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
44#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
45#include "../bridge/br_private.h"
46#endif
47
48#define NFULNL_NLBUFSIZ_DEFAULT	NLMSG_GOODSIZE
49#define NFULNL_TIMEOUT_DEFAULT 	100	/* every second */
50#define NFULNL_QTHRESH_DEFAULT 	100	/* 100 packets */
51/* max packet size is limited by 16-bit struct nfattr nfa_len field */
52#define NFULNL_COPY_RANGE_MAX	(0xFFFF - NLA_HDRLEN)
53
54#define PRINTR(x, args...)	do { if (net_ratelimit()) \
55				     printk(x, ## args); } while (0);
56
57struct nfulnl_instance {
58	struct hlist_node hlist;	/* global list of instances */
59	spinlock_t lock;
60	atomic_t use;			/* use count */
61
62	unsigned int qlen;		/* number of nlmsgs in skb */
63	struct sk_buff *skb;		/* pre-allocatd skb */
64	struct timer_list timer;
65	struct net *net;
66	struct user_namespace *peer_user_ns;	/* User namespace of the peer process */
67	u32 peer_portid;		/* PORTID of the peer process */
68
69	/* configurable parameters */
70	unsigned int flushtimeout;	/* timeout until queue flush */
71	unsigned int nlbufsiz;		/* netlink buffer allocation size */
72	unsigned int qthreshold;	/* threshold of the queue */
73	u_int32_t copy_range;
74	u_int32_t seq;			/* instance-local sequential counter */
75	u_int16_t group_num;		/* number of this queue */
76	u_int16_t flags;
77	u_int8_t copy_mode;
78	struct rcu_head rcu;
79};
80
81#define INSTANCE_BUCKETS	16
82
83static int nfnl_log_net_id __read_mostly;
84
85struct nfnl_log_net {
86	spinlock_t instances_lock;
87	struct hlist_head instance_table[INSTANCE_BUCKETS];
88	atomic_t global_seq;
89};
90
91static struct nfnl_log_net *nfnl_log_pernet(struct net *net)
92{
93	return net_generic(net, nfnl_log_net_id);
94}
95
96static inline u_int8_t instance_hashfn(u_int16_t group_num)
97{
98	return ((group_num & 0xff) % INSTANCE_BUCKETS);
99}
100
101static struct nfulnl_instance *
102__instance_lookup(struct nfnl_log_net *log, u_int16_t group_num)
103{
104	struct hlist_head *head;
105	struct nfulnl_instance *inst;
106
107	head = &log->instance_table[instance_hashfn(group_num)];
108	hlist_for_each_entry_rcu(inst, head, hlist) {
109		if (inst->group_num == group_num)
110			return inst;
111	}
112	return NULL;
113}
114
115static inline void
116instance_get(struct nfulnl_instance *inst)
117{
118	atomic_inc(&inst->use);
119}
120
121static struct nfulnl_instance *
122instance_lookup_get(struct nfnl_log_net *log, u_int16_t group_num)
123{
124	struct nfulnl_instance *inst;
125
126	rcu_read_lock_bh();
127	inst = __instance_lookup(log, group_num);
128	if (inst && !atomic_inc_not_zero(&inst->use))
129		inst = NULL;
130	rcu_read_unlock_bh();
131
132	return inst;
133}
134
135static void nfulnl_instance_free_rcu(struct rcu_head *head)
136{
137	struct nfulnl_instance *inst =
138		container_of(head, struct nfulnl_instance, rcu);
139
140	put_net(inst->net);
141	kfree(inst);
142	module_put(THIS_MODULE);
143}
144
145static void
146instance_put(struct nfulnl_instance *inst)
147{
148	if (inst && atomic_dec_and_test(&inst->use))
149		call_rcu_bh(&inst->rcu, nfulnl_instance_free_rcu);
150}
151
152static void nfulnl_timer(unsigned long data);
153
154static struct nfulnl_instance *
155instance_create(struct net *net, u_int16_t group_num,
156		u32 portid, struct user_namespace *user_ns)
157{
158	struct nfulnl_instance *inst;
159	struct nfnl_log_net *log = nfnl_log_pernet(net);
160	int err;
161
162	spin_lock_bh(&log->instances_lock);
163	if (__instance_lookup(log, group_num)) {
164		err = -EEXIST;
165		goto out_unlock;
166	}
167
168	inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
169	if (!inst) {
170		err = -ENOMEM;
171		goto out_unlock;
172	}
173
174	if (!try_module_get(THIS_MODULE)) {
175		kfree(inst);
176		err = -EAGAIN;
177		goto out_unlock;
178	}
179
180	INIT_HLIST_NODE(&inst->hlist);
181	spin_lock_init(&inst->lock);
182	/* needs to be two, since we _put() after creation */
183	atomic_set(&inst->use, 2);
184
185	setup_timer(&inst->timer, nfulnl_timer, (unsigned long)inst);
186
187	inst->net = get_net(net);
188	inst->peer_user_ns = user_ns;
189	inst->peer_portid = portid;
190	inst->group_num = group_num;
191
192	inst->qthreshold 	= NFULNL_QTHRESH_DEFAULT;
193	inst->flushtimeout 	= NFULNL_TIMEOUT_DEFAULT;
194	inst->nlbufsiz 		= NFULNL_NLBUFSIZ_DEFAULT;
195	inst->copy_mode 	= NFULNL_COPY_PACKET;
196	inst->copy_range 	= NFULNL_COPY_RANGE_MAX;
197
198	hlist_add_head_rcu(&inst->hlist,
199		       &log->instance_table[instance_hashfn(group_num)]);
200
201
202	spin_unlock_bh(&log->instances_lock);
203
204	return inst;
205
206out_unlock:
207	spin_unlock_bh(&log->instances_lock);
208	return ERR_PTR(err);
209}
210
211static void __nfulnl_flush(struct nfulnl_instance *inst);
212
213/* called with BH disabled */
214static void
215__instance_destroy(struct nfulnl_instance *inst)
216{
217	/* first pull it out of the global list */
218	hlist_del_rcu(&inst->hlist);
219
220	/* then flush all pending packets from skb */
221
222	spin_lock(&inst->lock);
223
224	/* lockless readers wont be able to use us */
225	inst->copy_mode = NFULNL_COPY_DISABLED;
226
227	if (inst->skb)
228		__nfulnl_flush(inst);
229	spin_unlock(&inst->lock);
230
231	/* and finally put the refcount */
232	instance_put(inst);
233}
234
235static inline void
236instance_destroy(struct nfnl_log_net *log,
237		 struct nfulnl_instance *inst)
238{
239	spin_lock_bh(&log->instances_lock);
240	__instance_destroy(inst);
241	spin_unlock_bh(&log->instances_lock);
242}
243
244static int
245nfulnl_set_mode(struct nfulnl_instance *inst, u_int8_t mode,
246		  unsigned int range)
247{
248	int status = 0;
249
250	spin_lock_bh(&inst->lock);
251
252	switch (mode) {
253	case NFULNL_COPY_NONE:
254	case NFULNL_COPY_META:
255		inst->copy_mode = mode;
256		inst->copy_range = 0;
257		break;
258
259	case NFULNL_COPY_PACKET:
260		inst->copy_mode = mode;
261		if (range == 0)
262			range = NFULNL_COPY_RANGE_MAX;
263		inst->copy_range = min_t(unsigned int,
264					 range, NFULNL_COPY_RANGE_MAX);
265		break;
266
267	default:
268		status = -EINVAL;
269		break;
270	}
271
272	spin_unlock_bh(&inst->lock);
273
274	return status;
275}
276
277static int
278nfulnl_set_nlbufsiz(struct nfulnl_instance *inst, u_int32_t nlbufsiz)
279{
280	int status;
281
282	spin_lock_bh(&inst->lock);
283	if (nlbufsiz < NFULNL_NLBUFSIZ_DEFAULT)
284		status = -ERANGE;
285	else if (nlbufsiz > 131072)
286		status = -ERANGE;
287	else {
288		inst->nlbufsiz = nlbufsiz;
289		status = 0;
290	}
291	spin_unlock_bh(&inst->lock);
292
293	return status;
294}
295
296static int
297nfulnl_set_timeout(struct nfulnl_instance *inst, u_int32_t timeout)
298{
299	spin_lock_bh(&inst->lock);
300	inst->flushtimeout = timeout;
301	spin_unlock_bh(&inst->lock);
302
303	return 0;
304}
305
306static int
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	return 0;
314}
315
316static int
317nfulnl_set_flags(struct nfulnl_instance *inst, u_int16_t flags)
318{
319	spin_lock_bh(&inst->lock);
320	inst->flags = flags;
321	spin_unlock_bh(&inst->lock);
322
323	return 0;
324}
325
326static struct sk_buff *
327nfulnl_alloc_skb(struct net *net, u32 peer_portid, unsigned int inst_size,
328		 unsigned int pkt_size)
329{
330	struct sk_buff *skb;
331	unsigned int n;
332
333	/* alloc skb which should be big enough for a whole multipart
334	 * message.  WARNING: has to be <= 128k due to slab restrictions */
335
336	n = max(inst_size, pkt_size);
337	skb = nfnetlink_alloc_skb(net, n, peer_portid, GFP_ATOMIC);
338	if (!skb) {
339		if (n > pkt_size) {
340			/* try to allocate only as much as we need for current
341			 * packet */
342
343			skb = nfnetlink_alloc_skb(net, pkt_size,
344						  peer_portid, GFP_ATOMIC);
345		}
346	}
347
348	return skb;
349}
350
351static void
352__nfulnl_send(struct nfulnl_instance *inst)
353{
354	if (inst->qlen > 1) {
355		struct nlmsghdr *nlh = nlmsg_put(inst->skb, 0, 0,
356						 NLMSG_DONE,
357						 sizeof(struct nfgenmsg),
358						 0);
359		if (WARN_ONCE(!nlh, "bad nlskb size: %u, tailroom %d\n",
360			      inst->skb->len, skb_tailroom(inst->skb))) {
361			kfree_skb(inst->skb);
362			goto out;
363		}
364	}
365	nfnetlink_unicast(inst->skb, inst->net, inst->peer_portid,
366			  MSG_DONTWAIT);
367out:
368	inst->qlen = 0;
369	inst->skb = NULL;
370}
371
372static void
373__nfulnl_flush(struct nfulnl_instance *inst)
374{
375	/* timer holds a reference */
376	if (del_timer(&inst->timer))
377		instance_put(inst);
378	if (inst->skb)
379		__nfulnl_send(inst);
380}
381
382static void
383nfulnl_timer(unsigned long data)
384{
385	struct nfulnl_instance *inst = (struct nfulnl_instance *)data;
386
387	spin_lock_bh(&inst->lock);
388	if (inst->skb)
389		__nfulnl_send(inst);
390	spin_unlock_bh(&inst->lock);
391	instance_put(inst);
392}
393
394/* This is an inline function, we don't really care about a long
395 * list of arguments */
396static inline int
397__build_packet_message(struct nfnl_log_net *log,
398			struct nfulnl_instance *inst,
399			const struct sk_buff *skb,
400			unsigned int data_len,
401			u_int8_t pf,
402			unsigned int hooknum,
403			const struct net_device *indev,
404			const struct net_device *outdev,
405			const char *prefix, unsigned int plen,
406			const struct nfnl_ct_hook *nfnl_ct,
407			struct nf_conn *ct, enum ip_conntrack_info ctinfo)
408{
409	struct nfulnl_msg_packet_hdr pmsg;
410	struct nlmsghdr *nlh;
411	struct nfgenmsg *nfmsg;
412	sk_buff_data_t old_tail = inst->skb->tail;
413	struct sock *sk;
414	const unsigned char *hwhdrp;
415
416	nlh = nlmsg_put(inst->skb, 0, 0,
417			NFNL_SUBSYS_ULOG << 8 | NFULNL_MSG_PACKET,
418			sizeof(struct nfgenmsg), 0);
419	if (!nlh)
420		return -1;
421	nfmsg = nlmsg_data(nlh);
422	nfmsg->nfgen_family = pf;
423	nfmsg->version = NFNETLINK_V0;
424	nfmsg->res_id = htons(inst->group_num);
425
426	memset(&pmsg, 0, sizeof(pmsg));
427	pmsg.hw_protocol	= skb->protocol;
428	pmsg.hook		= hooknum;
429
430	if (nla_put(inst->skb, NFULA_PACKET_HDR, sizeof(pmsg), &pmsg))
431		goto nla_put_failure;
432
433	if (prefix &&
434	    nla_put(inst->skb, NFULA_PREFIX, plen, prefix))
435		goto nla_put_failure;
436
437	if (indev) {
438#if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
439		if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
440				 htonl(indev->ifindex)))
441			goto nla_put_failure;
442#else
443		if (pf == PF_BRIDGE) {
444			/* Case 1: outdev is physical input device, we need to
445			 * look for bridge group (when called from
446			 * netfilter_bridge) */
447			if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
448					 htonl(indev->ifindex)) ||
449			/* this is the bridge group "brX" */
450			/* rcu_read_lock()ed by nf_hook_slow or nf_log_packet */
451			    nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
452					 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
453				goto nla_put_failure;
454		} else {
455			struct net_device *physindev;
456
457			/* Case 2: indev is bridge group, we need to look for
458			 * physical device (when called from ipv4) */
459			if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
460					 htonl(indev->ifindex)))
461				goto nla_put_failure;
462
463			physindev = nf_bridge_get_physindev(skb);
464			if (physindev &&
465			    nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
466					 htonl(physindev->ifindex)))
467				goto nla_put_failure;
468		}
469#endif
470	}
471
472	if (outdev) {
473#if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
474		if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
475				 htonl(outdev->ifindex)))
476			goto nla_put_failure;
477#else
478		if (pf == PF_BRIDGE) {
479			/* Case 1: outdev is physical output device, we need to
480			 * look for bridge group (when called from
481			 * netfilter_bridge) */
482			if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
483					 htonl(outdev->ifindex)) ||
484			/* this is the bridge group "brX" */
485			/* rcu_read_lock()ed by nf_hook_slow or nf_log_packet */
486			    nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
487					 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
488				goto nla_put_failure;
489		} else {
490			struct net_device *physoutdev;
491
492			/* Case 2: indev is a bridge group, we need to look
493			 * for physical device (when called from ipv4) */
494			if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
495					 htonl(outdev->ifindex)))
496				goto nla_put_failure;
497
498			physoutdev = nf_bridge_get_physoutdev(skb);
499			if (physoutdev &&
500			    nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
501					 htonl(physoutdev->ifindex)))
502				goto nla_put_failure;
503		}
504#endif
505	}
506
507	if (skb->mark &&
508	    nla_put_be32(inst->skb, NFULA_MARK, htonl(skb->mark)))
509		goto nla_put_failure;
510
511	if (indev && skb->dev &&
512	    skb->mac_header != skb->network_header) {
513		struct nfulnl_msg_packet_hw phw;
514		int len;
515
516		memset(&phw, 0, sizeof(phw));
517		len = dev_parse_header(skb, phw.hw_addr);
518		if (len > 0) {
519			phw.hw_addrlen = htons(len);
520			if (nla_put(inst->skb, NFULA_HWADDR, sizeof(phw), &phw))
521				goto nla_put_failure;
522		}
523	}
524
525	if (indev && skb_mac_header_was_set(skb)) {
526		if (nla_put_be16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type)) ||
527		    nla_put_be16(inst->skb, NFULA_HWLEN,
528				 htons(skb->dev->hard_header_len)))
529			goto nla_put_failure;
530
531		hwhdrp = skb_mac_header(skb);
532
533		if (skb->dev->type == ARPHRD_SIT)
534			hwhdrp -= ETH_HLEN;
535
536		if (hwhdrp >= skb->head &&
537		    nla_put(inst->skb, NFULA_HWHEADER,
538			    skb->dev->hard_header_len, hwhdrp))
539			goto nla_put_failure;
540	}
541
542	if (skb->tstamp.tv64) {
543		struct nfulnl_msg_packet_timestamp ts;
544		struct timespec64 kts = ktime_to_timespec64(skb->tstamp);
545		ts.sec = cpu_to_be64(kts.tv_sec);
546		ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
547
548		if (nla_put(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts))
549			goto nla_put_failure;
550	}
551
552	/* UID */
553	sk = skb->sk;
554	if (sk && sk_fullsock(sk)) {
555		read_lock_bh(&sk->sk_callback_lock);
556		if (sk->sk_socket && sk->sk_socket->file) {
557			struct file *file = sk->sk_socket->file;
558			const struct cred *cred = file->f_cred;
559			struct user_namespace *user_ns = inst->peer_user_ns;
560			__be32 uid = htonl(from_kuid_munged(user_ns, cred->fsuid));
561			__be32 gid = htonl(from_kgid_munged(user_ns, cred->fsgid));
562			read_unlock_bh(&sk->sk_callback_lock);
563			if (nla_put_be32(inst->skb, NFULA_UID, uid) ||
564			    nla_put_be32(inst->skb, NFULA_GID, gid))
565				goto nla_put_failure;
566		} else
567			read_unlock_bh(&sk->sk_callback_lock);
568	}
569
570	/* local sequence number */
571	if ((inst->flags & NFULNL_CFG_F_SEQ) &&
572	    nla_put_be32(inst->skb, NFULA_SEQ, htonl(inst->seq++)))
573		goto nla_put_failure;
574
575	/* global sequence number */
576	if ((inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) &&
577	    nla_put_be32(inst->skb, NFULA_SEQ_GLOBAL,
578			 htonl(atomic_inc_return(&log->global_seq))))
579		goto nla_put_failure;
580
581	if (ct && nfnl_ct->build(inst->skb, ct, ctinfo,
582				 NFULA_CT, NFULA_CT_INFO) < 0)
583		goto nla_put_failure;
584
585	if (data_len) {
586		struct nlattr *nla;
587		int size = nla_attr_size(data_len);
588
589		if (skb_tailroom(inst->skb) < nla_total_size(data_len))
590			goto nla_put_failure;
591
592		nla = (struct nlattr *)skb_put(inst->skb, nla_total_size(data_len));
593		nla->nla_type = NFULA_PAYLOAD;
594		nla->nla_len = size;
595
596		if (skb_copy_bits(skb, 0, nla_data(nla), data_len))
597			BUG();
598	}
599
600	nlh->nlmsg_len = inst->skb->tail - old_tail;
601	return 0;
602
603nla_put_failure:
604	PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n");
605	return -1;
606}
607
608static struct nf_loginfo default_loginfo = {
609	.type =		NF_LOG_TYPE_ULOG,
610	.u = {
611		.ulog = {
612			.copy_len	= 0xffff,
613			.group		= 0,
614			.qthreshold	= 1,
615		},
616	},
617};
618
619/* log handler for internal netfilter logging api */
620void
621nfulnl_log_packet(struct net *net,
622		  u_int8_t pf,
623		  unsigned int hooknum,
624		  const struct sk_buff *skb,
625		  const struct net_device *in,
626		  const struct net_device *out,
627		  const struct nf_loginfo *li_user,
628		  const char *prefix)
629{
630	size_t size;
631	unsigned int data_len;
632	struct nfulnl_instance *inst;
633	const struct nf_loginfo *li;
634	unsigned int qthreshold;
635	unsigned int plen;
636	struct nfnl_log_net *log = nfnl_log_pernet(net);
637	const struct nfnl_ct_hook *nfnl_ct = NULL;
638	struct nf_conn *ct = NULL;
639	enum ip_conntrack_info uninitialized_var(ctinfo);
640
641	if (li_user && li_user->type == NF_LOG_TYPE_ULOG)
642		li = li_user;
643	else
644		li = &default_loginfo;
645
646	inst = instance_lookup_get(log, li->u.ulog.group);
647	if (!inst)
648		return;
649
650	plen = 0;
651	if (prefix)
652		plen = strlen(prefix) + 1;
653
654	/* FIXME: do we want to make the size calculation conditional based on
655	 * what is actually present?  way more branches and checks, but more
656	 * memory efficient... */
657	size =    nlmsg_total_size(sizeof(struct nfgenmsg))
658		+ nla_total_size(sizeof(struct nfulnl_msg_packet_hdr))
659		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
660		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
661#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
662		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
663		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
664#endif
665		+ nla_total_size(sizeof(u_int32_t))	/* mark */
666		+ nla_total_size(sizeof(u_int32_t))	/* uid */
667		+ nla_total_size(sizeof(u_int32_t))	/* gid */
668		+ nla_total_size(plen)			/* prefix */
669		+ nla_total_size(sizeof(struct nfulnl_msg_packet_hw))
670		+ nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp))
671		+ nla_total_size(sizeof(struct nfgenmsg));	/* NLMSG_DONE */
672
673	if (in && skb_mac_header_was_set(skb)) {
674		size +=   nla_total_size(skb->dev->hard_header_len)
675			+ nla_total_size(sizeof(u_int16_t))	/* hwtype */
676			+ nla_total_size(sizeof(u_int16_t));	/* hwlen */
677	}
678
679	spin_lock_bh(&inst->lock);
680
681	if (inst->flags & NFULNL_CFG_F_SEQ)
682		size += nla_total_size(sizeof(u_int32_t));
683	if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
684		size += nla_total_size(sizeof(u_int32_t));
685	if (inst->flags & NFULNL_CFG_F_CONNTRACK) {
686		nfnl_ct = rcu_dereference(nfnl_ct_hook);
687		if (nfnl_ct != NULL) {
688			ct = nfnl_ct->get_ct(skb, &ctinfo);
689			if (ct != NULL)
690				size += nfnl_ct->build_size(ct);
691		}
692	}
693
694	qthreshold = inst->qthreshold;
695	/* per-rule qthreshold overrides per-instance */
696	if (li->u.ulog.qthreshold)
697		if (qthreshold > li->u.ulog.qthreshold)
698			qthreshold = li->u.ulog.qthreshold;
699
700
701	switch (inst->copy_mode) {
702	case NFULNL_COPY_META:
703	case NFULNL_COPY_NONE:
704		data_len = 0;
705		break;
706
707	case NFULNL_COPY_PACKET:
708		if (inst->copy_range > skb->len)
709			data_len = skb->len;
710		else
711			data_len = inst->copy_range;
712
713		size += nla_total_size(data_len);
714		break;
715
716	case NFULNL_COPY_DISABLED:
717	default:
718		goto unlock_and_release;
719	}
720
721	if (inst->skb && size > skb_tailroom(inst->skb)) {
722		/* either the queue len is too high or we don't have
723		 * enough room in the skb left. flush to userspace. */
724		__nfulnl_flush(inst);
725	}
726
727	if (!inst->skb) {
728		inst->skb = nfulnl_alloc_skb(net, inst->peer_portid,
729					     inst->nlbufsiz, size);
730		if (!inst->skb)
731			goto alloc_failure;
732	}
733
734	inst->qlen++;
735
736	__build_packet_message(log, inst, skb, data_len, pf,
737				hooknum, in, out, prefix, plen,
738				nfnl_ct, ct, ctinfo);
739
740	if (inst->qlen >= qthreshold)
741		__nfulnl_flush(inst);
742	/* timer_pending always called within inst->lock, so there
743	 * is no chance of a race here */
744	else if (!timer_pending(&inst->timer)) {
745		instance_get(inst);
746		inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100);
747		add_timer(&inst->timer);
748	}
749
750unlock_and_release:
751	spin_unlock_bh(&inst->lock);
752	instance_put(inst);
753	return;
754
755alloc_failure:
756	/* FIXME: statistics */
757	goto unlock_and_release;
758}
759EXPORT_SYMBOL_GPL(nfulnl_log_packet);
760
761static int
762nfulnl_rcv_nl_event(struct notifier_block *this,
763		   unsigned long event, void *ptr)
764{
765	struct netlink_notify *n = ptr;
766	struct nfnl_log_net *log = nfnl_log_pernet(n->net);
767
768	if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
769		int i;
770
771		/* destroy all instances for this portid */
772		spin_lock_bh(&log->instances_lock);
773		for  (i = 0; i < INSTANCE_BUCKETS; i++) {
774			struct hlist_node *t2;
775			struct nfulnl_instance *inst;
776			struct hlist_head *head = &log->instance_table[i];
777
778			hlist_for_each_entry_safe(inst, t2, head, hlist) {
779				if (n->portid == inst->peer_portid)
780					__instance_destroy(inst);
781			}
782		}
783		spin_unlock_bh(&log->instances_lock);
784	}
785	return NOTIFY_DONE;
786}
787
788static struct notifier_block nfulnl_rtnl_notifier = {
789	.notifier_call	= nfulnl_rcv_nl_event,
790};
791
792static int
793nfulnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
794		   const struct nlmsghdr *nlh,
795		   const struct nlattr * const nfqa[])
796{
797	return -ENOTSUPP;
798}
799
800static struct nf_logger nfulnl_logger __read_mostly = {
801	.name	= "nfnetlink_log",
802	.type	= NF_LOG_TYPE_ULOG,
803	.logfn	= &nfulnl_log_packet,
804	.me	= THIS_MODULE,
805};
806
807static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = {
808	[NFULA_CFG_CMD]		= { .len = sizeof(struct nfulnl_msg_config_cmd) },
809	[NFULA_CFG_MODE]	= { .len = sizeof(struct nfulnl_msg_config_mode) },
810	[NFULA_CFG_TIMEOUT]	= { .type = NLA_U32 },
811	[NFULA_CFG_QTHRESH]	= { .type = NLA_U32 },
812	[NFULA_CFG_NLBUFSIZ]	= { .type = NLA_U32 },
813	[NFULA_CFG_FLAGS]	= { .type = NLA_U16 },
814};
815
816static int
817nfulnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
818		   const struct nlmsghdr *nlh,
819		   const struct nlattr * const nfula[])
820{
821	struct nfgenmsg *nfmsg = nlmsg_data(nlh);
822	u_int16_t group_num = ntohs(nfmsg->res_id);
823	struct nfulnl_instance *inst;
824	struct nfulnl_msg_config_cmd *cmd = NULL;
825	struct net *net = sock_net(ctnl);
826	struct nfnl_log_net *log = nfnl_log_pernet(net);
827	int ret = 0;
828	u16 flags = 0;
829
830	if (nfula[NFULA_CFG_CMD]) {
831		u_int8_t pf = nfmsg->nfgen_family;
832		cmd = nla_data(nfula[NFULA_CFG_CMD]);
833
834		/* Commands without queue context */
835		switch (cmd->command) {
836		case NFULNL_CFG_CMD_PF_BIND:
837			return nf_log_bind_pf(net, pf, &nfulnl_logger);
838		case NFULNL_CFG_CMD_PF_UNBIND:
839			nf_log_unbind_pf(net, pf);
840			return 0;
841		}
842	}
843
844	inst = instance_lookup_get(log, group_num);
845	if (inst && inst->peer_portid != NETLINK_CB(skb).portid) {
846		ret = -EPERM;
847		goto out_put;
848	}
849
850	/* Check if we support these flags in first place, dependencies should
851	 * be there too not to break atomicity.
852	 */
853	if (nfula[NFULA_CFG_FLAGS]) {
854		flags = ntohs(nla_get_be16(nfula[NFULA_CFG_FLAGS]));
855
856		if ((flags & NFULNL_CFG_F_CONNTRACK) &&
857		    !rcu_access_pointer(nfnl_ct_hook)) {
858#ifdef CONFIG_MODULES
859			nfnl_unlock(NFNL_SUBSYS_ULOG);
860			request_module("ip_conntrack_netlink");
861			nfnl_lock(NFNL_SUBSYS_ULOG);
862			if (rcu_access_pointer(nfnl_ct_hook)) {
863				ret = -EAGAIN;
864				goto out_put;
865			}
866#endif
867			ret = -EOPNOTSUPP;
868			goto out_put;
869		}
870	}
871
872	if (cmd != NULL) {
873		switch (cmd->command) {
874		case NFULNL_CFG_CMD_BIND:
875			if (inst) {
876				ret = -EBUSY;
877				goto out_put;
878			}
879
880			inst = instance_create(net, group_num,
881					       NETLINK_CB(skb).portid,
882					       sk_user_ns(NETLINK_CB(skb).sk));
883			if (IS_ERR(inst)) {
884				ret = PTR_ERR(inst);
885				goto out;
886			}
887			break;
888		case NFULNL_CFG_CMD_UNBIND:
889			if (!inst) {
890				ret = -ENODEV;
891				goto out;
892			}
893
894			instance_destroy(log, inst);
895			goto out_put;
896		default:
897			ret = -ENOTSUPP;
898			break;
899		}
900	} else if (!inst) {
901		ret = -ENODEV;
902		goto out;
903	}
904
905	if (nfula[NFULA_CFG_MODE]) {
906		struct nfulnl_msg_config_mode *params =
907			nla_data(nfula[NFULA_CFG_MODE]);
908
909		nfulnl_set_mode(inst, params->copy_mode,
910				ntohl(params->copy_range));
911	}
912
913	if (nfula[NFULA_CFG_TIMEOUT]) {
914		__be32 timeout = nla_get_be32(nfula[NFULA_CFG_TIMEOUT]);
915
916		nfulnl_set_timeout(inst, ntohl(timeout));
917	}
918
919	if (nfula[NFULA_CFG_NLBUFSIZ]) {
920		__be32 nlbufsiz = nla_get_be32(nfula[NFULA_CFG_NLBUFSIZ]);
921
922		nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz));
923	}
924
925	if (nfula[NFULA_CFG_QTHRESH]) {
926		__be32 qthresh = nla_get_be32(nfula[NFULA_CFG_QTHRESH]);
927
928		nfulnl_set_qthresh(inst, ntohl(qthresh));
929	}
930
931	if (nfula[NFULA_CFG_FLAGS])
932		nfulnl_set_flags(inst, flags);
933
934out_put:
935	instance_put(inst);
936out:
937	return ret;
938}
939
940static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = {
941	[NFULNL_MSG_PACKET]	= { .call = nfulnl_recv_unsupp,
942				    .attr_count = NFULA_MAX, },
943	[NFULNL_MSG_CONFIG]	= { .call = nfulnl_recv_config,
944				    .attr_count = NFULA_CFG_MAX,
945				    .policy = nfula_cfg_policy },
946};
947
948static const struct nfnetlink_subsystem nfulnl_subsys = {
949	.name		= "log",
950	.subsys_id	= NFNL_SUBSYS_ULOG,
951	.cb_count	= NFULNL_MSG_MAX,
952	.cb		= nfulnl_cb,
953};
954
955#ifdef CONFIG_PROC_FS
956struct iter_state {
957	struct seq_net_private p;
958	unsigned int bucket;
959};
960
961static struct hlist_node *get_first(struct net *net, struct iter_state *st)
962{
963	struct nfnl_log_net *log;
964	if (!st)
965		return NULL;
966
967	log = nfnl_log_pernet(net);
968
969	for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
970		struct hlist_head *head = &log->instance_table[st->bucket];
971
972		if (!hlist_empty(head))
973			return rcu_dereference_bh(hlist_first_rcu(head));
974	}
975	return NULL;
976}
977
978static struct hlist_node *get_next(struct net *net, struct iter_state *st,
979				   struct hlist_node *h)
980{
981	h = rcu_dereference_bh(hlist_next_rcu(h));
982	while (!h) {
983		struct nfnl_log_net *log;
984		struct hlist_head *head;
985
986		if (++st->bucket >= INSTANCE_BUCKETS)
987			return NULL;
988
989		log = nfnl_log_pernet(net);
990		head = &log->instance_table[st->bucket];
991		h = rcu_dereference_bh(hlist_first_rcu(head));
992	}
993	return h;
994}
995
996static struct hlist_node *get_idx(struct net *net, struct iter_state *st,
997				  loff_t pos)
998{
999	struct hlist_node *head;
1000	head = get_first(net, st);
1001
1002	if (head)
1003		while (pos && (head = get_next(net, st, head)))
1004			pos--;
1005	return pos ? NULL : head;
1006}
1007
1008static void *seq_start(struct seq_file *s, loff_t *pos)
1009	__acquires(rcu_bh)
1010{
1011	rcu_read_lock_bh();
1012	return get_idx(seq_file_net(s), s->private, *pos);
1013}
1014
1015static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1016{
1017	(*pos)++;
1018	return get_next(seq_file_net(s), s->private, v);
1019}
1020
1021static void seq_stop(struct seq_file *s, void *v)
1022	__releases(rcu_bh)
1023{
1024	rcu_read_unlock_bh();
1025}
1026
1027static int seq_show(struct seq_file *s, void *v)
1028{
1029	const struct nfulnl_instance *inst = v;
1030
1031	seq_printf(s, "%5u %6u %5u %1u %5u %6u %2u\n",
1032		   inst->group_num,
1033		   inst->peer_portid, inst->qlen,
1034		   inst->copy_mode, inst->copy_range,
1035		   inst->flushtimeout, atomic_read(&inst->use));
1036
1037	return 0;
1038}
1039
1040static const struct seq_operations nful_seq_ops = {
1041	.start	= seq_start,
1042	.next	= seq_next,
1043	.stop	= seq_stop,
1044	.show	= seq_show,
1045};
1046
1047static int nful_open(struct inode *inode, struct file *file)
1048{
1049	return seq_open_net(inode, file, &nful_seq_ops,
1050			    sizeof(struct iter_state));
1051}
1052
1053static const struct file_operations nful_file_ops = {
1054	.owner	 = THIS_MODULE,
1055	.open	 = nful_open,
1056	.read	 = seq_read,
1057	.llseek	 = seq_lseek,
1058	.release = seq_release_net,
1059};
1060
1061#endif /* PROC_FS */
1062
1063static int __net_init nfnl_log_net_init(struct net *net)
1064{
1065	unsigned int i;
1066	struct nfnl_log_net *log = nfnl_log_pernet(net);
1067
1068	for (i = 0; i < INSTANCE_BUCKETS; i++)
1069		INIT_HLIST_HEAD(&log->instance_table[i]);
1070	spin_lock_init(&log->instances_lock);
1071
1072#ifdef CONFIG_PROC_FS
1073	if (!proc_create("nfnetlink_log", 0440,
1074			 net->nf.proc_netfilter, &nful_file_ops))
1075		return -ENOMEM;
1076#endif
1077	return 0;
1078}
1079
1080static void __net_exit nfnl_log_net_exit(struct net *net)
1081{
1082#ifdef CONFIG_PROC_FS
1083	remove_proc_entry("nfnetlink_log", net->nf.proc_netfilter);
1084#endif
1085	nf_log_unset(net, &nfulnl_logger);
1086}
1087
1088static struct pernet_operations nfnl_log_net_ops = {
1089	.init	= nfnl_log_net_init,
1090	.exit	= nfnl_log_net_exit,
1091	.id	= &nfnl_log_net_id,
1092	.size	= sizeof(struct nfnl_log_net),
1093};
1094
1095static int __init nfnetlink_log_init(void)
1096{
1097	int status;
1098
1099	status = register_pernet_subsys(&nfnl_log_net_ops);
1100	if (status < 0) {
1101		pr_err("failed to register pernet ops\n");
1102		goto out;
1103	}
1104
1105	netlink_register_notifier(&nfulnl_rtnl_notifier);
1106	status = nfnetlink_subsys_register(&nfulnl_subsys);
1107	if (status < 0) {
1108		pr_err("failed to create netlink socket\n");
1109		goto cleanup_netlink_notifier;
1110	}
1111
1112	status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger);
1113	if (status < 0) {
1114		pr_err("failed to register logger\n");
1115		goto cleanup_subsys;
1116	}
1117
1118	return status;
1119
1120cleanup_subsys:
1121	nfnetlink_subsys_unregister(&nfulnl_subsys);
1122cleanup_netlink_notifier:
1123	netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1124	unregister_pernet_subsys(&nfnl_log_net_ops);
1125out:
1126	return status;
1127}
1128
1129static void __exit nfnetlink_log_fini(void)
1130{
1131	nf_log_unregister(&nfulnl_logger);
1132	nfnetlink_subsys_unregister(&nfulnl_subsys);
1133	netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1134	unregister_pernet_subsys(&nfnl_log_net_ops);
1135}
1136
1137MODULE_DESCRIPTION("netfilter userspace logging");
1138MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1139MODULE_LICENSE("GPL");
1140MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG);
1141MODULE_ALIAS_NF_LOGGER(AF_INET, 1);
1142MODULE_ALIAS_NF_LOGGER(AF_INET6, 1);
1143MODULE_ALIAS_NF_LOGGER(AF_BRIDGE, 1);
1144
1145module_init(nfnetlink_log_init);
1146module_exit(nfnetlink_log_fini);
1147