1 /*
2  * This is a module which is used for queueing packets and communicating with
3  * userspace via nfnetlink.
4  *
5  * (C) 2005 by Harald Welte <[email protected]>
6  * (C) 2007 by Patrick McHardy <[email protected]>
7  *
8  * Based on the old ipv4-only ip_queue.c:
9  * (C) 2000-2002 James Morris <[email protected]>
10  * (C) 2003-2005 Netfilter Core Team <[email protected]>
11  *
12  * This program is free software; you can redistribute it and/or modify
13  * it under the terms of the GNU General Public License version 2 as
14  * published by the Free Software Foundation.
15  *
16  */
17 #include <linux/module.h>
18 #include <linux/skbuff.h>
19 #include <linux/init.h>
20 #include <linux/spinlock.h>
21 #include <linux/slab.h>
22 #include <linux/notifier.h>
23 #include <linux/netdevice.h>
24 #include <linux/netfilter.h>
25 #include <linux/proc_fs.h>
26 #include <linux/netfilter_ipv4.h>
27 #include <linux/netfilter_ipv6.h>
28 #include <linux/netfilter/nfnetlink.h>
29 #include <linux/netfilter/nfnetlink_queue.h>
30 #include <linux/list.h>
31 #include <net/sock.h>
32 #include <net/netfilter/nf_queue.h>
33 
34 #include <asm/atomic.h>
35 
36 #ifdef CONFIG_BRIDGE_NETFILTER
37 #include "../bridge/br_private.h"
38 #endif
39 
40 #define NFQNL_QMAX_DEFAULT 1024
41 
42 struct nfqnl_instance {
43 	struct hlist_node hlist;		/* global list of queues */
44 	struct rcu_head rcu;
45 
46 	int peer_pid;
47 	unsigned int queue_maxlen;
48 	unsigned int copy_range;
49 	unsigned int queue_dropped;
50 	unsigned int queue_user_dropped;
51 
52 
53 	u_int16_t queue_num;			/* number of this queue */
54 	u_int8_t copy_mode;
55 /*
56  * Following fields are dirtied for each queued packet,
57  * keep them in same cache line if possible.
58  */
59 	spinlock_t	lock;
60 	unsigned int	queue_total;
61 	atomic_t	id_sequence;		/* 'sequence' of pkt ids */
62 	struct list_head queue_list;		/* packets in queue */
63 };
64 
65 typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
66 
67 static DEFINE_SPINLOCK(instances_lock);
68 
69 #define INSTANCE_BUCKETS	16
70 static struct hlist_head instance_table[INSTANCE_BUCKETS] __read_mostly;
71 
72 static inline u_int8_t instance_hashfn(u_int16_t queue_num)
73 {
74 	return ((queue_num >> 8) | queue_num) % INSTANCE_BUCKETS;
75 }
76 
77 static struct nfqnl_instance *
78 instance_lookup(u_int16_t queue_num)
79 {
80 	struct hlist_head *head;
81 	struct hlist_node *pos;
82 	struct nfqnl_instance *inst;
83 
84 	head = &instance_table[instance_hashfn(queue_num)];
85 	hlist_for_each_entry_rcu(inst, pos, head, hlist) {
86 		if (inst->queue_num == queue_num)
87 			return inst;
88 	}
89 	return NULL;
90 }
91 
92 static struct nfqnl_instance *
93 instance_create(u_int16_t queue_num, int pid)
94 {
95 	struct nfqnl_instance *inst;
96 	unsigned int h;
97 	int err;
98 
99 	spin_lock(&instances_lock);
100 	if (instance_lookup(queue_num)) {
101 		err = -EEXIST;
102 		goto out_unlock;
103 	}
104 
105 	inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
106 	if (!inst) {
107 		err = -ENOMEM;
108 		goto out_unlock;
109 	}
110 
111 	inst->queue_num = queue_num;
112 	inst->peer_pid = pid;
113 	inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
114 	inst->copy_range = 0xfffff;
115 	inst->copy_mode = NFQNL_COPY_NONE;
116 	spin_lock_init(&inst->lock);
117 	INIT_LIST_HEAD(&inst->queue_list);
118 
119 	if (!try_module_get(THIS_MODULE)) {
120 		err = -EAGAIN;
121 		goto out_free;
122 	}
123 
124 	h = instance_hashfn(queue_num);
125 	hlist_add_head_rcu(&inst->hlist, &instance_table[h]);
126 
127 	spin_unlock(&instances_lock);
128 
129 	return inst;
130 
131 out_free:
132 	kfree(inst);
133 out_unlock:
134 	spin_unlock(&instances_lock);
135 	return ERR_PTR(err);
136 }
137 
138 static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
139 			unsigned long data);
140 
141 static void
142 instance_destroy_rcu(struct rcu_head *head)
143 {
144 	struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
145 						   rcu);
146 
147 	nfqnl_flush(inst, NULL, 0);
148 	kfree(inst);
149 	module_put(THIS_MODULE);
150 }
151 
152 static void
153 __instance_destroy(struct nfqnl_instance *inst)
154 {
155 	hlist_del_rcu(&inst->hlist);
156 	call_rcu(&inst->rcu, instance_destroy_rcu);
157 }
158 
159 static void
160 instance_destroy(struct nfqnl_instance *inst)
161 {
162 	spin_lock(&instances_lock);
163 	__instance_destroy(inst);
164 	spin_unlock(&instances_lock);
165 }
166 
167 static inline void
168 __enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
169 {
170        list_add_tail(&entry->list, &queue->queue_list);
171        queue->queue_total++;
172 }
173 
174 static struct nf_queue_entry *
175 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
176 {
177 	struct nf_queue_entry *entry = NULL, *i;
178 
179 	spin_lock_bh(&queue->lock);
180 
181 	list_for_each_entry(i, &queue->queue_list, list) {
182 		if (i->id == id) {
183 			entry = i;
184 			break;
185 		}
186 	}
187 
188 	if (entry) {
189 		list_del(&entry->list);
190 		queue->queue_total--;
191 	}
192 
193 	spin_unlock_bh(&queue->lock);
194 
195 	return entry;
196 }
197 
198 static void
199 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
200 {
201 	struct nf_queue_entry *entry, *next;
202 
203 	spin_lock_bh(&queue->lock);
204 	list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
205 		if (!cmpfn || cmpfn(entry, data)) {
206 			list_del(&entry->list);
207 			queue->queue_total--;
208 			nf_reinject(entry, NF_DROP);
209 		}
210 	}
211 	spin_unlock_bh(&queue->lock);
212 }
213 
214 static struct sk_buff *
215 nfqnl_build_packet_message(struct nfqnl_instance *queue,
216 			   struct nf_queue_entry *entry)
217 {
218 	sk_buff_data_t old_tail;
219 	size_t size;
220 	size_t data_len = 0;
221 	struct sk_buff *skb;
222 	struct nfqnl_msg_packet_hdr pmsg;
223 	struct nlmsghdr *nlh;
224 	struct nfgenmsg *nfmsg;
225 	struct sk_buff *entskb = entry->skb;
226 	struct net_device *indev;
227 	struct net_device *outdev;
228 
229 	size =    NLMSG_SPACE(sizeof(struct nfgenmsg))
230 		+ nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
231 		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
232 		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
233 #ifdef CONFIG_BRIDGE_NETFILTER
234 		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
235 		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
236 #endif
237 		+ nla_total_size(sizeof(u_int32_t))	/* mark */
238 		+ nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
239 		+ nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
240 
241 	outdev = entry->outdev;
242 
243 	switch ((enum nfqnl_config_mode)ACCESS_ONCE(queue->copy_mode)) {
244 	case NFQNL_COPY_META:
245 	case NFQNL_COPY_NONE:
246 		break;
247 
248 	case NFQNL_COPY_PACKET:
249 		if (entskb->ip_summed == CHECKSUM_PARTIAL &&
250 		    skb_checksum_help(entskb))
251 			return NULL;
252 
253 		data_len = ACCESS_ONCE(queue->copy_range);
254 		if (data_len == 0 || data_len > entskb->len)
255 			data_len = entskb->len;
256 
257 		size += nla_total_size(data_len);
258 		break;
259 	}
260 
261 
262 	skb = alloc_skb(size, GFP_ATOMIC);
263 	if (!skb)
264 		goto nlmsg_failure;
265 
266 	old_tail = skb->tail;
267 	nlh = NLMSG_PUT(skb, 0, 0,
268 			NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
269 			sizeof(struct nfgenmsg));
270 	nfmsg = NLMSG_DATA(nlh);
271 	nfmsg->nfgen_family = entry->pf;
272 	nfmsg->version = NFNETLINK_V0;
273 	nfmsg->res_id = htons(queue->queue_num);
274 
275 	entry->id = atomic_inc_return(&queue->id_sequence);
276 	pmsg.packet_id 		= htonl(entry->id);
277 	pmsg.hw_protocol	= entskb->protocol;
278 	pmsg.hook		= entry->hook;
279 
280 	NLA_PUT(skb, NFQA_PACKET_HDR, sizeof(pmsg), &pmsg);
281 
282 	indev = entry->indev;
283 	if (indev) {
284 #ifndef CONFIG_BRIDGE_NETFILTER
285 		NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex));
286 #else
287 		if (entry->pf == PF_BRIDGE) {
288 			/* Case 1: indev is physical input device, we need to
289 			 * look for bridge group (when called from
290 			 * netfilter_bridge) */
291 			NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSINDEV,
292 				     htonl(indev->ifindex));
293 			/* this is the bridge group "brX" */
294 			/* rcu_read_lock()ed by __nf_queue */
295 			NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV,
296 				     htonl(br_port_get_rcu(indev)->br->dev->ifindex));
297 		} else {
298 			/* Case 2: indev is bridge group, we need to look for
299 			 * physical device (when called from ipv4) */
300 			NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV,
301 				     htonl(indev->ifindex));
302 			if (entskb->nf_bridge && entskb->nf_bridge->physindev)
303 				NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSINDEV,
304 					     htonl(entskb->nf_bridge->physindev->ifindex));
305 		}
306 #endif
307 	}
308 
309 	if (outdev) {
310 #ifndef CONFIG_BRIDGE_NETFILTER
311 		NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex));
312 #else
313 		if (entry->pf == PF_BRIDGE) {
314 			/* Case 1: outdev is physical output device, we need to
315 			 * look for bridge group (when called from
316 			 * netfilter_bridge) */
317 			NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSOUTDEV,
318 				     htonl(outdev->ifindex));
319 			/* this is the bridge group "brX" */
320 			/* rcu_read_lock()ed by __nf_queue */
321 			NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV,
322 				     htonl(br_port_get_rcu(outdev)->br->dev->ifindex));
323 		} else {
324 			/* Case 2: outdev is bridge group, we need to look for
325 			 * physical output device (when called from ipv4) */
326 			NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV,
327 				     htonl(outdev->ifindex));
328 			if (entskb->nf_bridge && entskb->nf_bridge->physoutdev)
329 				NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSOUTDEV,
330 					     htonl(entskb->nf_bridge->physoutdev->ifindex));
331 		}
332 #endif
333 	}
334 
335 	if (entskb->mark)
336 		NLA_PUT_BE32(skb, NFQA_MARK, htonl(entskb->mark));
337 
338 	if (indev && entskb->dev) {
339 		struct nfqnl_msg_packet_hw phw;
340 		int len = dev_parse_header(entskb, phw.hw_addr);
341 		if (len) {
342 			phw.hw_addrlen = htons(len);
343 			NLA_PUT(skb, NFQA_HWADDR, sizeof(phw), &phw);
344 		}
345 	}
346 
347 	if (entskb->tstamp.tv64) {
348 		struct nfqnl_msg_packet_timestamp ts;
349 		struct timeval tv = ktime_to_timeval(entskb->tstamp);
350 		ts.sec = cpu_to_be64(tv.tv_sec);
351 		ts.usec = cpu_to_be64(tv.tv_usec);
352 
353 		NLA_PUT(skb, NFQA_TIMESTAMP, sizeof(ts), &ts);
354 	}
355 
356 	if (data_len) {
357 		struct nlattr *nla;
358 		int sz = nla_attr_size(data_len);
359 
360 		if (skb_tailroom(skb) < nla_total_size(data_len)) {
361 			printk(KERN_WARNING "nf_queue: no tailroom!\n");
362 			goto nlmsg_failure;
363 		}
364 
365 		nla = (struct nlattr *)skb_put(skb, nla_total_size(data_len));
366 		nla->nla_type = NFQA_PAYLOAD;
367 		nla->nla_len = sz;
368 
369 		if (skb_copy_bits(entskb, 0, nla_data(nla), data_len))
370 			BUG();
371 	}
372 
373 	nlh->nlmsg_len = skb->tail - old_tail;
374 	return skb;
375 
376 nlmsg_failure:
377 nla_put_failure:
378 	if (skb)
379 		kfree_skb(skb);
380 	if (net_ratelimit())
381 		printk(KERN_ERR "nf_queue: error creating packet message\n");
382 	return NULL;
383 }
384 
385 static int
386 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
387 {
388 	struct sk_buff *nskb;
389 	struct nfqnl_instance *queue;
390 	int err = -ENOBUFS;
391 
392 	/* rcu_read_lock()ed by nf_hook_slow() */
393 	queue = instance_lookup(queuenum);
394 	if (!queue) {
395 		err = -ESRCH;
396 		goto err_out;
397 	}
398 
399 	if (queue->copy_mode == NFQNL_COPY_NONE) {
400 		err = -EINVAL;
401 		goto err_out;
402 	}
403 
404 	nskb = nfqnl_build_packet_message(queue, entry);
405 	if (nskb == NULL) {
406 		err = -ENOMEM;
407 		goto err_out;
408 	}
409 	spin_lock_bh(&queue->lock);
410 
411 	if (!queue->peer_pid) {
412 		err = -EINVAL;
413 		goto err_out_free_nskb;
414 	}
415 	if (queue->queue_total >= queue->queue_maxlen) {
416 		queue->queue_dropped++;
417 		if (net_ratelimit())
418 			  printk(KERN_WARNING "nf_queue: full at %d entries, "
419 				 "dropping packets(s).\n",
420 				 queue->queue_total);
421 		goto err_out_free_nskb;
422 	}
423 
424 	/* nfnetlink_unicast will either free the nskb or add it to a socket */
425 	err = nfnetlink_unicast(nskb, &init_net, queue->peer_pid, MSG_DONTWAIT);
426 	if (err < 0) {
427 		queue->queue_user_dropped++;
428 		goto err_out_unlock;
429 	}
430 
431 	__enqueue_entry(queue, entry);
432 
433 	spin_unlock_bh(&queue->lock);
434 	return 0;
435 
436 err_out_free_nskb:
437 	kfree_skb(nskb);
438 err_out_unlock:
439 	spin_unlock_bh(&queue->lock);
440 err_out:
441 	return err;
442 }
443 
444 static int
445 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e)
446 {
447 	struct sk_buff *nskb;
448 	int diff;
449 
450 	diff = data_len - e->skb->len;
451 	if (diff < 0) {
452 		if (pskb_trim(e->skb, data_len))
453 			return -ENOMEM;
454 	} else if (diff > 0) {
455 		if (data_len > 0xFFFF)
456 			return -EINVAL;
457 		if (diff > skb_tailroom(e->skb)) {
458 			nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
459 					       diff, GFP_ATOMIC);
460 			if (!nskb) {
461 				printk(KERN_WARNING "nf_queue: OOM "
462 				      "in mangle, dropping packet\n");
463 				return -ENOMEM;
464 			}
465 			kfree_skb(e->skb);
466 			e->skb = nskb;
467 		}
468 		skb_put(e->skb, diff);
469 	}
470 	if (!skb_make_writable(e->skb, data_len))
471 		return -ENOMEM;
472 	skb_copy_to_linear_data(e->skb, data, data_len);
473 	e->skb->ip_summed = CHECKSUM_NONE;
474 	return 0;
475 }
476 
477 static int
478 nfqnl_set_mode(struct nfqnl_instance *queue,
479 	       unsigned char mode, unsigned int range)
480 {
481 	int status = 0;
482 
483 	spin_lock_bh(&queue->lock);
484 	switch (mode) {
485 	case NFQNL_COPY_NONE:
486 	case NFQNL_COPY_META:
487 		queue->copy_mode = mode;
488 		queue->copy_range = 0;
489 		break;
490 
491 	case NFQNL_COPY_PACKET:
492 		queue->copy_mode = mode;
493 		/* we're using struct nlattr which has 16bit nla_len */
494 		if (range > 0xffff)
495 			queue->copy_range = 0xffff;
496 		else
497 			queue->copy_range = range;
498 		break;
499 
500 	default:
501 		status = -EINVAL;
502 
503 	}
504 	spin_unlock_bh(&queue->lock);
505 
506 	return status;
507 }
508 
509 static int
510 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
511 {
512 	if (entry->indev)
513 		if (entry->indev->ifindex == ifindex)
514 			return 1;
515 	if (entry->outdev)
516 		if (entry->outdev->ifindex == ifindex)
517 			return 1;
518 #ifdef CONFIG_BRIDGE_NETFILTER
519 	if (entry->skb->nf_bridge) {
520 		if (entry->skb->nf_bridge->physindev &&
521 		    entry->skb->nf_bridge->physindev->ifindex == ifindex)
522 			return 1;
523 		if (entry->skb->nf_bridge->physoutdev &&
524 		    entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
525 			return 1;
526 	}
527 #endif
528 	return 0;
529 }
530 
531 /* drop all packets with either indev or outdev == ifindex from all queue
532  * instances */
533 static void
534 nfqnl_dev_drop(int ifindex)
535 {
536 	int i;
537 
538 	rcu_read_lock();
539 
540 	for (i = 0; i < INSTANCE_BUCKETS; i++) {
541 		struct hlist_node *tmp;
542 		struct nfqnl_instance *inst;
543 		struct hlist_head *head = &instance_table[i];
544 
545 		hlist_for_each_entry_rcu(inst, tmp, head, hlist)
546 			nfqnl_flush(inst, dev_cmp, ifindex);
547 	}
548 
549 	rcu_read_unlock();
550 }
551 
552 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
553 
554 static int
555 nfqnl_rcv_dev_event(struct notifier_block *this,
556 		    unsigned long event, void *ptr)
557 {
558 	struct net_device *dev = ptr;
559 
560 	if (!net_eq(dev_net(dev), &init_net))
561 		return NOTIFY_DONE;
562 
563 	/* Drop any packets associated with the downed device */
564 	if (event == NETDEV_DOWN)
565 		nfqnl_dev_drop(dev->ifindex);
566 	return NOTIFY_DONE;
567 }
568 
569 static struct notifier_block nfqnl_dev_notifier = {
570 	.notifier_call	= nfqnl_rcv_dev_event,
571 };
572 
573 static int
574 nfqnl_rcv_nl_event(struct notifier_block *this,
575 		   unsigned long event, void *ptr)
576 {
577 	struct netlink_notify *n = ptr;
578 
579 	if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
580 		int i;
581 
582 		/* destroy all instances for this pid */
583 		spin_lock(&instances_lock);
584 		for (i = 0; i < INSTANCE_BUCKETS; i++) {
585 			struct hlist_node *tmp, *t2;
586 			struct nfqnl_instance *inst;
587 			struct hlist_head *head = &instance_table[i];
588 
589 			hlist_for_each_entry_safe(inst, tmp, t2, head, hlist) {
590 				if ((n->net == &init_net) &&
591 				    (n->pid == inst->peer_pid))
592 					__instance_destroy(inst);
593 			}
594 		}
595 		spin_unlock(&instances_lock);
596 	}
597 	return NOTIFY_DONE;
598 }
599 
600 static struct notifier_block nfqnl_rtnl_notifier = {
601 	.notifier_call	= nfqnl_rcv_nl_event,
602 };
603 
604 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
605 	[NFQA_VERDICT_HDR]	= { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
606 	[NFQA_MARK]		= { .type = NLA_U32 },
607 	[NFQA_PAYLOAD]		= { .type = NLA_UNSPEC },
608 };
609 
610 static int
611 nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
612 		   const struct nlmsghdr *nlh,
613 		   const struct nlattr * const nfqa[])
614 {
615 	struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
616 	u_int16_t queue_num = ntohs(nfmsg->res_id);
617 
618 	struct nfqnl_msg_verdict_hdr *vhdr;
619 	struct nfqnl_instance *queue;
620 	unsigned int verdict;
621 	struct nf_queue_entry *entry;
622 	int err;
623 
624 	rcu_read_lock();
625 	queue = instance_lookup(queue_num);
626 	if (!queue) {
627 		err = -ENODEV;
628 		goto err_out_unlock;
629 	}
630 
631 	if (queue->peer_pid != NETLINK_CB(skb).pid) {
632 		err = -EPERM;
633 		goto err_out_unlock;
634 	}
635 
636 	if (!nfqa[NFQA_VERDICT_HDR]) {
637 		err = -EINVAL;
638 		goto err_out_unlock;
639 	}
640 
641 	vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
642 	verdict = ntohl(vhdr->verdict);
643 
644 	if ((verdict & NF_VERDICT_MASK) > NF_MAX_VERDICT) {
645 		err = -EINVAL;
646 		goto err_out_unlock;
647 	}
648 
649 	entry = find_dequeue_entry(queue, ntohl(vhdr->id));
650 	if (entry == NULL) {
651 		err = -ENOENT;
652 		goto err_out_unlock;
653 	}
654 	rcu_read_unlock();
655 
656 	if (nfqa[NFQA_PAYLOAD]) {
657 		if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
658 				 nla_len(nfqa[NFQA_PAYLOAD]), entry) < 0)
659 			verdict = NF_DROP;
660 	}
661 
662 	if (nfqa[NFQA_MARK])
663 		entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
664 
665 	nf_reinject(entry, verdict);
666 	return 0;
667 
668 err_out_unlock:
669 	rcu_read_unlock();
670 	return err;
671 }
672 
673 static int
674 nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
675 		  const struct nlmsghdr *nlh,
676 		  const struct nlattr * const nfqa[])
677 {
678 	return -ENOTSUPP;
679 }
680 
681 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
682 	[NFQA_CFG_CMD]		= { .len = sizeof(struct nfqnl_msg_config_cmd) },
683 	[NFQA_CFG_PARAMS]	= { .len = sizeof(struct nfqnl_msg_config_params) },
684 };
685 
686 static const struct nf_queue_handler nfqh = {
687 	.name 	= "nf_queue",
688 	.outfn	= &nfqnl_enqueue_packet,
689 };
690 
691 static int
692 nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
693 		  const struct nlmsghdr *nlh,
694 		  const struct nlattr * const nfqa[])
695 {
696 	struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
697 	u_int16_t queue_num = ntohs(nfmsg->res_id);
698 	struct nfqnl_instance *queue;
699 	struct nfqnl_msg_config_cmd *cmd = NULL;
700 	int ret = 0;
701 
702 	if (nfqa[NFQA_CFG_CMD]) {
703 		cmd = nla_data(nfqa[NFQA_CFG_CMD]);
704 
705 		/* Commands without queue context - might sleep */
706 		switch (cmd->command) {
707 		case NFQNL_CFG_CMD_PF_BIND:
708 			return nf_register_queue_handler(ntohs(cmd->pf),
709 							 &nfqh);
710 		case NFQNL_CFG_CMD_PF_UNBIND:
711 			return nf_unregister_queue_handler(ntohs(cmd->pf),
712 							   &nfqh);
713 		}
714 	}
715 
716 	rcu_read_lock();
717 	queue = instance_lookup(queue_num);
718 	if (queue && queue->peer_pid != NETLINK_CB(skb).pid) {
719 		ret = -EPERM;
720 		goto err_out_unlock;
721 	}
722 
723 	if (cmd != NULL) {
724 		switch (cmd->command) {
725 		case NFQNL_CFG_CMD_BIND:
726 			if (queue) {
727 				ret = -EBUSY;
728 				goto err_out_unlock;
729 			}
730 			queue = instance_create(queue_num, NETLINK_CB(skb).pid);
731 			if (IS_ERR(queue)) {
732 				ret = PTR_ERR(queue);
733 				goto err_out_unlock;
734 			}
735 			break;
736 		case NFQNL_CFG_CMD_UNBIND:
737 			if (!queue) {
738 				ret = -ENODEV;
739 				goto err_out_unlock;
740 			}
741 			instance_destroy(queue);
742 			break;
743 		case NFQNL_CFG_CMD_PF_BIND:
744 		case NFQNL_CFG_CMD_PF_UNBIND:
745 			break;
746 		default:
747 			ret = -ENOTSUPP;
748 			break;
749 		}
750 	}
751 
752 	if (nfqa[NFQA_CFG_PARAMS]) {
753 		struct nfqnl_msg_config_params *params;
754 
755 		if (!queue) {
756 			ret = -ENODEV;
757 			goto err_out_unlock;
758 		}
759 		params = nla_data(nfqa[NFQA_CFG_PARAMS]);
760 		nfqnl_set_mode(queue, params->copy_mode,
761 				ntohl(params->copy_range));
762 	}
763 
764 	if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
765 		__be32 *queue_maxlen;
766 
767 		if (!queue) {
768 			ret = -ENODEV;
769 			goto err_out_unlock;
770 		}
771 		queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
772 		spin_lock_bh(&queue->lock);
773 		queue->queue_maxlen = ntohl(*queue_maxlen);
774 		spin_unlock_bh(&queue->lock);
775 	}
776 
777 err_out_unlock:
778 	rcu_read_unlock();
779 	return ret;
780 }
781 
782 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
783 	[NFQNL_MSG_PACKET]	= { .call = nfqnl_recv_unsupp,
784 				    .attr_count = NFQA_MAX, },
785 	[NFQNL_MSG_VERDICT]	= { .call = nfqnl_recv_verdict,
786 				    .attr_count = NFQA_MAX,
787 				    .policy = nfqa_verdict_policy },
788 	[NFQNL_MSG_CONFIG]	= { .call = nfqnl_recv_config,
789 				    .attr_count = NFQA_CFG_MAX,
790 				    .policy = nfqa_cfg_policy },
791 };
792 
793 static const struct nfnetlink_subsystem nfqnl_subsys = {
794 	.name		= "nf_queue",
795 	.subsys_id	= NFNL_SUBSYS_QUEUE,
796 	.cb_count	= NFQNL_MSG_MAX,
797 	.cb		= nfqnl_cb,
798 };
799 
800 #ifdef CONFIG_PROC_FS
801 struct iter_state {
802 	unsigned int bucket;
803 };
804 
805 static struct hlist_node *get_first(struct seq_file *seq)
806 {
807 	struct iter_state *st = seq->private;
808 
809 	if (!st)
810 		return NULL;
811 
812 	for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
813 		if (!hlist_empty(&instance_table[st->bucket]))
814 			return instance_table[st->bucket].first;
815 	}
816 	return NULL;
817 }
818 
819 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
820 {
821 	struct iter_state *st = seq->private;
822 
823 	h = h->next;
824 	while (!h) {
825 		if (++st->bucket >= INSTANCE_BUCKETS)
826 			return NULL;
827 
828 		h = instance_table[st->bucket].first;
829 	}
830 	return h;
831 }
832 
833 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
834 {
835 	struct hlist_node *head;
836 	head = get_first(seq);
837 
838 	if (head)
839 		while (pos && (head = get_next(seq, head)))
840 			pos--;
841 	return pos ? NULL : head;
842 }
843 
844 static void *seq_start(struct seq_file *seq, loff_t *pos)
845 	__acquires(instances_lock)
846 {
847 	spin_lock(&instances_lock);
848 	return get_idx(seq, *pos);
849 }
850 
851 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
852 {
853 	(*pos)++;
854 	return get_next(s, v);
855 }
856 
857 static void seq_stop(struct seq_file *s, void *v)
858 	__releases(instances_lock)
859 {
860 	spin_unlock(&instances_lock);
861 }
862 
863 static int seq_show(struct seq_file *s, void *v)
864 {
865 	const struct nfqnl_instance *inst = v;
866 
867 	return seq_printf(s, "%5d %6d %5d %1d %5d %5d %5d %8d %2d\n",
868 			  inst->queue_num,
869 			  inst->peer_pid, inst->queue_total,
870 			  inst->copy_mode, inst->copy_range,
871 			  inst->queue_dropped, inst->queue_user_dropped,
872 			  atomic_read(&inst->id_sequence), 1);
873 }
874 
875 static const struct seq_operations nfqnl_seq_ops = {
876 	.start	= seq_start,
877 	.next	= seq_next,
878 	.stop	= seq_stop,
879 	.show	= seq_show,
880 };
881 
882 static int nfqnl_open(struct inode *inode, struct file *file)
883 {
884 	return seq_open_private(file, &nfqnl_seq_ops,
885 			sizeof(struct iter_state));
886 }
887 
888 static const struct file_operations nfqnl_file_ops = {
889 	.owner	 = THIS_MODULE,
890 	.open	 = nfqnl_open,
891 	.read	 = seq_read,
892 	.llseek	 = seq_lseek,
893 	.release = seq_release_private,
894 };
895 
896 #endif /* PROC_FS */
897 
898 static int __init nfnetlink_queue_init(void)
899 {
900 	int i, status = -ENOMEM;
901 
902 	for (i = 0; i < INSTANCE_BUCKETS; i++)
903 		INIT_HLIST_HEAD(&instance_table[i]);
904 
905 	netlink_register_notifier(&nfqnl_rtnl_notifier);
906 	status = nfnetlink_subsys_register(&nfqnl_subsys);
907 	if (status < 0) {
908 		printk(KERN_ERR "nf_queue: failed to create netlink socket\n");
909 		goto cleanup_netlink_notifier;
910 	}
911 
912 #ifdef CONFIG_PROC_FS
913 	if (!proc_create("nfnetlink_queue", 0440,
914 			 proc_net_netfilter, &nfqnl_file_ops))
915 		goto cleanup_subsys;
916 #endif
917 
918 	register_netdevice_notifier(&nfqnl_dev_notifier);
919 	return status;
920 
921 #ifdef CONFIG_PROC_FS
922 cleanup_subsys:
923 	nfnetlink_subsys_unregister(&nfqnl_subsys);
924 #endif
925 cleanup_netlink_notifier:
926 	netlink_unregister_notifier(&nfqnl_rtnl_notifier);
927 	return status;
928 }
929 
930 static void __exit nfnetlink_queue_fini(void)
931 {
932 	nf_unregister_queue_handlers(&nfqh);
933 	unregister_netdevice_notifier(&nfqnl_dev_notifier);
934 #ifdef CONFIG_PROC_FS
935 	remove_proc_entry("nfnetlink_queue", proc_net_netfilter);
936 #endif
937 	nfnetlink_subsys_unregister(&nfqnl_subsys);
938 	netlink_unregister_notifier(&nfqnl_rtnl_notifier);
939 
940 	rcu_barrier(); /* Wait for completion of call_rcu()'s */
941 }
942 
943 MODULE_DESCRIPTION("netfilter packet queue handler");
944 MODULE_AUTHOR("Harald Welte <[email protected]>");
945 MODULE_LICENSE("GPL");
946 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
947 
948 module_init(nfnetlink_queue_init);
949 module_exit(nfnetlink_queue_fini);
950