1
2
3
4
5
6
7
8
9
10
11
12
13
14
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/tcp_states.h>
33#include <net/netfilter/nf_queue.h>
34#include <net/netns/generic.h>
35#include <net/netfilter/nfnetlink_queue.h>
36
37#include <linux/atomic.h>
38
39#ifdef CONFIG_BRIDGE_NETFILTER
40#include "../bridge/br_private.h"
41#endif
42
43#define NFQNL_QMAX_DEFAULT 1024
44
45
46
47
48
49
50
51#define NFQNL_MAX_COPY_RANGE (0xffff - NLA_HDRLEN)
52
53struct nfqnl_instance {
54 struct hlist_node hlist;
55 struct rcu_head rcu;
56
57 int peer_portid;
58 unsigned int queue_maxlen;
59 unsigned int copy_range;
60 unsigned int queue_dropped;
61 unsigned int queue_user_dropped;
62
63
64 u_int16_t queue_num;
65 u_int8_t copy_mode;
66 u_int32_t flags;
67
68
69
70
71 spinlock_t lock;
72 unsigned int queue_total;
73 unsigned int id_sequence;
74 struct list_head queue_list;
75};
76
77typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
78
79static int nfnl_queue_net_id __read_mostly;
80
81#define INSTANCE_BUCKETS 16
82struct nfnl_queue_net {
83 spinlock_t instances_lock;
84 struct hlist_head instance_table[INSTANCE_BUCKETS];
85};
86
87static struct nfnl_queue_net *nfnl_queue_pernet(struct net *net)
88{
89 return net_generic(net, nfnl_queue_net_id);
90}
91
92static inline u_int8_t instance_hashfn(u_int16_t queue_num)
93{
94 return ((queue_num >> 8) ^ queue_num) % INSTANCE_BUCKETS;
95}
96
97static struct nfqnl_instance *
98instance_lookup(struct nfnl_queue_net *q, u_int16_t queue_num)
99{
100 struct hlist_head *head;
101 struct nfqnl_instance *inst;
102
103 head = &q->instance_table[instance_hashfn(queue_num)];
104 hlist_for_each_entry_rcu(inst, head, hlist) {
105 if (inst->queue_num == queue_num)
106 return inst;
107 }
108 return NULL;
109}
110
111static struct nfqnl_instance *
112instance_create(struct nfnl_queue_net *q, u_int16_t queue_num,
113 int portid)
114{
115 struct nfqnl_instance *inst;
116 unsigned int h;
117 int err;
118
119 spin_lock(&q->instances_lock);
120 if (instance_lookup(q, queue_num)) {
121 err = -EEXIST;
122 goto out_unlock;
123 }
124
125 inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
126 if (!inst) {
127 err = -ENOMEM;
128 goto out_unlock;
129 }
130
131 inst->queue_num = queue_num;
132 inst->peer_portid = portid;
133 inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
134 inst->copy_range = NFQNL_MAX_COPY_RANGE;
135 inst->copy_mode = NFQNL_COPY_NONE;
136 spin_lock_init(&inst->lock);
137 INIT_LIST_HEAD(&inst->queue_list);
138
139 if (!try_module_get(THIS_MODULE)) {
140 err = -EAGAIN;
141 goto out_free;
142 }
143
144 h = instance_hashfn(queue_num);
145 hlist_add_head_rcu(&inst->hlist, &q->instance_table[h]);
146
147 spin_unlock(&q->instances_lock);
148
149 return inst;
150
151out_free:
152 kfree(inst);
153out_unlock:
154 spin_unlock(&q->instances_lock);
155 return ERR_PTR(err);
156}
157
158static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
159 unsigned long data);
160
161static void
162instance_destroy_rcu(struct rcu_head *head)
163{
164 struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
165 rcu);
166
167 nfqnl_flush(inst, NULL, 0);
168 kfree(inst);
169 module_put(THIS_MODULE);
170}
171
172static void
173__instance_destroy(struct nfqnl_instance *inst)
174{
175 hlist_del_rcu(&inst->hlist);
176 call_rcu(&inst->rcu, instance_destroy_rcu);
177}
178
179static void
180instance_destroy(struct nfnl_queue_net *q, struct nfqnl_instance *inst)
181{
182 spin_lock(&q->instances_lock);
183 __instance_destroy(inst);
184 spin_unlock(&q->instances_lock);
185}
186
187static inline void
188__enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
189{
190 list_add_tail(&entry->list, &queue->queue_list);
191 queue->queue_total++;
192}
193
194static void
195__dequeue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
196{
197 list_del(&entry->list);
198 queue->queue_total--;
199}
200
201static struct nf_queue_entry *
202find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
203{
204 struct nf_queue_entry *entry = NULL, *i;
205
206 spin_lock_bh(&queue->lock);
207
208 list_for_each_entry(i, &queue->queue_list, list) {
209 if (i->id == id) {
210 entry = i;
211 break;
212 }
213 }
214
215 if (entry)
216 __dequeue_entry(queue, entry);
217
218 spin_unlock_bh(&queue->lock);
219
220 return entry;
221}
222
223static void
224nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
225{
226 struct nf_queue_entry *entry, *next;
227
228 spin_lock_bh(&queue->lock);
229 list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
230 if (!cmpfn || cmpfn(entry, data)) {
231 list_del(&entry->list);
232 queue->queue_total--;
233 nf_reinject(entry, NF_DROP);
234 }
235 }
236 spin_unlock_bh(&queue->lock);
237}
238
239static int
240nfqnl_put_packet_info(struct sk_buff *nlskb, struct sk_buff *packet,
241 bool csum_verify)
242{
243 __u32 flags = 0;
244
245 if (packet->ip_summed == CHECKSUM_PARTIAL)
246 flags = NFQA_SKB_CSUMNOTREADY;
247 else if (csum_verify)
248 flags = NFQA_SKB_CSUM_NOTVERIFIED;
249
250 if (skb_is_gso(packet))
251 flags |= NFQA_SKB_GSO;
252
253 return flags ? nla_put_be32(nlskb, NFQA_SKB_INFO, htonl(flags)) : 0;
254}
255
256static int nfqnl_put_sk_uidgid(struct sk_buff *skb, struct sock *sk)
257{
258 const struct cred *cred;
259
260 if (sk->sk_state == TCP_TIME_WAIT)
261 return 0;
262
263 read_lock_bh(&sk->sk_callback_lock);
264 if (sk->sk_socket && sk->sk_socket->file) {
265 cred = sk->sk_socket->file->f_cred;
266 if (nla_put_be32(skb, NFQA_UID,
267 htonl(from_kuid_munged(&init_user_ns, cred->fsuid))))
268 goto nla_put_failure;
269 if (nla_put_be32(skb, NFQA_GID,
270 htonl(from_kgid_munged(&init_user_ns, cred->fsgid))))
271 goto nla_put_failure;
272 }
273 read_unlock_bh(&sk->sk_callback_lock);
274 return 0;
275
276nla_put_failure:
277 read_unlock_bh(&sk->sk_callback_lock);
278 return -1;
279}
280
281static struct sk_buff *
282nfqnl_build_packet_message(struct net *net, struct nfqnl_instance *queue,
283 struct nf_queue_entry *entry,
284 __be32 **packet_id_ptr)
285{
286 size_t size;
287 size_t data_len = 0, cap_len = 0;
288 unsigned int hlen = 0;
289 struct sk_buff *skb;
290 struct nlattr *nla;
291 struct nfqnl_msg_packet_hdr *pmsg;
292 struct nlmsghdr *nlh;
293 struct nfgenmsg *nfmsg;
294 struct sk_buff *entskb = entry->skb;
295 struct net_device *indev;
296 struct net_device *outdev;
297 struct nf_conn *ct = NULL;
298 enum ip_conntrack_info uninitialized_var(ctinfo);
299 bool csum_verify;
300
301 size = nlmsg_total_size(sizeof(struct nfgenmsg))
302 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
303 + nla_total_size(sizeof(u_int32_t))
304 + nla_total_size(sizeof(u_int32_t))
305#ifdef CONFIG_BRIDGE_NETFILTER
306 + nla_total_size(sizeof(u_int32_t))
307 + nla_total_size(sizeof(u_int32_t))
308#endif
309 + nla_total_size(sizeof(u_int32_t))
310 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
311 + nla_total_size(sizeof(u_int32_t))
312 + nla_total_size(sizeof(u_int32_t));
313
314 if (entskb->tstamp.tv64)
315 size += nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
316
317 if (entry->hook <= NF_INET_FORWARD ||
318 (entry->hook == NF_INET_POST_ROUTING && entskb->sk == NULL))
319 csum_verify = !skb_csum_unnecessary(entskb);
320 else
321 csum_verify = false;
322
323 outdev = entry->outdev;
324
325 switch ((enum nfqnl_config_mode)ACCESS_ONCE(queue->copy_mode)) {
326 case NFQNL_COPY_META:
327 case NFQNL_COPY_NONE:
328 break;
329
330 case NFQNL_COPY_PACKET:
331 if (!(queue->flags & NFQA_CFG_F_GSO) &&
332 entskb->ip_summed == CHECKSUM_PARTIAL &&
333 skb_checksum_help(entskb))
334 return NULL;
335
336 data_len = ACCESS_ONCE(queue->copy_range);
337 if (data_len > entskb->len)
338 data_len = entskb->len;
339
340 hlen = skb_zerocopy_headlen(entskb);
341 hlen = min_t(unsigned int, hlen, data_len);
342 size += sizeof(struct nlattr) + hlen;
343 cap_len = entskb->len;
344 break;
345 }
346
347 if (queue->flags & NFQA_CFG_F_CONNTRACK)
348 ct = nfqnl_ct_get(entskb, &size, &ctinfo);
349
350 if (queue->flags & NFQA_CFG_F_UID_GID) {
351 size += (nla_total_size(sizeof(u_int32_t))
352 + nla_total_size(sizeof(u_int32_t)));
353 }
354
355 skb = nfnetlink_alloc_skb(net, size, queue->peer_portid,
356 GFP_ATOMIC);
357 if (!skb) {
358 skb_tx_error(entskb);
359 return NULL;
360 }
361
362 nlh = nlmsg_put(skb, 0, 0,
363 NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
364 sizeof(struct nfgenmsg), 0);
365 if (!nlh) {
366 skb_tx_error(entskb);
367 kfree_skb(skb);
368 return NULL;
369 }
370 nfmsg = nlmsg_data(nlh);
371 nfmsg->nfgen_family = entry->pf;
372 nfmsg->version = NFNETLINK_V0;
373 nfmsg->res_id = htons(queue->queue_num);
374
375 nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
376 pmsg = nla_data(nla);
377 pmsg->hw_protocol = entskb->protocol;
378 pmsg->hook = entry->hook;
379 *packet_id_ptr = &pmsg->packet_id;
380
381 indev = entry->indev;
382 if (indev) {
383#ifndef CONFIG_BRIDGE_NETFILTER
384 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
385 goto nla_put_failure;
386#else
387 if (entry->pf == PF_BRIDGE) {
388
389
390
391 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
392 htonl(indev->ifindex)) ||
393
394
395 nla_put_be32(skb, NFQA_IFINDEX_INDEV,
396 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
397 goto nla_put_failure;
398 } else {
399
400
401 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
402 htonl(indev->ifindex)))
403 goto nla_put_failure;
404 if (entskb->nf_bridge && entskb->nf_bridge->physindev &&
405 nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
406 htonl(entskb->nf_bridge->physindev->ifindex)))
407 goto nla_put_failure;
408 }
409#endif
410 }
411
412 if (outdev) {
413#ifndef CONFIG_BRIDGE_NETFILTER
414 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
415 goto nla_put_failure;
416#else
417 if (entry->pf == PF_BRIDGE) {
418
419
420
421 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
422 htonl(outdev->ifindex)) ||
423
424
425 nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
426 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
427 goto nla_put_failure;
428 } else {
429
430
431 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
432 htonl(outdev->ifindex)))
433 goto nla_put_failure;
434 if (entskb->nf_bridge && entskb->nf_bridge->physoutdev &&
435 nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
436 htonl(entskb->nf_bridge->physoutdev->ifindex)))
437 goto nla_put_failure;
438 }
439#endif
440 }
441
442 if (entskb->mark &&
443 nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
444 goto nla_put_failure;
445
446 if (indev && entskb->dev &&
447 entskb->mac_header != entskb->network_header) {
448 struct nfqnl_msg_packet_hw phw;
449 int len;
450
451 memset(&phw, 0, sizeof(phw));
452 len = dev_parse_header(entskb, phw.hw_addr);
453 if (len) {
454 phw.hw_addrlen = htons(len);
455 if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
456 goto nla_put_failure;
457 }
458 }
459
460 if (entskb->tstamp.tv64) {
461 struct nfqnl_msg_packet_timestamp ts;
462 struct timeval tv = ktime_to_timeval(entskb->tstamp);
463 ts.sec = cpu_to_be64(tv.tv_sec);
464 ts.usec = cpu_to_be64(tv.tv_usec);
465
466 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
467 goto nla_put_failure;
468 }
469
470 if ((queue->flags & NFQA_CFG_F_UID_GID) && entskb->sk &&
471 nfqnl_put_sk_uidgid(skb, entskb->sk) < 0)
472 goto nla_put_failure;
473
474 if (ct && nfqnl_ct_put(skb, ct, ctinfo) < 0)
475 goto nla_put_failure;
476
477 if (cap_len > data_len &&
478 nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
479 goto nla_put_failure;
480
481 if (nfqnl_put_packet_info(skb, entskb, csum_verify))
482 goto nla_put_failure;
483
484 if (data_len) {
485 struct nlattr *nla;
486
487 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
488 goto nla_put_failure;
489
490 nla = (struct nlattr *)skb_put(skb, sizeof(*nla));
491 nla->nla_type = NFQA_PAYLOAD;
492 nla->nla_len = nla_attr_size(data_len);
493
494 if (skb_zerocopy(skb, entskb, data_len, hlen))
495 goto nla_put_failure;
496 }
497
498 nlh->nlmsg_len = skb->len;
499 return skb;
500
501nla_put_failure:
502 skb_tx_error(entskb);
503 kfree_skb(skb);
504 net_err_ratelimited("nf_queue: error creating packet message\n");
505 return NULL;
506}
507
508static int
509__nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
510 struct nf_queue_entry *entry)
511{
512 struct sk_buff *nskb;
513 int err = -ENOBUFS;
514 __be32 *packet_id_ptr;
515 int failopen = 0;
516
517 nskb = nfqnl_build_packet_message(net, queue, entry, &packet_id_ptr);
518 if (nskb == NULL) {
519 err = -ENOMEM;
520 goto err_out;
521 }
522 spin_lock_bh(&queue->lock);
523
524 if (queue->queue_total >= queue->queue_maxlen) {
525 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
526 failopen = 1;
527 err = 0;
528 } else {
529 queue->queue_dropped++;
530 net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
531 queue->queue_total);
532 }
533 goto err_out_free_nskb;
534 }
535 entry->id = ++queue->id_sequence;
536 *packet_id_ptr = htonl(entry->id);
537
538
539 err = nfnetlink_unicast(nskb, net, queue->peer_portid, MSG_DONTWAIT);
540 if (err < 0) {
541 queue->queue_user_dropped++;
542 goto err_out_unlock;
543 }
544
545 __enqueue_entry(queue, entry);
546
547 spin_unlock_bh(&queue->lock);
548 return 0;
549
550err_out_free_nskb:
551 kfree_skb(nskb);
552err_out_unlock:
553 spin_unlock_bh(&queue->lock);
554 if (failopen)
555 nf_reinject(entry, NF_ACCEPT);
556err_out:
557 return err;
558}
559
560static struct nf_queue_entry *
561nf_queue_entry_dup(struct nf_queue_entry *e)
562{
563 struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
564 if (entry) {
565 if (nf_queue_entry_get_refs(entry))
566 return entry;
567 kfree(entry);
568 }
569 return NULL;
570}
571
572#ifdef CONFIG_BRIDGE_NETFILTER
573
574
575
576
577static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
578{
579 if (skb->nf_bridge)
580 __skb_push(skb, skb->network_header - skb->mac_header);
581}
582
583static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
584{
585 if (skb->nf_bridge)
586 __skb_pull(skb, skb->network_header - skb->mac_header);
587}
588#else
589#define nf_bridge_adjust_skb_data(s) do {} while (0)
590#define nf_bridge_adjust_segmented_data(s) do {} while (0)
591#endif
592
593static void free_entry(struct nf_queue_entry *entry)
594{
595 nf_queue_entry_release_refs(entry);
596 kfree(entry);
597}
598
599static int
600__nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
601 struct sk_buff *skb, struct nf_queue_entry *entry)
602{
603 int ret = -ENOMEM;
604 struct nf_queue_entry *entry_seg;
605
606 nf_bridge_adjust_segmented_data(skb);
607
608 if (skb->next == NULL) {
609 struct sk_buff *gso_skb = entry->skb;
610 entry->skb = skb;
611 ret = __nfqnl_enqueue_packet(net, queue, entry);
612 if (ret)
613 entry->skb = gso_skb;
614 return ret;
615 }
616
617 skb->next = NULL;
618
619 entry_seg = nf_queue_entry_dup(entry);
620 if (entry_seg) {
621 entry_seg->skb = skb;
622 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
623 if (ret)
624 free_entry(entry_seg);
625 }
626 return ret;
627}
628
629static int
630nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
631{
632 unsigned int queued;
633 struct nfqnl_instance *queue;
634 struct sk_buff *skb, *segs;
635 int err = -ENOBUFS;
636 struct net *net = dev_net(entry->indev ?
637 entry->indev : entry->outdev);
638 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
639
640
641 queue = instance_lookup(q, queuenum);
642 if (!queue)
643 return -ESRCH;
644
645 if (queue->copy_mode == NFQNL_COPY_NONE)
646 return -EINVAL;
647
648 skb = entry->skb;
649
650 switch (entry->pf) {
651 case NFPROTO_IPV4:
652 skb->protocol = htons(ETH_P_IP);
653 break;
654 case NFPROTO_IPV6:
655 skb->protocol = htons(ETH_P_IPV6);
656 break;
657 }
658
659 if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(skb))
660 return __nfqnl_enqueue_packet(net, queue, entry);
661
662 nf_bridge_adjust_skb_data(skb);
663 segs = skb_gso_segment(skb, 0);
664
665
666
667
668 if (IS_ERR(segs))
669 goto out_err;
670 queued = 0;
671 err = 0;
672 do {
673 struct sk_buff *nskb = segs->next;
674 if (err == 0)
675 err = __nfqnl_enqueue_packet_gso(net, queue,
676 segs, entry);
677 if (err == 0)
678 queued++;
679 else
680 kfree_skb(segs);
681 segs = nskb;
682 } while (segs);
683
684 if (queued) {
685 if (err)
686 free_entry(entry);
687 kfree_skb(skb);
688 return 0;
689 }
690 out_err:
691 nf_bridge_adjust_segmented_data(skb);
692 return err;
693}
694
695static int
696nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e, int diff)
697{
698 struct sk_buff *nskb;
699
700 if (diff < 0) {
701 if (pskb_trim(e->skb, data_len))
702 return -ENOMEM;
703 } else if (diff > 0) {
704 if (data_len > 0xFFFF)
705 return -EINVAL;
706 if (diff > skb_tailroom(e->skb)) {
707 nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
708 diff, GFP_ATOMIC);
709 if (!nskb) {
710 printk(KERN_WARNING "nf_queue: OOM "
711 "in mangle, dropping packet\n");
712 return -ENOMEM;
713 }
714 kfree_skb(e->skb);
715 e->skb = nskb;
716 }
717 skb_put(e->skb, diff);
718 }
719 if (!skb_make_writable(e->skb, data_len))
720 return -ENOMEM;
721 skb_copy_to_linear_data(e->skb, data, data_len);
722 e->skb->ip_summed = CHECKSUM_NONE;
723 return 0;
724}
725
726static int
727nfqnl_set_mode(struct nfqnl_instance *queue,
728 unsigned char mode, unsigned int range)
729{
730 int status = 0;
731
732 spin_lock_bh(&queue->lock);
733 switch (mode) {
734 case NFQNL_COPY_NONE:
735 case NFQNL_COPY_META:
736 queue->copy_mode = mode;
737 queue->copy_range = 0;
738 break;
739
740 case NFQNL_COPY_PACKET:
741 queue->copy_mode = mode;
742 if (range == 0 || range > NFQNL_MAX_COPY_RANGE)
743 queue->copy_range = NFQNL_MAX_COPY_RANGE;
744 else
745 queue->copy_range = range;
746 break;
747
748 default:
749 status = -EINVAL;
750
751 }
752 spin_unlock_bh(&queue->lock);
753
754 return status;
755}
756
757static int
758dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
759{
760 if (entry->indev)
761 if (entry->indev->ifindex == ifindex)
762 return 1;
763 if (entry->outdev)
764 if (entry->outdev->ifindex == ifindex)
765 return 1;
766#ifdef CONFIG_BRIDGE_NETFILTER
767 if (entry->skb->nf_bridge) {
768 if (entry->skb->nf_bridge->physindev &&
769 entry->skb->nf_bridge->physindev->ifindex == ifindex)
770 return 1;
771 if (entry->skb->nf_bridge->physoutdev &&
772 entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
773 return 1;
774 }
775#endif
776 return 0;
777}
778
779
780
781static void
782nfqnl_dev_drop(struct net *net, int ifindex)
783{
784 int i;
785 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
786
787 rcu_read_lock();
788
789 for (i = 0; i < INSTANCE_BUCKETS; i++) {
790 struct nfqnl_instance *inst;
791 struct hlist_head *head = &q->instance_table[i];
792
793 hlist_for_each_entry_rcu(inst, head, hlist)
794 nfqnl_flush(inst, dev_cmp, ifindex);
795 }
796
797 rcu_read_unlock();
798}
799
800#define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
801
802static int
803nfqnl_rcv_dev_event(struct notifier_block *this,
804 unsigned long event, void *ptr)
805{
806 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
807
808
809 if (event == NETDEV_DOWN)
810 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
811 return NOTIFY_DONE;
812}
813
814static struct notifier_block nfqnl_dev_notifier = {
815 .notifier_call = nfqnl_rcv_dev_event,
816};
817
818static int
819nfqnl_rcv_nl_event(struct notifier_block *this,
820 unsigned long event, void *ptr)
821{
822 struct netlink_notify *n = ptr;
823 struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
824
825 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
826 int i;
827
828
829 spin_lock(&q->instances_lock);
830 for (i = 0; i < INSTANCE_BUCKETS; i++) {
831 struct hlist_node *t2;
832 struct nfqnl_instance *inst;
833 struct hlist_head *head = &q->instance_table[i];
834
835 hlist_for_each_entry_safe(inst, t2, head, hlist) {
836 if (n->portid == inst->peer_portid)
837 __instance_destroy(inst);
838 }
839 }
840 spin_unlock(&q->instances_lock);
841 }
842 return NOTIFY_DONE;
843}
844
845static struct notifier_block nfqnl_rtnl_notifier = {
846 .notifier_call = nfqnl_rcv_nl_event,
847};
848
849static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
850 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
851 [NFQA_MARK] = { .type = NLA_U32 },
852 [NFQA_PAYLOAD] = { .type = NLA_UNSPEC },
853 [NFQA_CT] = { .type = NLA_UNSPEC },
854 [NFQA_EXP] = { .type = NLA_UNSPEC },
855};
856
857static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
858 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
859 [NFQA_MARK] = { .type = NLA_U32 },
860};
861
862static struct nfqnl_instance *
863verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, int nlportid)
864{
865 struct nfqnl_instance *queue;
866
867 queue = instance_lookup(q, queue_num);
868 if (!queue)
869 return ERR_PTR(-ENODEV);
870
871 if (queue->peer_portid != nlportid)
872 return ERR_PTR(-EPERM);
873
874 return queue;
875}
876
877static struct nfqnl_msg_verdict_hdr*
878verdicthdr_get(const struct nlattr * const nfqa[])
879{
880 struct nfqnl_msg_verdict_hdr *vhdr;
881 unsigned int verdict;
882
883 if (!nfqa[NFQA_VERDICT_HDR])
884 return NULL;
885
886 vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
887 verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
888 if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
889 return NULL;
890 return vhdr;
891}
892
893static int nfq_id_after(unsigned int id, unsigned int max)
894{
895 return (int)(id - max) > 0;
896}
897
898static int
899nfqnl_recv_verdict_batch(struct sock *ctnl, struct sk_buff *skb,
900 const struct nlmsghdr *nlh,
901 const struct nlattr * const nfqa[])
902{
903 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
904 struct nf_queue_entry *entry, *tmp;
905 unsigned int verdict, maxid;
906 struct nfqnl_msg_verdict_hdr *vhdr;
907 struct nfqnl_instance *queue;
908 LIST_HEAD(batch_list);
909 u16 queue_num = ntohs(nfmsg->res_id);
910
911 struct net *net = sock_net(ctnl);
912 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
913
914 queue = verdict_instance_lookup(q, queue_num,
915 NETLINK_CB(skb).portid);
916 if (IS_ERR(queue))
917 return PTR_ERR(queue);
918
919 vhdr = verdicthdr_get(nfqa);
920 if (!vhdr)
921 return -EINVAL;
922
923 verdict = ntohl(vhdr->verdict);
924 maxid = ntohl(vhdr->id);
925
926 spin_lock_bh(&queue->lock);
927
928 list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
929 if (nfq_id_after(entry->id, maxid))
930 break;
931 __dequeue_entry(queue, entry);
932 list_add_tail(&entry->list, &batch_list);
933 }
934
935 spin_unlock_bh(&queue->lock);
936
937 if (list_empty(&batch_list))
938 return -ENOENT;
939
940 list_for_each_entry_safe(entry, tmp, &batch_list, list) {
941 if (nfqa[NFQA_MARK])
942 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
943 nf_reinject(entry, verdict);
944 }
945 return 0;
946}
947
948static int
949nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
950 const struct nlmsghdr *nlh,
951 const struct nlattr * const nfqa[])
952{
953 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
954 u_int16_t queue_num = ntohs(nfmsg->res_id);
955
956 struct nfqnl_msg_verdict_hdr *vhdr;
957 struct nfqnl_instance *queue;
958 unsigned int verdict;
959 struct nf_queue_entry *entry;
960 enum ip_conntrack_info uninitialized_var(ctinfo);
961 struct nf_conn *ct = NULL;
962
963 struct net *net = sock_net(ctnl);
964 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
965
966 queue = instance_lookup(q, queue_num);
967 if (!queue)
968 queue = verdict_instance_lookup(q, queue_num,
969 NETLINK_CB(skb).portid);
970 if (IS_ERR(queue))
971 return PTR_ERR(queue);
972
973 vhdr = verdicthdr_get(nfqa);
974 if (!vhdr)
975 return -EINVAL;
976
977 verdict = ntohl(vhdr->verdict);
978
979 entry = find_dequeue_entry(queue, ntohl(vhdr->id));
980 if (entry == NULL)
981 return -ENOENT;
982
983 if (nfqa[NFQA_CT]) {
984 ct = nfqnl_ct_parse(entry->skb, nfqa[NFQA_CT], &ctinfo);
985 if (ct && nfqa[NFQA_EXP]) {
986 nfqnl_attach_expect(ct, nfqa[NFQA_EXP],
987 NETLINK_CB(skb).portid,
988 nlmsg_report(nlh));
989 }
990 }
991
992 if (nfqa[NFQA_PAYLOAD]) {
993 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
994 int diff = payload_len - entry->skb->len;
995
996 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
997 payload_len, entry, diff) < 0)
998 verdict = NF_DROP;
999
1000 if (ct)
1001 nfqnl_ct_seq_adjust(entry->skb, ct, ctinfo, diff);
1002 }
1003
1004 if (nfqa[NFQA_MARK])
1005 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1006
1007 nf_reinject(entry, verdict);
1008 return 0;
1009}
1010
1011static int
1012nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
1013 const struct nlmsghdr *nlh,
1014 const struct nlattr * const nfqa[])
1015{
1016 return -ENOTSUPP;
1017}
1018
1019static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1020 [NFQA_CFG_CMD] = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1021 [NFQA_CFG_PARAMS] = { .len = sizeof(struct nfqnl_msg_config_params) },
1022};
1023
1024static const struct nf_queue_handler nfqh = {
1025 .outfn = &nfqnl_enqueue_packet,
1026};
1027
1028static int
1029nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
1030 const struct nlmsghdr *nlh,
1031 const struct nlattr * const nfqa[])
1032{
1033 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1034 u_int16_t queue_num = ntohs(nfmsg->res_id);
1035 struct nfqnl_instance *queue;
1036 struct nfqnl_msg_config_cmd *cmd = NULL;
1037 struct net *net = sock_net(ctnl);
1038 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1039 int ret = 0;
1040
1041 if (nfqa[NFQA_CFG_CMD]) {
1042 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1043
1044
1045 switch (cmd->command) {
1046 case NFQNL_CFG_CMD_PF_BIND: return 0;
1047 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1048 }
1049 }
1050
1051 rcu_read_lock();
1052 queue = instance_lookup(q, queue_num);
1053 if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1054 ret = -EPERM;
1055 goto err_out_unlock;
1056 }
1057
1058 if (cmd != NULL) {
1059 switch (cmd->command) {
1060 case NFQNL_CFG_CMD_BIND:
1061 if (queue) {
1062 ret = -EBUSY;
1063 goto err_out_unlock;
1064 }
1065 queue = instance_create(q, queue_num,
1066 NETLINK_CB(skb).portid);
1067 if (IS_ERR(queue)) {
1068 ret = PTR_ERR(queue);
1069 goto err_out_unlock;
1070 }
1071 break;
1072 case NFQNL_CFG_CMD_UNBIND:
1073 if (!queue) {
1074 ret = -ENODEV;
1075 goto err_out_unlock;
1076 }
1077 instance_destroy(q, queue);
1078 break;
1079 case NFQNL_CFG_CMD_PF_BIND:
1080 case NFQNL_CFG_CMD_PF_UNBIND:
1081 break;
1082 default:
1083 ret = -ENOTSUPP;
1084 break;
1085 }
1086 }
1087
1088 if (nfqa[NFQA_CFG_PARAMS]) {
1089 struct nfqnl_msg_config_params *params;
1090
1091 if (!queue) {
1092 ret = -ENODEV;
1093 goto err_out_unlock;
1094 }
1095 params = nla_data(nfqa[NFQA_CFG_PARAMS]);
1096 nfqnl_set_mode(queue, params->copy_mode,
1097 ntohl(params->copy_range));
1098 }
1099
1100 if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1101 __be32 *queue_maxlen;
1102
1103 if (!queue) {
1104 ret = -ENODEV;
1105 goto err_out_unlock;
1106 }
1107 queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1108 spin_lock_bh(&queue->lock);
1109 queue->queue_maxlen = ntohl(*queue_maxlen);
1110 spin_unlock_bh(&queue->lock);
1111 }
1112
1113 if (nfqa[NFQA_CFG_FLAGS]) {
1114 __u32 flags, mask;
1115
1116 if (!queue) {
1117 ret = -ENODEV;
1118 goto err_out_unlock;
1119 }
1120
1121 if (!nfqa[NFQA_CFG_MASK]) {
1122
1123
1124
1125 ret = -EINVAL;
1126 goto err_out_unlock;
1127 }
1128
1129 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1130 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1131
1132 if (flags >= NFQA_CFG_F_MAX) {
1133 ret = -EOPNOTSUPP;
1134 goto err_out_unlock;
1135 }
1136
1137 spin_lock_bh(&queue->lock);
1138 queue->flags &= ~mask;
1139 queue->flags |= flags & mask;
1140 spin_unlock_bh(&queue->lock);
1141 }
1142
1143err_out_unlock:
1144 rcu_read_unlock();
1145 return ret;
1146}
1147
1148static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1149 [NFQNL_MSG_PACKET] = { .call_rcu = nfqnl_recv_unsupp,
1150 .attr_count = NFQA_MAX, },
1151 [NFQNL_MSG_VERDICT] = { .call_rcu = nfqnl_recv_verdict,
1152 .attr_count = NFQA_MAX,
1153 .policy = nfqa_verdict_policy },
1154 [NFQNL_MSG_CONFIG] = { .call = nfqnl_recv_config,
1155 .attr_count = NFQA_CFG_MAX,
1156 .policy = nfqa_cfg_policy },
1157 [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
1158 .attr_count = NFQA_MAX,
1159 .policy = nfqa_verdict_batch_policy },
1160};
1161
1162static const struct nfnetlink_subsystem nfqnl_subsys = {
1163 .name = "nf_queue",
1164 .subsys_id = NFNL_SUBSYS_QUEUE,
1165 .cb_count = NFQNL_MSG_MAX,
1166 .cb = nfqnl_cb,
1167};
1168
1169#ifdef CONFIG_PROC_FS
1170struct iter_state {
1171 struct seq_net_private p;
1172 unsigned int bucket;
1173};
1174
1175static struct hlist_node *get_first(struct seq_file *seq)
1176{
1177 struct iter_state *st = seq->private;
1178 struct net *net;
1179 struct nfnl_queue_net *q;
1180
1181 if (!st)
1182 return NULL;
1183
1184 net = seq_file_net(seq);
1185 q = nfnl_queue_pernet(net);
1186 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1187 if (!hlist_empty(&q->instance_table[st->bucket]))
1188 return q->instance_table[st->bucket].first;
1189 }
1190 return NULL;
1191}
1192
1193static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1194{
1195 struct iter_state *st = seq->private;
1196 struct net *net = seq_file_net(seq);
1197
1198 h = h->next;
1199 while (!h) {
1200 struct nfnl_queue_net *q;
1201
1202 if (++st->bucket >= INSTANCE_BUCKETS)
1203 return NULL;
1204
1205 q = nfnl_queue_pernet(net);
1206 h = q->instance_table[st->bucket].first;
1207 }
1208 return h;
1209}
1210
1211static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1212{
1213 struct hlist_node *head;
1214 head = get_first(seq);
1215
1216 if (head)
1217 while (pos && (head = get_next(seq, head)))
1218 pos--;
1219 return pos ? NULL : head;
1220}
1221
1222static void *seq_start(struct seq_file *s, loff_t *pos)
1223 __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1224{
1225 spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1226 return get_idx(s, *pos);
1227}
1228
1229static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1230{
1231 (*pos)++;
1232 return get_next(s, v);
1233}
1234
1235static void seq_stop(struct seq_file *s, void *v)
1236 __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1237{
1238 spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1239}
1240
1241static int seq_show(struct seq_file *s, void *v)
1242{
1243 const struct nfqnl_instance *inst = v;
1244
1245 return seq_printf(s, "%5d %6d %5d %1d %5d %5d %5d %8d %2d\n",
1246 inst->queue_num,
1247 inst->peer_portid, inst->queue_total,
1248 inst->copy_mode, inst->copy_range,
1249 inst->queue_dropped, inst->queue_user_dropped,
1250 inst->id_sequence, 1);
1251}
1252
1253static const struct seq_operations nfqnl_seq_ops = {
1254 .start = seq_start,
1255 .next = seq_next,
1256 .stop = seq_stop,
1257 .show = seq_show,
1258};
1259
1260static int nfqnl_open(struct inode *inode, struct file *file)
1261{
1262 return seq_open_net(inode, file, &nfqnl_seq_ops,
1263 sizeof(struct iter_state));
1264}
1265
1266static const struct file_operations nfqnl_file_ops = {
1267 .owner = THIS_MODULE,
1268 .open = nfqnl_open,
1269 .read = seq_read,
1270 .llseek = seq_lseek,
1271 .release = seq_release_net,
1272};
1273
1274#endif
1275
1276static int __net_init nfnl_queue_net_init(struct net *net)
1277{
1278 unsigned int i;
1279 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1280
1281 for (i = 0; i < INSTANCE_BUCKETS; i++)
1282 INIT_HLIST_HEAD(&q->instance_table[i]);
1283
1284 spin_lock_init(&q->instances_lock);
1285
1286#ifdef CONFIG_PROC_FS
1287 if (!proc_create("nfnetlink_queue", 0440,
1288 net->nf.proc_netfilter, &nfqnl_file_ops))
1289 return -ENOMEM;
1290#endif
1291 return 0;
1292}
1293
1294static void __net_exit nfnl_queue_net_exit(struct net *net)
1295{
1296#ifdef CONFIG_PROC_FS
1297 remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1298#endif
1299}
1300
1301static struct pernet_operations nfnl_queue_net_ops = {
1302 .init = nfnl_queue_net_init,
1303 .exit = nfnl_queue_net_exit,
1304 .id = &nfnl_queue_net_id,
1305 .size = sizeof(struct nfnl_queue_net),
1306};
1307
1308static int __init nfnetlink_queue_init(void)
1309{
1310 int status = -ENOMEM;
1311
1312 netlink_register_notifier(&nfqnl_rtnl_notifier);
1313 status = nfnetlink_subsys_register(&nfqnl_subsys);
1314 if (status < 0) {
1315 pr_err("nf_queue: failed to create netlink socket\n");
1316 goto cleanup_netlink_notifier;
1317 }
1318
1319 status = register_pernet_subsys(&nfnl_queue_net_ops);
1320 if (status < 0) {
1321 pr_err("nf_queue: failed to register pernet ops\n");
1322 goto cleanup_subsys;
1323 }
1324 register_netdevice_notifier(&nfqnl_dev_notifier);
1325 nf_register_queue_handler(&nfqh);
1326 return status;
1327
1328cleanup_subsys:
1329 nfnetlink_subsys_unregister(&nfqnl_subsys);
1330cleanup_netlink_notifier:
1331 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1332 return status;
1333}
1334
1335static void __exit nfnetlink_queue_fini(void)
1336{
1337 nf_unregister_queue_handler();
1338 unregister_netdevice_notifier(&nfqnl_dev_notifier);
1339 unregister_pernet_subsys(&nfnl_queue_net_ops);
1340 nfnetlink_subsys_unregister(&nfqnl_subsys);
1341 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1342
1343 rcu_barrier();
1344}
1345
1346MODULE_DESCRIPTION("netfilter packet queue handler");
1347MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1348MODULE_LICENSE("GPL");
1349MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1350
1351module_init(nfnetlink_queue_init);
1352module_exit(nfnetlink_queue_fini);
1353
