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// SPDX-License-Identifier: GPL-2.0-only
2/*
3 * This is a module which is used for queueing packets and communicating with
4 * userspace via nfnetlink.
5 *
6 * (C) 2005 by Harald Welte <laforge@netfilter.org>
7 * (C) 2007 by Patrick McHardy <kaber@trash.net>
8 *
9 * Based on the old ipv4-only ip_queue.c:
10 * (C) 2000-2002 James Morris <jmorris@intercode.com.au>
11 * (C) 2003-2005 Netfilter Core Team <coreteam@netfilter.org>
12 */
13
14#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
15
16#include <linux/module.h>
17#include <linux/skbuff.h>
18#include <linux/init.h>
19#include <linux/spinlock.h>
20#include <linux/slab.h>
21#include <linux/notifier.h>
22#include <linux/netdevice.h>
23#include <linux/netfilter.h>
24#include <linux/proc_fs.h>
25#include <linux/netfilter_ipv4.h>
26#include <linux/netfilter_ipv6.h>
27#include <linux/netfilter_bridge.h>
28#include <linux/netfilter/nfnetlink.h>
29#include <linux/netfilter/nfnetlink_queue.h>
30#include <linux/netfilter/nf_conntrack_common.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
37#include <linux/atomic.h>
38
39#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
40#include "../bridge/br_private.h"
41#endif
42
43#if IS_ENABLED(CONFIG_NF_CONNTRACK)
44#include <net/netfilter/nf_conntrack.h>
45#endif
46
47#define NFQNL_QMAX_DEFAULT 1024
48
49/* We're using struct nlattr which has 16bit nla_len. Note that nla_len
50 * includes the header length. Thus, the maximum packet length that we
51 * support is 65531 bytes. We send truncated packets if the specified length
52 * is larger than that. Userspace can check for presence of NFQA_CAP_LEN
53 * attribute to detect truncation.
54 */
55#define NFQNL_MAX_COPY_RANGE (0xffff - NLA_HDRLEN)
56
57struct nfqnl_instance {
58 struct hlist_node hlist; /* global list of queues */
59 struct rcu_head rcu;
60
61 u32 peer_portid;
62 unsigned int queue_maxlen;
63 unsigned int copy_range;
64 unsigned int queue_dropped;
65 unsigned int queue_user_dropped;
66
67
68 u_int16_t queue_num; /* number of this queue */
69 u_int8_t copy_mode;
70 u_int32_t flags; /* Set using NFQA_CFG_FLAGS */
71/*
72 * Following fields are dirtied for each queued packet,
73 * keep them in same cache line if possible.
74 */
75 spinlock_t lock ____cacheline_aligned_in_smp;
76 unsigned int queue_total;
77 unsigned int id_sequence; /* 'sequence' of pkt ids */
78 struct list_head queue_list; /* packets in queue */
79};
80
81typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
82
83static unsigned int nfnl_queue_net_id __read_mostly;
84
85#define INSTANCE_BUCKETS 16
86struct nfnl_queue_net {
87 spinlock_t instances_lock;
88 struct hlist_head instance_table[INSTANCE_BUCKETS];
89};
90
91static struct nfnl_queue_net *nfnl_queue_pernet(struct net *net)
92{
93 return net_generic(net, nfnl_queue_net_id);
94}
95
96static inline u_int8_t instance_hashfn(u_int16_t queue_num)
97{
98 return ((queue_num >> 8) ^ queue_num) % INSTANCE_BUCKETS;
99}
100
101static struct nfqnl_instance *
102instance_lookup(struct nfnl_queue_net *q, u_int16_t queue_num)
103{
104 struct hlist_head *head;
105 struct nfqnl_instance *inst;
106
107 head = &q->instance_table[instance_hashfn(queue_num)];
108 hlist_for_each_entry_rcu(inst, head, hlist) {
109 if (inst->queue_num == queue_num)
110 return inst;
111 }
112 return NULL;
113}
114
115static struct nfqnl_instance *
116instance_create(struct nfnl_queue_net *q, u_int16_t queue_num, u32 portid)
117{
118 struct nfqnl_instance *inst;
119 unsigned int h;
120 int err;
121
122 spin_lock(&q->instances_lock);
123 if (instance_lookup(q, queue_num)) {
124 err = -EEXIST;
125 goto out_unlock;
126 }
127
128 inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
129 if (!inst) {
130 err = -ENOMEM;
131 goto out_unlock;
132 }
133
134 inst->queue_num = queue_num;
135 inst->peer_portid = portid;
136 inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
137 inst->copy_range = NFQNL_MAX_COPY_RANGE;
138 inst->copy_mode = NFQNL_COPY_NONE;
139 spin_lock_init(&inst->lock);
140 INIT_LIST_HEAD(&inst->queue_list);
141
142 if (!try_module_get(THIS_MODULE)) {
143 err = -EAGAIN;
144 goto out_free;
145 }
146
147 h = instance_hashfn(queue_num);
148 hlist_add_head_rcu(&inst->hlist, &q->instance_table[h]);
149
150 spin_unlock(&q->instances_lock);
151
152 return inst;
153
154out_free:
155 kfree(inst);
156out_unlock:
157 spin_unlock(&q->instances_lock);
158 return ERR_PTR(err);
159}
160
161static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
162 unsigned long data);
163
164static void
165instance_destroy_rcu(struct rcu_head *head)
166{
167 struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
168 rcu);
169
170 nfqnl_flush(inst, NULL, 0);
171 kfree(inst);
172 module_put(THIS_MODULE);
173}
174
175static void
176__instance_destroy(struct nfqnl_instance *inst)
177{
178 hlist_del_rcu(&inst->hlist);
179 call_rcu(&inst->rcu, instance_destroy_rcu);
180}
181
182static void
183instance_destroy(struct nfnl_queue_net *q, struct nfqnl_instance *inst)
184{
185 spin_lock(&q->instances_lock);
186 __instance_destroy(inst);
187 spin_unlock(&q->instances_lock);
188}
189
190static inline void
191__enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
192{
193 list_add_tail(&entry->list, &queue->queue_list);
194 queue->queue_total++;
195}
196
197static void
198__dequeue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
199{
200 list_del(&entry->list);
201 queue->queue_total--;
202}
203
204static struct nf_queue_entry *
205find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
206{
207 struct nf_queue_entry *entry = NULL, *i;
208
209 spin_lock_bh(&queue->lock);
210
211 list_for_each_entry(i, &queue->queue_list, list) {
212 if (i->id == id) {
213 entry = i;
214 break;
215 }
216 }
217
218 if (entry)
219 __dequeue_entry(queue, entry);
220
221 spin_unlock_bh(&queue->lock);
222
223 return entry;
224}
225
226static void nfqnl_reinject(struct nf_queue_entry *entry, unsigned int verdict)
227{
228 const struct nf_ct_hook *ct_hook;
229 int err;
230
231 if (verdict == NF_ACCEPT ||
232 verdict == NF_REPEAT ||
233 verdict == NF_STOP) {
234 rcu_read_lock();
235 ct_hook = rcu_dereference(nf_ct_hook);
236 if (ct_hook) {
237 err = ct_hook->update(entry->state.net, entry->skb);
238 if (err < 0)
239 verdict = NF_DROP;
240 }
241 rcu_read_unlock();
242 }
243 nf_reinject(entry, verdict);
244}
245
246static void
247nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
248{
249 struct nf_queue_entry *entry, *next;
250
251 spin_lock_bh(&queue->lock);
252 list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
253 if (!cmpfn || cmpfn(entry, data)) {
254 list_del(&entry->list);
255 queue->queue_total--;
256 nfqnl_reinject(entry, NF_DROP);
257 }
258 }
259 spin_unlock_bh(&queue->lock);
260}
261
262static int
263nfqnl_put_packet_info(struct sk_buff *nlskb, struct sk_buff *packet,
264 bool csum_verify)
265{
266 __u32 flags = 0;
267
268 if (packet->ip_summed == CHECKSUM_PARTIAL)
269 flags = NFQA_SKB_CSUMNOTREADY;
270 else if (csum_verify)
271 flags = NFQA_SKB_CSUM_NOTVERIFIED;
272
273 if (skb_is_gso(packet))
274 flags |= NFQA_SKB_GSO;
275
276 return flags ? nla_put_be32(nlskb, NFQA_SKB_INFO, htonl(flags)) : 0;
277}
278
279static int nfqnl_put_sk_uidgid(struct sk_buff *skb, struct sock *sk)
280{
281 const struct cred *cred;
282
283 if (!sk_fullsock(sk))
284 return 0;
285
286 read_lock_bh(&sk->sk_callback_lock);
287 if (sk->sk_socket && sk->sk_socket->file) {
288 cred = sk->sk_socket->file->f_cred;
289 if (nla_put_be32(skb, NFQA_UID,
290 htonl(from_kuid_munged(&init_user_ns, cred->fsuid))))
291 goto nla_put_failure;
292 if (nla_put_be32(skb, NFQA_GID,
293 htonl(from_kgid_munged(&init_user_ns, cred->fsgid))))
294 goto nla_put_failure;
295 }
296 read_unlock_bh(&sk->sk_callback_lock);
297 return 0;
298
299nla_put_failure:
300 read_unlock_bh(&sk->sk_callback_lock);
301 return -1;
302}
303
304static u32 nfqnl_get_sk_secctx(struct sk_buff *skb, char **secdata)
305{
306 u32 seclen = 0;
307#if IS_ENABLED(CONFIG_NETWORK_SECMARK)
308 if (!skb || !sk_fullsock(skb->sk))
309 return 0;
310
311 read_lock_bh(&skb->sk->sk_callback_lock);
312
313 if (skb->secmark)
314 security_secid_to_secctx(skb->secmark, secdata, &seclen);
315
316 read_unlock_bh(&skb->sk->sk_callback_lock);
317#endif
318 return seclen;
319}
320
321static u32 nfqnl_get_bridge_size(struct nf_queue_entry *entry)
322{
323 struct sk_buff *entskb = entry->skb;
324 u32 nlalen = 0;
325
326 if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb))
327 return 0;
328
329 if (skb_vlan_tag_present(entskb))
330 nlalen += nla_total_size(nla_total_size(sizeof(__be16)) +
331 nla_total_size(sizeof(__be16)));
332
333 if (entskb->network_header > entskb->mac_header)
334 nlalen += nla_total_size((entskb->network_header -
335 entskb->mac_header));
336
337 return nlalen;
338}
339
340static int nfqnl_put_bridge(struct nf_queue_entry *entry, struct sk_buff *skb)
341{
342 struct sk_buff *entskb = entry->skb;
343
344 if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb))
345 return 0;
346
347 if (skb_vlan_tag_present(entskb)) {
348 struct nlattr *nest;
349
350 nest = nla_nest_start(skb, NFQA_VLAN);
351 if (!nest)
352 goto nla_put_failure;
353
354 if (nla_put_be16(skb, NFQA_VLAN_TCI, htons(entskb->vlan_tci)) ||
355 nla_put_be16(skb, NFQA_VLAN_PROTO, entskb->vlan_proto))
356 goto nla_put_failure;
357
358 nla_nest_end(skb, nest);
359 }
360
361 if (entskb->mac_header < entskb->network_header) {
362 int len = (int)(entskb->network_header - entskb->mac_header);
363
364 if (nla_put(skb, NFQA_L2HDR, len, skb_mac_header(entskb)))
365 goto nla_put_failure;
366 }
367
368 return 0;
369
370nla_put_failure:
371 return -1;
372}
373
374static struct sk_buff *
375nfqnl_build_packet_message(struct net *net, struct nfqnl_instance *queue,
376 struct nf_queue_entry *entry,
377 __be32 **packet_id_ptr)
378{
379 size_t size;
380 size_t data_len = 0, cap_len = 0;
381 unsigned int hlen = 0;
382 struct sk_buff *skb;
383 struct nlattr *nla;
384 struct nfqnl_msg_packet_hdr *pmsg;
385 struct nlmsghdr *nlh;
386 struct sk_buff *entskb = entry->skb;
387 struct net_device *indev;
388 struct net_device *outdev;
389 struct nf_conn *ct = NULL;
390 enum ip_conntrack_info ctinfo = 0;
391 const struct nfnl_ct_hook *nfnl_ct;
392 bool csum_verify;
393 char *secdata = NULL;
394 u32 seclen = 0;
395 ktime_t tstamp;
396
397 size = nlmsg_total_size(sizeof(struct nfgenmsg))
398 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
399 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
400 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
401#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
402 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
403 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
404#endif
405 + nla_total_size(sizeof(u_int32_t)) /* mark */
406 + nla_total_size(sizeof(u_int32_t)) /* priority */
407 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
408 + nla_total_size(sizeof(u_int32_t)) /* skbinfo */
409 + nla_total_size(sizeof(u_int32_t)); /* cap_len */
410
411 tstamp = skb_tstamp_cond(entskb, false);
412 if (tstamp)
413 size += nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
414
415 size += nfqnl_get_bridge_size(entry);
416
417 if (entry->state.hook <= NF_INET_FORWARD ||
418 (entry->state.hook == NF_INET_POST_ROUTING && entskb->sk == NULL))
419 csum_verify = !skb_csum_unnecessary(entskb);
420 else
421 csum_verify = false;
422
423 outdev = entry->state.out;
424
425 switch ((enum nfqnl_config_mode)READ_ONCE(queue->copy_mode)) {
426 case NFQNL_COPY_META:
427 case NFQNL_COPY_NONE:
428 break;
429
430 case NFQNL_COPY_PACKET:
431 if (!(queue->flags & NFQA_CFG_F_GSO) &&
432 entskb->ip_summed == CHECKSUM_PARTIAL &&
433 skb_checksum_help(entskb))
434 return NULL;
435
436 data_len = READ_ONCE(queue->copy_range);
437 if (data_len > entskb->len)
438 data_len = entskb->len;
439
440 hlen = skb_zerocopy_headlen(entskb);
441 hlen = min_t(unsigned int, hlen, data_len);
442 size += sizeof(struct nlattr) + hlen;
443 cap_len = entskb->len;
444 break;
445 }
446
447 nfnl_ct = rcu_dereference(nfnl_ct_hook);
448
449#if IS_ENABLED(CONFIG_NF_CONNTRACK)
450 if (queue->flags & NFQA_CFG_F_CONNTRACK) {
451 if (nfnl_ct != NULL) {
452 ct = nf_ct_get(entskb, &ctinfo);
453 if (ct != NULL)
454 size += nfnl_ct->build_size(ct);
455 }
456 }
457#endif
458
459 if (queue->flags & NFQA_CFG_F_UID_GID) {
460 size += (nla_total_size(sizeof(u_int32_t)) /* uid */
461 + nla_total_size(sizeof(u_int32_t))); /* gid */
462 }
463
464 if ((queue->flags & NFQA_CFG_F_SECCTX) && entskb->sk) {
465 seclen = nfqnl_get_sk_secctx(entskb, &secdata);
466 if (seclen)
467 size += nla_total_size(seclen);
468 }
469
470 skb = alloc_skb(size, GFP_ATOMIC);
471 if (!skb) {
472 skb_tx_error(entskb);
473 goto nlmsg_failure;
474 }
475
476 nlh = nfnl_msg_put(skb, 0, 0,
477 nfnl_msg_type(NFNL_SUBSYS_QUEUE, NFQNL_MSG_PACKET),
478 0, entry->state.pf, NFNETLINK_V0,
479 htons(queue->queue_num));
480 if (!nlh) {
481 skb_tx_error(entskb);
482 kfree_skb(skb);
483 goto nlmsg_failure;
484 }
485
486 nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
487 pmsg = nla_data(nla);
488 pmsg->hw_protocol = entskb->protocol;
489 pmsg->hook = entry->state.hook;
490 *packet_id_ptr = &pmsg->packet_id;
491
492 indev = entry->state.in;
493 if (indev) {
494#if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
495 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
496 goto nla_put_failure;
497#else
498 if (entry->state.pf == PF_BRIDGE) {
499 /* Case 1: indev is physical input device, we need to
500 * look for bridge group (when called from
501 * netfilter_bridge) */
502 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
503 htonl(indev->ifindex)) ||
504 /* this is the bridge group "brX" */
505 /* rcu_read_lock()ed by __nf_queue */
506 nla_put_be32(skb, NFQA_IFINDEX_INDEV,
507 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
508 goto nla_put_failure;
509 } else {
510 int physinif;
511
512 /* Case 2: indev is bridge group, we need to look for
513 * physical device (when called from ipv4) */
514 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
515 htonl(indev->ifindex)))
516 goto nla_put_failure;
517
518 physinif = nf_bridge_get_physinif(entskb);
519 if (physinif &&
520 nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
521 htonl(physinif)))
522 goto nla_put_failure;
523 }
524#endif
525 }
526
527 if (outdev) {
528#if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
529 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
530 goto nla_put_failure;
531#else
532 if (entry->state.pf == PF_BRIDGE) {
533 /* Case 1: outdev is physical output device, we need to
534 * look for bridge group (when called from
535 * netfilter_bridge) */
536 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
537 htonl(outdev->ifindex)) ||
538 /* this is the bridge group "brX" */
539 /* rcu_read_lock()ed by __nf_queue */
540 nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
541 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
542 goto nla_put_failure;
543 } else {
544 int physoutif;
545
546 /* Case 2: outdev is bridge group, we need to look for
547 * physical output device (when called from ipv4) */
548 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
549 htonl(outdev->ifindex)))
550 goto nla_put_failure;
551
552 physoutif = nf_bridge_get_physoutif(entskb);
553 if (physoutif &&
554 nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
555 htonl(physoutif)))
556 goto nla_put_failure;
557 }
558#endif
559 }
560
561 if (entskb->mark &&
562 nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
563 goto nla_put_failure;
564
565 if (entskb->priority &&
566 nla_put_be32(skb, NFQA_PRIORITY, htonl(entskb->priority)))
567 goto nla_put_failure;
568
569 if (indev && entskb->dev &&
570 skb_mac_header_was_set(entskb) &&
571 skb_mac_header_len(entskb) != 0) {
572 struct nfqnl_msg_packet_hw phw;
573 int len;
574
575 memset(&phw, 0, sizeof(phw));
576 len = dev_parse_header(entskb, phw.hw_addr);
577 if (len) {
578 phw.hw_addrlen = htons(len);
579 if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
580 goto nla_put_failure;
581 }
582 }
583
584 if (nfqnl_put_bridge(entry, skb) < 0)
585 goto nla_put_failure;
586
587 if (entry->state.hook <= NF_INET_FORWARD && tstamp) {
588 struct nfqnl_msg_packet_timestamp ts;
589 struct timespec64 kts = ktime_to_timespec64(tstamp);
590
591 ts.sec = cpu_to_be64(kts.tv_sec);
592 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
593
594 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
595 goto nla_put_failure;
596 }
597
598 if ((queue->flags & NFQA_CFG_F_UID_GID) && entskb->sk &&
599 nfqnl_put_sk_uidgid(skb, entskb->sk) < 0)
600 goto nla_put_failure;
601
602 if (seclen && nla_put(skb, NFQA_SECCTX, seclen, secdata))
603 goto nla_put_failure;
604
605 if (ct && nfnl_ct->build(skb, ct, ctinfo, NFQA_CT, NFQA_CT_INFO) < 0)
606 goto nla_put_failure;
607
608 if (cap_len > data_len &&
609 nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
610 goto nla_put_failure;
611
612 if (nfqnl_put_packet_info(skb, entskb, csum_verify))
613 goto nla_put_failure;
614
615 if (data_len) {
616 struct nlattr *nla;
617
618 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
619 goto nla_put_failure;
620
621 nla = skb_put(skb, sizeof(*nla));
622 nla->nla_type = NFQA_PAYLOAD;
623 nla->nla_len = nla_attr_size(data_len);
624
625 if (skb_zerocopy(skb, entskb, data_len, hlen))
626 goto nla_put_failure;
627 }
628
629 nlh->nlmsg_len = skb->len;
630 if (seclen)
631 security_release_secctx(secdata, seclen);
632 return skb;
633
634nla_put_failure:
635 skb_tx_error(entskb);
636 kfree_skb(skb);
637 net_err_ratelimited("nf_queue: error creating packet message\n");
638nlmsg_failure:
639 if (seclen)
640 security_release_secctx(secdata, seclen);
641 return NULL;
642}
643
644static bool nf_ct_drop_unconfirmed(const struct nf_queue_entry *entry)
645{
646#if IS_ENABLED(CONFIG_NF_CONNTRACK)
647 static const unsigned long flags = IPS_CONFIRMED | IPS_DYING;
648 const struct nf_conn *ct = (void *)skb_nfct(entry->skb);
649
650 if (ct && ((ct->status & flags) == IPS_DYING))
651 return true;
652#endif
653 return false;
654}
655
656static int
657__nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
658 struct nf_queue_entry *entry)
659{
660 struct sk_buff *nskb;
661 int err = -ENOBUFS;
662 __be32 *packet_id_ptr;
663 int failopen = 0;
664
665 nskb = nfqnl_build_packet_message(net, queue, entry, &packet_id_ptr);
666 if (nskb == NULL) {
667 err = -ENOMEM;
668 goto err_out;
669 }
670 spin_lock_bh(&queue->lock);
671
672 if (nf_ct_drop_unconfirmed(entry))
673 goto err_out_free_nskb;
674
675 if (queue->queue_total >= queue->queue_maxlen) {
676 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
677 failopen = 1;
678 err = 0;
679 } else {
680 queue->queue_dropped++;
681 net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
682 queue->queue_total);
683 }
684 goto err_out_free_nskb;
685 }
686 entry->id = ++queue->id_sequence;
687 *packet_id_ptr = htonl(entry->id);
688
689 /* nfnetlink_unicast will either free the nskb or add it to a socket */
690 err = nfnetlink_unicast(nskb, net, queue->peer_portid);
691 if (err < 0) {
692 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
693 failopen = 1;
694 err = 0;
695 } else {
696 queue->queue_user_dropped++;
697 }
698 goto err_out_unlock;
699 }
700
701 __enqueue_entry(queue, entry);
702
703 spin_unlock_bh(&queue->lock);
704 return 0;
705
706err_out_free_nskb:
707 kfree_skb(nskb);
708err_out_unlock:
709 spin_unlock_bh(&queue->lock);
710 if (failopen)
711 nfqnl_reinject(entry, NF_ACCEPT);
712err_out:
713 return err;
714}
715
716static struct nf_queue_entry *
717nf_queue_entry_dup(struct nf_queue_entry *e)
718{
719 struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
720
721 if (!entry)
722 return NULL;
723
724 if (nf_queue_entry_get_refs(entry))
725 return entry;
726
727 kfree(entry);
728 return NULL;
729}
730
731#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
732/* When called from bridge netfilter, skb->data must point to MAC header
733 * before calling skb_gso_segment(). Else, original MAC header is lost
734 * and segmented skbs will be sent to wrong destination.
735 */
736static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
737{
738 if (nf_bridge_info_get(skb))
739 __skb_push(skb, skb->network_header - skb->mac_header);
740}
741
742static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
743{
744 if (nf_bridge_info_get(skb))
745 __skb_pull(skb, skb->network_header - skb->mac_header);
746}
747#else
748#define nf_bridge_adjust_skb_data(s) do {} while (0)
749#define nf_bridge_adjust_segmented_data(s) do {} while (0)
750#endif
751
752static int
753__nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
754 struct sk_buff *skb, struct nf_queue_entry *entry)
755{
756 int ret = -ENOMEM;
757 struct nf_queue_entry *entry_seg;
758
759 nf_bridge_adjust_segmented_data(skb);
760
761 if (skb->next == NULL) { /* last packet, no need to copy entry */
762 struct sk_buff *gso_skb = entry->skb;
763 entry->skb = skb;
764 ret = __nfqnl_enqueue_packet(net, queue, entry);
765 if (ret)
766 entry->skb = gso_skb;
767 return ret;
768 }
769
770 skb_mark_not_on_list(skb);
771
772 entry_seg = nf_queue_entry_dup(entry);
773 if (entry_seg) {
774 entry_seg->skb = skb;
775 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
776 if (ret)
777 nf_queue_entry_free(entry_seg);
778 }
779 return ret;
780}
781
782static int
783nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
784{
785 unsigned int queued;
786 struct nfqnl_instance *queue;
787 struct sk_buff *skb, *segs, *nskb;
788 int err = -ENOBUFS;
789 struct net *net = entry->state.net;
790 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
791
792 /* rcu_read_lock()ed by nf_hook_thresh */
793 queue = instance_lookup(q, queuenum);
794 if (!queue)
795 return -ESRCH;
796
797 if (queue->copy_mode == NFQNL_COPY_NONE)
798 return -EINVAL;
799
800 skb = entry->skb;
801
802 switch (entry->state.pf) {
803 case NFPROTO_IPV4:
804 skb->protocol = htons(ETH_P_IP);
805 break;
806 case NFPROTO_IPV6:
807 skb->protocol = htons(ETH_P_IPV6);
808 break;
809 }
810
811 if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(skb))
812 return __nfqnl_enqueue_packet(net, queue, entry);
813
814 nf_bridge_adjust_skb_data(skb);
815 segs = skb_gso_segment(skb, 0);
816 /* Does not use PTR_ERR to limit the number of error codes that can be
817 * returned by nf_queue. For instance, callers rely on -ESRCH to
818 * mean 'ignore this hook'.
819 */
820 if (IS_ERR_OR_NULL(segs))
821 goto out_err;
822 queued = 0;
823 err = 0;
824 skb_list_walk_safe(segs, segs, nskb) {
825 if (err == 0)
826 err = __nfqnl_enqueue_packet_gso(net, queue,
827 segs, entry);
828 if (err == 0)
829 queued++;
830 else
831 kfree_skb(segs);
832 }
833
834 if (queued) {
835 if (err) /* some segments are already queued */
836 nf_queue_entry_free(entry);
837 kfree_skb(skb);
838 return 0;
839 }
840 out_err:
841 nf_bridge_adjust_segmented_data(skb);
842 return err;
843}
844
845static int
846nfqnl_mangle(void *data, unsigned int data_len, struct nf_queue_entry *e, int diff)
847{
848 struct sk_buff *nskb;
849
850 if (diff < 0) {
851 unsigned int min_len = skb_transport_offset(e->skb);
852
853 if (data_len < min_len)
854 return -EINVAL;
855
856 if (pskb_trim(e->skb, data_len))
857 return -ENOMEM;
858 } else if (diff > 0) {
859 if (data_len > 0xFFFF)
860 return -EINVAL;
861 if (diff > skb_tailroom(e->skb)) {
862 nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
863 diff, GFP_ATOMIC);
864 if (!nskb)
865 return -ENOMEM;
866 kfree_skb(e->skb);
867 e->skb = nskb;
868 }
869 skb_put(e->skb, diff);
870 }
871 if (skb_ensure_writable(e->skb, data_len))
872 return -ENOMEM;
873 skb_copy_to_linear_data(e->skb, data, data_len);
874 e->skb->ip_summed = CHECKSUM_NONE;
875 return 0;
876}
877
878static int
879nfqnl_set_mode(struct nfqnl_instance *queue,
880 unsigned char mode, unsigned int range)
881{
882 int status = 0;
883
884 spin_lock_bh(&queue->lock);
885 switch (mode) {
886 case NFQNL_COPY_NONE:
887 case NFQNL_COPY_META:
888 queue->copy_mode = mode;
889 queue->copy_range = 0;
890 break;
891
892 case NFQNL_COPY_PACKET:
893 queue->copy_mode = mode;
894 if (range == 0 || range > NFQNL_MAX_COPY_RANGE)
895 queue->copy_range = NFQNL_MAX_COPY_RANGE;
896 else
897 queue->copy_range = range;
898 break;
899
900 default:
901 status = -EINVAL;
902
903 }
904 spin_unlock_bh(&queue->lock);
905
906 return status;
907}
908
909static int
910dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
911{
912#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
913 int physinif, physoutif;
914
915 physinif = nf_bridge_get_physinif(entry->skb);
916 physoutif = nf_bridge_get_physoutif(entry->skb);
917
918 if (physinif == ifindex || physoutif == ifindex)
919 return 1;
920#endif
921 if (entry->state.in)
922 if (entry->state.in->ifindex == ifindex)
923 return 1;
924 if (entry->state.out)
925 if (entry->state.out->ifindex == ifindex)
926 return 1;
927
928 return 0;
929}
930
931/* drop all packets with either indev or outdev == ifindex from all queue
932 * instances */
933static void
934nfqnl_dev_drop(struct net *net, int ifindex)
935{
936 int i;
937 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
938
939 rcu_read_lock();
940
941 for (i = 0; i < INSTANCE_BUCKETS; i++) {
942 struct nfqnl_instance *inst;
943 struct hlist_head *head = &q->instance_table[i];
944
945 hlist_for_each_entry_rcu(inst, head, hlist)
946 nfqnl_flush(inst, dev_cmp, ifindex);
947 }
948
949 rcu_read_unlock();
950}
951
952static int
953nfqnl_rcv_dev_event(struct notifier_block *this,
954 unsigned long event, void *ptr)
955{
956 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
957
958 /* Drop any packets associated with the downed device */
959 if (event == NETDEV_DOWN)
960 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
961 return NOTIFY_DONE;
962}
963
964static struct notifier_block nfqnl_dev_notifier = {
965 .notifier_call = nfqnl_rcv_dev_event,
966};
967
968static void nfqnl_nf_hook_drop(struct net *net)
969{
970 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
971 int i;
972
973 /* This function is also called on net namespace error unwind,
974 * when pernet_ops->init() failed and ->exit() functions of the
975 * previous pernet_ops gets called.
976 *
977 * This may result in a call to nfqnl_nf_hook_drop() before
978 * struct nfnl_queue_net was allocated.
979 */
980 if (!q)
981 return;
982
983 for (i = 0; i < INSTANCE_BUCKETS; i++) {
984 struct nfqnl_instance *inst;
985 struct hlist_head *head = &q->instance_table[i];
986
987 hlist_for_each_entry_rcu(inst, head, hlist)
988 nfqnl_flush(inst, NULL, 0);
989 }
990}
991
992static int
993nfqnl_rcv_nl_event(struct notifier_block *this,
994 unsigned long event, void *ptr)
995{
996 struct netlink_notify *n = ptr;
997 struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
998
999 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
1000 int i;
1001
1002 /* destroy all instances for this portid */
1003 spin_lock(&q->instances_lock);
1004 for (i = 0; i < INSTANCE_BUCKETS; i++) {
1005 struct hlist_node *t2;
1006 struct nfqnl_instance *inst;
1007 struct hlist_head *head = &q->instance_table[i];
1008
1009 hlist_for_each_entry_safe(inst, t2, head, hlist) {
1010 if (n->portid == inst->peer_portid)
1011 __instance_destroy(inst);
1012 }
1013 }
1014 spin_unlock(&q->instances_lock);
1015 }
1016 return NOTIFY_DONE;
1017}
1018
1019static struct notifier_block nfqnl_rtnl_notifier = {
1020 .notifier_call = nfqnl_rcv_nl_event,
1021};
1022
1023static const struct nla_policy nfqa_vlan_policy[NFQA_VLAN_MAX + 1] = {
1024 [NFQA_VLAN_TCI] = { .type = NLA_U16},
1025 [NFQA_VLAN_PROTO] = { .type = NLA_U16},
1026};
1027
1028static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
1029 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
1030 [NFQA_MARK] = { .type = NLA_U32 },
1031 [NFQA_PAYLOAD] = { .type = NLA_UNSPEC },
1032 [NFQA_CT] = { .type = NLA_UNSPEC },
1033 [NFQA_EXP] = { .type = NLA_UNSPEC },
1034 [NFQA_VLAN] = { .type = NLA_NESTED },
1035 [NFQA_PRIORITY] = { .type = NLA_U32 },
1036};
1037
1038static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
1039 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
1040 [NFQA_MARK] = { .type = NLA_U32 },
1041 [NFQA_PRIORITY] = { .type = NLA_U32 },
1042};
1043
1044static struct nfqnl_instance *
1045verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, u32 nlportid)
1046{
1047 struct nfqnl_instance *queue;
1048
1049 queue = instance_lookup(q, queue_num);
1050 if (!queue)
1051 return ERR_PTR(-ENODEV);
1052
1053 if (queue->peer_portid != nlportid)
1054 return ERR_PTR(-EPERM);
1055
1056 return queue;
1057}
1058
1059static struct nfqnl_msg_verdict_hdr*
1060verdicthdr_get(const struct nlattr * const nfqa[])
1061{
1062 struct nfqnl_msg_verdict_hdr *vhdr;
1063 unsigned int verdict;
1064
1065 if (!nfqa[NFQA_VERDICT_HDR])
1066 return NULL;
1067
1068 vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
1069 verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
1070 if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
1071 return NULL;
1072 return vhdr;
1073}
1074
1075static int nfq_id_after(unsigned int id, unsigned int max)
1076{
1077 return (int)(id - max) > 0;
1078}
1079
1080static int nfqnl_recv_verdict_batch(struct sk_buff *skb,
1081 const struct nfnl_info *info,
1082 const struct nlattr * const nfqa[])
1083{
1084 struct nfnl_queue_net *q = nfnl_queue_pernet(info->net);
1085 u16 queue_num = ntohs(info->nfmsg->res_id);
1086 struct nf_queue_entry *entry, *tmp;
1087 struct nfqnl_msg_verdict_hdr *vhdr;
1088 struct nfqnl_instance *queue;
1089 unsigned int verdict, maxid;
1090 LIST_HEAD(batch_list);
1091
1092 queue = verdict_instance_lookup(q, queue_num,
1093 NETLINK_CB(skb).portid);
1094 if (IS_ERR(queue))
1095 return PTR_ERR(queue);
1096
1097 vhdr = verdicthdr_get(nfqa);
1098 if (!vhdr)
1099 return -EINVAL;
1100
1101 verdict = ntohl(vhdr->verdict);
1102 maxid = ntohl(vhdr->id);
1103
1104 spin_lock_bh(&queue->lock);
1105
1106 list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
1107 if (nfq_id_after(entry->id, maxid))
1108 break;
1109 __dequeue_entry(queue, entry);
1110 list_add_tail(&entry->list, &batch_list);
1111 }
1112
1113 spin_unlock_bh(&queue->lock);
1114
1115 if (list_empty(&batch_list))
1116 return -ENOENT;
1117
1118 list_for_each_entry_safe(entry, tmp, &batch_list, list) {
1119 if (nfqa[NFQA_MARK])
1120 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1121
1122 if (nfqa[NFQA_PRIORITY])
1123 entry->skb->priority = ntohl(nla_get_be32(nfqa[NFQA_PRIORITY]));
1124
1125 nfqnl_reinject(entry, verdict);
1126 }
1127 return 0;
1128}
1129
1130static struct nf_conn *nfqnl_ct_parse(const struct nfnl_ct_hook *nfnl_ct,
1131 const struct nlmsghdr *nlh,
1132 const struct nlattr * const nfqa[],
1133 struct nf_queue_entry *entry,
1134 enum ip_conntrack_info *ctinfo)
1135{
1136#if IS_ENABLED(CONFIG_NF_CONNTRACK)
1137 struct nf_conn *ct;
1138
1139 ct = nf_ct_get(entry->skb, ctinfo);
1140 if (ct == NULL)
1141 return NULL;
1142
1143 if (nfnl_ct->parse(nfqa[NFQA_CT], ct) < 0)
1144 return NULL;
1145
1146 if (nfqa[NFQA_EXP])
1147 nfnl_ct->attach_expect(nfqa[NFQA_EXP], ct,
1148 NETLINK_CB(entry->skb).portid,
1149 nlmsg_report(nlh));
1150 return ct;
1151#else
1152 return NULL;
1153#endif
1154}
1155
1156static int nfqa_parse_bridge(struct nf_queue_entry *entry,
1157 const struct nlattr * const nfqa[])
1158{
1159 if (nfqa[NFQA_VLAN]) {
1160 struct nlattr *tb[NFQA_VLAN_MAX + 1];
1161 int err;
1162
1163 err = nla_parse_nested_deprecated(tb, NFQA_VLAN_MAX,
1164 nfqa[NFQA_VLAN],
1165 nfqa_vlan_policy, NULL);
1166 if (err < 0)
1167 return err;
1168
1169 if (!tb[NFQA_VLAN_TCI] || !tb[NFQA_VLAN_PROTO])
1170 return -EINVAL;
1171
1172 __vlan_hwaccel_put_tag(entry->skb,
1173 nla_get_be16(tb[NFQA_VLAN_PROTO]),
1174 ntohs(nla_get_be16(tb[NFQA_VLAN_TCI])));
1175 }
1176
1177 if (nfqa[NFQA_L2HDR]) {
1178 int mac_header_len = entry->skb->network_header -
1179 entry->skb->mac_header;
1180
1181 if (mac_header_len != nla_len(nfqa[NFQA_L2HDR]))
1182 return -EINVAL;
1183 else if (mac_header_len > 0)
1184 memcpy(skb_mac_header(entry->skb),
1185 nla_data(nfqa[NFQA_L2HDR]),
1186 mac_header_len);
1187 }
1188
1189 return 0;
1190}
1191
1192static int nfqnl_recv_verdict(struct sk_buff *skb, const struct nfnl_info *info,
1193 const struct nlattr * const nfqa[])
1194{
1195 struct nfnl_queue_net *q = nfnl_queue_pernet(info->net);
1196 u_int16_t queue_num = ntohs(info->nfmsg->res_id);
1197 const struct nfnl_ct_hook *nfnl_ct;
1198 struct nfqnl_msg_verdict_hdr *vhdr;
1199 enum ip_conntrack_info ctinfo;
1200 struct nfqnl_instance *queue;
1201 struct nf_queue_entry *entry;
1202 struct nf_conn *ct = NULL;
1203 unsigned int verdict;
1204 int err;
1205
1206 queue = verdict_instance_lookup(q, queue_num,
1207 NETLINK_CB(skb).portid);
1208 if (IS_ERR(queue))
1209 return PTR_ERR(queue);
1210
1211 vhdr = verdicthdr_get(nfqa);
1212 if (!vhdr)
1213 return -EINVAL;
1214
1215 verdict = ntohl(vhdr->verdict);
1216
1217 entry = find_dequeue_entry(queue, ntohl(vhdr->id));
1218 if (entry == NULL)
1219 return -ENOENT;
1220
1221 /* rcu lock already held from nfnl->call_rcu. */
1222 nfnl_ct = rcu_dereference(nfnl_ct_hook);
1223
1224 if (nfqa[NFQA_CT]) {
1225 if (nfnl_ct != NULL)
1226 ct = nfqnl_ct_parse(nfnl_ct, info->nlh, nfqa, entry,
1227 &ctinfo);
1228 }
1229
1230 if (entry->state.pf == PF_BRIDGE) {
1231 err = nfqa_parse_bridge(entry, nfqa);
1232 if (err < 0)
1233 return err;
1234 }
1235
1236 if (nfqa[NFQA_PAYLOAD]) {
1237 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
1238 int diff = payload_len - entry->skb->len;
1239
1240 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
1241 payload_len, entry, diff) < 0)
1242 verdict = NF_DROP;
1243
1244 if (ct && diff)
1245 nfnl_ct->seq_adjust(entry->skb, ct, ctinfo, diff);
1246 }
1247
1248 if (nfqa[NFQA_MARK])
1249 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1250
1251 if (nfqa[NFQA_PRIORITY])
1252 entry->skb->priority = ntohl(nla_get_be32(nfqa[NFQA_PRIORITY]));
1253
1254 nfqnl_reinject(entry, verdict);
1255 return 0;
1256}
1257
1258static int nfqnl_recv_unsupp(struct sk_buff *skb, const struct nfnl_info *info,
1259 const struct nlattr * const cda[])
1260{
1261 return -ENOTSUPP;
1262}
1263
1264static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1265 [NFQA_CFG_CMD] = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1266 [NFQA_CFG_PARAMS] = { .len = sizeof(struct nfqnl_msg_config_params) },
1267 [NFQA_CFG_QUEUE_MAXLEN] = { .type = NLA_U32 },
1268 [NFQA_CFG_MASK] = { .type = NLA_U32 },
1269 [NFQA_CFG_FLAGS] = { .type = NLA_U32 },
1270};
1271
1272static const struct nf_queue_handler nfqh = {
1273 .outfn = nfqnl_enqueue_packet,
1274 .nf_hook_drop = nfqnl_nf_hook_drop,
1275};
1276
1277static int nfqnl_recv_config(struct sk_buff *skb, const struct nfnl_info *info,
1278 const struct nlattr * const nfqa[])
1279{
1280 struct nfnl_queue_net *q = nfnl_queue_pernet(info->net);
1281 u_int16_t queue_num = ntohs(info->nfmsg->res_id);
1282 struct nfqnl_msg_config_cmd *cmd = NULL;
1283 struct nfqnl_instance *queue;
1284 __u32 flags = 0, mask = 0;
1285 int ret = 0;
1286
1287 if (nfqa[NFQA_CFG_CMD]) {
1288 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1289
1290 /* Obsolete commands without queue context */
1291 switch (cmd->command) {
1292 case NFQNL_CFG_CMD_PF_BIND: return 0;
1293 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1294 }
1295 }
1296
1297 /* Check if we support these flags in first place, dependencies should
1298 * be there too not to break atomicity.
1299 */
1300 if (nfqa[NFQA_CFG_FLAGS]) {
1301 if (!nfqa[NFQA_CFG_MASK]) {
1302 /* A mask is needed to specify which flags are being
1303 * changed.
1304 */
1305 return -EINVAL;
1306 }
1307
1308 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1309 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1310
1311 if (flags >= NFQA_CFG_F_MAX)
1312 return -EOPNOTSUPP;
1313
1314#if !IS_ENABLED(CONFIG_NETWORK_SECMARK)
1315 if (flags & mask & NFQA_CFG_F_SECCTX)
1316 return -EOPNOTSUPP;
1317#endif
1318 if ((flags & mask & NFQA_CFG_F_CONNTRACK) &&
1319 !rcu_access_pointer(nfnl_ct_hook)) {
1320#ifdef CONFIG_MODULES
1321 nfnl_unlock(NFNL_SUBSYS_QUEUE);
1322 request_module("ip_conntrack_netlink");
1323 nfnl_lock(NFNL_SUBSYS_QUEUE);
1324 if (rcu_access_pointer(nfnl_ct_hook))
1325 return -EAGAIN;
1326#endif
1327 return -EOPNOTSUPP;
1328 }
1329 }
1330
1331 rcu_read_lock();
1332 queue = instance_lookup(q, queue_num);
1333 if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1334 ret = -EPERM;
1335 goto err_out_unlock;
1336 }
1337
1338 if (cmd != NULL) {
1339 switch (cmd->command) {
1340 case NFQNL_CFG_CMD_BIND:
1341 if (queue) {
1342 ret = -EBUSY;
1343 goto err_out_unlock;
1344 }
1345 queue = instance_create(q, queue_num,
1346 NETLINK_CB(skb).portid);
1347 if (IS_ERR(queue)) {
1348 ret = PTR_ERR(queue);
1349 goto err_out_unlock;
1350 }
1351 break;
1352 case NFQNL_CFG_CMD_UNBIND:
1353 if (!queue) {
1354 ret = -ENODEV;
1355 goto err_out_unlock;
1356 }
1357 instance_destroy(q, queue);
1358 goto err_out_unlock;
1359 case NFQNL_CFG_CMD_PF_BIND:
1360 case NFQNL_CFG_CMD_PF_UNBIND:
1361 break;
1362 default:
1363 ret = -ENOTSUPP;
1364 goto err_out_unlock;
1365 }
1366 }
1367
1368 if (!queue) {
1369 ret = -ENODEV;
1370 goto err_out_unlock;
1371 }
1372
1373 if (nfqa[NFQA_CFG_PARAMS]) {
1374 struct nfqnl_msg_config_params *params =
1375 nla_data(nfqa[NFQA_CFG_PARAMS]);
1376
1377 nfqnl_set_mode(queue, params->copy_mode,
1378 ntohl(params->copy_range));
1379 }
1380
1381 if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1382 __be32 *queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1383
1384 spin_lock_bh(&queue->lock);
1385 queue->queue_maxlen = ntohl(*queue_maxlen);
1386 spin_unlock_bh(&queue->lock);
1387 }
1388
1389 if (nfqa[NFQA_CFG_FLAGS]) {
1390 spin_lock_bh(&queue->lock);
1391 queue->flags &= ~mask;
1392 queue->flags |= flags & mask;
1393 spin_unlock_bh(&queue->lock);
1394 }
1395
1396err_out_unlock:
1397 rcu_read_unlock();
1398 return ret;
1399}
1400
1401static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1402 [NFQNL_MSG_PACKET] = {
1403 .call = nfqnl_recv_unsupp,
1404 .type = NFNL_CB_RCU,
1405 .attr_count = NFQA_MAX,
1406 },
1407 [NFQNL_MSG_VERDICT] = {
1408 .call = nfqnl_recv_verdict,
1409 .type = NFNL_CB_RCU,
1410 .attr_count = NFQA_MAX,
1411 .policy = nfqa_verdict_policy
1412 },
1413 [NFQNL_MSG_CONFIG] = {
1414 .call = nfqnl_recv_config,
1415 .type = NFNL_CB_MUTEX,
1416 .attr_count = NFQA_CFG_MAX,
1417 .policy = nfqa_cfg_policy
1418 },
1419 [NFQNL_MSG_VERDICT_BATCH] = {
1420 .call = nfqnl_recv_verdict_batch,
1421 .type = NFNL_CB_RCU,
1422 .attr_count = NFQA_MAX,
1423 .policy = nfqa_verdict_batch_policy
1424 },
1425};
1426
1427static const struct nfnetlink_subsystem nfqnl_subsys = {
1428 .name = "nf_queue",
1429 .subsys_id = NFNL_SUBSYS_QUEUE,
1430 .cb_count = NFQNL_MSG_MAX,
1431 .cb = nfqnl_cb,
1432};
1433
1434#ifdef CONFIG_PROC_FS
1435struct iter_state {
1436 struct seq_net_private p;
1437 unsigned int bucket;
1438};
1439
1440static struct hlist_node *get_first(struct seq_file *seq)
1441{
1442 struct iter_state *st = seq->private;
1443 struct net *net;
1444 struct nfnl_queue_net *q;
1445
1446 if (!st)
1447 return NULL;
1448
1449 net = seq_file_net(seq);
1450 q = nfnl_queue_pernet(net);
1451 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1452 if (!hlist_empty(&q->instance_table[st->bucket]))
1453 return q->instance_table[st->bucket].first;
1454 }
1455 return NULL;
1456}
1457
1458static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1459{
1460 struct iter_state *st = seq->private;
1461 struct net *net = seq_file_net(seq);
1462
1463 h = h->next;
1464 while (!h) {
1465 struct nfnl_queue_net *q;
1466
1467 if (++st->bucket >= INSTANCE_BUCKETS)
1468 return NULL;
1469
1470 q = nfnl_queue_pernet(net);
1471 h = q->instance_table[st->bucket].first;
1472 }
1473 return h;
1474}
1475
1476static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1477{
1478 struct hlist_node *head;
1479 head = get_first(seq);
1480
1481 if (head)
1482 while (pos && (head = get_next(seq, head)))
1483 pos--;
1484 return pos ? NULL : head;
1485}
1486
1487static void *seq_start(struct seq_file *s, loff_t *pos)
1488 __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1489{
1490 spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1491 return get_idx(s, *pos);
1492}
1493
1494static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1495{
1496 (*pos)++;
1497 return get_next(s, v);
1498}
1499
1500static void seq_stop(struct seq_file *s, void *v)
1501 __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1502{
1503 spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1504}
1505
1506static int seq_show(struct seq_file *s, void *v)
1507{
1508 const struct nfqnl_instance *inst = v;
1509
1510 seq_printf(s, "%5u %6u %5u %1u %5u %5u %5u %8u %2d\n",
1511 inst->queue_num,
1512 inst->peer_portid, inst->queue_total,
1513 inst->copy_mode, inst->copy_range,
1514 inst->queue_dropped, inst->queue_user_dropped,
1515 inst->id_sequence, 1);
1516 return 0;
1517}
1518
1519static const struct seq_operations nfqnl_seq_ops = {
1520 .start = seq_start,
1521 .next = seq_next,
1522 .stop = seq_stop,
1523 .show = seq_show,
1524};
1525#endif /* PROC_FS */
1526
1527static int __net_init nfnl_queue_net_init(struct net *net)
1528{
1529 unsigned int i;
1530 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1531
1532 for (i = 0; i < INSTANCE_BUCKETS; i++)
1533 INIT_HLIST_HEAD(&q->instance_table[i]);
1534
1535 spin_lock_init(&q->instances_lock);
1536
1537#ifdef CONFIG_PROC_FS
1538 if (!proc_create_net("nfnetlink_queue", 0440, net->nf.proc_netfilter,
1539 &nfqnl_seq_ops, sizeof(struct iter_state)))
1540 return -ENOMEM;
1541#endif
1542 return 0;
1543}
1544
1545static void __net_exit nfnl_queue_net_exit(struct net *net)
1546{
1547 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1548 unsigned int i;
1549
1550#ifdef CONFIG_PROC_FS
1551 remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1552#endif
1553 for (i = 0; i < INSTANCE_BUCKETS; i++)
1554 WARN_ON_ONCE(!hlist_empty(&q->instance_table[i]));
1555}
1556
1557static struct pernet_operations nfnl_queue_net_ops = {
1558 .init = nfnl_queue_net_init,
1559 .exit = nfnl_queue_net_exit,
1560 .id = &nfnl_queue_net_id,
1561 .size = sizeof(struct nfnl_queue_net),
1562};
1563
1564static int __init nfnetlink_queue_init(void)
1565{
1566 int status;
1567
1568 status = register_pernet_subsys(&nfnl_queue_net_ops);
1569 if (status < 0) {
1570 pr_err("failed to register pernet ops\n");
1571 goto out;
1572 }
1573
1574 netlink_register_notifier(&nfqnl_rtnl_notifier);
1575 status = nfnetlink_subsys_register(&nfqnl_subsys);
1576 if (status < 0) {
1577 pr_err("failed to create netlink socket\n");
1578 goto cleanup_netlink_notifier;
1579 }
1580
1581 status = register_netdevice_notifier(&nfqnl_dev_notifier);
1582 if (status < 0) {
1583 pr_err("failed to register netdevice notifier\n");
1584 goto cleanup_netlink_subsys;
1585 }
1586
1587 nf_register_queue_handler(&nfqh);
1588
1589 return status;
1590
1591cleanup_netlink_subsys:
1592 nfnetlink_subsys_unregister(&nfqnl_subsys);
1593cleanup_netlink_notifier:
1594 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1595 unregister_pernet_subsys(&nfnl_queue_net_ops);
1596out:
1597 return status;
1598}
1599
1600static void __exit nfnetlink_queue_fini(void)
1601{
1602 nf_unregister_queue_handler();
1603 unregister_netdevice_notifier(&nfqnl_dev_notifier);
1604 nfnetlink_subsys_unregister(&nfqnl_subsys);
1605 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1606 unregister_pernet_subsys(&nfnl_queue_net_ops);
1607
1608 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1609}
1610
1611MODULE_DESCRIPTION("netfilter packet queue handler");
1612MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1613MODULE_LICENSE("GPL");
1614MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1615
1616module_init(nfnetlink_queue_init);
1617module_exit(nfnetlink_queue_fini);