netfilter: netns nf_conntrack: per-netns statistics
[linux-2.6] / net / netfilter / nfnetlink_log.c
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 <net/sock.h>
32 #include <net/netfilter/nf_log.h>
33
34 #include <asm/atomic.h>
35
36 #ifdef CONFIG_BRIDGE_NETFILTER
37 #include "../bridge/br_private.h"
38 #endif
39
40 #define NFULNL_NLBUFSIZ_DEFAULT NLMSG_GOODSIZE
41 #define NFULNL_TIMEOUT_DEFAULT  HZ      /* every second */
42 #define NFULNL_QTHRESH_DEFAULT  100     /* 100 packets */
43 #define NFULNL_COPY_RANGE_MAX   0xFFFF  /* max packet size is limited by 16-bit struct nfattr nfa_len field */
44
45 #define PRINTR(x, args...)      do { if (net_ratelimit()) \
46                                      printk(x, ## args); } while (0);
47
48 struct nfulnl_instance {
49         struct hlist_node hlist;        /* global list of instances */
50         spinlock_t lock;
51         atomic_t use;                   /* use count */
52
53         unsigned int qlen;              /* number of nlmsgs in skb */
54         struct sk_buff *skb;            /* pre-allocatd skb */
55         struct timer_list timer;
56         int peer_pid;                   /* PID of the peer process */
57
58         /* configurable parameters */
59         unsigned int flushtimeout;      /* timeout until queue flush */
60         unsigned int nlbufsiz;          /* netlink buffer allocation size */
61         unsigned int qthreshold;        /* threshold of the queue */
62         u_int32_t copy_range;
63         u_int32_t seq;                  /* instance-local sequential counter */
64         u_int16_t group_num;            /* number of this queue */
65         u_int16_t flags;
66         u_int8_t copy_mode;
67 };
68
69 static DEFINE_RWLOCK(instances_lock);
70 static atomic_t global_seq;
71
72 #define INSTANCE_BUCKETS        16
73 static struct hlist_head instance_table[INSTANCE_BUCKETS];
74 static unsigned int hash_init;
75
76 static inline u_int8_t instance_hashfn(u_int16_t group_num)
77 {
78         return ((group_num & 0xff) % INSTANCE_BUCKETS);
79 }
80
81 static struct nfulnl_instance *
82 __instance_lookup(u_int16_t group_num)
83 {
84         struct hlist_head *head;
85         struct hlist_node *pos;
86         struct nfulnl_instance *inst;
87
88         head = &instance_table[instance_hashfn(group_num)];
89         hlist_for_each_entry(inst, pos, head, hlist) {
90                 if (inst->group_num == group_num)
91                         return inst;
92         }
93         return NULL;
94 }
95
96 static inline void
97 instance_get(struct nfulnl_instance *inst)
98 {
99         atomic_inc(&inst->use);
100 }
101
102 static struct nfulnl_instance *
103 instance_lookup_get(u_int16_t group_num)
104 {
105         struct nfulnl_instance *inst;
106
107         read_lock_bh(&instances_lock);
108         inst = __instance_lookup(group_num);
109         if (inst)
110                 instance_get(inst);
111         read_unlock_bh(&instances_lock);
112
113         return inst;
114 }
115
116 static void
117 instance_put(struct nfulnl_instance *inst)
118 {
119         if (inst && atomic_dec_and_test(&inst->use)) {
120                 kfree(inst);
121                 module_put(THIS_MODULE);
122         }
123 }
124
125 static void nfulnl_timer(unsigned long data);
126
127 static struct nfulnl_instance *
128 instance_create(u_int16_t group_num, int pid)
129 {
130         struct nfulnl_instance *inst;
131         int err;
132
133         write_lock_bh(&instances_lock);
134         if (__instance_lookup(group_num)) {
135                 err = -EEXIST;
136                 goto out_unlock;
137         }
138
139         inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
140         if (!inst) {
141                 err = -ENOMEM;
142                 goto out_unlock;
143         }
144
145         if (!try_module_get(THIS_MODULE)) {
146                 kfree(inst);
147                 err = -EAGAIN;
148                 goto out_unlock;
149         }
150
151         INIT_HLIST_NODE(&inst->hlist);
152         spin_lock_init(&inst->lock);
153         /* needs to be two, since we _put() after creation */
154         atomic_set(&inst->use, 2);
155
156         setup_timer(&inst->timer, nfulnl_timer, (unsigned long)inst);
157
158         inst->peer_pid = pid;
159         inst->group_num = group_num;
160
161         inst->qthreshold        = NFULNL_QTHRESH_DEFAULT;
162         inst->flushtimeout      = NFULNL_TIMEOUT_DEFAULT;
163         inst->nlbufsiz          = NFULNL_NLBUFSIZ_DEFAULT;
164         inst->copy_mode         = NFULNL_COPY_PACKET;
165         inst->copy_range        = NFULNL_COPY_RANGE_MAX;
166
167         hlist_add_head(&inst->hlist,
168                        &instance_table[instance_hashfn(group_num)]);
169
170         write_unlock_bh(&instances_lock);
171
172         return inst;
173
174 out_unlock:
175         write_unlock_bh(&instances_lock);
176         return ERR_PTR(err);
177 }
178
179 static void __nfulnl_flush(struct nfulnl_instance *inst);
180
181 static void
182 __instance_destroy(struct nfulnl_instance *inst)
183 {
184         /* first pull it out of the global list */
185         hlist_del(&inst->hlist);
186
187         /* then flush all pending packets from skb */
188
189         spin_lock_bh(&inst->lock);
190         if (inst->skb)
191                 __nfulnl_flush(inst);
192         spin_unlock_bh(&inst->lock);
193
194         /* and finally put the refcount */
195         instance_put(inst);
196 }
197
198 static inline void
199 instance_destroy(struct nfulnl_instance *inst)
200 {
201         write_lock_bh(&instances_lock);
202         __instance_destroy(inst);
203         write_unlock_bh(&instances_lock);
204 }
205
206 static int
207 nfulnl_set_mode(struct nfulnl_instance *inst, u_int8_t mode,
208                   unsigned int range)
209 {
210         int status = 0;
211
212         spin_lock_bh(&inst->lock);
213
214         switch (mode) {
215         case NFULNL_COPY_NONE:
216         case NFULNL_COPY_META:
217                 inst->copy_mode = mode;
218                 inst->copy_range = 0;
219                 break;
220
221         case NFULNL_COPY_PACKET:
222                 inst->copy_mode = mode;
223                 inst->copy_range = min_t(unsigned int,
224                                          range, NFULNL_COPY_RANGE_MAX);
225                 break;
226
227         default:
228                 status = -EINVAL;
229                 break;
230         }
231
232         spin_unlock_bh(&inst->lock);
233
234         return status;
235 }
236
237 static int
238 nfulnl_set_nlbufsiz(struct nfulnl_instance *inst, u_int32_t nlbufsiz)
239 {
240         int status;
241
242         spin_lock_bh(&inst->lock);
243         if (nlbufsiz < NFULNL_NLBUFSIZ_DEFAULT)
244                 status = -ERANGE;
245         else if (nlbufsiz > 131072)
246                 status = -ERANGE;
247         else {
248                 inst->nlbufsiz = nlbufsiz;
249                 status = 0;
250         }
251         spin_unlock_bh(&inst->lock);
252
253         return status;
254 }
255
256 static int
257 nfulnl_set_timeout(struct nfulnl_instance *inst, u_int32_t timeout)
258 {
259         spin_lock_bh(&inst->lock);
260         inst->flushtimeout = timeout;
261         spin_unlock_bh(&inst->lock);
262
263         return 0;
264 }
265
266 static int
267 nfulnl_set_qthresh(struct nfulnl_instance *inst, u_int32_t qthresh)
268 {
269         spin_lock_bh(&inst->lock);
270         inst->qthreshold = qthresh;
271         spin_unlock_bh(&inst->lock);
272
273         return 0;
274 }
275
276 static int
277 nfulnl_set_flags(struct nfulnl_instance *inst, u_int16_t flags)
278 {
279         spin_lock_bh(&inst->lock);
280         inst->flags = flags;
281         spin_unlock_bh(&inst->lock);
282
283         return 0;
284 }
285
286 static struct sk_buff *
287 nfulnl_alloc_skb(unsigned int inst_size, unsigned int pkt_size)
288 {
289         struct sk_buff *skb;
290         unsigned int n;
291
292         /* alloc skb which should be big enough for a whole multipart
293          * message.  WARNING: has to be <= 128k due to slab restrictions */
294
295         n = max(inst_size, pkt_size);
296         skb = alloc_skb(n, GFP_ATOMIC);
297         if (!skb) {
298                 PRINTR("nfnetlink_log: can't alloc whole buffer (%u bytes)\n",
299                         inst_size);
300
301                 if (n > pkt_size) {
302                         /* try to allocate only as much as we need for current
303                          * packet */
304
305                         skb = alloc_skb(pkt_size, GFP_ATOMIC);
306                         if (!skb)
307                                 PRINTR("nfnetlink_log: can't even alloc %u "
308                                        "bytes\n", pkt_size);
309                 }
310         }
311
312         return skb;
313 }
314
315 static int
316 __nfulnl_send(struct nfulnl_instance *inst)
317 {
318         int status = -1;
319
320         if (inst->qlen > 1)
321                 NLMSG_PUT(inst->skb, 0, 0,
322                           NLMSG_DONE,
323                           sizeof(struct nfgenmsg));
324
325         status = nfnetlink_unicast(inst->skb, inst->peer_pid, MSG_DONTWAIT);
326
327         inst->qlen = 0;
328         inst->skb = NULL;
329
330 nlmsg_failure:
331         return status;
332 }
333
334 static void
335 __nfulnl_flush(struct nfulnl_instance *inst)
336 {
337         /* timer holds a reference */
338         if (del_timer(&inst->timer))
339                 instance_put(inst);
340         if (inst->skb)
341                 __nfulnl_send(inst);
342 }
343
344 static void
345 nfulnl_timer(unsigned long data)
346 {
347         struct nfulnl_instance *inst = (struct nfulnl_instance *)data;
348
349         spin_lock_bh(&inst->lock);
350         if (inst->skb)
351                 __nfulnl_send(inst);
352         spin_unlock_bh(&inst->lock);
353         instance_put(inst);
354 }
355
356 /* This is an inline function, we don't really care about a long
357  * list of arguments */
358 static inline int
359 __build_packet_message(struct nfulnl_instance *inst,
360                         const struct sk_buff *skb,
361                         unsigned int data_len,
362                         u_int8_t pf,
363                         unsigned int hooknum,
364                         const struct net_device *indev,
365                         const struct net_device *outdev,
366                         const struct nf_loginfo *li,
367                         const char *prefix, unsigned int plen)
368 {
369         struct nfulnl_msg_packet_hdr pmsg;
370         struct nlmsghdr *nlh;
371         struct nfgenmsg *nfmsg;
372         __be32 tmp_uint;
373         sk_buff_data_t old_tail = inst->skb->tail;
374
375         nlh = NLMSG_PUT(inst->skb, 0, 0,
376                         NFNL_SUBSYS_ULOG << 8 | NFULNL_MSG_PACKET,
377                         sizeof(struct nfgenmsg));
378         nfmsg = NLMSG_DATA(nlh);
379         nfmsg->nfgen_family = pf;
380         nfmsg->version = NFNETLINK_V0;
381         nfmsg->res_id = htons(inst->group_num);
382
383         pmsg.hw_protocol        = skb->protocol;
384         pmsg.hook               = hooknum;
385
386         NLA_PUT(inst->skb, NFULA_PACKET_HDR, sizeof(pmsg), &pmsg);
387
388         if (prefix)
389                 NLA_PUT(inst->skb, NFULA_PREFIX, plen, prefix);
390
391         if (indev) {
392 #ifndef CONFIG_BRIDGE_NETFILTER
393                 NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_INDEV,
394                              htonl(indev->ifindex));
395 #else
396                 if (pf == PF_BRIDGE) {
397                         /* Case 1: outdev is physical input device, we need to
398                          * look for bridge group (when called from
399                          * netfilter_bridge) */
400                         NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
401                                      htonl(indev->ifindex));
402                         /* this is the bridge group "brX" */
403                         NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_INDEV,
404                                      htonl(indev->br_port->br->dev->ifindex));
405                 } else {
406                         /* Case 2: indev is bridge group, we need to look for
407                          * physical device (when called from ipv4) */
408                         NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_INDEV,
409                                      htonl(indev->ifindex));
410                         if (skb->nf_bridge && skb->nf_bridge->physindev)
411                                 NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
412                                              htonl(skb->nf_bridge->physindev->ifindex));
413                 }
414 #endif
415         }
416
417         if (outdev) {
418                 tmp_uint = htonl(outdev->ifindex);
419 #ifndef CONFIG_BRIDGE_NETFILTER
420                 NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_OUTDEV,
421                              htonl(outdev->ifindex));
422 #else
423                 if (pf == PF_BRIDGE) {
424                         /* Case 1: outdev is physical output device, we need to
425                          * look for bridge group (when called from
426                          * netfilter_bridge) */
427                         NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
428                                      htonl(outdev->ifindex));
429                         /* this is the bridge group "brX" */
430                         NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_OUTDEV,
431                                      htonl(outdev->br_port->br->dev->ifindex));
432                 } else {
433                         /* Case 2: indev is a bridge group, we need to look
434                          * for physical device (when called from ipv4) */
435                         NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_OUTDEV,
436                                      htonl(outdev->ifindex));
437                         if (skb->nf_bridge && skb->nf_bridge->physoutdev)
438                                 NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
439                                              htonl(skb->nf_bridge->physoutdev->ifindex));
440                 }
441 #endif
442         }
443
444         if (skb->mark)
445                 NLA_PUT_BE32(inst->skb, NFULA_MARK, htonl(skb->mark));
446
447         if (indev && skb->dev) {
448                 struct nfulnl_msg_packet_hw phw;
449                 int len = dev_parse_header(skb, phw.hw_addr);
450                 if (len > 0) {
451                         phw.hw_addrlen = htons(len);
452                         NLA_PUT(inst->skb, NFULA_HWADDR, sizeof(phw), &phw);
453                 }
454         }
455
456         if (indev && skb_mac_header_was_set(skb)) {
457                 NLA_PUT_BE16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type));
458                 NLA_PUT_BE16(inst->skb, NFULA_HWLEN,
459                              htons(skb->dev->hard_header_len));
460                 NLA_PUT(inst->skb, NFULA_HWHEADER, skb->dev->hard_header_len,
461                         skb_mac_header(skb));
462         }
463
464         if (skb->tstamp.tv64) {
465                 struct nfulnl_msg_packet_timestamp ts;
466                 struct timeval tv = ktime_to_timeval(skb->tstamp);
467                 ts.sec = cpu_to_be64(tv.tv_sec);
468                 ts.usec = cpu_to_be64(tv.tv_usec);
469
470                 NLA_PUT(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts);
471         }
472
473         /* UID */
474         if (skb->sk) {
475                 read_lock_bh(&skb->sk->sk_callback_lock);
476                 if (skb->sk->sk_socket && skb->sk->sk_socket->file) {
477                         __be32 uid = htonl(skb->sk->sk_socket->file->f_uid);
478                         __be32 gid = htonl(skb->sk->sk_socket->file->f_gid);
479                         /* need to unlock here since NLA_PUT may goto */
480                         read_unlock_bh(&skb->sk->sk_callback_lock);
481                         NLA_PUT_BE32(inst->skb, NFULA_UID, uid);
482                         NLA_PUT_BE32(inst->skb, NFULA_GID, gid);
483                 } else
484                         read_unlock_bh(&skb->sk->sk_callback_lock);
485         }
486
487         /* local sequence number */
488         if (inst->flags & NFULNL_CFG_F_SEQ)
489                 NLA_PUT_BE32(inst->skb, NFULA_SEQ, htonl(inst->seq++));
490
491         /* global sequence number */
492         if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
493                 NLA_PUT_BE32(inst->skb, NFULA_SEQ_GLOBAL,
494                              htonl(atomic_inc_return(&global_seq)));
495
496         if (data_len) {
497                 struct nlattr *nla;
498                 int size = nla_attr_size(data_len);
499
500                 if (skb_tailroom(inst->skb) < nla_total_size(data_len)) {
501                         printk(KERN_WARNING "nfnetlink_log: no tailroom!\n");
502                         goto nlmsg_failure;
503                 }
504
505                 nla = (struct nlattr *)skb_put(inst->skb, nla_total_size(data_len));
506                 nla->nla_type = NFULA_PAYLOAD;
507                 nla->nla_len = size;
508
509                 if (skb_copy_bits(skb, 0, nla_data(nla), data_len))
510                         BUG();
511         }
512
513         nlh->nlmsg_len = inst->skb->tail - old_tail;
514         return 0;
515
516 nlmsg_failure:
517 nla_put_failure:
518         PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n");
519         return -1;
520 }
521
522 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
523
524 static struct nf_loginfo default_loginfo = {
525         .type =         NF_LOG_TYPE_ULOG,
526         .u = {
527                 .ulog = {
528                         .copy_len       = 0xffff,
529                         .group          = 0,
530                         .qthreshold     = 1,
531                 },
532         },
533 };
534
535 /* log handler for internal netfilter logging api */
536 static void
537 nfulnl_log_packet(u_int8_t pf,
538                   unsigned int hooknum,
539                   const struct sk_buff *skb,
540                   const struct net_device *in,
541                   const struct net_device *out,
542                   const struct nf_loginfo *li_user,
543                   const char *prefix)
544 {
545         unsigned int size, data_len;
546         struct nfulnl_instance *inst;
547         const struct nf_loginfo *li;
548         unsigned int qthreshold;
549         unsigned int plen;
550
551         if (li_user && li_user->type == NF_LOG_TYPE_ULOG)
552                 li = li_user;
553         else
554                 li = &default_loginfo;
555
556         inst = instance_lookup_get(li->u.ulog.group);
557         if (!inst)
558                 return;
559
560         plen = 0;
561         if (prefix)
562                 plen = strlen(prefix) + 1;
563
564         /* FIXME: do we want to make the size calculation conditional based on
565          * what is actually present?  way more branches and checks, but more
566          * memory efficient... */
567         size =    NLMSG_SPACE(sizeof(struct nfgenmsg))
568                 + nla_total_size(sizeof(struct nfulnl_msg_packet_hdr))
569                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
570                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
571 #ifdef CONFIG_BRIDGE_NETFILTER
572                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
573                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
574 #endif
575                 + nla_total_size(sizeof(u_int32_t))     /* mark */
576                 + nla_total_size(sizeof(u_int32_t))     /* uid */
577                 + nla_total_size(sizeof(u_int32_t))     /* gid */
578                 + nla_total_size(plen)                  /* prefix */
579                 + nla_total_size(sizeof(struct nfulnl_msg_packet_hw))
580                 + nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp));
581
582         spin_lock_bh(&inst->lock);
583
584         if (inst->flags & NFULNL_CFG_F_SEQ)
585                 size += nla_total_size(sizeof(u_int32_t));
586         if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
587                 size += nla_total_size(sizeof(u_int32_t));
588
589         qthreshold = inst->qthreshold;
590         /* per-rule qthreshold overrides per-instance */
591         if (qthreshold > li->u.ulog.qthreshold)
592                 qthreshold = li->u.ulog.qthreshold;
593
594         switch (inst->copy_mode) {
595         case NFULNL_COPY_META:
596         case NFULNL_COPY_NONE:
597                 data_len = 0;
598                 break;
599
600         case NFULNL_COPY_PACKET:
601                 if (inst->copy_range == 0
602                     || inst->copy_range > skb->len)
603                         data_len = skb->len;
604                 else
605                         data_len = inst->copy_range;
606
607                 size += nla_total_size(data_len);
608                 break;
609
610         default:
611                 goto unlock_and_release;
612         }
613
614         if (inst->skb &&
615             size > skb_tailroom(inst->skb) - sizeof(struct nfgenmsg)) {
616                 /* either the queue len is too high or we don't have
617                  * enough room in the skb left. flush to userspace. */
618                 __nfulnl_flush(inst);
619         }
620
621         if (!inst->skb) {
622                 inst->skb = nfulnl_alloc_skb(inst->nlbufsiz, size);
623                 if (!inst->skb)
624                         goto alloc_failure;
625         }
626
627         inst->qlen++;
628
629         __build_packet_message(inst, skb, data_len, pf,
630                                 hooknum, in, out, li, prefix, plen);
631
632         if (inst->qlen >= qthreshold)
633                 __nfulnl_flush(inst);
634         /* timer_pending always called within inst->lock, so there
635          * is no chance of a race here */
636         else if (!timer_pending(&inst->timer)) {
637                 instance_get(inst);
638                 inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100);
639                 add_timer(&inst->timer);
640         }
641
642 unlock_and_release:
643         spin_unlock_bh(&inst->lock);
644         instance_put(inst);
645         return;
646
647 alloc_failure:
648         /* FIXME: statistics */
649         goto unlock_and_release;
650 }
651
652 static int
653 nfulnl_rcv_nl_event(struct notifier_block *this,
654                    unsigned long event, void *ptr)
655 {
656         struct netlink_notify *n = ptr;
657
658         if (event == NETLINK_URELEASE &&
659             n->protocol == NETLINK_NETFILTER && n->pid) {
660                 int i;
661
662                 /* destroy all instances for this pid */
663                 write_lock_bh(&instances_lock);
664                 for  (i = 0; i < INSTANCE_BUCKETS; i++) {
665                         struct hlist_node *tmp, *t2;
666                         struct nfulnl_instance *inst;
667                         struct hlist_head *head = &instance_table[i];
668
669                         hlist_for_each_entry_safe(inst, tmp, t2, head, hlist) {
670                                 if ((n->net == &init_net) &&
671                                     (n->pid == inst->peer_pid))
672                                         __instance_destroy(inst);
673                         }
674                 }
675                 write_unlock_bh(&instances_lock);
676         }
677         return NOTIFY_DONE;
678 }
679
680 static struct notifier_block nfulnl_rtnl_notifier = {
681         .notifier_call  = nfulnl_rcv_nl_event,
682 };
683
684 static int
685 nfulnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
686                   struct nlmsghdr *nlh, struct nlattr *nfqa[])
687 {
688         return -ENOTSUPP;
689 }
690
691 static const struct nf_logger nfulnl_logger = {
692         .name   = "nfnetlink_log",
693         .logfn  = &nfulnl_log_packet,
694         .me     = THIS_MODULE,
695 };
696
697 static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = {
698         [NFULA_CFG_CMD]         = { .len = sizeof(struct nfulnl_msg_config_cmd) },
699         [NFULA_CFG_MODE]        = { .len = sizeof(struct nfulnl_msg_config_mode) },
700         [NFULA_CFG_TIMEOUT]     = { .type = NLA_U32 },
701         [NFULA_CFG_QTHRESH]     = { .type = NLA_U32 },
702         [NFULA_CFG_NLBUFSIZ]    = { .type = NLA_U32 },
703         [NFULA_CFG_FLAGS]       = { .type = NLA_U16 },
704 };
705
706 static int
707 nfulnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
708                    struct nlmsghdr *nlh, struct nlattr *nfula[])
709 {
710         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
711         u_int16_t group_num = ntohs(nfmsg->res_id);
712         struct nfulnl_instance *inst;
713         struct nfulnl_msg_config_cmd *cmd = NULL;
714         int ret = 0;
715
716         if (nfula[NFULA_CFG_CMD]) {
717                 u_int8_t pf = nfmsg->nfgen_family;
718                 cmd = nla_data(nfula[NFULA_CFG_CMD]);
719
720                 /* Commands without queue context */
721                 switch (cmd->command) {
722                 case NFULNL_CFG_CMD_PF_BIND:
723                         return nf_log_register(pf, &nfulnl_logger);
724                 case NFULNL_CFG_CMD_PF_UNBIND:
725                         nf_log_unregister_pf(pf);
726                         return 0;
727                 }
728         }
729
730         inst = instance_lookup_get(group_num);
731         if (inst && inst->peer_pid != NETLINK_CB(skb).pid) {
732                 ret = -EPERM;
733                 goto out_put;
734         }
735
736         if (cmd != NULL) {
737                 switch (cmd->command) {
738                 case NFULNL_CFG_CMD_BIND:
739                         if (inst) {
740                                 ret = -EBUSY;
741                                 goto out_put;
742                         }
743
744                         inst = instance_create(group_num,
745                                                NETLINK_CB(skb).pid);
746                         if (IS_ERR(inst)) {
747                                 ret = PTR_ERR(inst);
748                                 goto out;
749                         }
750                         break;
751                 case NFULNL_CFG_CMD_UNBIND:
752                         if (!inst) {
753                                 ret = -ENODEV;
754                                 goto out;
755                         }
756
757                         instance_destroy(inst);
758                         goto out;
759                 default:
760                         ret = -ENOTSUPP;
761                         break;
762                 }
763         }
764
765         if (nfula[NFULA_CFG_MODE]) {
766                 struct nfulnl_msg_config_mode *params;
767                 params = nla_data(nfula[NFULA_CFG_MODE]);
768
769                 if (!inst) {
770                         ret = -ENODEV;
771                         goto out;
772                 }
773                 nfulnl_set_mode(inst, params->copy_mode,
774                                 ntohl(params->copy_range));
775         }
776
777         if (nfula[NFULA_CFG_TIMEOUT]) {
778                 __be32 timeout = nla_get_be32(nfula[NFULA_CFG_TIMEOUT]);
779
780                 if (!inst) {
781                         ret = -ENODEV;
782                         goto out;
783                 }
784                 nfulnl_set_timeout(inst, ntohl(timeout));
785         }
786
787         if (nfula[NFULA_CFG_NLBUFSIZ]) {
788                 __be32 nlbufsiz = nla_get_be32(nfula[NFULA_CFG_NLBUFSIZ]);
789
790                 if (!inst) {
791                         ret = -ENODEV;
792                         goto out;
793                 }
794                 nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz));
795         }
796
797         if (nfula[NFULA_CFG_QTHRESH]) {
798                 __be32 qthresh = nla_get_be32(nfula[NFULA_CFG_QTHRESH]);
799
800                 if (!inst) {
801                         ret = -ENODEV;
802                         goto out;
803                 }
804                 nfulnl_set_qthresh(inst, ntohl(qthresh));
805         }
806
807         if (nfula[NFULA_CFG_FLAGS]) {
808                 __be16 flags = nla_get_be16(nfula[NFULA_CFG_FLAGS]);
809
810                 if (!inst) {
811                         ret = -ENODEV;
812                         goto out;
813                 }
814                 nfulnl_set_flags(inst, ntohs(flags));
815         }
816
817 out_put:
818         instance_put(inst);
819 out:
820         return ret;
821 }
822
823 static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = {
824         [NFULNL_MSG_PACKET]     = { .call = nfulnl_recv_unsupp,
825                                     .attr_count = NFULA_MAX, },
826         [NFULNL_MSG_CONFIG]     = { .call = nfulnl_recv_config,
827                                     .attr_count = NFULA_CFG_MAX,
828                                     .policy = nfula_cfg_policy },
829 };
830
831 static const struct nfnetlink_subsystem nfulnl_subsys = {
832         .name           = "log",
833         .subsys_id      = NFNL_SUBSYS_ULOG,
834         .cb_count       = NFULNL_MSG_MAX,
835         .cb             = nfulnl_cb,
836 };
837
838 #ifdef CONFIG_PROC_FS
839 struct iter_state {
840         unsigned int bucket;
841 };
842
843 static struct hlist_node *get_first(struct iter_state *st)
844 {
845         if (!st)
846                 return NULL;
847
848         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
849                 if (!hlist_empty(&instance_table[st->bucket]))
850                         return instance_table[st->bucket].first;
851         }
852         return NULL;
853 }
854
855 static struct hlist_node *get_next(struct iter_state *st, struct hlist_node *h)
856 {
857         h = h->next;
858         while (!h) {
859                 if (++st->bucket >= INSTANCE_BUCKETS)
860                         return NULL;
861
862                 h = instance_table[st->bucket].first;
863         }
864         return h;
865 }
866
867 static struct hlist_node *get_idx(struct iter_state *st, loff_t pos)
868 {
869         struct hlist_node *head;
870         head = get_first(st);
871
872         if (head)
873                 while (pos && (head = get_next(st, head)))
874                         pos--;
875         return pos ? NULL : head;
876 }
877
878 static void *seq_start(struct seq_file *seq, loff_t *pos)
879         __acquires(instances_lock)
880 {
881         read_lock_bh(&instances_lock);
882         return get_idx(seq->private, *pos);
883 }
884
885 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
886 {
887         (*pos)++;
888         return get_next(s->private, v);
889 }
890
891 static void seq_stop(struct seq_file *s, void *v)
892         __releases(instances_lock)
893 {
894         read_unlock_bh(&instances_lock);
895 }
896
897 static int seq_show(struct seq_file *s, void *v)
898 {
899         const struct nfulnl_instance *inst = v;
900
901         return seq_printf(s, "%5d %6d %5d %1d %5d %6d %2d\n",
902                           inst->group_num,
903                           inst->peer_pid, inst->qlen,
904                           inst->copy_mode, inst->copy_range,
905                           inst->flushtimeout, atomic_read(&inst->use));
906 }
907
908 static const struct seq_operations nful_seq_ops = {
909         .start  = seq_start,
910         .next   = seq_next,
911         .stop   = seq_stop,
912         .show   = seq_show,
913 };
914
915 static int nful_open(struct inode *inode, struct file *file)
916 {
917         return seq_open_private(file, &nful_seq_ops,
918                         sizeof(struct iter_state));
919 }
920
921 static const struct file_operations nful_file_ops = {
922         .owner   = THIS_MODULE,
923         .open    = nful_open,
924         .read    = seq_read,
925         .llseek  = seq_lseek,
926         .release = seq_release_private,
927 };
928
929 #endif /* PROC_FS */
930
931 static int __init nfnetlink_log_init(void)
932 {
933         int i, status = -ENOMEM;
934
935         for (i = 0; i < INSTANCE_BUCKETS; i++)
936                 INIT_HLIST_HEAD(&instance_table[i]);
937
938         /* it's not really all that important to have a random value, so
939          * we can do this from the init function, even if there hasn't
940          * been that much entropy yet */
941         get_random_bytes(&hash_init, sizeof(hash_init));
942
943         netlink_register_notifier(&nfulnl_rtnl_notifier);
944         status = nfnetlink_subsys_register(&nfulnl_subsys);
945         if (status < 0) {
946                 printk(KERN_ERR "log: failed to create netlink socket\n");
947                 goto cleanup_netlink_notifier;
948         }
949
950 #ifdef CONFIG_PROC_FS
951         if (!proc_create("nfnetlink_log", 0440,
952                          proc_net_netfilter, &nful_file_ops))
953                 goto cleanup_subsys;
954 #endif
955         return status;
956
957 #ifdef CONFIG_PROC_FS
958 cleanup_subsys:
959         nfnetlink_subsys_unregister(&nfulnl_subsys);
960 #endif
961 cleanup_netlink_notifier:
962         netlink_unregister_notifier(&nfulnl_rtnl_notifier);
963         return status;
964 }
965
966 static void __exit nfnetlink_log_fini(void)
967 {
968         nf_log_unregister(&nfulnl_logger);
969 #ifdef CONFIG_PROC_FS
970         remove_proc_entry("nfnetlink_log", proc_net_netfilter);
971 #endif
972         nfnetlink_subsys_unregister(&nfulnl_subsys);
973         netlink_unregister_notifier(&nfulnl_rtnl_notifier);
974 }
975
976 MODULE_DESCRIPTION("netfilter userspace logging");
977 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
978 MODULE_LICENSE("GPL");
979 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG);
980
981 module_init(nfnetlink_log_init);
982 module_exit(nfnetlink_log_fini);