Loading...
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);
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 *
7 * Based on the old ipv4-only ipt_ULOG.c:
8 * (C) 2000-2004 by Harald Welte <laforge@netfilter.org>
9 *
10 * This program is free software; you can redistribute it and/or modify
11 * it under the terms of the GNU General Public License version 2 as
12 * published by the Free Software Foundation.
13 */
14#include <linux/module.h>
15#include <linux/skbuff.h>
16#include <linux/init.h>
17#include <linux/ip.h>
18#include <linux/ipv6.h>
19#include <linux/netdevice.h>
20#include <linux/netfilter.h>
21#include <linux/netlink.h>
22#include <linux/netfilter/nfnetlink.h>
23#include <linux/netfilter/nfnetlink_log.h>
24#include <linux/spinlock.h>
25#include <linux/sysctl.h>
26#include <linux/proc_fs.h>
27#include <linux/security.h>
28#include <linux/list.h>
29#include <linux/jhash.h>
30#include <linux/random.h>
31#include <linux/slab.h>
32#include <net/sock.h>
33#include <net/netfilter/nf_log.h>
34#include <net/netfilter/nfnetlink_log.h>
35
36#include <linux/atomic.h>
37
38#ifdef CONFIG_BRIDGE_NETFILTER
39#include "../bridge/br_private.h"
40#endif
41
42#define NFULNL_NLBUFSIZ_DEFAULT NLMSG_GOODSIZE
43#define NFULNL_TIMEOUT_DEFAULT 100 /* every second */
44#define NFULNL_QTHRESH_DEFAULT 100 /* 100 packets */
45#define NFULNL_COPY_RANGE_MAX 0xFFFF /* max packet size is limited by 16-bit struct nfattr nfa_len field */
46
47#define PRINTR(x, args...) do { if (net_ratelimit()) \
48 printk(x, ## args); } while (0);
49
50struct nfulnl_instance {
51 struct hlist_node hlist; /* global list of instances */
52 spinlock_t lock;
53 atomic_t use; /* use count */
54
55 unsigned int qlen; /* number of nlmsgs in skb */
56 struct sk_buff *skb; /* pre-allocatd skb */
57 struct timer_list timer;
58 int peer_pid; /* PID of the peer process */
59
60 /* configurable parameters */
61 unsigned int flushtimeout; /* timeout until queue flush */
62 unsigned int nlbufsiz; /* netlink buffer allocation size */
63 unsigned int qthreshold; /* threshold of the queue */
64 u_int32_t copy_range;
65 u_int32_t seq; /* instance-local sequential counter */
66 u_int16_t group_num; /* number of this queue */
67 u_int16_t flags;
68 u_int8_t copy_mode;
69 struct rcu_head rcu;
70};
71
72static DEFINE_SPINLOCK(instances_lock);
73static atomic_t global_seq;
74
75#define INSTANCE_BUCKETS 16
76static struct hlist_head instance_table[INSTANCE_BUCKETS];
77static unsigned int hash_init;
78
79static inline u_int8_t instance_hashfn(u_int16_t group_num)
80{
81 return ((group_num & 0xff) % INSTANCE_BUCKETS);
82}
83
84static struct nfulnl_instance *
85__instance_lookup(u_int16_t group_num)
86{
87 struct hlist_head *head;
88 struct hlist_node *pos;
89 struct nfulnl_instance *inst;
90
91 head = &instance_table[instance_hashfn(group_num)];
92 hlist_for_each_entry_rcu(inst, pos, head, hlist) {
93 if (inst->group_num == group_num)
94 return inst;
95 }
96 return NULL;
97}
98
99static inline void
100instance_get(struct nfulnl_instance *inst)
101{
102 atomic_inc(&inst->use);
103}
104
105static struct nfulnl_instance *
106instance_lookup_get(u_int16_t group_num)
107{
108 struct nfulnl_instance *inst;
109
110 rcu_read_lock_bh();
111 inst = __instance_lookup(group_num);
112 if (inst && !atomic_inc_not_zero(&inst->use))
113 inst = NULL;
114 rcu_read_unlock_bh();
115
116 return inst;
117}
118
119static void nfulnl_instance_free_rcu(struct rcu_head *head)
120{
121 kfree(container_of(head, struct nfulnl_instance, rcu));
122 module_put(THIS_MODULE);
123}
124
125static void
126instance_put(struct nfulnl_instance *inst)
127{
128 if (inst && atomic_dec_and_test(&inst->use))
129 call_rcu_bh(&inst->rcu, nfulnl_instance_free_rcu);
130}
131
132static void nfulnl_timer(unsigned long data);
133
134static struct nfulnl_instance *
135instance_create(u_int16_t group_num, int pid)
136{
137 struct nfulnl_instance *inst;
138 int err;
139
140 spin_lock_bh(&instances_lock);
141 if (__instance_lookup(group_num)) {
142 err = -EEXIST;
143 goto out_unlock;
144 }
145
146 inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
147 if (!inst) {
148 err = -ENOMEM;
149 goto out_unlock;
150 }
151
152 if (!try_module_get(THIS_MODULE)) {
153 kfree(inst);
154 err = -EAGAIN;
155 goto out_unlock;
156 }
157
158 INIT_HLIST_NODE(&inst->hlist);
159 spin_lock_init(&inst->lock);
160 /* needs to be two, since we _put() after creation */
161 atomic_set(&inst->use, 2);
162
163 setup_timer(&inst->timer, nfulnl_timer, (unsigned long)inst);
164
165 inst->peer_pid = pid;
166 inst->group_num = group_num;
167
168 inst->qthreshold = NFULNL_QTHRESH_DEFAULT;
169 inst->flushtimeout = NFULNL_TIMEOUT_DEFAULT;
170 inst->nlbufsiz = NFULNL_NLBUFSIZ_DEFAULT;
171 inst->copy_mode = NFULNL_COPY_PACKET;
172 inst->copy_range = NFULNL_COPY_RANGE_MAX;
173
174 hlist_add_head_rcu(&inst->hlist,
175 &instance_table[instance_hashfn(group_num)]);
176
177 spin_unlock_bh(&instances_lock);
178
179 return inst;
180
181out_unlock:
182 spin_unlock_bh(&instances_lock);
183 return ERR_PTR(err);
184}
185
186static void __nfulnl_flush(struct nfulnl_instance *inst);
187
188/* called with BH disabled */
189static void
190__instance_destroy(struct nfulnl_instance *inst)
191{
192 /* first pull it out of the global list */
193 hlist_del_rcu(&inst->hlist);
194
195 /* then flush all pending packets from skb */
196
197 spin_lock(&inst->lock);
198
199 /* lockless readers wont be able to use us */
200 inst->copy_mode = NFULNL_COPY_DISABLED;
201
202 if (inst->skb)
203 __nfulnl_flush(inst);
204 spin_unlock(&inst->lock);
205
206 /* and finally put the refcount */
207 instance_put(inst);
208}
209
210static inline void
211instance_destroy(struct nfulnl_instance *inst)
212{
213 spin_lock_bh(&instances_lock);
214 __instance_destroy(inst);
215 spin_unlock_bh(&instances_lock);
216}
217
218static int
219nfulnl_set_mode(struct nfulnl_instance *inst, u_int8_t mode,
220 unsigned int range)
221{
222 int status = 0;
223
224 spin_lock_bh(&inst->lock);
225
226 switch (mode) {
227 case NFULNL_COPY_NONE:
228 case NFULNL_COPY_META:
229 inst->copy_mode = mode;
230 inst->copy_range = 0;
231 break;
232
233 case NFULNL_COPY_PACKET:
234 inst->copy_mode = mode;
235 inst->copy_range = min_t(unsigned int,
236 range, NFULNL_COPY_RANGE_MAX);
237 break;
238
239 default:
240 status = -EINVAL;
241 break;
242 }
243
244 spin_unlock_bh(&inst->lock);
245
246 return status;
247}
248
249static int
250nfulnl_set_nlbufsiz(struct nfulnl_instance *inst, u_int32_t nlbufsiz)
251{
252 int status;
253
254 spin_lock_bh(&inst->lock);
255 if (nlbufsiz < NFULNL_NLBUFSIZ_DEFAULT)
256 status = -ERANGE;
257 else if (nlbufsiz > 131072)
258 status = -ERANGE;
259 else {
260 inst->nlbufsiz = nlbufsiz;
261 status = 0;
262 }
263 spin_unlock_bh(&inst->lock);
264
265 return status;
266}
267
268static int
269nfulnl_set_timeout(struct nfulnl_instance *inst, u_int32_t timeout)
270{
271 spin_lock_bh(&inst->lock);
272 inst->flushtimeout = timeout;
273 spin_unlock_bh(&inst->lock);
274
275 return 0;
276}
277
278static int
279nfulnl_set_qthresh(struct nfulnl_instance *inst, u_int32_t qthresh)
280{
281 spin_lock_bh(&inst->lock);
282 inst->qthreshold = qthresh;
283 spin_unlock_bh(&inst->lock);
284
285 return 0;
286}
287
288static int
289nfulnl_set_flags(struct nfulnl_instance *inst, u_int16_t flags)
290{
291 spin_lock_bh(&inst->lock);
292 inst->flags = flags;
293 spin_unlock_bh(&inst->lock);
294
295 return 0;
296}
297
298static struct sk_buff *
299nfulnl_alloc_skb(unsigned int inst_size, unsigned int pkt_size)
300{
301 struct sk_buff *skb;
302 unsigned int n;
303
304 /* alloc skb which should be big enough for a whole multipart
305 * message. WARNING: has to be <= 128k due to slab restrictions */
306
307 n = max(inst_size, pkt_size);
308 skb = alloc_skb(n, GFP_ATOMIC);
309 if (!skb) {
310 pr_notice("nfnetlink_log: can't alloc whole buffer (%u bytes)\n",
311 inst_size);
312
313 if (n > pkt_size) {
314 /* try to allocate only as much as we need for current
315 * packet */
316
317 skb = alloc_skb(pkt_size, GFP_ATOMIC);
318 if (!skb)
319 pr_err("nfnetlink_log: can't even alloc %u "
320 "bytes\n", pkt_size);
321 }
322 }
323
324 return skb;
325}
326
327static int
328__nfulnl_send(struct nfulnl_instance *inst)
329{
330 int status = -1;
331
332 if (inst->qlen > 1)
333 NLMSG_PUT(inst->skb, 0, 0,
334 NLMSG_DONE,
335 sizeof(struct nfgenmsg));
336
337 status = nfnetlink_unicast(inst->skb, &init_net, inst->peer_pid,
338 MSG_DONTWAIT);
339
340 inst->qlen = 0;
341 inst->skb = NULL;
342
343nlmsg_failure:
344 return status;
345}
346
347static void
348__nfulnl_flush(struct nfulnl_instance *inst)
349{
350 /* timer holds a reference */
351 if (del_timer(&inst->timer))
352 instance_put(inst);
353 if (inst->skb)
354 __nfulnl_send(inst);
355}
356
357static void
358nfulnl_timer(unsigned long data)
359{
360 struct nfulnl_instance *inst = (struct nfulnl_instance *)data;
361
362 spin_lock_bh(&inst->lock);
363 if (inst->skb)
364 __nfulnl_send(inst);
365 spin_unlock_bh(&inst->lock);
366 instance_put(inst);
367}
368
369/* This is an inline function, we don't really care about a long
370 * list of arguments */
371static inline int
372__build_packet_message(struct nfulnl_instance *inst,
373 const struct sk_buff *skb,
374 unsigned int data_len,
375 u_int8_t pf,
376 unsigned int hooknum,
377 const struct net_device *indev,
378 const struct net_device *outdev,
379 const char *prefix, unsigned int plen)
380{
381 struct nfulnl_msg_packet_hdr pmsg;
382 struct nlmsghdr *nlh;
383 struct nfgenmsg *nfmsg;
384 sk_buff_data_t old_tail = inst->skb->tail;
385
386 nlh = NLMSG_PUT(inst->skb, 0, 0,
387 NFNL_SUBSYS_ULOG << 8 | NFULNL_MSG_PACKET,
388 sizeof(struct nfgenmsg));
389 nfmsg = NLMSG_DATA(nlh);
390 nfmsg->nfgen_family = pf;
391 nfmsg->version = NFNETLINK_V0;
392 nfmsg->res_id = htons(inst->group_num);
393
394 pmsg.hw_protocol = skb->protocol;
395 pmsg.hook = hooknum;
396
397 NLA_PUT(inst->skb, NFULA_PACKET_HDR, sizeof(pmsg), &pmsg);
398
399 if (prefix)
400 NLA_PUT(inst->skb, NFULA_PREFIX, plen, prefix);
401
402 if (indev) {
403#ifndef CONFIG_BRIDGE_NETFILTER
404 NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_INDEV,
405 htonl(indev->ifindex));
406#else
407 if (pf == PF_BRIDGE) {
408 /* Case 1: outdev is physical input device, we need to
409 * look for bridge group (when called from
410 * netfilter_bridge) */
411 NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
412 htonl(indev->ifindex));
413 /* this is the bridge group "brX" */
414 /* rcu_read_lock()ed by nf_hook_slow or nf_log_packet */
415 NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_INDEV,
416 htonl(br_port_get_rcu(indev)->br->dev->ifindex));
417 } else {
418 /* Case 2: indev is bridge group, we need to look for
419 * physical device (when called from ipv4) */
420 NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_INDEV,
421 htonl(indev->ifindex));
422 if (skb->nf_bridge && skb->nf_bridge->physindev)
423 NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
424 htonl(skb->nf_bridge->physindev->ifindex));
425 }
426#endif
427 }
428
429 if (outdev) {
430#ifndef CONFIG_BRIDGE_NETFILTER
431 NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_OUTDEV,
432 htonl(outdev->ifindex));
433#else
434 if (pf == PF_BRIDGE) {
435 /* Case 1: outdev is physical output device, we need to
436 * look for bridge group (when called from
437 * netfilter_bridge) */
438 NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
439 htonl(outdev->ifindex));
440 /* this is the bridge group "brX" */
441 /* rcu_read_lock()ed by nf_hook_slow or nf_log_packet */
442 NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_OUTDEV,
443 htonl(br_port_get_rcu(outdev)->br->dev->ifindex));
444 } else {
445 /* Case 2: indev is a bridge group, we need to look
446 * for physical device (when called from ipv4) */
447 NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_OUTDEV,
448 htonl(outdev->ifindex));
449 if (skb->nf_bridge && skb->nf_bridge->physoutdev)
450 NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
451 htonl(skb->nf_bridge->physoutdev->ifindex));
452 }
453#endif
454 }
455
456 if (skb->mark)
457 NLA_PUT_BE32(inst->skb, NFULA_MARK, htonl(skb->mark));
458
459 if (indev && skb->dev &&
460 skb->mac_header != skb->network_header) {
461 struct nfulnl_msg_packet_hw phw;
462 int len = dev_parse_header(skb, phw.hw_addr);
463 if (len > 0) {
464 phw.hw_addrlen = htons(len);
465 NLA_PUT(inst->skb, NFULA_HWADDR, sizeof(phw), &phw);
466 }
467 }
468
469 if (indev && skb_mac_header_was_set(skb)) {
470 NLA_PUT_BE16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type));
471 NLA_PUT_BE16(inst->skb, NFULA_HWLEN,
472 htons(skb->dev->hard_header_len));
473 NLA_PUT(inst->skb, NFULA_HWHEADER, skb->dev->hard_header_len,
474 skb_mac_header(skb));
475 }
476
477 if (skb->tstamp.tv64) {
478 struct nfulnl_msg_packet_timestamp ts;
479 struct timeval tv = ktime_to_timeval(skb->tstamp);
480 ts.sec = cpu_to_be64(tv.tv_sec);
481 ts.usec = cpu_to_be64(tv.tv_usec);
482
483 NLA_PUT(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts);
484 }
485
486 /* UID */
487 if (skb->sk) {
488 read_lock_bh(&skb->sk->sk_callback_lock);
489 if (skb->sk->sk_socket && skb->sk->sk_socket->file) {
490 struct file *file = skb->sk->sk_socket->file;
491 __be32 uid = htonl(file->f_cred->fsuid);
492 __be32 gid = htonl(file->f_cred->fsgid);
493 /* need to unlock here since NLA_PUT may goto */
494 read_unlock_bh(&skb->sk->sk_callback_lock);
495 NLA_PUT_BE32(inst->skb, NFULA_UID, uid);
496 NLA_PUT_BE32(inst->skb, NFULA_GID, gid);
497 } else
498 read_unlock_bh(&skb->sk->sk_callback_lock);
499 }
500
501 /* local sequence number */
502 if (inst->flags & NFULNL_CFG_F_SEQ)
503 NLA_PUT_BE32(inst->skb, NFULA_SEQ, htonl(inst->seq++));
504
505 /* global sequence number */
506 if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
507 NLA_PUT_BE32(inst->skb, NFULA_SEQ_GLOBAL,
508 htonl(atomic_inc_return(&global_seq)));
509
510 if (data_len) {
511 struct nlattr *nla;
512 int size = nla_attr_size(data_len);
513
514 if (skb_tailroom(inst->skb) < nla_total_size(data_len)) {
515 printk(KERN_WARNING "nfnetlink_log: no tailroom!\n");
516 goto nlmsg_failure;
517 }
518
519 nla = (struct nlattr *)skb_put(inst->skb, nla_total_size(data_len));
520 nla->nla_type = NFULA_PAYLOAD;
521 nla->nla_len = size;
522
523 if (skb_copy_bits(skb, 0, nla_data(nla), data_len))
524 BUG();
525 }
526
527 nlh->nlmsg_len = inst->skb->tail - old_tail;
528 return 0;
529
530nlmsg_failure:
531nla_put_failure:
532 PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n");
533 return -1;
534}
535
536#define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
537
538static struct nf_loginfo default_loginfo = {
539 .type = NF_LOG_TYPE_ULOG,
540 .u = {
541 .ulog = {
542 .copy_len = 0xffff,
543 .group = 0,
544 .qthreshold = 1,
545 },
546 },
547};
548
549/* log handler for internal netfilter logging api */
550void
551nfulnl_log_packet(u_int8_t pf,
552 unsigned int hooknum,
553 const struct sk_buff *skb,
554 const struct net_device *in,
555 const struct net_device *out,
556 const struct nf_loginfo *li_user,
557 const char *prefix)
558{
559 unsigned int size, data_len;
560 struct nfulnl_instance *inst;
561 const struct nf_loginfo *li;
562 unsigned int qthreshold;
563 unsigned int plen;
564
565 if (li_user && li_user->type == NF_LOG_TYPE_ULOG)
566 li = li_user;
567 else
568 li = &default_loginfo;
569
570 inst = instance_lookup_get(li->u.ulog.group);
571 if (!inst)
572 return;
573
574 plen = 0;
575 if (prefix)
576 plen = strlen(prefix) + 1;
577
578 /* FIXME: do we want to make the size calculation conditional based on
579 * what is actually present? way more branches and checks, but more
580 * memory efficient... */
581 size = NLMSG_SPACE(sizeof(struct nfgenmsg))
582 + nla_total_size(sizeof(struct nfulnl_msg_packet_hdr))
583 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
584 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
585#ifdef CONFIG_BRIDGE_NETFILTER
586 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
587 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
588#endif
589 + nla_total_size(sizeof(u_int32_t)) /* mark */
590 + nla_total_size(sizeof(u_int32_t)) /* uid */
591 + nla_total_size(sizeof(u_int32_t)) /* gid */
592 + nla_total_size(plen) /* prefix */
593 + nla_total_size(sizeof(struct nfulnl_msg_packet_hw))
594 + nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp));
595
596 if (in && skb_mac_header_was_set(skb)) {
597 size += nla_total_size(skb->dev->hard_header_len)
598 + nla_total_size(sizeof(u_int16_t)) /* hwtype */
599 + nla_total_size(sizeof(u_int16_t)); /* hwlen */
600 }
601
602 spin_lock_bh(&inst->lock);
603
604 if (inst->flags & NFULNL_CFG_F_SEQ)
605 size += nla_total_size(sizeof(u_int32_t));
606 if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
607 size += nla_total_size(sizeof(u_int32_t));
608
609 qthreshold = inst->qthreshold;
610 /* per-rule qthreshold overrides per-instance */
611 if (li->u.ulog.qthreshold)
612 if (qthreshold > li->u.ulog.qthreshold)
613 qthreshold = li->u.ulog.qthreshold;
614
615
616 switch (inst->copy_mode) {
617 case NFULNL_COPY_META:
618 case NFULNL_COPY_NONE:
619 data_len = 0;
620 break;
621
622 case NFULNL_COPY_PACKET:
623 if (inst->copy_range == 0
624 || inst->copy_range > skb->len)
625 data_len = skb->len;
626 else
627 data_len = inst->copy_range;
628
629 size += nla_total_size(data_len);
630 break;
631
632 case NFULNL_COPY_DISABLED:
633 default:
634 goto unlock_and_release;
635 }
636
637 if (inst->skb &&
638 size > skb_tailroom(inst->skb) - sizeof(struct nfgenmsg)) {
639 /* either the queue len is too high or we don't have
640 * enough room in the skb left. flush to userspace. */
641 __nfulnl_flush(inst);
642 }
643
644 if (!inst->skb) {
645 inst->skb = nfulnl_alloc_skb(inst->nlbufsiz, size);
646 if (!inst->skb)
647 goto alloc_failure;
648 }
649
650 inst->qlen++;
651
652 __build_packet_message(inst, skb, data_len, pf,
653 hooknum, in, out, prefix, plen);
654
655 if (inst->qlen >= qthreshold)
656 __nfulnl_flush(inst);
657 /* timer_pending always called within inst->lock, so there
658 * is no chance of a race here */
659 else if (!timer_pending(&inst->timer)) {
660 instance_get(inst);
661 inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100);
662 add_timer(&inst->timer);
663 }
664
665unlock_and_release:
666 spin_unlock_bh(&inst->lock);
667 instance_put(inst);
668 return;
669
670alloc_failure:
671 /* FIXME: statistics */
672 goto unlock_and_release;
673}
674EXPORT_SYMBOL_GPL(nfulnl_log_packet);
675
676static int
677nfulnl_rcv_nl_event(struct notifier_block *this,
678 unsigned long event, void *ptr)
679{
680 struct netlink_notify *n = ptr;
681
682 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
683 int i;
684
685 /* destroy all instances for this pid */
686 spin_lock_bh(&instances_lock);
687 for (i = 0; i < INSTANCE_BUCKETS; i++) {
688 struct hlist_node *tmp, *t2;
689 struct nfulnl_instance *inst;
690 struct hlist_head *head = &instance_table[i];
691
692 hlist_for_each_entry_safe(inst, tmp, t2, head, hlist) {
693 if ((net_eq(n->net, &init_net)) &&
694 (n->pid == inst->peer_pid))
695 __instance_destroy(inst);
696 }
697 }
698 spin_unlock_bh(&instances_lock);
699 }
700 return NOTIFY_DONE;
701}
702
703static struct notifier_block nfulnl_rtnl_notifier = {
704 .notifier_call = nfulnl_rcv_nl_event,
705};
706
707static int
708nfulnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
709 const struct nlmsghdr *nlh,
710 const struct nlattr * const nfqa[])
711{
712 return -ENOTSUPP;
713}
714
715static struct nf_logger nfulnl_logger __read_mostly = {
716 .name = "nfnetlink_log",
717 .logfn = &nfulnl_log_packet,
718 .me = THIS_MODULE,
719};
720
721static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = {
722 [NFULA_CFG_CMD] = { .len = sizeof(struct nfulnl_msg_config_cmd) },
723 [NFULA_CFG_MODE] = { .len = sizeof(struct nfulnl_msg_config_mode) },
724 [NFULA_CFG_TIMEOUT] = { .type = NLA_U32 },
725 [NFULA_CFG_QTHRESH] = { .type = NLA_U32 },
726 [NFULA_CFG_NLBUFSIZ] = { .type = NLA_U32 },
727 [NFULA_CFG_FLAGS] = { .type = NLA_U16 },
728};
729
730static int
731nfulnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
732 const struct nlmsghdr *nlh,
733 const struct nlattr * const nfula[])
734{
735 struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
736 u_int16_t group_num = ntohs(nfmsg->res_id);
737 struct nfulnl_instance *inst;
738 struct nfulnl_msg_config_cmd *cmd = NULL;
739 int ret = 0;
740
741 if (nfula[NFULA_CFG_CMD]) {
742 u_int8_t pf = nfmsg->nfgen_family;
743 cmd = nla_data(nfula[NFULA_CFG_CMD]);
744
745 /* Commands without queue context */
746 switch (cmd->command) {
747 case NFULNL_CFG_CMD_PF_BIND:
748 return nf_log_bind_pf(pf, &nfulnl_logger);
749 case NFULNL_CFG_CMD_PF_UNBIND:
750 nf_log_unbind_pf(pf);
751 return 0;
752 }
753 }
754
755 inst = instance_lookup_get(group_num);
756 if (inst && inst->peer_pid != NETLINK_CB(skb).pid) {
757 ret = -EPERM;
758 goto out_put;
759 }
760
761 if (cmd != NULL) {
762 switch (cmd->command) {
763 case NFULNL_CFG_CMD_BIND:
764 if (inst) {
765 ret = -EBUSY;
766 goto out_put;
767 }
768
769 inst = instance_create(group_num,
770 NETLINK_CB(skb).pid);
771 if (IS_ERR(inst)) {
772 ret = PTR_ERR(inst);
773 goto out;
774 }
775 break;
776 case NFULNL_CFG_CMD_UNBIND:
777 if (!inst) {
778 ret = -ENODEV;
779 goto out;
780 }
781
782 instance_destroy(inst);
783 goto out_put;
784 default:
785 ret = -ENOTSUPP;
786 break;
787 }
788 }
789
790 if (nfula[NFULA_CFG_MODE]) {
791 struct nfulnl_msg_config_mode *params;
792 params = nla_data(nfula[NFULA_CFG_MODE]);
793
794 if (!inst) {
795 ret = -ENODEV;
796 goto out;
797 }
798 nfulnl_set_mode(inst, params->copy_mode,
799 ntohl(params->copy_range));
800 }
801
802 if (nfula[NFULA_CFG_TIMEOUT]) {
803 __be32 timeout = nla_get_be32(nfula[NFULA_CFG_TIMEOUT]);
804
805 if (!inst) {
806 ret = -ENODEV;
807 goto out;
808 }
809 nfulnl_set_timeout(inst, ntohl(timeout));
810 }
811
812 if (nfula[NFULA_CFG_NLBUFSIZ]) {
813 __be32 nlbufsiz = nla_get_be32(nfula[NFULA_CFG_NLBUFSIZ]);
814
815 if (!inst) {
816 ret = -ENODEV;
817 goto out;
818 }
819 nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz));
820 }
821
822 if (nfula[NFULA_CFG_QTHRESH]) {
823 __be32 qthresh = nla_get_be32(nfula[NFULA_CFG_QTHRESH]);
824
825 if (!inst) {
826 ret = -ENODEV;
827 goto out;
828 }
829 nfulnl_set_qthresh(inst, ntohl(qthresh));
830 }
831
832 if (nfula[NFULA_CFG_FLAGS]) {
833 __be16 flags = nla_get_be16(nfula[NFULA_CFG_FLAGS]);
834
835 if (!inst) {
836 ret = -ENODEV;
837 goto out;
838 }
839 nfulnl_set_flags(inst, ntohs(flags));
840 }
841
842out_put:
843 instance_put(inst);
844out:
845 return ret;
846}
847
848static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = {
849 [NFULNL_MSG_PACKET] = { .call = nfulnl_recv_unsupp,
850 .attr_count = NFULA_MAX, },
851 [NFULNL_MSG_CONFIG] = { .call = nfulnl_recv_config,
852 .attr_count = NFULA_CFG_MAX,
853 .policy = nfula_cfg_policy },
854};
855
856static const struct nfnetlink_subsystem nfulnl_subsys = {
857 .name = "log",
858 .subsys_id = NFNL_SUBSYS_ULOG,
859 .cb_count = NFULNL_MSG_MAX,
860 .cb = nfulnl_cb,
861};
862
863#ifdef CONFIG_PROC_FS
864struct iter_state {
865 unsigned int bucket;
866};
867
868static struct hlist_node *get_first(struct iter_state *st)
869{
870 if (!st)
871 return NULL;
872
873 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
874 if (!hlist_empty(&instance_table[st->bucket]))
875 return rcu_dereference_bh(hlist_first_rcu(&instance_table[st->bucket]));
876 }
877 return NULL;
878}
879
880static struct hlist_node *get_next(struct iter_state *st, struct hlist_node *h)
881{
882 h = rcu_dereference_bh(hlist_next_rcu(h));
883 while (!h) {
884 if (++st->bucket >= INSTANCE_BUCKETS)
885 return NULL;
886
887 h = rcu_dereference_bh(hlist_first_rcu(&instance_table[st->bucket]));
888 }
889 return h;
890}
891
892static struct hlist_node *get_idx(struct iter_state *st, loff_t pos)
893{
894 struct hlist_node *head;
895 head = get_first(st);
896
897 if (head)
898 while (pos && (head = get_next(st, head)))
899 pos--;
900 return pos ? NULL : head;
901}
902
903static void *seq_start(struct seq_file *seq, loff_t *pos)
904 __acquires(rcu_bh)
905{
906 rcu_read_lock_bh();
907 return get_idx(seq->private, *pos);
908}
909
910static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
911{
912 (*pos)++;
913 return get_next(s->private, v);
914}
915
916static void seq_stop(struct seq_file *s, void *v)
917 __releases(rcu_bh)
918{
919 rcu_read_unlock_bh();
920}
921
922static int seq_show(struct seq_file *s, void *v)
923{
924 const struct nfulnl_instance *inst = v;
925
926 return seq_printf(s, "%5d %6d %5d %1d %5d %6d %2d\n",
927 inst->group_num,
928 inst->peer_pid, inst->qlen,
929 inst->copy_mode, inst->copy_range,
930 inst->flushtimeout, atomic_read(&inst->use));
931}
932
933static const struct seq_operations nful_seq_ops = {
934 .start = seq_start,
935 .next = seq_next,
936 .stop = seq_stop,
937 .show = seq_show,
938};
939
940static int nful_open(struct inode *inode, struct file *file)
941{
942 return seq_open_private(file, &nful_seq_ops,
943 sizeof(struct iter_state));
944}
945
946static const struct file_operations nful_file_ops = {
947 .owner = THIS_MODULE,
948 .open = nful_open,
949 .read = seq_read,
950 .llseek = seq_lseek,
951 .release = seq_release_private,
952};
953
954#endif /* PROC_FS */
955
956static int __init nfnetlink_log_init(void)
957{
958 int i, status = -ENOMEM;
959
960 for (i = 0; i < INSTANCE_BUCKETS; i++)
961 INIT_HLIST_HEAD(&instance_table[i]);
962
963 /* it's not really all that important to have a random value, so
964 * we can do this from the init function, even if there hasn't
965 * been that much entropy yet */
966 get_random_bytes(&hash_init, sizeof(hash_init));
967
968 netlink_register_notifier(&nfulnl_rtnl_notifier);
969 status = nfnetlink_subsys_register(&nfulnl_subsys);
970 if (status < 0) {
971 printk(KERN_ERR "log: failed to create netlink socket\n");
972 goto cleanup_netlink_notifier;
973 }
974
975 status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger);
976 if (status < 0) {
977 printk(KERN_ERR "log: failed to register logger\n");
978 goto cleanup_subsys;
979 }
980
981#ifdef CONFIG_PROC_FS
982 if (!proc_create("nfnetlink_log", 0440,
983 proc_net_netfilter, &nful_file_ops))
984 goto cleanup_logger;
985#endif
986 return status;
987
988#ifdef CONFIG_PROC_FS
989cleanup_logger:
990 nf_log_unregister(&nfulnl_logger);
991#endif
992cleanup_subsys:
993 nfnetlink_subsys_unregister(&nfulnl_subsys);
994cleanup_netlink_notifier:
995 netlink_unregister_notifier(&nfulnl_rtnl_notifier);
996 return status;
997}
998
999static void __exit nfnetlink_log_fini(void)
1000{
1001 nf_log_unregister(&nfulnl_logger);
1002#ifdef CONFIG_PROC_FS
1003 remove_proc_entry("nfnetlink_log", proc_net_netfilter);
1004#endif
1005 nfnetlink_subsys_unregister(&nfulnl_subsys);
1006 netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1007}
1008
1009MODULE_DESCRIPTION("netfilter userspace logging");
1010MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1011MODULE_LICENSE("GPL");
1012MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG);
1013
1014module_init(nfnetlink_log_init);
1015module_exit(nfnetlink_log_fini);