Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/kaber/nf-next-2.6
[linux-2.6.git] / 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 #include <net/netfilter/nfnetlink_log.h>
34
35 #include <asm/atomic.h>
36
37 #ifdef CONFIG_BRIDGE_NETFILTER
38 #include "../bridge/br_private.h"
39 #endif
40
41 #define NFULNL_NLBUFSIZ_DEFAULT NLMSG_GOODSIZE
42 #define NFULNL_TIMEOUT_DEFAULT  100     /* every second */
43 #define NFULNL_QTHRESH_DEFAULT  100     /* 100 packets */
44 #define NFULNL_COPY_RANGE_MAX   0xFFFF  /* max packet size is limited by 16-bit struct nfattr nfa_len field */
45
46 #define PRINTR(x, args...)      do { if (net_ratelimit()) \
47                                      printk(x, ## args); } while (0);
48
49 struct nfulnl_instance {
50         struct hlist_node hlist;        /* global list of instances */
51         spinlock_t lock;
52         atomic_t use;                   /* use count */
53
54         unsigned int qlen;              /* number of nlmsgs in skb */
55         struct sk_buff *skb;            /* pre-allocatd skb */
56         struct timer_list timer;
57         int peer_pid;                   /* PID of the peer process */
58
59         /* configurable parameters */
60         unsigned int flushtimeout;      /* timeout until queue flush */
61         unsigned int nlbufsiz;          /* netlink buffer allocation size */
62         unsigned int qthreshold;        /* threshold of the queue */
63         u_int32_t copy_range;
64         u_int32_t seq;                  /* instance-local sequential counter */
65         u_int16_t group_num;            /* number of this queue */
66         u_int16_t flags;
67         u_int8_t copy_mode;
68 };
69
70 static DEFINE_RWLOCK(instances_lock);
71 static atomic_t global_seq;
72
73 #define INSTANCE_BUCKETS        16
74 static struct hlist_head instance_table[INSTANCE_BUCKETS];
75 static unsigned int hash_init;
76
77 static inline u_int8_t instance_hashfn(u_int16_t group_num)
78 {
79         return ((group_num & 0xff) % INSTANCE_BUCKETS);
80 }
81
82 static struct nfulnl_instance *
83 __instance_lookup(u_int16_t group_num)
84 {
85         struct hlist_head *head;
86         struct hlist_node *pos;
87         struct nfulnl_instance *inst;
88
89         head = &instance_table[instance_hashfn(group_num)];
90         hlist_for_each_entry(inst, pos, head, hlist) {
91                 if (inst->group_num == group_num)
92                         return inst;
93         }
94         return NULL;
95 }
96
97 static inline void
98 instance_get(struct nfulnl_instance *inst)
99 {
100         atomic_inc(&inst->use);
101 }
102
103 static struct nfulnl_instance *
104 instance_lookup_get(u_int16_t group_num)
105 {
106         struct nfulnl_instance *inst;
107
108         read_lock_bh(&instances_lock);
109         inst = __instance_lookup(group_num);
110         if (inst)
111                 instance_get(inst);
112         read_unlock_bh(&instances_lock);
113
114         return inst;
115 }
116
117 static void
118 instance_put(struct nfulnl_instance *inst)
119 {
120         if (inst && atomic_dec_and_test(&inst->use)) {
121                 kfree(inst);
122                 module_put(THIS_MODULE);
123         }
124 }
125
126 static void nfulnl_timer(unsigned long data);
127
128 static struct nfulnl_instance *
129 instance_create(u_int16_t group_num, int pid)
130 {
131         struct nfulnl_instance *inst;
132         int err;
133
134         write_lock_bh(&instances_lock);
135         if (__instance_lookup(group_num)) {
136                 err = -EEXIST;
137                 goto out_unlock;
138         }
139
140         inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
141         if (!inst) {
142                 err = -ENOMEM;
143                 goto out_unlock;
144         }
145
146         if (!try_module_get(THIS_MODULE)) {
147                 kfree(inst);
148                 err = -EAGAIN;
149                 goto out_unlock;
150         }
151
152         INIT_HLIST_NODE(&inst->hlist);
153         spin_lock_init(&inst->lock);
154         /* needs to be two, since we _put() after creation */
155         atomic_set(&inst->use, 2);
156
157         setup_timer(&inst->timer, nfulnl_timer, (unsigned long)inst);
158
159         inst->peer_pid = pid;
160         inst->group_num = group_num;
161
162         inst->qthreshold        = NFULNL_QTHRESH_DEFAULT;
163         inst->flushtimeout      = NFULNL_TIMEOUT_DEFAULT;
164         inst->nlbufsiz          = NFULNL_NLBUFSIZ_DEFAULT;
165         inst->copy_mode         = NFULNL_COPY_PACKET;
166         inst->copy_range        = NFULNL_COPY_RANGE_MAX;
167
168         hlist_add_head(&inst->hlist,
169                        &instance_table[instance_hashfn(group_num)]);
170
171         write_unlock_bh(&instances_lock);
172
173         return inst;
174
175 out_unlock:
176         write_unlock_bh(&instances_lock);
177         return ERR_PTR(err);
178 }
179
180 static void __nfulnl_flush(struct nfulnl_instance *inst);
181
182 static void
183 __instance_destroy(struct nfulnl_instance *inst)
184 {
185         /* first pull it out of the global list */
186         hlist_del(&inst->hlist);
187
188         /* then flush all pending packets from skb */
189
190         spin_lock_bh(&inst->lock);
191         if (inst->skb)
192                 __nfulnl_flush(inst);
193         spin_unlock_bh(&inst->lock);
194
195         /* and finally put the refcount */
196         instance_put(inst);
197 }
198
199 static inline void
200 instance_destroy(struct nfulnl_instance *inst)
201 {
202         write_lock_bh(&instances_lock);
203         __instance_destroy(inst);
204         write_unlock_bh(&instances_lock);
205 }
206
207 static int
208 nfulnl_set_mode(struct nfulnl_instance *inst, u_int8_t mode,
209                   unsigned int range)
210 {
211         int status = 0;
212
213         spin_lock_bh(&inst->lock);
214
215         switch (mode) {
216         case NFULNL_COPY_NONE:
217         case NFULNL_COPY_META:
218                 inst->copy_mode = mode;
219                 inst->copy_range = 0;
220                 break;
221
222         case NFULNL_COPY_PACKET:
223                 inst->copy_mode = mode;
224                 inst->copy_range = min_t(unsigned int,
225                                          range, NFULNL_COPY_RANGE_MAX);
226                 break;
227
228         default:
229                 status = -EINVAL;
230                 break;
231         }
232
233         spin_unlock_bh(&inst->lock);
234
235         return status;
236 }
237
238 static int
239 nfulnl_set_nlbufsiz(struct nfulnl_instance *inst, u_int32_t nlbufsiz)
240 {
241         int status;
242
243         spin_lock_bh(&inst->lock);
244         if (nlbufsiz < NFULNL_NLBUFSIZ_DEFAULT)
245                 status = -ERANGE;
246         else if (nlbufsiz > 131072)
247                 status = -ERANGE;
248         else {
249                 inst->nlbufsiz = nlbufsiz;
250                 status = 0;
251         }
252         spin_unlock_bh(&inst->lock);
253
254         return status;
255 }
256
257 static int
258 nfulnl_set_timeout(struct nfulnl_instance *inst, u_int32_t timeout)
259 {
260         spin_lock_bh(&inst->lock);
261         inst->flushtimeout = timeout;
262         spin_unlock_bh(&inst->lock);
263
264         return 0;
265 }
266
267 static int
268 nfulnl_set_qthresh(struct nfulnl_instance *inst, u_int32_t qthresh)
269 {
270         spin_lock_bh(&inst->lock);
271         inst->qthreshold = qthresh;
272         spin_unlock_bh(&inst->lock);
273
274         return 0;
275 }
276
277 static int
278 nfulnl_set_flags(struct nfulnl_instance *inst, u_int16_t flags)
279 {
280         spin_lock_bh(&inst->lock);
281         inst->flags = flags;
282         spin_unlock_bh(&inst->lock);
283
284         return 0;
285 }
286
287 static struct sk_buff *
288 nfulnl_alloc_skb(unsigned int inst_size, unsigned int pkt_size)
289 {
290         struct sk_buff *skb;
291         unsigned int n;
292
293         /* alloc skb which should be big enough for a whole multipart
294          * message.  WARNING: has to be <= 128k due to slab restrictions */
295
296         n = max(inst_size, pkt_size);
297         skb = alloc_skb(n, GFP_ATOMIC);
298         if (!skb) {
299                 PRINTR("nfnetlink_log: can't alloc whole buffer (%u bytes)\n",
300                         inst_size);
301
302                 if (n > pkt_size) {
303                         /* try to allocate only as much as we need for current
304                          * packet */
305
306                         skb = alloc_skb(pkt_size, GFP_ATOMIC);
307                         if (!skb)
308                                 PRINTR("nfnetlink_log: can't even alloc %u "
309                                        "bytes\n", pkt_size);
310                 }
311         }
312
313         return skb;
314 }
315
316 static int
317 __nfulnl_send(struct nfulnl_instance *inst)
318 {
319         int status = -1;
320
321         if (inst->qlen > 1)
322                 NLMSG_PUT(inst->skb, 0, 0,
323                           NLMSG_DONE,
324                           sizeof(struct nfgenmsg));
325
326         status = nfnetlink_unicast(inst->skb, inst->peer_pid, MSG_DONTWAIT);
327
328         inst->qlen = 0;
329         inst->skb = NULL;
330
331 nlmsg_failure:
332         return status;
333 }
334
335 static void
336 __nfulnl_flush(struct nfulnl_instance *inst)
337 {
338         /* timer holds a reference */
339         if (del_timer(&inst->timer))
340                 instance_put(inst);
341         if (inst->skb)
342                 __nfulnl_send(inst);
343 }
344
345 static void
346 nfulnl_timer(unsigned long data)
347 {
348         struct nfulnl_instance *inst = (struct nfulnl_instance *)data;
349
350         spin_lock_bh(&inst->lock);
351         if (inst->skb)
352                 __nfulnl_send(inst);
353         spin_unlock_bh(&inst->lock);
354         instance_put(inst);
355 }
356
357 /* This is an inline function, we don't really care about a long
358  * list of arguments */
359 static inline int
360 __build_packet_message(struct nfulnl_instance *inst,
361                         const struct sk_buff *skb,
362                         unsigned int data_len,
363                         u_int8_t pf,
364                         unsigned int hooknum,
365                         const struct net_device *indev,
366                         const struct net_device *outdev,
367                         const struct nf_loginfo *li,
368                         const char *prefix, unsigned int plen)
369 {
370         struct nfulnl_msg_packet_hdr pmsg;
371         struct nlmsghdr *nlh;
372         struct nfgenmsg *nfmsg;
373         __be32 tmp_uint;
374         sk_buff_data_t old_tail = inst->skb->tail;
375
376         nlh = NLMSG_PUT(inst->skb, 0, 0,
377                         NFNL_SUBSYS_ULOG << 8 | NFULNL_MSG_PACKET,
378                         sizeof(struct nfgenmsg));
379         nfmsg = NLMSG_DATA(nlh);
380         nfmsg->nfgen_family = pf;
381         nfmsg->version = NFNETLINK_V0;
382         nfmsg->res_id = htons(inst->group_num);
383
384         pmsg.hw_protocol        = skb->protocol;
385         pmsg.hook               = hooknum;
386
387         NLA_PUT(inst->skb, NFULA_PACKET_HDR, sizeof(pmsg), &pmsg);
388
389         if (prefix)
390                 NLA_PUT(inst->skb, NFULA_PREFIX, plen, prefix);
391
392         if (indev) {
393 #ifndef CONFIG_BRIDGE_NETFILTER
394                 NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_INDEV,
395                              htonl(indev->ifindex));
396 #else
397                 if (pf == PF_BRIDGE) {
398                         /* Case 1: outdev is physical input device, we need to
399                          * look for bridge group (when called from
400                          * netfilter_bridge) */
401                         NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
402                                      htonl(indev->ifindex));
403                         /* this is the bridge group "brX" */
404                         NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_INDEV,
405                                      htonl(indev->br_port->br->dev->ifindex));
406                 } else {
407                         /* Case 2: indev is bridge group, we need to look for
408                          * physical device (when called from ipv4) */
409                         NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_INDEV,
410                                      htonl(indev->ifindex));
411                         if (skb->nf_bridge && skb->nf_bridge->physindev)
412                                 NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
413                                              htonl(skb->nf_bridge->physindev->ifindex));
414                 }
415 #endif
416         }
417
418         if (outdev) {
419                 tmp_uint = htonl(outdev->ifindex);
420 #ifndef CONFIG_BRIDGE_NETFILTER
421                 NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_OUTDEV,
422                              htonl(outdev->ifindex));
423 #else
424                 if (pf == PF_BRIDGE) {
425                         /* Case 1: outdev is physical output device, we need to
426                          * look for bridge group (when called from
427                          * netfilter_bridge) */
428                         NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
429                                      htonl(outdev->ifindex));
430                         /* this is the bridge group "brX" */
431                         NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_OUTDEV,
432                                      htonl(outdev->br_port->br->dev->ifindex));
433                 } else {
434                         /* Case 2: indev is a bridge group, we need to look
435                          * for physical device (when called from ipv4) */
436                         NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_OUTDEV,
437                                      htonl(outdev->ifindex));
438                         if (skb->nf_bridge && skb->nf_bridge->physoutdev)
439                                 NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
440                                              htonl(skb->nf_bridge->physoutdev->ifindex));
441                 }
442 #endif
443         }
444
445         if (skb->mark)
446                 NLA_PUT_BE32(inst->skb, NFULA_MARK, htonl(skb->mark));
447
448         if (indev && skb->dev) {
449                 struct nfulnl_msg_packet_hw phw;
450                 int len = dev_parse_header(skb, phw.hw_addr);
451                 if (len > 0) {
452                         phw.hw_addrlen = htons(len);
453                         NLA_PUT(inst->skb, NFULA_HWADDR, sizeof(phw), &phw);
454                 }
455         }
456
457         if (indev && skb_mac_header_was_set(skb)) {
458                 NLA_PUT_BE16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type));
459                 NLA_PUT_BE16(inst->skb, NFULA_HWLEN,
460                              htons(skb->dev->hard_header_len));
461                 NLA_PUT(inst->skb, NFULA_HWHEADER, skb->dev->hard_header_len,
462                         skb_mac_header(skb));
463         }
464
465         if (skb->tstamp.tv64) {
466                 struct nfulnl_msg_packet_timestamp ts;
467                 struct timeval tv = ktime_to_timeval(skb->tstamp);
468                 ts.sec = cpu_to_be64(tv.tv_sec);
469                 ts.usec = cpu_to_be64(tv.tv_usec);
470
471                 NLA_PUT(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts);
472         }
473
474         /* UID */
475         if (skb->sk) {
476                 read_lock_bh(&skb->sk->sk_callback_lock);
477                 if (skb->sk->sk_socket && skb->sk->sk_socket->file) {
478                         struct file *file = skb->sk->sk_socket->file;
479                         __be32 uid = htonl(file->f_cred->fsuid);
480                         __be32 gid = htonl(file->f_cred->fsgid);
481                         /* need to unlock here since NLA_PUT may goto */
482                         read_unlock_bh(&skb->sk->sk_callback_lock);
483                         NLA_PUT_BE32(inst->skb, NFULA_UID, uid);
484                         NLA_PUT_BE32(inst->skb, NFULA_GID, gid);
485                 } else
486                         read_unlock_bh(&skb->sk->sk_callback_lock);
487         }
488
489         /* local sequence number */
490         if (inst->flags & NFULNL_CFG_F_SEQ)
491                 NLA_PUT_BE32(inst->skb, NFULA_SEQ, htonl(inst->seq++));
492
493         /* global sequence number */
494         if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
495                 NLA_PUT_BE32(inst->skb, NFULA_SEQ_GLOBAL,
496                              htonl(atomic_inc_return(&global_seq)));
497
498         if (data_len) {
499                 struct nlattr *nla;
500                 int size = nla_attr_size(data_len);
501
502                 if (skb_tailroom(inst->skb) < nla_total_size(data_len)) {
503                         printk(KERN_WARNING "nfnetlink_log: no tailroom!\n");
504                         goto nlmsg_failure;
505                 }
506
507                 nla = (struct nlattr *)skb_put(inst->skb, nla_total_size(data_len));
508                 nla->nla_type = NFULA_PAYLOAD;
509                 nla->nla_len = size;
510
511                 if (skb_copy_bits(skb, 0, nla_data(nla), data_len))
512                         BUG();
513         }
514
515         nlh->nlmsg_len = inst->skb->tail - old_tail;
516         return 0;
517
518 nlmsg_failure:
519 nla_put_failure:
520         PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n");
521         return -1;
522 }
523
524 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
525
526 static struct nf_loginfo default_loginfo = {
527         .type =         NF_LOG_TYPE_ULOG,
528         .u = {
529                 .ulog = {
530                         .copy_len       = 0xffff,
531                         .group          = 0,
532                         .qthreshold     = 1,
533                 },
534         },
535 };
536
537 /* log handler for internal netfilter logging api */
538 void
539 nfulnl_log_packet(u_int8_t pf,
540                   unsigned int hooknum,
541                   const struct sk_buff *skb,
542                   const struct net_device *in,
543                   const struct net_device *out,
544                   const struct nf_loginfo *li_user,
545                   const char *prefix)
546 {
547         unsigned int size, data_len;
548         struct nfulnl_instance *inst;
549         const struct nf_loginfo *li;
550         unsigned int qthreshold;
551         unsigned int plen;
552
553         if (li_user && li_user->type == NF_LOG_TYPE_ULOG)
554                 li = li_user;
555         else
556                 li = &default_loginfo;
557
558         inst = instance_lookup_get(li->u.ulog.group);
559         if (!inst)
560                 return;
561
562         plen = 0;
563         if (prefix)
564                 plen = strlen(prefix) + 1;
565
566         /* FIXME: do we want to make the size calculation conditional based on
567          * what is actually present?  way more branches and checks, but more
568          * memory efficient... */
569         size =    NLMSG_SPACE(sizeof(struct nfgenmsg))
570                 + nla_total_size(sizeof(struct nfulnl_msg_packet_hdr))
571                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
572                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
573 #ifdef CONFIG_BRIDGE_NETFILTER
574                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
575                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
576 #endif
577                 + nla_total_size(sizeof(u_int32_t))     /* mark */
578                 + nla_total_size(sizeof(u_int32_t))     /* uid */
579                 + nla_total_size(sizeof(u_int32_t))     /* gid */
580                 + nla_total_size(plen)                  /* prefix */
581                 + nla_total_size(sizeof(struct nfulnl_msg_packet_hw))
582                 + nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp));
583
584         if (in && skb_mac_header_was_set(skb)) {
585                 size +=   nla_total_size(skb->dev->hard_header_len)
586                         + nla_total_size(sizeof(u_int16_t))     /* hwtype */
587                         + nla_total_size(sizeof(u_int16_t));    /* hwlen */
588         }
589
590         spin_lock_bh(&inst->lock);
591
592         if (inst->flags & NFULNL_CFG_F_SEQ)
593                 size += nla_total_size(sizeof(u_int32_t));
594         if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
595                 size += nla_total_size(sizeof(u_int32_t));
596
597         qthreshold = inst->qthreshold;
598         /* per-rule qthreshold overrides per-instance */
599         if (li->u.ulog.qthreshold)
600                 if (qthreshold > li->u.ulog.qthreshold)
601                         qthreshold = li->u.ulog.qthreshold;
602
603
604         switch (inst->copy_mode) {
605         case NFULNL_COPY_META:
606         case NFULNL_COPY_NONE:
607                 data_len = 0;
608                 break;
609
610         case NFULNL_COPY_PACKET:
611                 if (inst->copy_range == 0
612                     || inst->copy_range > skb->len)
613                         data_len = skb->len;
614                 else
615                         data_len = inst->copy_range;
616
617                 size += nla_total_size(data_len);
618                 break;
619
620         default:
621                 goto unlock_and_release;
622         }
623
624         if (inst->skb &&
625             size > skb_tailroom(inst->skb) - sizeof(struct nfgenmsg)) {
626                 /* either the queue len is too high or we don't have
627                  * enough room in the skb left. flush to userspace. */
628                 __nfulnl_flush(inst);
629         }
630
631         if (!inst->skb) {
632                 inst->skb = nfulnl_alloc_skb(inst->nlbufsiz, size);
633                 if (!inst->skb)
634                         goto alloc_failure;
635         }
636
637         inst->qlen++;
638
639         __build_packet_message(inst, skb, data_len, pf,
640                                 hooknum, in, out, li, prefix, plen);
641
642         if (inst->qlen >= qthreshold)
643                 __nfulnl_flush(inst);
644         /* timer_pending always called within inst->lock, so there
645          * is no chance of a race here */
646         else if (!timer_pending(&inst->timer)) {
647                 instance_get(inst);
648                 inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100);
649                 add_timer(&inst->timer);
650         }
651
652 unlock_and_release:
653         spin_unlock_bh(&inst->lock);
654         instance_put(inst);
655         return;
656
657 alloc_failure:
658         /* FIXME: statistics */
659         goto unlock_and_release;
660 }
661 EXPORT_SYMBOL_GPL(nfulnl_log_packet);
662
663 static int
664 nfulnl_rcv_nl_event(struct notifier_block *this,
665                    unsigned long event, void *ptr)
666 {
667         struct netlink_notify *n = ptr;
668
669         if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
670                 int i;
671
672                 /* destroy all instances for this pid */
673                 write_lock_bh(&instances_lock);
674                 for  (i = 0; i < INSTANCE_BUCKETS; i++) {
675                         struct hlist_node *tmp, *t2;
676                         struct nfulnl_instance *inst;
677                         struct hlist_head *head = &instance_table[i];
678
679                         hlist_for_each_entry_safe(inst, tmp, t2, head, hlist) {
680                                 if ((net_eq(n->net, &init_net)) &&
681                                     (n->pid == inst->peer_pid))
682                                         __instance_destroy(inst);
683                         }
684                 }
685                 write_unlock_bh(&instances_lock);
686         }
687         return NOTIFY_DONE;
688 }
689
690 static struct notifier_block nfulnl_rtnl_notifier = {
691         .notifier_call  = nfulnl_rcv_nl_event,
692 };
693
694 static int
695 nfulnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
696                    const struct nlmsghdr *nlh,
697                    const struct nlattr * const nfqa[])
698 {
699         return -ENOTSUPP;
700 }
701
702 static struct nf_logger nfulnl_logger __read_mostly = {
703         .name   = "nfnetlink_log",
704         .logfn  = &nfulnl_log_packet,
705         .me     = THIS_MODULE,
706 };
707
708 static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = {
709         [NFULA_CFG_CMD]         = { .len = sizeof(struct nfulnl_msg_config_cmd) },
710         [NFULA_CFG_MODE]        = { .len = sizeof(struct nfulnl_msg_config_mode) },
711         [NFULA_CFG_TIMEOUT]     = { .type = NLA_U32 },
712         [NFULA_CFG_QTHRESH]     = { .type = NLA_U32 },
713         [NFULA_CFG_NLBUFSIZ]    = { .type = NLA_U32 },
714         [NFULA_CFG_FLAGS]       = { .type = NLA_U16 },
715 };
716
717 static int
718 nfulnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
719                    const struct nlmsghdr *nlh,
720                    const struct nlattr * const nfula[])
721 {
722         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
723         u_int16_t group_num = ntohs(nfmsg->res_id);
724         struct nfulnl_instance *inst;
725         struct nfulnl_msg_config_cmd *cmd = NULL;
726         int ret = 0;
727
728         if (nfula[NFULA_CFG_CMD]) {
729                 u_int8_t pf = nfmsg->nfgen_family;
730                 cmd = nla_data(nfula[NFULA_CFG_CMD]);
731
732                 /* Commands without queue context */
733                 switch (cmd->command) {
734                 case NFULNL_CFG_CMD_PF_BIND:
735                         return nf_log_bind_pf(pf, &nfulnl_logger);
736                 case NFULNL_CFG_CMD_PF_UNBIND:
737                         nf_log_unbind_pf(pf);
738                         return 0;
739                 }
740         }
741
742         inst = instance_lookup_get(group_num);
743         if (inst && inst->peer_pid != NETLINK_CB(skb).pid) {
744                 ret = -EPERM;
745                 goto out_put;
746         }
747
748         if (cmd != NULL) {
749                 switch (cmd->command) {
750                 case NFULNL_CFG_CMD_BIND:
751                         if (inst) {
752                                 ret = -EBUSY;
753                                 goto out_put;
754                         }
755
756                         inst = instance_create(group_num,
757                                                NETLINK_CB(skb).pid);
758                         if (IS_ERR(inst)) {
759                                 ret = PTR_ERR(inst);
760                                 goto out;
761                         }
762                         break;
763                 case NFULNL_CFG_CMD_UNBIND:
764                         if (!inst) {
765                                 ret = -ENODEV;
766                                 goto out;
767                         }
768
769                         instance_destroy(inst);
770                         goto out;
771                 default:
772                         ret = -ENOTSUPP;
773                         break;
774                 }
775         }
776
777         if (nfula[NFULA_CFG_MODE]) {
778                 struct nfulnl_msg_config_mode *params;
779                 params = nla_data(nfula[NFULA_CFG_MODE]);
780
781                 if (!inst) {
782                         ret = -ENODEV;
783                         goto out;
784                 }
785                 nfulnl_set_mode(inst, params->copy_mode,
786                                 ntohl(params->copy_range));
787         }
788
789         if (nfula[NFULA_CFG_TIMEOUT]) {
790                 __be32 timeout = nla_get_be32(nfula[NFULA_CFG_TIMEOUT]);
791
792                 if (!inst) {
793                         ret = -ENODEV;
794                         goto out;
795                 }
796                 nfulnl_set_timeout(inst, ntohl(timeout));
797         }
798
799         if (nfula[NFULA_CFG_NLBUFSIZ]) {
800                 __be32 nlbufsiz = nla_get_be32(nfula[NFULA_CFG_NLBUFSIZ]);
801
802                 if (!inst) {
803                         ret = -ENODEV;
804                         goto out;
805                 }
806                 nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz));
807         }
808
809         if (nfula[NFULA_CFG_QTHRESH]) {
810                 __be32 qthresh = nla_get_be32(nfula[NFULA_CFG_QTHRESH]);
811
812                 if (!inst) {
813                         ret = -ENODEV;
814                         goto out;
815                 }
816                 nfulnl_set_qthresh(inst, ntohl(qthresh));
817         }
818
819         if (nfula[NFULA_CFG_FLAGS]) {
820                 __be16 flags = nla_get_be16(nfula[NFULA_CFG_FLAGS]);
821
822                 if (!inst) {
823                         ret = -ENODEV;
824                         goto out;
825                 }
826                 nfulnl_set_flags(inst, ntohs(flags));
827         }
828
829 out_put:
830         instance_put(inst);
831 out:
832         return ret;
833 }
834
835 static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = {
836         [NFULNL_MSG_PACKET]     = { .call = nfulnl_recv_unsupp,
837                                     .attr_count = NFULA_MAX, },
838         [NFULNL_MSG_CONFIG]     = { .call = nfulnl_recv_config,
839                                     .attr_count = NFULA_CFG_MAX,
840                                     .policy = nfula_cfg_policy },
841 };
842
843 static const struct nfnetlink_subsystem nfulnl_subsys = {
844         .name           = "log",
845         .subsys_id      = NFNL_SUBSYS_ULOG,
846         .cb_count       = NFULNL_MSG_MAX,
847         .cb             = nfulnl_cb,
848 };
849
850 #ifdef CONFIG_PROC_FS
851 struct iter_state {
852         unsigned int bucket;
853 };
854
855 static struct hlist_node *get_first(struct iter_state *st)
856 {
857         if (!st)
858                 return NULL;
859
860         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
861                 if (!hlist_empty(&instance_table[st->bucket]))
862                         return instance_table[st->bucket].first;
863         }
864         return NULL;
865 }
866
867 static struct hlist_node *get_next(struct iter_state *st, struct hlist_node *h)
868 {
869         h = h->next;
870         while (!h) {
871                 if (++st->bucket >= INSTANCE_BUCKETS)
872                         return NULL;
873
874                 h = instance_table[st->bucket].first;
875         }
876         return h;
877 }
878
879 static struct hlist_node *get_idx(struct iter_state *st, loff_t pos)
880 {
881         struct hlist_node *head;
882         head = get_first(st);
883
884         if (head)
885                 while (pos && (head = get_next(st, head)))
886                         pos--;
887         return pos ? NULL : head;
888 }
889
890 static void *seq_start(struct seq_file *seq, loff_t *pos)
891         __acquires(instances_lock)
892 {
893         read_lock_bh(&instances_lock);
894         return get_idx(seq->private, *pos);
895 }
896
897 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
898 {
899         (*pos)++;
900         return get_next(s->private, v);
901 }
902
903 static void seq_stop(struct seq_file *s, void *v)
904         __releases(instances_lock)
905 {
906         read_unlock_bh(&instances_lock);
907 }
908
909 static int seq_show(struct seq_file *s, void *v)
910 {
911         const struct nfulnl_instance *inst = v;
912
913         return seq_printf(s, "%5d %6d %5d %1d %5d %6d %2d\n",
914                           inst->group_num,
915                           inst->peer_pid, inst->qlen,
916                           inst->copy_mode, inst->copy_range,
917                           inst->flushtimeout, atomic_read(&inst->use));
918 }
919
920 static const struct seq_operations nful_seq_ops = {
921         .start  = seq_start,
922         .next   = seq_next,
923         .stop   = seq_stop,
924         .show   = seq_show,
925 };
926
927 static int nful_open(struct inode *inode, struct file *file)
928 {
929         return seq_open_private(file, &nful_seq_ops,
930                         sizeof(struct iter_state));
931 }
932
933 static const struct file_operations nful_file_ops = {
934         .owner   = THIS_MODULE,
935         .open    = nful_open,
936         .read    = seq_read,
937         .llseek  = seq_lseek,
938         .release = seq_release_private,
939 };
940
941 #endif /* PROC_FS */
942
943 static int __init nfnetlink_log_init(void)
944 {
945         int i, status = -ENOMEM;
946
947         for (i = 0; i < INSTANCE_BUCKETS; i++)
948                 INIT_HLIST_HEAD(&instance_table[i]);
949
950         /* it's not really all that important to have a random value, so
951          * we can do this from the init function, even if there hasn't
952          * been that much entropy yet */
953         get_random_bytes(&hash_init, sizeof(hash_init));
954
955         netlink_register_notifier(&nfulnl_rtnl_notifier);
956         status = nfnetlink_subsys_register(&nfulnl_subsys);
957         if (status < 0) {
958                 printk(KERN_ERR "log: failed to create netlink socket\n");
959                 goto cleanup_netlink_notifier;
960         }
961
962         status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger);
963         if (status < 0) {
964                 printk(KERN_ERR "log: failed to register logger\n");
965                 goto cleanup_subsys;
966         }
967
968 #ifdef CONFIG_PROC_FS
969         if (!proc_create("nfnetlink_log", 0440,
970                          proc_net_netfilter, &nful_file_ops))
971                 goto cleanup_logger;
972 #endif
973         return status;
974
975 #ifdef CONFIG_PROC_FS
976 cleanup_logger:
977         nf_log_unregister(&nfulnl_logger);
978 #endif
979 cleanup_subsys:
980         nfnetlink_subsys_unregister(&nfulnl_subsys);
981 cleanup_netlink_notifier:
982         netlink_unregister_notifier(&nfulnl_rtnl_notifier);
983         return status;
984 }
985
986 static void __exit nfnetlink_log_fini(void)
987 {
988         nf_log_unregister(&nfulnl_logger);
989 #ifdef CONFIG_PROC_FS
990         remove_proc_entry("nfnetlink_log", proc_net_netfilter);
991 #endif
992         nfnetlink_subsys_unregister(&nfulnl_subsys);
993         netlink_unregister_notifier(&nfulnl_rtnl_notifier);
994 }
995
996 MODULE_DESCRIPTION("netfilter userspace logging");
997 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
998 MODULE_LICENSE("GPL");
999 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG);
1000
1001 module_init(nfnetlink_log_init);
1002 module_exit(nfnetlink_log_fini);