netfilter: nfnetlink: netns support
[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, &init_net, inst->peer_pid,
327                                    MSG_DONTWAIT);
328
329         inst->qlen = 0;
330         inst->skb = NULL;
331
332 nlmsg_failure:
333         return status;
334 }
335
336 static void
337 __nfulnl_flush(struct nfulnl_instance *inst)
338 {
339         /* timer holds a reference */
340         if (del_timer(&inst->timer))
341                 instance_put(inst);
342         if (inst->skb)
343                 __nfulnl_send(inst);
344 }
345
346 static void
347 nfulnl_timer(unsigned long data)
348 {
349         struct nfulnl_instance *inst = (struct nfulnl_instance *)data;
350
351         spin_lock_bh(&inst->lock);
352         if (inst->skb)
353                 __nfulnl_send(inst);
354         spin_unlock_bh(&inst->lock);
355         instance_put(inst);
356 }
357
358 /* This is an inline function, we don't really care about a long
359  * list of arguments */
360 static inline int
361 __build_packet_message(struct nfulnl_instance *inst,
362                         const struct sk_buff *skb,
363                         unsigned int data_len,
364                         u_int8_t pf,
365                         unsigned int hooknum,
366                         const struct net_device *indev,
367                         const struct net_device *outdev,
368                         const struct nf_loginfo *li,
369                         const char *prefix, unsigned int plen)
370 {
371         struct nfulnl_msg_packet_hdr pmsg;
372         struct nlmsghdr *nlh;
373         struct nfgenmsg *nfmsg;
374         __be32 tmp_uint;
375         sk_buff_data_t old_tail = inst->skb->tail;
376
377         nlh = NLMSG_PUT(inst->skb, 0, 0,
378                         NFNL_SUBSYS_ULOG << 8 | NFULNL_MSG_PACKET,
379                         sizeof(struct nfgenmsg));
380         nfmsg = NLMSG_DATA(nlh);
381         nfmsg->nfgen_family = pf;
382         nfmsg->version = NFNETLINK_V0;
383         nfmsg->res_id = htons(inst->group_num);
384
385         pmsg.hw_protocol        = skb->protocol;
386         pmsg.hook               = hooknum;
387
388         NLA_PUT(inst->skb, NFULA_PACKET_HDR, sizeof(pmsg), &pmsg);
389
390         if (prefix)
391                 NLA_PUT(inst->skb, NFULA_PREFIX, plen, prefix);
392
393         if (indev) {
394 #ifndef CONFIG_BRIDGE_NETFILTER
395                 NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_INDEV,
396                              htonl(indev->ifindex));
397 #else
398                 if (pf == PF_BRIDGE) {
399                         /* Case 1: outdev is physical input device, we need to
400                          * look for bridge group (when called from
401                          * netfilter_bridge) */
402                         NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
403                                      htonl(indev->ifindex));
404                         /* this is the bridge group "brX" */
405                         NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_INDEV,
406                                      htonl(indev->br_port->br->dev->ifindex));
407                 } else {
408                         /* Case 2: indev is bridge group, we need to look for
409                          * physical device (when called from ipv4) */
410                         NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_INDEV,
411                                      htonl(indev->ifindex));
412                         if (skb->nf_bridge && skb->nf_bridge->physindev)
413                                 NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
414                                              htonl(skb->nf_bridge->physindev->ifindex));
415                 }
416 #endif
417         }
418
419         if (outdev) {
420                 tmp_uint = htonl(outdev->ifindex);
421 #ifndef CONFIG_BRIDGE_NETFILTER
422                 NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_OUTDEV,
423                              htonl(outdev->ifindex));
424 #else
425                 if (pf == PF_BRIDGE) {
426                         /* Case 1: outdev is physical output device, we need to
427                          * look for bridge group (when called from
428                          * netfilter_bridge) */
429                         NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
430                                      htonl(outdev->ifindex));
431                         /* this is the bridge group "brX" */
432                         NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_OUTDEV,
433                                      htonl(outdev->br_port->br->dev->ifindex));
434                 } else {
435                         /* Case 2: indev is a bridge group, we need to look
436                          * for physical device (when called from ipv4) */
437                         NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_OUTDEV,
438                                      htonl(outdev->ifindex));
439                         if (skb->nf_bridge && skb->nf_bridge->physoutdev)
440                                 NLA_PUT_BE32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
441                                              htonl(skb->nf_bridge->physoutdev->ifindex));
442                 }
443 #endif
444         }
445
446         if (skb->mark)
447                 NLA_PUT_BE32(inst->skb, NFULA_MARK, htonl(skb->mark));
448
449         if (indev && skb->dev) {
450                 struct nfulnl_msg_packet_hw phw;
451                 int len = dev_parse_header(skb, phw.hw_addr);
452                 if (len > 0) {
453                         phw.hw_addrlen = htons(len);
454                         NLA_PUT(inst->skb, NFULA_HWADDR, sizeof(phw), &phw);
455                 }
456         }
457
458         if (indev && skb_mac_header_was_set(skb)) {
459                 NLA_PUT_BE16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type));
460                 NLA_PUT_BE16(inst->skb, NFULA_HWLEN,
461                              htons(skb->dev->hard_header_len));
462                 NLA_PUT(inst->skb, NFULA_HWHEADER, skb->dev->hard_header_len,
463                         skb_mac_header(skb));
464         }
465
466         if (skb->tstamp.tv64) {
467                 struct nfulnl_msg_packet_timestamp ts;
468                 struct timeval tv = ktime_to_timeval(skb->tstamp);
469                 ts.sec = cpu_to_be64(tv.tv_sec);
470                 ts.usec = cpu_to_be64(tv.tv_usec);
471
472                 NLA_PUT(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts);
473         }
474
475         /* UID */
476         if (skb->sk) {
477                 read_lock_bh(&skb->sk->sk_callback_lock);
478                 if (skb->sk->sk_socket && skb->sk->sk_socket->file) {
479                         struct file *file = skb->sk->sk_socket->file;
480                         __be32 uid = htonl(file->f_cred->fsuid);
481                         __be32 gid = htonl(file->f_cred->fsgid);
482                         /* need to unlock here since NLA_PUT may goto */
483                         read_unlock_bh(&skb->sk->sk_callback_lock);
484                         NLA_PUT_BE32(inst->skb, NFULA_UID, uid);
485                         NLA_PUT_BE32(inst->skb, NFULA_GID, gid);
486                 } else
487                         read_unlock_bh(&skb->sk->sk_callback_lock);
488         }
489
490         /* local sequence number */
491         if (inst->flags & NFULNL_CFG_F_SEQ)
492                 NLA_PUT_BE32(inst->skb, NFULA_SEQ, htonl(inst->seq++));
493
494         /* global sequence number */
495         if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
496                 NLA_PUT_BE32(inst->skb, NFULA_SEQ_GLOBAL,
497                              htonl(atomic_inc_return(&global_seq)));
498
499         if (data_len) {
500                 struct nlattr *nla;
501                 int size = nla_attr_size(data_len);
502
503                 if (skb_tailroom(inst->skb) < nla_total_size(data_len)) {
504                         printk(KERN_WARNING "nfnetlink_log: no tailroom!\n");
505                         goto nlmsg_failure;
506                 }
507
508                 nla = (struct nlattr *)skb_put(inst->skb, nla_total_size(data_len));
509                 nla->nla_type = NFULA_PAYLOAD;
510                 nla->nla_len = size;
511
512                 if (skb_copy_bits(skb, 0, nla_data(nla), data_len))
513                         BUG();
514         }
515
516         nlh->nlmsg_len = inst->skb->tail - old_tail;
517         return 0;
518
519 nlmsg_failure:
520 nla_put_failure:
521         PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n");
522         return -1;
523 }
524
525 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
526
527 static struct nf_loginfo default_loginfo = {
528         .type =         NF_LOG_TYPE_ULOG,
529         .u = {
530                 .ulog = {
531                         .copy_len       = 0xffff,
532                         .group          = 0,
533                         .qthreshold     = 1,
534                 },
535         },
536 };
537
538 /* log handler for internal netfilter logging api */
539 void
540 nfulnl_log_packet(u_int8_t pf,
541                   unsigned int hooknum,
542                   const struct sk_buff *skb,
543                   const struct net_device *in,
544                   const struct net_device *out,
545                   const struct nf_loginfo *li_user,
546                   const char *prefix)
547 {
548         unsigned int size, data_len;
549         struct nfulnl_instance *inst;
550         const struct nf_loginfo *li;
551         unsigned int qthreshold;
552         unsigned int plen;
553
554         if (li_user && li_user->type == NF_LOG_TYPE_ULOG)
555                 li = li_user;
556         else
557                 li = &default_loginfo;
558
559         inst = instance_lookup_get(li->u.ulog.group);
560         if (!inst)
561                 return;
562
563         plen = 0;
564         if (prefix)
565                 plen = strlen(prefix) + 1;
566
567         /* FIXME: do we want to make the size calculation conditional based on
568          * what is actually present?  way more branches and checks, but more
569          * memory efficient... */
570         size =    NLMSG_SPACE(sizeof(struct nfgenmsg))
571                 + nla_total_size(sizeof(struct nfulnl_msg_packet_hdr))
572                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
573                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
574 #ifdef CONFIG_BRIDGE_NETFILTER
575                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
576                 + nla_total_size(sizeof(u_int32_t))     /* ifindex */
577 #endif
578                 + nla_total_size(sizeof(u_int32_t))     /* mark */
579                 + nla_total_size(sizeof(u_int32_t))     /* uid */
580                 + nla_total_size(sizeof(u_int32_t))     /* gid */
581                 + nla_total_size(plen)                  /* prefix */
582                 + nla_total_size(sizeof(struct nfulnl_msg_packet_hw))
583                 + nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp));
584
585         if (in && skb_mac_header_was_set(skb)) {
586                 size +=   nla_total_size(skb->dev->hard_header_len)
587                         + nla_total_size(sizeof(u_int16_t))     /* hwtype */
588                         + nla_total_size(sizeof(u_int16_t));    /* hwlen */
589         }
590
591         spin_lock_bh(&inst->lock);
592
593         if (inst->flags & NFULNL_CFG_F_SEQ)
594                 size += nla_total_size(sizeof(u_int32_t));
595         if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
596                 size += nla_total_size(sizeof(u_int32_t));
597
598         qthreshold = inst->qthreshold;
599         /* per-rule qthreshold overrides per-instance */
600         if (li->u.ulog.qthreshold)
601                 if (qthreshold > li->u.ulog.qthreshold)
602                         qthreshold = li->u.ulog.qthreshold;
603
604
605         switch (inst->copy_mode) {
606         case NFULNL_COPY_META:
607         case NFULNL_COPY_NONE:
608                 data_len = 0;
609                 break;
610
611         case NFULNL_COPY_PACKET:
612                 if (inst->copy_range == 0
613                     || inst->copy_range > skb->len)
614                         data_len = skb->len;
615                 else
616                         data_len = inst->copy_range;
617
618                 size += nla_total_size(data_len);
619                 break;
620
621         default:
622                 goto unlock_and_release;
623         }
624
625         if (inst->skb &&
626             size > skb_tailroom(inst->skb) - sizeof(struct nfgenmsg)) {
627                 /* either the queue len is too high or we don't have
628                  * enough room in the skb left. flush to userspace. */
629                 __nfulnl_flush(inst);
630         }
631
632         if (!inst->skb) {
633                 inst->skb = nfulnl_alloc_skb(inst->nlbufsiz, size);
634                 if (!inst->skb)
635                         goto alloc_failure;
636         }
637
638         inst->qlen++;
639
640         __build_packet_message(inst, skb, data_len, pf,
641                                 hooknum, in, out, li, prefix, plen);
642
643         if (inst->qlen >= qthreshold)
644                 __nfulnl_flush(inst);
645         /* timer_pending always called within inst->lock, so there
646          * is no chance of a race here */
647         else if (!timer_pending(&inst->timer)) {
648                 instance_get(inst);
649                 inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100);
650                 add_timer(&inst->timer);
651         }
652
653 unlock_and_release:
654         spin_unlock_bh(&inst->lock);
655         instance_put(inst);
656         return;
657
658 alloc_failure:
659         /* FIXME: statistics */
660         goto unlock_and_release;
661 }
662 EXPORT_SYMBOL_GPL(nfulnl_log_packet);
663
664 static int
665 nfulnl_rcv_nl_event(struct notifier_block *this,
666                    unsigned long event, void *ptr)
667 {
668         struct netlink_notify *n = ptr;
669
670         if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
671                 int i;
672
673                 /* destroy all instances for this pid */
674                 write_lock_bh(&instances_lock);
675                 for  (i = 0; i < INSTANCE_BUCKETS; i++) {
676                         struct hlist_node *tmp, *t2;
677                         struct nfulnl_instance *inst;
678                         struct hlist_head *head = &instance_table[i];
679
680                         hlist_for_each_entry_safe(inst, tmp, t2, head, hlist) {
681                                 if ((net_eq(n->net, &init_net)) &&
682                                     (n->pid == inst->peer_pid))
683                                         __instance_destroy(inst);
684                         }
685                 }
686                 write_unlock_bh(&instances_lock);
687         }
688         return NOTIFY_DONE;
689 }
690
691 static struct notifier_block nfulnl_rtnl_notifier = {
692         .notifier_call  = nfulnl_rcv_nl_event,
693 };
694
695 static int
696 nfulnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
697                    const struct nlmsghdr *nlh,
698                    const struct nlattr * const nfqa[])
699 {
700         return -ENOTSUPP;
701 }
702
703 static struct nf_logger nfulnl_logger __read_mostly = {
704         .name   = "nfnetlink_log",
705         .logfn  = &nfulnl_log_packet,
706         .me     = THIS_MODULE,
707 };
708
709 static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = {
710         [NFULA_CFG_CMD]         = { .len = sizeof(struct nfulnl_msg_config_cmd) },
711         [NFULA_CFG_MODE]        = { .len = sizeof(struct nfulnl_msg_config_mode) },
712         [NFULA_CFG_TIMEOUT]     = { .type = NLA_U32 },
713         [NFULA_CFG_QTHRESH]     = { .type = NLA_U32 },
714         [NFULA_CFG_NLBUFSIZ]    = { .type = NLA_U32 },
715         [NFULA_CFG_FLAGS]       = { .type = NLA_U16 },
716 };
717
718 static int
719 nfulnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
720                    const struct nlmsghdr *nlh,
721                    const struct nlattr * const nfula[])
722 {
723         struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
724         u_int16_t group_num = ntohs(nfmsg->res_id);
725         struct nfulnl_instance *inst;
726         struct nfulnl_msg_config_cmd *cmd = NULL;
727         int ret = 0;
728
729         if (nfula[NFULA_CFG_CMD]) {
730                 u_int8_t pf = nfmsg->nfgen_family;
731                 cmd = nla_data(nfula[NFULA_CFG_CMD]);
732
733                 /* Commands without queue context */
734                 switch (cmd->command) {
735                 case NFULNL_CFG_CMD_PF_BIND:
736                         return nf_log_bind_pf(pf, &nfulnl_logger);
737                 case NFULNL_CFG_CMD_PF_UNBIND:
738                         nf_log_unbind_pf(pf);
739                         return 0;
740                 }
741         }
742
743         inst = instance_lookup_get(group_num);
744         if (inst && inst->peer_pid != NETLINK_CB(skb).pid) {
745                 ret = -EPERM;
746                 goto out_put;
747         }
748
749         if (cmd != NULL) {
750                 switch (cmd->command) {
751                 case NFULNL_CFG_CMD_BIND:
752                         if (inst) {
753                                 ret = -EBUSY;
754                                 goto out_put;
755                         }
756
757                         inst = instance_create(group_num,
758                                                NETLINK_CB(skb).pid);
759                         if (IS_ERR(inst)) {
760                                 ret = PTR_ERR(inst);
761                                 goto out;
762                         }
763                         break;
764                 case NFULNL_CFG_CMD_UNBIND:
765                         if (!inst) {
766                                 ret = -ENODEV;
767                                 goto out;
768                         }
769
770                         instance_destroy(inst);
771                         goto out;
772                 default:
773                         ret = -ENOTSUPP;
774                         break;
775                 }
776         }
777
778         if (nfula[NFULA_CFG_MODE]) {
779                 struct nfulnl_msg_config_mode *params;
780                 params = nla_data(nfula[NFULA_CFG_MODE]);
781
782                 if (!inst) {
783                         ret = -ENODEV;
784                         goto out;
785                 }
786                 nfulnl_set_mode(inst, params->copy_mode,
787                                 ntohl(params->copy_range));
788         }
789
790         if (nfula[NFULA_CFG_TIMEOUT]) {
791                 __be32 timeout = nla_get_be32(nfula[NFULA_CFG_TIMEOUT]);
792
793                 if (!inst) {
794                         ret = -ENODEV;
795                         goto out;
796                 }
797                 nfulnl_set_timeout(inst, ntohl(timeout));
798         }
799
800         if (nfula[NFULA_CFG_NLBUFSIZ]) {
801                 __be32 nlbufsiz = nla_get_be32(nfula[NFULA_CFG_NLBUFSIZ]);
802
803                 if (!inst) {
804                         ret = -ENODEV;
805                         goto out;
806                 }
807                 nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz));
808         }
809
810         if (nfula[NFULA_CFG_QTHRESH]) {
811                 __be32 qthresh = nla_get_be32(nfula[NFULA_CFG_QTHRESH]);
812
813                 if (!inst) {
814                         ret = -ENODEV;
815                         goto out;
816                 }
817                 nfulnl_set_qthresh(inst, ntohl(qthresh));
818         }
819
820         if (nfula[NFULA_CFG_FLAGS]) {
821                 __be16 flags = nla_get_be16(nfula[NFULA_CFG_FLAGS]);
822
823                 if (!inst) {
824                         ret = -ENODEV;
825                         goto out;
826                 }
827                 nfulnl_set_flags(inst, ntohs(flags));
828         }
829
830 out_put:
831         instance_put(inst);
832 out:
833         return ret;
834 }
835
836 static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = {
837         [NFULNL_MSG_PACKET]     = { .call = nfulnl_recv_unsupp,
838                                     .attr_count = NFULA_MAX, },
839         [NFULNL_MSG_CONFIG]     = { .call = nfulnl_recv_config,
840                                     .attr_count = NFULA_CFG_MAX,
841                                     .policy = nfula_cfg_policy },
842 };
843
844 static const struct nfnetlink_subsystem nfulnl_subsys = {
845         .name           = "log",
846         .subsys_id      = NFNL_SUBSYS_ULOG,
847         .cb_count       = NFULNL_MSG_MAX,
848         .cb             = nfulnl_cb,
849 };
850
851 #ifdef CONFIG_PROC_FS
852 struct iter_state {
853         unsigned int bucket;
854 };
855
856 static struct hlist_node *get_first(struct iter_state *st)
857 {
858         if (!st)
859                 return NULL;
860
861         for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
862                 if (!hlist_empty(&instance_table[st->bucket]))
863                         return instance_table[st->bucket].first;
864         }
865         return NULL;
866 }
867
868 static struct hlist_node *get_next(struct iter_state *st, struct hlist_node *h)
869 {
870         h = h->next;
871         while (!h) {
872                 if (++st->bucket >= INSTANCE_BUCKETS)
873                         return NULL;
874
875                 h = instance_table[st->bucket].first;
876         }
877         return h;
878 }
879
880 static struct hlist_node *get_idx(struct iter_state *st, loff_t pos)
881 {
882         struct hlist_node *head;
883         head = get_first(st);
884
885         if (head)
886                 while (pos && (head = get_next(st, head)))
887                         pos--;
888         return pos ? NULL : head;
889 }
890
891 static void *seq_start(struct seq_file *seq, loff_t *pos)
892         __acquires(instances_lock)
893 {
894         read_lock_bh(&instances_lock);
895         return get_idx(seq->private, *pos);
896 }
897
898 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
899 {
900         (*pos)++;
901         return get_next(s->private, v);
902 }
903
904 static void seq_stop(struct seq_file *s, void *v)
905         __releases(instances_lock)
906 {
907         read_unlock_bh(&instances_lock);
908 }
909
910 static int seq_show(struct seq_file *s, void *v)
911 {
912         const struct nfulnl_instance *inst = v;
913
914         return seq_printf(s, "%5d %6d %5d %1d %5d %6d %2d\n",
915                           inst->group_num,
916                           inst->peer_pid, inst->qlen,
917                           inst->copy_mode, inst->copy_range,
918                           inst->flushtimeout, atomic_read(&inst->use));
919 }
920
921 static const struct seq_operations nful_seq_ops = {
922         .start  = seq_start,
923         .next   = seq_next,
924         .stop   = seq_stop,
925         .show   = seq_show,
926 };
927
928 static int nful_open(struct inode *inode, struct file *file)
929 {
930         return seq_open_private(file, &nful_seq_ops,
931                         sizeof(struct iter_state));
932 }
933
934 static const struct file_operations nful_file_ops = {
935         .owner   = THIS_MODULE,
936         .open    = nful_open,
937         .read    = seq_read,
938         .llseek  = seq_lseek,
939         .release = seq_release_private,
940 };
941
942 #endif /* PROC_FS */
943
944 static int __init nfnetlink_log_init(void)
945 {
946         int i, status = -ENOMEM;
947
948         for (i = 0; i < INSTANCE_BUCKETS; i++)
949                 INIT_HLIST_HEAD(&instance_table[i]);
950
951         /* it's not really all that important to have a random value, so
952          * we can do this from the init function, even if there hasn't
953          * been that much entropy yet */
954         get_random_bytes(&hash_init, sizeof(hash_init));
955
956         netlink_register_notifier(&nfulnl_rtnl_notifier);
957         status = nfnetlink_subsys_register(&nfulnl_subsys);
958         if (status < 0) {
959                 printk(KERN_ERR "log: failed to create netlink socket\n");
960                 goto cleanup_netlink_notifier;
961         }
962
963         status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger);
964         if (status < 0) {
965                 printk(KERN_ERR "log: failed to register logger\n");
966                 goto cleanup_subsys;
967         }
968
969 #ifdef CONFIG_PROC_FS
970         if (!proc_create("nfnetlink_log", 0440,
971                          proc_net_netfilter, &nful_file_ops))
972                 goto cleanup_logger;
973 #endif
974         return status;
975
976 #ifdef CONFIG_PROC_FS
977 cleanup_logger:
978         nf_log_unregister(&nfulnl_logger);
979 #endif
980 cleanup_subsys:
981         nfnetlink_subsys_unregister(&nfulnl_subsys);
982 cleanup_netlink_notifier:
983         netlink_unregister_notifier(&nfulnl_rtnl_notifier);
984         return status;
985 }
986
987 static void __exit nfnetlink_log_fini(void)
988 {
989         nf_log_unregister(&nfulnl_logger);
990 #ifdef CONFIG_PROC_FS
991         remove_proc_entry("nfnetlink_log", proc_net_netfilter);
992 #endif
993         nfnetlink_subsys_unregister(&nfulnl_subsys);
994         netlink_unregister_notifier(&nfulnl_rtnl_notifier);
995 }
996
997 MODULE_DESCRIPTION("netfilter userspace logging");
998 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
999 MODULE_LICENSE("GPL");
1000 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG);
1001
1002 module_init(nfnetlink_log_init);
1003 module_exit(nfnetlink_log_fini);