nfnetlink_log.c

   1// SPDX-License-Identifier: GPL-2.0-only
   2/*
   3 * This is a module which is used for logging packets to userspace via
   4 * nfetlink.
   5 *
   6 * (C) 2005 by Harald Welte <laforge@netfilter.org>
   7 * (C) 2006-2012 Patrick McHardy <kaber@trash.net>
   8 *
   9 * Based on the old ipv4-only ipt_ULOG.c:
  10 * (C) 2000-2004 by Harald Welte <laforge@netfilter.org>
  11 */
  12
  13#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  14
  15#include <linux/module.h>
  16#include <linux/skbuff.h>
  17#include <linux/if_arp.h>
  18#include <linux/init.h>
  19#include <linux/ip.h>
  20#include <linux/ipv6.h>
  21#include <linux/netdevice.h>
  22#include <linux/netfilter.h>
  23#include <linux/netfilter_bridge.h>
  24#include <net/netlink.h>
  25#include <linux/netfilter/nfnetlink.h>
  26#include <linux/netfilter/nfnetlink_log.h>
  27#include <linux/netfilter/nf_conntrack_common.h>
  28#include <linux/spinlock.h>
  29#include <linux/sysctl.h>
  30#include <linux/proc_fs.h>
  31#include <linux/security.h>
  32#include <linux/list.h>
  33#include <linux/slab.h>
  34#include <net/sock.h>
  35#include <net/netfilter/nf_log.h>
  36#include <net/netns/generic.h>
  37
  38#include <linux/atomic.h>
  39#include <linux/refcount.h>
  40
  41
  42#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  43#include "../bridge/br_private.h"
  44#endif
  45
  46#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(const struct nfnl_log_net *log, u16 group_num)
 107{
 108	const 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_rcu(const struct nfnl_log_net *log, u16 group_num)
 127{
 128	struct nfulnl_instance *inst;
 129
 130	inst = __instance_lookup(log, group_num);
 131	if (inst && !refcount_inc_not_zero(&inst->use))
 132		inst = NULL;
 133
 134	return inst;
 135}
 136
 137static struct nfulnl_instance *
 138instance_lookup_get(const struct nfnl_log_net *log, u16 group_num)
 139{
 140	struct nfulnl_instance *inst;
 141
 142	rcu_read_lock();
 143	inst = instance_lookup_get_rcu(log, group_num);
 144	rcu_read_unlock();
 145
 146	return inst;
 147}
 148
 149static void nfulnl_instance_free_rcu(struct rcu_head *head)
 150{
 151	struct nfulnl_instance *inst =
 152		container_of(head, struct nfulnl_instance, rcu);
 153
 154	put_net_track(inst->net, &inst->ns_tracker);
 155	kfree(inst);
 156	module_put(THIS_MODULE);
 157}
 158
 159static void
 160instance_put(struct nfulnl_instance *inst)
 161{
 162	if (inst && refcount_dec_and_test(&inst->use))
 163		call_rcu(&inst->rcu, nfulnl_instance_free_rcu);
 164}
 165
 166static void nfulnl_timer(struct timer_list *t);
 167
 168static struct nfulnl_instance *
 169instance_create(struct net *net, u_int16_t group_num,
 170		u32 portid, struct user_namespace *user_ns)
 171{
 172	struct nfulnl_instance *inst;
 173	struct nfnl_log_net *log = nfnl_log_pernet(net);
 174	int err;
 175
 176	spin_lock_bh(&log->instances_lock);
 177	if (__instance_lookup(log, group_num)) {
 178		err = -EEXIST;
 179		goto out_unlock;
 180	}
 181
 182	inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
 183	if (!inst) {
 184		err = -ENOMEM;
 185		goto out_unlock;
 186	}
 187
 188	if (!try_module_get(THIS_MODULE)) {
 189		kfree(inst);
 190		err = -EAGAIN;
 191		goto out_unlock;
 192	}
 193
 194	INIT_HLIST_NODE(&inst->hlist);
 195	spin_lock_init(&inst->lock);
 196	/* needs to be two, since we _put() after creation */
 197	refcount_set(&inst->use, 2);
 198
 199	timer_setup(&inst->timer, nfulnl_timer, 0);
 200
 201	inst->net = get_net_track(net, &inst->ns_tracker, GFP_ATOMIC);
 202	inst->peer_user_ns = user_ns;
 203	inst->peer_portid = portid;
 204	inst->group_num = group_num;
 205
 206	inst->qthreshold 	= NFULNL_QTHRESH_DEFAULT;
 207	inst->flushtimeout 	= NFULNL_TIMEOUT_DEFAULT;
 208	inst->nlbufsiz 		= NFULNL_NLBUFSIZ_DEFAULT;
 209	inst->copy_mode 	= NFULNL_COPY_PACKET;
 210	inst->copy_range 	= NFULNL_COPY_RANGE_MAX;
 211
 212	hlist_add_head_rcu(&inst->hlist,
 213		       &log->instance_table[instance_hashfn(group_num)]);
 214
 215
 216	spin_unlock_bh(&log->instances_lock);
 217
 218	return inst;
 219
 220out_unlock:
 221	spin_unlock_bh(&log->instances_lock);
 222	return ERR_PTR(err);
 223}
 224
 225static void __nfulnl_flush(struct nfulnl_instance *inst);
 226
 227/* called with BH disabled */
 228static void
 229__instance_destroy(struct nfulnl_instance *inst)
 230{
 231	/* first pull it out of the global list */
 232	hlist_del_rcu(&inst->hlist);
 233
 234	/* then flush all pending packets from skb */
 235
 236	spin_lock(&inst->lock);
 237
 238	/* lockless readers wont be able to use us */
 239	inst->copy_mode = NFULNL_COPY_DISABLED;
 240
 241	if (inst->skb)
 242		__nfulnl_flush(inst);
 243	spin_unlock(&inst->lock);
 244
 245	/* and finally put the refcount */
 246	instance_put(inst);
 247}
 248
 249static inline void
 250instance_destroy(struct nfnl_log_net *log,
 251		 struct nfulnl_instance *inst)
 252{
 253	spin_lock_bh(&log->instances_lock);
 254	__instance_destroy(inst);
 255	spin_unlock_bh(&log->instances_lock);
 256}
 257
 258static int
 259nfulnl_set_mode(struct nfulnl_instance *inst, u_int8_t mode,
 260		  unsigned int range)
 261{
 262	int status = 0;
 263
 264	spin_lock_bh(&inst->lock);
 265
 266	switch (mode) {
 267	case NFULNL_COPY_NONE:
 268	case NFULNL_COPY_META:
 269		inst->copy_mode = mode;
 270		inst->copy_range = 0;
 271		break;
 272
 273	case NFULNL_COPY_PACKET:
 274		inst->copy_mode = mode;
 275		if (range == 0)
 276			range = NFULNL_COPY_RANGE_MAX;
 277		inst->copy_range = min_t(unsigned int,
 278					 range, NFULNL_COPY_RANGE_MAX);
 279		break;
 280
 281	default:
 282		status = -EINVAL;
 283		break;
 284	}
 285
 286	spin_unlock_bh(&inst->lock);
 287
 288	return status;
 289}
 290
 291static int
 292nfulnl_set_nlbufsiz(struct nfulnl_instance *inst, u_int32_t nlbufsiz)
 293{
 294	int status;
 295
 296	spin_lock_bh(&inst->lock);
 297	if (nlbufsiz < NFULNL_NLBUFSIZ_DEFAULT)
 298		status = -ERANGE;
 299	else if (nlbufsiz > 131072)
 300		status = -ERANGE;
 301	else {
 302		inst->nlbufsiz = nlbufsiz;
 303		status = 0;
 304	}
 305	spin_unlock_bh(&inst->lock);
 306
 307	return status;
 308}
 309
 310static void
 311nfulnl_set_timeout(struct nfulnl_instance *inst, u_int32_t timeout)
 312{
 313	spin_lock_bh(&inst->lock);
 314	inst->flushtimeout = timeout;
 315	spin_unlock_bh(&inst->lock);
 316}
 317
 318static void
 319nfulnl_set_qthresh(struct nfulnl_instance *inst, u_int32_t qthresh)
 320{
 321	spin_lock_bh(&inst->lock);
 322	inst->qthreshold = qthresh;
 323	spin_unlock_bh(&inst->lock);
 324}
 325
 326static int
 327nfulnl_set_flags(struct nfulnl_instance *inst, u_int16_t flags)
 328{
 329	spin_lock_bh(&inst->lock);
 330	inst->flags = flags;
 331	spin_unlock_bh(&inst->lock);
 332
 333	return 0;
 334}
 335
 336static struct sk_buff *
 337nfulnl_alloc_skb(struct net *net, u32 peer_portid, unsigned int inst_size,
 338		 unsigned int pkt_size)
 339{
 340	struct sk_buff *skb;
 341	unsigned int n;
 342
 343	/* alloc skb which should be big enough for a whole multipart
 344	 * message.  WARNING: has to be <= 128k due to slab restrictions */
 345
 346	n = max(inst_size, pkt_size);
 347	skb = alloc_skb(n, GFP_ATOMIC | __GFP_NOWARN);
 348	if (!skb) {
 349		if (n > pkt_size) {
 350			/* try to allocate only as much as we need for current
 351			 * packet */
 352
 353			skb = alloc_skb(pkt_size, GFP_ATOMIC);
 354		}
 355	}
 356
 357	return skb;
 358}
 359
 360static void
 361__nfulnl_send(struct nfulnl_instance *inst)
 362{
 363	if (inst->qlen > 1) {
 364		struct nlmsghdr *nlh = nfnl_msg_put(inst->skb, 0, 0,
 365						    NLMSG_DONE, 0,
 366						    AF_UNSPEC, NFNETLINK_V0,
 367						    htons(inst->group_num));
 368		if (WARN_ONCE(!nlh, "bad nlskb size: %u, tailroom %d\n",
 369			      inst->skb->len, skb_tailroom(inst->skb))) {
 370			kfree_skb(inst->skb);
 371			goto out;
 372		}
 373	}
 374	nfnetlink_unicast(inst->skb, inst->net, inst->peer_portid);
 375out:
 376	inst->qlen = 0;
 377	inst->skb = NULL;
 378}
 379
 380static void
 381__nfulnl_flush(struct nfulnl_instance *inst)
 382{
 383	/* timer holds a reference */
 384	if (timer_delete(&inst->timer))
 385		instance_put(inst);
 386	if (inst->skb)
 387		__nfulnl_send(inst);
 388}
 389
 390static void
 391nfulnl_timer(struct timer_list *t)
 392{
 393	struct nfulnl_instance *inst = timer_container_of(inst, t, timer);
 394
 395	spin_lock_bh(&inst->lock);
 396	if (inst->skb)
 397		__nfulnl_send(inst);
 398	spin_unlock_bh(&inst->lock);
 399	instance_put(inst);
 400}
 401
 402static u32 nfulnl_get_bridge_size(const struct sk_buff *skb)
 403{
 404	u32 size = 0;
 405
 406	if (!skb_mac_header_was_set(skb))
 407		return 0;
 408
 409	if (skb_vlan_tag_present(skb)) {
 410		size += nla_total_size(0); /* nested */
 411		size += nla_total_size(sizeof(u16)); /* id */
 412		size += nla_total_size(sizeof(u16)); /* tag */
 413	}
 414
 415	if (skb->network_header > skb->mac_header)
 416		size += nla_total_size(skb->network_header - skb->mac_header);
 417
 418	return size;
 419}
 420
 421static int nfulnl_put_bridge(struct nfulnl_instance *inst, const struct sk_buff *skb)
 422{
 423	if (!skb_mac_header_was_set(skb))
 424		return 0;
 425
 426	if (skb_vlan_tag_present(skb)) {
 427		struct nlattr *nest;
 428
 429		nest = nla_nest_start(inst->skb, NFULA_VLAN);
 430		if (!nest)
 431			goto nla_put_failure;
 432
 433		if (nla_put_be16(inst->skb, NFULA_VLAN_TCI, htons(skb->vlan_tci)) ||
 434		    nla_put_be16(inst->skb, NFULA_VLAN_PROTO, skb->vlan_proto))
 435			goto nla_put_failure;
 436
 437		nla_nest_end(inst->skb, nest);
 438	}
 439
 440	if (skb->mac_header < skb->network_header) {
 441		int len = (int)(skb->network_header - skb->mac_header);
 442
 443		if (nla_put(inst->skb, NFULA_L2HDR, len, skb_mac_header(skb)))
 444			goto nla_put_failure;
 445	}
 446
 447	return 0;
 448
 449nla_put_failure:
 450	return -1;
 451}
 452
 453/* This is an inline function, we don't really care about a long
 454 * list of arguments */
 455static inline int
 456__build_packet_message(struct nfnl_log_net *log,
 457			struct nfulnl_instance *inst,
 458			const struct sk_buff *skb,
 459			unsigned int data_len,
 460			u_int8_t pf,
 461			unsigned int hooknum,
 462			const struct net_device *indev,
 463			const struct net_device *outdev,
 464			const char *prefix, unsigned int plen,
 465			const struct nfnl_ct_hook *nfnl_ct,
 466			struct nf_conn *ct, enum ip_conntrack_info ctinfo)
 467{
 468	struct nfulnl_msg_packet_hdr pmsg;
 469	struct nlmsghdr *nlh;
 470	sk_buff_data_t old_tail = inst->skb->tail;
 471	struct sock *sk;
 472	const unsigned char *hwhdrp;
 473
 474	nlh = nfnl_msg_put(inst->skb, 0, 0,
 475			   nfnl_msg_type(NFNL_SUBSYS_ULOG, NFULNL_MSG_PACKET),
 476			   0, pf, NFNETLINK_V0, htons(inst->group_num));
 477	if (!nlh)
 478		return -1;
 479
 480	memset(&pmsg, 0, sizeof(pmsg));
 481	pmsg.hw_protocol	= skb->protocol;
 482	pmsg.hook		= hooknum;
 483
 484	if (nla_put(inst->skb, NFULA_PACKET_HDR, sizeof(pmsg), &pmsg))
 485		goto nla_put_failure;
 486
 487	if (prefix &&
 488	    nla_put(inst->skb, NFULA_PREFIX, plen, prefix))
 489		goto nla_put_failure;
 490
 491	if (indev) {
 492#if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 493		if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 494				 htonl(indev->ifindex)))
 495			goto nla_put_failure;
 496#else
 497		if (pf == PF_BRIDGE) {
 498			/* Case 1: outdev is physical input device, we need to
 499			 * look for bridge group (when called from
 500			 * netfilter_bridge) */
 501			if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
 502					 htonl(indev->ifindex)) ||
 503			/* this is the bridge group "brX" */
 504			/* rcu_read_lock()ed by nf_hook_thresh or
 505			 * nf_log_packet.
 506			 */
 507			    nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 508					 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
 509				goto nla_put_failure;
 510		} else {
 511			int physinif;
 512
 513			/* Case 2: indev is bridge group, we need to look for
 514			 * physical device (when called from ipv4) */
 515			if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 516					 htonl(indev->ifindex)))
 517				goto nla_put_failure;
 518
 519			physinif = nf_bridge_get_physinif(skb);
 520			if (physinif &&
 521			    nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
 522					 htonl(physinif)))
 523				goto nla_put_failure;
 524		}
 525#endif
 526	}
 527
 528	if (outdev) {
 529#if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 530		if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 531				 htonl(outdev->ifindex)))
 532			goto nla_put_failure;
 533#else
 534		if (pf == PF_BRIDGE) {
 535			/* Case 1: outdev is physical output device, we need to
 536			 * look for bridge group (when called from
 537			 * netfilter_bridge) */
 538			if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
 539					 htonl(outdev->ifindex)) ||
 540			/* this is the bridge group "brX" */
 541			/* rcu_read_lock()ed by nf_hook_thresh or
 542			 * nf_log_packet.
 543			 */
 544			    nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 545					 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
 546				goto nla_put_failure;
 547		} else {
 548			struct net_device *physoutdev;
 549
 550			/* Case 2: indev is a bridge group, we need to look
 551			 * for physical device (when called from ipv4) */
 552			if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 553					 htonl(outdev->ifindex)))
 554				goto nla_put_failure;
 555
 556			physoutdev = nf_bridge_get_physoutdev(skb);
 557			if (physoutdev &&
 558			    nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
 559					 htonl(physoutdev->ifindex)))
 560				goto nla_put_failure;
 561		}
 562#endif
 563	}
 564
 565	if (skb->mark &&
 566	    nla_put_be32(inst->skb, NFULA_MARK, htonl(skb->mark)))
 567		goto nla_put_failure;
 568
 569	if (indev && skb->dev &&
 570	    skb_mac_header_was_set(skb) &&
 571	    skb_mac_header_len(skb) != 0) {
 572		struct nfulnl_msg_packet_hw phw;
 573		int len;
 574
 575		memset(&phw, 0, sizeof(phw));
 576		len = dev_parse_header(skb, phw.hw_addr);
 577		if (len > 0) {
 578			phw.hw_addrlen = htons(len);
 579			if (nla_put(inst->skb, NFULA_HWADDR, sizeof(phw), &phw))
 580				goto nla_put_failure;
 581		}
 582	}
 583
 584	if (indev && skb_mac_header_was_set(skb)) {
 585		if (nla_put_be16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type)) ||
 586		    nla_put_be16(inst->skb, NFULA_HWLEN,
 587				 htons(skb->dev->hard_header_len)))
 588			goto nla_put_failure;
 589
 590		hwhdrp = skb_mac_header(skb);
 591
 592		if (skb->dev->type == ARPHRD_SIT)
 593			hwhdrp -= ETH_HLEN;
 594
 595		if (hwhdrp >= skb->head &&
 596		    nla_put(inst->skb, NFULA_HWHEADER,
 597			    skb->dev->hard_header_len, hwhdrp))
 598			goto nla_put_failure;
 599	}
 600
 601	if (hooknum <= NF_INET_FORWARD) {
 602		struct timespec64 kts = ktime_to_timespec64(skb_tstamp_cond(skb, true));
 603		struct nfulnl_msg_packet_timestamp ts;
 604		ts.sec = cpu_to_be64(kts.tv_sec);
 605		ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
 606
 607		if (nla_put(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts))
 608			goto nla_put_failure;
 609	}
 610
 611	/* UID */
 612	sk = skb->sk;
 613	if (sk && sk_fullsock(sk)) {
 614		read_lock_bh(&sk->sk_callback_lock);
 615		if (sk->sk_socket && sk->sk_socket->file) {
 616			struct file *file = sk->sk_socket->file;
 617			const struct cred *cred = file->f_cred;
 618			struct user_namespace *user_ns = inst->peer_user_ns;
 619			__be32 uid = htonl(from_kuid_munged(user_ns, cred->fsuid));
 620			__be32 gid = htonl(from_kgid_munged(user_ns, cred->fsgid));
 621			read_unlock_bh(&sk->sk_callback_lock);
 622			if (nla_put_be32(inst->skb, NFULA_UID, uid) ||
 623			    nla_put_be32(inst->skb, NFULA_GID, gid))
 624				goto nla_put_failure;
 625		} else
 626			read_unlock_bh(&sk->sk_callback_lock);
 627	}
 628
 629	/* local sequence number */
 630	if ((inst->flags & NFULNL_CFG_F_SEQ) &&
 631	    nla_put_be32(inst->skb, NFULA_SEQ, htonl(inst->seq++)))
 632		goto nla_put_failure;
 633
 634	/* global sequence number */
 635	if ((inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) &&
 636	    nla_put_be32(inst->skb, NFULA_SEQ_GLOBAL,
 637			 htonl(atomic_inc_return(&log->global_seq))))
 638		goto nla_put_failure;
 639
 640	if (ct && nfnl_ct->build(inst->skb, ct, ctinfo,
 641				 NFULA_CT, NFULA_CT_INFO) < 0)
 642		goto nla_put_failure;
 643
 644	if ((pf == NFPROTO_NETDEV || pf == NFPROTO_BRIDGE) &&
 645	    nfulnl_put_bridge(inst, skb) < 0)
 646		goto nla_put_failure;
 647
 648	if (data_len) {
 649		struct nlattr *nla;
 650
 651		nla = nla_reserve(inst->skb, NFULA_PAYLOAD, data_len);
 652		if (!nla)
 653			goto nla_put_failure;
 654
 655		if (skb_copy_bits(skb, 0, nla_data(nla), data_len))
 656			BUG();
 657	}
 658
 659	nlh->nlmsg_len = inst->skb->tail - old_tail;
 660	return 0;
 661
 662nla_put_failure:
 663	PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n");
 664	return -1;
 665}
 666
 667static const struct nf_loginfo default_loginfo = {
 668	.type =		NF_LOG_TYPE_ULOG,
 669	.u = {
 670		.ulog = {
 671			.copy_len	= 0xffff,
 672			.group		= 0,
 673			.qthreshold	= 1,
 674		},
 675	},
 676};
 677
 678/* log handler for internal netfilter logging api */
 679static void
 680nfulnl_log_packet(struct net *net,
 681		  u_int8_t pf,
 682		  unsigned int hooknum,
 683		  const struct sk_buff *skb,
 684		  const struct net_device *in,
 685		  const struct net_device *out,
 686		  const struct nf_loginfo *li_user,
 687		  const char *prefix)
 688{
 689	size_t size;
 690	unsigned int data_len;
 691	struct nfulnl_instance *inst;
 692	const struct nf_loginfo *li;
 693	unsigned int qthreshold;
 694	unsigned int plen = 0;
 695	struct nfnl_log_net *log = nfnl_log_pernet(net);
 696	const struct nfnl_ct_hook *nfnl_ct = NULL;
 697	enum ip_conntrack_info ctinfo = 0;
 698	struct nf_conn *ct = NULL;
 699
 700	if (li_user && li_user->type == NF_LOG_TYPE_ULOG)
 701		li = li_user;
 702	else
 703		li = &default_loginfo;
 704
 705	inst = instance_lookup_get_rcu(log, li->u.ulog.group);
 706	if (!inst)
 707		return;
 708
 709	if (prefix)
 710		plen = strlen(prefix) + 1;
 711
 712	/* FIXME: do we want to make the size calculation conditional based on
 713	 * what is actually present?  way more branches and checks, but more
 714	 * memory efficient... */
 715	size = nlmsg_total_size(sizeof(struct nfgenmsg))
 716		+ nla_total_size(sizeof(struct nfulnl_msg_packet_hdr))
 717		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
 718		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
 719#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 720		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
 721		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
 722#endif
 723		+ nla_total_size(sizeof(u_int32_t))	/* mark */
 724		+ nla_total_size(sizeof(u_int32_t))	/* uid */
 725		+ nla_total_size(sizeof(u_int32_t))	/* gid */
 726		+ nla_total_size(plen)			/* prefix */
 727		+ nla_total_size(sizeof(struct nfulnl_msg_packet_hw))
 728		+ nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp))
 729		+ nlmsg_total_size(sizeof(struct nfgenmsg));	/* NLMSG_DONE */
 730
 731	if (in && skb_mac_header_was_set(skb)) {
 732		size += nla_total_size(skb->dev->hard_header_len)
 733			+ nla_total_size(sizeof(u_int16_t))	/* hwtype */
 734			+ nla_total_size(sizeof(u_int16_t));	/* hwlen */
 735	}
 736
 737	spin_lock_bh(&inst->lock);
 738
 739	if (inst->flags & NFULNL_CFG_F_SEQ)
 740		size += nla_total_size(sizeof(u_int32_t));
 741	if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
 742		size += nla_total_size(sizeof(u_int32_t));
 743#if IS_ENABLED(CONFIG_NF_CONNTRACK)
 744	if (inst->flags & NFULNL_CFG_F_CONNTRACK) {
 745		nfnl_ct = rcu_dereference(nfnl_ct_hook);
 746		if (nfnl_ct != NULL) {
 747			ct = nf_ct_get(skb, &ctinfo);
 748			if (ct != NULL)
 749				size += nfnl_ct->build_size(ct);
 750		}
 751	}
 752#endif
 753	if (pf == NFPROTO_NETDEV || pf == NFPROTO_BRIDGE)
 754		size += nfulnl_get_bridge_size(skb);
 755
 756	qthreshold = inst->qthreshold;
 757	/* per-rule qthreshold overrides per-instance */
 758	if (li->u.ulog.qthreshold)
 759		if (qthreshold > li->u.ulog.qthreshold)
 760			qthreshold = li->u.ulog.qthreshold;
 761
 762
 763	switch (inst->copy_mode) {
 764	case NFULNL_COPY_META:
 765	case NFULNL_COPY_NONE:
 766		data_len = 0;
 767		break;
 768
 769	case NFULNL_COPY_PACKET:
 770		data_len = inst->copy_range;
 771		if ((li->u.ulog.flags & NF_LOG_F_COPY_LEN) &&
 772		    (li->u.ulog.copy_len < data_len))
 773			data_len = li->u.ulog.copy_len;
 774
 775		if (data_len > skb->len)
 776			data_len = skb->len;
 777
 778		size += nla_total_size(data_len);
 779		break;
 780
 781	case NFULNL_COPY_DISABLED:
 782	default:
 783		goto unlock_and_release;
 784	}
 785
 786	if (inst->skb && size > skb_tailroom(inst->skb)) {
 787		/* either the queue len is too high or we don't have
 788		 * enough room in the skb left. flush to userspace. */
 789		__nfulnl_flush(inst);
 790	}
 791
 792	if (!inst->skb) {
 793		inst->skb = nfulnl_alloc_skb(net, inst->peer_portid,
 794					     inst->nlbufsiz, size);
 795		if (!inst->skb)
 796			goto alloc_failure;
 797	}
 798
 799	inst->qlen++;
 800
 801	__build_packet_message(log, inst, skb, data_len, pf,
 802				hooknum, in, out, prefix, plen,
 803				nfnl_ct, ct, ctinfo);
 804
 805	if (inst->qlen >= qthreshold)
 806		__nfulnl_flush(inst);
 807	/* timer_pending always called within inst->lock, so there
 808	 * is no chance of a race here */
 809	else if (!timer_pending(&inst->timer)) {
 810		instance_get(inst);
 811		inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100);
 812		add_timer(&inst->timer);
 813	}
 814
 815unlock_and_release:
 816	spin_unlock_bh(&inst->lock);
 817	instance_put(inst);
 818	return;
 819
 820alloc_failure:
 821	/* FIXME: statistics */
 822	goto unlock_and_release;
 823}
 824
 825static int
 826nfulnl_rcv_nl_event(struct notifier_block *this,
 827		   unsigned long event, void *ptr)
 828{
 829	struct netlink_notify *n = ptr;
 830	struct nfnl_log_net *log = nfnl_log_pernet(n->net);
 831
 832	if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
 833		int i;
 834
 835		/* destroy all instances for this portid */
 836		spin_lock_bh(&log->instances_lock);
 837		for  (i = 0; i < INSTANCE_BUCKETS; i++) {
 838			struct hlist_node *t2;
 839			struct nfulnl_instance *inst;
 840			struct hlist_head *head = &log->instance_table[i];
 841
 842			hlist_for_each_entry_safe(inst, t2, head, hlist) {
 843				if (n->portid == inst->peer_portid)
 844					__instance_destroy(inst);
 845			}
 846		}
 847		spin_unlock_bh(&log->instances_lock);
 848	}
 849	return NOTIFY_DONE;
 850}
 851
 852static struct notifier_block nfulnl_rtnl_notifier = {
 853	.notifier_call	= nfulnl_rcv_nl_event,
 854};
 855
 856static int nfulnl_recv_unsupp(struct sk_buff *skb, const struct nfnl_info *info,
 857			      const struct nlattr * const nfula[])
 858{
 859	return -ENOTSUPP;
 860}
 861
 862static struct nf_logger nfulnl_logger __read_mostly = {
 863	.name	= "nfnetlink_log",
 864	.type	= NF_LOG_TYPE_ULOG,
 865	.logfn	= nfulnl_log_packet,
 866	.me	= THIS_MODULE,
 867};
 868
 869static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = {
 870	[NFULA_CFG_CMD]		= { .len = sizeof(struct nfulnl_msg_config_cmd) },
 871	[NFULA_CFG_MODE]	= { .len = sizeof(struct nfulnl_msg_config_mode) },
 872	[NFULA_CFG_TIMEOUT]	= { .type = NLA_U32 },
 873	[NFULA_CFG_QTHRESH]	= { .type = NLA_U32 },
 874	[NFULA_CFG_NLBUFSIZ]	= { .type = NLA_U32 },
 875	[NFULA_CFG_FLAGS]	= { .type = NLA_U16 },
 876};
 877
 878static int nfulnl_recv_config(struct sk_buff *skb, const struct nfnl_info *info,
 879			      const struct nlattr * const nfula[])
 880{
 881	struct nfnl_log_net *log = nfnl_log_pernet(info->net);
 882	u_int16_t group_num = ntohs(info->nfmsg->res_id);
 883	struct nfulnl_msg_config_cmd *cmd = NULL;
 884	struct nfulnl_instance *inst;
 885	u16 flags = 0;
 886	int ret = 0;
 887
 888	if (nfula[NFULA_CFG_CMD]) {
 889		u_int8_t pf = info->nfmsg->nfgen_family;
 890		cmd = nla_data(nfula[NFULA_CFG_CMD]);
 891
 892		/* Commands without queue context */
 893		switch (cmd->command) {
 894		case NFULNL_CFG_CMD_PF_BIND:
 895			return nf_log_bind_pf(info->net, pf, &nfulnl_logger);
 896		case NFULNL_CFG_CMD_PF_UNBIND:
 897			nf_log_unbind_pf(info->net, pf);
 898			return 0;
 899		}
 900	}
 901
 902	inst = instance_lookup_get(log, group_num);
 903	if (inst && inst->peer_portid != NETLINK_CB(skb).portid) {
 904		ret = -EPERM;
 905		goto out_put;
 906	}
 907
 908	/* Check if we support these flags in first place, dependencies should
 909	 * be there too not to break atomicity.
 910	 */
 911	if (nfula[NFULA_CFG_FLAGS]) {
 912		flags = ntohs(nla_get_be16(nfula[NFULA_CFG_FLAGS]));
 913
 914		if ((flags & NFULNL_CFG_F_CONNTRACK) &&
 915		    !rcu_access_pointer(nfnl_ct_hook)) {
 916#ifdef CONFIG_MODULES
 917			nfnl_unlock(NFNL_SUBSYS_ULOG);
 918			request_module("ip_conntrack_netlink");
 919			nfnl_lock(NFNL_SUBSYS_ULOG);
 920			if (rcu_access_pointer(nfnl_ct_hook)) {
 921				ret = -EAGAIN;
 922				goto out_put;
 923			}
 924#endif
 925			ret = -EOPNOTSUPP;
 926			goto out_put;
 927		}
 928	}
 929
 930	if (cmd != NULL) {
 931		switch (cmd->command) {
 932		case NFULNL_CFG_CMD_BIND:
 933			if (inst) {
 934				ret = -EBUSY;
 935				goto out_put;
 936			}
 937
 938			inst = instance_create(info->net, group_num,
 939					       NETLINK_CB(skb).portid,
 940					       sk_user_ns(NETLINK_CB(skb).sk));
 941			if (IS_ERR(inst)) {
 942				ret = PTR_ERR(inst);
 943				goto out;
 944			}
 945			break;
 946		case NFULNL_CFG_CMD_UNBIND:
 947			if (!inst) {
 948				ret = -ENODEV;
 949				goto out;
 950			}
 951
 952			instance_destroy(log, inst);
 953			goto out_put;
 954		default:
 955			ret = -ENOTSUPP;
 956			goto out_put;
 957		}
 958	} else if (!inst) {
 959		ret = -ENODEV;
 960		goto out;
 961	}
 962
 963	if (nfula[NFULA_CFG_MODE]) {
 964		struct nfulnl_msg_config_mode *params =
 965			nla_data(nfula[NFULA_CFG_MODE]);
 966
 967		nfulnl_set_mode(inst, params->copy_mode,
 968				ntohl(params->copy_range));
 969	}
 970
 971	if (nfula[NFULA_CFG_TIMEOUT]) {
 972		__be32 timeout = nla_get_be32(nfula[NFULA_CFG_TIMEOUT]);
 973
 974		nfulnl_set_timeout(inst, ntohl(timeout));
 975	}
 976
 977	if (nfula[NFULA_CFG_NLBUFSIZ]) {
 978		__be32 nlbufsiz = nla_get_be32(nfula[NFULA_CFG_NLBUFSIZ]);
 979
 980		nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz));
 981	}
 982
 983	if (nfula[NFULA_CFG_QTHRESH]) {
 984		__be32 qthresh = nla_get_be32(nfula[NFULA_CFG_QTHRESH]);
 985
 986		nfulnl_set_qthresh(inst, ntohl(qthresh));
 987	}
 988
 989	if (nfula[NFULA_CFG_FLAGS])
 990		nfulnl_set_flags(inst, flags);
 991
 992out_put:
 993	instance_put(inst);
 994out:
 995	return ret;
 996}
 997
 998static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = {
 999	[NFULNL_MSG_PACKET]	= {
1000		.call		= nfulnl_recv_unsupp,
1001		.type		= NFNL_CB_MUTEX,
1002		.attr_count	= NFULA_MAX,
1003	},
1004	[NFULNL_MSG_CONFIG]	= {
1005		.call		= nfulnl_recv_config,
1006		.type		= NFNL_CB_MUTEX,
1007		.attr_count	= NFULA_CFG_MAX,
1008		.policy		= nfula_cfg_policy
1009	},
1010};
1011
1012static const struct nfnetlink_subsystem nfulnl_subsys = {
1013	.name		= "log",
1014	.subsys_id	= NFNL_SUBSYS_ULOG,
1015	.cb_count	= NFULNL_MSG_MAX,
1016	.cb		= nfulnl_cb,
1017};
1018
1019#ifdef CONFIG_PROC_FS
1020struct iter_state {
1021	struct seq_net_private p;
1022	unsigned int bucket;
1023};
1024
1025static struct hlist_node *get_first(struct net *net, struct iter_state *st)
1026{
1027	struct nfnl_log_net *log;
1028	if (!st)
1029		return NULL;
1030
1031	log = nfnl_log_pernet(net);
1032
1033	for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1034		struct hlist_head *head = &log->instance_table[st->bucket];
1035
1036		if (!hlist_empty(head))
1037			return rcu_dereference(hlist_first_rcu(head));
1038	}
1039	return NULL;
1040}
1041
1042static struct hlist_node *get_next(struct net *net, struct iter_state *st,
1043				   struct hlist_node *h)
1044{
1045	h = rcu_dereference(hlist_next_rcu(h));
1046	while (!h) {
1047		struct nfnl_log_net *log;
1048		struct hlist_head *head;
1049
1050		if (++st->bucket >= INSTANCE_BUCKETS)
1051			return NULL;
1052
1053		log = nfnl_log_pernet(net);
1054		head = &log->instance_table[st->bucket];
1055		h = rcu_dereference(hlist_first_rcu(head));
1056	}
1057	return h;
1058}
1059
1060static struct hlist_node *get_idx(struct net *net, struct iter_state *st,
1061				  loff_t pos)
1062{
1063	struct hlist_node *head;
1064	head = get_first(net, st);
1065
1066	if (head)
1067		while (pos && (head = get_next(net, st, head)))
1068			pos--;
1069	return pos ? NULL : head;
1070}
1071
1072static void *seq_start(struct seq_file *s, loff_t *pos)
1073	__acquires(rcu)
1074{
1075	rcu_read_lock();
1076	return get_idx(seq_file_net(s), s->private, *pos);
1077}
1078
1079static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1080{
1081	(*pos)++;
1082	return get_next(seq_file_net(s), s->private, v);
1083}
1084
1085static void seq_stop(struct seq_file *s, void *v)
1086	__releases(rcu)
1087{
1088	rcu_read_unlock();
1089}
1090
1091static int seq_show(struct seq_file *s, void *v)
1092{
1093	const struct nfulnl_instance *inst = v;
1094
1095	seq_printf(s, "%5u %6u %5u %1u %5u %6u %2u\n",
1096		   inst->group_num,
1097		   inst->peer_portid, inst->qlen,
1098		   inst->copy_mode, inst->copy_range,
1099		   inst->flushtimeout, refcount_read(&inst->use));
1100
1101	return 0;
1102}
1103
1104static const struct seq_operations nful_seq_ops = {
1105	.start	= seq_start,
1106	.next	= seq_next,
1107	.stop	= seq_stop,
1108	.show	= seq_show,
1109};
1110#endif /* PROC_FS */
1111
1112static int __net_init nfnl_log_net_init(struct net *net)
1113{
1114	unsigned int i;
1115	struct nfnl_log_net *log = nfnl_log_pernet(net);
1116#ifdef CONFIG_PROC_FS
1117	struct proc_dir_entry *proc;
1118	kuid_t root_uid;
1119	kgid_t root_gid;
1120#endif
1121
1122	for (i = 0; i < INSTANCE_BUCKETS; i++)
1123		INIT_HLIST_HEAD(&log->instance_table[i]);
1124	spin_lock_init(&log->instances_lock);
1125
1126#ifdef CONFIG_PROC_FS
1127	proc = proc_create_net("nfnetlink_log", 0440, net->nf.proc_netfilter,
1128			&nful_seq_ops, sizeof(struct iter_state));
1129	if (!proc)
1130		return -ENOMEM;
1131
1132	root_uid = make_kuid(net->user_ns, 0);
1133	root_gid = make_kgid(net->user_ns, 0);
1134	if (uid_valid(root_uid) && gid_valid(root_gid))
1135		proc_set_user(proc, root_uid, root_gid);
1136#endif
1137	return 0;
1138}
1139
1140static void __net_exit nfnl_log_net_exit(struct net *net)
1141{
1142	struct nfnl_log_net *log = nfnl_log_pernet(net);
1143	unsigned int i;
1144
1145#ifdef CONFIG_PROC_FS
1146	remove_proc_entry("nfnetlink_log", net->nf.proc_netfilter);
1147#endif
1148	nf_log_unset(net, &nfulnl_logger);
1149	for (i = 0; i < INSTANCE_BUCKETS; i++)
1150		WARN_ON_ONCE(!hlist_empty(&log->instance_table[i]));
1151}
1152
1153static struct pernet_operations nfnl_log_net_ops = {
1154	.init	= nfnl_log_net_init,
1155	.exit	= nfnl_log_net_exit,
1156	.id	= &nfnl_log_net_id,
1157	.size	= sizeof(struct nfnl_log_net),
1158};
1159
1160static int __init nfnetlink_log_init(void)
1161{
1162	int status;
1163
1164	status = register_pernet_subsys(&nfnl_log_net_ops);
1165	if (status < 0) {
1166		pr_err("failed to register pernet ops\n");
1167		goto out;
1168	}
1169
1170	netlink_register_notifier(&nfulnl_rtnl_notifier);
1171	status = nfnetlink_subsys_register(&nfulnl_subsys);
1172	if (status < 0) {
1173		pr_err("failed to create netlink socket\n");
1174		goto cleanup_netlink_notifier;
1175	}
1176
1177	status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger);
1178	if (status < 0) {
1179		pr_err("failed to register logger\n");
1180		goto cleanup_subsys;
1181	}
1182
1183	return status;
1184
1185cleanup_subsys:
1186	nfnetlink_subsys_unregister(&nfulnl_subsys);
1187cleanup_netlink_notifier:
1188	netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1189	unregister_pernet_subsys(&nfnl_log_net_ops);
1190out:
1191	return status;
1192}
1193
1194static void __exit nfnetlink_log_fini(void)
1195{
1196	nfnetlink_subsys_unregister(&nfulnl_subsys);
1197	netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1198	unregister_pernet_subsys(&nfnl_log_net_ops);
1199	nf_log_unregister(&nfulnl_logger);
1200}
1201
1202MODULE_DESCRIPTION("netfilter userspace logging");
1203MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1204MODULE_LICENSE("GPL");
1205MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG);
1206MODULE_ALIAS_NF_LOGGER(AF_INET, 1);
1207MODULE_ALIAS_NF_LOGGER(AF_INET6, 1);
1208MODULE_ALIAS_NF_LOGGER(AF_BRIDGE, 1);
1209MODULE_ALIAS_NF_LOGGER(3, 1); /* NFPROTO_ARP */
1210MODULE_ALIAS_NF_LOGGER(5, 1); /* NFPROTO_NETDEV */
1211
1212module_init(nfnetlink_log_init);
1213module_exit(nfnetlink_log_fini);