linux/net/netfilter/nfnetlink_log.c
<<
>>
Prefs
   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#if IS_ENABLED(CONFIG_NF_CONNTRACK)
  47#include <net/netfilter/nf_conntrack.h>
  48#endif
  49
  50#define NFULNL_COPY_DISABLED    0xff
  51#define NFULNL_NLBUFSIZ_DEFAULT NLMSG_GOODSIZE
  52#define NFULNL_TIMEOUT_DEFAULT  100     /* every second */
  53#define NFULNL_QTHRESH_DEFAULT  100     /* 100 packets */
  54/* max packet size is limited by 16-bit struct nfattr nfa_len field */
  55#define NFULNL_COPY_RANGE_MAX   (0xFFFF - NLA_HDRLEN)
  56
  57#define PRINTR(x, args...)      do { if (net_ratelimit()) \
  58                                     printk(x, ## args); } while (0);
  59
  60struct nfulnl_instance {
  61        struct hlist_node hlist;        /* global list of instances */
  62        spinlock_t lock;
  63        refcount_t use;                 /* use count */
  64
  65        unsigned int qlen;              /* number of nlmsgs in skb */
  66        struct sk_buff *skb;            /* pre-allocatd skb */
  67        struct timer_list timer;
  68        struct net *net;
  69        netns_tracker ns_tracker;
  70        struct user_namespace *peer_user_ns;    /* User namespace of the peer process */
  71        u32 peer_portid;                /* PORTID of the peer process */
  72
  73        /* configurable parameters */
  74        unsigned int flushtimeout;      /* timeout until queue flush */
  75        unsigned int nlbufsiz;          /* netlink buffer allocation size */
  76        unsigned int qthreshold;        /* threshold of the queue */
  77        u_int32_t copy_range;
  78        u_int32_t seq;                  /* instance-local sequential counter */
  79        u_int16_t group_num;            /* number of this queue */
  80        u_int16_t flags;
  81        u_int8_t copy_mode;
  82        struct rcu_head rcu;
  83};
  84
  85#define INSTANCE_BUCKETS        16
  86
  87static unsigned int nfnl_log_net_id __read_mostly;
  88
  89struct nfnl_log_net {
  90        spinlock_t instances_lock;
  91        struct hlist_head instance_table[INSTANCE_BUCKETS];
  92        atomic_t global_seq;
  93};
  94
  95static struct nfnl_log_net *nfnl_log_pernet(struct net *net)
  96{
  97        return net_generic(net, nfnl_log_net_id);
  98}
  99
 100static inline u_int8_t instance_hashfn(u_int16_t group_num)
 101{
 102        return ((group_num & 0xff) % INSTANCE_BUCKETS);
 103}
 104
 105static struct nfulnl_instance *
 106__instance_lookup(struct nfnl_log_net *log, u_int16_t group_num)
 107{
 108        struct hlist_head *head;
 109        struct nfulnl_instance *inst;
 110
 111        head = &log->instance_table[instance_hashfn(group_num)];
 112        hlist_for_each_entry_rcu(inst, head, hlist) {
 113                if (inst->group_num == group_num)
 114                        return inst;
 115        }
 116        return NULL;
 117}
 118
 119static inline void
 120instance_get(struct nfulnl_instance *inst)
 121{
 122        refcount_inc(&inst->use);
 123}
 124
 125static struct nfulnl_instance *
 126instance_lookup_get(struct nfnl_log_net *log, u_int16_t group_num)
 127{
 128        struct nfulnl_instance *inst;
 129
 130        rcu_read_lock_bh();
 131        inst = __instance_lookup(log, group_num);
 132        if (inst && !refcount_inc_not_zero(&inst->use))
 133                inst = NULL;
 134        rcu_read_unlock_bh();
 135
 136        return inst;
 137}
 138
 139static void nfulnl_instance_free_rcu(struct rcu_head *head)
 140{
 141        struct nfulnl_instance *inst =
 142                container_of(head, struct nfulnl_instance, rcu);
 143
 144        put_net_track(inst->net, &inst->ns_tracker);
 145        kfree(inst);
 146        module_put(THIS_MODULE);
 147}
 148
 149static void
 150instance_put(struct nfulnl_instance *inst)
 151{
 152        if (inst && refcount_dec_and_test(&inst->use))
 153                call_rcu(&inst->rcu, nfulnl_instance_free_rcu);
 154}
 155
 156static void nfulnl_timer(struct timer_list *t);
 157
 158static struct nfulnl_instance *
 159instance_create(struct net *net, u_int16_t group_num,
 160                u32 portid, struct user_namespace *user_ns)
 161{
 162        struct nfulnl_instance *inst;
 163        struct nfnl_log_net *log = nfnl_log_pernet(net);
 164        int err;
 165
 166        spin_lock_bh(&log->instances_lock);
 167        if (__instance_lookup(log, group_num)) {
 168                err = -EEXIST;
 169                goto out_unlock;
 170        }
 171
 172        inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
 173        if (!inst) {
 174                err = -ENOMEM;
 175                goto out_unlock;
 176        }
 177
 178        if (!try_module_get(THIS_MODULE)) {
 179                kfree(inst);
 180                err = -EAGAIN;
 181                goto out_unlock;
 182        }
 183
 184        INIT_HLIST_NODE(&inst->hlist);
 185        spin_lock_init(&inst->lock);
 186        /* needs to be two, since we _put() after creation */
 187        refcount_set(&inst->use, 2);
 188
 189        timer_setup(&inst->timer, nfulnl_timer, 0);
 190
 191        inst->net = get_net_track(net, &inst->ns_tracker, GFP_ATOMIC);
 192        inst->peer_user_ns = user_ns;
 193        inst->peer_portid = portid;
 194        inst->group_num = group_num;
 195
 196        inst->qthreshold        = NFULNL_QTHRESH_DEFAULT;
 197        inst->flushtimeout      = NFULNL_TIMEOUT_DEFAULT;
 198        inst->nlbufsiz          = NFULNL_NLBUFSIZ_DEFAULT;
 199        inst->copy_mode         = NFULNL_COPY_PACKET;
 200        inst->copy_range        = NFULNL_COPY_RANGE_MAX;
 201
 202        hlist_add_head_rcu(&inst->hlist,
 203                       &log->instance_table[instance_hashfn(group_num)]);
 204
 205
 206        spin_unlock_bh(&log->instances_lock);
 207
 208        return inst;
 209
 210out_unlock:
 211        spin_unlock_bh(&log->instances_lock);
 212        return ERR_PTR(err);
 213}
 214
 215static void __nfulnl_flush(struct nfulnl_instance *inst);
 216
 217/* called with BH disabled */
 218static void
 219__instance_destroy(struct nfulnl_instance *inst)
 220{
 221        /* first pull it out of the global list */
 222        hlist_del_rcu(&inst->hlist);
 223
 224        /* then flush all pending packets from skb */
 225
 226        spin_lock(&inst->lock);
 227
 228        /* lockless readers wont be able to use us */
 229        inst->copy_mode = NFULNL_COPY_DISABLED;
 230
 231        if (inst->skb)
 232                __nfulnl_flush(inst);
 233        spin_unlock(&inst->lock);
 234
 235        /* and finally put the refcount */
 236        instance_put(inst);
 237}
 238
 239static inline void
 240instance_destroy(struct nfnl_log_net *log,
 241                 struct nfulnl_instance *inst)
 242{
 243        spin_lock_bh(&log->instances_lock);
 244        __instance_destroy(inst);
 245        spin_unlock_bh(&log->instances_lock);
 246}
 247
 248static int
 249nfulnl_set_mode(struct nfulnl_instance *inst, u_int8_t mode,
 250                  unsigned int range)
 251{
 252        int status = 0;
 253
 254        spin_lock_bh(&inst->lock);
 255
 256        switch (mode) {
 257        case NFULNL_COPY_NONE:
 258        case NFULNL_COPY_META:
 259                inst->copy_mode = mode;
 260                inst->copy_range = 0;
 261                break;
 262
 263        case NFULNL_COPY_PACKET:
 264                inst->copy_mode = mode;
 265                if (range == 0)
 266                        range = NFULNL_COPY_RANGE_MAX;
 267                inst->copy_range = min_t(unsigned int,
 268                                         range, NFULNL_COPY_RANGE_MAX);
 269                break;
 270
 271        default:
 272                status = -EINVAL;
 273                break;
 274        }
 275
 276        spin_unlock_bh(&inst->lock);
 277
 278        return status;
 279}
 280
 281static int
 282nfulnl_set_nlbufsiz(struct nfulnl_instance *inst, u_int32_t nlbufsiz)
 283{
 284        int status;
 285
 286        spin_lock_bh(&inst->lock);
 287        if (nlbufsiz < NFULNL_NLBUFSIZ_DEFAULT)
 288                status = -ERANGE;
 289        else if (nlbufsiz > 131072)
 290                status = -ERANGE;
 291        else {
 292                inst->nlbufsiz = nlbufsiz;
 293                status = 0;
 294        }
 295        spin_unlock_bh(&inst->lock);
 296
 297        return status;
 298}
 299
 300static void
 301nfulnl_set_timeout(struct nfulnl_instance *inst, u_int32_t timeout)
 302{
 303        spin_lock_bh(&inst->lock);
 304        inst->flushtimeout = timeout;
 305        spin_unlock_bh(&inst->lock);
 306}
 307
 308static void
 309nfulnl_set_qthresh(struct nfulnl_instance *inst, u_int32_t qthresh)
 310{
 311        spin_lock_bh(&inst->lock);
 312        inst->qthreshold = qthresh;
 313        spin_unlock_bh(&inst->lock);
 314}
 315
 316static int
 317nfulnl_set_flags(struct nfulnl_instance *inst, u_int16_t flags)
 318{
 319        spin_lock_bh(&inst->lock);
 320        inst->flags = flags;
 321        spin_unlock_bh(&inst->lock);
 322
 323        return 0;
 324}
 325
 326static struct sk_buff *
 327nfulnl_alloc_skb(struct net *net, u32 peer_portid, unsigned int inst_size,
 328                 unsigned int pkt_size)
 329{
 330        struct sk_buff *skb;
 331        unsigned int n;
 332
 333        /* alloc skb which should be big enough for a whole multipart
 334         * message.  WARNING: has to be <= 128k due to slab restrictions */
 335
 336        n = max(inst_size, pkt_size);
 337        skb = alloc_skb(n, GFP_ATOMIC | __GFP_NOWARN);
 338        if (!skb) {
 339                if (n > pkt_size) {
 340                        /* try to allocate only as much as we need for current
 341                         * packet */
 342
 343                        skb = alloc_skb(pkt_size, GFP_ATOMIC);
 344                }
 345        }
 346
 347        return skb;
 348}
 349
 350static void
 351__nfulnl_send(struct nfulnl_instance *inst)
 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 (WARN_ONCE(!nlh, "bad nlskb size: %u, tailroom %d\n",
 359                              inst->skb->len, skb_tailroom(inst->skb))) {
 360                        kfree_skb(inst->skb);
 361                        goto out;
 362                }
 363        }
 364        nfnetlink_unicast(inst->skb, inst->net, inst->peer_portid);
 365out:
 366        inst->qlen = 0;
 367        inst->skb = NULL;
 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(struct timer_list *t)
 382{
 383        struct nfulnl_instance *inst = from_timer(inst, t, timer);
 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
 392static u32 nfulnl_get_bridge_size(const struct sk_buff *skb)
 393{
 394        u32 size = 0;
 395
 396        if (!skb_mac_header_was_set(skb))
 397                return 0;
 398
 399        if (skb_vlan_tag_present(skb)) {
 400                size += nla_total_size(0); /* nested */
 401                size += nla_total_size(sizeof(u16)); /* id */
 402                size += nla_total_size(sizeof(u16)); /* tag */
 403        }
 404
 405        if (skb->network_header > skb->mac_header)
 406                size += nla_total_size(skb->network_header - skb->mac_header);
 407
 408        return size;
 409}
 410
 411static int nfulnl_put_bridge(struct nfulnl_instance *inst, const struct sk_buff *skb)
 412{
 413        if (!skb_mac_header_was_set(skb))
 414                return 0;
 415
 416        if (skb_vlan_tag_present(skb)) {
 417                struct nlattr *nest;
 418
 419                nest = nla_nest_start(inst->skb, NFULA_VLAN);
 420                if (!nest)
 421                        goto nla_put_failure;
 422
 423                if (nla_put_be16(inst->skb, NFULA_VLAN_TCI, htons(skb->vlan_tci)) ||
 424                    nla_put_be16(inst->skb, NFULA_VLAN_PROTO, skb->vlan_proto))
 425                        goto nla_put_failure;
 426
 427                nla_nest_end(inst->skb, nest);
 428        }
 429
 430        if (skb->mac_header < skb->network_header) {
 431                int len = (int)(skb->network_header - skb->mac_header);
 432
 433                if (nla_put(inst->skb, NFULA_L2HDR, len, skb_mac_header(skb)))
 434                        goto nla_put_failure;
 435        }
 436
 437        return 0;
 438
 439nla_put_failure:
 440        return -1;
 441}
 442
 443/* This is an inline function, we don't really care about a long
 444 * list of arguments */
 445static inline int
 446__build_packet_message(struct nfnl_log_net *log,
 447                        struct nfulnl_instance *inst,
 448                        const struct sk_buff *skb,
 449                        unsigned int data_len,
 450                        u_int8_t pf,
 451                        unsigned int hooknum,
 452                        const struct net_device *indev,
 453                        const struct net_device *outdev,
 454                        const char *prefix, unsigned int plen,
 455                        const struct nfnl_ct_hook *nfnl_ct,
 456                        struct nf_conn *ct, enum ip_conntrack_info ctinfo)
 457{
 458        struct nfulnl_msg_packet_hdr pmsg;
 459        struct nlmsghdr *nlh;
 460        sk_buff_data_t old_tail = inst->skb->tail;
 461        struct sock *sk;
 462        const unsigned char *hwhdrp;
 463        ktime_t tstamp;
 464
 465        nlh = nfnl_msg_put(inst->skb, 0, 0,
 466                           nfnl_msg_type(NFNL_SUBSYS_ULOG, NFULNL_MSG_PACKET),
 467                           0, pf, NFNETLINK_V0, htons(inst->group_num));
 468        if (!nlh)
 469                return -1;
 470
 471        memset(&pmsg, 0, sizeof(pmsg));
 472        pmsg.hw_protocol        = skb->protocol;
 473        pmsg.hook               = hooknum;
 474
 475        if (nla_put(inst->skb, NFULA_PACKET_HDR, sizeof(pmsg), &pmsg))
 476                goto nla_put_failure;
 477
 478        if (prefix &&
 479            nla_put(inst->skb, NFULA_PREFIX, plen, prefix))
 480                goto nla_put_failure;
 481
 482        if (indev) {
 483#if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 484                if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 485                                 htonl(indev->ifindex)))
 486                        goto nla_put_failure;
 487#else
 488                if (pf == PF_BRIDGE) {
 489                        /* Case 1: outdev is physical input device, we need to
 490                         * look for bridge group (when called from
 491                         * netfilter_bridge) */
 492                        if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
 493                                         htonl(indev->ifindex)) ||
 494                        /* this is the bridge group "brX" */
 495                        /* rcu_read_lock()ed by nf_hook_thresh or
 496                         * nf_log_packet.
 497                         */
 498                            nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 499                                         htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
 500                                goto nla_put_failure;
 501                } else {
 502                        struct net_device *physindev;
 503
 504                        /* Case 2: indev is bridge group, we need to look for
 505                         * physical device (when called from ipv4) */
 506                        if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 507                                         htonl(indev->ifindex)))
 508                                goto nla_put_failure;
 509
 510                        physindev = nf_bridge_get_physindev(skb);
 511                        if (physindev &&
 512                            nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
 513                                         htonl(physindev->ifindex)))
 514                                goto nla_put_failure;
 515                }
 516#endif
 517        }
 518
 519        if (outdev) {
 520#if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 521                if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 522                                 htonl(outdev->ifindex)))
 523                        goto nla_put_failure;
 524#else
 525                if (pf == PF_BRIDGE) {
 526                        /* Case 1: outdev is physical output device, we need to
 527                         * look for bridge group (when called from
 528                         * netfilter_bridge) */
 529                        if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
 530                                         htonl(outdev->ifindex)) ||
 531                        /* this is the bridge group "brX" */
 532                        /* rcu_read_lock()ed by nf_hook_thresh or
 533                         * nf_log_packet.
 534                         */
 535                            nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 536                                         htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
 537                                goto nla_put_failure;
 538                } else {
 539                        struct net_device *physoutdev;
 540
 541                        /* Case 2: indev is a bridge group, we need to look
 542                         * for physical device (when called from ipv4) */
 543                        if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 544                                         htonl(outdev->ifindex)))
 545                                goto nla_put_failure;
 546
 547                        physoutdev = nf_bridge_get_physoutdev(skb);
 548                        if (physoutdev &&
 549                            nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
 550                                         htonl(physoutdev->ifindex)))
 551                                goto nla_put_failure;
 552                }
 553#endif
 554        }
 555
 556        if (skb->mark &&
 557            nla_put_be32(inst->skb, NFULA_MARK, htonl(skb->mark)))
 558                goto nla_put_failure;
 559
 560        if (indev && skb->dev &&
 561            skb_mac_header_was_set(skb) &&
 562            skb_mac_header_len(skb) != 0) {
 563                struct nfulnl_msg_packet_hw phw;
 564                int len;
 565
 566                memset(&phw, 0, sizeof(phw));
 567                len = dev_parse_header(skb, phw.hw_addr);
 568                if (len > 0) {
 569                        phw.hw_addrlen = htons(len);
 570                        if (nla_put(inst->skb, NFULA_HWADDR, sizeof(phw), &phw))
 571                                goto nla_put_failure;
 572                }
 573        }
 574
 575        if (indev && skb_mac_header_was_set(skb)) {
 576                if (nla_put_be16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type)) ||
 577                    nla_put_be16(inst->skb, NFULA_HWLEN,
 578                                 htons(skb->dev->hard_header_len)))
 579                        goto nla_put_failure;
 580
 581                hwhdrp = skb_mac_header(skb);
 582
 583                if (skb->dev->type == ARPHRD_SIT)
 584                        hwhdrp -= ETH_HLEN;
 585
 586                if (hwhdrp >= skb->head &&
 587                    nla_put(inst->skb, NFULA_HWHEADER,
 588                            skb->dev->hard_header_len, hwhdrp))
 589                        goto nla_put_failure;
 590        }
 591
 592        tstamp = skb_tstamp_cond(skb, false);
 593        if (hooknum <= NF_INET_FORWARD && tstamp) {
 594                struct nfulnl_msg_packet_timestamp ts;
 595                struct timespec64 kts = ktime_to_timespec64(tstamp);
 596                ts.sec = cpu_to_be64(kts.tv_sec);
 597                ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
 598
 599                if (nla_put(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts))
 600                        goto nla_put_failure;
 601        }
 602
 603        /* UID */
 604        sk = skb->sk;
 605        if (sk && sk_fullsock(sk)) {
 606                read_lock_bh(&sk->sk_callback_lock);
 607                if (sk->sk_socket && sk->sk_socket->file) {
 608                        struct file *file = sk->sk_socket->file;
 609                        const struct cred *cred = file->f_cred;
 610                        struct user_namespace *user_ns = inst->peer_user_ns;
 611                        __be32 uid = htonl(from_kuid_munged(user_ns, cred->fsuid));
 612                        __be32 gid = htonl(from_kgid_munged(user_ns, cred->fsgid));
 613                        read_unlock_bh(&sk->sk_callback_lock);
 614                        if (nla_put_be32(inst->skb, NFULA_UID, uid) ||
 615                            nla_put_be32(inst->skb, NFULA_GID, gid))
 616                                goto nla_put_failure;
 617                } else
 618                        read_unlock_bh(&sk->sk_callback_lock);
 619        }
 620
 621        /* local sequence number */
 622        if ((inst->flags & NFULNL_CFG_F_SEQ) &&
 623            nla_put_be32(inst->skb, NFULA_SEQ, htonl(inst->seq++)))
 624                goto nla_put_failure;
 625
 626        /* global sequence number */
 627        if ((inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) &&
 628            nla_put_be32(inst->skb, NFULA_SEQ_GLOBAL,
 629                         htonl(atomic_inc_return(&log->global_seq))))
 630                goto nla_put_failure;
 631
 632        if (ct && nfnl_ct->build(inst->skb, ct, ctinfo,
 633                                 NFULA_CT, NFULA_CT_INFO) < 0)
 634                goto nla_put_failure;
 635
 636        if ((pf == NFPROTO_NETDEV || pf == NFPROTO_BRIDGE) &&
 637            nfulnl_put_bridge(inst, skb) < 0)
 638                goto nla_put_failure;
 639
 640        if (data_len) {
 641                struct nlattr *nla;
 642                int size = nla_attr_size(data_len);
 643
 644                if (skb_tailroom(inst->skb) < nla_total_size(data_len))
 645                        goto nla_put_failure;
 646
 647                nla = skb_put(inst->skb, nla_total_size(data_len));
 648                nla->nla_type = NFULA_PAYLOAD;
 649                nla->nla_len = size;
 650
 651                if (skb_copy_bits(skb, 0, nla_data(nla), data_len))
 652                        BUG();
 653        }
 654
 655        nlh->nlmsg_len = inst->skb->tail - old_tail;
 656        return 0;
 657
 658nla_put_failure:
 659        PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n");
 660        return -1;
 661}
 662
 663static const struct nf_loginfo default_loginfo = {
 664        .type =         NF_LOG_TYPE_ULOG,
 665        .u = {
 666                .ulog = {
 667                        .copy_len       = 0xffff,
 668                        .group          = 0,
 669                        .qthreshold     = 1,
 670                },
 671        },
 672};
 673
 674/* log handler for internal netfilter logging api */
 675static void
 676nfulnl_log_packet(struct net *net,
 677                  u_int8_t pf,
 678                  unsigned int hooknum,
 679                  const struct sk_buff *skb,
 680                  const struct net_device *in,
 681                  const struct net_device *out,
 682                  const struct nf_loginfo *li_user,
 683                  const char *prefix)
 684{
 685        size_t size;
 686        unsigned int data_len;
 687        struct nfulnl_instance *inst;
 688        const struct nf_loginfo *li;
 689        unsigned int qthreshold;
 690        unsigned int plen = 0;
 691        struct nfnl_log_net *log = nfnl_log_pernet(net);
 692        const struct nfnl_ct_hook *nfnl_ct = NULL;
 693        struct nf_conn *ct = NULL;
 694        enum ip_conntrack_info ctinfo;
 695
 696        if (li_user && li_user->type == NF_LOG_TYPE_ULOG)
 697                li = li_user;
 698        else
 699                li = &default_loginfo;
 700
 701        inst = instance_lookup_get(log, li->u.ulog.group);
 702        if (!inst)
 703                return;
 704
 705        if (prefix)
 706                plen = strlen(prefix) + 1;
 707
 708        /* FIXME: do we want to make the size calculation conditional based on
 709         * what is actually present?  way more branches and checks, but more
 710         * memory efficient... */
 711        size = nlmsg_total_size(sizeof(struct nfgenmsg))
 712                + nla_total_size(sizeof(struct nfulnl_msg_packet_hdr))
 713                + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 714                + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 715#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 716                + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 717                + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 718#endif
 719                + nla_total_size(sizeof(u_int32_t))     /* mark */
 720                + nla_total_size(sizeof(u_int32_t))     /* uid */
 721                + nla_total_size(sizeof(u_int32_t))     /* gid */
 722                + nla_total_size(plen)                  /* prefix */
 723                + nla_total_size(sizeof(struct nfulnl_msg_packet_hw))
 724                + nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp))
 725                + nla_total_size(sizeof(struct nfgenmsg));      /* NLMSG_DONE */
 726
 727        if (in && skb_mac_header_was_set(skb)) {
 728                size += nla_total_size(skb->dev->hard_header_len)
 729                        + nla_total_size(sizeof(u_int16_t))     /* hwtype */
 730                        + nla_total_size(sizeof(u_int16_t));    /* hwlen */
 731        }
 732
 733        spin_lock_bh(&inst->lock);
 734
 735        if (inst->flags & NFULNL_CFG_F_SEQ)
 736                size += nla_total_size(sizeof(u_int32_t));
 737        if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
 738                size += nla_total_size(sizeof(u_int32_t));
 739#if IS_ENABLED(CONFIG_NF_CONNTRACK)
 740        if (inst->flags & NFULNL_CFG_F_CONNTRACK) {
 741                nfnl_ct = rcu_dereference(nfnl_ct_hook);
 742                if (nfnl_ct != NULL) {
 743                        ct = nf_ct_get(skb, &ctinfo);
 744                        if (ct != NULL)
 745                                size += nfnl_ct->build_size(ct);
 746                }
 747        }
 748#endif
 749        if (pf == NFPROTO_NETDEV || pf == NFPROTO_BRIDGE)
 750                size += nfulnl_get_bridge_size(skb);
 751
 752        qthreshold = inst->qthreshold;
 753        /* per-rule qthreshold overrides per-instance */
 754        if (li->u.ulog.qthreshold)
 755                if (qthreshold > li->u.ulog.qthreshold)
 756                        qthreshold = li->u.ulog.qthreshold;
 757
 758
 759        switch (inst->copy_mode) {
 760        case NFULNL_COPY_META:
 761        case NFULNL_COPY_NONE:
 762                data_len = 0;
 763                break;
 764
 765        case NFULNL_COPY_PACKET:
 766                data_len = inst->copy_range;
 767                if ((li->u.ulog.flags & NF_LOG_F_COPY_LEN) &&
 768                    (li->u.ulog.copy_len < data_len))
 769                        data_len = li->u.ulog.copy_len;
 770
 771                if (data_len > skb->len)
 772                        data_len = skb->len;
 773
 774                size += nla_total_size(data_len);
 775                break;
 776
 777        case NFULNL_COPY_DISABLED:
 778        default:
 779                goto unlock_and_release;
 780        }
 781
 782        if (inst->skb && size > skb_tailroom(inst->skb)) {
 783                /* either the queue len is too high or we don't have
 784                 * enough room in the skb left. flush to userspace. */
 785                __nfulnl_flush(inst);
 786        }
 787
 788        if (!inst->skb) {
 789                inst->skb = nfulnl_alloc_skb(net, inst->peer_portid,
 790                                             inst->nlbufsiz, size);
 791                if (!inst->skb)
 792                        goto alloc_failure;
 793        }
 794
 795        inst->qlen++;
 796
 797        __build_packet_message(log, inst, skb, data_len, pf,
 798                                hooknum, in, out, prefix, plen,
 799                                nfnl_ct, ct, ctinfo);
 800
 801        if (inst->qlen >= qthreshold)
 802                __nfulnl_flush(inst);
 803        /* timer_pending always called within inst->lock, so there
 804         * is no chance of a race here */
 805        else if (!timer_pending(&inst->timer)) {
 806                instance_get(inst);
 807                inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100);
 808                add_timer(&inst->timer);
 809        }
 810
 811unlock_and_release:
 812        spin_unlock_bh(&inst->lock);
 813        instance_put(inst);
 814        return;
 815
 816alloc_failure:
 817        /* FIXME: statistics */
 818        goto unlock_and_release;
 819}
 820
 821static int
 822nfulnl_rcv_nl_event(struct notifier_block *this,
 823                   unsigned long event, void *ptr)
 824{
 825        struct netlink_notify *n = ptr;
 826        struct nfnl_log_net *log = nfnl_log_pernet(n->net);
 827
 828        if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
 829                int i;
 830
 831                /* destroy all instances for this portid */
 832                spin_lock_bh(&log->instances_lock);
 833                for  (i = 0; i < INSTANCE_BUCKETS; i++) {
 834                        struct hlist_node *t2;
 835                        struct nfulnl_instance *inst;
 836                        struct hlist_head *head = &log->instance_table[i];
 837
 838                        hlist_for_each_entry_safe(inst, t2, head, hlist) {
 839                                if (n->portid == inst->peer_portid)
 840                                        __instance_destroy(inst);
 841                        }
 842                }
 843                spin_unlock_bh(&log->instances_lock);
 844        }
 845        return NOTIFY_DONE;
 846}
 847
 848static struct notifier_block nfulnl_rtnl_notifier = {
 849        .notifier_call  = nfulnl_rcv_nl_event,
 850};
 851
 852static int nfulnl_recv_unsupp(struct sk_buff *skb, const struct nfnl_info *info,
 853                              const struct nlattr * const nfula[])
 854{
 855        return -ENOTSUPP;
 856}
 857
 858static struct nf_logger nfulnl_logger __read_mostly = {
 859        .name   = "nfnetlink_log",
 860        .type   = NF_LOG_TYPE_ULOG,
 861        .logfn  = nfulnl_log_packet,
 862        .me     = THIS_MODULE,
 863};
 864
 865static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = {
 866        [NFULA_CFG_CMD]         = { .len = sizeof(struct nfulnl_msg_config_cmd) },
 867        [NFULA_CFG_MODE]        = { .len = sizeof(struct nfulnl_msg_config_mode) },
 868        [NFULA_CFG_TIMEOUT]     = { .type = NLA_U32 },
 869        [NFULA_CFG_QTHRESH]     = { .type = NLA_U32 },
 870        [NFULA_CFG_NLBUFSIZ]    = { .type = NLA_U32 },
 871        [NFULA_CFG_FLAGS]       = { .type = NLA_U16 },
 872};
 873
 874static int nfulnl_recv_config(struct sk_buff *skb, const struct nfnl_info *info,
 875                              const struct nlattr * const nfula[])
 876{
 877        struct nfnl_log_net *log = nfnl_log_pernet(info->net);
 878        u_int16_t group_num = ntohs(info->nfmsg->res_id);
 879        struct nfulnl_msg_config_cmd *cmd = NULL;
 880        struct nfulnl_instance *inst;
 881        u16 flags = 0;
 882        int ret = 0;
 883
 884        if (nfula[NFULA_CFG_CMD]) {
 885                u_int8_t pf = info->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(info->net, pf, &nfulnl_logger);
 892                case NFULNL_CFG_CMD_PF_UNBIND:
 893                        nf_log_unbind_pf(info->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(info->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
 988out_put:
 989        instance_put(inst);
 990out:
 991        return ret;
 992}
 993
 994static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = {
 995        [NFULNL_MSG_PACKET]     = {
 996                .call           = nfulnl_recv_unsupp,
 997                .type           = NFNL_CB_MUTEX,
 998                .attr_count     = NFULA_MAX,
 999        },
1000        [NFULNL_MSG_CONFIG]     = {
1001                .call           = nfulnl_recv_config,
1002                .type           = NFNL_CB_MUTEX,
1003                .attr_count     = NFULA_CFG_MAX,
1004                .policy         = nfula_cfg_policy
1005        },
1006};
1007
1008static const struct nfnetlink_subsystem nfulnl_subsys = {
1009        .name           = "log",
1010        .subsys_id      = NFNL_SUBSYS_ULOG,
1011        .cb_count       = NFULNL_MSG_MAX,
1012        .cb             = nfulnl_cb,
1013};
1014
1015#ifdef CONFIG_PROC_FS
1016struct iter_state {
1017        struct seq_net_private p;
1018        unsigned int bucket;
1019};
1020
1021static struct hlist_node *get_first(struct net *net, struct iter_state *st)
1022{
1023        struct nfnl_log_net *log;
1024        if (!st)
1025                return NULL;
1026
1027        log = nfnl_log_pernet(net);
1028
1029        for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1030                struct hlist_head *head = &log->instance_table[st->bucket];
1031
1032                if (!hlist_empty(head))
1033                        return rcu_dereference_bh(hlist_first_rcu(head));
1034        }
1035        return NULL;
1036}
1037
1038static struct hlist_node *get_next(struct net *net, struct iter_state *st,
1039                                   struct hlist_node *h)
1040{
1041        h = rcu_dereference_bh(hlist_next_rcu(h));
1042        while (!h) {
1043                struct nfnl_log_net *log;
1044                struct hlist_head *head;
1045
1046                if (++st->bucket >= INSTANCE_BUCKETS)
1047                        return NULL;
1048
1049                log = nfnl_log_pernet(net);
1050                head = &log->instance_table[st->bucket];
1051                h = rcu_dereference_bh(hlist_first_rcu(head));
1052        }
1053        return h;
1054}
1055
1056static struct hlist_node *get_idx(struct net *net, struct iter_state *st,
1057                                  loff_t pos)
1058{
1059        struct hlist_node *head;
1060        head = get_first(net, st);
1061
1062        if (head)
1063                while (pos && (head = get_next(net, st, head)))
1064                        pos--;
1065        return pos ? NULL : head;
1066}
1067
1068static void *seq_start(struct seq_file *s, loff_t *pos)
1069        __acquires(rcu_bh)
1070{
1071        rcu_read_lock_bh();
1072        return get_idx(seq_file_net(s), s->private, *pos);
1073}
1074
1075static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1076{
1077        (*pos)++;
1078        return get_next(seq_file_net(s), s->private, v);
1079}
1080
1081static void seq_stop(struct seq_file *s, void *v)
1082        __releases(rcu_bh)
1083{
1084        rcu_read_unlock_bh();
1085}
1086
1087static int seq_show(struct seq_file *s, void *v)
1088{
1089        const struct nfulnl_instance *inst = v;
1090
1091        seq_printf(s, "%5u %6u %5u %1u %5u %6u %2u\n",
1092                   inst->group_num,
1093                   inst->peer_portid, inst->qlen,
1094                   inst->copy_mode, inst->copy_range,
1095                   inst->flushtimeout, refcount_read(&inst->use));
1096
1097        return 0;
1098}
1099
1100static const struct seq_operations nful_seq_ops = {
1101        .start  = seq_start,
1102        .next   = seq_next,
1103        .stop   = seq_stop,
1104        .show   = seq_show,
1105};
1106#endif /* PROC_FS */
1107
1108static int __net_init nfnl_log_net_init(struct net *net)
1109{
1110        unsigned int i;
1111        struct nfnl_log_net *log = nfnl_log_pernet(net);
1112#ifdef CONFIG_PROC_FS
1113        struct proc_dir_entry *proc;
1114        kuid_t root_uid;
1115        kgid_t root_gid;
1116#endif
1117
1118        for (i = 0; i < INSTANCE_BUCKETS; i++)
1119                INIT_HLIST_HEAD(&log->instance_table[i]);
1120        spin_lock_init(&log->instances_lock);
1121
1122#ifdef CONFIG_PROC_FS
1123        proc = proc_create_net("nfnetlink_log", 0440, net->nf.proc_netfilter,
1124                        &nful_seq_ops, sizeof(struct iter_state));
1125        if (!proc)
1126                return -ENOMEM;
1127
1128        root_uid = make_kuid(net->user_ns, 0);
1129        root_gid = make_kgid(net->user_ns, 0);
1130        if (uid_valid(root_uid) && gid_valid(root_gid))
1131                proc_set_user(proc, root_uid, root_gid);
1132#endif
1133        return 0;
1134}
1135
1136static void __net_exit nfnl_log_net_exit(struct net *net)
1137{
1138        struct nfnl_log_net *log = nfnl_log_pernet(net);
1139        unsigned int i;
1140
1141#ifdef CONFIG_PROC_FS
1142        remove_proc_entry("nfnetlink_log", net->nf.proc_netfilter);
1143#endif
1144        nf_log_unset(net, &nfulnl_logger);
1145        for (i = 0; i < INSTANCE_BUCKETS; i++)
1146                WARN_ON_ONCE(!hlist_empty(&log->instance_table[i]));
1147}
1148
1149static struct pernet_operations nfnl_log_net_ops = {
1150        .init   = nfnl_log_net_init,
1151        .exit   = nfnl_log_net_exit,
1152        .id     = &nfnl_log_net_id,
1153        .size   = sizeof(struct nfnl_log_net),
1154};
1155
1156static int __init nfnetlink_log_init(void)
1157{
1158        int status;
1159
1160        status = register_pernet_subsys(&nfnl_log_net_ops);
1161        if (status < 0) {
1162                pr_err("failed to register pernet ops\n");
1163                goto out;
1164        }
1165
1166        netlink_register_notifier(&nfulnl_rtnl_notifier);
1167        status = nfnetlink_subsys_register(&nfulnl_subsys);
1168        if (status < 0) {
1169                pr_err("failed to create netlink socket\n");
1170                goto cleanup_netlink_notifier;
1171        }
1172
1173        status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger);
1174        if (status < 0) {
1175                pr_err("failed to register logger\n");
1176                goto cleanup_subsys;
1177        }
1178
1179        return status;
1180
1181cleanup_subsys:
1182        nfnetlink_subsys_unregister(&nfulnl_subsys);
1183cleanup_netlink_notifier:
1184        netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1185        unregister_pernet_subsys(&nfnl_log_net_ops);
1186out:
1187        return status;
1188}
1189
1190static void __exit nfnetlink_log_fini(void)
1191{
1192        nfnetlink_subsys_unregister(&nfulnl_subsys);
1193        netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1194        unregister_pernet_subsys(&nfnl_log_net_ops);
1195        nf_log_unregister(&nfulnl_logger);
1196}
1197
1198MODULE_DESCRIPTION("netfilter userspace logging");
1199MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1200MODULE_LICENSE("GPL");
1201MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG);
1202MODULE_ALIAS_NF_LOGGER(AF_INET, 1);
1203MODULE_ALIAS_NF_LOGGER(AF_INET6, 1);
1204MODULE_ALIAS_NF_LOGGER(AF_BRIDGE, 1);
1205MODULE_ALIAS_NF_LOGGER(3, 1); /* NFPROTO_ARP */
1206MODULE_ALIAS_NF_LOGGER(5, 1); /* NFPROTO_NETDEV */
1207
1208module_init(nfnetlink_log_init);
1209module_exit(nfnetlink_log_fini);
1210