ops))
1045                return -ENOMEM;
1046#endif
1047        return 0;
1048}
1049
1050static void __net_exit nfnl_log_net_exit(struct net *net)
1051{
1052#ifdef CONFIG_PROC_FS
1053        remove_proc_entry("nfnetlink_log", net->nf.proc_netfilter);
1054#endif
1055}
1056
1057static struct pernet_operations nfnl_log_net_ops = {
1058        .init   = nfnl_log_net_init,
1059        .exit   = nfnl_log_net_exit,
1060        .id     = &nfnl_log_net_id,
1061        .size   = sizeof(struct nfnl_log_net),
1062};
1063
1064static int __init nfnetlink_log_init(void)
1065{
1066        int status = -ENOMEM;
1067
1068        /* it's not really all that important to have a random value, so
1069         * we can do this from the init function, even if there hasn't
1070         * been that much entropy yet */
1071        get_random_bytes(&hash_init, sizeof(hash_init));
1072
1073        netlink_register_notifier(&nfulnl_rtnl_notifier);
1074        status = nfnetlink_subsys_register(&nfulnl_subsys);
1075        if (status < 0) {
1076                pr_err("log: failed to create netlink socket\n");
1077                goto cleanup_netlink_notifier;
1078        }
1079
1080        status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger);
1081        if (status < 0) {
1082                pr_err("log: failed to register logger\n");
1083                goto cleanup_subsys;
1084        }
1085
1086        status = register_pernet_subsys(&nfnl_log_net_ops);
1087        if (status < 0) {
1088                pr_err("log: failed to register pernet ops\n");
1089                goto cleanup_logger;
1090        }
1091        return status;
1092
1093cleanup_logger:
1094        nf_log_unregister(&nfulnl_logger);
1095cleanup_subsys:
1096        nfnetlink_subsys_unregister(&nfulnl_subsys);
1097cleanup_netlink_notifier:
1098        netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1099        return status;
1100}
1101
1102static void __exit nfnetlink_log_fini(void)
1103{
1104        unregister_pernet_subsys(&nfnl_log_net_ops);
1105        nf_log_unregister(&nfulnl_logger);
1106        nfnetlink_subsys_unregister(&nfulnl_subsys);
1107        netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1108}
1109
1110MODULE_DESCRIPTION("netfilter userspace logging");
1111MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1112MODULE_LICENSE("GPL");
1113MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG);
1114
1115module_init(nfnetlink_log_init);
1116module_exit(nfnetlink_log_fini);
1117
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.