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#define NFULNL_COPY_DISABLED    0xff
  47#define NFULNL_NLBUFSIZ_DEFAULT NLMSG_GOODSIZE
  48#define NFULNL_TIMEOUT_DEFAULT  100     /* every second */
  49#define NFULNL_QTHRESH_DEFAULT  100     /* 100 packets */
  50/* max packet size is limited by 16-bit struct nfattr nfa_len field */
  51#define NFULNL_COPY_RANGE_MAX   (0xFFFF - NLA_HDRLEN)
  52
  53#define PRINTR(x, args...)      do { if (net_ratelimit()) \
  54                                     printk(x, ## args); } while (0);
  55
  56struct nfulnl_instance {
  57        struct hlist_node hlist;        /* global list of instances */
  58        spinlock_t lock;
  59        refcount_t use;                 /* use count */
  60
  61        unsigned int qlen;              /* number of nlmsgs in skb */
  62        struct sk_buff *skb;            /* pre-allocatd skb */
  63        struct timer_list timer;
  64        struct net *net;
  65        struct user_namespace *peer_user_ns;    /* User namespace of the peer process */
  66        u32 peer_portid;                /* PORTID of the peer process */
  67
  68        /* configurable parameters */
  69        unsigned int flushtimeout;      /* timeout until queue flush */
  70        unsigned int nlbufsiz;          /* netlink buffer allocation size */
  71        unsigned int qthreshold;        /* threshold of the queue */
  72        u_int32_t copy_range;
  73        u_int32_t seq;                  /* instance-local sequential counter */
  74        u_int16_t group_num;            /* number of this queue */
  75        u_int16_t flags;
  76        u_int8_t copy_mode;
  77        struct rcu_head rcu;
  78};
  79
  80#define INSTANCE_BUCKETS        16
  81
  82static unsigned int nfnl_log_net_id __read_mostly;
  83
  84struct nfnl_log_net {
  85        spinlock_t instances_lock;
  86        struct hlist_head instance_table[INSTANCE_BUCKETS];
  87        atomic_t global_seq;
  88};
  89
  90static struct nfnl_log_net *nfnl_log_pernet(struct net *net)
  91{
  92        return net_generic(net, nfnl_log_net_id);
  93}
  94
  95static inline u_int8_t instance_hashfn(u_int16_t group_num)
  96{
  97        return ((group_num & 0xff) % INSTANCE_BUCKETS);
  98}
  99
 100static struct nfulnl_instance *
 101__instance_lookup(struct nfnl_log_net *log, u_int16_t group_num)
 102{
 103        struct hlist_head *head;
 104        struct nfulnl_instance *inst;
 105
 106        head = &log->instance_table[instance_hashfn(group_num)];
 107        hlist_for_each_entry_rcu(inst, head, hlist) {
 108                if (inst->group_num == group_num)
 109                        return inst;
 110        }
 111        return NULL;
 112}
 113
 114static inline void
 115instance_get(struct nfulnl_instance *inst)
 116{
 117        refcount_inc(&inst->use);
 118}
 119
 120static struct nfulnl_instance *
 121instance_lookup_get(struct nfnl_log_net *log, u_int16_t group_num)
 122{
 123        struct nfulnl_instance *inst;
 124
 125        rcu_read_lock_bh();
 126        inst = __instance_lookup(log, group_num);
 127        if (inst && !refcount_inc_not_zero(&inst->use))
 128                inst = NULL;
 129        rcu_read_unlock_bh();
 130
 131        return inst;
 132}
 133
 134static void nfulnl_instance_free_rcu(struct rcu_head *head)
 135{
 136        struct nfulnl_instance *inst =
 137                container_of(head, struct nfulnl_instance, rcu);
 138
 139        put_net(inst->net);
 140        kfree(inst);
 141        module_put(THIS_MODULE);
 142}
 143
 144static void
 145instance_put(struct nfulnl_instance *inst)
 146{
 147        if (inst && refcount_dec_and_test(&inst->use))
 148                call_rcu(&inst->rcu, nfulnl_instance_free_rcu);
 149}
 150
 151static void nfulnl_timer(struct timer_list *t);
 152
 153static struct nfulnl_instance *
 154instance_create(struct net *net, u_int16_t group_num,
 155                u32 portid, struct user_namespace *user_ns)
 156{
 157        struct nfulnl_instance *inst;
 158        struct nfnl_log_net *log = nfnl_log_pernet(net);
 159        int err;
 160
 161        spin_lock_bh(&log->instances_lock);
 162        if (__instance_lookup(log, group_num)) {
 163                err = -EEXIST;
 164                goto out_unlock;
 165        }
 166
 167        inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
 168        if (!inst) {
 169                err = -ENOMEM;
 170                goto out_unlock;
 171        }
 172
 173        if (!try_module_get(THIS_MODULE)) {
 174                kfree(inst);
 175                err = -EAGAIN;
 176                goto out_unlock;
 177        }
 178
 179        INIT_HLIST_NODE(&inst->hlist);
 180        spin_lock_init(&inst->lock);
 181        /* needs to be two, since we _put() after creation */
 182        refcount_set(&inst->use, 2);
 183
 184        timer_setup(&inst->timer, nfulnl_timer, 0);
 185
 186        inst->net = get_net(net);
 187        inst->peer_user_ns = user_ns;
 188        inst->peer_portid = portid;
 189        inst->group_num = group_num;
 190
 191        inst->qthreshold        = NFULNL_QTHRESH_DEFAULT;
 192        inst->flushtimeout      = NFULNL_TIMEOUT_DEFAULT;
 193        inst->nlbufsiz          = NFULNL_NLBUFSIZ_DEFAULT;
 194        inst->copy_mode         = NFULNL_COPY_PACKET;
 195        inst->copy_range        = NFULNL_COPY_RANGE_MAX;
 196
 197        hlist_add_head_rcu(&inst->hlist,
 198                       &log->instance_table[instance_hashfn(group_num)]);
 199
 200
 201        spin_unlock_bh(&log->instances_lock);
 202
 203        return inst;
 204
 205out_unlock:
 206        spin_unlock_bh(&log->instances_lock);
 207        return ERR_PTR(err);
 208}
 209
 210static void __nfulnl_flush(struct nfulnl_instance *inst);
 211
 212/* called with BH disabled */
 213static void
 214__instance_destroy(struct nfulnl_instance *inst)
 215{
 216        /* first pull it out of the global list */
 217        hlist_del_rcu(&inst->hlist);
 218
 219        /* then flush all pending packets from skb */
 220
 221        spin_lock(&inst->lock);
 222
 223        /* lockless readers wont be able to use us */
 224        inst->copy_mode = NFULNL_COPY_DISABLED;
 225
 226        if (inst->skb)
 227                __nfulnl_flush(inst);
 228        spin_unlock(&inst->lock);
 229
 230        /* and finally put the refcount */
 231        instance_put(inst);
 232}
 233
 234static inline void
 235instance_destroy(struct nfnl_log_net *log,
 236                 struct nfulnl_instance *inst)
 237{
 238        spin_lock_bh(&log->instances_lock);
 239        __instance_destroy(inst);
 240        spin_unlock_bh(&log->instances_lock);
 241}
 242
 243static int
 244nfulnl_set_mode(struct nfulnl_instance *inst, u_int8_t mode,
 245                  unsigned int range)
 246{
 247        int status = 0;
 248
 249        spin_lock_bh(&inst->lock);
 250
 251        switch (mode) {
 252        case NFULNL_COPY_NONE:
 253        case NFULNL_COPY_META:
 254                inst->copy_mode = mode;
 255                inst->copy_range = 0;
 256                break;
 257
 258        case NFULNL_COPY_PACKET:
 259                inst->copy_mode = mode;
 260                if (range == 0)
 261                        range = NFULNL_COPY_RANGE_MAX;
 262                inst->copy_range = min_t(unsigned int,
 263                                         range, NFULNL_COPY_RANGE_MAX);
 264                break;
 265
 266        default:
 267                status = -EINVAL;
 268                break;
 269        }
 270
 271        spin_unlock_bh(&inst->lock);
 272
 273        return status;
 274}
 275
 276static int
 277nfulnl_set_nlbufsiz(struct nfulnl_instance *inst, u_int32_t nlbufsiz)
 278{
 279        int status;
 280
 281        spin_lock_bh(&inst->lock);
 282        if (nlbufsiz < NFULNL_NLBUFSIZ_DEFAULT)
 283                status = -ERANGE;
 284        else if (nlbufsiz > 131072)
 285                status = -ERANGE;
 286        else {
 287                inst->nlbufsiz = nlbufsiz;
 288                status = 0;
 289        }
 290        spin_unlock_bh(&inst->lock);
 291
 292        return status;
 293}
 294
 295static void
 296nfulnl_set_timeout(struct nfulnl_instance *inst, u_int32_t timeout)
 297{
 298        spin_lock_bh(&inst->lock);
 299        inst->flushtimeout = timeout;
 300        spin_unlock_bh(&inst->lock);
 301}
 302
 303static void
 304nfulnl_set_qthresh(struct nfulnl_instance *inst, u_int32_t qthresh)
 305{
 306        spin_lock_bh(&inst->lock);
 307        inst->qthreshold = qthresh;
 308        spin_unlock_bh(&inst->lock);
 309}
 310
 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(struct net *net, u32 peer_portid, unsigned int inst_size,
 323                 unsigned int pkt_size)
 324{
 325        struct sk_buff *skb;
 326        unsigned int n;
 327
 328        /* alloc skb which should be big enough for a whole multipart
 329         * message.  WARNING: has to be <= 128k due to slab restrictions */
 330
 331        n = max(inst_size, pkt_size);
 332        skb = alloc_skb(n, GFP_ATOMIC | __GFP_NOWARN);
 333        if (!skb) {
 334                if (n > pkt_size) {
 335                        /* try to allocate only as much as we need for current
 336                         * packet */
 337
 338                        skb = alloc_skb(pkt_size, GFP_ATOMIC);
 339                }
 340        }
 341
 342        return skb;
 343}
 344
 345static void
 346__nfulnl_send(struct nfulnl_instance *inst)
 347{
 348        if (inst->qlen > 1) {
 349                struct nlmsghdr *nlh = nlmsg_put(inst->skb, 0, 0,
 350                                                 NLMSG_DONE,
 351                                                 sizeof(struct nfgenmsg),
 352                                                 0);
 353                if (WARN_ONCE(!nlh, "bad nlskb size: %u, tailroom %d\n",
 354                              inst->skb->len, skb_tailroom(inst->skb))) {
 355                        kfree_skb(inst->skb);
 356                        goto out;
 357                }
 358        }
 359        nfnetlink_unicast(inst->skb, inst->net, inst->peer_portid,
 360                          MSG_DONTWAIT);
 361out:
 362        inst->qlen = 0;
 363        inst->skb = NULL;
 364}
 365
 366static void
 367__nfulnl_flush(struct nfulnl_instance *inst)
 368{
 369        /* timer holds a reference */
 370        if (del_timer(&inst->timer))
 371                instance_put(inst);
 372        if (inst->skb)
 373                __nfulnl_send(inst);
 374}
 375
 376static void
 377nfulnl_timer(struct timer_list *t)
 378{
 379        struct nfulnl_instance *inst = from_timer(inst, t, timer);
 380
 381        spin_lock_bh(&inst->lock);
 382        if (inst->skb)
 383                __nfulnl_send(inst);
 384        spin_unlock_bh(&inst->lock);
 385        instance_put(inst);
 386}
 387
 388static u32 nfulnl_get_bridge_size(const struct sk_buff *skb)
 389{
 390        u32 size = 0;
 391
 392        if (!skb_mac_header_was_set(skb))
 393                return 0;
 394
 395        if (skb_vlan_tag_present(skb)) {
 396                size += nla_total_size(0); /* nested */
 397                size += nla_total_size(sizeof(u16)); /* id */
 398                size += nla_total_size(sizeof(u16)); /* tag */
 399        }
 400
 401        if (skb->network_header > skb->mac_header)
 402                size += nla_total_size(skb->network_header - skb->mac_header);
 403
 404        return size;
 405}
 406
 407static int nfulnl_put_bridge(struct nfulnl_instance *inst, const struct sk_buff *skb)
 408{
 409        if (!skb_mac_header_was_set(skb))
 410                return 0;
 411
 412        if (skb_vlan_tag_present(skb)) {
 413                struct nlattr *nest;
 414
 415                nest = nla_nest_start(inst->skb, NFULA_VLAN);
 416                if (!nest)
 417                        goto nla_put_failure;
 418
 419                if (nla_put_be16(inst->skb, NFULA_VLAN_TCI, htons(skb->vlan_tci)) ||
 420                    nla_put_be16(inst->skb, NFULA_VLAN_PROTO, skb->vlan_proto))
 421                        goto nla_put_failure;
 422
 423                nla_nest_end(inst->skb, nest);
 424        }
 425
 426        if (skb->mac_header < skb->network_header) {
 427                int len = (int)(skb->network_header - skb->mac_header);
 428
 429                if (nla_put(inst->skb, NFULA_L2HDR, len, skb_mac_header(skb)))
 430                        goto nla_put_failure;
 431        }
 432
 433        return 0;
 434
 435nla_put_failure:
 436        return -1;
 437}
 438
 439/* This is an inline function, we don't really care about a long
 440 * list of arguments */
 441static inline int
 442__build_packet_message(struct nfnl_log_net *log,
 443                        struct nfulnl_instance *inst,
 444                        const struct sk_buff *skb,
 445                        unsigned int data_len,
 446                        u_int8_t pf,
 447                        unsigned int hooknum,
 448                        const struct net_device *indev,
 449                        const struct net_device *outdev,
 450                        const char *prefix, unsigned int plen,
 451                        const struct nfnl_ct_hook *nfnl_ct,
 452                        struct nf_conn *ct, enum ip_conntrack_info ctinfo)
 453{
 454        struct nfulnl_msg_packet_hdr pmsg;
 455        struct nlmsghdr *nlh;
 456        struct nfgenmsg *nfmsg;
 457        sk_buff_data_t old_tail = inst->skb->tail;
 458        struct sock *sk;
 459        const unsigned char *hwhdrp;
 460
 461        nlh = nlmsg_put(inst->skb, 0, 0,
 462                        nfnl_msg_type(NFNL_SUBSYS_ULOG, NFULNL_MSG_PACKET),
 463                        sizeof(struct nfgenmsg), 0);
 464        if (!nlh)
 465                return -1;
 466        nfmsg = nlmsg_data(nlh);
 467        nfmsg->nfgen_family = pf;
 468        nfmsg->version = NFNETLINK_V0;
 469        nfmsg->res_id = htons(inst->group_num);
 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 != skb->network_header) {
 562                struct nfulnl_msg_packet_hw phw;
 563                int len;
 564
 565                memset(&phw, 0, sizeof(phw));
 566                len = dev_parse_header(skb, phw.hw_addr);
 567                if (len > 0) {
 568                        phw.hw_addrlen = htons(len);
 569                        if (nla_put(inst->skb, NFULA_HWADDR, sizeof(phw), &phw))
 570                                goto nla_put_failure;
 571                }
 572        }
 573
 574        if (indev && skb_mac_header_was_set(skb)) {
 575                if (nla_put_be16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type)) ||
 576                    nla_put_be16(inst->skb, NFULA_HWLEN,
 577                                 htons(skb->dev->hard_header_len)))
 578                        goto nla_put_failure;
 579
 580                hwhdrp = skb_mac_header(skb);
 581
 582                if (skb->dev->type == ARPHRD_SIT)
 583                        hwhdrp -= ETH_HLEN;
 584
 585                if (hwhdrp >= skb->head &&
 586                    nla_put(inst->skb, NFULA_HWHEADER,
 587                            skb->dev->hard_header_len, hwhdrp))
 588                        goto nla_put_failure;
 589        }
 590
 591        if (hooknum <= NF_INET_FORWARD && skb->tstamp) {
 592                struct nfulnl_msg_packet_timestamp ts;
 593                struct timespec64 kts = ktime_to_timespec64(skb->tstamp);
 594                ts.sec = cpu_to_be64(kts.tv_sec);
 595                ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
 596
 597                if (nla_put(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts))
 598                        goto nla_put_failure;
 599        }
 600
 601        /* UID */
 602        sk = skb->sk;
 603        if (sk && sk_fullsock(sk)) {
 604                read_lock_bh(&sk->sk_callback_lock);
 605                if (sk->sk_socket && sk->sk_socket->file) {
 606                        struct file *file = sk->sk_socket->file;
 607                        const struct cred *cred = file->f_cred;
 608                        struct user_namespace *user_ns = inst->peer_user_ns;
 609                        __be32 uid = htonl(from_kuid_munged(user_ns, cred->fsuid));
 610                        __be32 gid = htonl(from_kgid_munged(user_ns, cred->fsgid));
 611                        read_unlock_bh(&sk->sk_callback_lock);
 612                        if (nla_put_be32(inst->skb, NFULA_UID, uid) ||
 613                            nla_put_be32(inst->skb, NFULA_GID, gid))
 614                                goto nla_put_failure;
 615                } else
 616                        read_unlock_bh(&sk->sk_callback_lock);
 617        }
 618
 619        /* local sequence number */
 620        if ((inst->flags & NFULNL_CFG_F_SEQ) &&
 621            nla_put_be32(inst->skb, NFULA_SEQ, htonl(inst->seq++)))
 622                goto nla_put_failure;
 623
 624        /* global sequence number */
 625        if ((inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) &&
 626            nla_put_be32(inst->skb, NFULA_SEQ_GLOBAL,
 627                         htonl(atomic_inc_return(&log->global_seq))))
 628                goto nla_put_failure;
 629
 630        if (ct && nfnl_ct->build(inst->skb, ct, ctinfo,
 631                                 NFULA_CT, NFULA_CT_INFO) < 0)
 632                goto nla_put_failure;
 633
 634        if ((pf == NFPROTO_NETDEV || pf == NFPROTO_BRIDGE) &&
 635            nfulnl_put_bridge(inst, skb) < 0)
 636                goto nla_put_failure;
 637
 638        if (data_len) {
 639                struct nlattr *nla;
 640                int size = nla_attr_size(data_len);
 641
 642                if (skb_tailroom(inst->skb) < nla_total_size(data_len))
 643                        goto nla_put_failure;
 644
 645                nla = skb_put(inst->skb, nla_total_size(data_len));
 646                nla->nla_type = NFULA_PAYLOAD;
 647                nla->nla_len = size;
 648
 649                if (skb_copy_bits(skb, 0, nla_data(nla), data_len))
 650                        BUG();
 651        }
 652
 653        nlh->nlmsg_len = inst->skb->tail - old_tail;
 654        return 0;
 655
 656nla_put_failure:
 657        PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n");
 658        return -1;
 659}
 660
 661static const struct nf_loginfo default_loginfo = {
 662        .type =         NF_LOG_TYPE_ULOG,
 663        .u = {
 664                .ulog = {
 665                        .copy_len       = 0xffff,
 666                        .group          = 0,
 667                        .qthreshold     = 1,
 668                },
 669        },
 670};
 671
 672/* log handler for internal netfilter logging api */
 673static void
 674nfulnl_log_packet(struct net *net,
 675                  u_int8_t pf,
 676                  unsigned int hooknum,
 677                  const struct sk_buff *skb,
 678                  const struct net_device *in,
 679                  const struct net_device *out,
 680                  const struct nf_loginfo *li_user,
 681                  const char *prefix)
 682{
 683        size_t size;
 684        unsigned int data_len;
 685        struct nfulnl_instance *inst;
 686        const struct nf_loginfo *li;
 687        unsigned int qthreshold;
 688        unsigned int plen = 0;
 689        struct nfnl_log_net *log = nfnl_log_pernet(net);
 690        const struct nfnl_ct_hook *nfnl_ct = NULL;
 691        struct nf_conn *ct = NULL;
 692        enum ip_conntrack_info uninitialized_var(ctinfo);
 693
 694        if (li_user && li_user->type == NF_LOG_TYPE_ULOG)
 695                li = li_user;
 696        else
 697                li = &default_loginfo;
 698
 699        inst = instance_lookup_get(log, li->u.ulog.group);
 700        if (!inst)
 701                return;
 702
 703        if (prefix)
 704                plen = strlen(prefix) + 1;
 705
 706        /* FIXME: do we want to make the size calculation conditional based on
 707         * what is actually present?  way more branches and checks, but more
 708         * memory efficient... */
 709        size = nlmsg_total_size(sizeof(struct nfgenmsg))
 710                + nla_total_size(sizeof(struct nfulnl_msg_packet_hdr))
 711                + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 712                + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 713#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 714                + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 715                + nla_total_size(sizeof(u_int32_t))     /* ifindex */
 716#endif
 717                + nla_total_size(sizeof(u_int32_t))     /* mark */
 718                + nla_total_size(sizeof(u_int32_t))     /* uid */
 719                + nla_total_size(sizeof(u_int32_t))     /* gid */
 720                + nla_total_size(plen)                  /* prefix */
 721                + nla_total_size(sizeof(struct nfulnl_msg_packet_hw))
 722                + nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp))
 723                + nla_total_size(sizeof(struct nfgenmsg));      /* NLMSG_DONE */
 724
 725        if (in && skb_mac_header_was_set(skb)) {
 726                size += nla_total_size(skb->dev->hard_header_len)
 727                        + nla_total_size(sizeof(u_int16_t))     /* hwtype */
 728                        + nla_total_size(sizeof(u_int16_t));    /* hwlen */
 729        }
 730
 731        spin_lock_bh(&inst->lock);
 732
 733        if (inst->flags & NFULNL_CFG_F_SEQ)
 734                size += nla_total_size(sizeof(u_int32_t));
 735        if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
 736                size += nla_total_size(sizeof(u_int32_t));
 737        if (inst->flags & NFULNL_CFG_F_CONNTRACK) {
 738                nfnl_ct = rcu_dereference(nfnl_ct_hook);
 739                if (nfnl_ct != NULL) {
 740                        ct = nfnl_ct->get_ct(skb, &ctinfo);
 741                        if (ct != NULL)
 742                                size += nfnl_ct->build_size(ct);
 743                }
 744        }
 745        if (pf == NFPROTO_NETDEV || pf == NFPROTO_BRIDGE)
 746                size += nfulnl_get_bridge_size(skb);
 747
 748        qthreshold = inst->qthreshold;
 749        /* per-rule qthreshold overrides per-instance */
 750        if (li->u.ulog.qthreshold)
 751                if (qthreshold > li->u.ulog.qthreshold)
 752                        qthreshold = li->u.ulog.qthreshold;
 753
 754
 755        switch (inst->copy_mode) {
 756        case NFULNL_COPY_META:
 757        case NFULNL_COPY_NONE:
 758                data_len = 0;
 759                break;
 760
 761        case NFULNL_COPY_PACKET:
 762                data_len = inst->copy_range;
 763                if ((li->u.ulog.flags & NF_LOG_F_COPY_LEN) &&
 764                    (li->u.ulog.copy_len < data_len))
 765                        data_len = li->u.ulog.copy_len;
 766
 767                if (data_len > skb->len)
 768                        data_len = skb->len;
 769
 770                size += nla_total_size(data_len);
 771                break;
 772
 773        case NFULNL_COPY_DISABLED:
 774        default:
 775                goto unlock_and_release;
 776        }
 777
 778        if (inst->skb && size > skb_tailroom(inst->skb)) {
 779                /* either the queue len is too high or we don't have
 780                 * enough room in the skb left. flush to userspace. */
 781                __nfulnl_flush(inst);
 782        }
 783
 784        if (!inst->skb) {
 785                inst->skb = nfulnl_alloc_skb(net, inst->peer_portid,
 786                                             inst->nlbufsiz, size);
 787                if (!inst->skb)
 788                        goto alloc_failure;
 789        }
 790
 791        inst->qlen++;
 792
 793        __build_packet_message(log, inst, skb, data_len, pf,
 794                                hooknum, in, out, prefix, plen,
 795                                nfnl_ct, ct, ctinfo);
 796
 797        if (inst->qlen >= qthreshold)
 798                __nfulnl_flush(inst);
 799        /* timer_pending always called within inst->lock, so there
 800         * is no chance of a race here */
 801        else if (!timer_pending(&inst->timer)) {
 802                instance_get(inst);
 803                inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100);
 804                add_timer(&inst->timer);
 805        }
 806
 807unlock_and_release:
 808        spin_unlock_bh(&inst->lock);
 809        instance_put(inst);
 810        return;
 811
 812alloc_failure:
 813        /* FIXME: statistics */
 814        goto unlock_and_release;
 815}
 816
 817static int
 818nfulnl_rcv_nl_event(struct notifier_block *this,
 819                   unsigned long event, void *ptr)
 820{
 821        struct netlink_notify *n = ptr;
 822        struct nfnl_log_net *log = nfnl_log_pernet(n->net);
 823
 824        if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
 825                int i;
 826
 827                /* destroy all instances for this portid */
 828                spin_lock_bh(&log->instances_lock);
 829                for  (i = 0; i < INSTANCE_BUCKETS; i++) {
 830                        struct hlist_node *t2;
 831                        struct nfulnl_instance *inst;
 832                        struct hlist_head *head = &log->instance_table[i];
 833
 834                        hlist_for_each_entry_safe(inst, t2, head, hlist) {
 835                                if (n->portid == inst->peer_portid)
 836                                        __instance_destroy(inst);
 837                        }
 838                }
 839                spin_unlock_bh(&log->instances_lock);
 840        }
 841        return NOTIFY_DONE;
 842}
 843
 844static struct notifier_block nfulnl_rtnl_notifier = {
 845        .notifier_call  = nfulnl_rcv_nl_event,
 846};
 847
 848static int nfulnl_recv_unsupp(struct net *net, struct sock *ctnl,
 849                              struct sk_buff *skb, const struct nlmsghdr *nlh,
 850                              const struct nlattr * const nfqa[],
 851                              struct netlink_ext_ack *extack)
 852{
 853        return -ENOTSUPP;
 854}
 855
 856static struct nf_logger nfulnl_logger __read_mostly = {
 857        .name   = "nfnetlink_log",
 858        .type   = NF_LOG_TYPE_ULOG,
 859        .logfn  = nfulnl_log_packet,
 860        .me     = THIS_MODULE,
 861};
 862
 863static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = {
 864        [NFULA_CFG_CMD]         = { .len = sizeof(struct nfulnl_msg_config_cmd) },
 865        [NFULA_CFG_MODE]        = { .len = sizeof(struct nfulnl_msg_config_mode) },
 866        [NFULA_CFG_TIMEOUT]     = { .type = NLA_U32 },
 867        [NFULA_CFG_QTHRESH]     = { .type = NLA_U32 },
 868        [NFULA_CFG_NLBUFSIZ]    = { .type = NLA_U32 },
 869        [NFULA_CFG_FLAGS]       = { .type = NLA_U16 },
 870};
 871
 872static int nfulnl_recv_config(struct net *net, struct sock *ctnl,
 873                              struct sk_buff *skb, const struct nlmsghdr *nlh,
 874                              const struct nlattr * const nfula[],
 875                              struct netlink_ext_ack *extack)
 876{
 877        struct nfgenmsg *nfmsg = nlmsg_data(nlh);
 878        u_int16_t group_num = ntohs(nfmsg->res_id);
 879        struct nfulnl_instance *inst;
 880        struct nfulnl_msg_config_cmd *cmd = NULL;
 881        struct nfnl_log_net *log = nfnl_log_pernet(net);
 882        int ret = 0;
 883        u16 flags = 0;
 884
 885        if (nfula[NFULA_CFG_CMD]) {
 886                u_int8_t pf = nfmsg->nfgen_family;
 887                cmd = nla_data(nfula[NFULA_CFG_CMD]);
 888
 889                /* Commands without queue context */
 890                switch (cmd->command) {
 891                case NFULNL_CFG_CMD_PF_BIND:
 892                        return nf_log_bind_pf(net, pf, &nfulnl_logger);
 893                case NFULNL_CFG_CMD_PF_UNBIND:
 894                        nf_log_unbind_pf(net, pf);
 895                        return 0;
 896                }
 897        }
 898
 899        inst = instance_lookup_get(log, group_num);
 900        if (inst && inst->peer_portid != NETLINK_CB(skb).portid) {
 901                ret = -EPERM;
 902                goto out_put;
 903        }
 904
 905        /* Check if we support these flags in first place, dependencies should
 906         * be there too not to break atomicity.
 907         */
 908        if (nfula[NFULA_CFG_FLAGS]) {
 909                flags = ntohs(nla_get_be16(nfula[NFULA_CFG_FLAGS]));
 910
 911                if ((flags & NFULNL_CFG_F_CONNTRACK) &&
 912                    !rcu_access_pointer(nfnl_ct_hook)) {
 913#ifdef CONFIG_MODULES
 914                        nfnl_unlock(NFNL_SUBSYS_ULOG);
 915                        request_module("ip_conntrack_netlink");
 916                        nfnl_lock(NFNL_SUBSYS_ULOG);
 917                        if (rcu_access_pointer(nfnl_ct_hook)) {
 918                                ret = -EAGAIN;
 919                                goto out_put;
 920                        }
 921#endif
 922                        ret = -EOPNOTSUPP;
 923                        goto out_put;
 924                }
 925        }
 926
 927        if (cmd != NULL) {
 928                switch (cmd->command) {
 929                case NFULNL_CFG_CMD_BIND:
 930                        if (inst) {
 931                                ret = -EBUSY;
 932                                goto out_put;
 933                        }
 934
 935                        inst = instance_create(net, group_num,
 936                                               NETLINK_CB(skb).portid,
 937                                               sk_user_ns(NETLINK_CB(skb).sk));
 938                        if (IS_ERR(inst)) {
 939                                ret = PTR_ERR(inst);
 940                                goto out;
 941                        }
 942                        break;
 943                case NFULNL_CFG_CMD_UNBIND:
 944                        if (!inst) {
 945                                ret = -ENODEV;
 946                                goto out;
 947                        }
 948
 949                        instance_destroy(log, inst);
 950                        goto out_put;
 951                default:
 952                        ret = -ENOTSUPP;
 953                        goto out_put;
 954                }
 955        } else if (!inst) {
 956                ret = -ENODEV;
 957                goto out;
 958        }
 959
 960        if (nfula[NFULA_CFG_MODE]) {
 961                struct nfulnl_msg_config_mode *params =
 962                        nla_data(nfula[NFULA_CFG_MODE]);
 963
 964                nfulnl_set_mode(inst, params->copy_mode,
 965                                ntohl(params->copy_range));
 966        }
 967
 968        if (nfula[NFULA_CFG_TIMEOUT]) {
 969                __be32 timeout = nla_get_be32(nfula[NFULA_CFG_TIMEOUT]);
 970
 971                nfulnl_set_timeout(inst, ntohl(timeout));
 972        }
 973
 974        if (nfula[NFULA_CFG_NLBUFSIZ]) {
 975                __be32 nlbufsiz = nla_get_be32(nfula[NFULA_CFG_NLBUFSIZ]);
 976
 977                nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz));
 978        }
 979
 980        if (nfula[NFULA_CFG_QTHRESH]) {
 981                __be32 qthresh = nla_get_be32(nfula[NFULA_CFG_QTHRESH]);
 982
 983                nfulnl_set_qthresh(inst, ntohl(qthresh));
 984        }
 985
 986        if (nfula[NFULA_CFG_FLAGS])
 987                nfulnl_set_flags(inst, flags);
 988
 989out_put:
 990        instance_put(inst);
 991out:
 992        return ret;
 993}
 994
 995static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = {
 996        [NFULNL_MSG_PACKET]     = { .call = nfulnl_recv_unsupp,
 997                                    .attr_count = NFULA_MAX, },
 998        [NFULNL_MSG_CONFIG]     = { .call = nfulnl_recv_config,
 999                                    .attr_count = NFULA_CFG_MAX,
1000                                    .policy = nfula_cfg_policy },
1001};
1002
1003static const struct nfnetlink_subsystem nfulnl_subsys = {
1004        .name           = "log",
1005        .subsys_id      = NFNL_SUBSYS_ULOG,
1006        .cb_count       = NFULNL_MSG_MAX,
1007        .cb             = nfulnl_cb,
1008};
1009
1010#ifdef CONFIG_PROC_FS
1011struct iter_state {
1012        struct seq_net_private p;
1013        unsigned int bucket;
1014};
1015
1016static struct hlist_node *get_first(struct net *net, struct iter_state *st)
1017{
1018        struct nfnl_log_net *log;
1019        if (!st)
1020                return NULL;
1021
1022        log = nfnl_log_pernet(net);
1023
1024        for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1025                struct hlist_head *head = &log->instance_table[st->bucket];
1026
1027                if (!hlist_empty(head))
1028                        return rcu_dereference_bh(hlist_first_rcu(head));
1029        }
1030        return NULL;
1031}
1032
1033static struct hlist_node *get_next(struct net *net, struct iter_state *st,
1034                                   struct hlist_node *h)
1035{
1036        h = rcu_dereference_bh(hlist_next_rcu(h));
1037        while (!h) {
1038                struct nfnl_log_net *log;
1039                struct hlist_head *head;
1040
1041                if (++st->bucket >= INSTANCE_BUCKETS)
1042                        return NULL;
1043
1044                log = nfnl_log_pernet(net);
1045                head = &log->instance_table[st->bucket];
1046                h = rcu_dereference_bh(hlist_first_rcu(head));
1047        }
1048        return h;
1049}
1050
1051static struct hlist_node *get_idx(struct net *net, struct iter_state *st,
1052                                  loff_t pos)
1053{
1054        struct hlist_node *head;
1055        head = get_first(net, st);
1056
1057        if (head)
1058                while (pos && (head = get_next(net, st, head)))
1059                        pos--;
1060        return pos ? NULL : head;
1061}
1062
1063static void *seq_start(struct seq_file *s, loff_t *pos)
1064        __acquires(rcu_bh)
1065{
1066        rcu_read_lock_bh();
1067        return get_idx(seq_file_net(s), s->private, *pos);
1068}
1069
1070static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1071{
1072        (*pos)++;
1073        return get_next(seq_file_net(s), s->private, v);
1074}
1075
1076static void seq_stop(struct seq_file *s, void *v)
1077        __releases(rcu_bh)
1078{
1079        rcu_read_unlock_bh();
1080}
1081
1082static int seq_show(struct seq_file *s, void *v)
1083{
1084        const struct nfulnl_instance *inst = v;
1085
1086        seq_printf(s, "%5u %6u %5u %1u %5u %6u %2u\n",
1087                   inst->group_num,
1088                   inst->peer_portid, inst->qlen,
1089                   inst->copy_mode, inst->copy_range,
1090                   inst->flushtimeout, refcount_read(&inst->use));
1091
1092        return 0;
1093}
1094
1095static const struct seq_operations nful_seq_ops = {
1096        .start  = seq_start,
1097        .next   = seq_next,
1098        .stop   = seq_stop,
1099        .show   = seq_show,
1100};
1101#endif /* PROC_FS */
1102
1103static int __net_init nfnl_log_net_init(struct net *net)
1104{
1105        unsigned int i;
1106        struct nfnl_log_net *log = nfnl_log_pernet(net);
1107#ifdef CONFIG_PROC_FS
1108        struct proc_dir_entry *proc;
1109        kuid_t root_uid;
1110        kgid_t root_gid;
1111#endif
1112
1113        for (i = 0; i < INSTANCE_BUCKETS; i++)
1114                INIT_HLIST_HEAD(&log->instance_table[i]);
1115        spin_lock_init(&log->instances_lock);
1116
1117#ifdef CONFIG_PROC_FS
1118        proc = proc_create_net("nfnetlink_log", 0440, net->nf.proc_netfilter,
1119                        &nful_seq_ops, sizeof(struct iter_state));
1120        if (!proc)
1121                return -ENOMEM;
1122
1123        root_uid = make_kuid(net->user_ns, 0);
1124        root_gid = make_kgid(net->user_ns, 0);
1125        if (uid_valid(root_uid) && gid_valid(root_gid))
1126                proc_set_user(proc, root_uid, root_gid);
1127#endif
1128        return 0;
1129}
1130
1131static void __net_exit nfnl_log_net_exit(struct net *net)
1132{
1133        struct nfnl_log_net *log = nfnl_log_pernet(net);
1134        unsigned int i;
1135
1136#ifdef CONFIG_PROC_FS
1137        remove_proc_entry("nfnetlink_log", net->nf.proc_netfilter);
1138#endif
1139        nf_log_unset(net, &nfulnl_logger);
1140        for (i = 0; i < INSTANCE_BUCKETS; i++)
1141                WARN_ON_ONCE(!hlist_empty(&log->instance_table[i]));
1142}
1143
1144static struct pernet_operations nfnl_log_net_ops = {
1145        .init   = nfnl_log_net_init,
1146        .exit   = nfnl_log_net_exit,
1147        .id     = &nfnl_log_net_id,
1148        .size   = sizeof(struct nfnl_log_net),
1149};
1150
1151static int __init nfnetlink_log_init(void)
1152{
1153        int status;
1154
1155        status = register_pernet_subsys(&nfnl_log_net_ops);
1156        if (status < 0) {
1157                pr_err("failed to register pernet ops\n");
1158                goto out;
1159        }
1160
1161        netlink_register_notifier(&nfulnl_rtnl_notifier);
1162        status = nfnetlink_subsys_register(&nfulnl_subsys);
1163        if (status < 0) {
1164                pr_err("failed to create netlink socket\n");
1165                goto cleanup_netlink_notifier;
1166        }
1167
1168        status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger);
1169        if (status < 0) {
1170                pr_err("failed to register logger\n");
1171                goto cleanup_subsys;
1172        }
1173
1174        return status;
1175
1176cleanup_subsys:
1177        nfnetlink_subsys_unregister(&nfulnl_subsys);
1178cleanup_netlink_notifier:
1179        netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1180        unregister_pernet_subsys(&nfnl_log_net_ops);
1181out:
1182        return status;
1183}
1184
1185static void __exit nfnetlink_log_fini(void)
1186{
1187        nfnetlink_subsys_unregister(&nfulnl_subsys);
1188        netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1189        unregister_pernet_subsys(&nfnl_log_net_ops);
1190        nf_log_unregister(&nfulnl_logger);
1191}
1192
1193MODULE_DESCRIPTION("netfilter userspace logging");
1194MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1195MODULE_LICENSE("GPL");
1196MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG);
1197MODULE_ALIAS_NF_LOGGER(AF_INET, 1);
1198MODULE_ALIAS_NF_LOGGER(AF_INET6, 1);
1199MODULE_ALIAS_NF_LOGGER(AF_BRIDGE, 1);
1200MODULE_ALIAS_NF_LOGGER(3, 1); /* NFPROTO_ARP */
1201MODULE_ALIAS_NF_LOGGER(5, 1); /* NFPROTO_NETDEV */
1202
1203module_init(nfnetlink_log_init);
1204module_exit(nfnetlink_log_fini);
1205