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