Linux Audio

Check our new training course

Loading...
v5.14.15
   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	struct user_namespace *peer_user_ns;	/* User namespace of the peer process */
  70	u32 peer_portid;		/* PORTID of the peer process */
  71
  72	/* configurable parameters */
  73	unsigned int flushtimeout;	/* timeout until queue flush */
  74	unsigned int nlbufsiz;		/* netlink buffer allocation size */
  75	unsigned int qthreshold;	/* threshold of the queue */
  76	u_int32_t copy_range;
  77	u_int32_t seq;			/* instance-local sequential counter */
  78	u_int16_t group_num;		/* number of this queue */
  79	u_int16_t flags;
  80	u_int8_t copy_mode;
  81	struct rcu_head rcu;
  82};
  83
  84#define INSTANCE_BUCKETS	16
  85
  86static unsigned int nfnl_log_net_id __read_mostly;
  87
  88struct nfnl_log_net {
  89	spinlock_t instances_lock;
  90	struct hlist_head instance_table[INSTANCE_BUCKETS];
  91	atomic_t global_seq;
  92};
  93
  94static struct nfnl_log_net *nfnl_log_pernet(struct net *net)
  95{
  96	return net_generic(net, nfnl_log_net_id);
  97}
  98
  99static inline u_int8_t instance_hashfn(u_int16_t group_num)
 100{
 101	return ((group_num & 0xff) % INSTANCE_BUCKETS);
 102}
 103
 104static struct nfulnl_instance *
 105__instance_lookup(struct nfnl_log_net *log, u_int16_t group_num)
 106{
 107	struct hlist_head *head;
 108	struct nfulnl_instance *inst;
 109
 110	head = &log->instance_table[instance_hashfn(group_num)];
 111	hlist_for_each_entry_rcu(inst, head, hlist) {
 112		if (inst->group_num == group_num)
 113			return inst;
 114	}
 115	return NULL;
 116}
 117
 118static inline void
 119instance_get(struct nfulnl_instance *inst)
 120{
 121	refcount_inc(&inst->use);
 122}
 123
 124static struct nfulnl_instance *
 125instance_lookup_get(struct nfnl_log_net *log, u_int16_t group_num)
 126{
 127	struct nfulnl_instance *inst;
 128
 129	rcu_read_lock_bh();
 130	inst = __instance_lookup(log, group_num);
 131	if (inst && !refcount_inc_not_zero(&inst->use))
 132		inst = NULL;
 133	rcu_read_unlock_bh();
 134
 135	return inst;
 136}
 137
 138static void nfulnl_instance_free_rcu(struct rcu_head *head)
 139{
 140	struct nfulnl_instance *inst =
 141		container_of(head, struct nfulnl_instance, rcu);
 142
 143	put_net(inst->net);
 144	kfree(inst);
 145	module_put(THIS_MODULE);
 146}
 147
 148static void
 149instance_put(struct nfulnl_instance *inst)
 150{
 151	if (inst && refcount_dec_and_test(&inst->use))
 152		call_rcu(&inst->rcu, nfulnl_instance_free_rcu);
 153}
 154
 155static void nfulnl_timer(struct timer_list *t);
 156
 157static struct nfulnl_instance *
 158instance_create(struct net *net, u_int16_t group_num,
 159		u32 portid, struct user_namespace *user_ns)
 160{
 161	struct nfulnl_instance *inst;
 162	struct nfnl_log_net *log = nfnl_log_pernet(net);
 163	int err;
 164
 165	spin_lock_bh(&log->instances_lock);
 166	if (__instance_lookup(log, group_num)) {
 167		err = -EEXIST;
 168		goto out_unlock;
 169	}
 170
 171	inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
 172	if (!inst) {
 173		err = -ENOMEM;
 174		goto out_unlock;
 175	}
 176
 177	if (!try_module_get(THIS_MODULE)) {
 178		kfree(inst);
 179		err = -EAGAIN;
 180		goto out_unlock;
 181	}
 182
 183	INIT_HLIST_NODE(&inst->hlist);
 184	spin_lock_init(&inst->lock);
 185	/* needs to be two, since we _put() after creation */
 186	refcount_set(&inst->use, 2);
 187
 188	timer_setup(&inst->timer, nfulnl_timer, 0);
 189
 190	inst->net = get_net(net);
 191	inst->peer_user_ns = user_ns;
 192	inst->peer_portid = portid;
 193	inst->group_num = group_num;
 194
 195	inst->qthreshold 	= NFULNL_QTHRESH_DEFAULT;
 196	inst->flushtimeout 	= NFULNL_TIMEOUT_DEFAULT;
 197	inst->nlbufsiz 		= NFULNL_NLBUFSIZ_DEFAULT;
 198	inst->copy_mode 	= NFULNL_COPY_PACKET;
 199	inst->copy_range 	= NFULNL_COPY_RANGE_MAX;
 200
 201	hlist_add_head_rcu(&inst->hlist,
 202		       &log->instance_table[instance_hashfn(group_num)]);
 203
 204
 205	spin_unlock_bh(&log->instances_lock);
 206
 207	return inst;
 208
 209out_unlock:
 210	spin_unlock_bh(&log->instances_lock);
 211	return ERR_PTR(err);
 212}
 213
 214static void __nfulnl_flush(struct nfulnl_instance *inst);
 215
 216/* called with BH disabled */
 217static void
 218__instance_destroy(struct nfulnl_instance *inst)
 219{
 220	/* first pull it out of the global list */
 221	hlist_del_rcu(&inst->hlist);
 222
 223	/* then flush all pending packets from skb */
 224
 225	spin_lock(&inst->lock);
 226
 227	/* lockless readers wont be able to use us */
 228	inst->copy_mode = NFULNL_COPY_DISABLED;
 229
 230	if (inst->skb)
 231		__nfulnl_flush(inst);
 232	spin_unlock(&inst->lock);
 233
 234	/* and finally put the refcount */
 235	instance_put(inst);
 236}
 237
 238static inline void
 239instance_destroy(struct nfnl_log_net *log,
 240		 struct nfulnl_instance *inst)
 241{
 242	spin_lock_bh(&log->instances_lock);
 243	__instance_destroy(inst);
 244	spin_unlock_bh(&log->instances_lock);
 245}
 246
 247static int
 248nfulnl_set_mode(struct nfulnl_instance *inst, u_int8_t mode,
 249		  unsigned int range)
 250{
 251	int status = 0;
 252
 253	spin_lock_bh(&inst->lock);
 254
 255	switch (mode) {
 256	case NFULNL_COPY_NONE:
 257	case NFULNL_COPY_META:
 258		inst->copy_mode = mode;
 259		inst->copy_range = 0;
 260		break;
 261
 262	case NFULNL_COPY_PACKET:
 263		inst->copy_mode = mode;
 264		if (range == 0)
 265			range = NFULNL_COPY_RANGE_MAX;
 266		inst->copy_range = min_t(unsigned int,
 267					 range, NFULNL_COPY_RANGE_MAX);
 268		break;
 269
 270	default:
 271		status = -EINVAL;
 272		break;
 273	}
 274
 275	spin_unlock_bh(&inst->lock);
 276
 277	return status;
 278}
 279
 280static int
 281nfulnl_set_nlbufsiz(struct nfulnl_instance *inst, u_int32_t nlbufsiz)
 282{
 283	int status;
 284
 285	spin_lock_bh(&inst->lock);
 286	if (nlbufsiz < NFULNL_NLBUFSIZ_DEFAULT)
 287		status = -ERANGE;
 288	else if (nlbufsiz > 131072)
 289		status = -ERANGE;
 290	else {
 291		inst->nlbufsiz = nlbufsiz;
 292		status = 0;
 293	}
 294	spin_unlock_bh(&inst->lock);
 295
 296	return status;
 297}
 298
 299static void
 300nfulnl_set_timeout(struct nfulnl_instance *inst, u_int32_t timeout)
 301{
 302	spin_lock_bh(&inst->lock);
 303	inst->flushtimeout = timeout;
 304	spin_unlock_bh(&inst->lock);
 305}
 306
 307static void
 308nfulnl_set_qthresh(struct nfulnl_instance *inst, u_int32_t qthresh)
 309{
 310	spin_lock_bh(&inst->lock);
 311	inst->qthreshold = qthresh;
 312	spin_unlock_bh(&inst->lock);
 313}
 314
 315static int
 316nfulnl_set_flags(struct nfulnl_instance *inst, u_int16_t flags)
 317{
 318	spin_lock_bh(&inst->lock);
 319	inst->flags = flags;
 320	spin_unlock_bh(&inst->lock);
 321
 322	return 0;
 323}
 324
 325static struct sk_buff *
 326nfulnl_alloc_skb(struct net *net, u32 peer_portid, unsigned int inst_size,
 327		 unsigned int pkt_size)
 328{
 329	struct sk_buff *skb;
 330	unsigned int n;
 331
 332	/* alloc skb which should be big enough for a whole multipart
 333	 * message.  WARNING: has to be <= 128k due to slab restrictions */
 334
 335	n = max(inst_size, pkt_size);
 336	skb = alloc_skb(n, GFP_ATOMIC | __GFP_NOWARN);
 337	if (!skb) {
 338		if (n > pkt_size) {
 339			/* try to allocate only as much as we need for current
 340			 * packet */
 341
 342			skb = alloc_skb(pkt_size, GFP_ATOMIC);
 343		}
 344	}
 345
 346	return skb;
 347}
 348
 349static void
 350__nfulnl_send(struct nfulnl_instance *inst)
 351{
 352	if (inst->qlen > 1) {
 353		struct nlmsghdr *nlh = nlmsg_put(inst->skb, 0, 0,
 354						 NLMSG_DONE,
 355						 sizeof(struct nfgenmsg),
 356						 0);
 357		if (WARN_ONCE(!nlh, "bad nlskb size: %u, tailroom %d\n",
 358			      inst->skb->len, skb_tailroom(inst->skb))) {
 359			kfree_skb(inst->skb);
 360			goto out;
 361		}
 362	}
 363	nfnetlink_unicast(inst->skb, inst->net, inst->peer_portid);
 
 364out:
 365	inst->qlen = 0;
 366	inst->skb = NULL;
 367}
 368
 369static void
 370__nfulnl_flush(struct nfulnl_instance *inst)
 371{
 372	/* timer holds a reference */
 373	if (del_timer(&inst->timer))
 374		instance_put(inst);
 375	if (inst->skb)
 376		__nfulnl_send(inst);
 377}
 378
 379static void
 380nfulnl_timer(struct timer_list *t)
 381{
 382	struct nfulnl_instance *inst = from_timer(inst, t, timer);
 383
 384	spin_lock_bh(&inst->lock);
 385	if (inst->skb)
 386		__nfulnl_send(inst);
 387	spin_unlock_bh(&inst->lock);
 388	instance_put(inst);
 389}
 390
 391static u32 nfulnl_get_bridge_size(const struct sk_buff *skb)
 392{
 393	u32 size = 0;
 394
 395	if (!skb_mac_header_was_set(skb))
 396		return 0;
 397
 398	if (skb_vlan_tag_present(skb)) {
 399		size += nla_total_size(0); /* nested */
 400		size += nla_total_size(sizeof(u16)); /* id */
 401		size += nla_total_size(sizeof(u16)); /* tag */
 402	}
 403
 404	if (skb->network_header > skb->mac_header)
 405		size += nla_total_size(skb->network_header - skb->mac_header);
 406
 407	return size;
 408}
 409
 410static int nfulnl_put_bridge(struct nfulnl_instance *inst, const struct sk_buff *skb)
 411{
 412	if (!skb_mac_header_was_set(skb))
 413		return 0;
 414
 415	if (skb_vlan_tag_present(skb)) {
 416		struct nlattr *nest;
 417
 418		nest = nla_nest_start(inst->skb, NFULA_VLAN);
 419		if (!nest)
 420			goto nla_put_failure;
 421
 422		if (nla_put_be16(inst->skb, NFULA_VLAN_TCI, htons(skb->vlan_tci)) ||
 423		    nla_put_be16(inst->skb, NFULA_VLAN_PROTO, skb->vlan_proto))
 424			goto nla_put_failure;
 425
 426		nla_nest_end(inst->skb, nest);
 427	}
 428
 429	if (skb->mac_header < skb->network_header) {
 430		int len = (int)(skb->network_header - skb->mac_header);
 431
 432		if (nla_put(inst->skb, NFULA_L2HDR, len, skb_mac_header(skb)))
 433			goto nla_put_failure;
 434	}
 435
 436	return 0;
 437
 438nla_put_failure:
 439	return -1;
 440}
 441
 442/* This is an inline function, we don't really care about a long
 443 * list of arguments */
 444static inline int
 445__build_packet_message(struct nfnl_log_net *log,
 446			struct nfulnl_instance *inst,
 447			const struct sk_buff *skb,
 448			unsigned int data_len,
 449			u_int8_t pf,
 450			unsigned int hooknum,
 451			const struct net_device *indev,
 452			const struct net_device *outdev,
 453			const char *prefix, unsigned int plen,
 454			const struct nfnl_ct_hook *nfnl_ct,
 455			struct nf_conn *ct, enum ip_conntrack_info ctinfo)
 456{
 457	struct nfulnl_msg_packet_hdr pmsg;
 458	struct nlmsghdr *nlh;
 
 459	sk_buff_data_t old_tail = inst->skb->tail;
 460	struct sock *sk;
 461	const unsigned char *hwhdrp;
 462
 463	nlh = nfnl_msg_put(inst->skb, 0, 0,
 464			   nfnl_msg_type(NFNL_SUBSYS_ULOG, NFULNL_MSG_PACKET),
 465			   0, pf, NFNETLINK_V0, htons(inst->group_num));
 466	if (!nlh)
 467		return -1;
 
 
 
 
 468
 469	memset(&pmsg, 0, sizeof(pmsg));
 470	pmsg.hw_protocol	= skb->protocol;
 471	pmsg.hook		= hooknum;
 472
 473	if (nla_put(inst->skb, NFULA_PACKET_HDR, sizeof(pmsg), &pmsg))
 474		goto nla_put_failure;
 475
 476	if (prefix &&
 477	    nla_put(inst->skb, NFULA_PREFIX, plen, prefix))
 478		goto nla_put_failure;
 479
 480	if (indev) {
 481#if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 482		if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 483				 htonl(indev->ifindex)))
 484			goto nla_put_failure;
 485#else
 486		if (pf == PF_BRIDGE) {
 487			/* Case 1: outdev is physical input device, we need to
 488			 * look for bridge group (when called from
 489			 * netfilter_bridge) */
 490			if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
 491					 htonl(indev->ifindex)) ||
 492			/* this is the bridge group "brX" */
 493			/* rcu_read_lock()ed by nf_hook_thresh or
 494			 * nf_log_packet.
 495			 */
 496			    nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 497					 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
 498				goto nla_put_failure;
 499		} else {
 500			struct net_device *physindev;
 501
 502			/* Case 2: indev is bridge group, we need to look for
 503			 * physical device (when called from ipv4) */
 504			if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 505					 htonl(indev->ifindex)))
 506				goto nla_put_failure;
 507
 508			physindev = nf_bridge_get_physindev(skb);
 509			if (physindev &&
 510			    nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
 511					 htonl(physindev->ifindex)))
 512				goto nla_put_failure;
 513		}
 514#endif
 515	}
 516
 517	if (outdev) {
 518#if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 519		if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 520				 htonl(outdev->ifindex)))
 521			goto nla_put_failure;
 522#else
 523		if (pf == PF_BRIDGE) {
 524			/* Case 1: outdev is physical output device, we need to
 525			 * look for bridge group (when called from
 526			 * netfilter_bridge) */
 527			if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
 528					 htonl(outdev->ifindex)) ||
 529			/* this is the bridge group "brX" */
 530			/* rcu_read_lock()ed by nf_hook_thresh or
 531			 * nf_log_packet.
 532			 */
 533			    nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 534					 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
 535				goto nla_put_failure;
 536		} else {
 537			struct net_device *physoutdev;
 538
 539			/* Case 2: indev is a bridge group, we need to look
 540			 * for physical device (when called from ipv4) */
 541			if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 542					 htonl(outdev->ifindex)))
 543				goto nla_put_failure;
 544
 545			physoutdev = nf_bridge_get_physoutdev(skb);
 546			if (physoutdev &&
 547			    nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
 548					 htonl(physoutdev->ifindex)))
 549				goto nla_put_failure;
 550		}
 551#endif
 552	}
 553
 554	if (skb->mark &&
 555	    nla_put_be32(inst->skb, NFULA_MARK, htonl(skb->mark)))
 556		goto nla_put_failure;
 557
 558	if (indev && skb->dev &&
 559	    skb->mac_header != skb->network_header) {
 560		struct nfulnl_msg_packet_hw phw;
 561		int len;
 562
 563		memset(&phw, 0, sizeof(phw));
 564		len = dev_parse_header(skb, phw.hw_addr);
 565		if (len > 0) {
 566			phw.hw_addrlen = htons(len);
 567			if (nla_put(inst->skb, NFULA_HWADDR, sizeof(phw), &phw))
 568				goto nla_put_failure;
 569		}
 570	}
 571
 572	if (indev && skb_mac_header_was_set(skb)) {
 573		if (nla_put_be16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type)) ||
 574		    nla_put_be16(inst->skb, NFULA_HWLEN,
 575				 htons(skb->dev->hard_header_len)))
 576			goto nla_put_failure;
 577
 578		hwhdrp = skb_mac_header(skb);
 579
 580		if (skb->dev->type == ARPHRD_SIT)
 581			hwhdrp -= ETH_HLEN;
 582
 583		if (hwhdrp >= skb->head &&
 584		    nla_put(inst->skb, NFULA_HWHEADER,
 585			    skb->dev->hard_header_len, hwhdrp))
 586			goto nla_put_failure;
 587	}
 588
 589	if (hooknum <= NF_INET_FORWARD && skb->tstamp) {
 590		struct nfulnl_msg_packet_timestamp ts;
 591		struct timespec64 kts = ktime_to_timespec64(skb->tstamp);
 592		ts.sec = cpu_to_be64(kts.tv_sec);
 593		ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
 594
 595		if (nla_put(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts))
 596			goto nla_put_failure;
 597	}
 598
 599	/* UID */
 600	sk = skb->sk;
 601	if (sk && sk_fullsock(sk)) {
 602		read_lock_bh(&sk->sk_callback_lock);
 603		if (sk->sk_socket && sk->sk_socket->file) {
 604			struct file *file = sk->sk_socket->file;
 605			const struct cred *cred = file->f_cred;
 606			struct user_namespace *user_ns = inst->peer_user_ns;
 607			__be32 uid = htonl(from_kuid_munged(user_ns, cred->fsuid));
 608			__be32 gid = htonl(from_kgid_munged(user_ns, cred->fsgid));
 609			read_unlock_bh(&sk->sk_callback_lock);
 610			if (nla_put_be32(inst->skb, NFULA_UID, uid) ||
 611			    nla_put_be32(inst->skb, NFULA_GID, gid))
 612				goto nla_put_failure;
 613		} else
 614			read_unlock_bh(&sk->sk_callback_lock);
 615	}
 616
 617	/* local sequence number */
 618	if ((inst->flags & NFULNL_CFG_F_SEQ) &&
 619	    nla_put_be32(inst->skb, NFULA_SEQ, htonl(inst->seq++)))
 620		goto nla_put_failure;
 621
 622	/* global sequence number */
 623	if ((inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) &&
 624	    nla_put_be32(inst->skb, NFULA_SEQ_GLOBAL,
 625			 htonl(atomic_inc_return(&log->global_seq))))
 626		goto nla_put_failure;
 627
 628	if (ct && nfnl_ct->build(inst->skb, ct, ctinfo,
 629				 NFULA_CT, NFULA_CT_INFO) < 0)
 630		goto nla_put_failure;
 631
 632	if ((pf == NFPROTO_NETDEV || pf == NFPROTO_BRIDGE) &&
 633	    nfulnl_put_bridge(inst, skb) < 0)
 634		goto nla_put_failure;
 635
 636	if (data_len) {
 637		struct nlattr *nla;
 638		int size = nla_attr_size(data_len);
 639
 640		if (skb_tailroom(inst->skb) < nla_total_size(data_len))
 641			goto nla_put_failure;
 642
 643		nla = skb_put(inst->skb, nla_total_size(data_len));
 644		nla->nla_type = NFULA_PAYLOAD;
 645		nla->nla_len = size;
 646
 647		if (skb_copy_bits(skb, 0, nla_data(nla), data_len))
 648			BUG();
 649	}
 650
 651	nlh->nlmsg_len = inst->skb->tail - old_tail;
 652	return 0;
 653
 654nla_put_failure:
 655	PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n");
 656	return -1;
 657}
 658
 659static const struct nf_loginfo default_loginfo = {
 660	.type =		NF_LOG_TYPE_ULOG,
 661	.u = {
 662		.ulog = {
 663			.copy_len	= 0xffff,
 664			.group		= 0,
 665			.qthreshold	= 1,
 666		},
 667	},
 668};
 669
 670/* log handler for internal netfilter logging api */
 671static void
 672nfulnl_log_packet(struct net *net,
 673		  u_int8_t pf,
 674		  unsigned int hooknum,
 675		  const struct sk_buff *skb,
 676		  const struct net_device *in,
 677		  const struct net_device *out,
 678		  const struct nf_loginfo *li_user,
 679		  const char *prefix)
 680{
 681	size_t size;
 682	unsigned int data_len;
 683	struct nfulnl_instance *inst;
 684	const struct nf_loginfo *li;
 685	unsigned int qthreshold;
 686	unsigned int plen = 0;
 687	struct nfnl_log_net *log = nfnl_log_pernet(net);
 688	const struct nfnl_ct_hook *nfnl_ct = NULL;
 689	struct nf_conn *ct = NULL;
 690	enum ip_conntrack_info ctinfo;
 691
 692	if (li_user && li_user->type == NF_LOG_TYPE_ULOG)
 693		li = li_user;
 694	else
 695		li = &default_loginfo;
 696
 697	inst = instance_lookup_get(log, li->u.ulog.group);
 698	if (!inst)
 699		return;
 700
 
 701	if (prefix)
 702		plen = strlen(prefix) + 1;
 703
 704	/* FIXME: do we want to make the size calculation conditional based on
 705	 * what is actually present?  way more branches and checks, but more
 706	 * memory efficient... */
 707	size = nlmsg_total_size(sizeof(struct nfgenmsg))
 708		+ nla_total_size(sizeof(struct nfulnl_msg_packet_hdr))
 709		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
 710		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
 711#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 712		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
 713		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
 714#endif
 715		+ nla_total_size(sizeof(u_int32_t))	/* mark */
 716		+ nla_total_size(sizeof(u_int32_t))	/* uid */
 717		+ nla_total_size(sizeof(u_int32_t))	/* gid */
 718		+ nla_total_size(plen)			/* prefix */
 719		+ nla_total_size(sizeof(struct nfulnl_msg_packet_hw))
 720		+ nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp))
 721		+ nla_total_size(sizeof(struct nfgenmsg));	/* NLMSG_DONE */
 722
 723	if (in && skb_mac_header_was_set(skb)) {
 724		size += nla_total_size(skb->dev->hard_header_len)
 725			+ nla_total_size(sizeof(u_int16_t))	/* hwtype */
 726			+ nla_total_size(sizeof(u_int16_t));	/* hwlen */
 727	}
 728
 729	spin_lock_bh(&inst->lock);
 730
 731	if (inst->flags & NFULNL_CFG_F_SEQ)
 732		size += nla_total_size(sizeof(u_int32_t));
 733	if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
 734		size += nla_total_size(sizeof(u_int32_t));
 735#if IS_ENABLED(CONFIG_NF_CONNTRACK)
 736	if (inst->flags & NFULNL_CFG_F_CONNTRACK) {
 737		nfnl_ct = rcu_dereference(nfnl_ct_hook);
 738		if (nfnl_ct != NULL) {
 739			ct = nf_ct_get(skb, &ctinfo);
 740			if (ct != NULL)
 741				size += nfnl_ct->build_size(ct);
 742		}
 743	}
 744#endif
 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 sk_buff *skb, const struct nfnl_info *info,
 849			      const struct nlattr * const nfula[])
 
 
 850{
 851	return -ENOTSUPP;
 852}
 853
 854static struct nf_logger nfulnl_logger __read_mostly = {
 855	.name	= "nfnetlink_log",
 856	.type	= NF_LOG_TYPE_ULOG,
 857	.logfn	= nfulnl_log_packet,
 858	.me	= THIS_MODULE,
 859};
 860
 861static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = {
 862	[NFULA_CFG_CMD]		= { .len = sizeof(struct nfulnl_msg_config_cmd) },
 863	[NFULA_CFG_MODE]	= { .len = sizeof(struct nfulnl_msg_config_mode) },
 864	[NFULA_CFG_TIMEOUT]	= { .type = NLA_U32 },
 865	[NFULA_CFG_QTHRESH]	= { .type = NLA_U32 },
 866	[NFULA_CFG_NLBUFSIZ]	= { .type = NLA_U32 },
 867	[NFULA_CFG_FLAGS]	= { .type = NLA_U16 },
 868};
 869
 870static int nfulnl_recv_config(struct sk_buff *skb, const struct nfnl_info *info,
 871			      const struct nlattr * const nfula[])
 
 
 872{
 873	struct nfnl_log_net *log = nfnl_log_pernet(info->net);
 874	u_int16_t group_num = ntohs(info->nfmsg->res_id);
 875	struct nfulnl_msg_config_cmd *cmd = NULL;
 876	struct nfulnl_instance *inst;
 877	u16 flags = 0;
 
 878	int ret = 0;
 
 879
 880	if (nfula[NFULA_CFG_CMD]) {
 881		u_int8_t pf = info->nfmsg->nfgen_family;
 882		cmd = nla_data(nfula[NFULA_CFG_CMD]);
 883
 884		/* Commands without queue context */
 885		switch (cmd->command) {
 886		case NFULNL_CFG_CMD_PF_BIND:
 887			return nf_log_bind_pf(info->net, pf, &nfulnl_logger);
 888		case NFULNL_CFG_CMD_PF_UNBIND:
 889			nf_log_unbind_pf(info->net, pf);
 890			return 0;
 891		}
 892	}
 893
 894	inst = instance_lookup_get(log, group_num);
 895	if (inst && inst->peer_portid != NETLINK_CB(skb).portid) {
 896		ret = -EPERM;
 897		goto out_put;
 898	}
 899
 900	/* Check if we support these flags in first place, dependencies should
 901	 * be there too not to break atomicity.
 902	 */
 903	if (nfula[NFULA_CFG_FLAGS]) {
 904		flags = ntohs(nla_get_be16(nfula[NFULA_CFG_FLAGS]));
 905
 906		if ((flags & NFULNL_CFG_F_CONNTRACK) &&
 907		    !rcu_access_pointer(nfnl_ct_hook)) {
 908#ifdef CONFIG_MODULES
 909			nfnl_unlock(NFNL_SUBSYS_ULOG);
 910			request_module("ip_conntrack_netlink");
 911			nfnl_lock(NFNL_SUBSYS_ULOG);
 912			if (rcu_access_pointer(nfnl_ct_hook)) {
 913				ret = -EAGAIN;
 914				goto out_put;
 915			}
 916#endif
 917			ret = -EOPNOTSUPP;
 918			goto out_put;
 919		}
 920	}
 921
 922	if (cmd != NULL) {
 923		switch (cmd->command) {
 924		case NFULNL_CFG_CMD_BIND:
 925			if (inst) {
 926				ret = -EBUSY;
 927				goto out_put;
 928			}
 929
 930			inst = instance_create(info->net, group_num,
 931					       NETLINK_CB(skb).portid,
 932					       sk_user_ns(NETLINK_CB(skb).sk));
 933			if (IS_ERR(inst)) {
 934				ret = PTR_ERR(inst);
 935				goto out;
 936			}
 937			break;
 938		case NFULNL_CFG_CMD_UNBIND:
 939			if (!inst) {
 940				ret = -ENODEV;
 941				goto out;
 942			}
 943
 944			instance_destroy(log, inst);
 945			goto out_put;
 946		default:
 947			ret = -ENOTSUPP;
 948			goto out_put;
 949		}
 950	} else if (!inst) {
 951		ret = -ENODEV;
 952		goto out;
 953	}
 954
 955	if (nfula[NFULA_CFG_MODE]) {
 956		struct nfulnl_msg_config_mode *params =
 957			nla_data(nfula[NFULA_CFG_MODE]);
 958
 959		nfulnl_set_mode(inst, params->copy_mode,
 960				ntohl(params->copy_range));
 961	}
 962
 963	if (nfula[NFULA_CFG_TIMEOUT]) {
 964		__be32 timeout = nla_get_be32(nfula[NFULA_CFG_TIMEOUT]);
 965
 966		nfulnl_set_timeout(inst, ntohl(timeout));
 967	}
 968
 969	if (nfula[NFULA_CFG_NLBUFSIZ]) {
 970		__be32 nlbufsiz = nla_get_be32(nfula[NFULA_CFG_NLBUFSIZ]);
 971
 972		nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz));
 973	}
 974
 975	if (nfula[NFULA_CFG_QTHRESH]) {
 976		__be32 qthresh = nla_get_be32(nfula[NFULA_CFG_QTHRESH]);
 977
 978		nfulnl_set_qthresh(inst, ntohl(qthresh));
 979	}
 980
 981	if (nfula[NFULA_CFG_FLAGS])
 982		nfulnl_set_flags(inst, flags);
 983
 984out_put:
 985	instance_put(inst);
 986out:
 987	return ret;
 988}
 989
 990static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = {
 991	[NFULNL_MSG_PACKET]	= {
 992		.call		= nfulnl_recv_unsupp,
 993		.type		= NFNL_CB_MUTEX,
 994		.attr_count	= NFULA_MAX,
 995	},
 996	[NFULNL_MSG_CONFIG]	= {
 997		.call		= nfulnl_recv_config,
 998		.type		= NFNL_CB_MUTEX,
 999		.attr_count	= NFULA_CFG_MAX,
1000		.policy		= nfula_cfg_policy
1001	},
1002};
1003
1004static const struct nfnetlink_subsystem nfulnl_subsys = {
1005	.name		= "log",
1006	.subsys_id	= NFNL_SUBSYS_ULOG,
1007	.cb_count	= NFULNL_MSG_MAX,
1008	.cb		= nfulnl_cb,
1009};
1010
1011#ifdef CONFIG_PROC_FS
1012struct iter_state {
1013	struct seq_net_private p;
1014	unsigned int bucket;
1015};
1016
1017static struct hlist_node *get_first(struct net *net, struct iter_state *st)
1018{
1019	struct nfnl_log_net *log;
1020	if (!st)
1021		return NULL;
1022
1023	log = nfnl_log_pernet(net);
1024
1025	for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1026		struct hlist_head *head = &log->instance_table[st->bucket];
1027
1028		if (!hlist_empty(head))
1029			return rcu_dereference_bh(hlist_first_rcu(head));
1030	}
1031	return NULL;
1032}
1033
1034static struct hlist_node *get_next(struct net *net, struct iter_state *st,
1035				   struct hlist_node *h)
1036{
1037	h = rcu_dereference_bh(hlist_next_rcu(h));
1038	while (!h) {
1039		struct nfnl_log_net *log;
1040		struct hlist_head *head;
1041
1042		if (++st->bucket >= INSTANCE_BUCKETS)
1043			return NULL;
1044
1045		log = nfnl_log_pernet(net);
1046		head = &log->instance_table[st->bucket];
1047		h = rcu_dereference_bh(hlist_first_rcu(head));
1048	}
1049	return h;
1050}
1051
1052static struct hlist_node *get_idx(struct net *net, struct iter_state *st,
1053				  loff_t pos)
1054{
1055	struct hlist_node *head;
1056	head = get_first(net, st);
1057
1058	if (head)
1059		while (pos && (head = get_next(net, st, head)))
1060			pos--;
1061	return pos ? NULL : head;
1062}
1063
1064static void *seq_start(struct seq_file *s, loff_t *pos)
1065	__acquires(rcu_bh)
1066{
1067	rcu_read_lock_bh();
1068	return get_idx(seq_file_net(s), s->private, *pos);
1069}
1070
1071static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1072{
1073	(*pos)++;
1074	return get_next(seq_file_net(s), s->private, v);
1075}
1076
1077static void seq_stop(struct seq_file *s, void *v)
1078	__releases(rcu_bh)
1079{
1080	rcu_read_unlock_bh();
1081}
1082
1083static int seq_show(struct seq_file *s, void *v)
1084{
1085	const struct nfulnl_instance *inst = v;
1086
1087	seq_printf(s, "%5u %6u %5u %1u %5u %6u %2u\n",
1088		   inst->group_num,
1089		   inst->peer_portid, inst->qlen,
1090		   inst->copy_mode, inst->copy_range,
1091		   inst->flushtimeout, refcount_read(&inst->use));
1092
1093	return 0;
1094}
1095
1096static const struct seq_operations nful_seq_ops = {
1097	.start	= seq_start,
1098	.next	= seq_next,
1099	.stop	= seq_stop,
1100	.show	= seq_show,
1101};
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1102#endif /* PROC_FS */
1103
1104static int __net_init nfnl_log_net_init(struct net *net)
1105{
1106	unsigned int i;
1107	struct nfnl_log_net *log = nfnl_log_pernet(net);
1108#ifdef CONFIG_PROC_FS
1109	struct proc_dir_entry *proc;
1110	kuid_t root_uid;
1111	kgid_t root_gid;
1112#endif
1113
1114	for (i = 0; i < INSTANCE_BUCKETS; i++)
1115		INIT_HLIST_HEAD(&log->instance_table[i]);
1116	spin_lock_init(&log->instances_lock);
1117
1118#ifdef CONFIG_PROC_FS
1119	proc = proc_create_net("nfnetlink_log", 0440, net->nf.proc_netfilter,
1120			&nful_seq_ops, sizeof(struct iter_state));
1121	if (!proc)
1122		return -ENOMEM;
1123
1124	root_uid = make_kuid(net->user_ns, 0);
1125	root_gid = make_kgid(net->user_ns, 0);
1126	if (uid_valid(root_uid) && gid_valid(root_gid))
1127		proc_set_user(proc, root_uid, root_gid);
1128#endif
1129	return 0;
1130}
1131
1132static void __net_exit nfnl_log_net_exit(struct net *net)
1133{
1134	struct nfnl_log_net *log = nfnl_log_pernet(net);
1135	unsigned int i;
1136
1137#ifdef CONFIG_PROC_FS
1138	remove_proc_entry("nfnetlink_log", net->nf.proc_netfilter);
1139#endif
1140	nf_log_unset(net, &nfulnl_logger);
1141	for (i = 0; i < INSTANCE_BUCKETS; i++)
1142		WARN_ON_ONCE(!hlist_empty(&log->instance_table[i]));
1143}
1144
1145static struct pernet_operations nfnl_log_net_ops = {
1146	.init	= nfnl_log_net_init,
1147	.exit	= nfnl_log_net_exit,
1148	.id	= &nfnl_log_net_id,
1149	.size	= sizeof(struct nfnl_log_net),
1150};
1151
1152static int __init nfnetlink_log_init(void)
1153{
1154	int status;
1155
1156	status = register_pernet_subsys(&nfnl_log_net_ops);
1157	if (status < 0) {
1158		pr_err("failed to register pernet ops\n");
1159		goto out;
1160	}
1161
1162	netlink_register_notifier(&nfulnl_rtnl_notifier);
1163	status = nfnetlink_subsys_register(&nfulnl_subsys);
1164	if (status < 0) {
1165		pr_err("failed to create netlink socket\n");
1166		goto cleanup_netlink_notifier;
1167	}
1168
1169	status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger);
1170	if (status < 0) {
1171		pr_err("failed to register logger\n");
1172		goto cleanup_subsys;
1173	}
1174
1175	return status;
1176
1177cleanup_subsys:
1178	nfnetlink_subsys_unregister(&nfulnl_subsys);
1179cleanup_netlink_notifier:
1180	netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1181	unregister_pernet_subsys(&nfnl_log_net_ops);
1182out:
1183	return status;
1184}
1185
1186static void __exit nfnetlink_log_fini(void)
1187{
1188	nfnetlink_subsys_unregister(&nfulnl_subsys);
1189	netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1190	unregister_pernet_subsys(&nfnl_log_net_ops);
1191	nf_log_unregister(&nfulnl_logger);
1192}
1193
1194MODULE_DESCRIPTION("netfilter userspace logging");
1195MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1196MODULE_LICENSE("GPL");
1197MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG);
1198MODULE_ALIAS_NF_LOGGER(AF_INET, 1);
1199MODULE_ALIAS_NF_LOGGER(AF_INET6, 1);
1200MODULE_ALIAS_NF_LOGGER(AF_BRIDGE, 1);
1201MODULE_ALIAS_NF_LOGGER(3, 1); /* NFPROTO_ARP */
1202MODULE_ALIAS_NF_LOGGER(5, 1); /* NFPROTO_NETDEV */
1203
1204module_init(nfnetlink_log_init);
1205module_exit(nfnetlink_log_fini);
v4.17
 
   1/*
   2 * This is a module which is used for logging packets to userspace via
   3 * nfetlink.
   4 *
   5 * (C) 2005 by Harald Welte <laforge@netfilter.org>
   6 * (C) 2006-2012 Patrick McHardy <kaber@trash.net>
   7 *
   8 * Based on the old ipv4-only ipt_ULOG.c:
   9 * (C) 2000-2004 by Harald Welte <laforge@netfilter.org>
  10 *
  11 * This program is free software; you can redistribute it and/or modify
  12 * it under the terms of the GNU General Public License version 2 as
  13 * published by the Free Software Foundation.
  14 */
  15
  16#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  17
  18#include <linux/module.h>
  19#include <linux/skbuff.h>
  20#include <linux/if_arp.h>
  21#include <linux/init.h>
  22#include <linux/ip.h>
  23#include <linux/ipv6.h>
  24#include <linux/netdevice.h>
  25#include <linux/netfilter.h>
  26#include <linux/netfilter_bridge.h>
  27#include <net/netlink.h>
  28#include <linux/netfilter/nfnetlink.h>
  29#include <linux/netfilter/nfnetlink_log.h>
  30#include <linux/netfilter/nf_conntrack_common.h>
  31#include <linux/spinlock.h>
  32#include <linux/sysctl.h>
  33#include <linux/proc_fs.h>
  34#include <linux/security.h>
  35#include <linux/list.h>
  36#include <linux/slab.h>
  37#include <net/sock.h>
  38#include <net/netfilter/nf_log.h>
  39#include <net/netns/generic.h>
  40#include <net/netfilter/nfnetlink_log.h>
  41
  42#include <linux/atomic.h>
  43#include <linux/refcount.h>
  44
  45
  46#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  47#include "../bridge/br_private.h"
  48#endif
  49
 
 
 
 
 
  50#define NFULNL_NLBUFSIZ_DEFAULT	NLMSG_GOODSIZE
  51#define NFULNL_TIMEOUT_DEFAULT 	100	/* every second */
  52#define NFULNL_QTHRESH_DEFAULT 	100	/* 100 packets */
  53/* max packet size is limited by 16-bit struct nfattr nfa_len field */
  54#define NFULNL_COPY_RANGE_MAX	(0xFFFF - NLA_HDRLEN)
  55
  56#define PRINTR(x, args...)	do { if (net_ratelimit()) \
  57				     printk(x, ## args); } while (0);
  58
  59struct nfulnl_instance {
  60	struct hlist_node hlist;	/* global list of instances */
  61	spinlock_t lock;
  62	refcount_t use;			/* use count */
  63
  64	unsigned int qlen;		/* number of nlmsgs in skb */
  65	struct sk_buff *skb;		/* pre-allocatd skb */
  66	struct timer_list timer;
  67	struct net *net;
  68	struct user_namespace *peer_user_ns;	/* User namespace of the peer process */
  69	u32 peer_portid;		/* PORTID of the peer process */
  70
  71	/* configurable parameters */
  72	unsigned int flushtimeout;	/* timeout until queue flush */
  73	unsigned int nlbufsiz;		/* netlink buffer allocation size */
  74	unsigned int qthreshold;	/* threshold of the queue */
  75	u_int32_t copy_range;
  76	u_int32_t seq;			/* instance-local sequential counter */
  77	u_int16_t group_num;		/* number of this queue */
  78	u_int16_t flags;
  79	u_int8_t copy_mode;
  80	struct rcu_head rcu;
  81};
  82
  83#define INSTANCE_BUCKETS	16
  84
  85static unsigned int nfnl_log_net_id __read_mostly;
  86
  87struct nfnl_log_net {
  88	spinlock_t instances_lock;
  89	struct hlist_head instance_table[INSTANCE_BUCKETS];
  90	atomic_t global_seq;
  91};
  92
  93static struct nfnl_log_net *nfnl_log_pernet(struct net *net)
  94{
  95	return net_generic(net, nfnl_log_net_id);
  96}
  97
  98static inline u_int8_t instance_hashfn(u_int16_t group_num)
  99{
 100	return ((group_num & 0xff) % INSTANCE_BUCKETS);
 101}
 102
 103static struct nfulnl_instance *
 104__instance_lookup(struct nfnl_log_net *log, u_int16_t group_num)
 105{
 106	struct hlist_head *head;
 107	struct nfulnl_instance *inst;
 108
 109	head = &log->instance_table[instance_hashfn(group_num)];
 110	hlist_for_each_entry_rcu(inst, head, hlist) {
 111		if (inst->group_num == group_num)
 112			return inst;
 113	}
 114	return NULL;
 115}
 116
 117static inline void
 118instance_get(struct nfulnl_instance *inst)
 119{
 120	refcount_inc(&inst->use);
 121}
 122
 123static struct nfulnl_instance *
 124instance_lookup_get(struct nfnl_log_net *log, u_int16_t group_num)
 125{
 126	struct nfulnl_instance *inst;
 127
 128	rcu_read_lock_bh();
 129	inst = __instance_lookup(log, group_num);
 130	if (inst && !refcount_inc_not_zero(&inst->use))
 131		inst = NULL;
 132	rcu_read_unlock_bh();
 133
 134	return inst;
 135}
 136
 137static void nfulnl_instance_free_rcu(struct rcu_head *head)
 138{
 139	struct nfulnl_instance *inst =
 140		container_of(head, struct nfulnl_instance, rcu);
 141
 142	put_net(inst->net);
 143	kfree(inst);
 144	module_put(THIS_MODULE);
 145}
 146
 147static void
 148instance_put(struct nfulnl_instance *inst)
 149{
 150	if (inst && refcount_dec_and_test(&inst->use))
 151		call_rcu_bh(&inst->rcu, nfulnl_instance_free_rcu);
 152}
 153
 154static void nfulnl_timer(struct timer_list *t);
 155
 156static struct nfulnl_instance *
 157instance_create(struct net *net, u_int16_t group_num,
 158		u32 portid, struct user_namespace *user_ns)
 159{
 160	struct nfulnl_instance *inst;
 161	struct nfnl_log_net *log = nfnl_log_pernet(net);
 162	int err;
 163
 164	spin_lock_bh(&log->instances_lock);
 165	if (__instance_lookup(log, group_num)) {
 166		err = -EEXIST;
 167		goto out_unlock;
 168	}
 169
 170	inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
 171	if (!inst) {
 172		err = -ENOMEM;
 173		goto out_unlock;
 174	}
 175
 176	if (!try_module_get(THIS_MODULE)) {
 177		kfree(inst);
 178		err = -EAGAIN;
 179		goto out_unlock;
 180	}
 181
 182	INIT_HLIST_NODE(&inst->hlist);
 183	spin_lock_init(&inst->lock);
 184	/* needs to be two, since we _put() after creation */
 185	refcount_set(&inst->use, 2);
 186
 187	timer_setup(&inst->timer, nfulnl_timer, 0);
 188
 189	inst->net = get_net(net);
 190	inst->peer_user_ns = user_ns;
 191	inst->peer_portid = portid;
 192	inst->group_num = group_num;
 193
 194	inst->qthreshold 	= NFULNL_QTHRESH_DEFAULT;
 195	inst->flushtimeout 	= NFULNL_TIMEOUT_DEFAULT;
 196	inst->nlbufsiz 		= NFULNL_NLBUFSIZ_DEFAULT;
 197	inst->copy_mode 	= NFULNL_COPY_PACKET;
 198	inst->copy_range 	= NFULNL_COPY_RANGE_MAX;
 199
 200	hlist_add_head_rcu(&inst->hlist,
 201		       &log->instance_table[instance_hashfn(group_num)]);
 202
 203
 204	spin_unlock_bh(&log->instances_lock);
 205
 206	return inst;
 207
 208out_unlock:
 209	spin_unlock_bh(&log->instances_lock);
 210	return ERR_PTR(err);
 211}
 212
 213static void __nfulnl_flush(struct nfulnl_instance *inst);
 214
 215/* called with BH disabled */
 216static void
 217__instance_destroy(struct nfulnl_instance *inst)
 218{
 219	/* first pull it out of the global list */
 220	hlist_del_rcu(&inst->hlist);
 221
 222	/* then flush all pending packets from skb */
 223
 224	spin_lock(&inst->lock);
 225
 226	/* lockless readers wont be able to use us */
 227	inst->copy_mode = NFULNL_COPY_DISABLED;
 228
 229	if (inst->skb)
 230		__nfulnl_flush(inst);
 231	spin_unlock(&inst->lock);
 232
 233	/* and finally put the refcount */
 234	instance_put(inst);
 235}
 236
 237static inline void
 238instance_destroy(struct nfnl_log_net *log,
 239		 struct nfulnl_instance *inst)
 240{
 241	spin_lock_bh(&log->instances_lock);
 242	__instance_destroy(inst);
 243	spin_unlock_bh(&log->instances_lock);
 244}
 245
 246static int
 247nfulnl_set_mode(struct nfulnl_instance *inst, u_int8_t mode,
 248		  unsigned int range)
 249{
 250	int status = 0;
 251
 252	spin_lock_bh(&inst->lock);
 253
 254	switch (mode) {
 255	case NFULNL_COPY_NONE:
 256	case NFULNL_COPY_META:
 257		inst->copy_mode = mode;
 258		inst->copy_range = 0;
 259		break;
 260
 261	case NFULNL_COPY_PACKET:
 262		inst->copy_mode = mode;
 263		if (range == 0)
 264			range = NFULNL_COPY_RANGE_MAX;
 265		inst->copy_range = min_t(unsigned int,
 266					 range, NFULNL_COPY_RANGE_MAX);
 267		break;
 268
 269	default:
 270		status = -EINVAL;
 271		break;
 272	}
 273
 274	spin_unlock_bh(&inst->lock);
 275
 276	return status;
 277}
 278
 279static int
 280nfulnl_set_nlbufsiz(struct nfulnl_instance *inst, u_int32_t nlbufsiz)
 281{
 282	int status;
 283
 284	spin_lock_bh(&inst->lock);
 285	if (nlbufsiz < NFULNL_NLBUFSIZ_DEFAULT)
 286		status = -ERANGE;
 287	else if (nlbufsiz > 131072)
 288		status = -ERANGE;
 289	else {
 290		inst->nlbufsiz = nlbufsiz;
 291		status = 0;
 292	}
 293	spin_unlock_bh(&inst->lock);
 294
 295	return status;
 296}
 297
 298static void
 299nfulnl_set_timeout(struct nfulnl_instance *inst, u_int32_t timeout)
 300{
 301	spin_lock_bh(&inst->lock);
 302	inst->flushtimeout = timeout;
 303	spin_unlock_bh(&inst->lock);
 304}
 305
 306static void
 307nfulnl_set_qthresh(struct nfulnl_instance *inst, u_int32_t qthresh)
 308{
 309	spin_lock_bh(&inst->lock);
 310	inst->qthreshold = qthresh;
 311	spin_unlock_bh(&inst->lock);
 312}
 313
 314static int
 315nfulnl_set_flags(struct nfulnl_instance *inst, u_int16_t flags)
 316{
 317	spin_lock_bh(&inst->lock);
 318	inst->flags = flags;
 319	spin_unlock_bh(&inst->lock);
 320
 321	return 0;
 322}
 323
 324static struct sk_buff *
 325nfulnl_alloc_skb(struct net *net, u32 peer_portid, unsigned int inst_size,
 326		 unsigned int pkt_size)
 327{
 328	struct sk_buff *skb;
 329	unsigned int n;
 330
 331	/* alloc skb which should be big enough for a whole multipart
 332	 * message.  WARNING: has to be <= 128k due to slab restrictions */
 333
 334	n = max(inst_size, pkt_size);
 335	skb = alloc_skb(n, GFP_ATOMIC | __GFP_NOWARN);
 336	if (!skb) {
 337		if (n > pkt_size) {
 338			/* try to allocate only as much as we need for current
 339			 * packet */
 340
 341			skb = alloc_skb(pkt_size, GFP_ATOMIC);
 342		}
 343	}
 344
 345	return skb;
 346}
 347
 348static void
 349__nfulnl_send(struct nfulnl_instance *inst)
 350{
 351	if (inst->qlen > 1) {
 352		struct nlmsghdr *nlh = nlmsg_put(inst->skb, 0, 0,
 353						 NLMSG_DONE,
 354						 sizeof(struct nfgenmsg),
 355						 0);
 356		if (WARN_ONCE(!nlh, "bad nlskb size: %u, tailroom %d\n",
 357			      inst->skb->len, skb_tailroom(inst->skb))) {
 358			kfree_skb(inst->skb);
 359			goto out;
 360		}
 361	}
 362	nfnetlink_unicast(inst->skb, inst->net, inst->peer_portid,
 363			  MSG_DONTWAIT);
 364out:
 365	inst->qlen = 0;
 366	inst->skb = NULL;
 367}
 368
 369static void
 370__nfulnl_flush(struct nfulnl_instance *inst)
 371{
 372	/* timer holds a reference */
 373	if (del_timer(&inst->timer))
 374		instance_put(inst);
 375	if (inst->skb)
 376		__nfulnl_send(inst);
 377}
 378
 379static void
 380nfulnl_timer(struct timer_list *t)
 381{
 382	struct nfulnl_instance *inst = from_timer(inst, t, timer);
 383
 384	spin_lock_bh(&inst->lock);
 385	if (inst->skb)
 386		__nfulnl_send(inst);
 387	spin_unlock_bh(&inst->lock);
 388	instance_put(inst);
 389}
 390
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 391/* This is an inline function, we don't really care about a long
 392 * list of arguments */
 393static inline int
 394__build_packet_message(struct nfnl_log_net *log,
 395			struct nfulnl_instance *inst,
 396			const struct sk_buff *skb,
 397			unsigned int data_len,
 398			u_int8_t pf,
 399			unsigned int hooknum,
 400			const struct net_device *indev,
 401			const struct net_device *outdev,
 402			const char *prefix, unsigned int plen,
 403			const struct nfnl_ct_hook *nfnl_ct,
 404			struct nf_conn *ct, enum ip_conntrack_info ctinfo)
 405{
 406	struct nfulnl_msg_packet_hdr pmsg;
 407	struct nlmsghdr *nlh;
 408	struct nfgenmsg *nfmsg;
 409	sk_buff_data_t old_tail = inst->skb->tail;
 410	struct sock *sk;
 411	const unsigned char *hwhdrp;
 412
 413	nlh = nlmsg_put(inst->skb, 0, 0,
 414			nfnl_msg_type(NFNL_SUBSYS_ULOG, NFULNL_MSG_PACKET),
 415			sizeof(struct nfgenmsg), 0);
 416	if (!nlh)
 417		return -1;
 418	nfmsg = nlmsg_data(nlh);
 419	nfmsg->nfgen_family = pf;
 420	nfmsg->version = NFNETLINK_V0;
 421	nfmsg->res_id = htons(inst->group_num);
 422
 423	memset(&pmsg, 0, sizeof(pmsg));
 424	pmsg.hw_protocol	= skb->protocol;
 425	pmsg.hook		= hooknum;
 426
 427	if (nla_put(inst->skb, NFULA_PACKET_HDR, sizeof(pmsg), &pmsg))
 428		goto nla_put_failure;
 429
 430	if (prefix &&
 431	    nla_put(inst->skb, NFULA_PREFIX, plen, prefix))
 432		goto nla_put_failure;
 433
 434	if (indev) {
 435#if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 436		if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 437				 htonl(indev->ifindex)))
 438			goto nla_put_failure;
 439#else
 440		if (pf == PF_BRIDGE) {
 441			/* Case 1: outdev is physical input device, we need to
 442			 * look for bridge group (when called from
 443			 * netfilter_bridge) */
 444			if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
 445					 htonl(indev->ifindex)) ||
 446			/* this is the bridge group "brX" */
 447			/* rcu_read_lock()ed by nf_hook_thresh or
 448			 * nf_log_packet.
 449			 */
 450			    nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 451					 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
 452				goto nla_put_failure;
 453		} else {
 454			struct net_device *physindev;
 455
 456			/* Case 2: indev is bridge group, we need to look for
 457			 * physical device (when called from ipv4) */
 458			if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
 459					 htonl(indev->ifindex)))
 460				goto nla_put_failure;
 461
 462			physindev = nf_bridge_get_physindev(skb);
 463			if (physindev &&
 464			    nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
 465					 htonl(physindev->ifindex)))
 466				goto nla_put_failure;
 467		}
 468#endif
 469	}
 470
 471	if (outdev) {
 472#if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 473		if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 474				 htonl(outdev->ifindex)))
 475			goto nla_put_failure;
 476#else
 477		if (pf == PF_BRIDGE) {
 478			/* Case 1: outdev is physical output device, we need to
 479			 * look for bridge group (when called from
 480			 * netfilter_bridge) */
 481			if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
 482					 htonl(outdev->ifindex)) ||
 483			/* this is the bridge group "brX" */
 484			/* rcu_read_lock()ed by nf_hook_thresh or
 485			 * nf_log_packet.
 486			 */
 487			    nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 488					 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
 489				goto nla_put_failure;
 490		} else {
 491			struct net_device *physoutdev;
 492
 493			/* Case 2: indev is a bridge group, we need to look
 494			 * for physical device (when called from ipv4) */
 495			if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
 496					 htonl(outdev->ifindex)))
 497				goto nla_put_failure;
 498
 499			physoutdev = nf_bridge_get_physoutdev(skb);
 500			if (physoutdev &&
 501			    nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
 502					 htonl(physoutdev->ifindex)))
 503				goto nla_put_failure;
 504		}
 505#endif
 506	}
 507
 508	if (skb->mark &&
 509	    nla_put_be32(inst->skb, NFULA_MARK, htonl(skb->mark)))
 510		goto nla_put_failure;
 511
 512	if (indev && skb->dev &&
 513	    skb->mac_header != skb->network_header) {
 514		struct nfulnl_msg_packet_hw phw;
 515		int len;
 516
 517		memset(&phw, 0, sizeof(phw));
 518		len = dev_parse_header(skb, phw.hw_addr);
 519		if (len > 0) {
 520			phw.hw_addrlen = htons(len);
 521			if (nla_put(inst->skb, NFULA_HWADDR, sizeof(phw), &phw))
 522				goto nla_put_failure;
 523		}
 524	}
 525
 526	if (indev && skb_mac_header_was_set(skb)) {
 527		if (nla_put_be16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type)) ||
 528		    nla_put_be16(inst->skb, NFULA_HWLEN,
 529				 htons(skb->dev->hard_header_len)))
 530			goto nla_put_failure;
 531
 532		hwhdrp = skb_mac_header(skb);
 533
 534		if (skb->dev->type == ARPHRD_SIT)
 535			hwhdrp -= ETH_HLEN;
 536
 537		if (hwhdrp >= skb->head &&
 538		    nla_put(inst->skb, NFULA_HWHEADER,
 539			    skb->dev->hard_header_len, hwhdrp))
 540			goto nla_put_failure;
 541	}
 542
 543	if (skb->tstamp) {
 544		struct nfulnl_msg_packet_timestamp ts;
 545		struct timespec64 kts = ktime_to_timespec64(skb->tstamp);
 546		ts.sec = cpu_to_be64(kts.tv_sec);
 547		ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
 548
 549		if (nla_put(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts))
 550			goto nla_put_failure;
 551	}
 552
 553	/* UID */
 554	sk = skb->sk;
 555	if (sk && sk_fullsock(sk)) {
 556		read_lock_bh(&sk->sk_callback_lock);
 557		if (sk->sk_socket && sk->sk_socket->file) {
 558			struct file *file = sk->sk_socket->file;
 559			const struct cred *cred = file->f_cred;
 560			struct user_namespace *user_ns = inst->peer_user_ns;
 561			__be32 uid = htonl(from_kuid_munged(user_ns, cred->fsuid));
 562			__be32 gid = htonl(from_kgid_munged(user_ns, cred->fsgid));
 563			read_unlock_bh(&sk->sk_callback_lock);
 564			if (nla_put_be32(inst->skb, NFULA_UID, uid) ||
 565			    nla_put_be32(inst->skb, NFULA_GID, gid))
 566				goto nla_put_failure;
 567		} else
 568			read_unlock_bh(&sk->sk_callback_lock);
 569	}
 570
 571	/* local sequence number */
 572	if ((inst->flags & NFULNL_CFG_F_SEQ) &&
 573	    nla_put_be32(inst->skb, NFULA_SEQ, htonl(inst->seq++)))
 574		goto nla_put_failure;
 575
 576	/* global sequence number */
 577	if ((inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) &&
 578	    nla_put_be32(inst->skb, NFULA_SEQ_GLOBAL,
 579			 htonl(atomic_inc_return(&log->global_seq))))
 580		goto nla_put_failure;
 581
 582	if (ct && nfnl_ct->build(inst->skb, ct, ctinfo,
 583				 NFULA_CT, NFULA_CT_INFO) < 0)
 584		goto nla_put_failure;
 585
 
 
 
 
 586	if (data_len) {
 587		struct nlattr *nla;
 588		int size = nla_attr_size(data_len);
 589
 590		if (skb_tailroom(inst->skb) < nla_total_size(data_len))
 591			goto nla_put_failure;
 592
 593		nla = skb_put(inst->skb, nla_total_size(data_len));
 594		nla->nla_type = NFULA_PAYLOAD;
 595		nla->nla_len = size;
 596
 597		if (skb_copy_bits(skb, 0, nla_data(nla), data_len))
 598			BUG();
 599	}
 600
 601	nlh->nlmsg_len = inst->skb->tail - old_tail;
 602	return 0;
 603
 604nla_put_failure:
 605	PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n");
 606	return -1;
 607}
 608
 609static const struct nf_loginfo default_loginfo = {
 610	.type =		NF_LOG_TYPE_ULOG,
 611	.u = {
 612		.ulog = {
 613			.copy_len	= 0xffff,
 614			.group		= 0,
 615			.qthreshold	= 1,
 616		},
 617	},
 618};
 619
 620/* log handler for internal netfilter logging api */
 621void
 622nfulnl_log_packet(struct net *net,
 623		  u_int8_t pf,
 624		  unsigned int hooknum,
 625		  const struct sk_buff *skb,
 626		  const struct net_device *in,
 627		  const struct net_device *out,
 628		  const struct nf_loginfo *li_user,
 629		  const char *prefix)
 630{
 631	size_t size;
 632	unsigned int data_len;
 633	struct nfulnl_instance *inst;
 634	const struct nf_loginfo *li;
 635	unsigned int qthreshold;
 636	unsigned int plen;
 637	struct nfnl_log_net *log = nfnl_log_pernet(net);
 638	const struct nfnl_ct_hook *nfnl_ct = NULL;
 639	struct nf_conn *ct = NULL;
 640	enum ip_conntrack_info uninitialized_var(ctinfo);
 641
 642	if (li_user && li_user->type == NF_LOG_TYPE_ULOG)
 643		li = li_user;
 644	else
 645		li = &default_loginfo;
 646
 647	inst = instance_lookup_get(log, li->u.ulog.group);
 648	if (!inst)
 649		return;
 650
 651	plen = 0;
 652	if (prefix)
 653		plen = strlen(prefix) + 1;
 654
 655	/* FIXME: do we want to make the size calculation conditional based on
 656	 * what is actually present?  way more branches and checks, but more
 657	 * memory efficient... */
 658	size =    nlmsg_total_size(sizeof(struct nfgenmsg))
 659		+ nla_total_size(sizeof(struct nfulnl_msg_packet_hdr))
 660		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
 661		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
 662#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 663		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
 664		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
 665#endif
 666		+ nla_total_size(sizeof(u_int32_t))	/* mark */
 667		+ nla_total_size(sizeof(u_int32_t))	/* uid */
 668		+ nla_total_size(sizeof(u_int32_t))	/* gid */
 669		+ nla_total_size(plen)			/* prefix */
 670		+ nla_total_size(sizeof(struct nfulnl_msg_packet_hw))
 671		+ nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp))
 672		+ nla_total_size(sizeof(struct nfgenmsg));	/* NLMSG_DONE */
 673
 674	if (in && skb_mac_header_was_set(skb)) {
 675		size +=   nla_total_size(skb->dev->hard_header_len)
 676			+ nla_total_size(sizeof(u_int16_t))	/* hwtype */
 677			+ nla_total_size(sizeof(u_int16_t));	/* hwlen */
 678	}
 679
 680	spin_lock_bh(&inst->lock);
 681
 682	if (inst->flags & NFULNL_CFG_F_SEQ)
 683		size += nla_total_size(sizeof(u_int32_t));
 684	if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
 685		size += nla_total_size(sizeof(u_int32_t));
 
 686	if (inst->flags & NFULNL_CFG_F_CONNTRACK) {
 687		nfnl_ct = rcu_dereference(nfnl_ct_hook);
 688		if (nfnl_ct != NULL) {
 689			ct = nfnl_ct->get_ct(skb, &ctinfo);
 690			if (ct != NULL)
 691				size += nfnl_ct->build_size(ct);
 692		}
 693	}
 
 
 
 694
 695	qthreshold = inst->qthreshold;
 696	/* per-rule qthreshold overrides per-instance */
 697	if (li->u.ulog.qthreshold)
 698		if (qthreshold > li->u.ulog.qthreshold)
 699			qthreshold = li->u.ulog.qthreshold;
 700
 701
 702	switch (inst->copy_mode) {
 703	case NFULNL_COPY_META:
 704	case NFULNL_COPY_NONE:
 705		data_len = 0;
 706		break;
 707
 708	case NFULNL_COPY_PACKET:
 709		data_len = inst->copy_range;
 710		if ((li->u.ulog.flags & NF_LOG_F_COPY_LEN) &&
 711		    (li->u.ulog.copy_len < data_len))
 712			data_len = li->u.ulog.copy_len;
 713
 714		if (data_len > skb->len)
 715			data_len = skb->len;
 716
 717		size += nla_total_size(data_len);
 718		break;
 719
 720	case NFULNL_COPY_DISABLED:
 721	default:
 722		goto unlock_and_release;
 723	}
 724
 725	if (inst->skb && size > skb_tailroom(inst->skb)) {
 726		/* either the queue len is too high or we don't have
 727		 * enough room in the skb left. flush to userspace. */
 728		__nfulnl_flush(inst);
 729	}
 730
 731	if (!inst->skb) {
 732		inst->skb = nfulnl_alloc_skb(net, inst->peer_portid,
 733					     inst->nlbufsiz, size);
 734		if (!inst->skb)
 735			goto alloc_failure;
 736	}
 737
 738	inst->qlen++;
 739
 740	__build_packet_message(log, inst, skb, data_len, pf,
 741				hooknum, in, out, prefix, plen,
 742				nfnl_ct, ct, ctinfo);
 743
 744	if (inst->qlen >= qthreshold)
 745		__nfulnl_flush(inst);
 746	/* timer_pending always called within inst->lock, so there
 747	 * is no chance of a race here */
 748	else if (!timer_pending(&inst->timer)) {
 749		instance_get(inst);
 750		inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100);
 751		add_timer(&inst->timer);
 752	}
 753
 754unlock_and_release:
 755	spin_unlock_bh(&inst->lock);
 756	instance_put(inst);
 757	return;
 758
 759alloc_failure:
 760	/* FIXME: statistics */
 761	goto unlock_and_release;
 762}
 763EXPORT_SYMBOL_GPL(nfulnl_log_packet);
 764
 765static int
 766nfulnl_rcv_nl_event(struct notifier_block *this,
 767		   unsigned long event, void *ptr)
 768{
 769	struct netlink_notify *n = ptr;
 770	struct nfnl_log_net *log = nfnl_log_pernet(n->net);
 771
 772	if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
 773		int i;
 774
 775		/* destroy all instances for this portid */
 776		spin_lock_bh(&log->instances_lock);
 777		for  (i = 0; i < INSTANCE_BUCKETS; i++) {
 778			struct hlist_node *t2;
 779			struct nfulnl_instance *inst;
 780			struct hlist_head *head = &log->instance_table[i];
 781
 782			hlist_for_each_entry_safe(inst, t2, head, hlist) {
 783				if (n->portid == inst->peer_portid)
 784					__instance_destroy(inst);
 785			}
 786		}
 787		spin_unlock_bh(&log->instances_lock);
 788	}
 789	return NOTIFY_DONE;
 790}
 791
 792static struct notifier_block nfulnl_rtnl_notifier = {
 793	.notifier_call	= nfulnl_rcv_nl_event,
 794};
 795
 796static int nfulnl_recv_unsupp(struct net *net, struct sock *ctnl,
 797			      struct sk_buff *skb, const struct nlmsghdr *nlh,
 798			      const struct nlattr * const nfqa[],
 799			      struct netlink_ext_ack *extack)
 800{
 801	return -ENOTSUPP;
 802}
 803
 804static struct nf_logger nfulnl_logger __read_mostly = {
 805	.name	= "nfnetlink_log",
 806	.type	= NF_LOG_TYPE_ULOG,
 807	.logfn	= nfulnl_log_packet,
 808	.me	= THIS_MODULE,
 809};
 810
 811static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = {
 812	[NFULA_CFG_CMD]		= { .len = sizeof(struct nfulnl_msg_config_cmd) },
 813	[NFULA_CFG_MODE]	= { .len = sizeof(struct nfulnl_msg_config_mode) },
 814	[NFULA_CFG_TIMEOUT]	= { .type = NLA_U32 },
 815	[NFULA_CFG_QTHRESH]	= { .type = NLA_U32 },
 816	[NFULA_CFG_NLBUFSIZ]	= { .type = NLA_U32 },
 817	[NFULA_CFG_FLAGS]	= { .type = NLA_U16 },
 818};
 819
 820static int nfulnl_recv_config(struct net *net, struct sock *ctnl,
 821			      struct sk_buff *skb, const struct nlmsghdr *nlh,
 822			      const struct nlattr * const nfula[],
 823			      struct netlink_ext_ack *extack)
 824{
 825	struct nfgenmsg *nfmsg = nlmsg_data(nlh);
 826	u_int16_t group_num = ntohs(nfmsg->res_id);
 
 827	struct nfulnl_instance *inst;
 828	struct nfulnl_msg_config_cmd *cmd = NULL;
 829	struct nfnl_log_net *log = nfnl_log_pernet(net);
 830	int ret = 0;
 831	u16 flags = 0;
 832
 833	if (nfula[NFULA_CFG_CMD]) {
 834		u_int8_t pf = nfmsg->nfgen_family;
 835		cmd = nla_data(nfula[NFULA_CFG_CMD]);
 836
 837		/* Commands without queue context */
 838		switch (cmd->command) {
 839		case NFULNL_CFG_CMD_PF_BIND:
 840			return nf_log_bind_pf(net, pf, &nfulnl_logger);
 841		case NFULNL_CFG_CMD_PF_UNBIND:
 842			nf_log_unbind_pf(net, pf);
 843			return 0;
 844		}
 845	}
 846
 847	inst = instance_lookup_get(log, group_num);
 848	if (inst && inst->peer_portid != NETLINK_CB(skb).portid) {
 849		ret = -EPERM;
 850		goto out_put;
 851	}
 852
 853	/* Check if we support these flags in first place, dependencies should
 854	 * be there too not to break atomicity.
 855	 */
 856	if (nfula[NFULA_CFG_FLAGS]) {
 857		flags = ntohs(nla_get_be16(nfula[NFULA_CFG_FLAGS]));
 858
 859		if ((flags & NFULNL_CFG_F_CONNTRACK) &&
 860		    !rcu_access_pointer(nfnl_ct_hook)) {
 861#ifdef CONFIG_MODULES
 862			nfnl_unlock(NFNL_SUBSYS_ULOG);
 863			request_module("ip_conntrack_netlink");
 864			nfnl_lock(NFNL_SUBSYS_ULOG);
 865			if (rcu_access_pointer(nfnl_ct_hook)) {
 866				ret = -EAGAIN;
 867				goto out_put;
 868			}
 869#endif
 870			ret = -EOPNOTSUPP;
 871			goto out_put;
 872		}
 873	}
 874
 875	if (cmd != NULL) {
 876		switch (cmd->command) {
 877		case NFULNL_CFG_CMD_BIND:
 878			if (inst) {
 879				ret = -EBUSY;
 880				goto out_put;
 881			}
 882
 883			inst = instance_create(net, group_num,
 884					       NETLINK_CB(skb).portid,
 885					       sk_user_ns(NETLINK_CB(skb).sk));
 886			if (IS_ERR(inst)) {
 887				ret = PTR_ERR(inst);
 888				goto out;
 889			}
 890			break;
 891		case NFULNL_CFG_CMD_UNBIND:
 892			if (!inst) {
 893				ret = -ENODEV;
 894				goto out;
 895			}
 896
 897			instance_destroy(log, inst);
 898			goto out_put;
 899		default:
 900			ret = -ENOTSUPP;
 901			goto out_put;
 902		}
 903	} else if (!inst) {
 904		ret = -ENODEV;
 905		goto out;
 906	}
 907
 908	if (nfula[NFULA_CFG_MODE]) {
 909		struct nfulnl_msg_config_mode *params =
 910			nla_data(nfula[NFULA_CFG_MODE]);
 911
 912		nfulnl_set_mode(inst, params->copy_mode,
 913				ntohl(params->copy_range));
 914	}
 915
 916	if (nfula[NFULA_CFG_TIMEOUT]) {
 917		__be32 timeout = nla_get_be32(nfula[NFULA_CFG_TIMEOUT]);
 918
 919		nfulnl_set_timeout(inst, ntohl(timeout));
 920	}
 921
 922	if (nfula[NFULA_CFG_NLBUFSIZ]) {
 923		__be32 nlbufsiz = nla_get_be32(nfula[NFULA_CFG_NLBUFSIZ]);
 924
 925		nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz));
 926	}
 927
 928	if (nfula[NFULA_CFG_QTHRESH]) {
 929		__be32 qthresh = nla_get_be32(nfula[NFULA_CFG_QTHRESH]);
 930
 931		nfulnl_set_qthresh(inst, ntohl(qthresh));
 932	}
 933
 934	if (nfula[NFULA_CFG_FLAGS])
 935		nfulnl_set_flags(inst, flags);
 936
 937out_put:
 938	instance_put(inst);
 939out:
 940	return ret;
 941}
 942
 943static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = {
 944	[NFULNL_MSG_PACKET]	= { .call = nfulnl_recv_unsupp,
 945				    .attr_count = NFULA_MAX, },
 946	[NFULNL_MSG_CONFIG]	= { .call = nfulnl_recv_config,
 947				    .attr_count = NFULA_CFG_MAX,
 948				    .policy = nfula_cfg_policy },
 
 
 
 
 
 
 949};
 950
 951static const struct nfnetlink_subsystem nfulnl_subsys = {
 952	.name		= "log",
 953	.subsys_id	= NFNL_SUBSYS_ULOG,
 954	.cb_count	= NFULNL_MSG_MAX,
 955	.cb		= nfulnl_cb,
 956};
 957
 958#ifdef CONFIG_PROC_FS
 959struct iter_state {
 960	struct seq_net_private p;
 961	unsigned int bucket;
 962};
 963
 964static struct hlist_node *get_first(struct net *net, struct iter_state *st)
 965{
 966	struct nfnl_log_net *log;
 967	if (!st)
 968		return NULL;
 969
 970	log = nfnl_log_pernet(net);
 971
 972	for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
 973		struct hlist_head *head = &log->instance_table[st->bucket];
 974
 975		if (!hlist_empty(head))
 976			return rcu_dereference_bh(hlist_first_rcu(head));
 977	}
 978	return NULL;
 979}
 980
 981static struct hlist_node *get_next(struct net *net, struct iter_state *st,
 982				   struct hlist_node *h)
 983{
 984	h = rcu_dereference_bh(hlist_next_rcu(h));
 985	while (!h) {
 986		struct nfnl_log_net *log;
 987		struct hlist_head *head;
 988
 989		if (++st->bucket >= INSTANCE_BUCKETS)
 990			return NULL;
 991
 992		log = nfnl_log_pernet(net);
 993		head = &log->instance_table[st->bucket];
 994		h = rcu_dereference_bh(hlist_first_rcu(head));
 995	}
 996	return h;
 997}
 998
 999static struct hlist_node *get_idx(struct net *net, struct iter_state *st,
1000				  loff_t pos)
1001{
1002	struct hlist_node *head;
1003	head = get_first(net, st);
1004
1005	if (head)
1006		while (pos && (head = get_next(net, st, head)))
1007			pos--;
1008	return pos ? NULL : head;
1009}
1010
1011static void *seq_start(struct seq_file *s, loff_t *pos)
1012	__acquires(rcu_bh)
1013{
1014	rcu_read_lock_bh();
1015	return get_idx(seq_file_net(s), s->private, *pos);
1016}
1017
1018static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1019{
1020	(*pos)++;
1021	return get_next(seq_file_net(s), s->private, v);
1022}
1023
1024static void seq_stop(struct seq_file *s, void *v)
1025	__releases(rcu_bh)
1026{
1027	rcu_read_unlock_bh();
1028}
1029
1030static int seq_show(struct seq_file *s, void *v)
1031{
1032	const struct nfulnl_instance *inst = v;
1033
1034	seq_printf(s, "%5u %6u %5u %1u %5u %6u %2u\n",
1035		   inst->group_num,
1036		   inst->peer_portid, inst->qlen,
1037		   inst->copy_mode, inst->copy_range,
1038		   inst->flushtimeout, refcount_read(&inst->use));
1039
1040	return 0;
1041}
1042
1043static const struct seq_operations nful_seq_ops = {
1044	.start	= seq_start,
1045	.next	= seq_next,
1046	.stop	= seq_stop,
1047	.show	= seq_show,
1048};
1049
1050static int nful_open(struct inode *inode, struct file *file)
1051{
1052	return seq_open_net(inode, file, &nful_seq_ops,
1053			    sizeof(struct iter_state));
1054}
1055
1056static const struct file_operations nful_file_ops = {
1057	.open	 = nful_open,
1058	.read	 = seq_read,
1059	.llseek	 = seq_lseek,
1060	.release = seq_release_net,
1061};
1062
1063#endif /* PROC_FS */
1064
1065static int __net_init nfnl_log_net_init(struct net *net)
1066{
1067	unsigned int i;
1068	struct nfnl_log_net *log = nfnl_log_pernet(net);
1069#ifdef CONFIG_PROC_FS
1070	struct proc_dir_entry *proc;
1071	kuid_t root_uid;
1072	kgid_t root_gid;
1073#endif
1074
1075	for (i = 0; i < INSTANCE_BUCKETS; i++)
1076		INIT_HLIST_HEAD(&log->instance_table[i]);
1077	spin_lock_init(&log->instances_lock);
1078
1079#ifdef CONFIG_PROC_FS
1080	proc = proc_create("nfnetlink_log", 0440,
1081			   net->nf.proc_netfilter, &nful_file_ops);
1082	if (!proc)
1083		return -ENOMEM;
1084
1085	root_uid = make_kuid(net->user_ns, 0);
1086	root_gid = make_kgid(net->user_ns, 0);
1087	if (uid_valid(root_uid) && gid_valid(root_gid))
1088		proc_set_user(proc, root_uid, root_gid);
1089#endif
1090	return 0;
1091}
1092
1093static void __net_exit nfnl_log_net_exit(struct net *net)
1094{
1095	struct nfnl_log_net *log = nfnl_log_pernet(net);
1096	unsigned int i;
1097
1098#ifdef CONFIG_PROC_FS
1099	remove_proc_entry("nfnetlink_log", net->nf.proc_netfilter);
1100#endif
1101	nf_log_unset(net, &nfulnl_logger);
1102	for (i = 0; i < INSTANCE_BUCKETS; i++)
1103		WARN_ON_ONCE(!hlist_empty(&log->instance_table[i]));
1104}
1105
1106static struct pernet_operations nfnl_log_net_ops = {
1107	.init	= nfnl_log_net_init,
1108	.exit	= nfnl_log_net_exit,
1109	.id	= &nfnl_log_net_id,
1110	.size	= sizeof(struct nfnl_log_net),
1111};
1112
1113static int __init nfnetlink_log_init(void)
1114{
1115	int status;
1116
1117	status = register_pernet_subsys(&nfnl_log_net_ops);
1118	if (status < 0) {
1119		pr_err("failed to register pernet ops\n");
1120		goto out;
1121	}
1122
1123	netlink_register_notifier(&nfulnl_rtnl_notifier);
1124	status = nfnetlink_subsys_register(&nfulnl_subsys);
1125	if (status < 0) {
1126		pr_err("failed to create netlink socket\n");
1127		goto cleanup_netlink_notifier;
1128	}
1129
1130	status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger);
1131	if (status < 0) {
1132		pr_err("failed to register logger\n");
1133		goto cleanup_subsys;
1134	}
1135
1136	return status;
1137
1138cleanup_subsys:
1139	nfnetlink_subsys_unregister(&nfulnl_subsys);
1140cleanup_netlink_notifier:
1141	netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1142	unregister_pernet_subsys(&nfnl_log_net_ops);
1143out:
1144	return status;
1145}
1146
1147static void __exit nfnetlink_log_fini(void)
1148{
1149	nfnetlink_subsys_unregister(&nfulnl_subsys);
1150	netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1151	unregister_pernet_subsys(&nfnl_log_net_ops);
1152	nf_log_unregister(&nfulnl_logger);
1153}
1154
1155MODULE_DESCRIPTION("netfilter userspace logging");
1156MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1157MODULE_LICENSE("GPL");
1158MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG);
1159MODULE_ALIAS_NF_LOGGER(AF_INET, 1);
1160MODULE_ALIAS_NF_LOGGER(AF_INET6, 1);
1161MODULE_ALIAS_NF_LOGGER(AF_BRIDGE, 1);
1162MODULE_ALIAS_NF_LOGGER(3, 1); /* NFPROTO_ARP */
1163MODULE_ALIAS_NF_LOGGER(5, 1); /* NFPROTO_NETDEV */
1164
1165module_init(nfnetlink_log_init);
1166module_exit(nfnetlink_log_fini);