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 <linux/rhashtable.h>
34 #include <linux/jhash.h>
35 #include <net/gso.h>
36 #include <net/sock.h>
37 #include <net/tcp_states.h>
38 #include <net/netfilter/nf_queue.h>
39 #include <net/netns/generic.h>
40
41 #include <linux/atomic.h>
42
43 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
44 #include "../bridge/br_private.h"
45 #endif
46
47 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
48 #include <net/netfilter/nf_conntrack.h>
49 #endif
50
51 #define NFQNL_QMAX_DEFAULT 1024
52 #define NFQNL_HASH_MIN 1024
53 #define NFQNL_HASH_MAX 1048576
54
55 /* We're using struct nlattr which has 16bit nla_len. Note that nla_len
56 * includes the header length. Thus, the maximum packet length that we
57 * support is 65531 bytes. We send truncated packets if the specified length
58 * is larger than that. Userspace can check for presence of NFQA_CAP_LEN
59 * attribute to detect truncation.
60 */
61 #define NFQNL_MAX_COPY_RANGE (0xffff - NLA_HDRLEN)
62
63 /* Composite key for packet lookup: (net, queue_num, packet_id) */
64 struct nfqnl_packet_key {
65 possible_net_t net;
66 u32 packet_id;
67 u16 queue_num;
68 } __aligned(sizeof(u32)); /* jhash2 requires 32-bit alignment */
69
70 /* Global rhashtable - one for entire system, all netns */
71 static struct rhashtable nfqnl_packet_map __read_mostly;
72
73 /* Helper to initialize composite key */
nfqnl_init_key(struct nfqnl_packet_key * key,struct net * net,u32 packet_id,u16 queue_num)74 static inline void nfqnl_init_key(struct nfqnl_packet_key *key,
75 struct net *net, u32 packet_id, u16 queue_num)
76 {
77 memset(key, 0, sizeof(*key));
78 write_pnet(&key->net, net);
79 key->packet_id = packet_id;
80 key->queue_num = queue_num;
81 }
82
83 struct nfqnl_instance {
84 struct hlist_node hlist; /* global list of queues */
85 struct rcu_head rcu;
86
87 u32 peer_portid;
88 unsigned int queue_maxlen;
89 unsigned int copy_range;
90 unsigned int queue_dropped;
91 unsigned int queue_user_dropped;
92
93
94 u_int16_t queue_num; /* number of this queue */
95 u_int8_t copy_mode;
96 u_int32_t flags; /* Set using NFQA_CFG_FLAGS */
97 /*
98 * Following fields are dirtied for each queued packet,
99 * keep them in same cache line if possible.
100 */
101 spinlock_t lock ____cacheline_aligned_in_smp;
102 unsigned int queue_total;
103 unsigned int id_sequence; /* 'sequence' of pkt ids */
104 struct list_head queue_list; /* packets in queue */
105 };
106
107 typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
108
109 static unsigned int nfnl_queue_net_id __read_mostly;
110
111 #define INSTANCE_BUCKETS 16
112 struct nfnl_queue_net {
113 spinlock_t instances_lock;
114 struct hlist_head instance_table[INSTANCE_BUCKETS];
115 };
116
nfnl_queue_pernet(struct net * net)117 static struct nfnl_queue_net *nfnl_queue_pernet(struct net *net)
118 {
119 return net_generic(net, nfnl_queue_net_id);
120 }
121
instance_hashfn(u_int16_t queue_num)122 static inline u_int8_t instance_hashfn(u_int16_t queue_num)
123 {
124 return ((queue_num >> 8) ^ queue_num) % INSTANCE_BUCKETS;
125 }
126
127 /* Extract composite key from nf_queue_entry for hashing */
nfqnl_packet_obj_hashfn(const void * data,u32 len,u32 seed)128 static u32 nfqnl_packet_obj_hashfn(const void *data, u32 len, u32 seed)
129 {
130 const struct nf_queue_entry *entry = data;
131 struct nfqnl_packet_key key;
132
133 nfqnl_init_key(&key, entry->state.net, entry->id, entry->queue_num);
134
135 return jhash2((u32 *)&key, sizeof(key) / sizeof(u32), seed);
136 }
137
138 /* Compare stack-allocated key against entry */
nfqnl_packet_obj_cmpfn(struct rhashtable_compare_arg * arg,const void * obj)139 static int nfqnl_packet_obj_cmpfn(struct rhashtable_compare_arg *arg,
140 const void *obj)
141 {
142 const struct nfqnl_packet_key *key = arg->key;
143 const struct nf_queue_entry *entry = obj;
144
145 return !net_eq(entry->state.net, read_pnet(&key->net)) ||
146 entry->queue_num != key->queue_num ||
147 entry->id != key->packet_id;
148 }
149
150 static const struct rhashtable_params nfqnl_rhashtable_params = {
151 .head_offset = offsetof(struct nf_queue_entry, hash_node),
152 .key_len = sizeof(struct nfqnl_packet_key),
153 .obj_hashfn = nfqnl_packet_obj_hashfn,
154 .obj_cmpfn = nfqnl_packet_obj_cmpfn,
155 .automatic_shrinking = true,
156 .min_size = NFQNL_HASH_MIN,
157 .max_size = NFQNL_HASH_MAX,
158 };
159
160 static struct nfqnl_instance *
instance_lookup(struct nfnl_queue_net * q,u_int16_t queue_num)161 instance_lookup(struct nfnl_queue_net *q, u_int16_t queue_num)
162 {
163 struct hlist_head *head;
164 struct nfqnl_instance *inst;
165
166 head = &q->instance_table[instance_hashfn(queue_num)];
167 hlist_for_each_entry_rcu(inst, head, hlist) {
168 if (inst->queue_num == queue_num)
169 return inst;
170 }
171 return NULL;
172 }
173
174 static struct nfqnl_instance *
instance_create(struct nfnl_queue_net * q,u_int16_t queue_num,u32 portid)175 instance_create(struct nfnl_queue_net *q, u_int16_t queue_num, u32 portid)
176 {
177 struct nfqnl_instance *inst;
178 unsigned int h;
179 int err;
180
181 inst = kzalloc_obj(*inst, GFP_KERNEL_ACCOUNT);
182 if (!inst)
183 return ERR_PTR(-ENOMEM);
184
185 inst->queue_num = queue_num;
186 inst->peer_portid = portid;
187 inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
188 inst->copy_range = NFQNL_MAX_COPY_RANGE;
189 inst->copy_mode = NFQNL_COPY_NONE;
190 spin_lock_init(&inst->lock);
191 INIT_LIST_HEAD(&inst->queue_list);
192
193 spin_lock(&q->instances_lock);
194 if (instance_lookup(q, queue_num)) {
195 err = -EEXIST;
196 goto out_unlock;
197 }
198
199 if (!try_module_get(THIS_MODULE)) {
200 err = -EAGAIN;
201 goto out_unlock;
202 }
203
204 h = instance_hashfn(queue_num);
205 hlist_add_head_rcu(&inst->hlist, &q->instance_table[h]);
206
207 spin_unlock(&q->instances_lock);
208
209 return inst;
210
211 out_unlock:
212 spin_unlock(&q->instances_lock);
213 kfree(inst);
214 return ERR_PTR(err);
215 }
216
217 static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
218 unsigned long data);
219
220 static void
instance_destroy_rcu(struct rcu_head * head)221 instance_destroy_rcu(struct rcu_head *head)
222 {
223 struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
224 rcu);
225
226 rcu_read_lock();
227 nfqnl_flush(inst, NULL, 0);
228 rcu_read_unlock();
229 kfree(inst);
230 module_put(THIS_MODULE);
231 }
232
233 static void
__instance_destroy(struct nfqnl_instance * inst)234 __instance_destroy(struct nfqnl_instance *inst)
235 {
236 hlist_del_rcu(&inst->hlist);
237 call_rcu(&inst->rcu, instance_destroy_rcu);
238 }
239
240 static void
instance_destroy(struct nfnl_queue_net * q,struct nfqnl_instance * inst)241 instance_destroy(struct nfnl_queue_net *q, struct nfqnl_instance *inst)
242 {
243 spin_lock(&q->instances_lock);
244 __instance_destroy(inst);
245 spin_unlock(&q->instances_lock);
246 }
247
248 static int
__enqueue_entry(struct nfqnl_instance * queue,struct nf_queue_entry * entry)249 __enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
250 {
251 int err;
252
253 entry->queue_num = queue->queue_num;
254
255 err = rhashtable_insert_fast(&nfqnl_packet_map, &entry->hash_node,
256 nfqnl_rhashtable_params);
257 if (unlikely(err))
258 return err;
259
260 list_add_tail(&entry->list, &queue->queue_list);
261 queue->queue_total++;
262
263 return 0;
264 }
265
266 static void
__dequeue_entry(struct nfqnl_instance * queue,struct nf_queue_entry * entry)267 __dequeue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
268 {
269 rhashtable_remove_fast(&nfqnl_packet_map, &entry->hash_node,
270 nfqnl_rhashtable_params);
271 list_del(&entry->list);
272 queue->queue_total--;
273 }
274
275 static struct nf_queue_entry *
find_dequeue_entry(struct nfqnl_instance * queue,unsigned int id,struct net * net)276 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id,
277 struct net *net)
278 {
279 struct nfqnl_packet_key key;
280 struct nf_queue_entry *entry;
281
282 nfqnl_init_key(&key, net, id, queue->queue_num);
283
284 spin_lock_bh(&queue->lock);
285 entry = rhashtable_lookup_fast(&nfqnl_packet_map, &key,
286 nfqnl_rhashtable_params);
287
288 if (entry)
289 __dequeue_entry(queue, entry);
290
291 spin_unlock_bh(&queue->lock);
292
293 return entry;
294 }
295
nf_iterate(struct sk_buff * skb,struct nf_hook_state * state,const struct nf_hook_entries * hooks,unsigned int * index)296 static unsigned int nf_iterate(struct sk_buff *skb,
297 struct nf_hook_state *state,
298 const struct nf_hook_entries *hooks,
299 unsigned int *index)
300 {
301 const struct nf_hook_entry *hook;
302 unsigned int verdict, i = *index;
303
304 while (i < hooks->num_hook_entries) {
305 hook = &hooks->hooks[i];
306 repeat:
307 verdict = nf_hook_entry_hookfn(hook, skb, state);
308 if (verdict != NF_ACCEPT) {
309 *index = i;
310 if (verdict != NF_REPEAT)
311 return verdict;
312 goto repeat;
313 }
314 i++;
315 }
316
317 *index = i;
318 return NF_ACCEPT;
319 }
320
nf_hook_entries_head(const struct net * net,u8 pf,u8 hooknum)321 static struct nf_hook_entries *nf_hook_entries_head(const struct net *net, u8 pf, u8 hooknum)
322 {
323 switch (pf) {
324 #ifdef CONFIG_NETFILTER_FAMILY_BRIDGE
325 case NFPROTO_BRIDGE:
326 return rcu_dereference(net->nf.hooks_bridge[hooknum]);
327 #endif
328 case NFPROTO_IPV4:
329 return rcu_dereference(net->nf.hooks_ipv4[hooknum]);
330 case NFPROTO_IPV6:
331 return rcu_dereference(net->nf.hooks_ipv6[hooknum]);
332 default:
333 WARN_ON_ONCE(1);
334 return NULL;
335 }
336
337 return NULL;
338 }
339
nf_ip_reroute(struct sk_buff * skb,const struct nf_queue_entry * entry)340 static int nf_ip_reroute(struct sk_buff *skb, const struct nf_queue_entry *entry)
341 {
342 #ifdef CONFIG_INET
343 const struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
344
345 if (entry->state.hook == NF_INET_LOCAL_OUT) {
346 const struct iphdr *iph = ip_hdr(skb);
347
348 if (!(iph->tos == rt_info->tos &&
349 skb->mark == rt_info->mark &&
350 iph->daddr == rt_info->daddr &&
351 iph->saddr == rt_info->saddr))
352 return ip_route_me_harder(entry->state.net, entry->state.sk,
353 skb, RTN_UNSPEC);
354 }
355 #endif
356 return 0;
357 }
358
nf_reroute(struct sk_buff * skb,struct nf_queue_entry * entry)359 static int nf_reroute(struct sk_buff *skb, struct nf_queue_entry *entry)
360 {
361 const struct nf_ipv6_ops *v6ops;
362 int ret = 0;
363
364 switch (entry->state.pf) {
365 case AF_INET:
366 ret = nf_ip_reroute(skb, entry);
367 break;
368 case AF_INET6:
369 v6ops = rcu_dereference(nf_ipv6_ops);
370 if (v6ops)
371 ret = v6ops->reroute(skb, entry);
372 break;
373 }
374 return ret;
375 }
376
377 /* caller must hold rcu read-side lock */
nf_reinject(struct nf_queue_entry * entry,unsigned int verdict)378 static void nf_reinject(struct nf_queue_entry *entry, unsigned int verdict)
379 {
380 const struct nf_hook_entry *hook_entry;
381 const struct nf_hook_entries *hooks;
382 struct sk_buff *skb = entry->skb;
383 const struct net *net;
384 unsigned int i;
385 int err;
386 u8 pf;
387
388 net = entry->state.net;
389 pf = entry->state.pf;
390
391 hooks = nf_hook_entries_head(net, pf, entry->state.hook);
392
393 i = entry->hook_index;
394 if (!hooks || i >= hooks->num_hook_entries) {
395 kfree_skb_reason(skb, SKB_DROP_REASON_NETFILTER_DROP);
396 nf_queue_entry_free(entry);
397 return;
398 }
399
400 hook_entry = &hooks->hooks[i];
401
402 /* Continue traversal iff userspace said ok... */
403 if (verdict == NF_REPEAT)
404 verdict = nf_hook_entry_hookfn(hook_entry, skb, &entry->state);
405
406 if (verdict == NF_ACCEPT) {
407 if (nf_reroute(skb, entry) < 0)
408 verdict = NF_DROP;
409 }
410
411 if (verdict == NF_ACCEPT) {
412 next_hook:
413 ++i;
414 verdict = nf_iterate(skb, &entry->state, hooks, &i);
415 }
416
417 switch (verdict & NF_VERDICT_MASK) {
418 case NF_ACCEPT:
419 case NF_STOP:
420 local_bh_disable();
421 entry->state.okfn(entry->state.net, entry->state.sk, skb);
422 local_bh_enable();
423 break;
424 case NF_QUEUE:
425 err = nf_queue(skb, &entry->state, i, verdict);
426 if (err == 1)
427 goto next_hook;
428 break;
429 case NF_STOLEN:
430 break;
431 default:
432 kfree_skb(skb);
433 }
434
435 nf_queue_entry_free(entry);
436 }
437
438 /* return true if the entry has an unconfirmed conntrack attached that isn't owned by us
439 * exclusively.
440 */
nf_ct_drop_unconfirmed(const struct nf_queue_entry * entry,bool * is_unconfirmed)441 static bool nf_ct_drop_unconfirmed(const struct nf_queue_entry *entry, bool *is_unconfirmed)
442 {
443 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
444 struct nf_conn *ct = (void *)skb_nfct(entry->skb);
445
446 if (!ct || nf_ct_is_confirmed(ct))
447 return false;
448
449 if (is_unconfirmed)
450 *is_unconfirmed = true;
451
452 /* in some cases skb_clone() can occur after initial conntrack
453 * pickup, but conntrack assumes exclusive skb->_nfct ownership for
454 * unconfirmed entries.
455 *
456 * This happens for br_netfilter and with ip multicast routing.
457 * This can't be solved with serialization here because one clone
458 * could have been queued for local delivery or could be transmitted
459 * in parallel on another CPU.
460 */
461 return refcount_read(&ct->ct_general.use) > 1;
462 #endif
463 return false;
464 }
465
nfqnl_reinject(struct nf_queue_entry * entry,unsigned int verdict)466 static void nfqnl_reinject(struct nf_queue_entry *entry, unsigned int verdict)
467 {
468 const struct nf_ct_hook *ct_hook;
469
470 if (verdict == NF_ACCEPT ||
471 verdict == NF_REPEAT ||
472 verdict == NF_STOP) {
473 unsigned int ct_verdict = verdict;
474
475 rcu_read_lock();
476 ct_hook = rcu_dereference(nf_ct_hook);
477 if (ct_hook)
478 ct_verdict = ct_hook->update(entry->state.net, entry->skb);
479 rcu_read_unlock();
480
481 switch (ct_verdict & NF_VERDICT_MASK) {
482 case NF_ACCEPT:
483 /* follow userspace verdict, could be REPEAT */
484 break;
485 case NF_STOLEN:
486 nf_queue_entry_free(entry);
487 return;
488 default:
489 verdict = ct_verdict & NF_VERDICT_MASK;
490 break;
491 }
492 }
493
494 if (verdict != NF_DROP && entry->nf_ct_is_unconfirmed) {
495 /* If first queued segment was already reinjected then
496 * there is a good chance the ct entry is now confirmed.
497 *
498 * Handle the rare cases:
499 * - out-of-order verdict
500 * - threaded userspace reinjecting in parallel
501 * - first segment was dropped
502 *
503 * In all of those cases we can't handle this packet
504 * because we can't be sure that another CPU won't modify
505 * nf_conn->ext in parallel which isn't allowed.
506 */
507 if (nf_ct_drop_unconfirmed(entry, NULL))
508 verdict = NF_DROP;
509 }
510
511 nf_reinject(entry, verdict);
512 }
513
514 static void
nfqnl_flush(struct nfqnl_instance * queue,nfqnl_cmpfn cmpfn,unsigned long data)515 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
516 {
517 struct nf_queue_entry *entry, *next;
518
519 spin_lock_bh(&queue->lock);
520 list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
521 if (!cmpfn || cmpfn(entry, data)) {
522 __dequeue_entry(queue, entry);
523 nfqnl_reinject(entry, NF_DROP);
524 }
525 }
526 spin_unlock_bh(&queue->lock);
527 }
528
529 static int
nfqnl_put_packet_info(struct sk_buff * nlskb,struct sk_buff * packet,bool csum_verify)530 nfqnl_put_packet_info(struct sk_buff *nlskb, struct sk_buff *packet,
531 bool csum_verify)
532 {
533 __u32 flags = 0;
534
535 if (packet->ip_summed == CHECKSUM_PARTIAL)
536 flags = NFQA_SKB_CSUMNOTREADY;
537 else if (csum_verify)
538 flags = NFQA_SKB_CSUM_NOTVERIFIED;
539
540 if (skb_is_gso(packet))
541 flags |= NFQA_SKB_GSO;
542
543 return flags ? nla_put_be32(nlskb, NFQA_SKB_INFO, htonl(flags)) : 0;
544 }
545
nfqnl_put_sk_uidgid(struct sk_buff * skb,struct sock * sk)546 static int nfqnl_put_sk_uidgid(struct sk_buff *skb, struct sock *sk)
547 {
548 const struct cred *cred;
549
550 if (!sk_fullsock(sk))
551 return 0;
552
553 read_lock_bh(&sk->sk_callback_lock);
554 if (sk->sk_socket && sk->sk_socket->file) {
555 cred = sk->sk_socket->file->f_cred;
556 if (nla_put_be32(skb, NFQA_UID,
557 htonl(from_kuid_munged(&init_user_ns, cred->fsuid))))
558 goto nla_put_failure;
559 if (nla_put_be32(skb, NFQA_GID,
560 htonl(from_kgid_munged(&init_user_ns, cred->fsgid))))
561 goto nla_put_failure;
562 }
563 read_unlock_bh(&sk->sk_callback_lock);
564 return 0;
565
566 nla_put_failure:
567 read_unlock_bh(&sk->sk_callback_lock);
568 return -1;
569 }
570
nfqnl_put_sk_classid(struct sk_buff * skb,struct sock * sk)571 static int nfqnl_put_sk_classid(struct sk_buff *skb, struct sock *sk)
572 {
573 #if IS_ENABLED(CONFIG_CGROUP_NET_CLASSID)
574 if (sk && sk_fullsock(sk)) {
575 u32 classid = sock_cgroup_classid(&sk->sk_cgrp_data);
576
577 if (classid && nla_put_be32(skb, NFQA_CGROUP_CLASSID, htonl(classid)))
578 return -1;
579 }
580 #endif
581 return 0;
582 }
583
nfqnl_get_sk_secctx(struct sk_buff * skb,struct lsm_context * ctx)584 static int nfqnl_get_sk_secctx(struct sk_buff *skb, struct lsm_context *ctx)
585 {
586 int seclen = 0;
587 #if IS_ENABLED(CONFIG_NETWORK_SECMARK)
588
589 if (!skb || !sk_fullsock(skb->sk))
590 return 0;
591
592 read_lock_bh(&skb->sk->sk_callback_lock);
593
594 if (skb->secmark)
595 seclen = security_secid_to_secctx(skb->secmark, ctx);
596 read_unlock_bh(&skb->sk->sk_callback_lock);
597 #endif
598 return seclen;
599 }
600
nfqnl_get_bridge_size(struct nf_queue_entry * entry)601 static u32 nfqnl_get_bridge_size(struct nf_queue_entry *entry)
602 {
603 struct sk_buff *entskb = entry->skb;
604 u32 nlalen = 0;
605
606 if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb))
607 return 0;
608
609 if (skb_vlan_tag_present(entskb))
610 nlalen += nla_total_size(nla_total_size(sizeof(__be16)) +
611 nla_total_size(sizeof(__be16)));
612
613 if (entskb->network_header > entskb->mac_header)
614 nlalen += nla_total_size((entskb->network_header -
615 entskb->mac_header));
616
617 return nlalen;
618 }
619
nfqnl_put_bridge(struct nf_queue_entry * entry,struct sk_buff * skb)620 static int nfqnl_put_bridge(struct nf_queue_entry *entry, struct sk_buff *skb)
621 {
622 struct sk_buff *entskb = entry->skb;
623
624 if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb))
625 return 0;
626
627 if (skb_vlan_tag_present(entskb)) {
628 struct nlattr *nest;
629
630 nest = nla_nest_start(skb, NFQA_VLAN);
631 if (!nest)
632 goto nla_put_failure;
633
634 if (nla_put_be16(skb, NFQA_VLAN_TCI, htons(entskb->vlan_tci)) ||
635 nla_put_be16(skb, NFQA_VLAN_PROTO, entskb->vlan_proto))
636 goto nla_put_failure;
637
638 nla_nest_end(skb, nest);
639 }
640
641 if (entskb->mac_header < entskb->network_header) {
642 int len = (int)(entskb->network_header - entskb->mac_header);
643
644 if (nla_put(skb, NFQA_L2HDR, len, skb_mac_header(entskb)))
645 goto nla_put_failure;
646 }
647
648 return 0;
649
650 nla_put_failure:
651 return -1;
652 }
653
nf_queue_checksum_help(struct sk_buff * entskb)654 static int nf_queue_checksum_help(struct sk_buff *entskb)
655 {
656 if (skb_csum_is_sctp(entskb))
657 return skb_crc32c_csum_help(entskb);
658
659 return skb_checksum_help(entskb);
660 }
661
662 static struct sk_buff *
nfqnl_build_packet_message(struct net * net,struct nfqnl_instance * queue,struct nf_queue_entry * entry,__be32 ** packet_id_ptr)663 nfqnl_build_packet_message(struct net *net, struct nfqnl_instance *queue,
664 struct nf_queue_entry *entry,
665 __be32 **packet_id_ptr)
666 {
667 size_t size;
668 size_t data_len = 0, cap_len = 0;
669 unsigned int hlen = 0;
670 struct sk_buff *skb;
671 struct nlattr *nla;
672 struct nfqnl_msg_packet_hdr *pmsg;
673 struct nlmsghdr *nlh;
674 struct sk_buff *entskb = entry->skb;
675 struct net_device *indev;
676 struct net_device *outdev;
677 struct nf_conn *ct = NULL;
678 enum ip_conntrack_info ctinfo = 0;
679 const struct nfnl_ct_hook *nfnl_ct;
680 bool csum_verify;
681 struct lsm_context ctx = { NULL, 0, 0 };
682 int seclen = 0;
683 ktime_t tstamp;
684
685 size = nlmsg_total_size(sizeof(struct nfgenmsg))
686 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
687 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
688 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
689 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
690 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
691 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
692 #endif
693 + nla_total_size(sizeof(u_int32_t)) /* mark */
694 + nla_total_size(sizeof(u_int32_t)) /* priority */
695 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
696 + nla_total_size(sizeof(u_int32_t)) /* skbinfo */
697 #if IS_ENABLED(CONFIG_CGROUP_NET_CLASSID)
698 + nla_total_size(sizeof(u_int32_t)) /* classid */
699 #endif
700 + nla_total_size(sizeof(u_int32_t)); /* cap_len */
701
702 tstamp = skb_tstamp_cond(entskb, false);
703 if (tstamp)
704 size += nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
705
706 size += nfqnl_get_bridge_size(entry);
707
708 if (entry->state.hook <= NF_INET_FORWARD ||
709 (entry->state.hook == NF_INET_POST_ROUTING && entskb->sk == NULL))
710 csum_verify = !skb_csum_unnecessary(entskb);
711 else
712 csum_verify = false;
713
714 outdev = entry->state.out;
715
716 switch ((enum nfqnl_config_mode)READ_ONCE(queue->copy_mode)) {
717 case NFQNL_COPY_META:
718 case NFQNL_COPY_NONE:
719 break;
720
721 case NFQNL_COPY_PACKET:
722 if (!(queue->flags & NFQA_CFG_F_GSO) &&
723 entskb->ip_summed == CHECKSUM_PARTIAL &&
724 nf_queue_checksum_help(entskb))
725 return NULL;
726
727 data_len = READ_ONCE(queue->copy_range);
728 if (data_len > entskb->len)
729 data_len = entskb->len;
730
731 hlen = skb_zerocopy_headlen(entskb);
732 hlen = min_t(unsigned int, hlen, data_len);
733 size += sizeof(struct nlattr) + hlen;
734 cap_len = entskb->len;
735 break;
736 }
737
738 nfnl_ct = rcu_dereference(nfnl_ct_hook);
739
740 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
741 if (queue->flags & NFQA_CFG_F_CONNTRACK) {
742 if (nfnl_ct != NULL) {
743 ct = nf_ct_get(entskb, &ctinfo);
744 if (ct != NULL)
745 size += nfnl_ct->build_size(ct);
746 }
747 }
748 #endif
749
750 if (queue->flags & NFQA_CFG_F_UID_GID) {
751 size += (nla_total_size(sizeof(u_int32_t)) /* uid */
752 + nla_total_size(sizeof(u_int32_t))); /* gid */
753 }
754
755 if ((queue->flags & NFQA_CFG_F_SECCTX) && entskb->sk) {
756 seclen = nfqnl_get_sk_secctx(entskb, &ctx);
757 if (seclen < 0)
758 return NULL;
759 if (seclen)
760 size += nla_total_size(seclen);
761 }
762
763 skb = alloc_skb(size, GFP_ATOMIC);
764 if (!skb) {
765 skb_tx_error(entskb);
766 goto nlmsg_failure;
767 }
768
769 nlh = nfnl_msg_put(skb, 0, 0,
770 nfnl_msg_type(NFNL_SUBSYS_QUEUE, NFQNL_MSG_PACKET),
771 0, entry->state.pf, NFNETLINK_V0,
772 htons(queue->queue_num));
773 if (!nlh) {
774 skb_tx_error(entskb);
775 kfree_skb(skb);
776 goto nlmsg_failure;
777 }
778
779 nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
780 pmsg = nla_data(nla);
781 pmsg->hw_protocol = entskb->protocol;
782 pmsg->hook = entry->state.hook;
783 *packet_id_ptr = &pmsg->packet_id;
784
785 indev = entry->state.in;
786 if (indev) {
787 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
788 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
789 goto nla_put_failure;
790 #else
791 if (entry->state.pf == PF_BRIDGE) {
792 /* Case 1: indev is physical input device, we need to
793 * look for bridge group (when called from
794 * netfilter_bridge) */
795 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
796 htonl(indev->ifindex)) ||
797 /* this is the bridge group "brX" */
798 /* rcu_read_lock()ed by __nf_queue */
799 nla_put_be32(skb, NFQA_IFINDEX_INDEV,
800 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
801 goto nla_put_failure;
802 } else {
803 int physinif;
804
805 /* Case 2: indev is bridge group, we need to look for
806 * physical device (when called from ipv4) */
807 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
808 htonl(indev->ifindex)))
809 goto nla_put_failure;
810
811 physinif = nf_bridge_get_physinif(entskb);
812 if (physinif &&
813 nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
814 htonl(physinif)))
815 goto nla_put_failure;
816 }
817 #endif
818 }
819
820 if (outdev) {
821 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
822 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
823 goto nla_put_failure;
824 #else
825 if (entry->state.pf == PF_BRIDGE) {
826 /* Case 1: outdev is physical output device, we need to
827 * look for bridge group (when called from
828 * netfilter_bridge) */
829 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
830 htonl(outdev->ifindex)) ||
831 /* this is the bridge group "brX" */
832 /* rcu_read_lock()ed by __nf_queue */
833 nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
834 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
835 goto nla_put_failure;
836 } else {
837 int physoutif;
838
839 /* Case 2: outdev is bridge group, we need to look for
840 * physical output device (when called from ipv4) */
841 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
842 htonl(outdev->ifindex)))
843 goto nla_put_failure;
844
845 physoutif = nf_bridge_get_physoutif(entskb);
846 if (physoutif &&
847 nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
848 htonl(physoutif)))
849 goto nla_put_failure;
850 }
851 #endif
852 }
853
854 if (entskb->mark &&
855 nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
856 goto nla_put_failure;
857
858 if (entskb->priority &&
859 nla_put_be32(skb, NFQA_PRIORITY, htonl(entskb->priority)))
860 goto nla_put_failure;
861
862 if (indev && entskb->dev &&
863 skb_mac_header_was_set(entskb) &&
864 skb_mac_header_len(entskb) != 0) {
865 struct nfqnl_msg_packet_hw phw;
866 int len;
867
868 memset(&phw, 0, sizeof(phw));
869 len = dev_parse_header(entskb, phw.hw_addr);
870 if (len) {
871 phw.hw_addrlen = htons(len);
872 if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
873 goto nla_put_failure;
874 }
875 }
876
877 if (nfqnl_put_bridge(entry, skb) < 0)
878 goto nla_put_failure;
879
880 if (entry->state.hook <= NF_INET_FORWARD && tstamp) {
881 struct nfqnl_msg_packet_timestamp ts;
882 struct timespec64 kts = ktime_to_timespec64(tstamp);
883
884 ts.sec = cpu_to_be64(kts.tv_sec);
885 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
886
887 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
888 goto nla_put_failure;
889 }
890
891 if ((queue->flags & NFQA_CFG_F_UID_GID) && entskb->sk &&
892 nfqnl_put_sk_uidgid(skb, entskb->sk) < 0)
893 goto nla_put_failure;
894
895 if (nfqnl_put_sk_classid(skb, entskb->sk) < 0)
896 goto nla_put_failure;
897
898 if (seclen > 0 && nla_put(skb, NFQA_SECCTX, ctx.len, ctx.context))
899 goto nla_put_failure;
900
901 if (ct && nfnl_ct->build(skb, ct, ctinfo, NFQA_CT, NFQA_CT_INFO) < 0)
902 goto nla_put_failure;
903
904 if (cap_len > data_len &&
905 nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
906 goto nla_put_failure;
907
908 if (nfqnl_put_packet_info(skb, entskb, csum_verify))
909 goto nla_put_failure;
910
911 if (data_len) {
912 struct nlattr *nla;
913
914 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
915 goto nla_put_failure;
916
917 nla = skb_put(skb, sizeof(*nla));
918 nla->nla_type = NFQA_PAYLOAD;
919 nla->nla_len = nla_attr_size(data_len);
920
921 if (skb_zerocopy(skb, entskb, data_len, hlen))
922 goto nla_put_failure;
923 }
924
925 nlh->nlmsg_len = skb->len;
926 if (seclen >= 0)
927 security_release_secctx(&ctx);
928 return skb;
929
930 nla_put_failure:
931 skb_tx_error(entskb);
932 kfree_skb(skb);
933 net_err_ratelimited("nf_queue: error creating packet message\n");
934 nlmsg_failure:
935 if (seclen >= 0)
936 security_release_secctx(&ctx);
937 return NULL;
938 }
939
940 static int
__nfqnl_enqueue_packet(struct net * net,struct nfqnl_instance * queue,struct nf_queue_entry * entry)941 __nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
942 struct nf_queue_entry *entry)
943 {
944 struct sk_buff *nskb;
945 int err = -ENOBUFS;
946 __be32 *packet_id_ptr;
947 int failopen = 0;
948
949 nskb = nfqnl_build_packet_message(net, queue, entry, &packet_id_ptr);
950 if (nskb == NULL) {
951 err = -ENOMEM;
952 goto err_out;
953 }
954 spin_lock_bh(&queue->lock);
955
956 if (queue->queue_total >= queue->queue_maxlen)
957 goto err_out_queue_drop;
958
959 entry->id = ++queue->id_sequence;
960 *packet_id_ptr = htonl(entry->id);
961
962 /* Insert into hash BEFORE unicast. If failure don't send to userspace. */
963 err = __enqueue_entry(queue, entry);
964 if (unlikely(err))
965 goto err_out_queue_drop;
966
967 /* nfnetlink_unicast will either free the nskb or add it to a socket */
968 err = nfnetlink_unicast(nskb, net, queue->peer_portid);
969 if (err < 0) {
970 /* Unicast failed - remove entry we just inserted */
971 __dequeue_entry(queue, entry);
972
973 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
974 failopen = 1;
975 err = 0;
976 } else {
977 queue->queue_user_dropped++;
978 }
979 goto err_out_unlock;
980 }
981
982 spin_unlock_bh(&queue->lock);
983 return 0;
984
985 err_out_queue_drop:
986 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
987 failopen = 1;
988 err = 0;
989 } else {
990 queue->queue_dropped++;
991
992 if (queue->queue_total >= queue->queue_maxlen)
993 net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
994 queue->queue_total);
995 else
996 net_warn_ratelimited("nf_queue: hash insert failed: %d\n", err);
997 }
998 kfree_skb(nskb);
999 err_out_unlock:
1000 spin_unlock_bh(&queue->lock);
1001 if (failopen)
1002 nfqnl_reinject(entry, NF_ACCEPT);
1003 err_out:
1004 return err;
1005 }
1006
1007 static struct nf_queue_entry *
nf_queue_entry_dup(struct nf_queue_entry * e)1008 nf_queue_entry_dup(struct nf_queue_entry *e)
1009 {
1010 struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
1011
1012 if (!entry)
1013 return NULL;
1014
1015 if (nf_queue_entry_get_refs(entry))
1016 return entry;
1017
1018 kfree(entry);
1019 return NULL;
1020 }
1021
1022 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1023 /* When called from bridge netfilter, skb->data must point to MAC header
1024 * before calling skb_gso_segment(). Else, original MAC header is lost
1025 * and segmented skbs will be sent to wrong destination.
1026 */
nf_bridge_adjust_skb_data(struct sk_buff * skb)1027 static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
1028 {
1029 if (nf_bridge_info_get(skb))
1030 __skb_push(skb, skb->network_header - skb->mac_header);
1031 }
1032
nf_bridge_adjust_segmented_data(struct sk_buff * skb)1033 static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
1034 {
1035 if (nf_bridge_info_get(skb))
1036 __skb_pull(skb, skb->network_header - skb->mac_header);
1037 }
1038 #else
1039 #define nf_bridge_adjust_skb_data(s) do {} while (0)
1040 #define nf_bridge_adjust_segmented_data(s) do {} while (0)
1041 #endif
1042
1043 static int
__nfqnl_enqueue_packet_gso(struct net * net,struct nfqnl_instance * queue,struct sk_buff * skb,struct nf_queue_entry * entry)1044 __nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
1045 struct sk_buff *skb, struct nf_queue_entry *entry)
1046 {
1047 int ret = -ENOMEM;
1048 struct nf_queue_entry *entry_seg;
1049
1050 nf_bridge_adjust_segmented_data(skb);
1051
1052 if (skb->next == NULL) { /* last packet, no need to copy entry */
1053 struct sk_buff *gso_skb = entry->skb;
1054 entry->skb = skb;
1055 ret = __nfqnl_enqueue_packet(net, queue, entry);
1056 if (ret)
1057 entry->skb = gso_skb;
1058 return ret;
1059 }
1060
1061 skb_mark_not_on_list(skb);
1062
1063 entry_seg = nf_queue_entry_dup(entry);
1064 if (entry_seg) {
1065 entry_seg->skb = skb;
1066 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
1067 if (ret)
1068 nf_queue_entry_free(entry_seg);
1069 }
1070 return ret;
1071 }
1072
1073 static int
nfqnl_enqueue_packet(struct nf_queue_entry * entry,unsigned int queuenum)1074 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
1075 {
1076 struct sk_buff *skb, *segs, *nskb;
1077 bool ct_is_unconfirmed = false;
1078 struct nfqnl_instance *queue;
1079 unsigned int queued;
1080 int err = -ENOBUFS;
1081 struct net *net = entry->state.net;
1082 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1083
1084 /* rcu_read_lock()ed by nf_hook_thresh */
1085 queue = instance_lookup(q, queuenum);
1086 if (!queue)
1087 return -ESRCH;
1088
1089 if (queue->copy_mode == NFQNL_COPY_NONE)
1090 return -EINVAL;
1091
1092 skb = entry->skb;
1093
1094 switch (entry->state.pf) {
1095 case NFPROTO_IPV4:
1096 skb->protocol = htons(ETH_P_IP);
1097 break;
1098 case NFPROTO_IPV6:
1099 skb->protocol = htons(ETH_P_IPV6);
1100 break;
1101 }
1102
1103 /* Check if someone already holds another reference to
1104 * unconfirmed ct. If so, we cannot queue the skb:
1105 * concurrent modifications of nf_conn->ext are not
1106 * allowed and we can't know if another CPU isn't
1107 * processing the same nf_conn entry in parallel.
1108 */
1109 if (nf_ct_drop_unconfirmed(entry, &ct_is_unconfirmed))
1110 return -EINVAL;
1111
1112 if (!skb_is_gso(skb) || ((queue->flags & NFQA_CFG_F_GSO) && !skb_is_gso_sctp(skb)))
1113 return __nfqnl_enqueue_packet(net, queue, entry);
1114
1115 nf_bridge_adjust_skb_data(skb);
1116 segs = skb_gso_segment(skb, 0);
1117 /* Does not use PTR_ERR to limit the number of error codes that can be
1118 * returned by nf_queue. For instance, callers rely on -ESRCH to
1119 * mean 'ignore this hook'.
1120 */
1121 if (IS_ERR_OR_NULL(segs))
1122 goto out_err;
1123 queued = 0;
1124 err = 0;
1125
1126 skb_list_walk_safe(segs, segs, nskb) {
1127 if (ct_is_unconfirmed && queued > 0) {
1128 /* skb_gso_segment() increments the ct refcount.
1129 * This is a problem for unconfirmed (not in hash)
1130 * entries, those can race when reinjections happen
1131 * in parallel.
1132 *
1133 * Annotate this for all queued entries except the
1134 * first one.
1135 *
1136 * As long as the first one is reinjected first it
1137 * will do the confirmation for us.
1138 */
1139 entry->nf_ct_is_unconfirmed = ct_is_unconfirmed;
1140 }
1141
1142 if (err == 0)
1143 err = __nfqnl_enqueue_packet_gso(net, queue,
1144 segs, entry);
1145 if (err == 0)
1146 queued++;
1147 else
1148 kfree_skb(segs);
1149 }
1150
1151 if (queued) {
1152 if (err) /* some segments are already queued */
1153 nf_queue_entry_free(entry);
1154 kfree_skb(skb);
1155 return 0;
1156 }
1157 out_err:
1158 nf_bridge_adjust_segmented_data(skb);
1159 return err;
1160 }
1161
1162 static int
nfqnl_mangle(void * data,unsigned int data_len,struct nf_queue_entry * e,int diff)1163 nfqnl_mangle(void *data, unsigned int data_len, struct nf_queue_entry *e, int diff)
1164 {
1165 struct sk_buff *nskb;
1166
1167 if (diff < 0) {
1168 unsigned int min_len = skb_transport_offset(e->skb);
1169
1170 if (data_len < min_len)
1171 return -EINVAL;
1172
1173 if (pskb_trim(e->skb, data_len))
1174 return -ENOMEM;
1175 } else if (diff > 0) {
1176 if (data_len > 0xFFFF)
1177 return -EINVAL;
1178 if (diff > skb_tailroom(e->skb)) {
1179 nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
1180 diff, GFP_ATOMIC);
1181 if (!nskb)
1182 return -ENOMEM;
1183 kfree_skb(e->skb);
1184 e->skb = nskb;
1185 }
1186 skb_put(e->skb, diff);
1187 }
1188 if (skb_ensure_writable(e->skb, data_len))
1189 return -ENOMEM;
1190 skb_copy_to_linear_data(e->skb, data, data_len);
1191 e->skb->ip_summed = CHECKSUM_NONE;
1192 return 0;
1193 }
1194
1195 static int
nfqnl_set_mode(struct nfqnl_instance * queue,unsigned char mode,unsigned int range)1196 nfqnl_set_mode(struct nfqnl_instance *queue,
1197 unsigned char mode, unsigned int range)
1198 {
1199 int status = 0;
1200
1201 spin_lock_bh(&queue->lock);
1202 switch (mode) {
1203 case NFQNL_COPY_NONE:
1204 case NFQNL_COPY_META:
1205 queue->copy_mode = mode;
1206 queue->copy_range = 0;
1207 break;
1208
1209 case NFQNL_COPY_PACKET:
1210 queue->copy_mode = mode;
1211 if (range == 0 || range > NFQNL_MAX_COPY_RANGE)
1212 queue->copy_range = NFQNL_MAX_COPY_RANGE;
1213 else
1214 queue->copy_range = range;
1215 break;
1216
1217 default:
1218 status = -EINVAL;
1219
1220 }
1221 spin_unlock_bh(&queue->lock);
1222
1223 return status;
1224 }
1225
1226 static int
dev_cmp(struct nf_queue_entry * entry,unsigned long ifindex)1227 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
1228 {
1229 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
1230 int physinif, physoutif;
1231
1232 physinif = nf_bridge_get_physinif(entry->skb);
1233 physoutif = nf_bridge_get_physoutif(entry->skb);
1234
1235 if (physinif == ifindex || physoutif == ifindex)
1236 return 1;
1237 #endif
1238 if (entry->state.in)
1239 if (entry->state.in->ifindex == ifindex)
1240 return 1;
1241 if (entry->state.out)
1242 if (entry->state.out->ifindex == ifindex)
1243 return 1;
1244
1245 return 0;
1246 }
1247
1248 /* drop all packets with either indev or outdev == ifindex from all queue
1249 * instances */
1250 static void
nfqnl_dev_drop(struct net * net,int ifindex)1251 nfqnl_dev_drop(struct net *net, int ifindex)
1252 {
1253 int i;
1254 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1255
1256 rcu_read_lock();
1257
1258 for (i = 0; i < INSTANCE_BUCKETS; i++) {
1259 struct nfqnl_instance *inst;
1260 struct hlist_head *head = &q->instance_table[i];
1261
1262 hlist_for_each_entry_rcu(inst, head, hlist)
1263 nfqnl_flush(inst, dev_cmp, ifindex);
1264 }
1265
1266 rcu_read_unlock();
1267 }
1268
1269 static int
nfqnl_rcv_dev_event(struct notifier_block * this,unsigned long event,void * ptr)1270 nfqnl_rcv_dev_event(struct notifier_block *this,
1271 unsigned long event, void *ptr)
1272 {
1273 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1274
1275 /* Drop any packets associated with the downed device */
1276 if (event == NETDEV_DOWN)
1277 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
1278 return NOTIFY_DONE;
1279 }
1280
1281 static struct notifier_block nfqnl_dev_notifier = {
1282 .notifier_call = nfqnl_rcv_dev_event,
1283 };
1284
nfqnl_nf_hook_drop(struct net * net)1285 static void nfqnl_nf_hook_drop(struct net *net)
1286 {
1287 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1288 int i;
1289
1290 /* This function is also called on net namespace error unwind,
1291 * when pernet_ops->init() failed and ->exit() functions of the
1292 * previous pernet_ops gets called.
1293 *
1294 * This may result in a call to nfqnl_nf_hook_drop() before
1295 * struct nfnl_queue_net was allocated.
1296 */
1297 if (!q)
1298 return;
1299
1300 for (i = 0; i < INSTANCE_BUCKETS; i++) {
1301 struct nfqnl_instance *inst;
1302 struct hlist_head *head = &q->instance_table[i];
1303
1304 hlist_for_each_entry_rcu(inst, head, hlist)
1305 nfqnl_flush(inst, NULL, 0);
1306 }
1307 }
1308
1309 static int
nfqnl_rcv_nl_event(struct notifier_block * this,unsigned long event,void * ptr)1310 nfqnl_rcv_nl_event(struct notifier_block *this,
1311 unsigned long event, void *ptr)
1312 {
1313 struct netlink_notify *n = ptr;
1314 struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
1315
1316 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
1317 int i;
1318
1319 /* destroy all instances for this portid */
1320 spin_lock(&q->instances_lock);
1321 for (i = 0; i < INSTANCE_BUCKETS; i++) {
1322 struct hlist_node *t2;
1323 struct nfqnl_instance *inst;
1324 struct hlist_head *head = &q->instance_table[i];
1325
1326 hlist_for_each_entry_safe(inst, t2, head, hlist) {
1327 if (n->portid == inst->peer_portid)
1328 __instance_destroy(inst);
1329 }
1330 }
1331 spin_unlock(&q->instances_lock);
1332 }
1333 return NOTIFY_DONE;
1334 }
1335
1336 static struct notifier_block nfqnl_rtnl_notifier = {
1337 .notifier_call = nfqnl_rcv_nl_event,
1338 };
1339
1340 static const struct nla_policy nfqa_vlan_policy[NFQA_VLAN_MAX + 1] = {
1341 [NFQA_VLAN_TCI] = { .type = NLA_U16},
1342 [NFQA_VLAN_PROTO] = { .type = NLA_U16},
1343 };
1344
1345 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
1346 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
1347 [NFQA_MARK] = { .type = NLA_U32 },
1348 [NFQA_PAYLOAD] = { .type = NLA_UNSPEC },
1349 [NFQA_CT] = { .type = NLA_UNSPEC },
1350 [NFQA_EXP] = { .type = NLA_UNSPEC },
1351 [NFQA_VLAN] = { .type = NLA_NESTED },
1352 [NFQA_PRIORITY] = { .type = NLA_U32 },
1353 };
1354
1355 static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
1356 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
1357 [NFQA_MARK] = { .type = NLA_U32 },
1358 [NFQA_PRIORITY] = { .type = NLA_U32 },
1359 };
1360
1361 static struct nfqnl_instance *
verdict_instance_lookup(struct nfnl_queue_net * q,u16 queue_num,u32 nlportid)1362 verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, u32 nlportid)
1363 {
1364 struct nfqnl_instance *queue;
1365
1366 queue = instance_lookup(q, queue_num);
1367 if (!queue)
1368 return ERR_PTR(-ENODEV);
1369
1370 if (queue->peer_portid != nlportid)
1371 return ERR_PTR(-EPERM);
1372
1373 return queue;
1374 }
1375
1376 static struct nfqnl_msg_verdict_hdr*
verdicthdr_get(const struct nlattr * const nfqa[])1377 verdicthdr_get(const struct nlattr * const nfqa[])
1378 {
1379 struct nfqnl_msg_verdict_hdr *vhdr;
1380 unsigned int verdict;
1381
1382 if (!nfqa[NFQA_VERDICT_HDR])
1383 return NULL;
1384
1385 vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
1386 verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
1387 if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
1388 return NULL;
1389 return vhdr;
1390 }
1391
nfq_id_after(unsigned int id,unsigned int max)1392 static int nfq_id_after(unsigned int id, unsigned int max)
1393 {
1394 return (int)(id - max) > 0;
1395 }
1396
nfqnl_recv_verdict_batch(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const nfqa[])1397 static int nfqnl_recv_verdict_batch(struct sk_buff *skb,
1398 const struct nfnl_info *info,
1399 const struct nlattr * const nfqa[])
1400 {
1401 struct nfnl_queue_net *q = nfnl_queue_pernet(info->net);
1402 u16 queue_num = ntohs(info->nfmsg->res_id);
1403 struct nf_queue_entry *entry, *tmp;
1404 struct nfqnl_msg_verdict_hdr *vhdr;
1405 struct nfqnl_instance *queue;
1406 unsigned int verdict, maxid;
1407 LIST_HEAD(batch_list);
1408
1409 queue = verdict_instance_lookup(q, queue_num,
1410 NETLINK_CB(skb).portid);
1411 if (IS_ERR(queue))
1412 return PTR_ERR(queue);
1413
1414 vhdr = verdicthdr_get(nfqa);
1415 if (!vhdr)
1416 return -EINVAL;
1417
1418 verdict = ntohl(vhdr->verdict);
1419 maxid = ntohl(vhdr->id);
1420
1421 spin_lock_bh(&queue->lock);
1422
1423 list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
1424 if (nfq_id_after(entry->id, maxid))
1425 break;
1426 __dequeue_entry(queue, entry);
1427 list_add_tail(&entry->list, &batch_list);
1428 }
1429
1430 spin_unlock_bh(&queue->lock);
1431
1432 if (list_empty(&batch_list))
1433 return -ENOENT;
1434
1435 list_for_each_entry_safe(entry, tmp, &batch_list, list) {
1436 if (nfqa[NFQA_MARK])
1437 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1438
1439 if (nfqa[NFQA_PRIORITY])
1440 entry->skb->priority = ntohl(nla_get_be32(nfqa[NFQA_PRIORITY]));
1441
1442 nfqnl_reinject(entry, verdict);
1443 }
1444 return 0;
1445 }
1446
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)1447 static struct nf_conn *nfqnl_ct_parse(const struct nfnl_ct_hook *nfnl_ct,
1448 const struct nlmsghdr *nlh,
1449 const struct nlattr * const nfqa[],
1450 struct nf_queue_entry *entry,
1451 enum ip_conntrack_info *ctinfo)
1452 {
1453 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
1454 struct nf_conn *ct;
1455
1456 ct = nf_ct_get(entry->skb, ctinfo);
1457 if (ct == NULL)
1458 return NULL;
1459
1460 if (nfnl_ct->parse(nfqa[NFQA_CT], ct) < 0)
1461 return NULL;
1462
1463 if (nfqa[NFQA_EXP])
1464 nfnl_ct->attach_expect(nfqa[NFQA_EXP], ct,
1465 NETLINK_CB(entry->skb).portid,
1466 nlmsg_report(nlh));
1467 return ct;
1468 #else
1469 return NULL;
1470 #endif
1471 }
1472
nfqa_parse_bridge(struct nf_queue_entry * entry,const struct nlattr * const nfqa[])1473 static int nfqa_parse_bridge(struct nf_queue_entry *entry,
1474 const struct nlattr * const nfqa[])
1475 {
1476 if (nfqa[NFQA_VLAN]) {
1477 struct nlattr *tb[NFQA_VLAN_MAX + 1];
1478 int err;
1479
1480 err = nla_parse_nested_deprecated(tb, NFQA_VLAN_MAX,
1481 nfqa[NFQA_VLAN],
1482 nfqa_vlan_policy, NULL);
1483 if (err < 0)
1484 return err;
1485
1486 if (!tb[NFQA_VLAN_TCI] || !tb[NFQA_VLAN_PROTO])
1487 return -EINVAL;
1488
1489 __vlan_hwaccel_put_tag(entry->skb,
1490 nla_get_be16(tb[NFQA_VLAN_PROTO]),
1491 ntohs(nla_get_be16(tb[NFQA_VLAN_TCI])));
1492 }
1493
1494 if (nfqa[NFQA_L2HDR]) {
1495 int mac_header_len = entry->skb->network_header -
1496 entry->skb->mac_header;
1497
1498 if (mac_header_len != nla_len(nfqa[NFQA_L2HDR]))
1499 return -EINVAL;
1500 else if (mac_header_len > 0)
1501 memcpy(skb_mac_header(entry->skb),
1502 nla_data(nfqa[NFQA_L2HDR]),
1503 mac_header_len);
1504 }
1505
1506 return 0;
1507 }
1508
nfqnl_recv_verdict(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const nfqa[])1509 static int nfqnl_recv_verdict(struct sk_buff *skb, const struct nfnl_info *info,
1510 const struct nlattr * const nfqa[])
1511 {
1512 struct nfnl_queue_net *q = nfnl_queue_pernet(info->net);
1513 u_int16_t queue_num = ntohs(info->nfmsg->res_id);
1514 const struct nfnl_ct_hook *nfnl_ct;
1515 struct nfqnl_msg_verdict_hdr *vhdr;
1516 enum ip_conntrack_info ctinfo;
1517 struct nfqnl_instance *queue;
1518 struct nf_queue_entry *entry;
1519 struct nf_conn *ct = NULL;
1520 unsigned int verdict;
1521 int err;
1522
1523 queue = verdict_instance_lookup(q, queue_num,
1524 NETLINK_CB(skb).portid);
1525 if (IS_ERR(queue))
1526 return PTR_ERR(queue);
1527
1528 vhdr = verdicthdr_get(nfqa);
1529 if (!vhdr)
1530 return -EINVAL;
1531
1532 verdict = ntohl(vhdr->verdict);
1533
1534 entry = find_dequeue_entry(queue, ntohl(vhdr->id), info->net);
1535 if (entry == NULL)
1536 return -ENOENT;
1537
1538 /* rcu lock already held from nfnl->call_rcu. */
1539 nfnl_ct = rcu_dereference(nfnl_ct_hook);
1540
1541 if (nfqa[NFQA_CT]) {
1542 if (nfnl_ct != NULL)
1543 ct = nfqnl_ct_parse(nfnl_ct, info->nlh, nfqa, entry,
1544 &ctinfo);
1545 }
1546
1547 if (entry->state.pf == PF_BRIDGE) {
1548 err = nfqa_parse_bridge(entry, nfqa);
1549 if (err < 0) {
1550 nfqnl_reinject(entry, NF_DROP);
1551 return err;
1552 }
1553 }
1554
1555 if (nfqa[NFQA_PAYLOAD]) {
1556 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
1557 int diff = payload_len - entry->skb->len;
1558
1559 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
1560 payload_len, entry, diff) < 0)
1561 verdict = NF_DROP;
1562
1563 if (ct && diff)
1564 nfnl_ct->seq_adjust(entry->skb, ct, ctinfo, diff);
1565 }
1566
1567 if (nfqa[NFQA_MARK])
1568 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1569
1570 if (nfqa[NFQA_PRIORITY])
1571 entry->skb->priority = ntohl(nla_get_be32(nfqa[NFQA_PRIORITY]));
1572
1573 nfqnl_reinject(entry, verdict);
1574 return 0;
1575 }
1576
nfqnl_recv_unsupp(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const cda[])1577 static int nfqnl_recv_unsupp(struct sk_buff *skb, const struct nfnl_info *info,
1578 const struct nlattr * const cda[])
1579 {
1580 return -ENOTSUPP;
1581 }
1582
1583 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1584 [NFQA_CFG_CMD] = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1585 [NFQA_CFG_PARAMS] = { .len = sizeof(struct nfqnl_msg_config_params) },
1586 [NFQA_CFG_QUEUE_MAXLEN] = { .type = NLA_U32 },
1587 [NFQA_CFG_MASK] = { .type = NLA_U32 },
1588 [NFQA_CFG_FLAGS] = { .type = NLA_U32 },
1589 };
1590
1591 static const struct nf_queue_handler nfqh = {
1592 .outfn = nfqnl_enqueue_packet,
1593 .nf_hook_drop = nfqnl_nf_hook_drop,
1594 };
1595
nfqnl_recv_config(struct sk_buff * skb,const struct nfnl_info * info,const struct nlattr * const nfqa[])1596 static int nfqnl_recv_config(struct sk_buff *skb, const struct nfnl_info *info,
1597 const struct nlattr * const nfqa[])
1598 {
1599 struct nfnl_queue_net *q = nfnl_queue_pernet(info->net);
1600 u_int16_t queue_num = ntohs(info->nfmsg->res_id);
1601 struct nfqnl_msg_config_cmd *cmd = NULL;
1602 struct nfqnl_instance *queue;
1603 __u32 flags = 0, mask = 0;
1604
1605 WARN_ON_ONCE(!lockdep_nfnl_is_held(NFNL_SUBSYS_QUEUE));
1606
1607 if (nfqa[NFQA_CFG_CMD]) {
1608 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1609
1610 /* Obsolete commands without queue context */
1611 switch (cmd->command) {
1612 case NFQNL_CFG_CMD_PF_BIND: return 0;
1613 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1614 }
1615 }
1616
1617 /* Check if we support these flags in first place, dependencies should
1618 * be there too not to break atomicity.
1619 */
1620 if (nfqa[NFQA_CFG_FLAGS]) {
1621 if (!nfqa[NFQA_CFG_MASK]) {
1622 /* A mask is needed to specify which flags are being
1623 * changed.
1624 */
1625 return -EINVAL;
1626 }
1627
1628 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1629 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1630
1631 if (flags >= NFQA_CFG_F_MAX)
1632 return -EOPNOTSUPP;
1633
1634 #if !IS_ENABLED(CONFIG_NETWORK_SECMARK)
1635 if (flags & mask & NFQA_CFG_F_SECCTX)
1636 return -EOPNOTSUPP;
1637 #endif
1638 if ((flags & mask & NFQA_CFG_F_CONNTRACK) &&
1639 !rcu_access_pointer(nfnl_ct_hook)) {
1640 #ifdef CONFIG_MODULES
1641 nfnl_unlock(NFNL_SUBSYS_QUEUE);
1642 request_module("ip_conntrack_netlink");
1643 nfnl_lock(NFNL_SUBSYS_QUEUE);
1644 if (rcu_access_pointer(nfnl_ct_hook))
1645 return -EAGAIN;
1646 #endif
1647 return -EOPNOTSUPP;
1648 }
1649 }
1650
1651 /* Lookup queue under RCU. After peer_portid check (or for new queue
1652 * in BIND case), the queue is owned by the socket sending this message.
1653 * A socket cannot simultaneously send a message and close, so while
1654 * processing this CONFIG message, nfqnl_rcv_nl_event() (triggered by
1655 * socket close) cannot destroy this queue. Safe to use without RCU.
1656 */
1657 rcu_read_lock();
1658 queue = instance_lookup(q, queue_num);
1659 if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1660 rcu_read_unlock();
1661 return -EPERM;
1662 }
1663 rcu_read_unlock();
1664
1665 if (cmd != NULL) {
1666 switch (cmd->command) {
1667 case NFQNL_CFG_CMD_BIND:
1668 if (queue)
1669 return -EBUSY;
1670 queue = instance_create(q, queue_num, NETLINK_CB(skb).portid);
1671 if (IS_ERR(queue))
1672 return PTR_ERR(queue);
1673 break;
1674 case NFQNL_CFG_CMD_UNBIND:
1675 if (!queue)
1676 return -ENODEV;
1677 instance_destroy(q, queue);
1678 return 0;
1679 case NFQNL_CFG_CMD_PF_BIND:
1680 case NFQNL_CFG_CMD_PF_UNBIND:
1681 break;
1682 default:
1683 return -EOPNOTSUPP;
1684 }
1685 }
1686
1687 if (!queue)
1688 return -ENODEV;
1689
1690 if (nfqa[NFQA_CFG_PARAMS]) {
1691 struct nfqnl_msg_config_params *params =
1692 nla_data(nfqa[NFQA_CFG_PARAMS]);
1693
1694 nfqnl_set_mode(queue, params->copy_mode,
1695 ntohl(params->copy_range));
1696 }
1697
1698 if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1699 __be32 *queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1700
1701 spin_lock_bh(&queue->lock);
1702 queue->queue_maxlen = ntohl(*queue_maxlen);
1703 spin_unlock_bh(&queue->lock);
1704 }
1705
1706 if (nfqa[NFQA_CFG_FLAGS]) {
1707 spin_lock_bh(&queue->lock);
1708 queue->flags &= ~mask;
1709 queue->flags |= flags & mask;
1710 spin_unlock_bh(&queue->lock);
1711 }
1712
1713 return 0;
1714 }
1715
1716 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1717 [NFQNL_MSG_PACKET] = {
1718 .call = nfqnl_recv_unsupp,
1719 .type = NFNL_CB_RCU,
1720 .attr_count = NFQA_MAX,
1721 },
1722 [NFQNL_MSG_VERDICT] = {
1723 .call = nfqnl_recv_verdict,
1724 .type = NFNL_CB_RCU,
1725 .attr_count = NFQA_MAX,
1726 .policy = nfqa_verdict_policy
1727 },
1728 [NFQNL_MSG_CONFIG] = {
1729 .call = nfqnl_recv_config,
1730 .type = NFNL_CB_MUTEX,
1731 .attr_count = NFQA_CFG_MAX,
1732 .policy = nfqa_cfg_policy
1733 },
1734 [NFQNL_MSG_VERDICT_BATCH] = {
1735 .call = nfqnl_recv_verdict_batch,
1736 .type = NFNL_CB_RCU,
1737 .attr_count = NFQA_MAX,
1738 .policy = nfqa_verdict_batch_policy
1739 },
1740 };
1741
1742 static const struct nfnetlink_subsystem nfqnl_subsys = {
1743 .name = "nf_queue",
1744 .subsys_id = NFNL_SUBSYS_QUEUE,
1745 .cb_count = NFQNL_MSG_MAX,
1746 .cb = nfqnl_cb,
1747 };
1748
1749 #ifdef CONFIG_PROC_FS
1750 struct iter_state {
1751 struct seq_net_private p;
1752 unsigned int bucket;
1753 };
1754
get_first(struct seq_file * seq)1755 static struct hlist_node *get_first(struct seq_file *seq)
1756 {
1757 struct iter_state *st = seq->private;
1758 struct net *net;
1759 struct nfnl_queue_net *q;
1760
1761 if (!st)
1762 return NULL;
1763
1764 net = seq_file_net(seq);
1765 q = nfnl_queue_pernet(net);
1766 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1767 if (!hlist_empty(&q->instance_table[st->bucket]))
1768 return q->instance_table[st->bucket].first;
1769 }
1770 return NULL;
1771 }
1772
get_next(struct seq_file * seq,struct hlist_node * h)1773 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1774 {
1775 struct iter_state *st = seq->private;
1776 struct net *net = seq_file_net(seq);
1777
1778 h = h->next;
1779 while (!h) {
1780 struct nfnl_queue_net *q;
1781
1782 if (++st->bucket >= INSTANCE_BUCKETS)
1783 return NULL;
1784
1785 q = nfnl_queue_pernet(net);
1786 h = q->instance_table[st->bucket].first;
1787 }
1788 return h;
1789 }
1790
get_idx(struct seq_file * seq,loff_t pos)1791 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1792 {
1793 struct hlist_node *head;
1794 head = get_first(seq);
1795
1796 if (head)
1797 while (pos && (head = get_next(seq, head)))
1798 pos--;
1799 return pos ? NULL : head;
1800 }
1801
seq_start(struct seq_file * s,loff_t * pos)1802 static void *seq_start(struct seq_file *s, loff_t *pos)
1803 __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1804 {
1805 spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1806 return get_idx(s, *pos);
1807 }
1808
seq_next(struct seq_file * s,void * v,loff_t * pos)1809 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1810 {
1811 (*pos)++;
1812 return get_next(s, v);
1813 }
1814
seq_stop(struct seq_file * s,void * v)1815 static void seq_stop(struct seq_file *s, void *v)
1816 __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1817 {
1818 spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1819 }
1820
seq_show(struct seq_file * s,void * v)1821 static int seq_show(struct seq_file *s, void *v)
1822 {
1823 const struct nfqnl_instance *inst = v;
1824
1825 seq_printf(s, "%5u %6u %5u %1u %5u %5u %5u %8u %2d\n",
1826 inst->queue_num,
1827 inst->peer_portid, inst->queue_total,
1828 inst->copy_mode, inst->copy_range,
1829 inst->queue_dropped, inst->queue_user_dropped,
1830 inst->id_sequence, 1);
1831 return 0;
1832 }
1833
1834 static const struct seq_operations nfqnl_seq_ops = {
1835 .start = seq_start,
1836 .next = seq_next,
1837 .stop = seq_stop,
1838 .show = seq_show,
1839 };
1840 #endif /* PROC_FS */
1841
nfnl_queue_net_init(struct net * net)1842 static int __net_init nfnl_queue_net_init(struct net *net)
1843 {
1844 unsigned int i;
1845 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1846
1847 for (i = 0; i < INSTANCE_BUCKETS; i++)
1848 INIT_HLIST_HEAD(&q->instance_table[i]);
1849
1850 spin_lock_init(&q->instances_lock);
1851
1852 #ifdef CONFIG_PROC_FS
1853 if (!proc_create_net("nfnetlink_queue", 0440, net->nf.proc_netfilter,
1854 &nfqnl_seq_ops, sizeof(struct iter_state)))
1855 return -ENOMEM;
1856 #endif
1857 return 0;
1858 }
1859
nfnl_queue_net_exit(struct net * net)1860 static void __net_exit nfnl_queue_net_exit(struct net *net)
1861 {
1862 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1863 unsigned int i;
1864
1865 #ifdef CONFIG_PROC_FS
1866 remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1867 #endif
1868 for (i = 0; i < INSTANCE_BUCKETS; i++)
1869 WARN_ON_ONCE(!hlist_empty(&q->instance_table[i]));
1870 }
1871
1872 static struct pernet_operations nfnl_queue_net_ops = {
1873 .init = nfnl_queue_net_init,
1874 .exit = nfnl_queue_net_exit,
1875 .id = &nfnl_queue_net_id,
1876 .size = sizeof(struct nfnl_queue_net),
1877 };
1878
nfnetlink_queue_init(void)1879 static int __init nfnetlink_queue_init(void)
1880 {
1881 int status;
1882
1883 status = rhashtable_init(&nfqnl_packet_map, &nfqnl_rhashtable_params);
1884 if (status < 0)
1885 return status;
1886
1887 status = register_pernet_subsys(&nfnl_queue_net_ops);
1888 if (status < 0) {
1889 pr_err("failed to register pernet ops\n");
1890 goto cleanup_rhashtable;
1891 }
1892
1893 netlink_register_notifier(&nfqnl_rtnl_notifier);
1894 status = nfnetlink_subsys_register(&nfqnl_subsys);
1895 if (status < 0) {
1896 pr_err("failed to create netlink socket\n");
1897 goto cleanup_netlink_notifier;
1898 }
1899
1900 status = register_netdevice_notifier(&nfqnl_dev_notifier);
1901 if (status < 0) {
1902 pr_err("failed to register netdevice notifier\n");
1903 goto cleanup_netlink_subsys;
1904 }
1905
1906 nf_register_queue_handler(&nfqh);
1907
1908 return status;
1909
1910 cleanup_netlink_subsys:
1911 nfnetlink_subsys_unregister(&nfqnl_subsys);
1912 cleanup_netlink_notifier:
1913 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1914 unregister_pernet_subsys(&nfnl_queue_net_ops);
1915 cleanup_rhashtable:
1916 rhashtable_destroy(&nfqnl_packet_map);
1917 return status;
1918 }
1919
nfnetlink_queue_fini(void)1920 static void __exit nfnetlink_queue_fini(void)
1921 {
1922 nf_unregister_queue_handler();
1923 unregister_netdevice_notifier(&nfqnl_dev_notifier);
1924 nfnetlink_subsys_unregister(&nfqnl_subsys);
1925 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1926 unregister_pernet_subsys(&nfnl_queue_net_ops);
1927
1928 rhashtable_destroy(&nfqnl_packet_map);
1929
1930 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1931 }
1932
1933 MODULE_DESCRIPTION("netfilter packet queue handler");
1934 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1935 MODULE_LICENSE("GPL");
1936 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1937
1938 module_init(nfnetlink_queue_init);
1939 module_exit(nfnetlink_queue_fini);
1940