1 /*
2  * This is a module which is used for queueing packets and communicating with
3  * userspace via nfetlink.
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/notifier.h>
22 #include <linux/netdevice.h>
23 #include <linux/netfilter.h>
24 #include <linux/proc_fs.h>
25 #include <linux/netfilter_ipv4.h>
26 #include <linux/netfilter_ipv6.h>
27 #include <linux/netfilter/nfnetlink.h>
28 #include <linux/netfilter/nfnetlink_queue.h>
29 #include <linux/list.h>
30 #include <net/sock.h>
31 #include <net/netfilter/nf_queue.h>
32 
33 #include <asm/atomic.h>
34 
35 #ifdef CONFIG_BRIDGE_NETFILTER
36 #include "../bridge/br_private.h"
37 #endif
38 
39 #define NFQNL_QMAX_DEFAULT 1024
40 
41 struct nfqnl_instance {
42 	struct hlist_node hlist;		/* global list of queues */
43 	struct rcu_head rcu;
44 
45 	int peer_pid;
46 	unsigned int queue_maxlen;
47 	unsigned int copy_range;
48 	unsigned int queue_total;
49 	unsigned int queue_dropped;
50 	unsigned int queue_user_dropped;
51 
52 	unsigned int id_sequence;		/* 'sequence' of pkt ids */
53 
54 	u_int16_t queue_num;			/* number of this queue */
55 	u_int8_t copy_mode;
56 
57 	spinlock_t lock;
58 
59 	struct list_head queue_list;		/* packets in queue */
60 };
61 
62 typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
63 
64 static DEFINE_SPINLOCK(instances_lock);
65 
66 #define INSTANCE_BUCKETS	16
67 static struct hlist_head instance_table[INSTANCE_BUCKETS] __read_mostly;
68 
69 static inline u_int8_t instance_hashfn(u_int16_t queue_num)
70 {
71 	return ((queue_num >> 8) | queue_num) % INSTANCE_BUCKETS;
72 }
73 
74 static struct nfqnl_instance *
75 instance_lookup(u_int16_t queue_num)
76 {
77 	struct hlist_head *head;
78 	struct hlist_node *pos;
79 	struct nfqnl_instance *inst;
80 
81 	head = &instance_table[instance_hashfn(queue_num)];
82 	hlist_for_each_entry_rcu(inst, pos, head, hlist) {
83 		if (inst->queue_num == queue_num)
84 			return inst;
85 	}
86 	return NULL;
87 }
88 
89 static struct nfqnl_instance *
90 instance_create(u_int16_t queue_num, int pid)
91 {
92 	struct nfqnl_instance *inst;
93 	unsigned int h;
94 	int err;
95 
96 	spin_lock(&instances_lock);
97 	if (instance_lookup(queue_num)) {
98 		err = -EEXIST;
99 		goto out_unlock;
100 	}
101 
102 	inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
103 	if (!inst) {
104 		err = -ENOMEM;
105 		goto out_unlock;
106 	}
107 
108 	inst->queue_num = queue_num;
109 	inst->peer_pid = pid;
110 	inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
111 	inst->copy_range = 0xfffff;
112 	inst->copy_mode = NFQNL_COPY_NONE;
113 	spin_lock_init(&inst->lock);
114 	INIT_LIST_HEAD(&inst->queue_list);
115 	INIT_RCU_HEAD(&inst->rcu);
116 
117 	if (!try_module_get(THIS_MODULE)) {
118 		err = -EAGAIN;
119 		goto out_free;
120 	}
121 
122 	h = instance_hashfn(queue_num);
123 	hlist_add_head_rcu(&inst->hlist, &instance_table[h]);
124 
125 	spin_unlock(&instances_lock);
126 
127 	return inst;
128 
129 out_free:
130 	kfree(inst);
131 out_unlock:
132 	spin_unlock(&instances_lock);
133 	return ERR_PTR(err);
134 }
135 
136 static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
137 			unsigned long data);
138 
139 static void
140 instance_destroy_rcu(struct rcu_head *head)
141 {
142 	struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
143 						   rcu);
144 
145 	nfqnl_flush(inst, NULL, 0);
146 	kfree(inst);
147 	module_put(THIS_MODULE);
148 }
149 
150 static void
151 __instance_destroy(struct nfqnl_instance *inst)
152 {
153 	hlist_del_rcu(&inst->hlist);
154 	call_rcu(&inst->rcu, instance_destroy_rcu);
155 }
156 
157 static void
158 instance_destroy(struct nfqnl_instance *inst)
159 {
160 	spin_lock(&instances_lock);
161 	__instance_destroy(inst);
162 	spin_unlock(&instances_lock);
163 }
164 
165 static inline void
166 __enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
167 {
168        list_add_tail(&entry->list, &queue->queue_list);
169        queue->queue_total++;
170 }
171 
172 static struct nf_queue_entry *
173 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
174 {
175 	struct nf_queue_entry *entry = NULL, *i;
176 
177 	spin_lock_bh(&queue->lock);
178 
179 	list_for_each_entry(i, &queue->queue_list, list) {
180 		if (i->id == id) {
181 			entry = i;
182 			break;
183 		}
184 	}
185 
186 	if (entry) {
187 		list_del(&entry->list);
188 		queue->queue_total--;
189 	}
190 
191 	spin_unlock_bh(&queue->lock);
192 
193 	return entry;
194 }
195 
196 static void
197 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
198 {
199 	struct nf_queue_entry *entry, *next;
200 
201 	spin_lock_bh(&queue->lock);
202 	list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
203 		if (!cmpfn || cmpfn(entry, data)) {
204 			list_del(&entry->list);
205 			queue->queue_total--;
206 			nf_reinject(entry, NF_DROP);
207 		}
208 	}
209 	spin_unlock_bh(&queue->lock);
210 }
211 
212 static struct sk_buff *
213 nfqnl_build_packet_message(struct nfqnl_instance *queue,
214 			   struct nf_queue_entry *entry)
215 {
216 	sk_buff_data_t old_tail;
217 	size_t size;
218 	size_t data_len = 0;
219 	struct sk_buff *skb;
220 	struct nfqnl_msg_packet_hdr pmsg;
221 	struct nlmsghdr *nlh;
222 	struct nfgenmsg *nfmsg;
223 	struct sk_buff *entskb = entry->skb;
224 	struct net_device *indev;
225 	struct net_device *outdev;
226 
227 	size =    NLMSG_ALIGN(sizeof(struct nfgenmsg))
228 		+ nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
229 		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
230 		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
231 #ifdef CONFIG_BRIDGE_NETFILTER
232 		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
233 		+ nla_total_size(sizeof(u_int32_t))	/* ifindex */
234 #endif
235 		+ nla_total_size(sizeof(u_int32_t))	/* mark */
236 		+ nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
237 		+ nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
238 
239 	outdev = entry->outdev;
240 
241 	spin_lock_bh(&queue->lock);
242 
243 	switch ((enum nfqnl_config_mode)queue->copy_mode) {
244 	case NFQNL_COPY_META:
245 	case NFQNL_COPY_NONE:
246 		data_len = 0;
247 		break;
248 
249 	case NFQNL_COPY_PACKET:
250 		if ((entskb->ip_summed == CHECKSUM_PARTIAL ||
251 		     entskb->ip_summed == CHECKSUM_COMPLETE) &&
252 		    skb_checksum_help(entskb)) {
253 			spin_unlock_bh(&queue->lock);
254 			return NULL;
255 		}
256 		if (queue->copy_range == 0
257 		    || queue->copy_range > entskb->len)
258 			data_len = entskb->len;
259 		else
260 			data_len = queue->copy_range;
261 
262 		size += nla_total_size(data_len);
263 		break;
264 	}
265 
266 	entry->id = queue->id_sequence++;
267 
268 	spin_unlock_bh(&queue->lock);
269 
270 	skb = alloc_skb(size, GFP_ATOMIC);
271 	if (!skb)
272 		goto nlmsg_failure;
273 
274 	old_tail = skb->tail;
275 	nlh = NLMSG_PUT(skb, 0, 0,
276 			NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
277 			sizeof(struct nfgenmsg));
278 	nfmsg = NLMSG_DATA(nlh);
279 	nfmsg->nfgen_family = entry->pf;
280 	nfmsg->version = NFNETLINK_V0;
281 	nfmsg->res_id = htons(queue->queue_num);
282 
283 	pmsg.packet_id 		= htonl(entry->id);
284 	pmsg.hw_protocol	= entskb->protocol;
285 	pmsg.hook		= entry->hook;
286 
287 	NLA_PUT(skb, NFQA_PACKET_HDR, sizeof(pmsg), &pmsg);
288 
289 	indev = entry->indev;
290 	if (indev) {
291 #ifndef CONFIG_BRIDGE_NETFILTER
292 		NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex));
293 #else
294 		if (entry->pf == PF_BRIDGE) {
295 			/* Case 1: indev is physical input device, we need to
296 			 * look for bridge group (when called from
297 			 * netfilter_bridge) */
298 			NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSINDEV,
299 				     htonl(indev->ifindex));
300 			/* this is the bridge group "brX" */
301 			NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV,
302 				     htonl(indev->br_port->br->dev->ifindex));
303 		} else {
304 			/* Case 2: indev is bridge group, we need to look for
305 			 * physical device (when called from ipv4) */
306 			NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV,
307 				     htonl(indev->ifindex));
308 			if (entskb->nf_bridge && entskb->nf_bridge->physindev)
309 				NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSINDEV,
310 					     htonl(entskb->nf_bridge->physindev->ifindex));
311 		}
312 #endif
313 	}
314 
315 	if (outdev) {
316 #ifndef CONFIG_BRIDGE_NETFILTER
317 		NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex));
318 #else
319 		if (entry->pf == PF_BRIDGE) {
320 			/* Case 1: outdev is physical output device, we need to
321 			 * look for bridge group (when called from
322 			 * netfilter_bridge) */
323 			NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSOUTDEV,
324 				     htonl(outdev->ifindex));
325 			/* this is the bridge group "brX" */
326 			NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV,
327 				     htonl(outdev->br_port->br->dev->ifindex));
328 		} else {
329 			/* Case 2: outdev is bridge group, we need to look for
330 			 * physical output device (when called from ipv4) */
331 			NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV,
332 				     htonl(outdev->ifindex));
333 			if (entskb->nf_bridge && entskb->nf_bridge->physoutdev)
334 				NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSOUTDEV,
335 					     htonl(entskb->nf_bridge->physoutdev->ifindex));
336 		}
337 #endif
338 	}
339 
340 	if (entskb->mark)
341 		NLA_PUT_BE32(skb, NFQA_MARK, htonl(entskb->mark));
342 
343 	if (indev && entskb->dev) {
344 		struct nfqnl_msg_packet_hw phw;
345 		int len = dev_parse_header(entskb, phw.hw_addr);
346 		if (len) {
347 			phw.hw_addrlen = htons(len);
348 			NLA_PUT(skb, NFQA_HWADDR, sizeof(phw), &phw);
349 		}
350 	}
351 
352 	if (entskb->tstamp.tv64) {
353 		struct nfqnl_msg_packet_timestamp ts;
354 		struct timeval tv = ktime_to_timeval(entskb->tstamp);
355 		ts.sec = cpu_to_be64(tv.tv_sec);
356 		ts.usec = cpu_to_be64(tv.tv_usec);
357 
358 		NLA_PUT(skb, NFQA_TIMESTAMP, sizeof(ts), &ts);
359 	}
360 
361 	if (data_len) {
362 		struct nlattr *nla;
363 		int sz = nla_attr_size(data_len);
364 
365 		if (skb_tailroom(skb) < nla_total_size(data_len)) {
366 			printk(KERN_WARNING "nf_queue: no tailroom!\n");
367 			goto nlmsg_failure;
368 		}
369 
370 		nla = (struct nlattr *)skb_put(skb, nla_total_size(data_len));
371 		nla->nla_type = NFQA_PAYLOAD;
372 		nla->nla_len = sz;
373 
374 		if (skb_copy_bits(entskb, 0, nla_data(nla), data_len))
375 			BUG();
376 	}
377 
378 	nlh->nlmsg_len = skb->tail - old_tail;
379 	return skb;
380 
381 nlmsg_failure:
382 nla_put_failure:
383 	if (skb)
384 		kfree_skb(skb);
385 	if (net_ratelimit())
386 		printk(KERN_ERR "nf_queue: error creating packet message\n");
387 	return NULL;
388 }
389 
390 static int
391 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
392 {
393 	struct sk_buff *nskb;
394 	struct nfqnl_instance *queue;
395 	int err;
396 
397 	/* rcu_read_lock()ed by nf_hook_slow() */
398 	queue = instance_lookup(queuenum);
399 	if (!queue)
400 		goto err_out;
401 
402 	if (queue->copy_mode == NFQNL_COPY_NONE)
403 		goto err_out;
404 
405 	nskb = nfqnl_build_packet_message(queue, entry);
406 	if (nskb == NULL)
407 		goto err_out;
408 
409 	spin_lock_bh(&queue->lock);
410 
411 	if (!queue->peer_pid)
412 		goto err_out_free_nskb;
413 
414 	if (queue->queue_total >= queue->queue_maxlen) {
415 		queue->queue_dropped++;
416 		if (net_ratelimit())
417 			  printk(KERN_WARNING "nf_queue: full at %d entries, "
418 				 "dropping packets(s). Dropped: %d\n",
419 				 queue->queue_total, queue->queue_dropped);
420 		goto err_out_free_nskb;
421 	}
422 
423 	/* nfnetlink_unicast will either free the nskb or add it to a socket */
424 	err = nfnetlink_unicast(nskb, queue->peer_pid, MSG_DONTWAIT);
425 	if (err < 0) {
426 		queue->queue_user_dropped++;
427 		goto err_out_unlock;
428 	}
429 
430 	__enqueue_entry(queue, entry);
431 
432 	spin_unlock_bh(&queue->lock);
433 	return 0;
434 
435 err_out_free_nskb:
436 	kfree_skb(nskb);
437 err_out_unlock:
438 	spin_unlock_bh(&queue->lock);
439 err_out:
440 	return -1;
441 }
442 
443 static int
444 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e)
445 {
446 	struct sk_buff *nskb;
447 	int diff;
448 
449 	diff = data_len - e->skb->len;
450 	if (diff < 0) {
451 		if (pskb_trim(e->skb, data_len))
452 			return -ENOMEM;
453 	} else if (diff > 0) {
454 		if (data_len > 0xFFFF)
455 			return -EINVAL;
456 		if (diff > skb_tailroom(e->skb)) {
457 			nskb = skb_copy_expand(e->skb, 0,
458 					       diff - skb_tailroom(e->skb),
459 					       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 (dev->nd_net != &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 &&
580 	    n->protocol == NETLINK_NETFILTER && n->pid) {
581 		int i;
582 
583 		/* destroy all instances for this pid */
584 		spin_lock(&instances_lock);
585 		for (i = 0; i < INSTANCE_BUCKETS; i++) {
586 			struct hlist_node *tmp, *t2;
587 			struct nfqnl_instance *inst;
588 			struct hlist_head *head = &instance_table[i];
589 
590 			hlist_for_each_entry_safe(inst, tmp, t2, head, hlist) {
591 				if ((n->net == &init_net) &&
592 				    (n->pid == inst->peer_pid))
593 					__instance_destroy(inst);
594 			}
595 		}
596 		spin_unlock(&instances_lock);
597 	}
598 	return NOTIFY_DONE;
599 }
600 
601 static struct notifier_block nfqnl_rtnl_notifier = {
602 	.notifier_call	= nfqnl_rcv_nl_event,
603 };
604 
605 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
606 	[NFQA_VERDICT_HDR]	= { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
607 	[NFQA_MARK]		= { .type = NLA_U32 },
608 	[NFQA_PAYLOAD]		= { .type = NLA_UNSPEC },
609 };
610 
611 static int
612 nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
613 		   struct nlmsghdr *nlh, struct nlattr *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 		  struct nlmsghdr *nlh, struct nlattr *nfqa[])
676 {
677 	return -ENOTSUPP;
678 }
679 
680 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
681 	[NFQA_CFG_CMD]		= { .len = sizeof(struct nfqnl_msg_config_cmd) },
682 	[NFQA_CFG_PARAMS]	= { .len = sizeof(struct nfqnl_msg_config_params) },
683 };
684 
685 static const struct nf_queue_handler nfqh = {
686 	.name 	= "nf_queue",
687 	.outfn	= &nfqnl_enqueue_packet,
688 };
689 
690 static int
691 nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
692 		  struct nlmsghdr *nlh, struct nlattr *nfqa[])
693 {
694 	struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
695 	u_int16_t queue_num = ntohs(nfmsg->res_id);
696 	struct nfqnl_instance *queue;
697 	struct nfqnl_msg_config_cmd *cmd = NULL;
698 	int ret = 0;
699 
700 	if (nfqa[NFQA_CFG_CMD]) {
701 		cmd = nla_data(nfqa[NFQA_CFG_CMD]);
702 
703 		/* Commands without queue context - might sleep */
704 		switch (cmd->command) {
705 		case NFQNL_CFG_CMD_PF_BIND:
706 			ret = nf_register_queue_handler(ntohs(cmd->pf),
707 							&nfqh);
708 			break;
709 		case NFQNL_CFG_CMD_PF_UNBIND:
710 			ret = nf_unregister_queue_handler(ntohs(cmd->pf),
711 							  &nfqh);
712 			break;
713 		default:
714 			break;
715 		}
716 
717 		if (ret < 0)
718 			return ret;
719 	}
720 
721 	rcu_read_lock();
722 	queue = instance_lookup(queue_num);
723 	if (queue && queue->peer_pid != NETLINK_CB(skb).pid) {
724 		ret = -EPERM;
725 		goto err_out_unlock;
726 	}
727 
728 	if (cmd != NULL) {
729 		switch (cmd->command) {
730 		case NFQNL_CFG_CMD_BIND:
731 			if (queue) {
732 				ret = -EBUSY;
733 				goto err_out_unlock;
734 			}
735 			queue = instance_create(queue_num, NETLINK_CB(skb).pid);
736 			if (IS_ERR(queue)) {
737 				ret = PTR_ERR(queue);
738 				goto err_out_unlock;
739 			}
740 			break;
741 		case NFQNL_CFG_CMD_UNBIND:
742 			if (!queue) {
743 				ret = -ENODEV;
744 				goto err_out_unlock;
745 			}
746 			instance_destroy(queue);
747 			break;
748 		case NFQNL_CFG_CMD_PF_BIND:
749 		case NFQNL_CFG_CMD_PF_UNBIND:
750 			break;
751 		default:
752 			ret = -ENOTSUPP;
753 			break;
754 		}
755 	}
756 
757 	if (nfqa[NFQA_CFG_PARAMS]) {
758 		struct nfqnl_msg_config_params *params;
759 
760 		if (!queue) {
761 			ret = -ENODEV;
762 			goto err_out_unlock;
763 		}
764 		params = nla_data(nfqa[NFQA_CFG_PARAMS]);
765 		nfqnl_set_mode(queue, params->copy_mode,
766 				ntohl(params->copy_range));
767 	}
768 
769 	if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
770 		__be32 *queue_maxlen;
771 
772 		if (!queue) {
773 			ret = -ENODEV;
774 			goto err_out_unlock;
775 		}
776 		queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
777 		spin_lock_bh(&queue->lock);
778 		queue->queue_maxlen = ntohl(*queue_maxlen);
779 		spin_unlock_bh(&queue->lock);
780 	}
781 
782 err_out_unlock:
783 	rcu_read_unlock();
784 	return ret;
785 }
786 
787 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
788 	[NFQNL_MSG_PACKET]	= { .call = nfqnl_recv_unsupp,
789 				    .attr_count = NFQA_MAX, },
790 	[NFQNL_MSG_VERDICT]	= { .call = nfqnl_recv_verdict,
791 				    .attr_count = NFQA_MAX,
792 				    .policy = nfqa_verdict_policy },
793 	[NFQNL_MSG_CONFIG]	= { .call = nfqnl_recv_config,
794 				    .attr_count = NFQA_CFG_MAX,
795 				    .policy = nfqa_cfg_policy },
796 };
797 
798 static const struct nfnetlink_subsystem nfqnl_subsys = {
799 	.name		= "nf_queue",
800 	.subsys_id	= NFNL_SUBSYS_QUEUE,
801 	.cb_count	= NFQNL_MSG_MAX,
802 	.cb		= nfqnl_cb,
803 };
804 
805 #ifdef CONFIG_PROC_FS
806 struct iter_state {
807 	unsigned int bucket;
808 };
809 
810 static struct hlist_node *get_first(struct seq_file *seq)
811 {
812 	struct iter_state *st = seq->private;
813 
814 	if (!st)
815 		return NULL;
816 
817 	for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
818 		if (!hlist_empty(&instance_table[st->bucket]))
819 			return instance_table[st->bucket].first;
820 	}
821 	return NULL;
822 }
823 
824 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
825 {
826 	struct iter_state *st = seq->private;
827 
828 	h = h->next;
829 	while (!h) {
830 		if (++st->bucket >= INSTANCE_BUCKETS)
831 			return NULL;
832 
833 		h = instance_table[st->bucket].first;
834 	}
835 	return h;
836 }
837 
838 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
839 {
840 	struct hlist_node *head;
841 	head = get_first(seq);
842 
843 	if (head)
844 		while (pos && (head = get_next(seq, head)))
845 			pos--;
846 	return pos ? NULL : head;
847 }
848 
849 static void *seq_start(struct seq_file *seq, loff_t *pos)
850 	__acquires(instances_lock)
851 {
852 	spin_lock(&instances_lock);
853 	return get_idx(seq, *pos);
854 }
855 
856 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
857 {
858 	(*pos)++;
859 	return get_next(s, v);
860 }
861 
862 static void seq_stop(struct seq_file *s, void *v)
863 	__releases(instances_lock)
864 {
865 	spin_unlock(&instances_lock);
866 }
867 
868 static int seq_show(struct seq_file *s, void *v)
869 {
870 	const struct nfqnl_instance *inst = v;
871 
872 	return seq_printf(s, "%5d %6d %5d %1d %5d %5d %5d %8d %2d\n",
873 			  inst->queue_num,
874 			  inst->peer_pid, inst->queue_total,
875 			  inst->copy_mode, inst->copy_range,
876 			  inst->queue_dropped, inst->queue_user_dropped,
877 			  inst->id_sequence, 1);
878 }
879 
880 static const struct seq_operations nfqnl_seq_ops = {
881 	.start	= seq_start,
882 	.next	= seq_next,
883 	.stop	= seq_stop,
884 	.show	= seq_show,
885 };
886 
887 static int nfqnl_open(struct inode *inode, struct file *file)
888 {
889 	return seq_open_private(file, &nfqnl_seq_ops,
890 			sizeof(struct iter_state));
891 }
892 
893 static const struct file_operations nfqnl_file_ops = {
894 	.owner	 = THIS_MODULE,
895 	.open	 = nfqnl_open,
896 	.read	 = seq_read,
897 	.llseek	 = seq_lseek,
898 	.release = seq_release_private,
899 };
900 
901 #endif /* PROC_FS */
902 
903 static int __init nfnetlink_queue_init(void)
904 {
905 	int i, status = -ENOMEM;
906 #ifdef CONFIG_PROC_FS
907 	struct proc_dir_entry *proc_nfqueue;
908 #endif
909 
910 	for (i = 0; i < INSTANCE_BUCKETS; i++)
911 		INIT_HLIST_HEAD(&instance_table[i]);
912 
913 	netlink_register_notifier(&nfqnl_rtnl_notifier);
914 	status = nfnetlink_subsys_register(&nfqnl_subsys);
915 	if (status < 0) {
916 		printk(KERN_ERR "nf_queue: failed to create netlink socket\n");
917 		goto cleanup_netlink_notifier;
918 	}
919 
920 #ifdef CONFIG_PROC_FS
921 	proc_nfqueue = create_proc_entry("nfnetlink_queue", 0440,
922 					 proc_net_netfilter);
923 	if (!proc_nfqueue)
924 		goto cleanup_subsys;
925 	proc_nfqueue->proc_fops = &nfqnl_file_ops;
926 #endif
927 
928 	register_netdevice_notifier(&nfqnl_dev_notifier);
929 	return status;
930 
931 #ifdef CONFIG_PROC_FS
932 cleanup_subsys:
933 	nfnetlink_subsys_unregister(&nfqnl_subsys);
934 #endif
935 cleanup_netlink_notifier:
936 	netlink_unregister_notifier(&nfqnl_rtnl_notifier);
937 	return status;
938 }
939 
940 static void __exit nfnetlink_queue_fini(void)
941 {
942 	nf_unregister_queue_handlers(&nfqh);
943 	unregister_netdevice_notifier(&nfqnl_dev_notifier);
944 #ifdef CONFIG_PROC_FS
945 	remove_proc_entry("nfnetlink_queue", proc_net_netfilter);
946 #endif
947 	nfnetlink_subsys_unregister(&nfqnl_subsys);
948 	netlink_unregister_notifier(&nfqnl_rtnl_notifier);
949 }
950 
951 MODULE_DESCRIPTION("netfilter packet queue handler");
952 MODULE_AUTHOR("Harald Welte <[email protected]>");
953 MODULE_LICENSE("GPL");
954 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
955 
956 module_init(nfnetlink_queue_init);
957 module_exit(nfnetlink_queue_fini);
958