netlink: Rename pid to portid to avoid confusion
[GitHub/mt8127/android_kernel_alcatel_ttab.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 <linux/slab.h>
32 #include <net/sock.h>
33 #include <net/netfilter/nf_log.h>
34 #include <net/netfilter/nfnetlink_log.h>
35
36 #include <linux/atomic.h>
37
38 #ifdef CONFIG_BRIDGE_NETFILTER
39 #include "../bridge/br_private.h"
40 #endif
41
42 #define NFULNL_NLBUFSIZ_DEFAULT NLMSG_GOODSIZE
43 #define NFULNL_TIMEOUT_DEFAULT 100 /* every second */
44 #define NFULNL_QTHRESH_DEFAULT 100 /* 100 packets */
45 #define NFULNL_COPY_RANGE_MAX 0xFFFF /* max packet size is limited by 16-bit struct nfattr nfa_len field */
46
47 #define PRINTR(x, args...) do { if (net_ratelimit()) \
48 printk(x, ## args); } while (0);
49
50 struct nfulnl_instance {
51 struct hlist_node hlist; /* global list of instances */
52 spinlock_t lock;
53 atomic_t use; /* use count */
54
55 unsigned int qlen; /* number of nlmsgs in skb */
56 struct sk_buff *skb; /* pre-allocatd skb */
57 struct timer_list timer;
58 struct user_namespace *peer_user_ns; /* User namespace of the peer process */
59 int peer_portid; /* PORTID of the peer process */
60
61 /* configurable parameters */
62 unsigned int flushtimeout; /* timeout until queue flush */
63 unsigned int nlbufsiz; /* netlink buffer allocation size */
64 unsigned int qthreshold; /* threshold of the queue */
65 u_int32_t copy_range;
66 u_int32_t seq; /* instance-local sequential counter */
67 u_int16_t group_num; /* number of this queue */
68 u_int16_t flags;
69 u_int8_t copy_mode;
70 struct rcu_head rcu;
71 };
72
73 static DEFINE_SPINLOCK(instances_lock);
74 static atomic_t global_seq;
75
76 #define INSTANCE_BUCKETS 16
77 static struct hlist_head instance_table[INSTANCE_BUCKETS];
78 static unsigned int hash_init;
79
80 static inline u_int8_t instance_hashfn(u_int16_t group_num)
81 {
82 return ((group_num & 0xff) % INSTANCE_BUCKETS);
83 }
84
85 static struct nfulnl_instance *
86 __instance_lookup(u_int16_t group_num)
87 {
88 struct hlist_head *head;
89 struct hlist_node *pos;
90 struct nfulnl_instance *inst;
91
92 head = &instance_table[instance_hashfn(group_num)];
93 hlist_for_each_entry_rcu(inst, pos, head, hlist) {
94 if (inst->group_num == group_num)
95 return inst;
96 }
97 return NULL;
98 }
99
100 static inline void
101 instance_get(struct nfulnl_instance *inst)
102 {
103 atomic_inc(&inst->use);
104 }
105
106 static struct nfulnl_instance *
107 instance_lookup_get(u_int16_t group_num)
108 {
109 struct nfulnl_instance *inst;
110
111 rcu_read_lock_bh();
112 inst = __instance_lookup(group_num);
113 if (inst && !atomic_inc_not_zero(&inst->use))
114 inst = NULL;
115 rcu_read_unlock_bh();
116
117 return inst;
118 }
119
120 static void nfulnl_instance_free_rcu(struct rcu_head *head)
121 {
122 kfree(container_of(head, struct nfulnl_instance, rcu));
123 module_put(THIS_MODULE);
124 }
125
126 static void
127 instance_put(struct nfulnl_instance *inst)
128 {
129 if (inst && atomic_dec_and_test(&inst->use))
130 call_rcu_bh(&inst->rcu, nfulnl_instance_free_rcu);
131 }
132
133 static void nfulnl_timer(unsigned long data);
134
135 static struct nfulnl_instance *
136 instance_create(u_int16_t group_num, int portid, struct user_namespace *user_ns)
137 {
138 struct nfulnl_instance *inst;
139 int err;
140
141 spin_lock_bh(&instances_lock);
142 if (__instance_lookup(group_num)) {
143 err = -EEXIST;
144 goto out_unlock;
145 }
146
147 inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
148 if (!inst) {
149 err = -ENOMEM;
150 goto out_unlock;
151 }
152
153 if (!try_module_get(THIS_MODULE)) {
154 kfree(inst);
155 err = -EAGAIN;
156 goto out_unlock;
157 }
158
159 INIT_HLIST_NODE(&inst->hlist);
160 spin_lock_init(&inst->lock);
161 /* needs to be two, since we _put() after creation */
162 atomic_set(&inst->use, 2);
163
164 setup_timer(&inst->timer, nfulnl_timer, (unsigned long)inst);
165
166 inst->peer_user_ns = user_ns;
167 inst->peer_portid = portid;
168 inst->group_num = group_num;
169
170 inst->qthreshold = NFULNL_QTHRESH_DEFAULT;
171 inst->flushtimeout = NFULNL_TIMEOUT_DEFAULT;
172 inst->nlbufsiz = NFULNL_NLBUFSIZ_DEFAULT;
173 inst->copy_mode = NFULNL_COPY_PACKET;
174 inst->copy_range = NFULNL_COPY_RANGE_MAX;
175
176 hlist_add_head_rcu(&inst->hlist,
177 &instance_table[instance_hashfn(group_num)]);
178
179 spin_unlock_bh(&instances_lock);
180
181 return inst;
182
183 out_unlock:
184 spin_unlock_bh(&instances_lock);
185 return ERR_PTR(err);
186 }
187
188 static void __nfulnl_flush(struct nfulnl_instance *inst);
189
190 /* called with BH disabled */
191 static void
192 __instance_destroy(struct nfulnl_instance *inst)
193 {
194 /* first pull it out of the global list */
195 hlist_del_rcu(&inst->hlist);
196
197 /* then flush all pending packets from skb */
198
199 spin_lock(&inst->lock);
200
201 /* lockless readers wont be able to use us */
202 inst->copy_mode = NFULNL_COPY_DISABLED;
203
204 if (inst->skb)
205 __nfulnl_flush(inst);
206 spin_unlock(&inst->lock);
207
208 /* and finally put the refcount */
209 instance_put(inst);
210 }
211
212 static inline void
213 instance_destroy(struct nfulnl_instance *inst)
214 {
215 spin_lock_bh(&instances_lock);
216 __instance_destroy(inst);
217 spin_unlock_bh(&instances_lock);
218 }
219
220 static int
221 nfulnl_set_mode(struct nfulnl_instance *inst, u_int8_t mode,
222 unsigned int range)
223 {
224 int status = 0;
225
226 spin_lock_bh(&inst->lock);
227
228 switch (mode) {
229 case NFULNL_COPY_NONE:
230 case NFULNL_COPY_META:
231 inst->copy_mode = mode;
232 inst->copy_range = 0;
233 break;
234
235 case NFULNL_COPY_PACKET:
236 inst->copy_mode = mode;
237 inst->copy_range = min_t(unsigned int,
238 range, NFULNL_COPY_RANGE_MAX);
239 break;
240
241 default:
242 status = -EINVAL;
243 break;
244 }
245
246 spin_unlock_bh(&inst->lock);
247
248 return status;
249 }
250
251 static int
252 nfulnl_set_nlbufsiz(struct nfulnl_instance *inst, u_int32_t nlbufsiz)
253 {
254 int status;
255
256 spin_lock_bh(&inst->lock);
257 if (nlbufsiz < NFULNL_NLBUFSIZ_DEFAULT)
258 status = -ERANGE;
259 else if (nlbufsiz > 131072)
260 status = -ERANGE;
261 else {
262 inst->nlbufsiz = nlbufsiz;
263 status = 0;
264 }
265 spin_unlock_bh(&inst->lock);
266
267 return status;
268 }
269
270 static int
271 nfulnl_set_timeout(struct nfulnl_instance *inst, u_int32_t timeout)
272 {
273 spin_lock_bh(&inst->lock);
274 inst->flushtimeout = timeout;
275 spin_unlock_bh(&inst->lock);
276
277 return 0;
278 }
279
280 static int
281 nfulnl_set_qthresh(struct nfulnl_instance *inst, u_int32_t qthresh)
282 {
283 spin_lock_bh(&inst->lock);
284 inst->qthreshold = qthresh;
285 spin_unlock_bh(&inst->lock);
286
287 return 0;
288 }
289
290 static int
291 nfulnl_set_flags(struct nfulnl_instance *inst, u_int16_t flags)
292 {
293 spin_lock_bh(&inst->lock);
294 inst->flags = flags;
295 spin_unlock_bh(&inst->lock);
296
297 return 0;
298 }
299
300 static struct sk_buff *
301 nfulnl_alloc_skb(unsigned int inst_size, unsigned int pkt_size)
302 {
303 struct sk_buff *skb;
304 unsigned int n;
305
306 /* alloc skb which should be big enough for a whole multipart
307 * message. WARNING: has to be <= 128k due to slab restrictions */
308
309 n = max(inst_size, pkt_size);
310 skb = alloc_skb(n, GFP_ATOMIC);
311 if (!skb) {
312 if (n > pkt_size) {
313 /* try to allocate only as much as we need for current
314 * packet */
315
316 skb = alloc_skb(pkt_size, GFP_ATOMIC);
317 if (!skb)
318 pr_err("nfnetlink_log: can't even alloc %u bytes\n",
319 pkt_size);
320 }
321 }
322
323 return skb;
324 }
325
326 static int
327 __nfulnl_send(struct nfulnl_instance *inst)
328 {
329 int status = -1;
330
331 if (inst->qlen > 1) {
332 struct nlmsghdr *nlh = nlmsg_put(inst->skb, 0, 0,
333 NLMSG_DONE,
334 sizeof(struct nfgenmsg),
335 0);
336 if (!nlh)
337 goto out;
338 }
339 status = nfnetlink_unicast(inst->skb, &init_net, inst->peer_portid,
340 MSG_DONTWAIT);
341
342 inst->qlen = 0;
343 inst->skb = NULL;
344 out:
345 return status;
346 }
347
348 static void
349 __nfulnl_flush(struct nfulnl_instance *inst)
350 {
351 /* timer holds a reference */
352 if (del_timer(&inst->timer))
353 instance_put(inst);
354 if (inst->skb)
355 __nfulnl_send(inst);
356 }
357
358 static void
359 nfulnl_timer(unsigned long data)
360 {
361 struct nfulnl_instance *inst = (struct nfulnl_instance *)data;
362
363 spin_lock_bh(&inst->lock);
364 if (inst->skb)
365 __nfulnl_send(inst);
366 spin_unlock_bh(&inst->lock);
367 instance_put(inst);
368 }
369
370 /* This is an inline function, we don't really care about a long
371 * list of arguments */
372 static inline int
373 __build_packet_message(struct nfulnl_instance *inst,
374 const struct sk_buff *skb,
375 unsigned int data_len,
376 u_int8_t pf,
377 unsigned int hooknum,
378 const struct net_device *indev,
379 const struct net_device *outdev,
380 const char *prefix, unsigned int plen)
381 {
382 struct nfulnl_msg_packet_hdr pmsg;
383 struct nlmsghdr *nlh;
384 struct nfgenmsg *nfmsg;
385 sk_buff_data_t old_tail = inst->skb->tail;
386
387 nlh = nlmsg_put(inst->skb, 0, 0,
388 NFNL_SUBSYS_ULOG << 8 | NFULNL_MSG_PACKET,
389 sizeof(struct nfgenmsg), 0);
390 if (!nlh)
391 return -1;
392 nfmsg = nlmsg_data(nlh);
393 nfmsg->nfgen_family = pf;
394 nfmsg->version = NFNETLINK_V0;
395 nfmsg->res_id = htons(inst->group_num);
396
397 pmsg.hw_protocol = skb->protocol;
398 pmsg.hook = hooknum;
399
400 if (nla_put(inst->skb, NFULA_PACKET_HDR, sizeof(pmsg), &pmsg))
401 goto nla_put_failure;
402
403 if (prefix &&
404 nla_put(inst->skb, NFULA_PREFIX, plen, prefix))
405 goto nla_put_failure;
406
407 if (indev) {
408 #ifndef CONFIG_BRIDGE_NETFILTER
409 if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
410 htonl(indev->ifindex)))
411 goto nla_put_failure;
412 #else
413 if (pf == PF_BRIDGE) {
414 /* Case 1: outdev is physical input device, we need to
415 * look for bridge group (when called from
416 * netfilter_bridge) */
417 if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
418 htonl(indev->ifindex)) ||
419 /* this is the bridge group "brX" */
420 /* rcu_read_lock()ed by nf_hook_slow or nf_log_packet */
421 nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
422 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
423 goto nla_put_failure;
424 } else {
425 /* Case 2: indev is bridge group, we need to look for
426 * physical device (when called from ipv4) */
427 if (nla_put_be32(inst->skb, NFULA_IFINDEX_INDEV,
428 htonl(indev->ifindex)))
429 goto nla_put_failure;
430 if (skb->nf_bridge && skb->nf_bridge->physindev &&
431 nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSINDEV,
432 htonl(skb->nf_bridge->physindev->ifindex)))
433 goto nla_put_failure;
434 }
435 #endif
436 }
437
438 if (outdev) {
439 #ifndef CONFIG_BRIDGE_NETFILTER
440 if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
441 htonl(outdev->ifindex)))
442 goto nla_put_failure;
443 #else
444 if (pf == PF_BRIDGE) {
445 /* Case 1: outdev is physical output device, we need to
446 * look for bridge group (when called from
447 * netfilter_bridge) */
448 if (nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
449 htonl(outdev->ifindex)) ||
450 /* this is the bridge group "brX" */
451 /* rcu_read_lock()ed by nf_hook_slow or nf_log_packet */
452 nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
453 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
454 goto nla_put_failure;
455 } else {
456 /* Case 2: indev is a bridge group, we need to look
457 * for physical device (when called from ipv4) */
458 if (nla_put_be32(inst->skb, NFULA_IFINDEX_OUTDEV,
459 htonl(outdev->ifindex)))
460 goto nla_put_failure;
461 if (skb->nf_bridge && skb->nf_bridge->physoutdev &&
462 nla_put_be32(inst->skb, NFULA_IFINDEX_PHYSOUTDEV,
463 htonl(skb->nf_bridge->physoutdev->ifindex)))
464 goto nla_put_failure;
465 }
466 #endif
467 }
468
469 if (skb->mark &&
470 nla_put_be32(inst->skb, NFULA_MARK, htonl(skb->mark)))
471 goto nla_put_failure;
472
473 if (indev && skb->dev &&
474 skb->mac_header != skb->network_header) {
475 struct nfulnl_msg_packet_hw phw;
476 int len = dev_parse_header(skb, phw.hw_addr);
477 if (len > 0) {
478 phw.hw_addrlen = htons(len);
479 if (nla_put(inst->skb, NFULA_HWADDR, sizeof(phw), &phw))
480 goto nla_put_failure;
481 }
482 }
483
484 if (indev && skb_mac_header_was_set(skb)) {
485 if (nla_put_be16(inst->skb, NFULA_HWTYPE, htons(skb->dev->type)) ||
486 nla_put_be16(inst->skb, NFULA_HWLEN,
487 htons(skb->dev->hard_header_len)) ||
488 nla_put(inst->skb, NFULA_HWHEADER, skb->dev->hard_header_len,
489 skb_mac_header(skb)))
490 goto nla_put_failure;
491 }
492
493 if (skb->tstamp.tv64) {
494 struct nfulnl_msg_packet_timestamp ts;
495 struct timeval tv = ktime_to_timeval(skb->tstamp);
496 ts.sec = cpu_to_be64(tv.tv_sec);
497 ts.usec = cpu_to_be64(tv.tv_usec);
498
499 if (nla_put(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts))
500 goto nla_put_failure;
501 }
502
503 /* UID */
504 if (skb->sk) {
505 read_lock_bh(&skb->sk->sk_callback_lock);
506 if (skb->sk->sk_socket && skb->sk->sk_socket->file) {
507 struct file *file = skb->sk->sk_socket->file;
508 __be32 uid = htonl(from_kuid_munged(inst->peer_user_ns,
509 file->f_cred->fsuid));
510 __be32 gid = htonl(from_kgid_munged(inst->peer_user_ns,
511 file->f_cred->fsgid));
512 /* need to unlock here since NLA_PUT may goto */
513 read_unlock_bh(&skb->sk->sk_callback_lock);
514 if (nla_put_be32(inst->skb, NFULA_UID, uid) ||
515 nla_put_be32(inst->skb, NFULA_GID, gid))
516 goto nla_put_failure;
517 } else
518 read_unlock_bh(&skb->sk->sk_callback_lock);
519 }
520
521 /* local sequence number */
522 if ((inst->flags & NFULNL_CFG_F_SEQ) &&
523 nla_put_be32(inst->skb, NFULA_SEQ, htonl(inst->seq++)))
524 goto nla_put_failure;
525
526 /* global sequence number */
527 if ((inst->flags & NFULNL_CFG_F_SEQ_GLOBAL) &&
528 nla_put_be32(inst->skb, NFULA_SEQ_GLOBAL,
529 htonl(atomic_inc_return(&global_seq))))
530 goto nla_put_failure;
531
532 if (data_len) {
533 struct nlattr *nla;
534 int size = nla_attr_size(data_len);
535
536 if (skb_tailroom(inst->skb) < nla_total_size(data_len)) {
537 printk(KERN_WARNING "nfnetlink_log: no tailroom!\n");
538 return -1;
539 }
540
541 nla = (struct nlattr *)skb_put(inst->skb, nla_total_size(data_len));
542 nla->nla_type = NFULA_PAYLOAD;
543 nla->nla_len = size;
544
545 if (skb_copy_bits(skb, 0, nla_data(nla), data_len))
546 BUG();
547 }
548
549 nlh->nlmsg_len = inst->skb->tail - old_tail;
550 return 0;
551
552 nla_put_failure:
553 PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n");
554 return -1;
555 }
556
557 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
558
559 static struct nf_loginfo default_loginfo = {
560 .type = NF_LOG_TYPE_ULOG,
561 .u = {
562 .ulog = {
563 .copy_len = 0xffff,
564 .group = 0,
565 .qthreshold = 1,
566 },
567 },
568 };
569
570 /* log handler for internal netfilter logging api */
571 void
572 nfulnl_log_packet(u_int8_t pf,
573 unsigned int hooknum,
574 const struct sk_buff *skb,
575 const struct net_device *in,
576 const struct net_device *out,
577 const struct nf_loginfo *li_user,
578 const char *prefix)
579 {
580 unsigned int size, data_len;
581 struct nfulnl_instance *inst;
582 const struct nf_loginfo *li;
583 unsigned int qthreshold;
584 unsigned int plen;
585
586 if (li_user && li_user->type == NF_LOG_TYPE_ULOG)
587 li = li_user;
588 else
589 li = &default_loginfo;
590
591 inst = instance_lookup_get(li->u.ulog.group);
592 if (!inst)
593 return;
594
595 plen = 0;
596 if (prefix)
597 plen = strlen(prefix) + 1;
598
599 /* FIXME: do we want to make the size calculation conditional based on
600 * what is actually present? way more branches and checks, but more
601 * memory efficient... */
602 size = NLMSG_SPACE(sizeof(struct nfgenmsg))
603 + nla_total_size(sizeof(struct nfulnl_msg_packet_hdr))
604 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
605 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
606 #ifdef CONFIG_BRIDGE_NETFILTER
607 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
608 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
609 #endif
610 + nla_total_size(sizeof(u_int32_t)) /* mark */
611 + nla_total_size(sizeof(u_int32_t)) /* uid */
612 + nla_total_size(sizeof(u_int32_t)) /* gid */
613 + nla_total_size(plen) /* prefix */
614 + nla_total_size(sizeof(struct nfulnl_msg_packet_hw))
615 + nla_total_size(sizeof(struct nfulnl_msg_packet_timestamp));
616
617 if (in && skb_mac_header_was_set(skb)) {
618 size += nla_total_size(skb->dev->hard_header_len)
619 + nla_total_size(sizeof(u_int16_t)) /* hwtype */
620 + nla_total_size(sizeof(u_int16_t)); /* hwlen */
621 }
622
623 spin_lock_bh(&inst->lock);
624
625 if (inst->flags & NFULNL_CFG_F_SEQ)
626 size += nla_total_size(sizeof(u_int32_t));
627 if (inst->flags & NFULNL_CFG_F_SEQ_GLOBAL)
628 size += nla_total_size(sizeof(u_int32_t));
629
630 qthreshold = inst->qthreshold;
631 /* per-rule qthreshold overrides per-instance */
632 if (li->u.ulog.qthreshold)
633 if (qthreshold > li->u.ulog.qthreshold)
634 qthreshold = li->u.ulog.qthreshold;
635
636
637 switch (inst->copy_mode) {
638 case NFULNL_COPY_META:
639 case NFULNL_COPY_NONE:
640 data_len = 0;
641 break;
642
643 case NFULNL_COPY_PACKET:
644 if (inst->copy_range == 0
645 || inst->copy_range > skb->len)
646 data_len = skb->len;
647 else
648 data_len = inst->copy_range;
649
650 size += nla_total_size(data_len);
651 break;
652
653 case NFULNL_COPY_DISABLED:
654 default:
655 goto unlock_and_release;
656 }
657
658 if (inst->skb &&
659 size > skb_tailroom(inst->skb) - sizeof(struct nfgenmsg)) {
660 /* either the queue len is too high or we don't have
661 * enough room in the skb left. flush to userspace. */
662 __nfulnl_flush(inst);
663 }
664
665 if (!inst->skb) {
666 inst->skb = nfulnl_alloc_skb(inst->nlbufsiz, size);
667 if (!inst->skb)
668 goto alloc_failure;
669 }
670
671 inst->qlen++;
672
673 __build_packet_message(inst, skb, data_len, pf,
674 hooknum, in, out, prefix, plen);
675
676 if (inst->qlen >= qthreshold)
677 __nfulnl_flush(inst);
678 /* timer_pending always called within inst->lock, so there
679 * is no chance of a race here */
680 else if (!timer_pending(&inst->timer)) {
681 instance_get(inst);
682 inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100);
683 add_timer(&inst->timer);
684 }
685
686 unlock_and_release:
687 spin_unlock_bh(&inst->lock);
688 instance_put(inst);
689 return;
690
691 alloc_failure:
692 /* FIXME: statistics */
693 goto unlock_and_release;
694 }
695 EXPORT_SYMBOL_GPL(nfulnl_log_packet);
696
697 static int
698 nfulnl_rcv_nl_event(struct notifier_block *this,
699 unsigned long event, void *ptr)
700 {
701 struct netlink_notify *n = ptr;
702
703 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
704 int i;
705
706 /* destroy all instances for this portid */
707 spin_lock_bh(&instances_lock);
708 for (i = 0; i < INSTANCE_BUCKETS; i++) {
709 struct hlist_node *tmp, *t2;
710 struct nfulnl_instance *inst;
711 struct hlist_head *head = &instance_table[i];
712
713 hlist_for_each_entry_safe(inst, tmp, t2, head, hlist) {
714 if ((net_eq(n->net, &init_net)) &&
715 (n->portid == inst->peer_portid))
716 __instance_destroy(inst);
717 }
718 }
719 spin_unlock_bh(&instances_lock);
720 }
721 return NOTIFY_DONE;
722 }
723
724 static struct notifier_block nfulnl_rtnl_notifier = {
725 .notifier_call = nfulnl_rcv_nl_event,
726 };
727
728 static int
729 nfulnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
730 const struct nlmsghdr *nlh,
731 const struct nlattr * const nfqa[])
732 {
733 return -ENOTSUPP;
734 }
735
736 static struct nf_logger nfulnl_logger __read_mostly = {
737 .name = "nfnetlink_log",
738 .logfn = &nfulnl_log_packet,
739 .me = THIS_MODULE,
740 };
741
742 static const struct nla_policy nfula_cfg_policy[NFULA_CFG_MAX+1] = {
743 [NFULA_CFG_CMD] = { .len = sizeof(struct nfulnl_msg_config_cmd) },
744 [NFULA_CFG_MODE] = { .len = sizeof(struct nfulnl_msg_config_mode) },
745 [NFULA_CFG_TIMEOUT] = { .type = NLA_U32 },
746 [NFULA_CFG_QTHRESH] = { .type = NLA_U32 },
747 [NFULA_CFG_NLBUFSIZ] = { .type = NLA_U32 },
748 [NFULA_CFG_FLAGS] = { .type = NLA_U16 },
749 };
750
751 static int
752 nfulnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
753 const struct nlmsghdr *nlh,
754 const struct nlattr * const nfula[])
755 {
756 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
757 u_int16_t group_num = ntohs(nfmsg->res_id);
758 struct nfulnl_instance *inst;
759 struct nfulnl_msg_config_cmd *cmd = NULL;
760 int ret = 0;
761
762 if (nfula[NFULA_CFG_CMD]) {
763 u_int8_t pf = nfmsg->nfgen_family;
764 cmd = nla_data(nfula[NFULA_CFG_CMD]);
765
766 /* Commands without queue context */
767 switch (cmd->command) {
768 case NFULNL_CFG_CMD_PF_BIND:
769 return nf_log_bind_pf(pf, &nfulnl_logger);
770 case NFULNL_CFG_CMD_PF_UNBIND:
771 nf_log_unbind_pf(pf);
772 return 0;
773 }
774 }
775
776 inst = instance_lookup_get(group_num);
777 if (inst && inst->peer_portid != NETLINK_CB(skb).portid) {
778 ret = -EPERM;
779 goto out_put;
780 }
781
782 if (cmd != NULL) {
783 switch (cmd->command) {
784 case NFULNL_CFG_CMD_BIND:
785 if (inst) {
786 ret = -EBUSY;
787 goto out_put;
788 }
789
790 inst = instance_create(group_num,
791 NETLINK_CB(skb).portid,
792 sk_user_ns(NETLINK_CB(skb).ssk));
793 if (IS_ERR(inst)) {
794 ret = PTR_ERR(inst);
795 goto out;
796 }
797 break;
798 case NFULNL_CFG_CMD_UNBIND:
799 if (!inst) {
800 ret = -ENODEV;
801 goto out;
802 }
803
804 instance_destroy(inst);
805 goto out_put;
806 default:
807 ret = -ENOTSUPP;
808 break;
809 }
810 }
811
812 if (nfula[NFULA_CFG_MODE]) {
813 struct nfulnl_msg_config_mode *params;
814 params = nla_data(nfula[NFULA_CFG_MODE]);
815
816 if (!inst) {
817 ret = -ENODEV;
818 goto out;
819 }
820 nfulnl_set_mode(inst, params->copy_mode,
821 ntohl(params->copy_range));
822 }
823
824 if (nfula[NFULA_CFG_TIMEOUT]) {
825 __be32 timeout = nla_get_be32(nfula[NFULA_CFG_TIMEOUT]);
826
827 if (!inst) {
828 ret = -ENODEV;
829 goto out;
830 }
831 nfulnl_set_timeout(inst, ntohl(timeout));
832 }
833
834 if (nfula[NFULA_CFG_NLBUFSIZ]) {
835 __be32 nlbufsiz = nla_get_be32(nfula[NFULA_CFG_NLBUFSIZ]);
836
837 if (!inst) {
838 ret = -ENODEV;
839 goto out;
840 }
841 nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz));
842 }
843
844 if (nfula[NFULA_CFG_QTHRESH]) {
845 __be32 qthresh = nla_get_be32(nfula[NFULA_CFG_QTHRESH]);
846
847 if (!inst) {
848 ret = -ENODEV;
849 goto out;
850 }
851 nfulnl_set_qthresh(inst, ntohl(qthresh));
852 }
853
854 if (nfula[NFULA_CFG_FLAGS]) {
855 __be16 flags = nla_get_be16(nfula[NFULA_CFG_FLAGS]);
856
857 if (!inst) {
858 ret = -ENODEV;
859 goto out;
860 }
861 nfulnl_set_flags(inst, ntohs(flags));
862 }
863
864 out_put:
865 instance_put(inst);
866 out:
867 return ret;
868 }
869
870 static const struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = {
871 [NFULNL_MSG_PACKET] = { .call = nfulnl_recv_unsupp,
872 .attr_count = NFULA_MAX, },
873 [NFULNL_MSG_CONFIG] = { .call = nfulnl_recv_config,
874 .attr_count = NFULA_CFG_MAX,
875 .policy = nfula_cfg_policy },
876 };
877
878 static const struct nfnetlink_subsystem nfulnl_subsys = {
879 .name = "log",
880 .subsys_id = NFNL_SUBSYS_ULOG,
881 .cb_count = NFULNL_MSG_MAX,
882 .cb = nfulnl_cb,
883 };
884
885 #ifdef CONFIG_PROC_FS
886 struct iter_state {
887 unsigned int bucket;
888 };
889
890 static struct hlist_node *get_first(struct iter_state *st)
891 {
892 if (!st)
893 return NULL;
894
895 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
896 if (!hlist_empty(&instance_table[st->bucket]))
897 return rcu_dereference_bh(hlist_first_rcu(&instance_table[st->bucket]));
898 }
899 return NULL;
900 }
901
902 static struct hlist_node *get_next(struct iter_state *st, struct hlist_node *h)
903 {
904 h = rcu_dereference_bh(hlist_next_rcu(h));
905 while (!h) {
906 if (++st->bucket >= INSTANCE_BUCKETS)
907 return NULL;
908
909 h = rcu_dereference_bh(hlist_first_rcu(&instance_table[st->bucket]));
910 }
911 return h;
912 }
913
914 static struct hlist_node *get_idx(struct iter_state *st, loff_t pos)
915 {
916 struct hlist_node *head;
917 head = get_first(st);
918
919 if (head)
920 while (pos && (head = get_next(st, head)))
921 pos--;
922 return pos ? NULL : head;
923 }
924
925 static void *seq_start(struct seq_file *seq, loff_t *pos)
926 __acquires(rcu_bh)
927 {
928 rcu_read_lock_bh();
929 return get_idx(seq->private, *pos);
930 }
931
932 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
933 {
934 (*pos)++;
935 return get_next(s->private, v);
936 }
937
938 static void seq_stop(struct seq_file *s, void *v)
939 __releases(rcu_bh)
940 {
941 rcu_read_unlock_bh();
942 }
943
944 static int seq_show(struct seq_file *s, void *v)
945 {
946 const struct nfulnl_instance *inst = v;
947
948 return seq_printf(s, "%5d %6d %5d %1d %5d %6d %2d\n",
949 inst->group_num,
950 inst->peer_portid, inst->qlen,
951 inst->copy_mode, inst->copy_range,
952 inst->flushtimeout, atomic_read(&inst->use));
953 }
954
955 static const struct seq_operations nful_seq_ops = {
956 .start = seq_start,
957 .next = seq_next,
958 .stop = seq_stop,
959 .show = seq_show,
960 };
961
962 static int nful_open(struct inode *inode, struct file *file)
963 {
964 return seq_open_private(file, &nful_seq_ops,
965 sizeof(struct iter_state));
966 }
967
968 static const struct file_operations nful_file_ops = {
969 .owner = THIS_MODULE,
970 .open = nful_open,
971 .read = seq_read,
972 .llseek = seq_lseek,
973 .release = seq_release_private,
974 };
975
976 #endif /* PROC_FS */
977
978 static int __init nfnetlink_log_init(void)
979 {
980 int i, status = -ENOMEM;
981
982 for (i = 0; i < INSTANCE_BUCKETS; i++)
983 INIT_HLIST_HEAD(&instance_table[i]);
984
985 /* it's not really all that important to have a random value, so
986 * we can do this from the init function, even if there hasn't
987 * been that much entropy yet */
988 get_random_bytes(&hash_init, sizeof(hash_init));
989
990 netlink_register_notifier(&nfulnl_rtnl_notifier);
991 status = nfnetlink_subsys_register(&nfulnl_subsys);
992 if (status < 0) {
993 printk(KERN_ERR "log: failed to create netlink socket\n");
994 goto cleanup_netlink_notifier;
995 }
996
997 status = nf_log_register(NFPROTO_UNSPEC, &nfulnl_logger);
998 if (status < 0) {
999 printk(KERN_ERR "log: failed to register logger\n");
1000 goto cleanup_subsys;
1001 }
1002
1003 #ifdef CONFIG_PROC_FS
1004 if (!proc_create("nfnetlink_log", 0440,
1005 proc_net_netfilter, &nful_file_ops)) {
1006 status = -ENOMEM;
1007 goto cleanup_logger;
1008 }
1009 #endif
1010 return status;
1011
1012 #ifdef CONFIG_PROC_FS
1013 cleanup_logger:
1014 nf_log_unregister(&nfulnl_logger);
1015 #endif
1016 cleanup_subsys:
1017 nfnetlink_subsys_unregister(&nfulnl_subsys);
1018 cleanup_netlink_notifier:
1019 netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1020 return status;
1021 }
1022
1023 static void __exit nfnetlink_log_fini(void)
1024 {
1025 nf_log_unregister(&nfulnl_logger);
1026 #ifdef CONFIG_PROC_FS
1027 remove_proc_entry("nfnetlink_log", proc_net_netfilter);
1028 #endif
1029 nfnetlink_subsys_unregister(&nfulnl_subsys);
1030 netlink_unregister_notifier(&nfulnl_rtnl_notifier);
1031 }
1032
1033 MODULE_DESCRIPTION("netfilter userspace logging");
1034 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1035 MODULE_LICENSE("GPL");
1036 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_ULOG);
1037
1038 module_init(nfnetlink_log_init);
1039 module_exit(nfnetlink_log_fini);