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/spinlock.h>
31#include <linux/sysctl.h>
32#include <linux/proc_fs.h>
33#include <linux/security.h>
34#include <linux/list.h>
35#include <linux/slab.h>
36#include <net/sock.h>
37#include <net/netfilter/nf_log.h>
38#include <net/netns/generic.h>
39#include <net/netfilter/nfnetlink_log.h>
40
41#include <linux/atomic.h>
42
43#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
44#include "../bridge/br_private.h"
45#endif
46
47#define NFULNL_NLBUFSIZ_DEFAULT	NLMSG_GOODSIZE
48#define NFULNL_TIMEOUT_DEFAULT 	100	/* every second */
49#define NFULNL_QTHRESH_DEFAULT 	100	/* 100 packets */
50/* max packet size is limited by 16-bit struct nfattr nfa_len field */
51#define NFULNL_COPY_RANGE_MAX	(0xFFFF - NLA_HDRLEN)
52
53#define PRINTR(x, args...)	do { if (net_ratelimit()) \
54				     printk(x, ## args); } while (0);
55
56struct nfulnl_instance {
57	struct hlist_node hlist;	/* global list of instances */
58	spinlock_t lock;
59	atomic_t use;			/* use count */
60
61	unsigned int qlen;		/* number of nlmsgs in skb */
62	struct sk_buff *skb;		/* pre-allocatd skb */
63	struct timer_list timer;
64	struct net *net;
65	struct user_namespace *peer_user_ns;	/* User namespace of the peer process */
66	u32 peer_portid;		/* PORTID of the peer process */
67
68	/* configurable parameters */
69	unsigned int flushtimeout;	/* timeout until queue flush */
70	unsigned int nlbufsiz;		/* netlink buffer allocation size */
71	unsigned int qthreshold;	/* threshold of the queue */
72	u_int32_t copy_range;
73	u_int32_t seq;			/* instance-local sequential counter */
74	u_int16_t group_num;		/* number of this queue */
75	u_int16_t flags;
76	u_int8_t copy_mode;
77	struct rcu_head rcu;
78};
79
80#define INSTANCE_BUCKETS	16
81
82static int nfnl_log_net_id __read_mostly;
83
84struct nfnl_log_net {
85	spinlock_t instances_lock;
86	struct hlist_head instance_table[INSTANCE_BUCKETS];
87	atomic_t global_seq;
88};
89
90static struct nfnl_log_net *nfnl_log_pernet(struct net *net)
91{
92	return net_generic(net, nfnl_log_net_id);
93}
94
95static inline u_int8_t instance_hashfn(u_int16_t group_num)
96{
97	return ((group_num & 0xff) % INSTANCE_BUCKETS);
98}
99
100static struct nfulnl_instance *
101__instance_lookup(struct nfnl_log_net *log, u_int16_t group_num)
102{
103	struct hlist_head *head;
104	struct nfulnl_instance *inst;
105
106	head = &log->instance_table[instance_hashfn(group_num)];
107	hlist_for_each_entry_rcu(inst, head, hlist) {
108		if (inst->group_num == group_num)
109			return inst;
110	}
111	return NULL;
112}
113
114static inline void
115instance_get(struct nfulnl_instance *inst)
116{
117	atomic_inc(&inst->use);
118}
119
120static struct nfulnl_instance *
121instance_lookup_get(struct nfnl_log_net *log, u_int16_t group_num)
122{
123	struct nfulnl_instance *inst;
124
125	rcu_read_lock_bh();
126	inst = __instance_lookup(log, group_num);
127	if (inst && !atomic_inc_not_zero(&inst->use))
128		inst = NULL;
129	rcu_read_unlock_bh();
130
131	return inst;
132}
133
134static void nfulnl_instance_free_rcu(struct rcu_head *head)
135{
136	struct nfulnl_instance *inst =
137		container_of(head, struct nfulnl_instance, rcu);
138
139	put_net(inst->net);
140	kfree(inst);
141	module_put(THIS_MODULE);
142}
143
144static void
145instance_put(struct nfulnl_instance *inst)
146{
147	if (inst && atomic_dec_and_test(&inst->use))
148		call_rcu_bh(&inst->rcu, nfulnl_instance_free_rcu);
149}
150
151static void nfulnl_timer(unsigned long data);
152
153static struct nfulnl_instance *
154instance_create(struct net *net, u_int16_t group_num,
155		u32 portid, struct user_namespace *user_ns)
156{
157	struct nfulnl_instance *inst;
158	struct nfnl_log_net *log = nfnl_log_pernet(net);
159	int err;
160
161	spin_lock_bh(&log->instances_lock);
162	if (__instance_lookup(log, group_num)) {
163		err = -EEXIST;
164		goto out_unlock;
165	}
166
167	inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
168	if (!inst) {
169		err = -ENOMEM;
170		goto out_unlock;
171	}
172
173	if (!try_module_get(THIS_MODULE)) {
174		kfree(inst);
175		err = -EAGAIN;
176		goto out_unlock;
177	}
178
179	INIT_HLIST_NODE(&inst->hlist);
180	spin_lock_init(&inst->lock);
181	/* needs to be two, since we _put() after creation */
182	atomic_set(&inst->use, 2);
183
184	setup_timer(&inst->timer, nfulnl_timer, (unsigned long)inst);
185
186	inst->net = get_net(net);
187	inst->peer_user_ns = user_ns;
188	inst->peer_portid = portid;
189	inst->group_num = group_num;
190
191	inst->qthreshold 	= NFULNL_QTHRESH_DEFAULT;
192	inst->flushtimeout 	= NFULNL_TIMEOUT_DEFAULT;
193	inst->nlbufsiz 		= NFULNL_NLBUFSIZ_DEFAULT;
194	inst->copy_mode 	= NFULNL_COPY_PACKET;
195	inst->copy_range 	= NFULNL_COPY_RANGE_MAX;
196
197	hlist_add_head_rcu(&inst->hlist,
198		       &log->instance_table[instance_hashfn(group_num)]);
199
200
201	spin_unlock_bh(&log->instances_lock);
202
203	return inst;
204
205out_unlock:
206	spin_unlock_bh(&log->instances_lock);
207	return ERR_PTR(err);
208}
209
210static void __nfulnl_flush(struct nfulnl_instance *inst);
211
212/* called with BH disabled */
213static void
214__instance_destroy(struct nfulnl_instance *inst)
215{
216	/* first pull it out of the global list */
217	hlist_del_rcu(&inst->hlist);
218
219	/* then flush all pending packets from skb */
220
221	spin_lock(&inst->lock);
222
223	/* lockless readers wont be able to use us */
224	inst->copy_mode = NFULNL_COPY_DISABLED;
225
226	if (inst->skb)
227		__nfulnl_flush(inst);
228	spin_unlock(&inst->lock);
229
230	/* and finally put the refcount */
231	instance_put(inst);
232}
233
234static inline void
235instance_destroy(struct nfnl_log_net *log,
236		 struct nfulnl_instance *inst)
237{
238	spin_lock_bh(&log->instances_lock);
239	__instance_destroy(inst);
240	spin_unlock_bh(&log->instances_lock);
241}
242
243static int
244nfulnl_set_mode(struct nfulnl_instance *inst, u_int8_t mode,
245		  unsigned int range)
246{
247	int status = 0;
248
249	spin_lock_bh(&inst->lock);
250
251	switch (mode) {
252	case NFULNL_COPY_NONE:
253	case NFULNL_COPY_META:
254		inst->copy_mode = mode;
255		inst->copy_range = 0;
256		break;
257
258	case NFULNL_COPY_PACKET:
259		inst->copy_mode = mode;
260		if (range == 0)
261			range = NFULNL_COPY_RANGE_MAX;
262		inst->copy_range = min_t(unsigned int,
263					 range, NFULNL_COPY_RANGE_MAX);
264		break;
265
266	default:
267		status = -EINVAL;
268		break;
269	}
270
271	spin_unlock_bh(&inst->lock);
272
273	return status;
274}
275
276static int
277nfulnl_set_nlbufsiz(struct nfulnl_instance *inst, u_int32_t nlbufsiz)
278{
279	int status;
280
281	spin_lock_bh(&inst->lock);
282	if (nlbufsiz < NFULNL_NLBUFSIZ_DEFAULT)
283		status = -ERANGE;
284	else if (nlbufsiz > 131072)
285		status = -ERANGE;
286	else {
287		inst->nlbufsiz = nlbufsiz;
288		status = 0;
289	}
290	spin_unlock_bh(&inst->lock);
291
292	return status;
293}
294
295static int
296nfulnl_set_timeout(struct nfulnl_instance *inst, u_int32_t timeout)
297{
298	spin_lock_bh(&inst->lock);
299	inst->flushtimeout = timeout;
300	spin_unlock_bh(&inst->lock);
301
302	return 0;
303}
304
305static int
306nfulnl_set_qthresh(struct nfulnl_instance *inst, u_int32_t qthresh)
307{
308	spin_lock_bh(&inst->lock);
309	inst->qthreshold = qthresh;
310	spin_unlock_bh(&inst->lock);
311
312	return 0;
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 = nfnetlink_alloc_skb(net, n, peer_portid, GFP_ATOMIC);
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 = nfnetlink_alloc_skb(net, pkt_size,
343						  peer_portid, GFP_ATOMIC);
344		}
345	}
346
347	return skb;
348}
349
350static void
351__nfulnl_send(struct nfulnl_instance *inst)
352{
353	if (inst->qlen > 1) {
354		struct nlmsghdr *nlh = nlmsg_put(inst->skb, 0, 0,
355						 NLMSG_DONE,
356						 sizeof(struct nfgenmsg),
357						 0);
358		if (WARN_ONCE(!nlh, "bad nlskb size: %u, tailroom %d\n",
359			      inst->skb->len, skb_tailroom(inst->skb))) {
360			kfree_skb(inst->skb);
361			goto out;
362		}
363	}
364	nfnetlink_unicast(inst->skb, inst->net, inst->peer_portid,
365			  MSG_DONTWAIT);
366out:
367	inst->qlen = 0;
368	inst->skb = NULL;
369}
370
371static void
372__nfulnl_flush(struct nfulnl_instance *inst)
373{
374	/* timer holds a reference */
375	if (del_timer(&inst->timer))
376		instance_put(inst);
377	if (inst->skb)
378		__nfulnl_send(inst);
379}
380
381static void
382nfulnl_timer(unsigned long data)
383{
384	struct nfulnl_instance *inst = (struct nfulnl_instance *)data;
385
386	spin_lock_bh(&inst->lock);
387	if (inst->skb)
388		__nfulnl_send(inst);
389	spin_unlock_bh(&inst->lock);
390	instance_put(inst);
391}
392
393/* This is an inline function, we don't really care about a long
394 * list of arguments */
395static inline int
396__build_packet_message(struct nfnl_log_net *log,
397			struct nfulnl_instance *inst,
398			const struct sk_buff *skb,
399			unsigned int data_len,
400			u_int8_t pf,
401			unsigned int hooknum,
402			const struct net_device *indev,
403			const struct net_device *outdev,
404			const char *prefix, unsigned int plen)
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_SUBSYS_ULOG << 8 | 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_slow or nf_log_packet */
448			    nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
449					 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
450				goto nla_put_failure;
451		} else {
452			struct net_device *physindev;
453
454			/* Case 2: indev is bridge group, we need to look for
455			 * physical device (when called from ipv4) */
456			if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
457					 htonl(indev->ifindex)))
458				goto nla_put_failure;
459
460			physindev = nf_bridge_get_physindev(skb);
461			if (physindev &&
462			    nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
463					 htonl(physindev->ifindex)))
464				goto nla_put_failure;
465		}
466#endif
467	}
468
469	if (outdev) {
470#if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
471		if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
472				 htonl(outdev->ifindex)))
473			goto nla_put_failure;
474#else
475		if (pf == PF_BRIDGE) {
476			/* Case 1: outdev is physical output device, we need to
477			 * look for bridge group (when called from
478			 * netfilter_bridge) */
479			if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
480					 htonl(outdev->ifindex)) ||
481			/* this is the bridge group "brX" */
482			/* rcu_read_lock()ed by nf_hook_slow or nf_log_packet */
483			    nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
484					 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
485				goto nla_put_failure;
486		} else {
487			struct net_device *physoutdev;
488
489			/* Case 2: indev is a bridge group, we need to look
490			 * for physical device (when called from ipv4) */
491			if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
492					 htonl(outdev->ifindex)))
493				goto nla_put_failure;
494
495			physoutdev = nf_bridge_get_physoutdev(skb);
496			if (physoutdev &&
497			    nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
498					 htonl(physoutdev->ifindex)))
499				goto nla_put_failure;
500		}
501#endif
502	}
503
504	if (skb->mark &&
505	    nla_put_be32(inst->skb, NFULA_MARK, htonl(skb->mark)))
506		goto nla_put_failure;
507
508	if (indev && skb->dev &&
509	    skb->mac_header != skb->network_header) {
510		struct nfulnl_msg_packet_hw phw;
511		int len;
512
513		memset(&phw, 0, sizeof(phw));
514		len = dev_parse_header(skb, phw.hw_addr);
515		if (len > 0) {
516			phw.hw_addrlen = htons(len);
517			if (nla_put(inst->skb, NFULA_HWADDR, sizeof(phw), &phw))
518				goto nla_put_failure;
519		}
520	}
521
522	if (indev && skb_mac_header_was_set(skb)) {
523		if (nla_put_be16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type)) ||
524		    nla_put_be16(inst->skb, NFULA_HWLEN,
525				 htons(skb->dev->hard_header_len)))
526			goto nla_put_failure;
527
528		hwhdrp = skb_mac_header(skb);
529
530		if (skb->dev->type == ARPHRD_SIT)
531			hwhdrp -= ETH_HLEN;
532
533		if (hwhdrp >= skb->head &&
534		    nla_put(inst->skb, NFULA_HWHEADER,
535			    skb->dev->hard_header_len, hwhdrp))
536			goto nla_put_failure;
537	}
538
539	if (skb->tstamp.tv64) {
540		struct nfulnl_msg_packet_timestamp ts;
541		struct timeval tv = ktime_to_timeval(skb->tstamp);
542		ts.sec = cpu_to_be64(tv.tv_sec);
543		ts.usec = cpu_to_be64(tv.tv_usec);
544
545		if (nla_put(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts))
546			goto nla_put_failure;
547	}
548
549	/* UID */
550	sk = skb->sk;
551	if (sk && sk_fullsock(sk)) {
552		read_lock_bh(&sk->sk_callback_lock);
553		if (sk->sk_socket && sk->sk_socket->file) {
554			struct file *file = sk->sk_socket->file;
555			const struct cred *cred = file->f_cred;
556			struct user_namespace *user_ns = inst->peer_user_ns;
557			__be32 uid = htonl(from_kuid_munged(user_ns, cred->fsuid));
558			__be32 gid = htonl(from_kgid_munged(user_ns, cred->fsgid));
559			read_unlock_bh(&sk->sk_callback_lock);
560			if (nla_put_be32(inst->skb, NFULA_UID, uid) ||
561			    nla_put_be32(inst->skb, NFULA_GID, gid))
562				goto nla_put_failure;
563		} else
564			read_unlock_bh(&sk->sk_callback_lock);
565	}
566
567	/* local sequence number */
568	if ((inst->flags & NFULNL_CFG_F_SEQ) &&
569	    nla_put_be32(inst->skb, NFULA_SEQ, htonl(inst->seq++)))
570		goto nla_put_failure;
571
572	/* global sequence number */
573	if ((inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) &&
574	    nla_put_be32(inst->skb, NFULA_SEQ_GLOBAL,
575			 htonl(atomic_inc_return(&log->global_seq))))
576		goto nla_put_failure;
577
578	if (data_len) {
579		struct nlattr *nla;
580		int size = nla_attr_size(data_len);
581
582		if (skb_tailroom(inst->skb) < nla_total_size(data_len))
583			goto nla_put_failure;
584
585		nla = (struct nlattr *)skb_put(inst->skb, nla_total_size(data_len));
586		nla->nla_type = NFULA_PAYLOAD;
587		nla->nla_len = size;
588
589		if (skb_copy_bits(skb, 0, nla_data(nla), data_len))
590			BUG();
591	}
592
593	nlh->nlmsg_len = inst->skb->tail - old_tail;
594	return 0;
595
596nla_put_failure:
597	PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n");
598	return -1;
599}
600
601#define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
602
603static struct nf_loginfo default_loginfo = {
604	.type =		NF_LOG_TYPE_ULOG,
605	.u = {
606		.ulog = {
607			.copy_len	= 0xffff,
608			.group		= 0,
609			.qthreshold	= 1,
610		},
611	},
612};
613
614/* log handler for internal netfilter logging api */
615void
616nfulnl_log_packet(struct net *net,
617		  u_int8_t pf,
618		  unsigned int hooknum,
619		  const struct sk_buff *skb,
620		  const struct net_device *in,
621		  const struct net_device *out,
622		  const struct nf_loginfo *li_user,
623		  const char *prefix)
624{
625	unsigned int size, data_len;
626	struct nfulnl_instance *inst;
627	const struct nf_loginfo *li;
628	unsigned int qthreshold;
629	unsigned int plen;
630	struct nfnl_log_net *log = nfnl_log_pernet(net);
631
632	if (li_user && li_user->type == NF_LOG_TYPE_ULOG)
633		li = li_user;
634	else
635		li = &default_loginfo;
636
637	inst = instance_lookup_get(log, li->u.ulog.group);
638	if (!inst)
639		return;
640
641	plen = 0;
642	if (prefix)
643		plen = strlen(prefix) + 1;
644
645	/* FIXME: do we want to make the size calculation conditional based on
646	 * what is actually present?  way more branches and checks, but more
647	 * memory efficient... */
648	size =    nlmsg_total_size(sizeof(struct nfgenmsg))
649		+ nla_total_size(sizeof(struct nfulnl_msg_packet_hdr))
650		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
651		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
652#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
653		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
654		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
655#endif
656		+ nla_total_size(sizeof(u_int32_t))	/* mark */
657		+ nla_total_size(sizeof(u_int32_t))	/* uid */
658		+ nla_total_size(sizeof(u_int32_t))	/* gid */
659		+ nla_total_size(plen)			/* prefix */
660		+ nla_total_size(sizeof(struct nfulnl_msg_packet_hw))
661		+ nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp))
662		+ nla_total_size(sizeof(struct nfgenmsg));	/* NLMSG_DONE */
663
664	if (in && skb_mac_header_was_set(skb)) {
665		size +=   nla_total_size(skb->dev->hard_header_len)
666			+ nla_total_size(sizeof(u_int16_t))	/* hwtype */
667			+ nla_total_size(sizeof(u_int16_t));	/* hwlen */
668	}
669
670	spin_lock_bh(&inst->lock);
671
672	if (inst->flags & NFULNL_CFG_F_SEQ)
673		size += nla_total_size(sizeof(u_int32_t));
674	if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
675		size += nla_total_size(sizeof(u_int32_t));
676
677	qthreshold = inst->qthreshold;
678	/* per-rule qthreshold overrides per-instance */
679	if (li->u.ulog.qthreshold)
680		if (qthreshold > li->u.ulog.qthreshold)
681			qthreshold = li->u.ulog.qthreshold;
682
683
684	switch (inst->copy_mode) {
685	case NFULNL_COPY_META:
686	case NFULNL_COPY_NONE:
687		data_len = 0;
688		break;
689
690	case NFULNL_COPY_PACKET:
691		if (inst->copy_range > skb->len)
692			data_len = skb->len;
693		else
694			data_len = inst->copy_range;
695
696		size += nla_total_size(data_len);
697		break;
698
699	case NFULNL_COPY_DISABLED:
700	default:
701		goto unlock_and_release;
702	}
703
704	if (inst->skb && size > skb_tailroom(inst->skb)) {
705		/* either the queue len is too high or we don't have
706		 * enough room in the skb left. flush to userspace. */
707		__nfulnl_flush(inst);
708	}
709
710	if (!inst->skb) {
711		inst->skb = nfulnl_alloc_skb(net, inst->peer_portid,
712					     inst->nlbufsiz, size);
713		if (!inst->skb)
714			goto alloc_failure;
715	}
716
717	inst->qlen++;
718
719	__build_packet_message(log, inst, skb, data_len, pf,
720				hooknum, in, out, prefix, plen);
721
722	if (inst->qlen >= qthreshold)
723		__nfulnl_flush(inst);
724	/* timer_pending always called within inst->lock, so there
725	 * is no chance of a race here */
726	else if (!timer_pending(&inst->timer)) {
727		instance_get(inst);
728		inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100);
729		add_timer(&inst->timer);
730	}
731
732unlock_and_release:
733	spin_unlock_bh(&inst->lock);
734	instance_put(inst);
735	return;
736
737alloc_failure:
738	/* FIXME: statistics */
739	goto unlock_and_release;
740}
741EXPORT_SYMBOL_GPL(nfulnl_log_packet);
742
743static int
744nfulnl_rcv_nl_event(struct notifier_block *this,
745		   unsigned long event, void *ptr)
746{
747	struct netlink_notify *n = ptr;
748	struct nfnl_log_net *log = nfnl_log_pernet(n->net);
749
750	if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
751		int i;
752
753		/* destroy all instances for this portid */
754		spin_lock_bh(&log->instances_lock);
755		for  (i = 0; i < INSTANCE_BUCKETS; i++) {
756			struct hlist_node *t2;
757			struct nfulnl_instance *inst;
758			struct hlist_head *head = &log->instance_table[i];
759
760			hlist_for_each_entry_safe(inst, t2, head, hlist) {
761				if (n->portid == inst->peer_portid)
762					__instance_destroy(inst);
763			}
764		}
765		spin_unlock_bh(&log->instances_lock);
766	}
767	return NOTIFY_DONE;
768}
769
770static struct notifier_block nfulnl_rtnl_notifier = {
771	.notifier_call	= nfulnl_rcv_nl_event,
772};
773
774static int
775nfulnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
776		   const struct nlmsghdr *nlh,
777		   const struct nlattr * const nfqa[])
778{
779	return -ENOTSUPP;
780}
781
782static struct nf_logger nfulnl_logger __read_mostly = {
783	.name	= "nfnetlink_log",
784	.type	= NF_LOG_TYPE_ULOG,
785	.logfn	= &nfulnl_log_packet,
786	.me	= THIS_MODULE,
787};
788
789static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = {
790	[NFULA_CFG_CMD]		= { .len = sizeof(struct nfulnl_msg_config_cmd) },
791	[NFULA_CFG_MODE]	= { .len = sizeof(struct nfulnl_msg_config_mode) },
792	[NFULA_CFG_TIMEOUT]	= { .type = NLA_U32 },
793	[NFULA_CFG_QTHRESH]	= { .type = NLA_U32 },
794	[NFULA_CFG_NLBUFSIZ]	= { .type = NLA_U32 },
795	[NFULA_CFG_FLAGS]	= { .type = NLA_U16 },
796};
797
798static int
799nfulnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
800		   const struct nlmsghdr *nlh,
801		   const struct nlattr * const nfula[])
802{
803	struct nfgenmsg *nfmsg = nlmsg_data(nlh);
804	u_int16_t group_num = ntohs(nfmsg->res_id);
805	struct nfulnl_instance *inst;
806	struct nfulnl_msg_config_cmd *cmd = NULL;
807	struct net *net = sock_net(ctnl);
808	struct nfnl_log_net *log = nfnl_log_pernet(net);
809	int ret = 0;
810
811	if (nfula[NFULA_CFG_CMD]) {
812		u_int8_t pf = nfmsg->nfgen_family;
813		cmd = nla_data(nfula[NFULA_CFG_CMD]);
814
815		/* Commands without queue context */
816		switch (cmd->command) {
817		case NFULNL_CFG_CMD_PF_BIND:
818			return nf_log_bind_pf(net, pf, &nfulnl_logger);
819		case NFULNL_CFG_CMD_PF_UNBIND:
820			nf_log_unbind_pf(net, pf);
821			return 0;
822		}
823	}
824
825	inst = instance_lookup_get(log, group_num);
826	if (inst && inst->peer_portid != NETLINK_CB(skb).portid) {
827		ret = -EPERM;
828		goto out_put;
829	}
830
831	if (cmd != NULL) {
832		switch (cmd->command) {
833		case NFULNL_CFG_CMD_BIND:
834			if (inst) {
835				ret = -EBUSY;
836				goto out_put;
837			}
838
839			inst = instance_create(net, group_num,
840					       NETLINK_CB(skb).portid,
841					       sk_user_ns(NETLINK_CB(skb).sk));
842			if (IS_ERR(inst)) {
843				ret = PTR_ERR(inst);
844				goto out;
845			}
846			break;
847		case NFULNL_CFG_CMD_UNBIND:
848			if (!inst) {
849				ret = -ENODEV;
850				goto out;
851			}
852
853			instance_destroy(log, inst);
854			goto out_put;
855		default:
856			ret = -ENOTSUPP;
857			break;
858		}
859	}
860
861	if (nfula[NFULA_CFG_MODE]) {
862		struct nfulnl_msg_config_mode *params;
863		params = nla_data(nfula[NFULA_CFG_MODE]);
864
865		if (!inst) {
866			ret = -ENODEV;
867			goto out;
868		}
869		nfulnl_set_mode(inst, params->copy_mode,
870				ntohl(params->copy_range));
871	}
872
873	if (nfula[NFULA_CFG_TIMEOUT]) {
874		__be32 timeout = nla_get_be32(nfula[NFULA_CFG_TIMEOUT]);
875
876		if (!inst) {
877			ret = -ENODEV;
878			goto out;
879		}
880		nfulnl_set_timeout(inst, ntohl(timeout));
881	}
882
883	if (nfula[NFULA_CFG_NLBUFSIZ]) {
884		__be32 nlbufsiz = nla_get_be32(nfula[NFULA_CFG_NLBUFSIZ]);
885
886		if (!inst) {
887			ret = -ENODEV;
888			goto out;
889		}
890		nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz));
891	}
892
893	if (nfula[NFULA_CFG_QTHRESH]) {
894		__be32 qthresh = nla_get_be32(nfula[NFULA_CFG_QTHRESH]);
895
896		if (!inst) {
897			ret = -ENODEV;
898			goto out;
899		}
900		nfulnl_set_qthresh(inst, ntohl(qthresh));
901	}
902
903	if (nfula[NFULA_CFG_FLAGS]) {
904		__be16 flags = nla_get_be16(nfula[NFULA_CFG_FLAGS]);
905
906		if (!inst) {
907			ret = -ENODEV;
908			goto out;
909		}
910		nfulnl_set_flags(inst, ntohs(flags));
911	}
912
913out_put:
914	instance_put(inst);
915out:
916	return ret;
917}
918
919static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = {
920	[NFULNL_MSG_PACKET]	= { .call = nfulnl_recv_unsupp,
921				    .attr_count = NFULA_MAX, },
922	[NFULNL_MSG_CONFIG]	= { .call = nfulnl_recv_config,
923				    .attr_count = NFULA_CFG_MAX,
924				    .policy = nfula_cfg_policy },
925};
926
927static const struct nfnetlink_subsystem nfulnl_subsys = {
928	.name		= "log",
929	.subsys_id	= NFNL_SUBSYS_ULOG,
930	.cb_count	= NFULNL_MSG_MAX,
931	.cb		= nfulnl_cb,
932};
933
934#ifdef CONFIG_PROC_FS
935struct iter_state {
936	struct seq_net_private p;
937	unsigned int bucket;
938};
939
940static struct hlist_node *get_first(struct net *net, struct iter_state *st)
941{
942	struct nfnl_log_net *log;
943	if (!st)
944		return NULL;
945
946	log = nfnl_log_pernet(net);
947
948	for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
949		struct hlist_head *head = &log->instance_table[st->bucket];
950
951		if (!hlist_empty(head))
952			return rcu_dereference_bh(hlist_first_rcu(head));
953	}
954	return NULL;
955}
956
957static struct hlist_node *get_next(struct net *net, struct iter_state *st,
958				   struct hlist_node *h)
959{
960	h = rcu_dereference_bh(hlist_next_rcu(h));
961	while (!h) {
962		struct nfnl_log_net *log;
963		struct hlist_head *head;
964
965		if (++st->bucket >= INSTANCE_BUCKETS)
966			return NULL;
967
968		log = nfnl_log_pernet(net);
969		head = &log->instance_table[st->bucket];
970		h = rcu_dereference_bh(hlist_first_rcu(head));
971	}
972	return h;
973}
974
975static struct hlist_node *get_idx(struct net *net, struct iter_state *st,
976				  loff_t pos)
977{
978	struct hlist_node *head;
979	head = get_first(net, st);
980
981	if (head)
982		while (pos && (head = get_next(net, st, head)))
983			pos--;
984	return pos ? NULL : head;
985}
986
987static void *seq_start(struct seq_file *s, loff_t *pos)
988	__acquires(rcu_bh)
989{
990	rcu_read_lock_bh();
991	return get_idx(seq_file_net(s), s->private, *pos);
992}
993
994static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
995{
996	(*pos)++;
997	return get_next(seq_file_net(s), s->private, v);
998}
999
1000static void seq_stop(struct seq_file *s, void *v)
1001	__releases(rcu_bh)
1002{
1003	rcu_read_unlock_bh();
1004}
1005
1006static int seq_show(struct seq_file *s, void *v)
1007{
1008	const struct nfulnl_instance *inst = v;
1009
1010	seq_printf(s, "%5u %6u %5u %1u %5u %6u %2u\n",
1011		   inst->group_num,
1012		   inst->peer_portid, inst->qlen,
1013		   inst->copy_mode, inst->copy_range,
1014		   inst->flushtimeout, atomic_read(&inst->use));
1015
1016	return 0;
1017}
1018
1019static const struct seq_operations nful_seq_ops = {
1020	.start	= seq_start,
1021	.next	= seq_next,
1022	.stop	= seq_stop,
1023	.show	= seq_show,
1024};
1025
1026static int nful_open(struct inode *inode, struct file *file)
1027{
1028	return seq_open_net(inode, file, &nful_seq_ops,
1029			    sizeof(struct iter_state));
1030}
1031
1032static const struct file_operations nful_file_ops = {
1033	.owner	 = THIS_MODULE,
1034	.open	 = nful_open,
1035	.read	 = seq_read,
1036	.llseek	 = seq_lseek,
1037	.release = seq_release_net,
1038};
1039
1040#endif /* PROC_FS */
1041
1042static int __net_init nfnl_log_net_init(struct net *net)
1043{
1044	unsigned int i;
1045	struct nfnl_log_net *log = nfnl_log_pernet(net);
1046
1047	for (i = 0; i < INSTANCE_BUCKETS; i++)
1048		INIT_HLIST_HEAD(&log->instance_table[i]);
1049	spin_lock_init(&log->instances_lock);
1050
1051#ifdef CONFIG_PROC_FS
1052	if (!proc_create("nfnetlink_log", 0440,
1053			 net->nf.proc_netfilter, &nful_file_ops))
1054		return -ENOMEM;
1055#endif
1056	return 0;
1057}
1058
1059static void __net_exit nfnl_log_net_exit(struct net *net)
1060{
1061#ifdef CONFIG_PROC_FS
1062	remove_proc_entry("nfnetlink_log", net->nf.proc_netfilter);
1063#endif
1064	nf_log_unset(net, &nfulnl_logger);
1065}
1066
1067static struct pernet_operations nfnl_log_net_ops = {
1068	.init	= nfnl_log_net_init,
1069	.exit	= nfnl_log_net_exit,
1070	.id	= &nfnl_log_net_id,
1071	.size	= sizeof(struct nfnl_log_net),
1072};
1073
1074static int __init nfnetlink_log_init(void)
1075{
1076	int status;
1077
1078	status = register_pernet_subsys(&nfnl_log_net_ops);
1079	if (status < 0) {
1080		pr_err("failed to register pernet ops\n");
1081		goto out;
1082	}
1083
1084	netlink_register_notifier(&nfulnl_rtnl_notifier);
1085	status = nfnetlink_subsys_register(&nfulnl_subsys);
1086	if (status < 0) {
1087		pr_err("failed to create netlink socket\n");
1088		goto cleanup_netlink_notifier;
1089	}
1090
1091	status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger);
1092	if (status < 0) {
1093		pr_err("failed to register logger\n");
1094		goto cleanup_subsys;
1095	}
1096
1097	return status;
1098
1099cleanup_subsys:
1100	nfnetlink_subsys_unregister(&nfulnl_subsys);
1101cleanup_netlink_notifier:
1102	netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1103	unregister_pernet_subsys(&nfnl_log_net_ops);
1104out:
1105	return status;
1106}
1107
1108static void __exit nfnetlink_log_fini(void)
1109{
1110	nf_log_unregister(&nfulnl_logger);
1111	nfnetlink_subsys_unregister(&nfulnl_subsys);
1112	netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1113	unregister_pernet_subsys(&nfnl_log_net_ops);
1114}
1115
1116MODULE_DESCRIPTION("netfilter userspace logging");
1117MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1118MODULE_LICENSE("GPL");
1119MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG);
1120MODULE_ALIAS_NF_LOGGER(AF_INET, 1);
1121MODULE_ALIAS_NF_LOGGER(AF_INET6, 1);
1122MODULE_ALIAS_NF_LOGGER(AF_BRIDGE, 1);
1123
1124module_init(nfnetlink_log_init);
1125module_exit(nfnetlink_log_fini);
1126