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_bridge.h>
29#include <linux/netfilter/nfnetlink.h>
30#include <linux/netfilter/nfnetlink_queue.h>
31#include <linux/list.h>
32#include <net/sock.h>
33#include <net/tcp_states.h>
34#include <net/netfilter/nf_queue.h>
35#include <net/netns/generic.h>
36#include <net/netfilter/nfnetlink_queue.h>
37
38#include <linux/atomic.h>
39
40#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
41#include "../bridge/br_private.h"
42#endif
43
44#define NFQNL_QMAX_DEFAULT 1024
45
46
47
48
49
50
51
52#define NFQNL_MAX_COPY_RANGE (0xffff - NLA_HDRLEN)
53
54struct nfqnl_instance {
55 struct hlist_node hlist;
56 struct rcu_head rcu;
57
58 u32 peer_portid;
59 unsigned int queue_maxlen;
60 unsigned int copy_range;
61 unsigned int queue_dropped;
62 unsigned int queue_user_dropped;
63
64
65 u_int16_t queue_num;
66 u_int8_t copy_mode;
67 u_int32_t flags;
68
69
70
71
72 spinlock_t lock;
73 unsigned int queue_total;
74 unsigned int id_sequence;
75 struct list_head queue_list;
76};
77
78typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
79
80static int nfnl_queue_net_id __read_mostly;
81
82#define INSTANCE_BUCKETS 16
83struct nfnl_queue_net {
84 spinlock_t instances_lock;
85 struct hlist_head instance_table[INSTANCE_BUCKETS];
86};
87
88static struct nfnl_queue_net *nfnl_queue_pernet(struct net *net)
89{
90 return net_generic(net, nfnl_queue_net_id);
91}
92
93static inline u_int8_t instance_hashfn(u_int16_t queue_num)
94{
95 return ((queue_num >> 8) ^ queue_num) % INSTANCE_BUCKETS;
96}
97
98static struct nfqnl_instance *
99instance_lookup(struct nfnl_queue_net *q, u_int16_t queue_num)
100{
101 struct hlist_head *head;
102 struct nfqnl_instance *inst;
103
104 head = &q->instance_table[instance_hashfn(queue_num)];
105 hlist_for_each_entry_rcu(inst, head, hlist) {
106 if (inst->queue_num == queue_num)
107 return inst;
108 }
109 return NULL;
110}
111
112static struct nfqnl_instance *
113instance_create(struct nfnl_queue_net *q, u_int16_t queue_num, u32 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_fullsock(sk))
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#if IS_ENABLED(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->state.hook <= NF_INET_FORWARD ||
318 (entry->state.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->state.out;
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->state.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->state.hook;
379 *packet_id_ptr = &pmsg->packet_id;
380
381 indev = entry->state.in;
382 if (indev) {
383#if !IS_ENABLED(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->state.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 int physinif;
400
401
402
403 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
404 htonl(indev->ifindex)))
405 goto nla_put_failure;
406
407 physinif = nf_bridge_get_physinif(entskb);
408 if (physinif &&
409 nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
410 htonl(physinif)))
411 goto nla_put_failure;
412 }
413#endif
414 }
415
416 if (outdev) {
417#if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
418 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
419 goto nla_put_failure;
420#else
421 if (entry->state.pf == PF_BRIDGE) {
422
423
424
425 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
426 htonl(outdev->ifindex)) ||
427
428
429 nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
430 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
431 goto nla_put_failure;
432 } else {
433 int physoutif;
434
435
436
437 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
438 htonl(outdev->ifindex)))
439 goto nla_put_failure;
440
441 physoutif = nf_bridge_get_physoutif(entskb);
442 if (physoutif &&
443 nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
444 htonl(physoutif)))
445 goto nla_put_failure;
446 }
447#endif
448 }
449
450 if (entskb->mark &&
451 nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
452 goto nla_put_failure;
453
454 if (indev && entskb->dev &&
455 entskb->mac_header != entskb->network_header) {
456 struct nfqnl_msg_packet_hw phw;
457 int len;
458
459 memset(&phw, 0, sizeof(phw));
460 len = dev_parse_header(entskb, phw.hw_addr);
461 if (len) {
462 phw.hw_addrlen = htons(len);
463 if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
464 goto nla_put_failure;
465 }
466 }
467
468 if (entskb->tstamp.tv64) {
469 struct nfqnl_msg_packet_timestamp ts;
470 struct timeval tv = ktime_to_timeval(entskb->tstamp);
471 ts.sec = cpu_to_be64(tv.tv_sec);
472 ts.usec = cpu_to_be64(tv.tv_usec);
473
474 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
475 goto nla_put_failure;
476 }
477
478 if ((queue->flags & NFQA_CFG_F_UID_GID) && entskb->sk &&
479 nfqnl_put_sk_uidgid(skb, entskb->sk) < 0)
480 goto nla_put_failure;
481
482 if (ct && nfqnl_ct_put(skb, ct, ctinfo) < 0)
483 goto nla_put_failure;
484
485 if (cap_len > data_len &&
486 nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
487 goto nla_put_failure;
488
489 if (nfqnl_put_packet_info(skb, entskb, csum_verify))
490 goto nla_put_failure;
491
492 if (data_len) {
493 struct nlattr *nla;
494
495 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
496 goto nla_put_failure;
497
498 nla = (struct nlattr *)skb_put(skb, sizeof(*nla));
499 nla->nla_type = NFQA_PAYLOAD;
500 nla->nla_len = nla_attr_size(data_len);
501
502 if (skb_zerocopy(skb, entskb, data_len, hlen))
503 goto nla_put_failure;
504 }
505
506 nlh->nlmsg_len = skb->len;
507 return skb;
508
509nla_put_failure:
510 skb_tx_error(entskb);
511 kfree_skb(skb);
512 net_err_ratelimited("nf_queue: error creating packet message\n");
513 return NULL;
514}
515
516static int
517__nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
518 struct nf_queue_entry *entry)
519{
520 struct sk_buff *nskb;
521 int err = -ENOBUFS;
522 __be32 *packet_id_ptr;
523 int failopen = 0;
524
525 nskb = nfqnl_build_packet_message(net, queue, entry, &packet_id_ptr);
526 if (nskb == NULL) {
527 err = -ENOMEM;
528 goto err_out;
529 }
530 spin_lock_bh(&queue->lock);
531
532 if (queue->queue_total >= queue->queue_maxlen) {
533 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
534 failopen = 1;
535 err = 0;
536 } else {
537 queue->queue_dropped++;
538 net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
539 queue->queue_total);
540 }
541 goto err_out_free_nskb;
542 }
543 entry->id = ++queue->id_sequence;
544 *packet_id_ptr = htonl(entry->id);
545
546
547 err = nfnetlink_unicast(nskb, net, queue->peer_portid, MSG_DONTWAIT);
548 if (err < 0) {
549 queue->queue_user_dropped++;
550 goto err_out_unlock;
551 }
552
553 __enqueue_entry(queue, entry);
554
555 spin_unlock_bh(&queue->lock);
556 return 0;
557
558err_out_free_nskb:
559 kfree_skb(nskb);
560err_out_unlock:
561 spin_unlock_bh(&queue->lock);
562 if (failopen)
563 nf_reinject(entry, NF_ACCEPT);
564err_out:
565 return err;
566}
567
568static struct nf_queue_entry *
569nf_queue_entry_dup(struct nf_queue_entry *e)
570{
571 struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
572 if (entry) {
573 if (nf_queue_entry_get_refs(entry))
574 return entry;
575 kfree(entry);
576 }
577 return NULL;
578}
579
580#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
581
582
583
584
585static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
586{
587 if (skb->nf_bridge)
588 __skb_push(skb, skb->network_header - skb->mac_header);
589}
590
591static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
592{
593 if (skb->nf_bridge)
594 __skb_pull(skb, skb->network_header - skb->mac_header);
595}
596#else
597#define nf_bridge_adjust_skb_data(s) do {} while (0)
598#define nf_bridge_adjust_segmented_data(s) do {} while (0)
599#endif
600
601static void free_entry(struct nf_queue_entry *entry)
602{
603 nf_queue_entry_release_refs(entry);
604 kfree(entry);
605}
606
607static int
608__nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
609 struct sk_buff *skb, struct nf_queue_entry *entry)
610{
611 int ret = -ENOMEM;
612 struct nf_queue_entry *entry_seg;
613
614 nf_bridge_adjust_segmented_data(skb);
615
616 if (skb->next == NULL) {
617 struct sk_buff *gso_skb = entry->skb;
618 entry->skb = skb;
619 ret = __nfqnl_enqueue_packet(net, queue, entry);
620 if (ret)
621 entry->skb = gso_skb;
622 return ret;
623 }
624
625 skb->next = NULL;
626
627 entry_seg = nf_queue_entry_dup(entry);
628 if (entry_seg) {
629 entry_seg->skb = skb;
630 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
631 if (ret)
632 free_entry(entry_seg);
633 }
634 return ret;
635}
636
637static int
638nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
639{
640 unsigned int queued;
641 struct nfqnl_instance *queue;
642 struct sk_buff *skb, *segs;
643 int err = -ENOBUFS;
644 struct net *net = dev_net(entry->state.in ?
645 entry->state.in : entry->state.out);
646 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
647
648
649 queue = instance_lookup(q, queuenum);
650 if (!queue)
651 return -ESRCH;
652
653 if (queue->copy_mode == NFQNL_COPY_NONE)
654 return -EINVAL;
655
656 skb = entry->skb;
657
658 switch (entry->state.pf) {
659 case NFPROTO_IPV4:
660 skb->protocol = htons(ETH_P_IP);
661 break;
662 case NFPROTO_IPV6:
663 skb->protocol = htons(ETH_P_IPV6);
664 break;
665 }
666
667 if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(skb))
668 return __nfqnl_enqueue_packet(net, queue, entry);
669
670 nf_bridge_adjust_skb_data(skb);
671 segs = skb_gso_segment(skb, 0);
672
673
674
675
676 if (IS_ERR_OR_NULL(segs))
677 goto out_err;
678 queued = 0;
679 err = 0;
680 do {
681 struct sk_buff *nskb = segs->next;
682 if (err == 0)
683 err = __nfqnl_enqueue_packet_gso(net, queue,
684 segs, entry);
685 if (err == 0)
686 queued++;
687 else
688 kfree_skb(segs);
689 segs = nskb;
690 } while (segs);
691
692 if (queued) {
693 if (err)
694 free_entry(entry);
695 kfree_skb(skb);
696 return 0;
697 }
698 out_err:
699 nf_bridge_adjust_segmented_data(skb);
700 return err;
701}
702
703static int
704nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e, int diff)
705{
706 struct sk_buff *nskb;
707
708 if (diff < 0) {
709 if (pskb_trim(e->skb, data_len))
710 return -ENOMEM;
711 } else if (diff > 0) {
712 if (data_len > 0xFFFF)
713 return -EINVAL;
714 if (diff > skb_tailroom(e->skb)) {
715 nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
716 diff, GFP_ATOMIC);
717 if (!nskb) {
718 printk(KERN_WARNING "nf_queue: OOM "
719 "in mangle, dropping packet\n");
720 return -ENOMEM;
721 }
722 kfree_skb(e->skb);
723 e->skb = nskb;
724 }
725 skb_put(e->skb, diff);
726 }
727 if (!skb_make_writable(e->skb, data_len))
728 return -ENOMEM;
729 skb_copy_to_linear_data(e->skb, data, data_len);
730 e->skb->ip_summed = CHECKSUM_NONE;
731 return 0;
732}
733
734static int
735nfqnl_set_mode(struct nfqnl_instance *queue,
736 unsigned char mode, unsigned int range)
737{
738 int status = 0;
739
740 spin_lock_bh(&queue->lock);
741 switch (mode) {
742 case NFQNL_COPY_NONE:
743 case NFQNL_COPY_META:
744 queue->copy_mode = mode;
745 queue->copy_range = 0;
746 break;
747
748 case NFQNL_COPY_PACKET:
749 queue->copy_mode = mode;
750 if (range == 0 || range > NFQNL_MAX_COPY_RANGE)
751 queue->copy_range = NFQNL_MAX_COPY_RANGE;
752 else
753 queue->copy_range = range;
754 break;
755
756 default:
757 status = -EINVAL;
758
759 }
760 spin_unlock_bh(&queue->lock);
761
762 return status;
763}
764
765static int
766dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
767{
768 if (entry->state.in)
769 if (entry->state.in->ifindex == ifindex)
770 return 1;
771 if (entry->state.out)
772 if (entry->state.out->ifindex == ifindex)
773 return 1;
774#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
775 if (entry->skb->nf_bridge) {
776 int physinif, physoutif;
777
778 physinif = nf_bridge_get_physinif(entry->skb);
779 physoutif = nf_bridge_get_physoutif(entry->skb);
780
781 if (physinif == ifindex || physoutif == ifindex)
782 return 1;
783 }
784#endif
785 return 0;
786}
787
788
789
790static void
791nfqnl_dev_drop(struct net *net, int ifindex)
792{
793 int i;
794 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
795
796 rcu_read_lock();
797
798 for (i = 0; i < INSTANCE_BUCKETS; i++) {
799 struct nfqnl_instance *inst;
800 struct hlist_head *head = &q->instance_table[i];
801
802 hlist_for_each_entry_rcu(inst, head, hlist)
803 nfqnl_flush(inst, dev_cmp, ifindex);
804 }
805
806 rcu_read_unlock();
807}
808
809#define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
810
811static int
812nfqnl_rcv_dev_event(struct notifier_block *this,
813 unsigned long event, void *ptr)
814{
815 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
816
817
818 if (event == NETDEV_DOWN)
819 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
820 return NOTIFY_DONE;
821}
822
823static struct notifier_block nfqnl_dev_notifier = {
824 .notifier_call = nfqnl_rcv_dev_event,
825};
826
827static int
828nfqnl_rcv_nl_event(struct notifier_block *this,
829 unsigned long event, void *ptr)
830{
831 struct netlink_notify *n = ptr;
832 struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
833
834 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
835 int i;
836
837
838 spin_lock(&q->instances_lock);
839 for (i = 0; i < INSTANCE_BUCKETS; i++) {
840 struct hlist_node *t2;
841 struct nfqnl_instance *inst;
842 struct hlist_head *head = &q->instance_table[i];
843
844 hlist_for_each_entry_safe(inst, t2, head, hlist) {
845 if (n->portid == inst->peer_portid)
846 __instance_destroy(inst);
847 }
848 }
849 spin_unlock(&q->instances_lock);
850 }
851 return NOTIFY_DONE;
852}
853
854static struct notifier_block nfqnl_rtnl_notifier = {
855 .notifier_call = nfqnl_rcv_nl_event,
856};
857
858static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
859 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
860 [NFQA_MARK] = { .type = NLA_U32 },
861 [NFQA_PAYLOAD] = { .type = NLA_UNSPEC },
862 [NFQA_CT] = { .type = NLA_UNSPEC },
863 [NFQA_EXP] = { .type = NLA_UNSPEC },
864};
865
866static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
867 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
868 [NFQA_MARK] = { .type = NLA_U32 },
869};
870
871static struct nfqnl_instance *
872verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, u32 nlportid)
873{
874 struct nfqnl_instance *queue;
875
876 queue = instance_lookup(q, queue_num);
877 if (!queue)
878 return ERR_PTR(-ENODEV);
879
880 if (queue->peer_portid != nlportid)
881 return ERR_PTR(-EPERM);
882
883 return queue;
884}
885
886static struct nfqnl_msg_verdict_hdr*
887verdicthdr_get(const struct nlattr * const nfqa[])
888{
889 struct nfqnl_msg_verdict_hdr *vhdr;
890 unsigned int verdict;
891
892 if (!nfqa[NFQA_VERDICT_HDR])
893 return NULL;
894
895 vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
896 verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
897 if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
898 return NULL;
899 return vhdr;
900}
901
902static int nfq_id_after(unsigned int id, unsigned int max)
903{
904 return (int)(id - max) > 0;
905}
906
907static int
908nfqnl_recv_verdict_batch(struct sock *ctnl, struct sk_buff *skb,
909 const struct nlmsghdr *nlh,
910 const struct nlattr * const nfqa[])
911{
912 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
913 struct nf_queue_entry *entry, *tmp;
914 unsigned int verdict, maxid;
915 struct nfqnl_msg_verdict_hdr *vhdr;
916 struct nfqnl_instance *queue;
917 LIST_HEAD(batch_list);
918 u16 queue_num = ntohs(nfmsg->res_id);
919
920 struct net *net = sock_net(ctnl);
921 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
922
923 queue = verdict_instance_lookup(q, queue_num,
924 NETLINK_CB(skb).portid);
925 if (IS_ERR(queue))
926 return PTR_ERR(queue);
927
928 vhdr = verdicthdr_get(nfqa);
929 if (!vhdr)
930 return -EINVAL;
931
932 verdict = ntohl(vhdr->verdict);
933 maxid = ntohl(vhdr->id);
934
935 spin_lock_bh(&queue->lock);
936
937 list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
938 if (nfq_id_after(entry->id, maxid))
939 break;
940 __dequeue_entry(queue, entry);
941 list_add_tail(&entry->list, &batch_list);
942 }
943
944 spin_unlock_bh(&queue->lock);
945
946 if (list_empty(&batch_list))
947 return -ENOENT;
948
949 list_for_each_entry_safe(entry, tmp, &batch_list, list) {
950 if (nfqa[NFQA_MARK])
951 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
952 nf_reinject(entry, verdict);
953 }
954 return 0;
955}
956
957static int
958nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
959 const struct nlmsghdr *nlh,
960 const struct nlattr * const nfqa[])
961{
962 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
963 u_int16_t queue_num = ntohs(nfmsg->res_id);
964
965 struct nfqnl_msg_verdict_hdr *vhdr;
966 struct nfqnl_instance *queue;
967 unsigned int verdict;
968 struct nf_queue_entry *entry;
969 enum ip_conntrack_info uninitialized_var(ctinfo);
970 struct nf_conn *ct = NULL;
971
972 struct net *net = sock_net(ctnl);
973 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
974
975 queue = instance_lookup(q, queue_num);
976 if (!queue)
977 queue = verdict_instance_lookup(q, queue_num,
978 NETLINK_CB(skb).portid);
979 if (IS_ERR(queue))
980 return PTR_ERR(queue);
981
982 vhdr = verdicthdr_get(nfqa);
983 if (!vhdr)
984 return -EINVAL;
985
986 verdict = ntohl(vhdr->verdict);
987
988 entry = find_dequeue_entry(queue, ntohl(vhdr->id));
989 if (entry == NULL)
990 return -ENOENT;
991
992 if (nfqa[NFQA_CT]) {
993 ct = nfqnl_ct_parse(entry->skb, nfqa[NFQA_CT], &ctinfo);
994 if (ct && nfqa[NFQA_EXP]) {
995 nfqnl_attach_expect(ct, nfqa[NFQA_EXP],
996 NETLINK_CB(skb).portid,
997 nlmsg_report(nlh));
998 }
999 }
1000
1001 if (nfqa[NFQA_PAYLOAD]) {
1002 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
1003 int diff = payload_len - entry->skb->len;
1004
1005 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
1006 payload_len, entry, diff) < 0)
1007 verdict = NF_DROP;
1008
1009 if (ct)
1010 nfqnl_ct_seq_adjust(entry->skb, ct, ctinfo, diff);
1011 }
1012
1013 if (nfqa[NFQA_MARK])
1014 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1015
1016 nf_reinject(entry, verdict);
1017 return 0;
1018}
1019
1020static int
1021nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
1022 const struct nlmsghdr *nlh,
1023 const struct nlattr * const nfqa[])
1024{
1025 return -ENOTSUPP;
1026}
1027
1028static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1029 [NFQA_CFG_CMD] = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1030 [NFQA_CFG_PARAMS] = { .len = sizeof(struct nfqnl_msg_config_params) },
1031};
1032
1033static const struct nf_queue_handler nfqh = {
1034 .outfn = &nfqnl_enqueue_packet,
1035};
1036
1037static int
1038nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
1039 const struct nlmsghdr *nlh,
1040 const struct nlattr * const nfqa[])
1041{
1042 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1043 u_int16_t queue_num = ntohs(nfmsg->res_id);
1044 struct nfqnl_instance *queue;
1045 struct nfqnl_msg_config_cmd *cmd = NULL;
1046 struct net *net = sock_net(ctnl);
1047 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1048 int ret = 0;
1049
1050 if (nfqa[NFQA_CFG_CMD]) {
1051 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1052
1053
1054 switch (cmd->command) {
1055 case NFQNL_CFG_CMD_PF_BIND: return 0;
1056 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1057 }
1058 }
1059
1060 rcu_read_lock();
1061 queue = instance_lookup(q, queue_num);
1062 if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1063 ret = -EPERM;
1064 goto err_out_unlock;
1065 }
1066
1067 if (cmd != NULL) {
1068 switch (cmd->command) {
1069 case NFQNL_CFG_CMD_BIND:
1070 if (queue) {
1071 ret = -EBUSY;
1072 goto err_out_unlock;
1073 }
1074 queue = instance_create(q, queue_num,
1075 NETLINK_CB(skb).portid);
1076 if (IS_ERR(queue)) {
1077 ret = PTR_ERR(queue);
1078 goto err_out_unlock;
1079 }
1080 break;
1081 case NFQNL_CFG_CMD_UNBIND:
1082 if (!queue) {
1083 ret = -ENODEV;
1084 goto err_out_unlock;
1085 }
1086 instance_destroy(q, queue);
1087 break;
1088 case NFQNL_CFG_CMD_PF_BIND:
1089 case NFQNL_CFG_CMD_PF_UNBIND:
1090 break;
1091 default:
1092 ret = -ENOTSUPP;
1093 break;
1094 }
1095 }
1096
1097 if (nfqa[NFQA_CFG_PARAMS]) {
1098 struct nfqnl_msg_config_params *params;
1099
1100 if (!queue) {
1101 ret = -ENODEV;
1102 goto err_out_unlock;
1103 }
1104 params = nla_data(nfqa[NFQA_CFG_PARAMS]);
1105 nfqnl_set_mode(queue, params->copy_mode,
1106 ntohl(params->copy_range));
1107 }
1108
1109 if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1110 __be32 *queue_maxlen;
1111
1112 if (!queue) {
1113 ret = -ENODEV;
1114 goto err_out_unlock;
1115 }
1116 queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1117 spin_lock_bh(&queue->lock);
1118 queue->queue_maxlen = ntohl(*queue_maxlen);
1119 spin_unlock_bh(&queue->lock);
1120 }
1121
1122 if (nfqa[NFQA_CFG_FLAGS]) {
1123 __u32 flags, mask;
1124
1125 if (!queue) {
1126 ret = -ENODEV;
1127 goto err_out_unlock;
1128 }
1129
1130 if (!nfqa[NFQA_CFG_MASK]) {
1131
1132
1133
1134 ret = -EINVAL;
1135 goto err_out_unlock;
1136 }
1137
1138 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1139 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1140
1141 if (flags >= NFQA_CFG_F_MAX) {
1142 ret = -EOPNOTSUPP;
1143 goto err_out_unlock;
1144 }
1145
1146 spin_lock_bh(&queue->lock);
1147 queue->flags &= ~mask;
1148 queue->flags |= flags & mask;
1149 spin_unlock_bh(&queue->lock);
1150 }
1151
1152err_out_unlock:
1153 rcu_read_unlock();
1154 return ret;
1155}
1156
1157static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1158 [NFQNL_MSG_PACKET] = { .call_rcu = nfqnl_recv_unsupp,
1159 .attr_count = NFQA_MAX, },
1160 [NFQNL_MSG_VERDICT] = { .call_rcu = nfqnl_recv_verdict,
1161 .attr_count = NFQA_MAX,
1162 .policy = nfqa_verdict_policy },
1163 [NFQNL_MSG_CONFIG] = { .call = nfqnl_recv_config,
1164 .attr_count = NFQA_CFG_MAX,
1165 .policy = nfqa_cfg_policy },
1166 [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
1167 .attr_count = NFQA_MAX,
1168 .policy = nfqa_verdict_batch_policy },
1169};
1170
1171static const struct nfnetlink_subsystem nfqnl_subsys = {
1172 .name = "nf_queue",
1173 .subsys_id = NFNL_SUBSYS_QUEUE,
1174 .cb_count = NFQNL_MSG_MAX,
1175 .cb = nfqnl_cb,
1176};
1177
1178#ifdef CONFIG_PROC_FS
1179struct iter_state {
1180 struct seq_net_private p;
1181 unsigned int bucket;
1182};
1183
1184static struct hlist_node *get_first(struct seq_file *seq)
1185{
1186 struct iter_state *st = seq->private;
1187 struct net *net;
1188 struct nfnl_queue_net *q;
1189
1190 if (!st)
1191 return NULL;
1192
1193 net = seq_file_net(seq);
1194 q = nfnl_queue_pernet(net);
1195 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1196 if (!hlist_empty(&q->instance_table[st->bucket]))
1197 return q->instance_table[st->bucket].first;
1198 }
1199 return NULL;
1200}
1201
1202static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1203{
1204 struct iter_state *st = seq->private;
1205 struct net *net = seq_file_net(seq);
1206
1207 h = h->next;
1208 while (!h) {
1209 struct nfnl_queue_net *q;
1210
1211 if (++st->bucket >= INSTANCE_BUCKETS)
1212 return NULL;
1213
1214 q = nfnl_queue_pernet(net);
1215 h = q->instance_table[st->bucket].first;
1216 }
1217 return h;
1218}
1219
1220static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1221{
1222 struct hlist_node *head;
1223 head = get_first(seq);
1224
1225 if (head)
1226 while (pos && (head = get_next(seq, head)))
1227 pos--;
1228 return pos ? NULL : head;
1229}
1230
1231static void *seq_start(struct seq_file *s, loff_t *pos)
1232 __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1233{
1234 spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1235 return get_idx(s, *pos);
1236}
1237
1238static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1239{
1240 (*pos)++;
1241 return get_next(s, v);
1242}
1243
1244static void seq_stop(struct seq_file *s, void *v)
1245 __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1246{
1247 spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1248}
1249
1250static int seq_show(struct seq_file *s, void *v)
1251{
1252 const struct nfqnl_instance *inst = v;
1253
1254 seq_printf(s, "%5u %6u %5u %1u %5u %5u %5u %8u %2d\n",
1255 inst->queue_num,
1256 inst->peer_portid, inst->queue_total,
1257 inst->copy_mode, inst->copy_range,
1258 inst->queue_dropped, inst->queue_user_dropped,
1259 inst->id_sequence, 1);
1260 return seq_has_overflowed(s);
1261}
1262
1263static const struct seq_operations nfqnl_seq_ops = {
1264 .start = seq_start,
1265 .next = seq_next,
1266 .stop = seq_stop,
1267 .show = seq_show,
1268};
1269
1270static int nfqnl_open(struct inode *inode, struct file *file)
1271{
1272 return seq_open_net(inode, file, &nfqnl_seq_ops,
1273 sizeof(struct iter_state));
1274}
1275
1276static const struct file_operations nfqnl_file_ops = {
1277 .owner = THIS_MODULE,
1278 .open = nfqnl_open,
1279 .read = seq_read,
1280 .llseek = seq_lseek,
1281 .release = seq_release_net,
1282};
1283
1284#endif
1285
1286static int __net_init nfnl_queue_net_init(struct net *net)
1287{
1288 unsigned int i;
1289 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1290
1291 for (i = 0; i < INSTANCE_BUCKETS; i++)
1292 INIT_HLIST_HEAD(&q->instance_table[i]);
1293
1294 spin_lock_init(&q->instances_lock);
1295
1296#ifdef CONFIG_PROC_FS
1297 if (!proc_create("nfnetlink_queue", 0440,
1298 net->nf.proc_netfilter, &nfqnl_file_ops))
1299 return -ENOMEM;
1300#endif
1301 return 0;
1302}
1303
1304static void __net_exit nfnl_queue_net_exit(struct net *net)
1305{
1306#ifdef CONFIG_PROC_FS
1307 remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1308#endif
1309}
1310
1311static struct pernet_operations nfnl_queue_net_ops = {
1312 .init = nfnl_queue_net_init,
1313 .exit = nfnl_queue_net_exit,
1314 .id = &nfnl_queue_net_id,
1315 .size = sizeof(struct nfnl_queue_net),
1316};
1317
1318static int __init nfnetlink_queue_init(void)
1319{
1320 int status;
1321
1322 status = register_pernet_subsys(&nfnl_queue_net_ops);
1323 if (status < 0) {
1324 pr_err("nf_queue: failed to register pernet ops\n");
1325 goto out;
1326 }
1327
1328 netlink_register_notifier(&nfqnl_rtnl_notifier);
1329 status = nfnetlink_subsys_register(&nfqnl_subsys);
1330 if (status < 0) {
1331 pr_err("nf_queue: failed to create netlink socket\n");
1332 goto cleanup_netlink_notifier;
1333 }
1334
1335 register_netdevice_notifier(&nfqnl_dev_notifier);
1336 nf_register_queue_handler(&nfqh);
1337 return status;
1338
1339cleanup_netlink_notifier:
1340 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1341out:
1342 return status;
1343}
1344
1345static void __exit nfnetlink_queue_fini(void)
1346{
1347 nf_unregister_queue_handler();
1348 unregister_netdevice_notifier(&nfqnl_dev_notifier);
1349 nfnetlink_subsys_unregister(&nfqnl_subsys);
1350 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1351 unregister_pernet_subsys(&nfnl_queue_net_ops);
1352
1353 rcu_barrier();
1354}
1355
1356MODULE_DESCRIPTION("netfilter packet queue handler");
1357MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1358MODULE_LICENSE("GPL");
1359MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1360
1361module_init(nfnetlink_queue_init);
1362module_exit(nfnetlink_queue_fini);
1363
