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// 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 <linux/cgroup-defs.h>
33#include <net/gso.h>
34#include <net/sock.h>
35#include <net/tcp_states.h>
36#include <net/netfilter/nf_queue.h>
37#include <net/netns/generic.h>
38
39#include <linux/atomic.h>
40
41#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
42#include "../bridge/br_private.h"
43#endif
44
45#if IS_ENABLED(CONFIG_NF_CONNTRACK)
46#include <net/netfilter/nf_conntrack.h>
47#endif
48
49#define NFQNL_QMAX_DEFAULT 1024
50
51/* We're using struct nlattr which has 16bit nla_len. Note that nla_len
52 * includes the header length. Thus, the maximum packet length that we
53 * support is 65531 bytes. We send truncated packets if the specified length
54 * is larger than that. Userspace can check for presence of NFQA_CAP_LEN
55 * attribute to detect truncation.
56 */
57#define NFQNL_MAX_COPY_RANGE (0xffff - NLA_HDRLEN)
58
59struct nfqnl_instance {
60 struct hlist_node hlist; /* global list of queues */
61 struct rcu_head rcu;
62
63 u32 peer_portid;
64 unsigned int queue_maxlen;
65 unsigned int copy_range;
66 unsigned int queue_dropped;
67 unsigned int queue_user_dropped;
68
69
70 u_int16_t queue_num; /* number of this queue */
71 u_int8_t copy_mode;
72 u_int32_t flags; /* Set using NFQA_CFG_FLAGS */
73/*
74 * Following fields are dirtied for each queued packet,
75 * keep them in same cache line if possible.
76 */
77 spinlock_t lock ____cacheline_aligned_in_smp;
78 unsigned int queue_total;
79 unsigned int id_sequence; /* 'sequence' of pkt ids */
80 struct list_head queue_list; /* packets in queue */
81};
82
83typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
84
85static unsigned int nfnl_queue_net_id __read_mostly;
86
87#define INSTANCE_BUCKETS 16
88struct nfnl_queue_net {
89 spinlock_t instances_lock;
90 struct hlist_head instance_table[INSTANCE_BUCKETS];
91};
92
93static struct nfnl_queue_net *nfnl_queue_pernet(struct net *net)
94{
95 return net_generic(net, nfnl_queue_net_id);
96}
97
98static inline u_int8_t instance_hashfn(u_int16_t queue_num)
99{
100 return ((queue_num >> 8) ^ queue_num) % INSTANCE_BUCKETS;
101}
102
103static struct nfqnl_instance *
104instance_lookup(struct nfnl_queue_net *q, u_int16_t queue_num)
105{
106 struct hlist_head *head;
107 struct nfqnl_instance *inst;
108
109 head = &q->instance_table[instance_hashfn(queue_num)];
110 hlist_for_each_entry_rcu(inst, head, hlist) {
111 if (inst->queue_num == queue_num)
112 return inst;
113 }
114 return NULL;
115}
116
117static struct nfqnl_instance *
118instance_create(struct nfnl_queue_net *q, u_int16_t queue_num, u32 portid)
119{
120 struct nfqnl_instance *inst;
121 unsigned int h;
122 int err;
123
124 spin_lock(&q->instances_lock);
125 if (instance_lookup(q, queue_num)) {
126 err = -EEXIST;
127 goto out_unlock;
128 }
129
130 inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
131 if (!inst) {
132 err = -ENOMEM;
133 goto out_unlock;
134 }
135
136 inst->queue_num = queue_num;
137 inst->peer_portid = portid;
138 inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
139 inst->copy_range = NFQNL_MAX_COPY_RANGE;
140 inst->copy_mode = NFQNL_COPY_NONE;
141 spin_lock_init(&inst->lock);
142 INIT_LIST_HEAD(&inst->queue_list);
143
144 if (!try_module_get(THIS_MODULE)) {
145 err = -EAGAIN;
146 goto out_free;
147 }
148
149 h = instance_hashfn(queue_num);
150 hlist_add_head_rcu(&inst->hlist, &q->instance_table[h]);
151
152 spin_unlock(&q->instances_lock);
153
154 return inst;
155
156out_free:
157 kfree(inst);
158out_unlock:
159 spin_unlock(&q->instances_lock);
160 return ERR_PTR(err);
161}
162
163static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
164 unsigned long data);
165
166static void
167instance_destroy_rcu(struct rcu_head *head)
168{
169 struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
170 rcu);
171
172 rcu_read_lock();
173 nfqnl_flush(inst, NULL, 0);
174 rcu_read_unlock();
175 kfree(inst);
176 module_put(THIS_MODULE);
177}
178
179static void
180__instance_destroy(struct nfqnl_instance *inst)
181{
182 hlist_del_rcu(&inst->hlist);
183 call_rcu(&inst->rcu, instance_destroy_rcu);
184}
185
186static void
187instance_destroy(struct nfnl_queue_net *q, struct nfqnl_instance *inst)
188{
189 spin_lock(&q->instances_lock);
190 __instance_destroy(inst);
191 spin_unlock(&q->instances_lock);
192}
193
194static inline void
195__enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
196{
197 list_add_tail(&entry->list, &queue->queue_list);
198 queue->queue_total++;
199}
200
201static void
202__dequeue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
203{
204 list_del(&entry->list);
205 queue->queue_total--;
206}
207
208static struct nf_queue_entry *
209find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
210{
211 struct nf_queue_entry *entry = NULL, *i;
212
213 spin_lock_bh(&queue->lock);
214
215 list_for_each_entry(i, &queue->queue_list, list) {
216 if (i->id == id) {
217 entry = i;
218 break;
219 }
220 }
221
222 if (entry)
223 __dequeue_entry(queue, entry);
224
225 spin_unlock_bh(&queue->lock);
226
227 return entry;
228}
229
230static unsigned int nf_iterate(struct sk_buff *skb,
231 struct nf_hook_state *state,
232 const struct nf_hook_entries *hooks,
233 unsigned int *index)
234{
235 const struct nf_hook_entry *hook;
236 unsigned int verdict, i = *index;
237
238 while (i < hooks->num_hook_entries) {
239 hook = &hooks->hooks[i];
240repeat:
241 verdict = nf_hook_entry_hookfn(hook, skb, state);
242 if (verdict != NF_ACCEPT) {
243 *index = i;
244 if (verdict != NF_REPEAT)
245 return verdict;
246 goto repeat;
247 }
248 i++;
249 }
250
251 *index = i;
252 return NF_ACCEPT;
253}
254
255static struct nf_hook_entries *nf_hook_entries_head(const struct net *net, u8 pf, u8 hooknum)
256{
257 switch (pf) {
258#ifdef CONFIG_NETFILTER_FAMILY_BRIDGE
259 case NFPROTO_BRIDGE:
260 return rcu_dereference(net->nf.hooks_bridge[hooknum]);
261#endif
262 case NFPROTO_IPV4:
263 return rcu_dereference(net->nf.hooks_ipv4[hooknum]);
264 case NFPROTO_IPV6:
265 return rcu_dereference(net->nf.hooks_ipv6[hooknum]);
266 default:
267 WARN_ON_ONCE(1);
268 return NULL;
269 }
270
271 return NULL;
272}
273
274static int nf_ip_reroute(struct sk_buff *skb, const struct nf_queue_entry *entry)
275{
276#ifdef CONFIG_INET
277 const struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
278
279 if (entry->state.hook == NF_INET_LOCAL_OUT) {
280 const struct iphdr *iph = ip_hdr(skb);
281
282 if (!(iph->tos == rt_info->tos &&
283 skb->mark == rt_info->mark &&
284 iph->daddr == rt_info->daddr &&
285 iph->saddr == rt_info->saddr))
286 return ip_route_me_harder(entry->state.net, entry->state.sk,
287 skb, RTN_UNSPEC);
288 }
289#endif
290 return 0;
291}
292
293static int nf_reroute(struct sk_buff *skb, struct nf_queue_entry *entry)
294{
295 const struct nf_ipv6_ops *v6ops;
296 int ret = 0;
297
298 switch (entry->state.pf) {
299 case AF_INET:
300 ret = nf_ip_reroute(skb, entry);
301 break;
302 case AF_INET6:
303 v6ops = rcu_dereference(nf_ipv6_ops);
304 if (v6ops)
305 ret = v6ops->reroute(skb, entry);
306 break;
307 }
308 return ret;
309}
310
311/* caller must hold rcu read-side lock */
312static void nf_reinject(struct nf_queue_entry *entry, unsigned int verdict)
313{
314 const struct nf_hook_entry *hook_entry;
315 const struct nf_hook_entries *hooks;
316 struct sk_buff *skb = entry->skb;
317 const struct net *net;
318 unsigned int i;
319 int err;
320 u8 pf;
321
322 net = entry->state.net;
323 pf = entry->state.pf;
324
325 hooks = nf_hook_entries_head(net, pf, entry->state.hook);
326
327 i = entry->hook_index;
328 if (WARN_ON_ONCE(!hooks || i >= hooks->num_hook_entries)) {
329 kfree_skb_reason(skb, SKB_DROP_REASON_NETFILTER_DROP);
330 nf_queue_entry_free(entry);
331 return;
332 }
333
334 hook_entry = &hooks->hooks[i];
335
336 /* Continue traversal iff userspace said ok... */
337 if (verdict == NF_REPEAT)
338 verdict = nf_hook_entry_hookfn(hook_entry, skb, &entry->state);
339
340 if (verdict == NF_ACCEPT) {
341 if (nf_reroute(skb, entry) < 0)
342 verdict = NF_DROP;
343 }
344
345 if (verdict == NF_ACCEPT) {
346next_hook:
347 ++i;
348 verdict = nf_iterate(skb, &entry->state, hooks, &i);
349 }
350
351 switch (verdict & NF_VERDICT_MASK) {
352 case NF_ACCEPT:
353 case NF_STOP:
354 local_bh_disable();
355 entry->state.okfn(entry->state.net, entry->state.sk, skb);
356 local_bh_enable();
357 break;
358 case NF_QUEUE:
359 err = nf_queue(skb, &entry->state, i, verdict);
360 if (err == 1)
361 goto next_hook;
362 break;
363 case NF_STOLEN:
364 break;
365 default:
366 kfree_skb(skb);
367 }
368
369 nf_queue_entry_free(entry);
370}
371
372static void nfqnl_reinject(struct nf_queue_entry *entry, unsigned int verdict)
373{
374 const struct nf_ct_hook *ct_hook;
375
376 if (verdict == NF_ACCEPT ||
377 verdict == NF_REPEAT ||
378 verdict == NF_STOP) {
379 unsigned int ct_verdict = verdict;
380
381 rcu_read_lock();
382 ct_hook = rcu_dereference(nf_ct_hook);
383 if (ct_hook)
384 ct_verdict = ct_hook->update(entry->state.net, entry->skb);
385 rcu_read_unlock();
386
387 switch (ct_verdict & NF_VERDICT_MASK) {
388 case NF_ACCEPT:
389 /* follow userspace verdict, could be REPEAT */
390 break;
391 case NF_STOLEN:
392 nf_queue_entry_free(entry);
393 return;
394 default:
395 verdict = ct_verdict & NF_VERDICT_MASK;
396 break;
397 }
398 }
399 nf_reinject(entry, verdict);
400}
401
402static void
403nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
404{
405 struct nf_queue_entry *entry, *next;
406
407 spin_lock_bh(&queue->lock);
408 list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
409 if (!cmpfn || cmpfn(entry, data)) {
410 list_del(&entry->list);
411 queue->queue_total--;
412 nfqnl_reinject(entry, NF_DROP);
413 }
414 }
415 spin_unlock_bh(&queue->lock);
416}
417
418static int
419nfqnl_put_packet_info(struct sk_buff *nlskb, struct sk_buff *packet,
420 bool csum_verify)
421{
422 __u32 flags = 0;
423
424 if (packet->ip_summed == CHECKSUM_PARTIAL)
425 flags = NFQA_SKB_CSUMNOTREADY;
426 else if (csum_verify)
427 flags = NFQA_SKB_CSUM_NOTVERIFIED;
428
429 if (skb_is_gso(packet))
430 flags |= NFQA_SKB_GSO;
431
432 return flags ? nla_put_be32(nlskb, NFQA_SKB_INFO, htonl(flags)) : 0;
433}
434
435static int nfqnl_put_sk_uidgid(struct sk_buff *skb, struct sock *sk)
436{
437 const struct cred *cred;
438
439 if (!sk_fullsock(sk))
440 return 0;
441
442 read_lock_bh(&sk->sk_callback_lock);
443 if (sk->sk_socket && sk->sk_socket->file) {
444 cred = sk->sk_socket->file->f_cred;
445 if (nla_put_be32(skb, NFQA_UID,
446 htonl(from_kuid_munged(&init_user_ns, cred->fsuid))))
447 goto nla_put_failure;
448 if (nla_put_be32(skb, NFQA_GID,
449 htonl(from_kgid_munged(&init_user_ns, cred->fsgid))))
450 goto nla_put_failure;
451 }
452 read_unlock_bh(&sk->sk_callback_lock);
453 return 0;
454
455nla_put_failure:
456 read_unlock_bh(&sk->sk_callback_lock);
457 return -1;
458}
459
460static int nfqnl_put_sk_classid(struct sk_buff *skb, struct sock *sk)
461{
462#if IS_ENABLED(CONFIG_CGROUP_NET_CLASSID)
463 if (sk && sk_fullsock(sk)) {
464 u32 classid = sock_cgroup_classid(&sk->sk_cgrp_data);
465
466 if (classid && nla_put_be32(skb, NFQA_CGROUP_CLASSID, htonl(classid)))
467 return -1;
468 }
469#endif
470 return 0;
471}
472
473static u32 nfqnl_get_sk_secctx(struct sk_buff *skb, char **secdata)
474{
475 u32 seclen = 0;
476#if IS_ENABLED(CONFIG_NETWORK_SECMARK)
477 if (!skb || !sk_fullsock(skb->sk))
478 return 0;
479
480 read_lock_bh(&skb->sk->sk_callback_lock);
481
482 if (skb->secmark)
483 security_secid_to_secctx(skb->secmark, secdata, &seclen);
484
485 read_unlock_bh(&skb->sk->sk_callback_lock);
486#endif
487 return seclen;
488}
489
490static u32 nfqnl_get_bridge_size(struct nf_queue_entry *entry)
491{
492 struct sk_buff *entskb = entry->skb;
493 u32 nlalen = 0;
494
495 if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb))
496 return 0;
497
498 if (skb_vlan_tag_present(entskb))
499 nlalen += nla_total_size(nla_total_size(sizeof(__be16)) +
500 nla_total_size(sizeof(__be16)));
501
502 if (entskb->network_header > entskb->mac_header)
503 nlalen += nla_total_size((entskb->network_header -
504 entskb->mac_header));
505
506 return nlalen;
507}
508
509static int nfqnl_put_bridge(struct nf_queue_entry *entry, struct sk_buff *skb)
510{
511 struct sk_buff *entskb = entry->skb;
512
513 if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb))
514 return 0;
515
516 if (skb_vlan_tag_present(entskb)) {
517 struct nlattr *nest;
518
519 nest = nla_nest_start(skb, NFQA_VLAN);
520 if (!nest)
521 goto nla_put_failure;
522
523 if (nla_put_be16(skb, NFQA_VLAN_TCI, htons(entskb->vlan_tci)) ||
524 nla_put_be16(skb, NFQA_VLAN_PROTO, entskb->vlan_proto))
525 goto nla_put_failure;
526
527 nla_nest_end(skb, nest);
528 }
529
530 if (entskb->mac_header < entskb->network_header) {
531 int len = (int)(entskb->network_header - entskb->mac_header);
532
533 if (nla_put(skb, NFQA_L2HDR, len, skb_mac_header(entskb)))
534 goto nla_put_failure;
535 }
536
537 return 0;
538
539nla_put_failure:
540 return -1;
541}
542
543static struct sk_buff *
544nfqnl_build_packet_message(struct net *net, struct nfqnl_instance *queue,
545 struct nf_queue_entry *entry,
546 __be32 **packet_id_ptr)
547{
548 size_t size;
549 size_t data_len = 0, cap_len = 0;
550 unsigned int hlen = 0;
551 struct sk_buff *skb;
552 struct nlattr *nla;
553 struct nfqnl_msg_packet_hdr *pmsg;
554 struct nlmsghdr *nlh;
555 struct sk_buff *entskb = entry->skb;
556 struct net_device *indev;
557 struct net_device *outdev;
558 struct nf_conn *ct = NULL;
559 enum ip_conntrack_info ctinfo = 0;
560 const struct nfnl_ct_hook *nfnl_ct;
561 bool csum_verify;
562 char *secdata = NULL;
563 u32 seclen = 0;
564 ktime_t tstamp;
565
566 size = nlmsg_total_size(sizeof(struct nfgenmsg))
567 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
568 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
569 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
570#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
571 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
572 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
573#endif
574 + nla_total_size(sizeof(u_int32_t)) /* mark */
575 + nla_total_size(sizeof(u_int32_t)) /* priority */
576 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
577 + nla_total_size(sizeof(u_int32_t)) /* skbinfo */
578#if IS_ENABLED(CONFIG_CGROUP_NET_CLASSID)
579 + nla_total_size(sizeof(u_int32_t)) /* classid */
580#endif
581 + nla_total_size(sizeof(u_int32_t)); /* cap_len */
582
583 tstamp = skb_tstamp_cond(entskb, false);
584 if (tstamp)
585 size += nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
586
587 size += nfqnl_get_bridge_size(entry);
588
589 if (entry->state.hook <= NF_INET_FORWARD ||
590 (entry->state.hook == NF_INET_POST_ROUTING && entskb->sk == NULL))
591 csum_verify = !skb_csum_unnecessary(entskb);
592 else
593 csum_verify = false;
594
595 outdev = entry->state.out;
596
597 switch ((enum nfqnl_config_mode)READ_ONCE(queue->copy_mode)) {
598 case NFQNL_COPY_META:
599 case NFQNL_COPY_NONE:
600 break;
601
602 case NFQNL_COPY_PACKET:
603 if (!(queue->flags & NFQA_CFG_F_GSO) &&
604 entskb->ip_summed == CHECKSUM_PARTIAL &&
605 skb_checksum_help(entskb))
606 return NULL;
607
608 data_len = READ_ONCE(queue->copy_range);
609 if (data_len > entskb->len)
610 data_len = entskb->len;
611
612 hlen = skb_zerocopy_headlen(entskb);
613 hlen = min_t(unsigned int, hlen, data_len);
614 size += sizeof(struct nlattr) + hlen;
615 cap_len = entskb->len;
616 break;
617 }
618
619 nfnl_ct = rcu_dereference(nfnl_ct_hook);
620
621#if IS_ENABLED(CONFIG_NF_CONNTRACK)
622 if (queue->flags & NFQA_CFG_F_CONNTRACK) {
623 if (nfnl_ct != NULL) {
624 ct = nf_ct_get(entskb, &ctinfo);
625 if (ct != NULL)
626 size += nfnl_ct->build_size(ct);
627 }
628 }
629#endif
630
631 if (queue->flags & NFQA_CFG_F_UID_GID) {
632 size += (nla_total_size(sizeof(u_int32_t)) /* uid */
633 + nla_total_size(sizeof(u_int32_t))); /* gid */
634 }
635
636 if ((queue->flags & NFQA_CFG_F_SECCTX) && entskb->sk) {
637 seclen = nfqnl_get_sk_secctx(entskb, &secdata);
638 if (seclen)
639 size += nla_total_size(seclen);
640 }
641
642 skb = alloc_skb(size, GFP_ATOMIC);
643 if (!skb) {
644 skb_tx_error(entskb);
645 goto nlmsg_failure;
646 }
647
648 nlh = nfnl_msg_put(skb, 0, 0,
649 nfnl_msg_type(NFNL_SUBSYS_QUEUE, NFQNL_MSG_PACKET),
650 0, entry->state.pf, NFNETLINK_V0,
651 htons(queue->queue_num));
652 if (!nlh) {
653 skb_tx_error(entskb);
654 kfree_skb(skb);
655 goto nlmsg_failure;
656 }
657
658 nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
659 pmsg = nla_data(nla);
660 pmsg->hw_protocol = entskb->protocol;
661 pmsg->hook = entry->state.hook;
662 *packet_id_ptr = &pmsg->packet_id;
663
664 indev = entry->state.in;
665 if (indev) {
666#if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
667 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
668 goto nla_put_failure;
669#else
670 if (entry->state.pf == PF_BRIDGE) {
671 /* Case 1: indev is physical input device, we need to
672 * look for bridge group (when called from
673 * netfilter_bridge) */
674 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
675 htonl(indev->ifindex)) ||
676 /* this is the bridge group "brX" */
677 /* rcu_read_lock()ed by __nf_queue */
678 nla_put_be32(skb, NFQA_IFINDEX_INDEV,
679 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
680 goto nla_put_failure;
681 } else {
682 int physinif;
683
684 /* Case 2: indev is bridge group, we need to look for
685 * physical device (when called from ipv4) */
686 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
687 htonl(indev->ifindex)))
688 goto nla_put_failure;
689
690 physinif = nf_bridge_get_physinif(entskb);
691 if (physinif &&
692 nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
693 htonl(physinif)))
694 goto nla_put_failure;
695 }
696#endif
697 }
698
699 if (outdev) {
700#if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
701 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
702 goto nla_put_failure;
703#else
704 if (entry->state.pf == PF_BRIDGE) {
705 /* Case 1: outdev is physical output device, we need to
706 * look for bridge group (when called from
707 * netfilter_bridge) */
708 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
709 htonl(outdev->ifindex)) ||
710 /* this is the bridge group "brX" */
711 /* rcu_read_lock()ed by __nf_queue */
712 nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
713 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
714 goto nla_put_failure;
715 } else {
716 int physoutif;
717
718 /* Case 2: outdev is bridge group, we need to look for
719 * physical output device (when called from ipv4) */
720 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
721 htonl(outdev->ifindex)))
722 goto nla_put_failure;
723
724 physoutif = nf_bridge_get_physoutif(entskb);
725 if (physoutif &&
726 nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
727 htonl(physoutif)))
728 goto nla_put_failure;
729 }
730#endif
731 }
732
733 if (entskb->mark &&
734 nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
735 goto nla_put_failure;
736
737 if (entskb->priority &&
738 nla_put_be32(skb, NFQA_PRIORITY, htonl(entskb->priority)))
739 goto nla_put_failure;
740
741 if (indev && entskb->dev &&
742 skb_mac_header_was_set(entskb) &&
743 skb_mac_header_len(entskb) != 0) {
744 struct nfqnl_msg_packet_hw phw;
745 int len;
746
747 memset(&phw, 0, sizeof(phw));
748 len = dev_parse_header(entskb, phw.hw_addr);
749 if (len) {
750 phw.hw_addrlen = htons(len);
751 if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
752 goto nla_put_failure;
753 }
754 }
755
756 if (nfqnl_put_bridge(entry, skb) < 0)
757 goto nla_put_failure;
758
759 if (entry->state.hook <= NF_INET_FORWARD && tstamp) {
760 struct nfqnl_msg_packet_timestamp ts;
761 struct timespec64 kts = ktime_to_timespec64(tstamp);
762
763 ts.sec = cpu_to_be64(kts.tv_sec);
764 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
765
766 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
767 goto nla_put_failure;
768 }
769
770 if ((queue->flags & NFQA_CFG_F_UID_GID) && entskb->sk &&
771 nfqnl_put_sk_uidgid(skb, entskb->sk) < 0)
772 goto nla_put_failure;
773
774 if (nfqnl_put_sk_classid(skb, entskb->sk) < 0)
775 goto nla_put_failure;
776
777 if (seclen && nla_put(skb, NFQA_SECCTX, seclen, secdata))
778 goto nla_put_failure;
779
780 if (ct && nfnl_ct->build(skb, ct, ctinfo, NFQA_CT, NFQA_CT_INFO) < 0)
781 goto nla_put_failure;
782
783 if (cap_len > data_len &&
784 nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
785 goto nla_put_failure;
786
787 if (nfqnl_put_packet_info(skb, entskb, csum_verify))
788 goto nla_put_failure;
789
790 if (data_len) {
791 struct nlattr *nla;
792
793 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
794 goto nla_put_failure;
795
796 nla = skb_put(skb, sizeof(*nla));
797 nla->nla_type = NFQA_PAYLOAD;
798 nla->nla_len = nla_attr_size(data_len);
799
800 if (skb_zerocopy(skb, entskb, data_len, hlen))
801 goto nla_put_failure;
802 }
803
804 nlh->nlmsg_len = skb->len;
805 if (seclen)
806 security_release_secctx(secdata, seclen);
807 return skb;
808
809nla_put_failure:
810 skb_tx_error(entskb);
811 kfree_skb(skb);
812 net_err_ratelimited("nf_queue: error creating packet message\n");
813nlmsg_failure:
814 if (seclen)
815 security_release_secctx(secdata, seclen);
816 return NULL;
817}
818
819static bool nf_ct_drop_unconfirmed(const struct nf_queue_entry *entry)
820{
821#if IS_ENABLED(CONFIG_NF_CONNTRACK)
822 static const unsigned long flags = IPS_CONFIRMED | IPS_DYING;
823 const struct nf_conn *ct = (void *)skb_nfct(entry->skb);
824
825 if (ct && ((ct->status & flags) == IPS_DYING))
826 return true;
827#endif
828 return false;
829}
830
831static int
832__nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
833 struct nf_queue_entry *entry)
834{
835 struct sk_buff *nskb;
836 int err = -ENOBUFS;
837 __be32 *packet_id_ptr;
838 int failopen = 0;
839
840 nskb = nfqnl_build_packet_message(net, queue, entry, &packet_id_ptr);
841 if (nskb == NULL) {
842 err = -ENOMEM;
843 goto err_out;
844 }
845 spin_lock_bh(&queue->lock);
846
847 if (nf_ct_drop_unconfirmed(entry))
848 goto err_out_free_nskb;
849
850 if (queue->queue_total >= queue->queue_maxlen) {
851 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
852 failopen = 1;
853 err = 0;
854 } else {
855 queue->queue_dropped++;
856 net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
857 queue->queue_total);
858 }
859 goto err_out_free_nskb;
860 }
861 entry->id = ++queue->id_sequence;
862 *packet_id_ptr = htonl(entry->id);
863
864 /* nfnetlink_unicast will either free the nskb or add it to a socket */
865 err = nfnetlink_unicast(nskb, net, queue->peer_portid);
866 if (err < 0) {
867 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
868 failopen = 1;
869 err = 0;
870 } else {
871 queue->queue_user_dropped++;
872 }
873 goto err_out_unlock;
874 }
875
876 __enqueue_entry(queue, entry);
877
878 spin_unlock_bh(&queue->lock);
879 return 0;
880
881err_out_free_nskb:
882 kfree_skb(nskb);
883err_out_unlock:
884 spin_unlock_bh(&queue->lock);
885 if (failopen)
886 nfqnl_reinject(entry, NF_ACCEPT);
887err_out:
888 return err;
889}
890
891static struct nf_queue_entry *
892nf_queue_entry_dup(struct nf_queue_entry *e)
893{
894 struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
895
896 if (!entry)
897 return NULL;
898
899 if (nf_queue_entry_get_refs(entry))
900 return entry;
901
902 kfree(entry);
903 return NULL;
904}
905
906#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
907/* When called from bridge netfilter, skb->data must point to MAC header
908 * before calling skb_gso_segment(). Else, original MAC header is lost
909 * and segmented skbs will be sent to wrong destination.
910 */
911static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
912{
913 if (nf_bridge_info_get(skb))
914 __skb_push(skb, skb->network_header - skb->mac_header);
915}
916
917static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
918{
919 if (nf_bridge_info_get(skb))
920 __skb_pull(skb, skb->network_header - skb->mac_header);
921}
922#else
923#define nf_bridge_adjust_skb_data(s) do {} while (0)
924#define nf_bridge_adjust_segmented_data(s) do {} while (0)
925#endif
926
927static int
928__nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
929 struct sk_buff *skb, struct nf_queue_entry *entry)
930{
931 int ret = -ENOMEM;
932 struct nf_queue_entry *entry_seg;
933
934 nf_bridge_adjust_segmented_data(skb);
935
936 if (skb->next == NULL) { /* last packet, no need to copy entry */
937 struct sk_buff *gso_skb = entry->skb;
938 entry->skb = skb;
939 ret = __nfqnl_enqueue_packet(net, queue, entry);
940 if (ret)
941 entry->skb = gso_skb;
942 return ret;
943 }
944
945 skb_mark_not_on_list(skb);
946
947 entry_seg = nf_queue_entry_dup(entry);
948 if (entry_seg) {
949 entry_seg->skb = skb;
950 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
951 if (ret)
952 nf_queue_entry_free(entry_seg);
953 }
954 return ret;
955}
956
957static int
958nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
959{
960 unsigned int queued;
961 struct nfqnl_instance *queue;
962 struct sk_buff *skb, *segs, *nskb;
963 int err = -ENOBUFS;
964 struct net *net = entry->state.net;
965 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
966
967 /* rcu_read_lock()ed by nf_hook_thresh */
968 queue = instance_lookup(q, queuenum);
969 if (!queue)
970 return -ESRCH;
971
972 if (queue->copy_mode == NFQNL_COPY_NONE)
973 return -EINVAL;
974
975 skb = entry->skb;
976
977 switch (entry->state.pf) {
978 case NFPROTO_IPV4:
979 skb->protocol = htons(ETH_P_IP);
980 break;
981 case NFPROTO_IPV6:
982 skb->protocol = htons(ETH_P_IPV6);
983 break;
984 }
985
986 if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(skb))
987 return __nfqnl_enqueue_packet(net, queue, entry);
988
989 nf_bridge_adjust_skb_data(skb);
990 segs = skb_gso_segment(skb, 0);
991 /* Does not use PTR_ERR to limit the number of error codes that can be
992 * returned by nf_queue. For instance, callers rely on -ESRCH to
993 * mean 'ignore this hook'.
994 */
995 if (IS_ERR_OR_NULL(segs))
996 goto out_err;
997 queued = 0;
998 err = 0;
999 skb_list_walk_safe(segs, segs, nskb) {
1000 if (err == 0)
1001 err = __nfqnl_enqueue_packet_gso(net, queue,
1002 segs, entry);
1003 if (err == 0)
1004 queued++;
1005 else
1006 kfree_skb(segs);
1007 }
1008
1009 if (queued) {
1010 if (err) /* some segments are already queued */
1011 nf_queue_entry_free(entry);
1012 kfree_skb(skb);
1013 return 0;
1014 }
1015 out_err:
1016 nf_bridge_adjust_segmented_data(skb);
1017 return err;
1018}
1019
1020static int
1021nfqnl_mangle(void *data, unsigned int data_len, struct nf_queue_entry *e, int diff)
1022{
1023 struct sk_buff *nskb;
1024
1025 if (diff < 0) {
1026 unsigned int min_len = skb_transport_offset(e->skb);
1027
1028 if (data_len < min_len)
1029 return -EINVAL;
1030
1031 if (pskb_trim(e->skb, data_len))
1032 return -ENOMEM;
1033 } else if (diff > 0) {
1034 if (data_len > 0xFFFF)
1035 return -EINVAL;
1036 if (diff > skb_tailroom(e->skb)) {
1037 nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
1038 diff, GFP_ATOMIC);
1039 if (!nskb)
1040 return -ENOMEM;
1041 kfree_skb(e->skb);
1042 e->skb = nskb;
1043 }
1044 skb_put(e->skb, diff);
1045 }
1046 if (skb_ensure_writable(e->skb, data_len))
1047 return -ENOMEM;
1048 skb_copy_to_linear_data(e->skb, data, data_len);
1049 e->skb->ip_summed = CHECKSUM_NONE;
1050 return 0;
1051}
1052
1053static int
1054nfqnl_set_mode(struct nfqnl_instance *queue,
1055 unsigned char mode, unsigned int range)
1056{
1057 int status = 0;
1058
1059 spin_lock_bh(&queue->lock);
1060 switch (mode) {
1061 case NFQNL_COPY_NONE:
1062 case NFQNL_COPY_META:
1063 queue->copy_mode = mode;
1064 queue->copy_range = 0;
1065 break;
1066
1067 case NFQNL_COPY_PACKET:
1068 queue->copy_mode = mode;
1069 if (range == 0 || range > NFQNL_MAX_COPY_RANGE)
1070 queue->copy_range = NFQNL_MAX_COPY_RANGE;
1071 else
1072 queue->copy_range = range;
1073 break;
1074
1075 default:
1076 status = -EINVAL;
1077
1078 }
1079 spin_unlock_bh(&queue->lock);
1080
1081 return status;
1082}
1083
1084static int
1085dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
1086{
1087#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1088 int physinif, physoutif;
1089
1090 physinif = nf_bridge_get_physinif(entry->skb);
1091 physoutif = nf_bridge_get_physoutif(entry->skb);
1092
1093 if (physinif == ifindex || physoutif == ifindex)
1094 return 1;
1095#endif
1096 if (entry->state.in)
1097 if (entry->state.in->ifindex == ifindex)
1098 return 1;
1099 if (entry->state.out)
1100 if (entry->state.out->ifindex == ifindex)
1101 return 1;
1102
1103 return 0;
1104}
1105
1106/* drop all packets with either indev or outdev == ifindex from all queue
1107 * instances */
1108static void
1109nfqnl_dev_drop(struct net *net, int ifindex)
1110{
1111 int i;
1112 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1113
1114 rcu_read_lock();
1115
1116 for (i = 0; i < INSTANCE_BUCKETS; i++) {
1117 struct nfqnl_instance *inst;
1118 struct hlist_head *head = &q->instance_table[i];
1119
1120 hlist_for_each_entry_rcu(inst, head, hlist)
1121 nfqnl_flush(inst, dev_cmp, ifindex);
1122 }
1123
1124 rcu_read_unlock();
1125}
1126
1127static int
1128nfqnl_rcv_dev_event(struct notifier_block *this,
1129 unsigned long event, void *ptr)
1130{
1131 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1132
1133 /* Drop any packets associated with the downed device */
1134 if (event == NETDEV_DOWN)
1135 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
1136 return NOTIFY_DONE;
1137}
1138
1139static struct notifier_block nfqnl_dev_notifier = {
1140 .notifier_call = nfqnl_rcv_dev_event,
1141};
1142
1143static void nfqnl_nf_hook_drop(struct net *net)
1144{
1145 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1146 int i;
1147
1148 /* This function is also called on net namespace error unwind,
1149 * when pernet_ops->init() failed and ->exit() functions of the
1150 * previous pernet_ops gets called.
1151 *
1152 * This may result in a call to nfqnl_nf_hook_drop() before
1153 * struct nfnl_queue_net was allocated.
1154 */
1155 if (!q)
1156 return;
1157
1158 for (i = 0; i < INSTANCE_BUCKETS; i++) {
1159 struct nfqnl_instance *inst;
1160 struct hlist_head *head = &q->instance_table[i];
1161
1162 hlist_for_each_entry_rcu(inst, head, hlist)
1163 nfqnl_flush(inst, NULL, 0);
1164 }
1165}
1166
1167static int
1168nfqnl_rcv_nl_event(struct notifier_block *this,
1169 unsigned long event, void *ptr)
1170{
1171 struct netlink_notify *n = ptr;
1172 struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
1173
1174 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
1175 int i;
1176
1177 /* destroy all instances for this portid */
1178 spin_lock(&q->instances_lock);
1179 for (i = 0; i < INSTANCE_BUCKETS; i++) {
1180 struct hlist_node *t2;
1181 struct nfqnl_instance *inst;
1182 struct hlist_head *head = &q->instance_table[i];
1183
1184 hlist_for_each_entry_safe(inst, t2, head, hlist) {
1185 if (n->portid == inst->peer_portid)
1186 __instance_destroy(inst);
1187 }
1188 }
1189 spin_unlock(&q->instances_lock);
1190 }
1191 return NOTIFY_DONE;
1192}
1193
1194static struct notifier_block nfqnl_rtnl_notifier = {
1195 .notifier_call = nfqnl_rcv_nl_event,
1196};
1197
1198static const struct nla_policy nfqa_vlan_policy[NFQA_VLAN_MAX + 1] = {
1199 [NFQA_VLAN_TCI] = { .type = NLA_U16},
1200 [NFQA_VLAN_PROTO] = { .type = NLA_U16},
1201};
1202
1203static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
1204 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
1205 [NFQA_MARK] = { .type = NLA_U32 },
1206 [NFQA_PAYLOAD] = { .type = NLA_UNSPEC },
1207 [NFQA_CT] = { .type = NLA_UNSPEC },
1208 [NFQA_EXP] = { .type = NLA_UNSPEC },
1209 [NFQA_VLAN] = { .type = NLA_NESTED },
1210 [NFQA_PRIORITY] = { .type = NLA_U32 },
1211};
1212
1213static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
1214 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
1215 [NFQA_MARK] = { .type = NLA_U32 },
1216 [NFQA_PRIORITY] = { .type = NLA_U32 },
1217};
1218
1219static struct nfqnl_instance *
1220verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, u32 nlportid)
1221{
1222 struct nfqnl_instance *queue;
1223
1224 queue = instance_lookup(q, queue_num);
1225 if (!queue)
1226 return ERR_PTR(-ENODEV);
1227
1228 if (queue->peer_portid != nlportid)
1229 return ERR_PTR(-EPERM);
1230
1231 return queue;
1232}
1233
1234static struct nfqnl_msg_verdict_hdr*
1235verdicthdr_get(const struct nlattr * const nfqa[])
1236{
1237 struct nfqnl_msg_verdict_hdr *vhdr;
1238 unsigned int verdict;
1239
1240 if (!nfqa[NFQA_VERDICT_HDR])
1241 return NULL;
1242
1243 vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
1244 verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
1245 if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
1246 return NULL;
1247 return vhdr;
1248}
1249
1250static int nfq_id_after(unsigned int id, unsigned int max)
1251{
1252 return (int)(id - max) > 0;
1253}
1254
1255static int nfqnl_recv_verdict_batch(struct sk_buff *skb,
1256 const struct nfnl_info *info,
1257 const struct nlattr * const nfqa[])
1258{
1259 struct nfnl_queue_net *q = nfnl_queue_pernet(info->net);
1260 u16 queue_num = ntohs(info->nfmsg->res_id);
1261 struct nf_queue_entry *entry, *tmp;
1262 struct nfqnl_msg_verdict_hdr *vhdr;
1263 struct nfqnl_instance *queue;
1264 unsigned int verdict, maxid;
1265 LIST_HEAD(batch_list);
1266
1267 queue = verdict_instance_lookup(q, queue_num,
1268 NETLINK_CB(skb).portid);
1269 if (IS_ERR(queue))
1270 return PTR_ERR(queue);
1271
1272 vhdr = verdicthdr_get(nfqa);
1273 if (!vhdr)
1274 return -EINVAL;
1275
1276 verdict = ntohl(vhdr->verdict);
1277 maxid = ntohl(vhdr->id);
1278
1279 spin_lock_bh(&queue->lock);
1280
1281 list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
1282 if (nfq_id_after(entry->id, maxid))
1283 break;
1284 __dequeue_entry(queue, entry);
1285 list_add_tail(&entry->list, &batch_list);
1286 }
1287
1288 spin_unlock_bh(&queue->lock);
1289
1290 if (list_empty(&batch_list))
1291 return -ENOENT;
1292
1293 list_for_each_entry_safe(entry, tmp, &batch_list, list) {
1294 if (nfqa[NFQA_MARK])
1295 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1296
1297 if (nfqa[NFQA_PRIORITY])
1298 entry->skb->priority = ntohl(nla_get_be32(nfqa[NFQA_PRIORITY]));
1299
1300 nfqnl_reinject(entry, verdict);
1301 }
1302 return 0;
1303}
1304
1305static struct nf_conn *nfqnl_ct_parse(const struct nfnl_ct_hook *nfnl_ct,
1306 const struct nlmsghdr *nlh,
1307 const struct nlattr * const nfqa[],
1308 struct nf_queue_entry *entry,
1309 enum ip_conntrack_info *ctinfo)
1310{
1311#if IS_ENABLED(CONFIG_NF_CONNTRACK)
1312 struct nf_conn *ct;
1313
1314 ct = nf_ct_get(entry->skb, ctinfo);
1315 if (ct == NULL)
1316 return NULL;
1317
1318 if (nfnl_ct->parse(nfqa[NFQA_CT], ct) < 0)
1319 return NULL;
1320
1321 if (nfqa[NFQA_EXP])
1322 nfnl_ct->attach_expect(nfqa[NFQA_EXP], ct,
1323 NETLINK_CB(entry->skb).portid,
1324 nlmsg_report(nlh));
1325 return ct;
1326#else
1327 return NULL;
1328#endif
1329}
1330
1331static int nfqa_parse_bridge(struct nf_queue_entry *entry,
1332 const struct nlattr * const nfqa[])
1333{
1334 if (nfqa[NFQA_VLAN]) {
1335 struct nlattr *tb[NFQA_VLAN_MAX + 1];
1336 int err;
1337
1338 err = nla_parse_nested_deprecated(tb, NFQA_VLAN_MAX,
1339 nfqa[NFQA_VLAN],
1340 nfqa_vlan_policy, NULL);
1341 if (err < 0)
1342 return err;
1343
1344 if (!tb[NFQA_VLAN_TCI] || !tb[NFQA_VLAN_PROTO])
1345 return -EINVAL;
1346
1347 __vlan_hwaccel_put_tag(entry->skb,
1348 nla_get_be16(tb[NFQA_VLAN_PROTO]),
1349 ntohs(nla_get_be16(tb[NFQA_VLAN_TCI])));
1350 }
1351
1352 if (nfqa[NFQA_L2HDR]) {
1353 int mac_header_len = entry->skb->network_header -
1354 entry->skb->mac_header;
1355
1356 if (mac_header_len != nla_len(nfqa[NFQA_L2HDR]))
1357 return -EINVAL;
1358 else if (mac_header_len > 0)
1359 memcpy(skb_mac_header(entry->skb),
1360 nla_data(nfqa[NFQA_L2HDR]),
1361 mac_header_len);
1362 }
1363
1364 return 0;
1365}
1366
1367static int nfqnl_recv_verdict(struct sk_buff *skb, const struct nfnl_info *info,
1368 const struct nlattr * const nfqa[])
1369{
1370 struct nfnl_queue_net *q = nfnl_queue_pernet(info->net);
1371 u_int16_t queue_num = ntohs(info->nfmsg->res_id);
1372 const struct nfnl_ct_hook *nfnl_ct;
1373 struct nfqnl_msg_verdict_hdr *vhdr;
1374 enum ip_conntrack_info ctinfo;
1375 struct nfqnl_instance *queue;
1376 struct nf_queue_entry *entry;
1377 struct nf_conn *ct = NULL;
1378 unsigned int verdict;
1379 int err;
1380
1381 queue = verdict_instance_lookup(q, queue_num,
1382 NETLINK_CB(skb).portid);
1383 if (IS_ERR(queue))
1384 return PTR_ERR(queue);
1385
1386 vhdr = verdicthdr_get(nfqa);
1387 if (!vhdr)
1388 return -EINVAL;
1389
1390 verdict = ntohl(vhdr->verdict);
1391
1392 entry = find_dequeue_entry(queue, ntohl(vhdr->id));
1393 if (entry == NULL)
1394 return -ENOENT;
1395
1396 /* rcu lock already held from nfnl->call_rcu. */
1397 nfnl_ct = rcu_dereference(nfnl_ct_hook);
1398
1399 if (nfqa[NFQA_CT]) {
1400 if (nfnl_ct != NULL)
1401 ct = nfqnl_ct_parse(nfnl_ct, info->nlh, nfqa, entry,
1402 &ctinfo);
1403 }
1404
1405 if (entry->state.pf == PF_BRIDGE) {
1406 err = nfqa_parse_bridge(entry, nfqa);
1407 if (err < 0)
1408 return err;
1409 }
1410
1411 if (nfqa[NFQA_PAYLOAD]) {
1412 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
1413 int diff = payload_len - entry->skb->len;
1414
1415 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
1416 payload_len, entry, diff) < 0)
1417 verdict = NF_DROP;
1418
1419 if (ct && diff)
1420 nfnl_ct->seq_adjust(entry->skb, ct, ctinfo, diff);
1421 }
1422
1423 if (nfqa[NFQA_MARK])
1424 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1425
1426 if (nfqa[NFQA_PRIORITY])
1427 entry->skb->priority = ntohl(nla_get_be32(nfqa[NFQA_PRIORITY]));
1428
1429 nfqnl_reinject(entry, verdict);
1430 return 0;
1431}
1432
1433static int nfqnl_recv_unsupp(struct sk_buff *skb, const struct nfnl_info *info,
1434 const struct nlattr * const cda[])
1435{
1436 return -ENOTSUPP;
1437}
1438
1439static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1440 [NFQA_CFG_CMD] = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1441 [NFQA_CFG_PARAMS] = { .len = sizeof(struct nfqnl_msg_config_params) },
1442 [NFQA_CFG_QUEUE_MAXLEN] = { .type = NLA_U32 },
1443 [NFQA_CFG_MASK] = { .type = NLA_U32 },
1444 [NFQA_CFG_FLAGS] = { .type = NLA_U32 },
1445};
1446
1447static const struct nf_queue_handler nfqh = {
1448 .outfn = nfqnl_enqueue_packet,
1449 .nf_hook_drop = nfqnl_nf_hook_drop,
1450};
1451
1452static int nfqnl_recv_config(struct sk_buff *skb, const struct nfnl_info *info,
1453 const struct nlattr * const nfqa[])
1454{
1455 struct nfnl_queue_net *q = nfnl_queue_pernet(info->net);
1456 u_int16_t queue_num = ntohs(info->nfmsg->res_id);
1457 struct nfqnl_msg_config_cmd *cmd = NULL;
1458 struct nfqnl_instance *queue;
1459 __u32 flags = 0, mask = 0;
1460 int ret = 0;
1461
1462 if (nfqa[NFQA_CFG_CMD]) {
1463 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1464
1465 /* Obsolete commands without queue context */
1466 switch (cmd->command) {
1467 case NFQNL_CFG_CMD_PF_BIND: return 0;
1468 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1469 }
1470 }
1471
1472 /* Check if we support these flags in first place, dependencies should
1473 * be there too not to break atomicity.
1474 */
1475 if (nfqa[NFQA_CFG_FLAGS]) {
1476 if (!nfqa[NFQA_CFG_MASK]) {
1477 /* A mask is needed to specify which flags are being
1478 * changed.
1479 */
1480 return -EINVAL;
1481 }
1482
1483 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1484 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1485
1486 if (flags >= NFQA_CFG_F_MAX)
1487 return -EOPNOTSUPP;
1488
1489#if !IS_ENABLED(CONFIG_NETWORK_SECMARK)
1490 if (flags & mask & NFQA_CFG_F_SECCTX)
1491 return -EOPNOTSUPP;
1492#endif
1493 if ((flags & mask & NFQA_CFG_F_CONNTRACK) &&
1494 !rcu_access_pointer(nfnl_ct_hook)) {
1495#ifdef CONFIG_MODULES
1496 nfnl_unlock(NFNL_SUBSYS_QUEUE);
1497 request_module("ip_conntrack_netlink");
1498 nfnl_lock(NFNL_SUBSYS_QUEUE);
1499 if (rcu_access_pointer(nfnl_ct_hook))
1500 return -EAGAIN;
1501#endif
1502 return -EOPNOTSUPP;
1503 }
1504 }
1505
1506 rcu_read_lock();
1507 queue = instance_lookup(q, queue_num);
1508 if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1509 ret = -EPERM;
1510 goto err_out_unlock;
1511 }
1512
1513 if (cmd != NULL) {
1514 switch (cmd->command) {
1515 case NFQNL_CFG_CMD_BIND:
1516 if (queue) {
1517 ret = -EBUSY;
1518 goto err_out_unlock;
1519 }
1520 queue = instance_create(q, queue_num,
1521 NETLINK_CB(skb).portid);
1522 if (IS_ERR(queue)) {
1523 ret = PTR_ERR(queue);
1524 goto err_out_unlock;
1525 }
1526 break;
1527 case NFQNL_CFG_CMD_UNBIND:
1528 if (!queue) {
1529 ret = -ENODEV;
1530 goto err_out_unlock;
1531 }
1532 instance_destroy(q, queue);
1533 goto err_out_unlock;
1534 case NFQNL_CFG_CMD_PF_BIND:
1535 case NFQNL_CFG_CMD_PF_UNBIND:
1536 break;
1537 default:
1538 ret = -ENOTSUPP;
1539 goto err_out_unlock;
1540 }
1541 }
1542
1543 if (!queue) {
1544 ret = -ENODEV;
1545 goto err_out_unlock;
1546 }
1547
1548 if (nfqa[NFQA_CFG_PARAMS]) {
1549 struct nfqnl_msg_config_params *params =
1550 nla_data(nfqa[NFQA_CFG_PARAMS]);
1551
1552 nfqnl_set_mode(queue, params->copy_mode,
1553 ntohl(params->copy_range));
1554 }
1555
1556 if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1557 __be32 *queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1558
1559 spin_lock_bh(&queue->lock);
1560 queue->queue_maxlen = ntohl(*queue_maxlen);
1561 spin_unlock_bh(&queue->lock);
1562 }
1563
1564 if (nfqa[NFQA_CFG_FLAGS]) {
1565 spin_lock_bh(&queue->lock);
1566 queue->flags &= ~mask;
1567 queue->flags |= flags & mask;
1568 spin_unlock_bh(&queue->lock);
1569 }
1570
1571err_out_unlock:
1572 rcu_read_unlock();
1573 return ret;
1574}
1575
1576static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1577 [NFQNL_MSG_PACKET] = {
1578 .call = nfqnl_recv_unsupp,
1579 .type = NFNL_CB_RCU,
1580 .attr_count = NFQA_MAX,
1581 },
1582 [NFQNL_MSG_VERDICT] = {
1583 .call = nfqnl_recv_verdict,
1584 .type = NFNL_CB_RCU,
1585 .attr_count = NFQA_MAX,
1586 .policy = nfqa_verdict_policy
1587 },
1588 [NFQNL_MSG_CONFIG] = {
1589 .call = nfqnl_recv_config,
1590 .type = NFNL_CB_MUTEX,
1591 .attr_count = NFQA_CFG_MAX,
1592 .policy = nfqa_cfg_policy
1593 },
1594 [NFQNL_MSG_VERDICT_BATCH] = {
1595 .call = nfqnl_recv_verdict_batch,
1596 .type = NFNL_CB_RCU,
1597 .attr_count = NFQA_MAX,
1598 .policy = nfqa_verdict_batch_policy
1599 },
1600};
1601
1602static const struct nfnetlink_subsystem nfqnl_subsys = {
1603 .name = "nf_queue",
1604 .subsys_id = NFNL_SUBSYS_QUEUE,
1605 .cb_count = NFQNL_MSG_MAX,
1606 .cb = nfqnl_cb,
1607};
1608
1609#ifdef CONFIG_PROC_FS
1610struct iter_state {
1611 struct seq_net_private p;
1612 unsigned int bucket;
1613};
1614
1615static struct hlist_node *get_first(struct seq_file *seq)
1616{
1617 struct iter_state *st = seq->private;
1618 struct net *net;
1619 struct nfnl_queue_net *q;
1620
1621 if (!st)
1622 return NULL;
1623
1624 net = seq_file_net(seq);
1625 q = nfnl_queue_pernet(net);
1626 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1627 if (!hlist_empty(&q->instance_table[st->bucket]))
1628 return q->instance_table[st->bucket].first;
1629 }
1630 return NULL;
1631}
1632
1633static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1634{
1635 struct iter_state *st = seq->private;
1636 struct net *net = seq_file_net(seq);
1637
1638 h = h->next;
1639 while (!h) {
1640 struct nfnl_queue_net *q;
1641
1642 if (++st->bucket >= INSTANCE_BUCKETS)
1643 return NULL;
1644
1645 q = nfnl_queue_pernet(net);
1646 h = q->instance_table[st->bucket].first;
1647 }
1648 return h;
1649}
1650
1651static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1652{
1653 struct hlist_node *head;
1654 head = get_first(seq);
1655
1656 if (head)
1657 while (pos && (head = get_next(seq, head)))
1658 pos--;
1659 return pos ? NULL : head;
1660}
1661
1662static void *seq_start(struct seq_file *s, loff_t *pos)
1663 __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1664{
1665 spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1666 return get_idx(s, *pos);
1667}
1668
1669static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1670{
1671 (*pos)++;
1672 return get_next(s, v);
1673}
1674
1675static void seq_stop(struct seq_file *s, void *v)
1676 __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1677{
1678 spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1679}
1680
1681static int seq_show(struct seq_file *s, void *v)
1682{
1683 const struct nfqnl_instance *inst = v;
1684
1685 seq_printf(s, "%5u %6u %5u %1u %5u %5u %5u %8u %2d\n",
1686 inst->queue_num,
1687 inst->peer_portid, inst->queue_total,
1688 inst->copy_mode, inst->copy_range,
1689 inst->queue_dropped, inst->queue_user_dropped,
1690 inst->id_sequence, 1);
1691 return 0;
1692}
1693
1694static const struct seq_operations nfqnl_seq_ops = {
1695 .start = seq_start,
1696 .next = seq_next,
1697 .stop = seq_stop,
1698 .show = seq_show,
1699};
1700#endif /* PROC_FS */
1701
1702static int __net_init nfnl_queue_net_init(struct net *net)
1703{
1704 unsigned int i;
1705 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1706
1707 for (i = 0; i < INSTANCE_BUCKETS; i++)
1708 INIT_HLIST_HEAD(&q->instance_table[i]);
1709
1710 spin_lock_init(&q->instances_lock);
1711
1712#ifdef CONFIG_PROC_FS
1713 if (!proc_create_net("nfnetlink_queue", 0440, net->nf.proc_netfilter,
1714 &nfqnl_seq_ops, sizeof(struct iter_state)))
1715 return -ENOMEM;
1716#endif
1717 return 0;
1718}
1719
1720static void __net_exit nfnl_queue_net_exit(struct net *net)
1721{
1722 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1723 unsigned int i;
1724
1725#ifdef CONFIG_PROC_FS
1726 remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1727#endif
1728 for (i = 0; i < INSTANCE_BUCKETS; i++)
1729 WARN_ON_ONCE(!hlist_empty(&q->instance_table[i]));
1730}
1731
1732static struct pernet_operations nfnl_queue_net_ops = {
1733 .init = nfnl_queue_net_init,
1734 .exit = nfnl_queue_net_exit,
1735 .id = &nfnl_queue_net_id,
1736 .size = sizeof(struct nfnl_queue_net),
1737};
1738
1739static int __init nfnetlink_queue_init(void)
1740{
1741 int status;
1742
1743 status = register_pernet_subsys(&nfnl_queue_net_ops);
1744 if (status < 0) {
1745 pr_err("failed to register pernet ops\n");
1746 goto out;
1747 }
1748
1749 netlink_register_notifier(&nfqnl_rtnl_notifier);
1750 status = nfnetlink_subsys_register(&nfqnl_subsys);
1751 if (status < 0) {
1752 pr_err("failed to create netlink socket\n");
1753 goto cleanup_netlink_notifier;
1754 }
1755
1756 status = register_netdevice_notifier(&nfqnl_dev_notifier);
1757 if (status < 0) {
1758 pr_err("failed to register netdevice notifier\n");
1759 goto cleanup_netlink_subsys;
1760 }
1761
1762 nf_register_queue_handler(&nfqh);
1763
1764 return status;
1765
1766cleanup_netlink_subsys:
1767 nfnetlink_subsys_unregister(&nfqnl_subsys);
1768cleanup_netlink_notifier:
1769 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1770 unregister_pernet_subsys(&nfnl_queue_net_ops);
1771out:
1772 return status;
1773}
1774
1775static void __exit nfnetlink_queue_fini(void)
1776{
1777 nf_unregister_queue_handler();
1778 unregister_netdevice_notifier(&nfqnl_dev_notifier);
1779 nfnetlink_subsys_unregister(&nfqnl_subsys);
1780 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1781 unregister_pernet_subsys(&nfnl_queue_net_ops);
1782
1783 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1784}
1785
1786MODULE_DESCRIPTION("netfilter packet queue handler");
1787MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1788MODULE_LICENSE("GPL");
1789MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1790
1791module_init(nfnetlink_queue_init);
1792module_exit(nfnetlink_queue_fini);