Loading...
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 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;
391 struct nfnl_ct_hook *nfnl_ct;
392 bool csum_verify;
393 char *secdata = NULL;
394 u32 seclen = 0;
395
396 size = nlmsg_total_size(sizeof(struct nfgenmsg))
397 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
398 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
399 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
400#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
401 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
402 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
403#endif
404 + nla_total_size(sizeof(u_int32_t)) /* mark */
405 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
406 + nla_total_size(sizeof(u_int32_t)) /* skbinfo */
407 + nla_total_size(sizeof(u_int32_t)); /* cap_len */
408
409 if (entskb->tstamp)
410 size += nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
411
412 size += nfqnl_get_bridge_size(entry);
413
414 if (entry->state.hook <= NF_INET_FORWARD ||
415 (entry->state.hook == NF_INET_POST_ROUTING && entskb->sk == NULL))
416 csum_verify = !skb_csum_unnecessary(entskb);
417 else
418 csum_verify = false;
419
420 outdev = entry->state.out;
421
422 switch ((enum nfqnl_config_mode)READ_ONCE(queue->copy_mode)) {
423 case NFQNL_COPY_META:
424 case NFQNL_COPY_NONE:
425 break;
426
427 case NFQNL_COPY_PACKET:
428 if (!(queue->flags & NFQA_CFG_F_GSO) &&
429 entskb->ip_summed == CHECKSUM_PARTIAL &&
430 skb_checksum_help(entskb))
431 return NULL;
432
433 data_len = READ_ONCE(queue->copy_range);
434 if (data_len > entskb->len)
435 data_len = entskb->len;
436
437 hlen = skb_zerocopy_headlen(entskb);
438 hlen = min_t(unsigned int, hlen, data_len);
439 size += sizeof(struct nlattr) + hlen;
440 cap_len = entskb->len;
441 break;
442 }
443
444 nfnl_ct = rcu_dereference(nfnl_ct_hook);
445
446#if IS_ENABLED(CONFIG_NF_CONNTRACK)
447 if (queue->flags & NFQA_CFG_F_CONNTRACK) {
448 if (nfnl_ct != NULL) {
449 ct = nf_ct_get(entskb, &ctinfo);
450 if (ct != NULL)
451 size += nfnl_ct->build_size(ct);
452 }
453 }
454#endif
455
456 if (queue->flags & NFQA_CFG_F_UID_GID) {
457 size += (nla_total_size(sizeof(u_int32_t)) /* uid */
458 + nla_total_size(sizeof(u_int32_t))); /* gid */
459 }
460
461 if ((queue->flags & NFQA_CFG_F_SECCTX) && entskb->sk) {
462 seclen = nfqnl_get_sk_secctx(entskb, &secdata);
463 if (seclen)
464 size += nla_total_size(seclen);
465 }
466
467 skb = alloc_skb(size, GFP_ATOMIC);
468 if (!skb) {
469 skb_tx_error(entskb);
470 goto nlmsg_failure;
471 }
472
473 nlh = nfnl_msg_put(skb, 0, 0,
474 nfnl_msg_type(NFNL_SUBSYS_QUEUE, NFQNL_MSG_PACKET),
475 0, entry->state.pf, NFNETLINK_V0,
476 htons(queue->queue_num));
477 if (!nlh) {
478 skb_tx_error(entskb);
479 kfree_skb(skb);
480 goto nlmsg_failure;
481 }
482
483 nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
484 pmsg = nla_data(nla);
485 pmsg->hw_protocol = entskb->protocol;
486 pmsg->hook = entry->state.hook;
487 *packet_id_ptr = &pmsg->packet_id;
488
489 indev = entry->state.in;
490 if (indev) {
491#if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
492 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
493 goto nla_put_failure;
494#else
495 if (entry->state.pf == PF_BRIDGE) {
496 /* Case 1: indev is physical input device, we need to
497 * look for bridge group (when called from
498 * netfilter_bridge) */
499 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
500 htonl(indev->ifindex)) ||
501 /* this is the bridge group "brX" */
502 /* rcu_read_lock()ed by __nf_queue */
503 nla_put_be32(skb, NFQA_IFINDEX_INDEV,
504 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
505 goto nla_put_failure;
506 } else {
507 int physinif;
508
509 /* Case 2: indev is bridge group, we need to look for
510 * physical device (when called from ipv4) */
511 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
512 htonl(indev->ifindex)))
513 goto nla_put_failure;
514
515 physinif = nf_bridge_get_physinif(entskb);
516 if (physinif &&
517 nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
518 htonl(physinif)))
519 goto nla_put_failure;
520 }
521#endif
522 }
523
524 if (outdev) {
525#if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
526 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
527 goto nla_put_failure;
528#else
529 if (entry->state.pf == PF_BRIDGE) {
530 /* Case 1: outdev is physical output device, we need to
531 * look for bridge group (when called from
532 * netfilter_bridge) */
533 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
534 htonl(outdev->ifindex)) ||
535 /* this is the bridge group "brX" */
536 /* rcu_read_lock()ed by __nf_queue */
537 nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
538 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
539 goto nla_put_failure;
540 } else {
541 int physoutif;
542
543 /* Case 2: outdev is bridge group, we need to look for
544 * physical output device (when called from ipv4) */
545 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
546 htonl(outdev->ifindex)))
547 goto nla_put_failure;
548
549 physoutif = nf_bridge_get_physoutif(entskb);
550 if (physoutif &&
551 nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
552 htonl(physoutif)))
553 goto nla_put_failure;
554 }
555#endif
556 }
557
558 if (entskb->mark &&
559 nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
560 goto nla_put_failure;
561
562 if (indev && entskb->dev &&
563 entskb->mac_header != entskb->network_header) {
564 struct nfqnl_msg_packet_hw phw;
565 int len;
566
567 memset(&phw, 0, sizeof(phw));
568 len = dev_parse_header(entskb, phw.hw_addr);
569 if (len) {
570 phw.hw_addrlen = htons(len);
571 if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
572 goto nla_put_failure;
573 }
574 }
575
576 if (nfqnl_put_bridge(entry, skb) < 0)
577 goto nla_put_failure;
578
579 if (entry->state.hook <= NF_INET_FORWARD && entskb->tstamp) {
580 struct nfqnl_msg_packet_timestamp ts;
581 struct timespec64 kts = ktime_to_timespec64(entskb->tstamp);
582
583 ts.sec = cpu_to_be64(kts.tv_sec);
584 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
585
586 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
587 goto nla_put_failure;
588 }
589
590 if ((queue->flags & NFQA_CFG_F_UID_GID) && entskb->sk &&
591 nfqnl_put_sk_uidgid(skb, entskb->sk) < 0)
592 goto nla_put_failure;
593
594 if (seclen && nla_put(skb, NFQA_SECCTX, seclen, secdata))
595 goto nla_put_failure;
596
597 if (ct && nfnl_ct->build(skb, ct, ctinfo, NFQA_CT, NFQA_CT_INFO) < 0)
598 goto nla_put_failure;
599
600 if (cap_len > data_len &&
601 nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
602 goto nla_put_failure;
603
604 if (nfqnl_put_packet_info(skb, entskb, csum_verify))
605 goto nla_put_failure;
606
607 if (data_len) {
608 struct nlattr *nla;
609
610 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
611 goto nla_put_failure;
612
613 nla = skb_put(skb, sizeof(*nla));
614 nla->nla_type = NFQA_PAYLOAD;
615 nla->nla_len = nla_attr_size(data_len);
616
617 if (skb_zerocopy(skb, entskb, data_len, hlen))
618 goto nla_put_failure;
619 }
620
621 nlh->nlmsg_len = skb->len;
622 if (seclen)
623 security_release_secctx(secdata, seclen);
624 return skb;
625
626nla_put_failure:
627 skb_tx_error(entskb);
628 kfree_skb(skb);
629 net_err_ratelimited("nf_queue: error creating packet message\n");
630nlmsg_failure:
631 if (seclen)
632 security_release_secctx(secdata, seclen);
633 return NULL;
634}
635
636static bool nf_ct_drop_unconfirmed(const struct nf_queue_entry *entry)
637{
638#if IS_ENABLED(CONFIG_NF_CONNTRACK)
639 static const unsigned long flags = IPS_CONFIRMED | IPS_DYING;
640 const struct nf_conn *ct = (void *)skb_nfct(entry->skb);
641
642 if (ct && ((ct->status & flags) == IPS_DYING))
643 return true;
644#endif
645 return false;
646}
647
648static int
649__nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
650 struct nf_queue_entry *entry)
651{
652 struct sk_buff *nskb;
653 int err = -ENOBUFS;
654 __be32 *packet_id_ptr;
655 int failopen = 0;
656
657 nskb = nfqnl_build_packet_message(net, queue, entry, &packet_id_ptr);
658 if (nskb == NULL) {
659 err = -ENOMEM;
660 goto err_out;
661 }
662 spin_lock_bh(&queue->lock);
663
664 if (nf_ct_drop_unconfirmed(entry))
665 goto err_out_free_nskb;
666
667 if (queue->queue_total >= queue->queue_maxlen) {
668 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
669 failopen = 1;
670 err = 0;
671 } else {
672 queue->queue_dropped++;
673 net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
674 queue->queue_total);
675 }
676 goto err_out_free_nskb;
677 }
678 entry->id = ++queue->id_sequence;
679 *packet_id_ptr = htonl(entry->id);
680
681 /* nfnetlink_unicast will either free the nskb or add it to a socket */
682 err = nfnetlink_unicast(nskb, net, queue->peer_portid);
683 if (err < 0) {
684 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
685 failopen = 1;
686 err = 0;
687 } else {
688 queue->queue_user_dropped++;
689 }
690 goto err_out_unlock;
691 }
692
693 __enqueue_entry(queue, entry);
694
695 spin_unlock_bh(&queue->lock);
696 return 0;
697
698err_out_free_nskb:
699 kfree_skb(nskb);
700err_out_unlock:
701 spin_unlock_bh(&queue->lock);
702 if (failopen)
703 nfqnl_reinject(entry, NF_ACCEPT);
704err_out:
705 return err;
706}
707
708static struct nf_queue_entry *
709nf_queue_entry_dup(struct nf_queue_entry *e)
710{
711 struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
712 if (entry)
713 nf_queue_entry_get_refs(entry);
714 return entry;
715}
716
717#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
718/* When called from bridge netfilter, skb->data must point to MAC header
719 * before calling skb_gso_segment(). Else, original MAC header is lost
720 * and segmented skbs will be sent to wrong destination.
721 */
722static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
723{
724 if (nf_bridge_info_get(skb))
725 __skb_push(skb, skb->network_header - skb->mac_header);
726}
727
728static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
729{
730 if (nf_bridge_info_get(skb))
731 __skb_pull(skb, skb->network_header - skb->mac_header);
732}
733#else
734#define nf_bridge_adjust_skb_data(s) do {} while (0)
735#define nf_bridge_adjust_segmented_data(s) do {} while (0)
736#endif
737
738static int
739__nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
740 struct sk_buff *skb, struct nf_queue_entry *entry)
741{
742 int ret = -ENOMEM;
743 struct nf_queue_entry *entry_seg;
744
745 nf_bridge_adjust_segmented_data(skb);
746
747 if (skb->next == NULL) { /* last packet, no need to copy entry */
748 struct sk_buff *gso_skb = entry->skb;
749 entry->skb = skb;
750 ret = __nfqnl_enqueue_packet(net, queue, entry);
751 if (ret)
752 entry->skb = gso_skb;
753 return ret;
754 }
755
756 skb_mark_not_on_list(skb);
757
758 entry_seg = nf_queue_entry_dup(entry);
759 if (entry_seg) {
760 entry_seg->skb = skb;
761 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
762 if (ret)
763 nf_queue_entry_free(entry_seg);
764 }
765 return ret;
766}
767
768static int
769nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
770{
771 unsigned int queued;
772 struct nfqnl_instance *queue;
773 struct sk_buff *skb, *segs, *nskb;
774 int err = -ENOBUFS;
775 struct net *net = entry->state.net;
776 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
777
778 /* rcu_read_lock()ed by nf_hook_thresh */
779 queue = instance_lookup(q, queuenum);
780 if (!queue)
781 return -ESRCH;
782
783 if (queue->copy_mode == NFQNL_COPY_NONE)
784 return -EINVAL;
785
786 skb = entry->skb;
787
788 switch (entry->state.pf) {
789 case NFPROTO_IPV4:
790 skb->protocol = htons(ETH_P_IP);
791 break;
792 case NFPROTO_IPV6:
793 skb->protocol = htons(ETH_P_IPV6);
794 break;
795 }
796
797 if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(skb))
798 return __nfqnl_enqueue_packet(net, queue, entry);
799
800 nf_bridge_adjust_skb_data(skb);
801 segs = skb_gso_segment(skb, 0);
802 /* Does not use PTR_ERR to limit the number of error codes that can be
803 * returned by nf_queue. For instance, callers rely on -ESRCH to
804 * mean 'ignore this hook'.
805 */
806 if (IS_ERR_OR_NULL(segs))
807 goto out_err;
808 queued = 0;
809 err = 0;
810 skb_list_walk_safe(segs, segs, nskb) {
811 if (err == 0)
812 err = __nfqnl_enqueue_packet_gso(net, queue,
813 segs, entry);
814 if (err == 0)
815 queued++;
816 else
817 kfree_skb(segs);
818 }
819
820 if (queued) {
821 if (err) /* some segments are already queued */
822 nf_queue_entry_free(entry);
823 kfree_skb(skb);
824 return 0;
825 }
826 out_err:
827 nf_bridge_adjust_segmented_data(skb);
828 return err;
829}
830
831static int
832nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e, int diff)
833{
834 struct sk_buff *nskb;
835
836 if (diff < 0) {
837 if (pskb_trim(e->skb, data_len))
838 return -ENOMEM;
839 } else if (diff > 0) {
840 if (data_len > 0xFFFF)
841 return -EINVAL;
842 if (diff > skb_tailroom(e->skb)) {
843 nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
844 diff, GFP_ATOMIC);
845 if (!nskb)
846 return -ENOMEM;
847 kfree_skb(e->skb);
848 e->skb = nskb;
849 }
850 skb_put(e->skb, diff);
851 }
852 if (skb_ensure_writable(e->skb, data_len))
853 return -ENOMEM;
854 skb_copy_to_linear_data(e->skb, data, data_len);
855 e->skb->ip_summed = CHECKSUM_NONE;
856 return 0;
857}
858
859static int
860nfqnl_set_mode(struct nfqnl_instance *queue,
861 unsigned char mode, unsigned int range)
862{
863 int status = 0;
864
865 spin_lock_bh(&queue->lock);
866 switch (mode) {
867 case NFQNL_COPY_NONE:
868 case NFQNL_COPY_META:
869 queue->copy_mode = mode;
870 queue->copy_range = 0;
871 break;
872
873 case NFQNL_COPY_PACKET:
874 queue->copy_mode = mode;
875 if (range == 0 || range > NFQNL_MAX_COPY_RANGE)
876 queue->copy_range = NFQNL_MAX_COPY_RANGE;
877 else
878 queue->copy_range = range;
879 break;
880
881 default:
882 status = -EINVAL;
883
884 }
885 spin_unlock_bh(&queue->lock);
886
887 return status;
888}
889
890static int
891dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
892{
893#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
894 int physinif, physoutif;
895
896 physinif = nf_bridge_get_physinif(entry->skb);
897 physoutif = nf_bridge_get_physoutif(entry->skb);
898
899 if (physinif == ifindex || physoutif == ifindex)
900 return 1;
901#endif
902 if (entry->state.in)
903 if (entry->state.in->ifindex == ifindex)
904 return 1;
905 if (entry->state.out)
906 if (entry->state.out->ifindex == ifindex)
907 return 1;
908
909 return 0;
910}
911
912/* drop all packets with either indev or outdev == ifindex from all queue
913 * instances */
914static void
915nfqnl_dev_drop(struct net *net, int ifindex)
916{
917 int i;
918 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
919
920 rcu_read_lock();
921
922 for (i = 0; i < INSTANCE_BUCKETS; i++) {
923 struct nfqnl_instance *inst;
924 struct hlist_head *head = &q->instance_table[i];
925
926 hlist_for_each_entry_rcu(inst, head, hlist)
927 nfqnl_flush(inst, dev_cmp, ifindex);
928 }
929
930 rcu_read_unlock();
931}
932
933static int
934nfqnl_rcv_dev_event(struct notifier_block *this,
935 unsigned long event, void *ptr)
936{
937 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
938
939 /* Drop any packets associated with the downed device */
940 if (event == NETDEV_DOWN)
941 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
942 return NOTIFY_DONE;
943}
944
945static struct notifier_block nfqnl_dev_notifier = {
946 .notifier_call = nfqnl_rcv_dev_event,
947};
948
949static void nfqnl_nf_hook_drop(struct net *net)
950{
951 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
952 int i;
953
954 for (i = 0; i < INSTANCE_BUCKETS; i++) {
955 struct nfqnl_instance *inst;
956 struct hlist_head *head = &q->instance_table[i];
957
958 hlist_for_each_entry_rcu(inst, head, hlist)
959 nfqnl_flush(inst, NULL, 0);
960 }
961}
962
963static int
964nfqnl_rcv_nl_event(struct notifier_block *this,
965 unsigned long event, void *ptr)
966{
967 struct netlink_notify *n = ptr;
968 struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
969
970 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
971 int i;
972
973 /* destroy all instances for this portid */
974 spin_lock(&q->instances_lock);
975 for (i = 0; i < INSTANCE_BUCKETS; i++) {
976 struct hlist_node *t2;
977 struct nfqnl_instance *inst;
978 struct hlist_head *head = &q->instance_table[i];
979
980 hlist_for_each_entry_safe(inst, t2, head, hlist) {
981 if (n->portid == inst->peer_portid)
982 __instance_destroy(inst);
983 }
984 }
985 spin_unlock(&q->instances_lock);
986 }
987 return NOTIFY_DONE;
988}
989
990static struct notifier_block nfqnl_rtnl_notifier = {
991 .notifier_call = nfqnl_rcv_nl_event,
992};
993
994static const struct nla_policy nfqa_vlan_policy[NFQA_VLAN_MAX + 1] = {
995 [NFQA_VLAN_TCI] = { .type = NLA_U16},
996 [NFQA_VLAN_PROTO] = { .type = NLA_U16},
997};
998
999static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
1000 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
1001 [NFQA_MARK] = { .type = NLA_U32 },
1002 [NFQA_PAYLOAD] = { .type = NLA_UNSPEC },
1003 [NFQA_CT] = { .type = NLA_UNSPEC },
1004 [NFQA_EXP] = { .type = NLA_UNSPEC },
1005 [NFQA_VLAN] = { .type = NLA_NESTED },
1006};
1007
1008static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
1009 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
1010 [NFQA_MARK] = { .type = NLA_U32 },
1011};
1012
1013static struct nfqnl_instance *
1014verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, u32 nlportid)
1015{
1016 struct nfqnl_instance *queue;
1017
1018 queue = instance_lookup(q, queue_num);
1019 if (!queue)
1020 return ERR_PTR(-ENODEV);
1021
1022 if (queue->peer_portid != nlportid)
1023 return ERR_PTR(-EPERM);
1024
1025 return queue;
1026}
1027
1028static struct nfqnl_msg_verdict_hdr*
1029verdicthdr_get(const struct nlattr * const nfqa[])
1030{
1031 struct nfqnl_msg_verdict_hdr *vhdr;
1032 unsigned int verdict;
1033
1034 if (!nfqa[NFQA_VERDICT_HDR])
1035 return NULL;
1036
1037 vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
1038 verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
1039 if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
1040 return NULL;
1041 return vhdr;
1042}
1043
1044static int nfq_id_after(unsigned int id, unsigned int max)
1045{
1046 return (int)(id - max) > 0;
1047}
1048
1049static int nfqnl_recv_verdict_batch(struct sk_buff *skb,
1050 const struct nfnl_info *info,
1051 const struct nlattr * const nfqa[])
1052{
1053 struct nfnl_queue_net *q = nfnl_queue_pernet(info->net);
1054 u16 queue_num = ntohs(info->nfmsg->res_id);
1055 struct nf_queue_entry *entry, *tmp;
1056 struct nfqnl_msg_verdict_hdr *vhdr;
1057 struct nfqnl_instance *queue;
1058 unsigned int verdict, maxid;
1059 LIST_HEAD(batch_list);
1060
1061 queue = verdict_instance_lookup(q, queue_num,
1062 NETLINK_CB(skb).portid);
1063 if (IS_ERR(queue))
1064 return PTR_ERR(queue);
1065
1066 vhdr = verdicthdr_get(nfqa);
1067 if (!vhdr)
1068 return -EINVAL;
1069
1070 verdict = ntohl(vhdr->verdict);
1071 maxid = ntohl(vhdr->id);
1072
1073 spin_lock_bh(&queue->lock);
1074
1075 list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
1076 if (nfq_id_after(entry->id, maxid))
1077 break;
1078 __dequeue_entry(queue, entry);
1079 list_add_tail(&entry->list, &batch_list);
1080 }
1081
1082 spin_unlock_bh(&queue->lock);
1083
1084 if (list_empty(&batch_list))
1085 return -ENOENT;
1086
1087 list_for_each_entry_safe(entry, tmp, &batch_list, list) {
1088 if (nfqa[NFQA_MARK])
1089 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1090
1091 nfqnl_reinject(entry, verdict);
1092 }
1093 return 0;
1094}
1095
1096static struct nf_conn *nfqnl_ct_parse(struct nfnl_ct_hook *nfnl_ct,
1097 const struct nlmsghdr *nlh,
1098 const struct nlattr * const nfqa[],
1099 struct nf_queue_entry *entry,
1100 enum ip_conntrack_info *ctinfo)
1101{
1102#if IS_ENABLED(CONFIG_NF_CONNTRACK)
1103 struct nf_conn *ct;
1104
1105 ct = nf_ct_get(entry->skb, ctinfo);
1106 if (ct == NULL)
1107 return NULL;
1108
1109 if (nfnl_ct->parse(nfqa[NFQA_CT], ct) < 0)
1110 return NULL;
1111
1112 if (nfqa[NFQA_EXP])
1113 nfnl_ct->attach_expect(nfqa[NFQA_EXP], ct,
1114 NETLINK_CB(entry->skb).portid,
1115 nlmsg_report(nlh));
1116 return ct;
1117#else
1118 return NULL;
1119#endif
1120}
1121
1122static int nfqa_parse_bridge(struct nf_queue_entry *entry,
1123 const struct nlattr * const nfqa[])
1124{
1125 if (nfqa[NFQA_VLAN]) {
1126 struct nlattr *tb[NFQA_VLAN_MAX + 1];
1127 int err;
1128
1129 err = nla_parse_nested_deprecated(tb, NFQA_VLAN_MAX,
1130 nfqa[NFQA_VLAN],
1131 nfqa_vlan_policy, NULL);
1132 if (err < 0)
1133 return err;
1134
1135 if (!tb[NFQA_VLAN_TCI] || !tb[NFQA_VLAN_PROTO])
1136 return -EINVAL;
1137
1138 __vlan_hwaccel_put_tag(entry->skb,
1139 nla_get_be16(tb[NFQA_VLAN_PROTO]),
1140 ntohs(nla_get_be16(tb[NFQA_VLAN_TCI])));
1141 }
1142
1143 if (nfqa[NFQA_L2HDR]) {
1144 int mac_header_len = entry->skb->network_header -
1145 entry->skb->mac_header;
1146
1147 if (mac_header_len != nla_len(nfqa[NFQA_L2HDR]))
1148 return -EINVAL;
1149 else if (mac_header_len > 0)
1150 memcpy(skb_mac_header(entry->skb),
1151 nla_data(nfqa[NFQA_L2HDR]),
1152 mac_header_len);
1153 }
1154
1155 return 0;
1156}
1157
1158static int nfqnl_recv_verdict(struct sk_buff *skb, const struct nfnl_info *info,
1159 const struct nlattr * const nfqa[])
1160{
1161 struct nfnl_queue_net *q = nfnl_queue_pernet(info->net);
1162 u_int16_t queue_num = ntohs(info->nfmsg->res_id);
1163 struct nfqnl_msg_verdict_hdr *vhdr;
1164 enum ip_conntrack_info ctinfo;
1165 struct nfqnl_instance *queue;
1166 struct nf_queue_entry *entry;
1167 struct nfnl_ct_hook *nfnl_ct;
1168 struct nf_conn *ct = NULL;
1169 unsigned int verdict;
1170 int err;
1171
1172 queue = verdict_instance_lookup(q, queue_num,
1173 NETLINK_CB(skb).portid);
1174 if (IS_ERR(queue))
1175 return PTR_ERR(queue);
1176
1177 vhdr = verdicthdr_get(nfqa);
1178 if (!vhdr)
1179 return -EINVAL;
1180
1181 verdict = ntohl(vhdr->verdict);
1182
1183 entry = find_dequeue_entry(queue, ntohl(vhdr->id));
1184 if (entry == NULL)
1185 return -ENOENT;
1186
1187 /* rcu lock already held from nfnl->call_rcu. */
1188 nfnl_ct = rcu_dereference(nfnl_ct_hook);
1189
1190 if (nfqa[NFQA_CT]) {
1191 if (nfnl_ct != NULL)
1192 ct = nfqnl_ct_parse(nfnl_ct, info->nlh, nfqa, entry,
1193 &ctinfo);
1194 }
1195
1196 if (entry->state.pf == PF_BRIDGE) {
1197 err = nfqa_parse_bridge(entry, nfqa);
1198 if (err < 0)
1199 return err;
1200 }
1201
1202 if (nfqa[NFQA_PAYLOAD]) {
1203 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
1204 int diff = payload_len - entry->skb->len;
1205
1206 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
1207 payload_len, entry, diff) < 0)
1208 verdict = NF_DROP;
1209
1210 if (ct && diff)
1211 nfnl_ct->seq_adjust(entry->skb, ct, ctinfo, diff);
1212 }
1213
1214 if (nfqa[NFQA_MARK])
1215 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1216
1217 nfqnl_reinject(entry, verdict);
1218 return 0;
1219}
1220
1221static int nfqnl_recv_unsupp(struct sk_buff *skb, const struct nfnl_info *info,
1222 const struct nlattr * const cda[])
1223{
1224 return -ENOTSUPP;
1225}
1226
1227static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1228 [NFQA_CFG_CMD] = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1229 [NFQA_CFG_PARAMS] = { .len = sizeof(struct nfqnl_msg_config_params) },
1230 [NFQA_CFG_QUEUE_MAXLEN] = { .type = NLA_U32 },
1231 [NFQA_CFG_MASK] = { .type = NLA_U32 },
1232 [NFQA_CFG_FLAGS] = { .type = NLA_U32 },
1233};
1234
1235static const struct nf_queue_handler nfqh = {
1236 .outfn = nfqnl_enqueue_packet,
1237 .nf_hook_drop = nfqnl_nf_hook_drop,
1238};
1239
1240static int nfqnl_recv_config(struct sk_buff *skb, const struct nfnl_info *info,
1241 const struct nlattr * const nfqa[])
1242{
1243 struct nfnl_queue_net *q = nfnl_queue_pernet(info->net);
1244 u_int16_t queue_num = ntohs(info->nfmsg->res_id);
1245 struct nfqnl_msg_config_cmd *cmd = NULL;
1246 struct nfqnl_instance *queue;
1247 __u32 flags = 0, mask = 0;
1248 int ret = 0;
1249
1250 if (nfqa[NFQA_CFG_CMD]) {
1251 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1252
1253 /* Obsolete commands without queue context */
1254 switch (cmd->command) {
1255 case NFQNL_CFG_CMD_PF_BIND: return 0;
1256 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1257 }
1258 }
1259
1260 /* Check if we support these flags in first place, dependencies should
1261 * be there too not to break atomicity.
1262 */
1263 if (nfqa[NFQA_CFG_FLAGS]) {
1264 if (!nfqa[NFQA_CFG_MASK]) {
1265 /* A mask is needed to specify which flags are being
1266 * changed.
1267 */
1268 return -EINVAL;
1269 }
1270
1271 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1272 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1273
1274 if (flags >= NFQA_CFG_F_MAX)
1275 return -EOPNOTSUPP;
1276
1277#if !IS_ENABLED(CONFIG_NETWORK_SECMARK)
1278 if (flags & mask & NFQA_CFG_F_SECCTX)
1279 return -EOPNOTSUPP;
1280#endif
1281 if ((flags & mask & NFQA_CFG_F_CONNTRACK) &&
1282 !rcu_access_pointer(nfnl_ct_hook)) {
1283#ifdef CONFIG_MODULES
1284 nfnl_unlock(NFNL_SUBSYS_QUEUE);
1285 request_module("ip_conntrack_netlink");
1286 nfnl_lock(NFNL_SUBSYS_QUEUE);
1287 if (rcu_access_pointer(nfnl_ct_hook))
1288 return -EAGAIN;
1289#endif
1290 return -EOPNOTSUPP;
1291 }
1292 }
1293
1294 rcu_read_lock();
1295 queue = instance_lookup(q, queue_num);
1296 if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1297 ret = -EPERM;
1298 goto err_out_unlock;
1299 }
1300
1301 if (cmd != NULL) {
1302 switch (cmd->command) {
1303 case NFQNL_CFG_CMD_BIND:
1304 if (queue) {
1305 ret = -EBUSY;
1306 goto err_out_unlock;
1307 }
1308 queue = instance_create(q, queue_num,
1309 NETLINK_CB(skb).portid);
1310 if (IS_ERR(queue)) {
1311 ret = PTR_ERR(queue);
1312 goto err_out_unlock;
1313 }
1314 break;
1315 case NFQNL_CFG_CMD_UNBIND:
1316 if (!queue) {
1317 ret = -ENODEV;
1318 goto err_out_unlock;
1319 }
1320 instance_destroy(q, queue);
1321 goto err_out_unlock;
1322 case NFQNL_CFG_CMD_PF_BIND:
1323 case NFQNL_CFG_CMD_PF_UNBIND:
1324 break;
1325 default:
1326 ret = -ENOTSUPP;
1327 goto err_out_unlock;
1328 }
1329 }
1330
1331 if (!queue) {
1332 ret = -ENODEV;
1333 goto err_out_unlock;
1334 }
1335
1336 if (nfqa[NFQA_CFG_PARAMS]) {
1337 struct nfqnl_msg_config_params *params =
1338 nla_data(nfqa[NFQA_CFG_PARAMS]);
1339
1340 nfqnl_set_mode(queue, params->copy_mode,
1341 ntohl(params->copy_range));
1342 }
1343
1344 if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1345 __be32 *queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1346
1347 spin_lock_bh(&queue->lock);
1348 queue->queue_maxlen = ntohl(*queue_maxlen);
1349 spin_unlock_bh(&queue->lock);
1350 }
1351
1352 if (nfqa[NFQA_CFG_FLAGS]) {
1353 spin_lock_bh(&queue->lock);
1354 queue->flags &= ~mask;
1355 queue->flags |= flags & mask;
1356 spin_unlock_bh(&queue->lock);
1357 }
1358
1359err_out_unlock:
1360 rcu_read_unlock();
1361 return ret;
1362}
1363
1364static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1365 [NFQNL_MSG_PACKET] = {
1366 .call = nfqnl_recv_unsupp,
1367 .type = NFNL_CB_RCU,
1368 .attr_count = NFQA_MAX,
1369 },
1370 [NFQNL_MSG_VERDICT] = {
1371 .call = nfqnl_recv_verdict,
1372 .type = NFNL_CB_RCU,
1373 .attr_count = NFQA_MAX,
1374 .policy = nfqa_verdict_policy
1375 },
1376 [NFQNL_MSG_CONFIG] = {
1377 .call = nfqnl_recv_config,
1378 .type = NFNL_CB_MUTEX,
1379 .attr_count = NFQA_CFG_MAX,
1380 .policy = nfqa_cfg_policy
1381 },
1382 [NFQNL_MSG_VERDICT_BATCH] = {
1383 .call = nfqnl_recv_verdict_batch,
1384 .type = NFNL_CB_RCU,
1385 .attr_count = NFQA_MAX,
1386 .policy = nfqa_verdict_batch_policy
1387 },
1388};
1389
1390static const struct nfnetlink_subsystem nfqnl_subsys = {
1391 .name = "nf_queue",
1392 .subsys_id = NFNL_SUBSYS_QUEUE,
1393 .cb_count = NFQNL_MSG_MAX,
1394 .cb = nfqnl_cb,
1395};
1396
1397#ifdef CONFIG_PROC_FS
1398struct iter_state {
1399 struct seq_net_private p;
1400 unsigned int bucket;
1401};
1402
1403static struct hlist_node *get_first(struct seq_file *seq)
1404{
1405 struct iter_state *st = seq->private;
1406 struct net *net;
1407 struct nfnl_queue_net *q;
1408
1409 if (!st)
1410 return NULL;
1411
1412 net = seq_file_net(seq);
1413 q = nfnl_queue_pernet(net);
1414 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1415 if (!hlist_empty(&q->instance_table[st->bucket]))
1416 return q->instance_table[st->bucket].first;
1417 }
1418 return NULL;
1419}
1420
1421static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1422{
1423 struct iter_state *st = seq->private;
1424 struct net *net = seq_file_net(seq);
1425
1426 h = h->next;
1427 while (!h) {
1428 struct nfnl_queue_net *q;
1429
1430 if (++st->bucket >= INSTANCE_BUCKETS)
1431 return NULL;
1432
1433 q = nfnl_queue_pernet(net);
1434 h = q->instance_table[st->bucket].first;
1435 }
1436 return h;
1437}
1438
1439static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1440{
1441 struct hlist_node *head;
1442 head = get_first(seq);
1443
1444 if (head)
1445 while (pos && (head = get_next(seq, head)))
1446 pos--;
1447 return pos ? NULL : head;
1448}
1449
1450static void *seq_start(struct seq_file *s, loff_t *pos)
1451 __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1452{
1453 spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1454 return get_idx(s, *pos);
1455}
1456
1457static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1458{
1459 (*pos)++;
1460 return get_next(s, v);
1461}
1462
1463static void seq_stop(struct seq_file *s, void *v)
1464 __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1465{
1466 spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1467}
1468
1469static int seq_show(struct seq_file *s, void *v)
1470{
1471 const struct nfqnl_instance *inst = v;
1472
1473 seq_printf(s, "%5u %6u %5u %1u %5u %5u %5u %8u %2d\n",
1474 inst->queue_num,
1475 inst->peer_portid, inst->queue_total,
1476 inst->copy_mode, inst->copy_range,
1477 inst->queue_dropped, inst->queue_user_dropped,
1478 inst->id_sequence, 1);
1479 return 0;
1480}
1481
1482static const struct seq_operations nfqnl_seq_ops = {
1483 .start = seq_start,
1484 .next = seq_next,
1485 .stop = seq_stop,
1486 .show = seq_show,
1487};
1488#endif /* PROC_FS */
1489
1490static int __net_init nfnl_queue_net_init(struct net *net)
1491{
1492 unsigned int i;
1493 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1494
1495 for (i = 0; i < INSTANCE_BUCKETS; i++)
1496 INIT_HLIST_HEAD(&q->instance_table[i]);
1497
1498 spin_lock_init(&q->instances_lock);
1499
1500#ifdef CONFIG_PROC_FS
1501 if (!proc_create_net("nfnetlink_queue", 0440, net->nf.proc_netfilter,
1502 &nfqnl_seq_ops, sizeof(struct iter_state)))
1503 return -ENOMEM;
1504#endif
1505 nf_register_queue_handler(net, &nfqh);
1506 return 0;
1507}
1508
1509static void __net_exit nfnl_queue_net_exit(struct net *net)
1510{
1511 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1512 unsigned int i;
1513
1514 nf_unregister_queue_handler(net);
1515#ifdef CONFIG_PROC_FS
1516 remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1517#endif
1518 for (i = 0; i < INSTANCE_BUCKETS; i++)
1519 WARN_ON_ONCE(!hlist_empty(&q->instance_table[i]));
1520}
1521
1522static void nfnl_queue_net_exit_batch(struct list_head *net_exit_list)
1523{
1524 synchronize_rcu();
1525}
1526
1527static struct pernet_operations nfnl_queue_net_ops = {
1528 .init = nfnl_queue_net_init,
1529 .exit = nfnl_queue_net_exit,
1530 .exit_batch = nfnl_queue_net_exit_batch,
1531 .id = &nfnl_queue_net_id,
1532 .size = sizeof(struct nfnl_queue_net),
1533};
1534
1535static int __init nfnetlink_queue_init(void)
1536{
1537 int status;
1538
1539 status = register_pernet_subsys(&nfnl_queue_net_ops);
1540 if (status < 0) {
1541 pr_err("failed to register pernet ops\n");
1542 goto out;
1543 }
1544
1545 netlink_register_notifier(&nfqnl_rtnl_notifier);
1546 status = nfnetlink_subsys_register(&nfqnl_subsys);
1547 if (status < 0) {
1548 pr_err("failed to create netlink socket\n");
1549 goto cleanup_netlink_notifier;
1550 }
1551
1552 status = register_netdevice_notifier(&nfqnl_dev_notifier);
1553 if (status < 0) {
1554 pr_err("failed to register netdevice notifier\n");
1555 goto cleanup_netlink_subsys;
1556 }
1557
1558 return status;
1559
1560cleanup_netlink_subsys:
1561 nfnetlink_subsys_unregister(&nfqnl_subsys);
1562cleanup_netlink_notifier:
1563 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1564 unregister_pernet_subsys(&nfnl_queue_net_ops);
1565out:
1566 return status;
1567}
1568
1569static void __exit nfnetlink_queue_fini(void)
1570{
1571 unregister_netdevice_notifier(&nfqnl_dev_notifier);
1572 nfnetlink_subsys_unregister(&nfqnl_subsys);
1573 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1574 unregister_pernet_subsys(&nfnl_queue_net_ops);
1575
1576 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1577}
1578
1579MODULE_DESCRIPTION("netfilter packet queue handler");
1580MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1581MODULE_LICENSE("GPL");
1582MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1583
1584module_init(nfnetlink_queue_init);
1585module_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 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
292 goto nla_put_failure;
293#else
294 if (entry->pf == PF_BRIDGE) {
295 /* Case 1: indev is physical input device, we need to
296 * look for bridge group (when called from
297 * netfilter_bridge) */
298 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
299 htonl(indev->ifindex)) ||
300 /* this is the bridge group "brX" */
301 /* rcu_read_lock()ed by __nf_queue */
302 nla_put_be32(skb, NFQA_IFINDEX_INDEV,
303 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
304 goto nla_put_failure;
305 } else {
306 /* Case 2: indev is bridge group, we need to look for
307 * physical device (when called from ipv4) */
308 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
309 htonl(indev->ifindex)))
310 goto nla_put_failure;
311 if (entskb->nf_bridge && entskb->nf_bridge->physindev &&
312 nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
313 htonl(entskb->nf_bridge->physindev->ifindex)))
314 goto nla_put_failure;
315 }
316#endif
317 }
318
319 if (outdev) {
320#ifndef CONFIG_BRIDGE_NETFILTER
321 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
322 goto nla_put_failure;
323#else
324 if (entry->pf == PF_BRIDGE) {
325 /* Case 1: outdev is physical output device, we need to
326 * look for bridge group (when called from
327 * netfilter_bridge) */
328 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
329 htonl(outdev->ifindex)) ||
330 /* this is the bridge group "brX" */
331 /* rcu_read_lock()ed by __nf_queue */
332 nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
333 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
334 goto nla_put_failure;
335 } else {
336 /* Case 2: outdev is bridge group, we need to look for
337 * physical output device (when called from ipv4) */
338 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
339 htonl(outdev->ifindex)))
340 goto nla_put_failure;
341 if (entskb->nf_bridge && entskb->nf_bridge->physoutdev &&
342 nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
343 htonl(entskb->nf_bridge->physoutdev->ifindex)))
344 goto nla_put_failure;
345 }
346#endif
347 }
348
349 if (entskb->mark &&
350 nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
351 goto nla_put_failure;
352
353 if (indev && entskb->dev &&
354 entskb->mac_header != entskb->network_header) {
355 struct nfqnl_msg_packet_hw phw;
356 int len = dev_parse_header(entskb, phw.hw_addr);
357 if (len) {
358 phw.hw_addrlen = htons(len);
359 if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
360 goto nla_put_failure;
361 }
362 }
363
364 if (entskb->tstamp.tv64) {
365 struct nfqnl_msg_packet_timestamp ts;
366 struct timeval tv = ktime_to_timeval(entskb->tstamp);
367 ts.sec = cpu_to_be64(tv.tv_sec);
368 ts.usec = cpu_to_be64(tv.tv_usec);
369
370 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
371 goto nla_put_failure;
372 }
373
374 if (data_len) {
375 struct nlattr *nla;
376 int sz = nla_attr_size(data_len);
377
378 if (skb_tailroom(skb) < nla_total_size(data_len)) {
379 printk(KERN_WARNING "nf_queue: no tailroom!\n");
380 goto nlmsg_failure;
381 }
382
383 nla = (struct nlattr *)skb_put(skb, nla_total_size(data_len));
384 nla->nla_type = NFQA_PAYLOAD;
385 nla->nla_len = sz;
386
387 if (skb_copy_bits(entskb, 0, nla_data(nla), data_len))
388 BUG();
389 }
390
391 nlh->nlmsg_len = skb->tail - old_tail;
392 return skb;
393
394nlmsg_failure:
395nla_put_failure:
396 if (skb)
397 kfree_skb(skb);
398 net_err_ratelimited("nf_queue: error creating packet message\n");
399 return NULL;
400}
401
402static int
403nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
404{
405 struct sk_buff *nskb;
406 struct nfqnl_instance *queue;
407 int err = -ENOBUFS;
408 __be32 *packet_id_ptr;
409
410 /* rcu_read_lock()ed by nf_hook_slow() */
411 queue = instance_lookup(queuenum);
412 if (!queue) {
413 err = -ESRCH;
414 goto err_out;
415 }
416
417 if (queue->copy_mode == NFQNL_COPY_NONE) {
418 err = -EINVAL;
419 goto err_out;
420 }
421
422 nskb = nfqnl_build_packet_message(queue, entry, &packet_id_ptr);
423 if (nskb == NULL) {
424 err = -ENOMEM;
425 goto err_out;
426 }
427 spin_lock_bh(&queue->lock);
428
429 if (!queue->peer_pid) {
430 err = -EINVAL;
431 goto err_out_free_nskb;
432 }
433 if (queue->queue_total >= queue->queue_maxlen) {
434 queue->queue_dropped++;
435 net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
436 queue->queue_total);
437 goto err_out_free_nskb;
438 }
439 entry->id = ++queue->id_sequence;
440 *packet_id_ptr = htonl(entry->id);
441
442 /* nfnetlink_unicast will either free the nskb or add it to a socket */
443 err = nfnetlink_unicast(nskb, &init_net, queue->peer_pid, MSG_DONTWAIT);
444 if (err < 0) {
445 queue->queue_user_dropped++;
446 goto err_out_unlock;
447 }
448
449 __enqueue_entry(queue, entry);
450
451 spin_unlock_bh(&queue->lock);
452 return 0;
453
454err_out_free_nskb:
455 kfree_skb(nskb);
456err_out_unlock:
457 spin_unlock_bh(&queue->lock);
458err_out:
459 return err;
460}
461
462static int
463nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e)
464{
465 struct sk_buff *nskb;
466 int diff;
467
468 diff = data_len - e->skb->len;
469 if (diff < 0) {
470 if (pskb_trim(e->skb, data_len))
471 return -ENOMEM;
472 } else if (diff > 0) {
473 if (data_len > 0xFFFF)
474 return -EINVAL;
475 if (diff > skb_tailroom(e->skb)) {
476 nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
477 diff, GFP_ATOMIC);
478 if (!nskb) {
479 printk(KERN_WARNING "nf_queue: OOM "
480 "in mangle, dropping packet\n");
481 return -ENOMEM;
482 }
483 kfree_skb(e->skb);
484 e->skb = nskb;
485 }
486 skb_put(e->skb, diff);
487 }
488 if (!skb_make_writable(e->skb, data_len))
489 return -ENOMEM;
490 skb_copy_to_linear_data(e->skb, data, data_len);
491 e->skb->ip_summed = CHECKSUM_NONE;
492 return 0;
493}
494
495static int
496nfqnl_set_mode(struct nfqnl_instance *queue,
497 unsigned char mode, unsigned int range)
498{
499 int status = 0;
500
501 spin_lock_bh(&queue->lock);
502 switch (mode) {
503 case NFQNL_COPY_NONE:
504 case NFQNL_COPY_META:
505 queue->copy_mode = mode;
506 queue->copy_range = 0;
507 break;
508
509 case NFQNL_COPY_PACKET:
510 queue->copy_mode = mode;
511 /* we're using struct nlattr which has 16bit nla_len */
512 if (range > 0xffff)
513 queue->copy_range = 0xffff;
514 else
515 queue->copy_range = range;
516 break;
517
518 default:
519 status = -EINVAL;
520
521 }
522 spin_unlock_bh(&queue->lock);
523
524 return status;
525}
526
527static int
528dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
529{
530 if (entry->indev)
531 if (entry->indev->ifindex == ifindex)
532 return 1;
533 if (entry->outdev)
534 if (entry->outdev->ifindex == ifindex)
535 return 1;
536#ifdef CONFIG_BRIDGE_NETFILTER
537 if (entry->skb->nf_bridge) {
538 if (entry->skb->nf_bridge->physindev &&
539 entry->skb->nf_bridge->physindev->ifindex == ifindex)
540 return 1;
541 if (entry->skb->nf_bridge->physoutdev &&
542 entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
543 return 1;
544 }
545#endif
546 return 0;
547}
548
549/* drop all packets with either indev or outdev == ifindex from all queue
550 * instances */
551static void
552nfqnl_dev_drop(int ifindex)
553{
554 int i;
555
556 rcu_read_lock();
557
558 for (i = 0; i < INSTANCE_BUCKETS; i++) {
559 struct hlist_node *tmp;
560 struct nfqnl_instance *inst;
561 struct hlist_head *head = &instance_table[i];
562
563 hlist_for_each_entry_rcu(inst, tmp, head, hlist)
564 nfqnl_flush(inst, dev_cmp, ifindex);
565 }
566
567 rcu_read_unlock();
568}
569
570#define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
571
572static int
573nfqnl_rcv_dev_event(struct notifier_block *this,
574 unsigned long event, void *ptr)
575{
576 struct net_device *dev = ptr;
577
578 if (!net_eq(dev_net(dev), &init_net))
579 return NOTIFY_DONE;
580
581 /* Drop any packets associated with the downed device */
582 if (event == NETDEV_DOWN)
583 nfqnl_dev_drop(dev->ifindex);
584 return NOTIFY_DONE;
585}
586
587static struct notifier_block nfqnl_dev_notifier = {
588 .notifier_call = nfqnl_rcv_dev_event,
589};
590
591static int
592nfqnl_rcv_nl_event(struct notifier_block *this,
593 unsigned long event, void *ptr)
594{
595 struct netlink_notify *n = ptr;
596
597 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
598 int i;
599
600 /* destroy all instances for this pid */
601 spin_lock(&instances_lock);
602 for (i = 0; i < INSTANCE_BUCKETS; i++) {
603 struct hlist_node *tmp, *t2;
604 struct nfqnl_instance *inst;
605 struct hlist_head *head = &instance_table[i];
606
607 hlist_for_each_entry_safe(inst, tmp, t2, head, hlist) {
608 if ((n->net == &init_net) &&
609 (n->pid == inst->peer_pid))
610 __instance_destroy(inst);
611 }
612 }
613 spin_unlock(&instances_lock);
614 }
615 return NOTIFY_DONE;
616}
617
618static struct notifier_block nfqnl_rtnl_notifier = {
619 .notifier_call = nfqnl_rcv_nl_event,
620};
621
622static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
623 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
624 [NFQA_MARK] = { .type = NLA_U32 },
625 [NFQA_PAYLOAD] = { .type = NLA_UNSPEC },
626};
627
628static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
629 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
630 [NFQA_MARK] = { .type = NLA_U32 },
631};
632
633static struct nfqnl_instance *verdict_instance_lookup(u16 queue_num, int nlpid)
634{
635 struct nfqnl_instance *queue;
636
637 queue = instance_lookup(queue_num);
638 if (!queue)
639 return ERR_PTR(-ENODEV);
640
641 if (queue->peer_pid != nlpid)
642 return ERR_PTR(-EPERM);
643
644 return queue;
645}
646
647static struct nfqnl_msg_verdict_hdr*
648verdicthdr_get(const struct nlattr * const nfqa[])
649{
650 struct nfqnl_msg_verdict_hdr *vhdr;
651 unsigned int verdict;
652
653 if (!nfqa[NFQA_VERDICT_HDR])
654 return NULL;
655
656 vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
657 verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
658 if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
659 return NULL;
660 return vhdr;
661}
662
663static int nfq_id_after(unsigned int id, unsigned int max)
664{
665 return (int)(id - max) > 0;
666}
667
668static int
669nfqnl_recv_verdict_batch(struct sock *ctnl, struct sk_buff *skb,
670 const struct nlmsghdr *nlh,
671 const struct nlattr * const nfqa[])
672{
673 struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
674 struct nf_queue_entry *entry, *tmp;
675 unsigned int verdict, maxid;
676 struct nfqnl_msg_verdict_hdr *vhdr;
677 struct nfqnl_instance *queue;
678 LIST_HEAD(batch_list);
679 u16 queue_num = ntohs(nfmsg->res_id);
680
681 queue = verdict_instance_lookup(queue_num, NETLINK_CB(skb).pid);
682 if (IS_ERR(queue))
683 return PTR_ERR(queue);
684
685 vhdr = verdicthdr_get(nfqa);
686 if (!vhdr)
687 return -EINVAL;
688
689 verdict = ntohl(vhdr->verdict);
690 maxid = ntohl(vhdr->id);
691
692 spin_lock_bh(&queue->lock);
693
694 list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
695 if (nfq_id_after(entry->id, maxid))
696 break;
697 __dequeue_entry(queue, entry);
698 list_add_tail(&entry->list, &batch_list);
699 }
700
701 spin_unlock_bh(&queue->lock);
702
703 if (list_empty(&batch_list))
704 return -ENOENT;
705
706 list_for_each_entry_safe(entry, tmp, &batch_list, list) {
707 if (nfqa[NFQA_MARK])
708 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
709 nf_reinject(entry, verdict);
710 }
711 return 0;
712}
713
714static int
715nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
716 const struct nlmsghdr *nlh,
717 const struct nlattr * const nfqa[])
718{
719 struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
720 u_int16_t queue_num = ntohs(nfmsg->res_id);
721
722 struct nfqnl_msg_verdict_hdr *vhdr;
723 struct nfqnl_instance *queue;
724 unsigned int verdict;
725 struct nf_queue_entry *entry;
726
727 queue = instance_lookup(queue_num);
728 if (!queue)
729
730 queue = verdict_instance_lookup(queue_num, NETLINK_CB(skb).pid);
731 if (IS_ERR(queue))
732 return PTR_ERR(queue);
733
734 vhdr = verdicthdr_get(nfqa);
735 if (!vhdr)
736 return -EINVAL;
737
738 verdict = ntohl(vhdr->verdict);
739
740 entry = find_dequeue_entry(queue, ntohl(vhdr->id));
741 if (entry == NULL)
742 return -ENOENT;
743
744 if (nfqa[NFQA_PAYLOAD]) {
745 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
746 nla_len(nfqa[NFQA_PAYLOAD]), entry) < 0)
747 verdict = NF_DROP;
748 }
749
750 if (nfqa[NFQA_MARK])
751 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
752
753 nf_reinject(entry, verdict);
754 return 0;
755}
756
757static int
758nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
759 const struct nlmsghdr *nlh,
760 const struct nlattr * const nfqa[])
761{
762 return -ENOTSUPP;
763}
764
765static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
766 [NFQA_CFG_CMD] = { .len = sizeof(struct nfqnl_msg_config_cmd) },
767 [NFQA_CFG_PARAMS] = { .len = sizeof(struct nfqnl_msg_config_params) },
768};
769
770static const struct nf_queue_handler nfqh = {
771 .name = "nf_queue",
772 .outfn = &nfqnl_enqueue_packet,
773};
774
775static int
776nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
777 const struct nlmsghdr *nlh,
778 const struct nlattr * const nfqa[])
779{
780 struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
781 u_int16_t queue_num = ntohs(nfmsg->res_id);
782 struct nfqnl_instance *queue;
783 struct nfqnl_msg_config_cmd *cmd = NULL;
784 int ret = 0;
785
786 if (nfqa[NFQA_CFG_CMD]) {
787 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
788
789 /* Commands without queue context - might sleep */
790 switch (cmd->command) {
791 case NFQNL_CFG_CMD_PF_BIND:
792 return nf_register_queue_handler(ntohs(cmd->pf),
793 &nfqh);
794 case NFQNL_CFG_CMD_PF_UNBIND:
795 return nf_unregister_queue_handler(ntohs(cmd->pf),
796 &nfqh);
797 }
798 }
799
800 rcu_read_lock();
801 queue = instance_lookup(queue_num);
802 if (queue && queue->peer_pid != NETLINK_CB(skb).pid) {
803 ret = -EPERM;
804 goto err_out_unlock;
805 }
806
807 if (cmd != NULL) {
808 switch (cmd->command) {
809 case NFQNL_CFG_CMD_BIND:
810 if (queue) {
811 ret = -EBUSY;
812 goto err_out_unlock;
813 }
814 queue = instance_create(queue_num, NETLINK_CB(skb).pid);
815 if (IS_ERR(queue)) {
816 ret = PTR_ERR(queue);
817 goto err_out_unlock;
818 }
819 break;
820 case NFQNL_CFG_CMD_UNBIND:
821 if (!queue) {
822 ret = -ENODEV;
823 goto err_out_unlock;
824 }
825 instance_destroy(queue);
826 break;
827 case NFQNL_CFG_CMD_PF_BIND:
828 case NFQNL_CFG_CMD_PF_UNBIND:
829 break;
830 default:
831 ret = -ENOTSUPP;
832 break;
833 }
834 }
835
836 if (nfqa[NFQA_CFG_PARAMS]) {
837 struct nfqnl_msg_config_params *params;
838
839 if (!queue) {
840 ret = -ENODEV;
841 goto err_out_unlock;
842 }
843 params = nla_data(nfqa[NFQA_CFG_PARAMS]);
844 nfqnl_set_mode(queue, params->copy_mode,
845 ntohl(params->copy_range));
846 }
847
848 if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
849 __be32 *queue_maxlen;
850
851 if (!queue) {
852 ret = -ENODEV;
853 goto err_out_unlock;
854 }
855 queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
856 spin_lock_bh(&queue->lock);
857 queue->queue_maxlen = ntohl(*queue_maxlen);
858 spin_unlock_bh(&queue->lock);
859 }
860
861err_out_unlock:
862 rcu_read_unlock();
863 return ret;
864}
865
866static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
867 [NFQNL_MSG_PACKET] = { .call_rcu = nfqnl_recv_unsupp,
868 .attr_count = NFQA_MAX, },
869 [NFQNL_MSG_VERDICT] = { .call_rcu = nfqnl_recv_verdict,
870 .attr_count = NFQA_MAX,
871 .policy = nfqa_verdict_policy },
872 [NFQNL_MSG_CONFIG] = { .call = nfqnl_recv_config,
873 .attr_count = NFQA_CFG_MAX,
874 .policy = nfqa_cfg_policy },
875 [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
876 .attr_count = NFQA_MAX,
877 .policy = nfqa_verdict_batch_policy },
878};
879
880static const struct nfnetlink_subsystem nfqnl_subsys = {
881 .name = "nf_queue",
882 .subsys_id = NFNL_SUBSYS_QUEUE,
883 .cb_count = NFQNL_MSG_MAX,
884 .cb = nfqnl_cb,
885};
886
887#ifdef CONFIG_PROC_FS
888struct iter_state {
889 unsigned int bucket;
890};
891
892static struct hlist_node *get_first(struct seq_file *seq)
893{
894 struct iter_state *st = seq->private;
895
896 if (!st)
897 return NULL;
898
899 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
900 if (!hlist_empty(&instance_table[st->bucket]))
901 return instance_table[st->bucket].first;
902 }
903 return NULL;
904}
905
906static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
907{
908 struct iter_state *st = seq->private;
909
910 h = h->next;
911 while (!h) {
912 if (++st->bucket >= INSTANCE_BUCKETS)
913 return NULL;
914
915 h = instance_table[st->bucket].first;
916 }
917 return h;
918}
919
920static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
921{
922 struct hlist_node *head;
923 head = get_first(seq);
924
925 if (head)
926 while (pos && (head = get_next(seq, head)))
927 pos--;
928 return pos ? NULL : head;
929}
930
931static void *seq_start(struct seq_file *seq, loff_t *pos)
932 __acquires(instances_lock)
933{
934 spin_lock(&instances_lock);
935 return get_idx(seq, *pos);
936}
937
938static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
939{
940 (*pos)++;
941 return get_next(s, v);
942}
943
944static void seq_stop(struct seq_file *s, void *v)
945 __releases(instances_lock)
946{
947 spin_unlock(&instances_lock);
948}
949
950static int seq_show(struct seq_file *s, void *v)
951{
952 const struct nfqnl_instance *inst = v;
953
954 return seq_printf(s, "%5d %6d %5d %1d %5d %5d %5d %8d %2d\n",
955 inst->queue_num,
956 inst->peer_pid, inst->queue_total,
957 inst->copy_mode, inst->copy_range,
958 inst->queue_dropped, inst->queue_user_dropped,
959 inst->id_sequence, 1);
960}
961
962static const struct seq_operations nfqnl_seq_ops = {
963 .start = seq_start,
964 .next = seq_next,
965 .stop = seq_stop,
966 .show = seq_show,
967};
968
969static int nfqnl_open(struct inode *inode, struct file *file)
970{
971 return seq_open_private(file, &nfqnl_seq_ops,
972 sizeof(struct iter_state));
973}
974
975static const struct file_operations nfqnl_file_ops = {
976 .owner = THIS_MODULE,
977 .open = nfqnl_open,
978 .read = seq_read,
979 .llseek = seq_lseek,
980 .release = seq_release_private,
981};
982
983#endif /* PROC_FS */
984
985static int __init nfnetlink_queue_init(void)
986{
987 int i, status = -ENOMEM;
988
989 for (i = 0; i < INSTANCE_BUCKETS; i++)
990 INIT_HLIST_HEAD(&instance_table[i]);
991
992 netlink_register_notifier(&nfqnl_rtnl_notifier);
993 status = nfnetlink_subsys_register(&nfqnl_subsys);
994 if (status < 0) {
995 printk(KERN_ERR "nf_queue: failed to create netlink socket\n");
996 goto cleanup_netlink_notifier;
997 }
998
999#ifdef CONFIG_PROC_FS
1000 if (!proc_create("nfnetlink_queue", 0440,
1001 proc_net_netfilter, &nfqnl_file_ops))
1002 goto cleanup_subsys;
1003#endif
1004
1005 register_netdevice_notifier(&nfqnl_dev_notifier);
1006 return status;
1007
1008#ifdef CONFIG_PROC_FS
1009cleanup_subsys:
1010 nfnetlink_subsys_unregister(&nfqnl_subsys);
1011#endif
1012cleanup_netlink_notifier:
1013 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1014 return status;
1015}
1016
1017static void __exit nfnetlink_queue_fini(void)
1018{
1019 nf_unregister_queue_handlers(&nfqh);
1020 unregister_netdevice_notifier(&nfqnl_dev_notifier);
1021#ifdef CONFIG_PROC_FS
1022 remove_proc_entry("nfnetlink_queue", proc_net_netfilter);
1023#endif
1024 nfnetlink_subsys_unregister(&nfqnl_subsys);
1025 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1026
1027 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1028}
1029
1030MODULE_DESCRIPTION("netfilter packet queue handler");
1031MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1032MODULE_LICENSE("GPL");
1033MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1034
1035module_init(nfnetlink_queue_init);
1036module_exit(nfnetlink_queue_fini);