Loading...
1/*
2 * This is a module which is used for queueing packets and communicating with
3 * userspace via nfnetlink.
4 *
5 * (C) 2005 by Harald Welte <laforge@netfilter.org>
6 * (C) 2007 by Patrick McHardy <kaber@trash.net>
7 *
8 * Based on the old ipv4-only ip_queue.c:
9 * (C) 2000-2002 James Morris <jmorris@intercode.com.au>
10 * (C) 2003-2005 Netfilter Core Team <coreteam@netfilter.org>
11 *
12 * This program is free software; you can redistribute it and/or modify
13 * it under the terms of the GNU General Public License version 2 as
14 * published by the Free Software Foundation.
15 *
16 */
17#include <linux/module.h>
18#include <linux/skbuff.h>
19#include <linux/init.h>
20#include <linux/spinlock.h>
21#include <linux/slab.h>
22#include <linux/notifier.h>
23#include <linux/netdevice.h>
24#include <linux/netfilter.h>
25#include <linux/proc_fs.h>
26#include <linux/netfilter_ipv4.h>
27#include <linux/netfilter_ipv6.h>
28#include <linux/netfilter_bridge.h>
29#include <linux/netfilter/nfnetlink.h>
30#include <linux/netfilter/nfnetlink_queue.h>
31#include <linux/netfilter/nf_conntrack_common.h>
32#include <linux/list.h>
33#include <net/sock.h>
34#include <net/tcp_states.h>
35#include <net/netfilter/nf_queue.h>
36#include <net/netns/generic.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/* We're using struct nlattr which has 16bit nla_len. Note that nla_len
47 * includes the header length. Thus, the maximum packet length that we
48 * support is 65531 bytes. We send truncated packets if the specified length
49 * is larger than that. Userspace can check for presence of NFQA_CAP_LEN
50 * attribute to detect truncation.
51 */
52#define NFQNL_MAX_COPY_RANGE (0xffff - NLA_HDRLEN)
53
54struct nfqnl_instance {
55 struct hlist_node hlist; /* global list of queues */
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; /* number of this queue */
66 u_int8_t copy_mode;
67 u_int32_t flags; /* Set using NFQA_CFG_FLAGS */
68/*
69 * Following fields are dirtied for each queued packet,
70 * keep them in same cache line if possible.
71 */
72 spinlock_t lock ____cacheline_aligned_in_smp;
73 unsigned int queue_total;
74 unsigned int id_sequence; /* 'sequence' of pkt ids */
75 struct list_head queue_list; /* packets in queue */
76};
77
78typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
79
80static unsigned 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 u32 nfqnl_get_sk_secctx(struct sk_buff *skb, char **secdata)
282{
283 u32 seclen = 0;
284#if IS_ENABLED(CONFIG_NETWORK_SECMARK)
285 if (!skb || !sk_fullsock(skb->sk))
286 return 0;
287
288 read_lock_bh(&skb->sk->sk_callback_lock);
289
290 if (skb->secmark)
291 security_secid_to_secctx(skb->secmark, secdata, &seclen);
292
293 read_unlock_bh(&skb->sk->sk_callback_lock);
294#endif
295 return seclen;
296}
297
298static u32 nfqnl_get_bridge_size(struct nf_queue_entry *entry)
299{
300 struct sk_buff *entskb = entry->skb;
301 u32 nlalen = 0;
302
303 if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb))
304 return 0;
305
306 if (skb_vlan_tag_present(entskb))
307 nlalen += nla_total_size(nla_total_size(sizeof(__be16)) +
308 nla_total_size(sizeof(__be16)));
309
310 if (entskb->network_header > entskb->mac_header)
311 nlalen += nla_total_size((entskb->network_header -
312 entskb->mac_header));
313
314 return nlalen;
315}
316
317static int nfqnl_put_bridge(struct nf_queue_entry *entry, struct sk_buff *skb)
318{
319 struct sk_buff *entskb = entry->skb;
320
321 if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb))
322 return 0;
323
324 if (skb_vlan_tag_present(entskb)) {
325 struct nlattr *nest;
326
327 nest = nla_nest_start(skb, NFQA_VLAN | NLA_F_NESTED);
328 if (!nest)
329 goto nla_put_failure;
330
331 if (nla_put_be16(skb, NFQA_VLAN_TCI, htons(entskb->vlan_tci)) ||
332 nla_put_be16(skb, NFQA_VLAN_PROTO, entskb->vlan_proto))
333 goto nla_put_failure;
334
335 nla_nest_end(skb, nest);
336 }
337
338 if (entskb->mac_header < entskb->network_header) {
339 int len = (int)(entskb->network_header - entskb->mac_header);
340
341 if (nla_put(skb, NFQA_L2HDR, len, skb_mac_header(entskb)))
342 goto nla_put_failure;
343 }
344
345 return 0;
346
347nla_put_failure:
348 return -1;
349}
350
351static struct sk_buff *
352nfqnl_build_packet_message(struct net *net, struct nfqnl_instance *queue,
353 struct nf_queue_entry *entry,
354 __be32 **packet_id_ptr)
355{
356 size_t size;
357 size_t data_len = 0, cap_len = 0;
358 unsigned int hlen = 0;
359 struct sk_buff *skb;
360 struct nlattr *nla;
361 struct nfqnl_msg_packet_hdr *pmsg;
362 struct nlmsghdr *nlh;
363 struct nfgenmsg *nfmsg;
364 struct sk_buff *entskb = entry->skb;
365 struct net_device *indev;
366 struct net_device *outdev;
367 struct nf_conn *ct = NULL;
368 enum ip_conntrack_info uninitialized_var(ctinfo);
369 struct nfnl_ct_hook *nfnl_ct;
370 bool csum_verify;
371 char *secdata = NULL;
372 u32 seclen = 0;
373
374 size = nlmsg_total_size(sizeof(struct nfgenmsg))
375 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
376 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
377 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
378#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
379 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
380 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
381#endif
382 + nla_total_size(sizeof(u_int32_t)) /* mark */
383 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
384 + nla_total_size(sizeof(u_int32_t)) /* skbinfo */
385 + nla_total_size(sizeof(u_int32_t)); /* cap_len */
386
387 if (entskb->tstamp)
388 size += nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
389
390 size += nfqnl_get_bridge_size(entry);
391
392 if (entry->state.hook <= NF_INET_FORWARD ||
393 (entry->state.hook == NF_INET_POST_ROUTING && entskb->sk == NULL))
394 csum_verify = !skb_csum_unnecessary(entskb);
395 else
396 csum_verify = false;
397
398 outdev = entry->state.out;
399
400 switch ((enum nfqnl_config_mode)ACCESS_ONCE(queue->copy_mode)) {
401 case NFQNL_COPY_META:
402 case NFQNL_COPY_NONE:
403 break;
404
405 case NFQNL_COPY_PACKET:
406 if (!(queue->flags & NFQA_CFG_F_GSO) &&
407 entskb->ip_summed == CHECKSUM_PARTIAL &&
408 skb_checksum_help(entskb))
409 return NULL;
410
411 data_len = ACCESS_ONCE(queue->copy_range);
412 if (data_len > entskb->len)
413 data_len = entskb->len;
414
415 hlen = skb_zerocopy_headlen(entskb);
416 hlen = min_t(unsigned int, hlen, data_len);
417 size += sizeof(struct nlattr) + hlen;
418 cap_len = entskb->len;
419 break;
420 }
421
422 nfnl_ct = rcu_dereference(nfnl_ct_hook);
423
424 if (queue->flags & NFQA_CFG_F_CONNTRACK) {
425 if (nfnl_ct != NULL) {
426 ct = nfnl_ct->get_ct(entskb, &ctinfo);
427 if (ct != NULL)
428 size += nfnl_ct->build_size(ct);
429 }
430 }
431
432 if (queue->flags & NFQA_CFG_F_UID_GID) {
433 size += (nla_total_size(sizeof(u_int32_t)) /* uid */
434 + nla_total_size(sizeof(u_int32_t))); /* gid */
435 }
436
437 if ((queue->flags & NFQA_CFG_F_SECCTX) && entskb->sk) {
438 seclen = nfqnl_get_sk_secctx(entskb, &secdata);
439 if (seclen)
440 size += nla_total_size(seclen);
441 }
442
443 skb = alloc_skb(size, GFP_ATOMIC);
444 if (!skb) {
445 skb_tx_error(entskb);
446 return NULL;
447 }
448
449 nlh = nlmsg_put(skb, 0, 0,
450 NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
451 sizeof(struct nfgenmsg), 0);
452 if (!nlh) {
453 skb_tx_error(entskb);
454 kfree_skb(skb);
455 return NULL;
456 }
457 nfmsg = nlmsg_data(nlh);
458 nfmsg->nfgen_family = entry->state.pf;
459 nfmsg->version = NFNETLINK_V0;
460 nfmsg->res_id = htons(queue->queue_num);
461
462 nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
463 pmsg = nla_data(nla);
464 pmsg->hw_protocol = entskb->protocol;
465 pmsg->hook = entry->state.hook;
466 *packet_id_ptr = &pmsg->packet_id;
467
468 indev = entry->state.in;
469 if (indev) {
470#if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
471 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
472 goto nla_put_failure;
473#else
474 if (entry->state.pf == PF_BRIDGE) {
475 /* Case 1: indev is physical input device, we need to
476 * look for bridge group (when called from
477 * netfilter_bridge) */
478 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
479 htonl(indev->ifindex)) ||
480 /* this is the bridge group "brX" */
481 /* rcu_read_lock()ed by __nf_queue */
482 nla_put_be32(skb, NFQA_IFINDEX_INDEV,
483 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
484 goto nla_put_failure;
485 } else {
486 int physinif;
487
488 /* Case 2: indev is bridge group, we need to look for
489 * physical device (when called from ipv4) */
490 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
491 htonl(indev->ifindex)))
492 goto nla_put_failure;
493
494 physinif = nf_bridge_get_physinif(entskb);
495 if (physinif &&
496 nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
497 htonl(physinif)))
498 goto nla_put_failure;
499 }
500#endif
501 }
502
503 if (outdev) {
504#if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
505 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
506 goto nla_put_failure;
507#else
508 if (entry->state.pf == PF_BRIDGE) {
509 /* Case 1: outdev is physical output device, we need to
510 * look for bridge group (when called from
511 * netfilter_bridge) */
512 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
513 htonl(outdev->ifindex)) ||
514 /* this is the bridge group "brX" */
515 /* rcu_read_lock()ed by __nf_queue */
516 nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
517 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
518 goto nla_put_failure;
519 } else {
520 int physoutif;
521
522 /* Case 2: outdev is bridge group, we need to look for
523 * physical output device (when called from ipv4) */
524 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
525 htonl(outdev->ifindex)))
526 goto nla_put_failure;
527
528 physoutif = nf_bridge_get_physoutif(entskb);
529 if (physoutif &&
530 nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
531 htonl(physoutif)))
532 goto nla_put_failure;
533 }
534#endif
535 }
536
537 if (entskb->mark &&
538 nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
539 goto nla_put_failure;
540
541 if (indev && entskb->dev &&
542 entskb->mac_header != entskb->network_header) {
543 struct nfqnl_msg_packet_hw phw;
544 int len;
545
546 memset(&phw, 0, sizeof(phw));
547 len = dev_parse_header(entskb, phw.hw_addr);
548 if (len) {
549 phw.hw_addrlen = htons(len);
550 if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
551 goto nla_put_failure;
552 }
553 }
554
555 if (nfqnl_put_bridge(entry, skb) < 0)
556 goto nla_put_failure;
557
558 if (entskb->tstamp) {
559 struct nfqnl_msg_packet_timestamp ts;
560 struct timespec64 kts = ktime_to_timespec64(entskb->tstamp);
561
562 ts.sec = cpu_to_be64(kts.tv_sec);
563 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
564
565 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
566 goto nla_put_failure;
567 }
568
569 if ((queue->flags & NFQA_CFG_F_UID_GID) && entskb->sk &&
570 nfqnl_put_sk_uidgid(skb, entskb->sk) < 0)
571 goto nla_put_failure;
572
573 if (seclen && nla_put(skb, NFQA_SECCTX, seclen, secdata))
574 goto nla_put_failure;
575
576 if (ct && nfnl_ct->build(skb, ct, ctinfo, NFQA_CT, NFQA_CT_INFO) < 0)
577 goto nla_put_failure;
578
579 if (cap_len > data_len &&
580 nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
581 goto nla_put_failure;
582
583 if (nfqnl_put_packet_info(skb, entskb, csum_verify))
584 goto nla_put_failure;
585
586 if (data_len) {
587 struct nlattr *nla;
588
589 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
590 goto nla_put_failure;
591
592 nla = (struct nlattr *)skb_put(skb, sizeof(*nla));
593 nla->nla_type = NFQA_PAYLOAD;
594 nla->nla_len = nla_attr_size(data_len);
595
596 if (skb_zerocopy(skb, entskb, data_len, hlen))
597 goto nla_put_failure;
598 }
599
600 nlh->nlmsg_len = skb->len;
601 return skb;
602
603nla_put_failure:
604 skb_tx_error(entskb);
605 kfree_skb(skb);
606 net_err_ratelimited("nf_queue: error creating packet message\n");
607 return NULL;
608}
609
610static int
611__nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
612 struct nf_queue_entry *entry)
613{
614 struct sk_buff *nskb;
615 int err = -ENOBUFS;
616 __be32 *packet_id_ptr;
617 int failopen = 0;
618
619 nskb = nfqnl_build_packet_message(net, queue, entry, &packet_id_ptr);
620 if (nskb == NULL) {
621 err = -ENOMEM;
622 goto err_out;
623 }
624 spin_lock_bh(&queue->lock);
625
626 if (queue->queue_total >= queue->queue_maxlen) {
627 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
628 failopen = 1;
629 err = 0;
630 } else {
631 queue->queue_dropped++;
632 net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
633 queue->queue_total);
634 }
635 goto err_out_free_nskb;
636 }
637 entry->id = ++queue->id_sequence;
638 *packet_id_ptr = htonl(entry->id);
639
640 /* nfnetlink_unicast will either free the nskb or add it to a socket */
641 err = nfnetlink_unicast(nskb, net, queue->peer_portid, MSG_DONTWAIT);
642 if (err < 0) {
643 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
644 failopen = 1;
645 err = 0;
646 } else {
647 queue->queue_user_dropped++;
648 }
649 goto err_out_unlock;
650 }
651
652 __enqueue_entry(queue, entry);
653
654 spin_unlock_bh(&queue->lock);
655 return 0;
656
657err_out_free_nskb:
658 kfree_skb(nskb);
659err_out_unlock:
660 spin_unlock_bh(&queue->lock);
661 if (failopen)
662 nf_reinject(entry, NF_ACCEPT);
663err_out:
664 return err;
665}
666
667static struct nf_queue_entry *
668nf_queue_entry_dup(struct nf_queue_entry *e)
669{
670 struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
671 if (entry)
672 nf_queue_entry_get_refs(entry);
673 return entry;
674}
675
676#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
677/* When called from bridge netfilter, skb->data must point to MAC header
678 * before calling skb_gso_segment(). Else, original MAC header is lost
679 * and segmented skbs will be sent to wrong destination.
680 */
681static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
682{
683 if (skb->nf_bridge)
684 __skb_push(skb, skb->network_header - skb->mac_header);
685}
686
687static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
688{
689 if (skb->nf_bridge)
690 __skb_pull(skb, skb->network_header - skb->mac_header);
691}
692#else
693#define nf_bridge_adjust_skb_data(s) do {} while (0)
694#define nf_bridge_adjust_segmented_data(s) do {} while (0)
695#endif
696
697static void free_entry(struct nf_queue_entry *entry)
698{
699 nf_queue_entry_release_refs(entry);
700 kfree(entry);
701}
702
703static int
704__nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
705 struct sk_buff *skb, struct nf_queue_entry *entry)
706{
707 int ret = -ENOMEM;
708 struct nf_queue_entry *entry_seg;
709
710 nf_bridge_adjust_segmented_data(skb);
711
712 if (skb->next == NULL) { /* last packet, no need to copy entry */
713 struct sk_buff *gso_skb = entry->skb;
714 entry->skb = skb;
715 ret = __nfqnl_enqueue_packet(net, queue, entry);
716 if (ret)
717 entry->skb = gso_skb;
718 return ret;
719 }
720
721 skb->next = NULL;
722
723 entry_seg = nf_queue_entry_dup(entry);
724 if (entry_seg) {
725 entry_seg->skb = skb;
726 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
727 if (ret)
728 free_entry(entry_seg);
729 }
730 return ret;
731}
732
733static int
734nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
735{
736 unsigned int queued;
737 struct nfqnl_instance *queue;
738 struct sk_buff *skb, *segs;
739 int err = -ENOBUFS;
740 struct net *net = entry->state.net;
741 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
742
743 /* rcu_read_lock()ed by nf_hook_thresh */
744 queue = instance_lookup(q, queuenum);
745 if (!queue)
746 return -ESRCH;
747
748 if (queue->copy_mode == NFQNL_COPY_NONE)
749 return -EINVAL;
750
751 skb = entry->skb;
752
753 switch (entry->state.pf) {
754 case NFPROTO_IPV4:
755 skb->protocol = htons(ETH_P_IP);
756 break;
757 case NFPROTO_IPV6:
758 skb->protocol = htons(ETH_P_IPV6);
759 break;
760 }
761
762 if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(skb))
763 return __nfqnl_enqueue_packet(net, queue, entry);
764
765 nf_bridge_adjust_skb_data(skb);
766 segs = skb_gso_segment(skb, 0);
767 /* Does not use PTR_ERR to limit the number of error codes that can be
768 * returned by nf_queue. For instance, callers rely on -ESRCH to
769 * mean 'ignore this hook'.
770 */
771 if (IS_ERR_OR_NULL(segs))
772 goto out_err;
773 queued = 0;
774 err = 0;
775 do {
776 struct sk_buff *nskb = segs->next;
777 if (err == 0)
778 err = __nfqnl_enqueue_packet_gso(net, queue,
779 segs, entry);
780 if (err == 0)
781 queued++;
782 else
783 kfree_skb(segs);
784 segs = nskb;
785 } while (segs);
786
787 if (queued) {
788 if (err) /* some segments are already queued */
789 free_entry(entry);
790 kfree_skb(skb);
791 return 0;
792 }
793 out_err:
794 nf_bridge_adjust_segmented_data(skb);
795 return err;
796}
797
798static int
799nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e, int diff)
800{
801 struct sk_buff *nskb;
802
803 if (diff < 0) {
804 if (pskb_trim(e->skb, data_len))
805 return -ENOMEM;
806 } else if (diff > 0) {
807 if (data_len > 0xFFFF)
808 return -EINVAL;
809 if (diff > skb_tailroom(e->skb)) {
810 nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
811 diff, GFP_ATOMIC);
812 if (!nskb) {
813 printk(KERN_WARNING "nf_queue: OOM "
814 "in mangle, dropping packet\n");
815 return -ENOMEM;
816 }
817 kfree_skb(e->skb);
818 e->skb = nskb;
819 }
820 skb_put(e->skb, diff);
821 }
822 if (!skb_make_writable(e->skb, data_len))
823 return -ENOMEM;
824 skb_copy_to_linear_data(e->skb, data, data_len);
825 e->skb->ip_summed = CHECKSUM_NONE;
826 return 0;
827}
828
829static int
830nfqnl_set_mode(struct nfqnl_instance *queue,
831 unsigned char mode, unsigned int range)
832{
833 int status = 0;
834
835 spin_lock_bh(&queue->lock);
836 switch (mode) {
837 case NFQNL_COPY_NONE:
838 case NFQNL_COPY_META:
839 queue->copy_mode = mode;
840 queue->copy_range = 0;
841 break;
842
843 case NFQNL_COPY_PACKET:
844 queue->copy_mode = mode;
845 if (range == 0 || range > NFQNL_MAX_COPY_RANGE)
846 queue->copy_range = NFQNL_MAX_COPY_RANGE;
847 else
848 queue->copy_range = range;
849 break;
850
851 default:
852 status = -EINVAL;
853
854 }
855 spin_unlock_bh(&queue->lock);
856
857 return status;
858}
859
860static int
861dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
862{
863 if (entry->state.in)
864 if (entry->state.in->ifindex == ifindex)
865 return 1;
866 if (entry->state.out)
867 if (entry->state.out->ifindex == ifindex)
868 return 1;
869#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
870 if (entry->skb->nf_bridge) {
871 int physinif, physoutif;
872
873 physinif = nf_bridge_get_physinif(entry->skb);
874 physoutif = nf_bridge_get_physoutif(entry->skb);
875
876 if (physinif == ifindex || physoutif == ifindex)
877 return 1;
878 }
879#endif
880 return 0;
881}
882
883/* drop all packets with either indev or outdev == ifindex from all queue
884 * instances */
885static void
886nfqnl_dev_drop(struct net *net, int ifindex)
887{
888 int i;
889 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
890
891 rcu_read_lock();
892
893 for (i = 0; i < INSTANCE_BUCKETS; i++) {
894 struct nfqnl_instance *inst;
895 struct hlist_head *head = &q->instance_table[i];
896
897 hlist_for_each_entry_rcu(inst, head, hlist)
898 nfqnl_flush(inst, dev_cmp, ifindex);
899 }
900
901 rcu_read_unlock();
902}
903
904static int
905nfqnl_rcv_dev_event(struct notifier_block *this,
906 unsigned long event, void *ptr)
907{
908 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
909
910 /* Drop any packets associated with the downed device */
911 if (event == NETDEV_DOWN)
912 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
913 return NOTIFY_DONE;
914}
915
916static struct notifier_block nfqnl_dev_notifier = {
917 .notifier_call = nfqnl_rcv_dev_event,
918};
919
920static int nf_hook_cmp(struct nf_queue_entry *entry, unsigned long entry_ptr)
921{
922 return rcu_access_pointer(entry->hook) ==
923 (struct nf_hook_entry *)entry_ptr;
924}
925
926static void nfqnl_nf_hook_drop(struct net *net,
927 const struct nf_hook_entry *hook)
928{
929 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
930 int i;
931
932 rcu_read_lock();
933 for (i = 0; i < INSTANCE_BUCKETS; i++) {
934 struct nfqnl_instance *inst;
935 struct hlist_head *head = &q->instance_table[i];
936
937 hlist_for_each_entry_rcu(inst, head, hlist)
938 nfqnl_flush(inst, nf_hook_cmp, (unsigned long)hook);
939 }
940 rcu_read_unlock();
941}
942
943static int
944nfqnl_rcv_nl_event(struct notifier_block *this,
945 unsigned long event, void *ptr)
946{
947 struct netlink_notify *n = ptr;
948 struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
949
950 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
951 int i;
952
953 /* destroy all instances for this portid */
954 spin_lock(&q->instances_lock);
955 for (i = 0; i < INSTANCE_BUCKETS; i++) {
956 struct hlist_node *t2;
957 struct nfqnl_instance *inst;
958 struct hlist_head *head = &q->instance_table[i];
959
960 hlist_for_each_entry_safe(inst, t2, head, hlist) {
961 if (n->portid == inst->peer_portid)
962 __instance_destroy(inst);
963 }
964 }
965 spin_unlock(&q->instances_lock);
966 }
967 return NOTIFY_DONE;
968}
969
970static struct notifier_block nfqnl_rtnl_notifier = {
971 .notifier_call = nfqnl_rcv_nl_event,
972};
973
974static const struct nla_policy nfqa_vlan_policy[NFQA_VLAN_MAX + 1] = {
975 [NFQA_VLAN_TCI] = { .type = NLA_U16},
976 [NFQA_VLAN_PROTO] = { .type = NLA_U16},
977};
978
979static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
980 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
981 [NFQA_MARK] = { .type = NLA_U32 },
982 [NFQA_PAYLOAD] = { .type = NLA_UNSPEC },
983 [NFQA_CT] = { .type = NLA_UNSPEC },
984 [NFQA_EXP] = { .type = NLA_UNSPEC },
985 [NFQA_VLAN] = { .type = NLA_NESTED },
986};
987
988static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
989 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
990 [NFQA_MARK] = { .type = NLA_U32 },
991};
992
993static struct nfqnl_instance *
994verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, u32 nlportid)
995{
996 struct nfqnl_instance *queue;
997
998 queue = instance_lookup(q, queue_num);
999 if (!queue)
1000 return ERR_PTR(-ENODEV);
1001
1002 if (queue->peer_portid != nlportid)
1003 return ERR_PTR(-EPERM);
1004
1005 return queue;
1006}
1007
1008static struct nfqnl_msg_verdict_hdr*
1009verdicthdr_get(const struct nlattr * const nfqa[])
1010{
1011 struct nfqnl_msg_verdict_hdr *vhdr;
1012 unsigned int verdict;
1013
1014 if (!nfqa[NFQA_VERDICT_HDR])
1015 return NULL;
1016
1017 vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
1018 verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
1019 if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
1020 return NULL;
1021 return vhdr;
1022}
1023
1024static int nfq_id_after(unsigned int id, unsigned int max)
1025{
1026 return (int)(id - max) > 0;
1027}
1028
1029static int nfqnl_recv_verdict_batch(struct net *net, struct sock *ctnl,
1030 struct sk_buff *skb,
1031 const struct nlmsghdr *nlh,
1032 const struct nlattr * const nfqa[])
1033{
1034 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1035 struct nf_queue_entry *entry, *tmp;
1036 unsigned int verdict, maxid;
1037 struct nfqnl_msg_verdict_hdr *vhdr;
1038 struct nfqnl_instance *queue;
1039 LIST_HEAD(batch_list);
1040 u16 queue_num = ntohs(nfmsg->res_id);
1041 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1042
1043 queue = verdict_instance_lookup(q, queue_num,
1044 NETLINK_CB(skb).portid);
1045 if (IS_ERR(queue))
1046 return PTR_ERR(queue);
1047
1048 vhdr = verdicthdr_get(nfqa);
1049 if (!vhdr)
1050 return -EINVAL;
1051
1052 verdict = ntohl(vhdr->verdict);
1053 maxid = ntohl(vhdr->id);
1054
1055 spin_lock_bh(&queue->lock);
1056
1057 list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
1058 if (nfq_id_after(entry->id, maxid))
1059 break;
1060 __dequeue_entry(queue, entry);
1061 list_add_tail(&entry->list, &batch_list);
1062 }
1063
1064 spin_unlock_bh(&queue->lock);
1065
1066 if (list_empty(&batch_list))
1067 return -ENOENT;
1068
1069 list_for_each_entry_safe(entry, tmp, &batch_list, list) {
1070 if (nfqa[NFQA_MARK])
1071 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1072 nf_reinject(entry, verdict);
1073 }
1074 return 0;
1075}
1076
1077static struct nf_conn *nfqnl_ct_parse(struct nfnl_ct_hook *nfnl_ct,
1078 const struct nlmsghdr *nlh,
1079 const struct nlattr * const nfqa[],
1080 struct nf_queue_entry *entry,
1081 enum ip_conntrack_info *ctinfo)
1082{
1083 struct nf_conn *ct;
1084
1085 ct = nfnl_ct->get_ct(entry->skb, ctinfo);
1086 if (ct == NULL)
1087 return NULL;
1088
1089 if (nfnl_ct->parse(nfqa[NFQA_CT], ct) < 0)
1090 return NULL;
1091
1092 if (nfqa[NFQA_EXP])
1093 nfnl_ct->attach_expect(nfqa[NFQA_EXP], ct,
1094 NETLINK_CB(entry->skb).portid,
1095 nlmsg_report(nlh));
1096 return ct;
1097}
1098
1099static int nfqa_parse_bridge(struct nf_queue_entry *entry,
1100 const struct nlattr * const nfqa[])
1101{
1102 if (nfqa[NFQA_VLAN]) {
1103 struct nlattr *tb[NFQA_VLAN_MAX + 1];
1104 int err;
1105
1106 err = nla_parse_nested(tb, NFQA_VLAN_MAX, nfqa[NFQA_VLAN],
1107 nfqa_vlan_policy);
1108 if (err < 0)
1109 return err;
1110
1111 if (!tb[NFQA_VLAN_TCI] || !tb[NFQA_VLAN_PROTO])
1112 return -EINVAL;
1113
1114 entry->skb->vlan_tci = ntohs(nla_get_be16(tb[NFQA_VLAN_TCI]));
1115 entry->skb->vlan_proto = nla_get_be16(tb[NFQA_VLAN_PROTO]);
1116 }
1117
1118 if (nfqa[NFQA_L2HDR]) {
1119 int mac_header_len = entry->skb->network_header -
1120 entry->skb->mac_header;
1121
1122 if (mac_header_len != nla_len(nfqa[NFQA_L2HDR]))
1123 return -EINVAL;
1124 else if (mac_header_len > 0)
1125 memcpy(skb_mac_header(entry->skb),
1126 nla_data(nfqa[NFQA_L2HDR]),
1127 mac_header_len);
1128 }
1129
1130 return 0;
1131}
1132
1133static int nfqnl_recv_verdict(struct net *net, struct sock *ctnl,
1134 struct sk_buff *skb,
1135 const struct nlmsghdr *nlh,
1136 const struct nlattr * const nfqa[])
1137{
1138 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1139 u_int16_t queue_num = ntohs(nfmsg->res_id);
1140 struct nfqnl_msg_verdict_hdr *vhdr;
1141 struct nfqnl_instance *queue;
1142 unsigned int verdict;
1143 struct nf_queue_entry *entry;
1144 enum ip_conntrack_info uninitialized_var(ctinfo);
1145 struct nfnl_ct_hook *nfnl_ct;
1146 struct nf_conn *ct = NULL;
1147 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1148 int err;
1149
1150 queue = verdict_instance_lookup(q, queue_num,
1151 NETLINK_CB(skb).portid);
1152 if (IS_ERR(queue))
1153 return PTR_ERR(queue);
1154
1155 vhdr = verdicthdr_get(nfqa);
1156 if (!vhdr)
1157 return -EINVAL;
1158
1159 verdict = ntohl(vhdr->verdict);
1160
1161 entry = find_dequeue_entry(queue, ntohl(vhdr->id));
1162 if (entry == NULL)
1163 return -ENOENT;
1164
1165 /* rcu lock already held from nfnl->call_rcu. */
1166 nfnl_ct = rcu_dereference(nfnl_ct_hook);
1167
1168 if (nfqa[NFQA_CT]) {
1169 if (nfnl_ct != NULL)
1170 ct = nfqnl_ct_parse(nfnl_ct, nlh, nfqa, entry, &ctinfo);
1171 }
1172
1173 if (entry->state.pf == PF_BRIDGE) {
1174 err = nfqa_parse_bridge(entry, nfqa);
1175 if (err < 0)
1176 return err;
1177 }
1178
1179 if (nfqa[NFQA_PAYLOAD]) {
1180 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
1181 int diff = payload_len - entry->skb->len;
1182
1183 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
1184 payload_len, entry, diff) < 0)
1185 verdict = NF_DROP;
1186
1187 if (ct && diff)
1188 nfnl_ct->seq_adjust(entry->skb, ct, ctinfo, diff);
1189 }
1190
1191 if (nfqa[NFQA_MARK])
1192 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1193
1194 nf_reinject(entry, verdict);
1195 return 0;
1196}
1197
1198static int nfqnl_recv_unsupp(struct net *net, struct sock *ctnl,
1199 struct sk_buff *skb, const struct nlmsghdr *nlh,
1200 const struct nlattr * const nfqa[])
1201{
1202 return -ENOTSUPP;
1203}
1204
1205static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1206 [NFQA_CFG_CMD] = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1207 [NFQA_CFG_PARAMS] = { .len = sizeof(struct nfqnl_msg_config_params) },
1208};
1209
1210static const struct nf_queue_handler nfqh = {
1211 .outfn = &nfqnl_enqueue_packet,
1212 .nf_hook_drop = &nfqnl_nf_hook_drop,
1213};
1214
1215static int nfqnl_recv_config(struct net *net, struct sock *ctnl,
1216 struct sk_buff *skb, const struct nlmsghdr *nlh,
1217 const struct nlattr * const nfqa[])
1218{
1219 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1220 u_int16_t queue_num = ntohs(nfmsg->res_id);
1221 struct nfqnl_instance *queue;
1222 struct nfqnl_msg_config_cmd *cmd = NULL;
1223 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1224 __u32 flags = 0, mask = 0;
1225 int ret = 0;
1226
1227 if (nfqa[NFQA_CFG_CMD]) {
1228 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1229
1230 /* Obsolete commands without queue context */
1231 switch (cmd->command) {
1232 case NFQNL_CFG_CMD_PF_BIND: return 0;
1233 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1234 }
1235 }
1236
1237 /* Check if we support these flags in first place, dependencies should
1238 * be there too not to break atomicity.
1239 */
1240 if (nfqa[NFQA_CFG_FLAGS]) {
1241 if (!nfqa[NFQA_CFG_MASK]) {
1242 /* A mask is needed to specify which flags are being
1243 * changed.
1244 */
1245 return -EINVAL;
1246 }
1247
1248 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1249 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1250
1251 if (flags >= NFQA_CFG_F_MAX)
1252 return -EOPNOTSUPP;
1253
1254#if !IS_ENABLED(CONFIG_NETWORK_SECMARK)
1255 if (flags & mask & NFQA_CFG_F_SECCTX)
1256 return -EOPNOTSUPP;
1257#endif
1258 if ((flags & mask & NFQA_CFG_F_CONNTRACK) &&
1259 !rcu_access_pointer(nfnl_ct_hook)) {
1260#ifdef CONFIG_MODULES
1261 nfnl_unlock(NFNL_SUBSYS_QUEUE);
1262 request_module("ip_conntrack_netlink");
1263 nfnl_lock(NFNL_SUBSYS_QUEUE);
1264 if (rcu_access_pointer(nfnl_ct_hook))
1265 return -EAGAIN;
1266#endif
1267 return -EOPNOTSUPP;
1268 }
1269 }
1270
1271 rcu_read_lock();
1272 queue = instance_lookup(q, queue_num);
1273 if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1274 ret = -EPERM;
1275 goto err_out_unlock;
1276 }
1277
1278 if (cmd != NULL) {
1279 switch (cmd->command) {
1280 case NFQNL_CFG_CMD_BIND:
1281 if (queue) {
1282 ret = -EBUSY;
1283 goto err_out_unlock;
1284 }
1285 queue = instance_create(q, queue_num,
1286 NETLINK_CB(skb).portid);
1287 if (IS_ERR(queue)) {
1288 ret = PTR_ERR(queue);
1289 goto err_out_unlock;
1290 }
1291 break;
1292 case NFQNL_CFG_CMD_UNBIND:
1293 if (!queue) {
1294 ret = -ENODEV;
1295 goto err_out_unlock;
1296 }
1297 instance_destroy(q, queue);
1298 goto err_out_unlock;
1299 case NFQNL_CFG_CMD_PF_BIND:
1300 case NFQNL_CFG_CMD_PF_UNBIND:
1301 break;
1302 default:
1303 ret = -ENOTSUPP;
1304 goto err_out_unlock;
1305 }
1306 }
1307
1308 if (!queue) {
1309 ret = -ENODEV;
1310 goto err_out_unlock;
1311 }
1312
1313 if (nfqa[NFQA_CFG_PARAMS]) {
1314 struct nfqnl_msg_config_params *params =
1315 nla_data(nfqa[NFQA_CFG_PARAMS]);
1316
1317 nfqnl_set_mode(queue, params->copy_mode,
1318 ntohl(params->copy_range));
1319 }
1320
1321 if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1322 __be32 *queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1323
1324 spin_lock_bh(&queue->lock);
1325 queue->queue_maxlen = ntohl(*queue_maxlen);
1326 spin_unlock_bh(&queue->lock);
1327 }
1328
1329 if (nfqa[NFQA_CFG_FLAGS]) {
1330 spin_lock_bh(&queue->lock);
1331 queue->flags &= ~mask;
1332 queue->flags |= flags & mask;
1333 spin_unlock_bh(&queue->lock);
1334 }
1335
1336err_out_unlock:
1337 rcu_read_unlock();
1338 return ret;
1339}
1340
1341static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1342 [NFQNL_MSG_PACKET] = { .call_rcu = nfqnl_recv_unsupp,
1343 .attr_count = NFQA_MAX, },
1344 [NFQNL_MSG_VERDICT] = { .call_rcu = nfqnl_recv_verdict,
1345 .attr_count = NFQA_MAX,
1346 .policy = nfqa_verdict_policy },
1347 [NFQNL_MSG_CONFIG] = { .call = nfqnl_recv_config,
1348 .attr_count = NFQA_CFG_MAX,
1349 .policy = nfqa_cfg_policy },
1350 [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
1351 .attr_count = NFQA_MAX,
1352 .policy = nfqa_verdict_batch_policy },
1353};
1354
1355static const struct nfnetlink_subsystem nfqnl_subsys = {
1356 .name = "nf_queue",
1357 .subsys_id = NFNL_SUBSYS_QUEUE,
1358 .cb_count = NFQNL_MSG_MAX,
1359 .cb = nfqnl_cb,
1360};
1361
1362#ifdef CONFIG_PROC_FS
1363struct iter_state {
1364 struct seq_net_private p;
1365 unsigned int bucket;
1366};
1367
1368static struct hlist_node *get_first(struct seq_file *seq)
1369{
1370 struct iter_state *st = seq->private;
1371 struct net *net;
1372 struct nfnl_queue_net *q;
1373
1374 if (!st)
1375 return NULL;
1376
1377 net = seq_file_net(seq);
1378 q = nfnl_queue_pernet(net);
1379 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1380 if (!hlist_empty(&q->instance_table[st->bucket]))
1381 return q->instance_table[st->bucket].first;
1382 }
1383 return NULL;
1384}
1385
1386static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1387{
1388 struct iter_state *st = seq->private;
1389 struct net *net = seq_file_net(seq);
1390
1391 h = h->next;
1392 while (!h) {
1393 struct nfnl_queue_net *q;
1394
1395 if (++st->bucket >= INSTANCE_BUCKETS)
1396 return NULL;
1397
1398 q = nfnl_queue_pernet(net);
1399 h = q->instance_table[st->bucket].first;
1400 }
1401 return h;
1402}
1403
1404static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1405{
1406 struct hlist_node *head;
1407 head = get_first(seq);
1408
1409 if (head)
1410 while (pos && (head = get_next(seq, head)))
1411 pos--;
1412 return pos ? NULL : head;
1413}
1414
1415static void *seq_start(struct seq_file *s, loff_t *pos)
1416 __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1417{
1418 spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1419 return get_idx(s, *pos);
1420}
1421
1422static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1423{
1424 (*pos)++;
1425 return get_next(s, v);
1426}
1427
1428static void seq_stop(struct seq_file *s, void *v)
1429 __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1430{
1431 spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1432}
1433
1434static int seq_show(struct seq_file *s, void *v)
1435{
1436 const struct nfqnl_instance *inst = v;
1437
1438 seq_printf(s, "%5u %6u %5u %1u %5u %5u %5u %8u %2d\n",
1439 inst->queue_num,
1440 inst->peer_portid, inst->queue_total,
1441 inst->copy_mode, inst->copy_range,
1442 inst->queue_dropped, inst->queue_user_dropped,
1443 inst->id_sequence, 1);
1444 return 0;
1445}
1446
1447static const struct seq_operations nfqnl_seq_ops = {
1448 .start = seq_start,
1449 .next = seq_next,
1450 .stop = seq_stop,
1451 .show = seq_show,
1452};
1453
1454static int nfqnl_open(struct inode *inode, struct file *file)
1455{
1456 return seq_open_net(inode, file, &nfqnl_seq_ops,
1457 sizeof(struct iter_state));
1458}
1459
1460static const struct file_operations nfqnl_file_ops = {
1461 .owner = THIS_MODULE,
1462 .open = nfqnl_open,
1463 .read = seq_read,
1464 .llseek = seq_lseek,
1465 .release = seq_release_net,
1466};
1467
1468#endif /* PROC_FS */
1469
1470static int __net_init nfnl_queue_net_init(struct net *net)
1471{
1472 unsigned int i;
1473 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1474
1475 for (i = 0; i < INSTANCE_BUCKETS; i++)
1476 INIT_HLIST_HEAD(&q->instance_table[i]);
1477
1478 spin_lock_init(&q->instances_lock);
1479
1480#ifdef CONFIG_PROC_FS
1481 if (!proc_create("nfnetlink_queue", 0440,
1482 net->nf.proc_netfilter, &nfqnl_file_ops))
1483 return -ENOMEM;
1484#endif
1485 nf_register_queue_handler(net, &nfqh);
1486 return 0;
1487}
1488
1489static void __net_exit nfnl_queue_net_exit(struct net *net)
1490{
1491 nf_unregister_queue_handler(net);
1492#ifdef CONFIG_PROC_FS
1493 remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1494#endif
1495}
1496
1497static void nfnl_queue_net_exit_batch(struct list_head *net_exit_list)
1498{
1499 synchronize_rcu();
1500}
1501
1502static struct pernet_operations nfnl_queue_net_ops = {
1503 .init = nfnl_queue_net_init,
1504 .exit = nfnl_queue_net_exit,
1505 .exit_batch = nfnl_queue_net_exit_batch,
1506 .id = &nfnl_queue_net_id,
1507 .size = sizeof(struct nfnl_queue_net),
1508};
1509
1510static int __init nfnetlink_queue_init(void)
1511{
1512 int status;
1513
1514 status = register_pernet_subsys(&nfnl_queue_net_ops);
1515 if (status < 0) {
1516 pr_err("nf_queue: failed to register pernet ops\n");
1517 goto out;
1518 }
1519
1520 netlink_register_notifier(&nfqnl_rtnl_notifier);
1521 status = nfnetlink_subsys_register(&nfqnl_subsys);
1522 if (status < 0) {
1523 pr_err("nf_queue: failed to create netlink socket\n");
1524 goto cleanup_netlink_notifier;
1525 }
1526
1527 status = register_netdevice_notifier(&nfqnl_dev_notifier);
1528 if (status < 0) {
1529 pr_err("nf_queue: failed to register netdevice notifier\n");
1530 goto cleanup_netlink_subsys;
1531 }
1532
1533 return status;
1534
1535cleanup_netlink_subsys:
1536 nfnetlink_subsys_unregister(&nfqnl_subsys);
1537cleanup_netlink_notifier:
1538 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1539 unregister_pernet_subsys(&nfnl_queue_net_ops);
1540out:
1541 return status;
1542}
1543
1544static void __exit nfnetlink_queue_fini(void)
1545{
1546 unregister_netdevice_notifier(&nfqnl_dev_notifier);
1547 nfnetlink_subsys_unregister(&nfqnl_subsys);
1548 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1549 unregister_pernet_subsys(&nfnl_queue_net_ops);
1550
1551 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1552}
1553
1554MODULE_DESCRIPTION("netfilter packet queue handler");
1555MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1556MODULE_LICENSE("GPL");
1557MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1558
1559module_init(nfnetlink_queue_init);
1560module_exit(nfnetlink_queue_fini);
1/*
2 * This is a module which is used for queueing packets and communicating with
3 * userspace via nfnetlink.
4 *
5 * (C) 2005 by Harald Welte <laforge@netfilter.org>
6 * (C) 2007 by Patrick McHardy <kaber@trash.net>
7 *
8 * Based on the old ipv4-only ip_queue.c:
9 * (C) 2000-2002 James Morris <jmorris@intercode.com.au>
10 * (C) 2003-2005 Netfilter Core Team <coreteam@netfilter.org>
11 *
12 * This program is free software; you can redistribute it and/or modify
13 * it under the terms of the GNU General Public License version 2 as
14 * published by the Free Software Foundation.
15 *
16 */
17#include <linux/module.h>
18#include <linux/skbuff.h>
19#include <linux/init.h>
20#include <linux/spinlock.h>
21#include <linux/slab.h>
22#include <linux/notifier.h>
23#include <linux/netdevice.h>
24#include <linux/netfilter.h>
25#include <linux/proc_fs.h>
26#include <linux/netfilter_ipv4.h>
27#include <linux/netfilter_ipv6.h>
28#include <linux/netfilter/nfnetlink.h>
29#include <linux/netfilter/nfnetlink_queue.h>
30#include <linux/list.h>
31#include <net/sock.h>
32#include <net/netfilter/nf_queue.h>
33
34#include <linux/atomic.h>
35
36#ifdef CONFIG_BRIDGE_NETFILTER
37#include "../bridge/br_private.h"
38#endif
39
40#define NFQNL_QMAX_DEFAULT 1024
41
42struct nfqnl_instance {
43 struct hlist_node hlist; /* global list of queues */
44 struct rcu_head rcu;
45
46 int peer_pid;
47 unsigned int queue_maxlen;
48 unsigned int copy_range;
49 unsigned int queue_dropped;
50 unsigned int queue_user_dropped;
51
52
53 u_int16_t queue_num; /* number of this queue */
54 u_int8_t copy_mode;
55/*
56 * Following fields are dirtied for each queued packet,
57 * keep them in same cache line if possible.
58 */
59 spinlock_t lock;
60 unsigned int queue_total;
61 unsigned int id_sequence; /* 'sequence' of pkt ids */
62 struct list_head queue_list; /* packets in queue */
63};
64
65typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
66
67static DEFINE_SPINLOCK(instances_lock);
68
69#define INSTANCE_BUCKETS 16
70static struct hlist_head instance_table[INSTANCE_BUCKETS] __read_mostly;
71
72static inline u_int8_t instance_hashfn(u_int16_t queue_num)
73{
74 return ((queue_num >> 8) | queue_num) % INSTANCE_BUCKETS;
75}
76
77static struct nfqnl_instance *
78instance_lookup(u_int16_t queue_num)
79{
80 struct hlist_head *head;
81 struct hlist_node *pos;
82 struct nfqnl_instance *inst;
83
84 head = &instance_table[instance_hashfn(queue_num)];
85 hlist_for_each_entry_rcu(inst, pos, head, hlist) {
86 if (inst->queue_num == queue_num)
87 return inst;
88 }
89 return NULL;
90}
91
92static struct nfqnl_instance *
93instance_create(u_int16_t queue_num, int pid)
94{
95 struct nfqnl_instance *inst;
96 unsigned int h;
97 int err;
98
99 spin_lock(&instances_lock);
100 if (instance_lookup(queue_num)) {
101 err = -EEXIST;
102 goto out_unlock;
103 }
104
105 inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
106 if (!inst) {
107 err = -ENOMEM;
108 goto out_unlock;
109 }
110
111 inst->queue_num = queue_num;
112 inst->peer_pid = pid;
113 inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
114 inst->copy_range = 0xfffff;
115 inst->copy_mode = NFQNL_COPY_NONE;
116 spin_lock_init(&inst->lock);
117 INIT_LIST_HEAD(&inst->queue_list);
118
119 if (!try_module_get(THIS_MODULE)) {
120 err = -EAGAIN;
121 goto out_free;
122 }
123
124 h = instance_hashfn(queue_num);
125 hlist_add_head_rcu(&inst->hlist, &instance_table[h]);
126
127 spin_unlock(&instances_lock);
128
129 return inst;
130
131out_free:
132 kfree(inst);
133out_unlock:
134 spin_unlock(&instances_lock);
135 return ERR_PTR(err);
136}
137
138static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
139 unsigned long data);
140
141static void
142instance_destroy_rcu(struct rcu_head *head)
143{
144 struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
145 rcu);
146
147 nfqnl_flush(inst, NULL, 0);
148 kfree(inst);
149 module_put(THIS_MODULE);
150}
151
152static void
153__instance_destroy(struct nfqnl_instance *inst)
154{
155 hlist_del_rcu(&inst->hlist);
156 call_rcu(&inst->rcu, instance_destroy_rcu);
157}
158
159static void
160instance_destroy(struct nfqnl_instance *inst)
161{
162 spin_lock(&instances_lock);
163 __instance_destroy(inst);
164 spin_unlock(&instances_lock);
165}
166
167static inline void
168__enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
169{
170 list_add_tail(&entry->list, &queue->queue_list);
171 queue->queue_total++;
172}
173
174static void
175__dequeue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
176{
177 list_del(&entry->list);
178 queue->queue_total--;
179}
180
181static struct nf_queue_entry *
182find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
183{
184 struct nf_queue_entry *entry = NULL, *i;
185
186 spin_lock_bh(&queue->lock);
187
188 list_for_each_entry(i, &queue->queue_list, list) {
189 if (i->id == id) {
190 entry = i;
191 break;
192 }
193 }
194
195 if (entry)
196 __dequeue_entry(queue, entry);
197
198 spin_unlock_bh(&queue->lock);
199
200 return entry;
201}
202
203static void
204nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
205{
206 struct nf_queue_entry *entry, *next;
207
208 spin_lock_bh(&queue->lock);
209 list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
210 if (!cmpfn || cmpfn(entry, data)) {
211 list_del(&entry->list);
212 queue->queue_total--;
213 nf_reinject(entry, NF_DROP);
214 }
215 }
216 spin_unlock_bh(&queue->lock);
217}
218
219static struct sk_buff *
220nfqnl_build_packet_message(struct nfqnl_instance *queue,
221 struct nf_queue_entry *entry,
222 __be32 **packet_id_ptr)
223{
224 sk_buff_data_t old_tail;
225 size_t size;
226 size_t data_len = 0;
227 struct sk_buff *skb;
228 struct nlattr *nla;
229 struct nfqnl_msg_packet_hdr *pmsg;
230 struct nlmsghdr *nlh;
231 struct nfgenmsg *nfmsg;
232 struct sk_buff *entskb = entry->skb;
233 struct net_device *indev;
234 struct net_device *outdev;
235
236 size = NLMSG_SPACE(sizeof(struct nfgenmsg))
237 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
238 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
239 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
240#ifdef CONFIG_BRIDGE_NETFILTER
241 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
242 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
243#endif
244 + nla_total_size(sizeof(u_int32_t)) /* mark */
245 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
246 + nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
247
248 outdev = entry->outdev;
249
250 switch ((enum nfqnl_config_mode)ACCESS_ONCE(queue->copy_mode)) {
251 case NFQNL_COPY_META:
252 case NFQNL_COPY_NONE:
253 break;
254
255 case NFQNL_COPY_PACKET:
256 if (entskb->ip_summed == CHECKSUM_PARTIAL &&
257 skb_checksum_help(entskb))
258 return NULL;
259
260 data_len = ACCESS_ONCE(queue->copy_range);
261 if (data_len == 0 || data_len > entskb->len)
262 data_len = entskb->len;
263
264 size += nla_total_size(data_len);
265 break;
266 }
267
268
269 skb = alloc_skb(size, GFP_ATOMIC);
270 if (!skb)
271 goto nlmsg_failure;
272
273 old_tail = skb->tail;
274 nlh = NLMSG_PUT(skb, 0, 0,
275 NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
276 sizeof(struct nfgenmsg));
277 nfmsg = NLMSG_DATA(nlh);
278 nfmsg->nfgen_family = entry->pf;
279 nfmsg->version = NFNETLINK_V0;
280 nfmsg->res_id = htons(queue->queue_num);
281
282 nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
283 pmsg = nla_data(nla);
284 pmsg->hw_protocol = entskb->protocol;
285 pmsg->hook = entry->hook;
286 *packet_id_ptr = &pmsg->packet_id;
287
288 indev = entry->indev;
289 if (indev) {
290#ifndef CONFIG_BRIDGE_NETFILTER
291 NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex));
292#else
293 if (entry->pf == PF_BRIDGE) {
294 /* Case 1: indev is physical input device, we need to
295 * look for bridge group (when called from
296 * netfilter_bridge) */
297 NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSINDEV,
298 htonl(indev->ifindex));
299 /* this is the bridge group "brX" */
300 /* rcu_read_lock()ed by __nf_queue */
301 NLA_PUT_BE32(skb, NFQA_IFINDEX_INDEV,
302 htonl(br_port_get_rcu(indev)->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 /* rcu_read_lock()ed by __nf_queue */
327 NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV,
328 htonl(br_port_get_rcu(outdev)->br->dev->ifindex));
329 } else {
330 /* Case 2: outdev is bridge group, we need to look for
331 * physical output device (when called from ipv4) */
332 NLA_PUT_BE32(skb, NFQA_IFINDEX_OUTDEV,
333 htonl(outdev->ifindex));
334 if (entskb->nf_bridge && entskb->nf_bridge->physoutdev)
335 NLA_PUT_BE32(skb, NFQA_IFINDEX_PHYSOUTDEV,
336 htonl(entskb->nf_bridge->physoutdev->ifindex));
337 }
338#endif
339 }
340
341 if (entskb->mark)
342 NLA_PUT_BE32(skb, NFQA_MARK, htonl(entskb->mark));
343
344 if (indev && entskb->dev &&
345 entskb->mac_header != entskb->network_header) {
346 struct nfqnl_msg_packet_hw phw;
347 int len = dev_parse_header(entskb, phw.hw_addr);
348 if (len) {
349 phw.hw_addrlen = htons(len);
350 NLA_PUT(skb, NFQA_HWADDR, sizeof(phw), &phw);
351 }
352 }
353
354 if (entskb->tstamp.tv64) {
355 struct nfqnl_msg_packet_timestamp ts;
356 struct timeval tv = ktime_to_timeval(entskb->tstamp);
357 ts.sec = cpu_to_be64(tv.tv_sec);
358 ts.usec = cpu_to_be64(tv.tv_usec);
359
360 NLA_PUT(skb, NFQA_TIMESTAMP, sizeof(ts), &ts);
361 }
362
363 if (data_len) {
364 struct nlattr *nla;
365 int sz = nla_attr_size(data_len);
366
367 if (skb_tailroom(skb) < nla_total_size(data_len)) {
368 printk(KERN_WARNING "nf_queue: no tailroom!\n");
369 goto nlmsg_failure;
370 }
371
372 nla = (struct nlattr *)skb_put(skb, nla_total_size(data_len));
373 nla->nla_type = NFQA_PAYLOAD;
374 nla->nla_len = sz;
375
376 if (skb_copy_bits(entskb, 0, nla_data(nla), data_len))
377 BUG();
378 }
379
380 nlh->nlmsg_len = skb->tail - old_tail;
381 return skb;
382
383nlmsg_failure:
384nla_put_failure:
385 if (skb)
386 kfree_skb(skb);
387 if (net_ratelimit())
388 printk(KERN_ERR "nf_queue: error creating packet message\n");
389 return NULL;
390}
391
392static int
393nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
394{
395 struct sk_buff *nskb;
396 struct nfqnl_instance *queue;
397 int err = -ENOBUFS;
398 __be32 *packet_id_ptr;
399
400 /* rcu_read_lock()ed by nf_hook_slow() */
401 queue = instance_lookup(queuenum);
402 if (!queue) {
403 err = -ESRCH;
404 goto err_out;
405 }
406
407 if (queue->copy_mode == NFQNL_COPY_NONE) {
408 err = -EINVAL;
409 goto err_out;
410 }
411
412 nskb = nfqnl_build_packet_message(queue, entry, &packet_id_ptr);
413 if (nskb == NULL) {
414 err = -ENOMEM;
415 goto err_out;
416 }
417 spin_lock_bh(&queue->lock);
418
419 if (!queue->peer_pid) {
420 err = -EINVAL;
421 goto err_out_free_nskb;
422 }
423 if (queue->queue_total >= queue->queue_maxlen) {
424 queue->queue_dropped++;
425 if (net_ratelimit())
426 printk(KERN_WARNING "nf_queue: full at %d entries, "
427 "dropping packets(s).\n",
428 queue->queue_total);
429 goto err_out_free_nskb;
430 }
431 entry->id = ++queue->id_sequence;
432 *packet_id_ptr = htonl(entry->id);
433
434 /* nfnetlink_unicast will either free the nskb or add it to a socket */
435 err = nfnetlink_unicast(nskb, &init_net, queue->peer_pid, MSG_DONTWAIT);
436 if (err < 0) {
437 queue->queue_user_dropped++;
438 goto err_out_unlock;
439 }
440
441 __enqueue_entry(queue, entry);
442
443 spin_unlock_bh(&queue->lock);
444 return 0;
445
446err_out_free_nskb:
447 kfree_skb(nskb);
448err_out_unlock:
449 spin_unlock_bh(&queue->lock);
450err_out:
451 return err;
452}
453
454static int
455nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e)
456{
457 struct sk_buff *nskb;
458 int diff;
459
460 diff = data_len - e->skb->len;
461 if (diff < 0) {
462 if (pskb_trim(e->skb, data_len))
463 return -ENOMEM;
464 } else if (diff > 0) {
465 if (data_len > 0xFFFF)
466 return -EINVAL;
467 if (diff > skb_tailroom(e->skb)) {
468 nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
469 diff, GFP_ATOMIC);
470 if (!nskb) {
471 printk(KERN_WARNING "nf_queue: OOM "
472 "in mangle, dropping packet\n");
473 return -ENOMEM;
474 }
475 kfree_skb(e->skb);
476 e->skb = nskb;
477 }
478 skb_put(e->skb, diff);
479 }
480 if (!skb_make_writable(e->skb, data_len))
481 return -ENOMEM;
482 skb_copy_to_linear_data(e->skb, data, data_len);
483 e->skb->ip_summed = CHECKSUM_NONE;
484 return 0;
485}
486
487static int
488nfqnl_set_mode(struct nfqnl_instance *queue,
489 unsigned char mode, unsigned int range)
490{
491 int status = 0;
492
493 spin_lock_bh(&queue->lock);
494 switch (mode) {
495 case NFQNL_COPY_NONE:
496 case NFQNL_COPY_META:
497 queue->copy_mode = mode;
498 queue->copy_range = 0;
499 break;
500
501 case NFQNL_COPY_PACKET:
502 queue->copy_mode = mode;
503 /* we're using struct nlattr which has 16bit nla_len */
504 if (range > 0xffff)
505 queue->copy_range = 0xffff;
506 else
507 queue->copy_range = range;
508 break;
509
510 default:
511 status = -EINVAL;
512
513 }
514 spin_unlock_bh(&queue->lock);
515
516 return status;
517}
518
519static int
520dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
521{
522 if (entry->indev)
523 if (entry->indev->ifindex == ifindex)
524 return 1;
525 if (entry->outdev)
526 if (entry->outdev->ifindex == ifindex)
527 return 1;
528#ifdef CONFIG_BRIDGE_NETFILTER
529 if (entry->skb->nf_bridge) {
530 if (entry->skb->nf_bridge->physindev &&
531 entry->skb->nf_bridge->physindev->ifindex == ifindex)
532 return 1;
533 if (entry->skb->nf_bridge->physoutdev &&
534 entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
535 return 1;
536 }
537#endif
538 return 0;
539}
540
541/* drop all packets with either indev or outdev == ifindex from all queue
542 * instances */
543static void
544nfqnl_dev_drop(int ifindex)
545{
546 int i;
547
548 rcu_read_lock();
549
550 for (i = 0; i < INSTANCE_BUCKETS; i++) {
551 struct hlist_node *tmp;
552 struct nfqnl_instance *inst;
553 struct hlist_head *head = &instance_table[i];
554
555 hlist_for_each_entry_rcu(inst, tmp, head, hlist)
556 nfqnl_flush(inst, dev_cmp, ifindex);
557 }
558
559 rcu_read_unlock();
560}
561
562#define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
563
564static int
565nfqnl_rcv_dev_event(struct notifier_block *this,
566 unsigned long event, void *ptr)
567{
568 struct net_device *dev = ptr;
569
570 if (!net_eq(dev_net(dev), &init_net))
571 return NOTIFY_DONE;
572
573 /* Drop any packets associated with the downed device */
574 if (event == NETDEV_DOWN)
575 nfqnl_dev_drop(dev->ifindex);
576 return NOTIFY_DONE;
577}
578
579static struct notifier_block nfqnl_dev_notifier = {
580 .notifier_call = nfqnl_rcv_dev_event,
581};
582
583static int
584nfqnl_rcv_nl_event(struct notifier_block *this,
585 unsigned long event, void *ptr)
586{
587 struct netlink_notify *n = ptr;
588
589 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
590 int i;
591
592 /* destroy all instances for this pid */
593 spin_lock(&instances_lock);
594 for (i = 0; i < INSTANCE_BUCKETS; i++) {
595 struct hlist_node *tmp, *t2;
596 struct nfqnl_instance *inst;
597 struct hlist_head *head = &instance_table[i];
598
599 hlist_for_each_entry_safe(inst, tmp, t2, head, hlist) {
600 if ((n->net == &init_net) &&
601 (n->pid == inst->peer_pid))
602 __instance_destroy(inst);
603 }
604 }
605 spin_unlock(&instances_lock);
606 }
607 return NOTIFY_DONE;
608}
609
610static struct notifier_block nfqnl_rtnl_notifier = {
611 .notifier_call = nfqnl_rcv_nl_event,
612};
613
614static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
615 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
616 [NFQA_MARK] = { .type = NLA_U32 },
617 [NFQA_PAYLOAD] = { .type = NLA_UNSPEC },
618};
619
620static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
621 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
622 [NFQA_MARK] = { .type = NLA_U32 },
623};
624
625static struct nfqnl_instance *verdict_instance_lookup(u16 queue_num, int nlpid)
626{
627 struct nfqnl_instance *queue;
628
629 queue = instance_lookup(queue_num);
630 if (!queue)
631 return ERR_PTR(-ENODEV);
632
633 if (queue->peer_pid != nlpid)
634 return ERR_PTR(-EPERM);
635
636 return queue;
637}
638
639static struct nfqnl_msg_verdict_hdr*
640verdicthdr_get(const struct nlattr * const nfqa[])
641{
642 struct nfqnl_msg_verdict_hdr *vhdr;
643 unsigned int verdict;
644
645 if (!nfqa[NFQA_VERDICT_HDR])
646 return NULL;
647
648 vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
649 verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
650 if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
651 return NULL;
652 return vhdr;
653}
654
655static int nfq_id_after(unsigned int id, unsigned int max)
656{
657 return (int)(id - max) > 0;
658}
659
660static int
661nfqnl_recv_verdict_batch(struct sock *ctnl, struct sk_buff *skb,
662 const struct nlmsghdr *nlh,
663 const struct nlattr * const nfqa[])
664{
665 struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
666 struct nf_queue_entry *entry, *tmp;
667 unsigned int verdict, maxid;
668 struct nfqnl_msg_verdict_hdr *vhdr;
669 struct nfqnl_instance *queue;
670 LIST_HEAD(batch_list);
671 u16 queue_num = ntohs(nfmsg->res_id);
672
673 queue = verdict_instance_lookup(queue_num, NETLINK_CB(skb).pid);
674 if (IS_ERR(queue))
675 return PTR_ERR(queue);
676
677 vhdr = verdicthdr_get(nfqa);
678 if (!vhdr)
679 return -EINVAL;
680
681 verdict = ntohl(vhdr->verdict);
682 maxid = ntohl(vhdr->id);
683
684 spin_lock_bh(&queue->lock);
685
686 list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
687 if (nfq_id_after(entry->id, maxid))
688 break;
689 __dequeue_entry(queue, entry);
690 list_add_tail(&entry->list, &batch_list);
691 }
692
693 spin_unlock_bh(&queue->lock);
694
695 if (list_empty(&batch_list))
696 return -ENOENT;
697
698 list_for_each_entry_safe(entry, tmp, &batch_list, list) {
699 if (nfqa[NFQA_MARK])
700 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
701 nf_reinject(entry, verdict);
702 }
703 return 0;
704}
705
706static int
707nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
708 const struct nlmsghdr *nlh,
709 const struct nlattr * const nfqa[])
710{
711 struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
712 u_int16_t queue_num = ntohs(nfmsg->res_id);
713
714 struct nfqnl_msg_verdict_hdr *vhdr;
715 struct nfqnl_instance *queue;
716 unsigned int verdict;
717 struct nf_queue_entry *entry;
718
719 queue = instance_lookup(queue_num);
720 if (!queue)
721
722 queue = verdict_instance_lookup(queue_num, NETLINK_CB(skb).pid);
723 if (IS_ERR(queue))
724 return PTR_ERR(queue);
725
726 vhdr = verdicthdr_get(nfqa);
727 if (!vhdr)
728 return -EINVAL;
729
730 verdict = ntohl(vhdr->verdict);
731
732 entry = find_dequeue_entry(queue, ntohl(vhdr->id));
733 if (entry == NULL)
734 return -ENOENT;
735
736 if (nfqa[NFQA_PAYLOAD]) {
737 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
738 nla_len(nfqa[NFQA_PAYLOAD]), entry) < 0)
739 verdict = NF_DROP;
740 }
741
742 if (nfqa[NFQA_MARK])
743 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
744
745 nf_reinject(entry, verdict);
746 return 0;
747}
748
749static int
750nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
751 const struct nlmsghdr *nlh,
752 const struct nlattr * const nfqa[])
753{
754 return -ENOTSUPP;
755}
756
757static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
758 [NFQA_CFG_CMD] = { .len = sizeof(struct nfqnl_msg_config_cmd) },
759 [NFQA_CFG_PARAMS] = { .len = sizeof(struct nfqnl_msg_config_params) },
760};
761
762static const struct nf_queue_handler nfqh = {
763 .name = "nf_queue",
764 .outfn = &nfqnl_enqueue_packet,
765};
766
767static int
768nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
769 const struct nlmsghdr *nlh,
770 const struct nlattr * const nfqa[])
771{
772 struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
773 u_int16_t queue_num = ntohs(nfmsg->res_id);
774 struct nfqnl_instance *queue;
775 struct nfqnl_msg_config_cmd *cmd = NULL;
776 int ret = 0;
777
778 if (nfqa[NFQA_CFG_CMD]) {
779 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
780
781 /* Commands without queue context - might sleep */
782 switch (cmd->command) {
783 case NFQNL_CFG_CMD_PF_BIND:
784 return nf_register_queue_handler(ntohs(cmd->pf),
785 &nfqh);
786 case NFQNL_CFG_CMD_PF_UNBIND:
787 return nf_unregister_queue_handler(ntohs(cmd->pf),
788 &nfqh);
789 }
790 }
791
792 rcu_read_lock();
793 queue = instance_lookup(queue_num);
794 if (queue && queue->peer_pid != NETLINK_CB(skb).pid) {
795 ret = -EPERM;
796 goto err_out_unlock;
797 }
798
799 if (cmd != NULL) {
800 switch (cmd->command) {
801 case NFQNL_CFG_CMD_BIND:
802 if (queue) {
803 ret = -EBUSY;
804 goto err_out_unlock;
805 }
806 queue = instance_create(queue_num, NETLINK_CB(skb).pid);
807 if (IS_ERR(queue)) {
808 ret = PTR_ERR(queue);
809 goto err_out_unlock;
810 }
811 break;
812 case NFQNL_CFG_CMD_UNBIND:
813 if (!queue) {
814 ret = -ENODEV;
815 goto err_out_unlock;
816 }
817 instance_destroy(queue);
818 break;
819 case NFQNL_CFG_CMD_PF_BIND:
820 case NFQNL_CFG_CMD_PF_UNBIND:
821 break;
822 default:
823 ret = -ENOTSUPP;
824 break;
825 }
826 }
827
828 if (nfqa[NFQA_CFG_PARAMS]) {
829 struct nfqnl_msg_config_params *params;
830
831 if (!queue) {
832 ret = -ENODEV;
833 goto err_out_unlock;
834 }
835 params = nla_data(nfqa[NFQA_CFG_PARAMS]);
836 nfqnl_set_mode(queue, params->copy_mode,
837 ntohl(params->copy_range));
838 }
839
840 if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
841 __be32 *queue_maxlen;
842
843 if (!queue) {
844 ret = -ENODEV;
845 goto err_out_unlock;
846 }
847 queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
848 spin_lock_bh(&queue->lock);
849 queue->queue_maxlen = ntohl(*queue_maxlen);
850 spin_unlock_bh(&queue->lock);
851 }
852
853err_out_unlock:
854 rcu_read_unlock();
855 return ret;
856}
857
858static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
859 [NFQNL_MSG_PACKET] = { .call_rcu = nfqnl_recv_unsupp,
860 .attr_count = NFQA_MAX, },
861 [NFQNL_MSG_VERDICT] = { .call_rcu = nfqnl_recv_verdict,
862 .attr_count = NFQA_MAX,
863 .policy = nfqa_verdict_policy },
864 [NFQNL_MSG_CONFIG] = { .call = nfqnl_recv_config,
865 .attr_count = NFQA_CFG_MAX,
866 .policy = nfqa_cfg_policy },
867 [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
868 .attr_count = NFQA_MAX,
869 .policy = nfqa_verdict_batch_policy },
870};
871
872static const struct nfnetlink_subsystem nfqnl_subsys = {
873 .name = "nf_queue",
874 .subsys_id = NFNL_SUBSYS_QUEUE,
875 .cb_count = NFQNL_MSG_MAX,
876 .cb = nfqnl_cb,
877};
878
879#ifdef CONFIG_PROC_FS
880struct iter_state {
881 unsigned int bucket;
882};
883
884static struct hlist_node *get_first(struct seq_file *seq)
885{
886 struct iter_state *st = seq->private;
887
888 if (!st)
889 return NULL;
890
891 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
892 if (!hlist_empty(&instance_table[st->bucket]))
893 return instance_table[st->bucket].first;
894 }
895 return NULL;
896}
897
898static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
899{
900 struct iter_state *st = seq->private;
901
902 h = h->next;
903 while (!h) {
904 if (++st->bucket >= INSTANCE_BUCKETS)
905 return NULL;
906
907 h = instance_table[st->bucket].first;
908 }
909 return h;
910}
911
912static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
913{
914 struct hlist_node *head;
915 head = get_first(seq);
916
917 if (head)
918 while (pos && (head = get_next(seq, head)))
919 pos--;
920 return pos ? NULL : head;
921}
922
923static void *seq_start(struct seq_file *seq, loff_t *pos)
924 __acquires(instances_lock)
925{
926 spin_lock(&instances_lock);
927 return get_idx(seq, *pos);
928}
929
930static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
931{
932 (*pos)++;
933 return get_next(s, v);
934}
935
936static void seq_stop(struct seq_file *s, void *v)
937 __releases(instances_lock)
938{
939 spin_unlock(&instances_lock);
940}
941
942static int seq_show(struct seq_file *s, void *v)
943{
944 const struct nfqnl_instance *inst = v;
945
946 return seq_printf(s, "%5d %6d %5d %1d %5d %5d %5d %8d %2d\n",
947 inst->queue_num,
948 inst->peer_pid, inst->queue_total,
949 inst->copy_mode, inst->copy_range,
950 inst->queue_dropped, inst->queue_user_dropped,
951 inst->id_sequence, 1);
952}
953
954static const struct seq_operations nfqnl_seq_ops = {
955 .start = seq_start,
956 .next = seq_next,
957 .stop = seq_stop,
958 .show = seq_show,
959};
960
961static int nfqnl_open(struct inode *inode, struct file *file)
962{
963 return seq_open_private(file, &nfqnl_seq_ops,
964 sizeof(struct iter_state));
965}
966
967static const struct file_operations nfqnl_file_ops = {
968 .owner = THIS_MODULE,
969 .open = nfqnl_open,
970 .read = seq_read,
971 .llseek = seq_lseek,
972 .release = seq_release_private,
973};
974
975#endif /* PROC_FS */
976
977static int __init nfnetlink_queue_init(void)
978{
979 int i, status = -ENOMEM;
980
981 for (i = 0; i < INSTANCE_BUCKETS; i++)
982 INIT_HLIST_HEAD(&instance_table[i]);
983
984 netlink_register_notifier(&nfqnl_rtnl_notifier);
985 status = nfnetlink_subsys_register(&nfqnl_subsys);
986 if (status < 0) {
987 printk(KERN_ERR "nf_queue: failed to create netlink socket\n");
988 goto cleanup_netlink_notifier;
989 }
990
991#ifdef CONFIG_PROC_FS
992 if (!proc_create("nfnetlink_queue", 0440,
993 proc_net_netfilter, &nfqnl_file_ops))
994 goto cleanup_subsys;
995#endif
996
997 register_netdevice_notifier(&nfqnl_dev_notifier);
998 return status;
999
1000#ifdef CONFIG_PROC_FS
1001cleanup_subsys:
1002 nfnetlink_subsys_unregister(&nfqnl_subsys);
1003#endif
1004cleanup_netlink_notifier:
1005 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1006 return status;
1007}
1008
1009static void __exit nfnetlink_queue_fini(void)
1010{
1011 nf_unregister_queue_handlers(&nfqh);
1012 unregister_netdevice_notifier(&nfqnl_dev_notifier);
1013#ifdef CONFIG_PROC_FS
1014 remove_proc_entry("nfnetlink_queue", proc_net_netfilter);
1015#endif
1016 nfnetlink_subsys_unregister(&nfqnl_subsys);
1017 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1018
1019 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1020}
1021
1022MODULE_DESCRIPTION("netfilter packet queue handler");
1023MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1024MODULE_LICENSE("GPL");
1025MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1026
1027module_init(nfnetlink_queue_init);
1028module_exit(nfnetlink_queue_fini);