?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge tag 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mst/vhost
[cris-mirror.git] / net / netfilter / nfnetlink_queue.c
blob8bba23160a68fde04624eac306731ebf64fb8600
1 /*
2 * This is a module which is used for queueing packets and communicating with
3 * userspace via nfnetlink.
4 *
5 * (C) 2005 by Harald Welte <laforge@netfilter.org>
6 * (C) 2007 by Patrick McHardy <kaber@trash.net>
7 *
8 * Based on the old ipv4-only ip_queue.c:
9 * (C) 2000-2002 James Morris <jmorris@intercode.com.au>
10 * (C) 2003-2005 Netfilter Core Team <coreteam@netfilter.org>
11 *
12 * This program is free software; you can redistribute it and/or modify
13 * it under the terms of the GNU General Public License version 2 as
14 * published by the Free Software Foundation.
15 *
16 */
17 #include <linux/module.h>
18 #include <linux/skbuff.h>
19 #include <linux/init.h>
20 #include <linux/spinlock.h>
21 #include <linux/slab.h>
22 #include <linux/notifier.h>
23 #include <linux/netdevice.h>
24 #include <linux/netfilter.h>
25 #include <linux/proc_fs.h>
26 #include <linux/netfilter_ipv4.h>
27 #include <linux/netfilter_ipv6.h>
28 #include <linux/netfilter_bridge.h>
29 #include <linux/netfilter/nfnetlink.h>
30 #include <linux/netfilter/nfnetlink_queue.h>
31 #include <linux/netfilter/nf_conntrack_common.h>
32 #include <linux/list.h>
33 #include <net/sock.h>
34 #include <net/tcp_states.h>
35 #include <net/netfilter/nf_queue.h>
36 #include <net/netns/generic.h>
37
38 #include <linux/atomic.h>
39
40 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
41 #include "../bridge/br_private.h"
42 #endif
43
44 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
45 #include <net/netfilter/nf_conntrack.h>
46 #endif
47
48 #define NFQNL_QMAX_DEFAULT 1024
49
50 /* We're using struct nlattr which has 16bit nla_len. Note that nla_len
51 * includes the header length. Thus, the maximum packet length that we
52 * support is 65531 bytes. We send truncated packets if the specified length
53 * is larger than that. Userspace can check for presence of NFQA_CAP_LEN
54 * attribute to detect truncation.
55 */
56 #define NFQNL_MAX_COPY_RANGE (0xffff - NLA_HDRLEN)
57
58 struct nfqnl_instance {
59 struct hlist_node hlist; /* global list of queues */
60 struct rcu_head rcu;
61
62 u32 peer_portid;
63 unsigned int queue_maxlen;
64 unsigned int copy_range;
65 unsigned int queue_dropped;
66 unsigned int queue_user_dropped;
67
68
69 u_int16_t queue_num; /* number of this queue */
70 u_int8_t copy_mode;
71 u_int32_t flags; /* Set using NFQA_CFG_FLAGS */
72 /*
73 * Following fields are dirtied for each queued packet,
74 * keep them in same cache line if possible.
75 */
76 spinlock_t lock ____cacheline_aligned_in_smp;
77 unsigned int queue_total;
78 unsigned int id_sequence; /* 'sequence' of pkt ids */
79 struct list_head queue_list; /* packets in queue */
80 };
81
82 typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
83
84 static unsigned int nfnl_queue_net_id __read_mostly;
85
86 #define INSTANCE_BUCKETS 16
87 struct nfnl_queue_net {
88 spinlock_t instances_lock;
89 struct hlist_head instance_table[INSTANCE_BUCKETS];
90 };
91
92 static struct nfnl_queue_net *nfnl_queue_pernet(struct net *net)
93 {
94 return net_generic(net, nfnl_queue_net_id);
95 }
96
97 static inline u_int8_t instance_hashfn(u_int16_t queue_num)
98 {
99 return ((queue_num >> 8) ^ queue_num) % INSTANCE_BUCKETS;
100 }
101
102 static struct nfqnl_instance *
103 instance_lookup(struct nfnl_queue_net *q, u_int16_t queue_num)
104 {
105 struct hlist_head *head;
106 struct nfqnl_instance *inst;
107
108 head = &q->instance_table[instance_hashfn(queue_num)];
109 hlist_for_each_entry_rcu(inst, head, hlist) {
110 if (inst->queue_num == queue_num)
111 return inst;
112 }
113 return NULL;
114 }
115
116 static struct nfqnl_instance *
117 instance_create(struct nfnl_queue_net *q, u_int16_t queue_num, u32 portid)
118 {
119 struct nfqnl_instance *inst;
120 unsigned int h;
121 int err;
122
123 spin_lock(&q->instances_lock);
124 if (instance_lookup(q, queue_num)) {
125 err = -EEXIST;
126 goto out_unlock;
127 }
128
129 inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
130 if (!inst) {
131 err = -ENOMEM;
132 goto out_unlock;
133 }
134
135 inst->queue_num = queue_num;
136 inst->peer_portid = portid;
137 inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
138 inst->copy_range = NFQNL_MAX_COPY_RANGE;
139 inst->copy_mode = NFQNL_COPY_NONE;
140 spin_lock_init(&inst->lock);
141 INIT_LIST_HEAD(&inst->queue_list);
142
143 if (!try_module_get(THIS_MODULE)) {
144 err = -EAGAIN;
145 goto out_free;
146 }
147
148 h = instance_hashfn(queue_num);
149 hlist_add_head_rcu(&inst->hlist, &q->instance_table[h]);
150
151 spin_unlock(&q->instances_lock);
152
153 return inst;
154
155 out_free:
156 kfree(inst);
157 out_unlock:
158 spin_unlock(&q->instances_lock);
159 return ERR_PTR(err);
160 }
161
162 static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
163 unsigned long data);
164
165 static void
166 instance_destroy_rcu(struct rcu_head *head)
167 {
168 struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
169 rcu);
170
171 nfqnl_flush(inst, NULL, 0);
172 kfree(inst);
173 module_put(THIS_MODULE);
174 }
175
176 static void
177 __instance_destroy(struct nfqnl_instance *inst)
178 {
179 hlist_del_rcu(&inst->hlist);
180 call_rcu(&inst->rcu, instance_destroy_rcu);
181 }
182
183 static void
184 instance_destroy(struct nfnl_queue_net *q, struct nfqnl_instance *inst)
185 {
186 spin_lock(&q->instances_lock);
187 __instance_destroy(inst);
188 spin_unlock(&q->instances_lock);
189 }
190
191 static inline void
192 __enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
193 {
194 list_add_tail(&entry->list, &queue->queue_list);
195 queue->queue_total++;
196 }
197
198 static void
199 __dequeue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
200 {
201 list_del(&entry->list);
202 queue->queue_total--;
203 }
204
205 static struct nf_queue_entry *
206 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
207 {
208 struct nf_queue_entry *entry = NULL, *i;
209
210 spin_lock_bh(&queue->lock);
211
212 list_for_each_entry(i, &queue->queue_list, list) {
213 if (i->id == id) {
214 entry = i;
215 break;
216 }
217 }
218
219 if (entry)
220 __dequeue_entry(queue, entry);
221
222 spin_unlock_bh(&queue->lock);
223
224 return entry;
225 }
226
227 static void
228 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
229 {
230 struct nf_queue_entry *entry, *next;
231
232 spin_lock_bh(&queue->lock);
233 list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
234 if (!cmpfn || cmpfn(entry, data)) {
235 list_del(&entry->list);
236 queue->queue_total--;
237 nf_reinject(entry, NF_DROP);
238 }
239 }
240 spin_unlock_bh(&queue->lock);
241 }
242
243 static int
244 nfqnl_put_packet_info(struct sk_buff *nlskb, struct sk_buff *packet,
245 bool csum_verify)
246 {
247 __u32 flags = 0;
248
249 if (packet->ip_summed == CHECKSUM_PARTIAL)
250 flags = NFQA_SKB_CSUMNOTREADY;
251 else if (csum_verify)
252 flags = NFQA_SKB_CSUM_NOTVERIFIED;
253
254 if (skb_is_gso(packet))
255 flags |= NFQA_SKB_GSO;
256
257 return flags ? nla_put_be32(nlskb, NFQA_SKB_INFO, htonl(flags)) : 0;
258 }
259
260 static int nfqnl_put_sk_uidgid(struct sk_buff *skb, struct sock *sk)
261 {
262 const struct cred *cred;
263
264 if (!sk_fullsock(sk))
265 return 0;
266
267 read_lock_bh(&sk->sk_callback_lock);
268 if (sk->sk_socket && sk->sk_socket->file) {
269 cred = sk->sk_socket->file->f_cred;
270 if (nla_put_be32(skb, NFQA_UID,
271 htonl(from_kuid_munged(&init_user_ns, cred->fsuid))))
272 goto nla_put_failure;
273 if (nla_put_be32(skb, NFQA_GID,
274 htonl(from_kgid_munged(&init_user_ns, cred->fsgid))))
275 goto nla_put_failure;
276 }
277 read_unlock_bh(&sk->sk_callback_lock);
278 return 0;
279
280 nla_put_failure:
281 read_unlock_bh(&sk->sk_callback_lock);
282 return -1;
283 }
284
285 static u32 nfqnl_get_sk_secctx(struct sk_buff *skb, char **secdata)
286 {
287 u32 seclen = 0;
288 #if IS_ENABLED(CONFIG_NETWORK_SECMARK)
289 if (!skb || !sk_fullsock(skb->sk))
290 return 0;
291
292 read_lock_bh(&skb->sk->sk_callback_lock);
293
294 if (skb->secmark)
295 security_secid_to_secctx(skb->secmark, secdata, &seclen);
296
297 read_unlock_bh(&skb->sk->sk_callback_lock);
298 #endif
299 return seclen;
300 }
301
302 static u32 nfqnl_get_bridge_size(struct nf_queue_entry *entry)
303 {
304 struct sk_buff *entskb = entry->skb;
305 u32 nlalen = 0;
306
307 if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb))
308 return 0;
309
310 if (skb_vlan_tag_present(entskb))
311 nlalen += nla_total_size(nla_total_size(sizeof(__be16)) +
312 nla_total_size(sizeof(__be16)));
313
314 if (entskb->network_header > entskb->mac_header)
315 nlalen += nla_total_size((entskb->network_header -
316 entskb->mac_header));
317
318 return nlalen;
319 }
320
321 static int nfqnl_put_bridge(struct nf_queue_entry *entry, struct sk_buff *skb)
322 {
323 struct sk_buff *entskb = entry->skb;
324
325 if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb))
326 return 0;
327
328 if (skb_vlan_tag_present(entskb)) {
329 struct nlattr *nest;
330
331 nest = nla_nest_start(skb, NFQA_VLAN | NLA_F_NESTED);
332 if (!nest)
333 goto nla_put_failure;
334
335 if (nla_put_be16(skb, NFQA_VLAN_TCI, htons(entskb->vlan_tci)) ||
336 nla_put_be16(skb, NFQA_VLAN_PROTO, entskb->vlan_proto))
337 goto nla_put_failure;
338
339 nla_nest_end(skb, nest);
340 }
341
342 if (entskb->mac_header < entskb->network_header) {
343 int len = (int)(entskb->network_header - entskb->mac_header);
344
345 if (nla_put(skb, NFQA_L2HDR, len, skb_mac_header(entskb)))
346 goto nla_put_failure;
347 }
348
349 return 0;
350
351 nla_put_failure:
352 return -1;
353 }
354
355 static struct sk_buff *
356 nfqnl_build_packet_message(struct net *net, struct nfqnl_instance *queue,
357 struct nf_queue_entry *entry,
358 __be32 **packet_id_ptr)
359 {
360 size_t size;
361 size_t data_len = 0, cap_len = 0;
362 unsigned int hlen = 0;
363 struct sk_buff *skb;
364 struct nlattr *nla;
365 struct nfqnl_msg_packet_hdr *pmsg;
366 struct nlmsghdr *nlh;
367 struct nfgenmsg *nfmsg;
368 struct sk_buff *entskb = entry->skb;
369 struct net_device *indev;
370 struct net_device *outdev;
371 struct nf_conn *ct = NULL;
372 enum ip_conntrack_info uninitialized_var(ctinfo);
373 struct nfnl_ct_hook *nfnl_ct;
374 bool csum_verify;
375 char *secdata = NULL;
376 u32 seclen = 0;
377
378 size = nlmsg_total_size(sizeof(struct nfgenmsg))
379 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
380 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
381 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
382 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
383 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
384 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
385 #endif
386 + nla_total_size(sizeof(u_int32_t)) /* mark */
387 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
388 + nla_total_size(sizeof(u_int32_t)) /* skbinfo */
389 + nla_total_size(sizeof(u_int32_t)); /* cap_len */
390
391 if (entskb->tstamp)
392 size += nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
393
394 size += nfqnl_get_bridge_size(entry);
395
396 if (entry->state.hook <= NF_INET_FORWARD ||
397 (entry->state.hook == NF_INET_POST_ROUTING && entskb->sk == NULL))
398 csum_verify = !skb_csum_unnecessary(entskb);
399 else
400 csum_verify = false;
401
402 outdev = entry->state.out;
403
404 switch ((enum nfqnl_config_mode)READ_ONCE(queue->copy_mode)) {
405 case NFQNL_COPY_META:
406 case NFQNL_COPY_NONE:
407 break;
408
409 case NFQNL_COPY_PACKET:
410 if (!(queue->flags & NFQA_CFG_F_GSO) &&
411 entskb->ip_summed == CHECKSUM_PARTIAL &&
412 skb_checksum_help(entskb))
413 return NULL;
414
415 data_len = READ_ONCE(queue->copy_range);
416 if (data_len > entskb->len)
417 data_len = entskb->len;
418
419 hlen = skb_zerocopy_headlen(entskb);
420 hlen = min_t(unsigned int, hlen, data_len);
421 size += sizeof(struct nlattr) + hlen;
422 cap_len = entskb->len;
423 break;
424 }
425
426 nfnl_ct = rcu_dereference(nfnl_ct_hook);
427
428 if (queue->flags & NFQA_CFG_F_CONNTRACK) {
429 if (nfnl_ct != NULL) {
430 ct = nfnl_ct->get_ct(entskb, &ctinfo);
431 if (ct != NULL)
432 size += nfnl_ct->build_size(ct);
433 }
434 }
435
436 if (queue->flags & NFQA_CFG_F_UID_GID) {
437 size += (nla_total_size(sizeof(u_int32_t)) /* uid */
438 + nla_total_size(sizeof(u_int32_t))); /* gid */
439 }
440
441 if ((queue->flags & NFQA_CFG_F_SECCTX) && entskb->sk) {
442 seclen = nfqnl_get_sk_secctx(entskb, &secdata);
443 if (seclen)
444 size += nla_total_size(seclen);
445 }
446
447 skb = alloc_skb(size, GFP_ATOMIC);
448 if (!skb) {
449 skb_tx_error(entskb);
450 goto nlmsg_failure;
451 }
452
453 nlh = nlmsg_put(skb, 0, 0,
454 nfnl_msg_type(NFNL_SUBSYS_QUEUE, NFQNL_MSG_PACKET),
455 sizeof(struct nfgenmsg), 0);
456 if (!nlh) {
457 skb_tx_error(entskb);
458 kfree_skb(skb);
459 goto nlmsg_failure;
460 }
461 nfmsg = nlmsg_data(nlh);
462 nfmsg->nfgen_family = entry->state.pf;
463 nfmsg->version = NFNETLINK_V0;
464 nfmsg->res_id = htons(queue->queue_num);
465
466 nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
467 pmsg = nla_data(nla);
468 pmsg->hw_protocol = entskb->protocol;
469 pmsg->hook = entry->state.hook;
470 *packet_id_ptr = &pmsg->packet_id;
471
472 indev = entry->state.in;
473 if (indev) {
474 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
475 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
476 goto nla_put_failure;
477 #else
478 if (entry->state.pf == PF_BRIDGE) {
479 /* Case 1: indev is physical input device, we need to
480 * look for bridge group (when called from
481 * netfilter_bridge) */
482 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
483 htonl(indev->ifindex)) ||
484 /* this is the bridge group "brX" */
485 /* rcu_read_lock()ed by __nf_queue */
486 nla_put_be32(skb, NFQA_IFINDEX_INDEV,
487 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
488 goto nla_put_failure;
489 } else {
490 int physinif;
491
492 /* Case 2: indev is bridge group, we need to look for
493 * physical device (when called from ipv4) */
494 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
495 htonl(indev->ifindex)))
496 goto nla_put_failure;
497
498 physinif = nf_bridge_get_physinif(entskb);
499 if (physinif &&
500 nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
501 htonl(physinif)))
502 goto nla_put_failure;
503 }
504 #endif
505 }
506
507 if (outdev) {
508 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
509 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
510 goto nla_put_failure;
511 #else
512 if (entry->state.pf == PF_BRIDGE) {
513 /* Case 1: outdev is physical output device, we need to
514 * look for bridge group (when called from
515 * netfilter_bridge) */
516 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
517 htonl(outdev->ifindex)) ||
518 /* this is the bridge group "brX" */
519 /* rcu_read_lock()ed by __nf_queue */
520 nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
521 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
522 goto nla_put_failure;
523 } else {
524 int physoutif;
525
526 /* Case 2: outdev is bridge group, we need to look for
527 * physical output device (when called from ipv4) */
528 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
529 htonl(outdev->ifindex)))
530 goto nla_put_failure;
531
532 physoutif = nf_bridge_get_physoutif(entskb);
533 if (physoutif &&
534 nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
535 htonl(physoutif)))
536 goto nla_put_failure;
537 }
538 #endif
539 }
540
541 if (entskb->mark &&
542 nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
543 goto nla_put_failure;
544
545 if (indev && entskb->dev &&
546 entskb->mac_header != entskb->network_header) {
547 struct nfqnl_msg_packet_hw phw;
548 int len;
549
550 memset(&phw, 0, sizeof(phw));
551 len = dev_parse_header(entskb, phw.hw_addr);
552 if (len) {
553 phw.hw_addrlen = htons(len);
554 if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
555 goto nla_put_failure;
556 }
557 }
558
559 if (nfqnl_put_bridge(entry, skb) < 0)
560 goto nla_put_failure;
561
562 if (entskb->tstamp) {
563 struct nfqnl_msg_packet_timestamp ts;
564 struct timespec64 kts = ktime_to_timespec64(entskb->tstamp);
565
566 ts.sec = cpu_to_be64(kts.tv_sec);
567 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
568
569 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
570 goto nla_put_failure;
571 }
572
573 if ((queue->flags & NFQA_CFG_F_UID_GID) && entskb->sk &&
574 nfqnl_put_sk_uidgid(skb, entskb->sk) < 0)
575 goto nla_put_failure;
576
577 if (seclen && nla_put(skb, NFQA_SECCTX, seclen, secdata))
578 goto nla_put_failure;
579
580 if (ct && nfnl_ct->build(skb, ct, ctinfo, NFQA_CT, NFQA_CT_INFO) < 0)
581 goto nla_put_failure;
582
583 if (cap_len > data_len &&
584 nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
585 goto nla_put_failure;
586
587 if (nfqnl_put_packet_info(skb, entskb, csum_verify))
588 goto nla_put_failure;
589
590 if (data_len) {
591 struct nlattr *nla;
592
593 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
594 goto nla_put_failure;
595
596 nla = skb_put(skb, sizeof(*nla));
597 nla->nla_type = NFQA_PAYLOAD;
598 nla->nla_len = nla_attr_size(data_len);
599
600 if (skb_zerocopy(skb, entskb, data_len, hlen))
601 goto nla_put_failure;
602 }
603
604 nlh->nlmsg_len = skb->len;
605 if (seclen)
606 security_release_secctx(secdata, seclen);
607 return skb;
608
609 nla_put_failure:
610 skb_tx_error(entskb);
611 kfree_skb(skb);
612 net_err_ratelimited("nf_queue: error creating packet message\n");
613 nlmsg_failure:
614 if (seclen)
615 security_release_secctx(secdata, seclen);
616 return NULL;
617 }
618
619 static bool nf_ct_drop_unconfirmed(const struct nf_queue_entry *entry)
620 {
621 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
622 static const unsigned long flags = IPS_CONFIRMED | IPS_DYING;
623 const struct nf_conn *ct = (void *)skb_nfct(entry->skb);
624
625 if (ct && ((ct->status & flags) == IPS_DYING))
626 return true;
627 #endif
628 return false;
629 }
630
631 static int
632 __nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
633 struct nf_queue_entry *entry)
634 {
635 struct sk_buff *nskb;
636 int err = -ENOBUFS;
637 __be32 *packet_id_ptr;
638 int failopen = 0;
639
640 nskb = nfqnl_build_packet_message(net, queue, entry, &packet_id_ptr);
641 if (nskb == NULL) {
642 err = -ENOMEM;
643 goto err_out;
644 }
645 spin_lock_bh(&queue->lock);
646
647 if (nf_ct_drop_unconfirmed(entry))
648 goto err_out_free_nskb;
649
650 if (queue->queue_total >= queue->queue_maxlen) {
651 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
652 failopen = 1;
653 err = 0;
654 } else {
655 queue->queue_dropped++;
656 net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
657 queue->queue_total);
658 }
659 goto err_out_free_nskb;
660 }
661 entry->id = ++queue->id_sequence;
662 *packet_id_ptr = htonl(entry->id);
663
664 /* nfnetlink_unicast will either free the nskb or add it to a socket */
665 err = nfnetlink_unicast(nskb, net, queue->peer_portid, MSG_DONTWAIT);
666 if (err < 0) {
667 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
668 failopen = 1;
669 err = 0;
670 } else {
671 queue->queue_user_dropped++;
672 }
673 goto err_out_unlock;
674 }
675
676 __enqueue_entry(queue, entry);
677
678 spin_unlock_bh(&queue->lock);
679 return 0;
680
681 err_out_free_nskb:
682 kfree_skb(nskb);
683 err_out_unlock:
684 spin_unlock_bh(&queue->lock);
685 if (failopen)
686 nf_reinject(entry, NF_ACCEPT);
687 err_out:
688 return err;
689 }
690
691 static struct nf_queue_entry *
692 nf_queue_entry_dup(struct nf_queue_entry *e)
693 {
694 struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
695 if (entry)
696 nf_queue_entry_get_refs(entry);
697 return entry;
698 }
699
700 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
701 /* When called from bridge netfilter, skb->data must point to MAC header
702 * before calling skb_gso_segment(). Else, original MAC header is lost
703 * and segmented skbs will be sent to wrong destination.
704 */
705 static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
706 {
707 if (skb->nf_bridge)
708 __skb_push(skb, skb->network_header - skb->mac_header);
709 }
710
711 static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
712 {
713 if (skb->nf_bridge)
714 __skb_pull(skb, skb->network_header - skb->mac_header);
715 }
716 #else
717 #define nf_bridge_adjust_skb_data(s) do {} while (0)
718 #define nf_bridge_adjust_segmented_data(s) do {} while (0)
719 #endif
720
721 static void free_entry(struct nf_queue_entry *entry)
722 {
723 nf_queue_entry_release_refs(entry);
724 kfree(entry);
725 }
726
727 static int
728 __nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
729 struct sk_buff *skb, struct nf_queue_entry *entry)
730 {
731 int ret = -ENOMEM;
732 struct nf_queue_entry *entry_seg;
733
734 nf_bridge_adjust_segmented_data(skb);
735
736 if (skb->next == NULL) { /* last packet, no need to copy entry */
737 struct sk_buff *gso_skb = entry->skb;
738 entry->skb = skb;
739 ret = __nfqnl_enqueue_packet(net, queue, entry);
740 if (ret)
741 entry->skb = gso_skb;
742 return ret;
743 }
744
745 skb->next = NULL;
746
747 entry_seg = nf_queue_entry_dup(entry);
748 if (entry_seg) {
749 entry_seg->skb = skb;
750 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
751 if (ret)
752 free_entry(entry_seg);
753 }
754 return ret;
755 }
756
757 static int
758 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
759 {
760 unsigned int queued;
761 struct nfqnl_instance *queue;
762 struct sk_buff *skb, *segs;
763 int err = -ENOBUFS;
764 struct net *net = entry->state.net;
765 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
766
767 /* rcu_read_lock()ed by nf_hook_thresh */
768 queue = instance_lookup(q, queuenum);
769 if (!queue)
770 return -ESRCH;
771
772 if (queue->copy_mode == NFQNL_COPY_NONE)
773 return -EINVAL;
774
775 skb = entry->skb;
776
777 switch (entry->state.pf) {
778 case NFPROTO_IPV4:
779 skb->protocol = htons(ETH_P_IP);
780 break;
781 case NFPROTO_IPV6:
782 skb->protocol = htons(ETH_P_IPV6);
783 break;
784 }
785
786 if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(skb))
787 return __nfqnl_enqueue_packet(net, queue, entry);
788
789 nf_bridge_adjust_skb_data(skb);
790 segs = skb_gso_segment(skb, 0);
791 /* Does not use PTR_ERR to limit the number of error codes that can be
792 * returned by nf_queue. For instance, callers rely on -ESRCH to
793 * mean 'ignore this hook'.
794 */
795 if (IS_ERR_OR_NULL(segs))
796 goto out_err;
797 queued = 0;
798 err = 0;
799 do {
800 struct sk_buff *nskb = segs->next;
801 if (err == 0)
802 err = __nfqnl_enqueue_packet_gso(net, queue,
803 segs, entry);
804 if (err == 0)
805 queued++;
806 else
807 kfree_skb(segs);
808 segs = nskb;
809 } while (segs);
810
811 if (queued) {
812 if (err) /* some segments are already queued */
813 free_entry(entry);
814 kfree_skb(skb);
815 return 0;
816 }
817 out_err:
818 nf_bridge_adjust_segmented_data(skb);
819 return err;
820 }
821
822 static int
823 nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e, int diff)
824 {
825 struct sk_buff *nskb;
826
827 if (diff < 0) {
828 if (pskb_trim(e->skb, data_len))
829 return -ENOMEM;
830 } else if (diff > 0) {
831 if (data_len > 0xFFFF)
832 return -EINVAL;
833 if (diff > skb_tailroom(e->skb)) {
834 nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
835 diff, GFP_ATOMIC);
836 if (!nskb) {
837 printk(KERN_WARNING "nf_queue: OOM "
838 "in mangle, dropping packet\n");
839 return -ENOMEM;
840 }
841 kfree_skb(e->skb);
842 e->skb = nskb;
843 }
844 skb_put(e->skb, diff);
845 }
846 if (!skb_make_writable(e->skb, data_len))
847 return -ENOMEM;
848 skb_copy_to_linear_data(e->skb, data, data_len);
849 e->skb->ip_summed = CHECKSUM_NONE;
850 return 0;
851 }
852
853 static int
854 nfqnl_set_mode(struct nfqnl_instance *queue,
855 unsigned char mode, unsigned int range)
856 {
857 int status = 0;
858
859 spin_lock_bh(&queue->lock);
860 switch (mode) {
861 case NFQNL_COPY_NONE:
862 case NFQNL_COPY_META:
863 queue->copy_mode = mode;
864 queue->copy_range = 0;
865 break;
866
867 case NFQNL_COPY_PACKET:
868 queue->copy_mode = mode;
869 if (range == 0 || range > NFQNL_MAX_COPY_RANGE)
870 queue->copy_range = NFQNL_MAX_COPY_RANGE;
871 else
872 queue->copy_range = range;
873 break;
874
875 default:
876 status = -EINVAL;
877
878 }
879 spin_unlock_bh(&queue->lock);
880
881 return status;
882 }
883
884 static int
885 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
886 {
887 if (entry->state.in)
888 if (entry->state.in->ifindex == ifindex)
889 return 1;
890 if (entry->state.out)
891 if (entry->state.out->ifindex == ifindex)
892 return 1;
893 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
894 if (entry->skb->nf_bridge) {
895 int physinif, physoutif;
896
897 physinif = nf_bridge_get_physinif(entry->skb);
898 physoutif = nf_bridge_get_physoutif(entry->skb);
899
900 if (physinif == ifindex || physoutif == ifindex)
901 return 1;
902 }
903 #endif
904 return 0;
905 }
906
907 /* drop all packets with either indev or outdev == ifindex from all queue
908 * instances */
909 static void
910 nfqnl_dev_drop(struct net *net, int ifindex)
911 {
912 int i;
913 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
914
915 rcu_read_lock();
916
917 for (i = 0; i < INSTANCE_BUCKETS; i++) {
918 struct nfqnl_instance *inst;
919 struct hlist_head *head = &q->instance_table[i];
920
921 hlist_for_each_entry_rcu(inst, head, hlist)
922 nfqnl_flush(inst, dev_cmp, ifindex);
923 }
924
925 rcu_read_unlock();
926 }
927
928 static int
929 nfqnl_rcv_dev_event(struct notifier_block *this,
930 unsigned long event, void *ptr)
931 {
932 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
933
934 /* Drop any packets associated with the downed device */
935 if (event == NETDEV_DOWN)
936 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
937 return NOTIFY_DONE;
938 }
939
940 static struct notifier_block nfqnl_dev_notifier = {
941 .notifier_call = nfqnl_rcv_dev_event,
942 };
943
944 static void nfqnl_nf_hook_drop(struct net *net)
945 {
946 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
947 int i;
948
949 for (i = 0; i < INSTANCE_BUCKETS; i++) {
950 struct nfqnl_instance *inst;
951 struct hlist_head *head = &q->instance_table[i];
952
953 hlist_for_each_entry_rcu(inst, head, hlist)
954 nfqnl_flush(inst, NULL, 0);
955 }
956 }
957
958 static int
959 nfqnl_rcv_nl_event(struct notifier_block *this,
960 unsigned long event, void *ptr)
961 {
962 struct netlink_notify *n = ptr;
963 struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
964
965 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
966 int i;
967
968 /* destroy all instances for this portid */
969 spin_lock(&q->instances_lock);
970 for (i = 0; i < INSTANCE_BUCKETS; i++) {
971 struct hlist_node *t2;
972 struct nfqnl_instance *inst;
973 struct hlist_head *head = &q->instance_table[i];
974
975 hlist_for_each_entry_safe(inst, t2, head, hlist) {
976 if (n->portid == inst->peer_portid)
977 __instance_destroy(inst);
978 }
979 }
980 spin_unlock(&q->instances_lock);
981 }
982 return NOTIFY_DONE;
983 }
984
985 static struct notifier_block nfqnl_rtnl_notifier = {
986 .notifier_call = nfqnl_rcv_nl_event,
987 };
988
989 static const struct nla_policy nfqa_vlan_policy[NFQA_VLAN_MAX + 1] = {
990 [NFQA_VLAN_TCI] = { .type = NLA_U16},
991 [NFQA_VLAN_PROTO] = { .type = NLA_U16},
992 };
993
994 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
995 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
996 [NFQA_MARK] = { .type = NLA_U32 },
997 [NFQA_PAYLOAD] = { .type = NLA_UNSPEC },
998 [NFQA_CT] = { .type = NLA_UNSPEC },
999 [NFQA_EXP] = { .type = NLA_UNSPEC },
1000 [NFQA_VLAN] = { .type = NLA_NESTED },
1001 };
1002
1003 static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
1004 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
1005 [NFQA_MARK] = { .type = NLA_U32 },
1006 };
1007
1008 static struct nfqnl_instance *
1009 verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, u32 nlportid)
1010 {
1011 struct nfqnl_instance *queue;
1012
1013 queue = instance_lookup(q, queue_num);
1014 if (!queue)
1015 return ERR_PTR(-ENODEV);
1016
1017 if (queue->peer_portid != nlportid)
1018 return ERR_PTR(-EPERM);
1019
1020 return queue;
1021 }
1022
1023 static struct nfqnl_msg_verdict_hdr*
1024 verdicthdr_get(const struct nlattr * const nfqa[])
1025 {
1026 struct nfqnl_msg_verdict_hdr *vhdr;
1027 unsigned int verdict;
1028
1029 if (!nfqa[NFQA_VERDICT_HDR])
1030 return NULL;
1031
1032 vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
1033 verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
1034 if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
1035 return NULL;
1036 return vhdr;
1037 }
1038
1039 static int nfq_id_after(unsigned int id, unsigned int max)
1040 {
1041 return (int)(id - max) > 0;
1042 }
1043
1044 static int nfqnl_recv_verdict_batch(struct net *net, struct sock *ctnl,
1045 struct sk_buff *skb,
1046 const struct nlmsghdr *nlh,
1047 const struct nlattr * const nfqa[],
1048 struct netlink_ext_ack *extack)
1049 {
1050 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1051 struct nf_queue_entry *entry, *tmp;
1052 unsigned int verdict, maxid;
1053 struct nfqnl_msg_verdict_hdr *vhdr;
1054 struct nfqnl_instance *queue;
1055 LIST_HEAD(batch_list);
1056 u16 queue_num = ntohs(nfmsg->res_id);
1057 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1058
1059 queue = verdict_instance_lookup(q, queue_num,
1060 NETLINK_CB(skb).portid);
1061 if (IS_ERR(queue))
1062 return PTR_ERR(queue);
1063
1064 vhdr = verdicthdr_get(nfqa);
1065 if (!vhdr)
1066 return -EINVAL;
1067
1068 verdict = ntohl(vhdr->verdict);
1069 maxid = ntohl(vhdr->id);
1070
1071 spin_lock_bh(&queue->lock);
1072
1073 list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
1074 if (nfq_id_after(entry->id, maxid))
1075 break;
1076 __dequeue_entry(queue, entry);
1077 list_add_tail(&entry->list, &batch_list);
1078 }
1079
1080 spin_unlock_bh(&queue->lock);
1081
1082 if (list_empty(&batch_list))
1083 return -ENOENT;
1084
1085 list_for_each_entry_safe(entry, tmp, &batch_list, list) {
1086 if (nfqa[NFQA_MARK])
1087 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1088 nf_reinject(entry, verdict);
1089 }
1090 return 0;
1091 }
1092
1093 static struct nf_conn *nfqnl_ct_parse(struct nfnl_ct_hook *nfnl_ct,
1094 const struct nlmsghdr *nlh,
1095 const struct nlattr * const nfqa[],
1096 struct nf_queue_entry *entry,
1097 enum ip_conntrack_info *ctinfo)
1098 {
1099 struct nf_conn *ct;
1100
1101 ct = nfnl_ct->get_ct(entry->skb, ctinfo);
1102 if (ct == NULL)
1103 return NULL;
1104
1105 if (nfnl_ct->parse(nfqa[NFQA_CT], ct) < 0)
1106 return NULL;
1107
1108 if (nfqa[NFQA_EXP])
1109 nfnl_ct->attach_expect(nfqa[NFQA_EXP], ct,
1110 NETLINK_CB(entry->skb).portid,
1111 nlmsg_report(nlh));
1112 return ct;
1113 }
1114
1115 static int nfqa_parse_bridge(struct nf_queue_entry *entry,
1116 const struct nlattr * const nfqa[])
1117 {
1118 if (nfqa[NFQA_VLAN]) {
1119 struct nlattr *tb[NFQA_VLAN_MAX + 1];
1120 int err;
1121
1122 err = nla_parse_nested(tb, NFQA_VLAN_MAX, nfqa[NFQA_VLAN],
1123 nfqa_vlan_policy, NULL);
1124 if (err < 0)
1125 return err;
1126
1127 if (!tb[NFQA_VLAN_TCI] || !tb[NFQA_VLAN_PROTO])
1128 return -EINVAL;
1129
1130 entry->skb->vlan_tci = ntohs(nla_get_be16(tb[NFQA_VLAN_TCI]));
1131 entry->skb->vlan_proto = nla_get_be16(tb[NFQA_VLAN_PROTO]);
1132 }
1133
1134 if (nfqa[NFQA_L2HDR]) {
1135 int mac_header_len = entry->skb->network_header -
1136 entry->skb->mac_header;
1137
1138 if (mac_header_len != nla_len(nfqa[NFQA_L2HDR]))
1139 return -EINVAL;
1140 else if (mac_header_len > 0)
1141 memcpy(skb_mac_header(entry->skb),
1142 nla_data(nfqa[NFQA_L2HDR]),
1143 mac_header_len);
1144 }
1145
1146 return 0;
1147 }
1148
1149 static int nfqnl_recv_verdict(struct net *net, struct sock *ctnl,
1150 struct sk_buff *skb,
1151 const struct nlmsghdr *nlh,
1152 const struct nlattr * const nfqa[],
1153 struct netlink_ext_ack *extack)
1154 {
1155 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1156 u_int16_t queue_num = ntohs(nfmsg->res_id);
1157 struct nfqnl_msg_verdict_hdr *vhdr;
1158 struct nfqnl_instance *queue;
1159 unsigned int verdict;
1160 struct nf_queue_entry *entry;
1161 enum ip_conntrack_info uninitialized_var(ctinfo);
1162 struct nfnl_ct_hook *nfnl_ct;
1163 struct nf_conn *ct = NULL;
1164 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1165 int err;
1166
1167 queue = verdict_instance_lookup(q, queue_num,
1168 NETLINK_CB(skb).portid);
1169 if (IS_ERR(queue))
1170 return PTR_ERR(queue);
1171
1172 vhdr = verdicthdr_get(nfqa);
1173 if (!vhdr)
1174 return -EINVAL;
1175
1176 verdict = ntohl(vhdr->verdict);
1177
1178 entry = find_dequeue_entry(queue, ntohl(vhdr->id));
1179 if (entry == NULL)
1180 return -ENOENT;
1181
1182 /* rcu lock already held from nfnl->call_rcu. */
1183 nfnl_ct = rcu_dereference(nfnl_ct_hook);
1184
1185 if (nfqa[NFQA_CT]) {
1186 if (nfnl_ct != NULL)
1187 ct = nfqnl_ct_parse(nfnl_ct, nlh, nfqa, entry, &ctinfo);
1188 }
1189
1190 if (entry->state.pf == PF_BRIDGE) {
1191 err = nfqa_parse_bridge(entry, nfqa);
1192 if (err < 0)
1193 return err;
1194 }
1195
1196 if (nfqa[NFQA_PAYLOAD]) {
1197 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
1198 int diff = payload_len - entry->skb->len;
1199
1200 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
1201 payload_len, entry, diff) < 0)
1202 verdict = NF_DROP;
1203
1204 if (ct && diff)
1205 nfnl_ct->seq_adjust(entry->skb, ct, ctinfo, diff);
1206 }
1207
1208 if (nfqa[NFQA_MARK])
1209 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1210
1211 nf_reinject(entry, verdict);
1212 return 0;
1213 }
1214
1215 static int nfqnl_recv_unsupp(struct net *net, struct sock *ctnl,
1216 struct sk_buff *skb, const struct nlmsghdr *nlh,
1217 const struct nlattr * const nfqa[],
1218 struct netlink_ext_ack *extack)
1219 {
1220 return -ENOTSUPP;
1221 }
1222
1223 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1224 [NFQA_CFG_CMD] = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1225 [NFQA_CFG_PARAMS] = { .len = sizeof(struct nfqnl_msg_config_params) },
1226 };
1227
1228 static const struct nf_queue_handler nfqh = {
1229 .outfn = nfqnl_enqueue_packet,
1230 .nf_hook_drop = nfqnl_nf_hook_drop,
1231 };
1232
1233 static int nfqnl_recv_config(struct net *net, struct sock *ctnl,
1234 struct sk_buff *skb, const struct nlmsghdr *nlh,
1235 const struct nlattr * const nfqa[],
1236 struct netlink_ext_ack *extack)
1237 {
1238 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1239 u_int16_t queue_num = ntohs(nfmsg->res_id);
1240 struct nfqnl_instance *queue;
1241 struct nfqnl_msg_config_cmd *cmd = NULL;
1242 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1243 __u32 flags = 0, mask = 0;
1244 int ret = 0;
1245
1246 if (nfqa[NFQA_CFG_CMD]) {
1247 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1248
1249 /* Obsolete commands without queue context */
1250 switch (cmd->command) {
1251 case NFQNL_CFG_CMD_PF_BIND: return 0;
1252 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1253 }
1254 }
1255
1256 /* Check if we support these flags in first place, dependencies should
1257 * be there too not to break atomicity.
1258 */
1259 if (nfqa[NFQA_CFG_FLAGS]) {
1260 if (!nfqa[NFQA_CFG_MASK]) {
1261 /* A mask is needed to specify which flags are being
1262 * changed.
1263 */
1264 return -EINVAL;
1265 }
1266
1267 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1268 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1269
1270 if (flags >= NFQA_CFG_F_MAX)
1271 return -EOPNOTSUPP;
1272
1273 #if !IS_ENABLED(CONFIG_NETWORK_SECMARK)
1274 if (flags & mask & NFQA_CFG_F_SECCTX)
1275 return -EOPNOTSUPP;
1276 #endif
1277 if ((flags & mask & NFQA_CFG_F_CONNTRACK) &&
1278 !rcu_access_pointer(nfnl_ct_hook)) {
1279 #ifdef CONFIG_MODULES
1280 nfnl_unlock(NFNL_SUBSYS_QUEUE);
1281 request_module("ip_conntrack_netlink");
1282 nfnl_lock(NFNL_SUBSYS_QUEUE);
1283 if (rcu_access_pointer(nfnl_ct_hook))
1284 return -EAGAIN;
1285 #endif
1286 return -EOPNOTSUPP;
1287 }
1288 }
1289
1290 rcu_read_lock();
1291 queue = instance_lookup(q, queue_num);
1292 if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1293 ret = -EPERM;
1294 goto err_out_unlock;
1295 }
1296
1297 if (cmd != NULL) {
1298 switch (cmd->command) {
1299 case NFQNL_CFG_CMD_BIND:
1300 if (queue) {
1301 ret = -EBUSY;
1302 goto err_out_unlock;
1303 }
1304 queue = instance_create(q, queue_num,
1305 NETLINK_CB(skb).portid);
1306 if (IS_ERR(queue)) {
1307 ret = PTR_ERR(queue);
1308 goto err_out_unlock;
1309 }
1310 break;
1311 case NFQNL_CFG_CMD_UNBIND:
1312 if (!queue) {
1313 ret = -ENODEV;
1314 goto err_out_unlock;
1315 }
1316 instance_destroy(q, queue);
1317 goto err_out_unlock;
1318 case NFQNL_CFG_CMD_PF_BIND:
1319 case NFQNL_CFG_CMD_PF_UNBIND:
1320 break;
1321 default:
1322 ret = -ENOTSUPP;
1323 goto err_out_unlock;
1324 }
1325 }
1326
1327 if (!queue) {
1328 ret = -ENODEV;
1329 goto err_out_unlock;
1330 }
1331
1332 if (nfqa[NFQA_CFG_PARAMS]) {
1333 struct nfqnl_msg_config_params *params =
1334 nla_data(nfqa[NFQA_CFG_PARAMS]);
1335
1336 nfqnl_set_mode(queue, params->copy_mode,
1337 ntohl(params->copy_range));
1338 }
1339
1340 if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1341 __be32 *queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1342
1343 spin_lock_bh(&queue->lock);
1344 queue->queue_maxlen = ntohl(*queue_maxlen);
1345 spin_unlock_bh(&queue->lock);
1346 }
1347
1348 if (nfqa[NFQA_CFG_FLAGS]) {
1349 spin_lock_bh(&queue->lock);
1350 queue->flags &= ~mask;
1351 queue->flags |= flags & mask;
1352 spin_unlock_bh(&queue->lock);
1353 }
1354
1355 err_out_unlock:
1356 rcu_read_unlock();
1357 return ret;
1358 }
1359
1360 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1361 [NFQNL_MSG_PACKET] = { .call_rcu = nfqnl_recv_unsupp,
1362 .attr_count = NFQA_MAX, },
1363 [NFQNL_MSG_VERDICT] = { .call_rcu = nfqnl_recv_verdict,
1364 .attr_count = NFQA_MAX,
1365 .policy = nfqa_verdict_policy },
1366 [NFQNL_MSG_CONFIG] = { .call = nfqnl_recv_config,
1367 .attr_count = NFQA_CFG_MAX,
1368 .policy = nfqa_cfg_policy },
1369 [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
1370 .attr_count = NFQA_MAX,
1371 .policy = nfqa_verdict_batch_policy },
1372 };
1373
1374 static const struct nfnetlink_subsystem nfqnl_subsys = {
1375 .name = "nf_queue",
1376 .subsys_id = NFNL_SUBSYS_QUEUE,
1377 .cb_count = NFQNL_MSG_MAX,
1378 .cb = nfqnl_cb,
1379 };
1380
1381 #ifdef CONFIG_PROC_FS
1382 struct iter_state {
1383 struct seq_net_private p;
1384 unsigned int bucket;
1385 };
1386
1387 static struct hlist_node *get_first(struct seq_file *seq)
1388 {
1389 struct iter_state *st = seq->private;
1390 struct net *net;
1391 struct nfnl_queue_net *q;
1392
1393 if (!st)
1394 return NULL;
1395
1396 net = seq_file_net(seq);
1397 q = nfnl_queue_pernet(net);
1398 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1399 if (!hlist_empty(&q->instance_table[st->bucket]))
1400 return q->instance_table[st->bucket].first;
1401 }
1402 return NULL;
1403 }
1404
1405 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1406 {
1407 struct iter_state *st = seq->private;
1408 struct net *net = seq_file_net(seq);
1409
1410 h = h->next;
1411 while (!h) {
1412 struct nfnl_queue_net *q;
1413
1414 if (++st->bucket >= INSTANCE_BUCKETS)
1415 return NULL;
1416
1417 q = nfnl_queue_pernet(net);
1418 h = q->instance_table[st->bucket].first;
1419 }
1420 return h;
1421 }
1422
1423 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1424 {
1425 struct hlist_node *head;
1426 head = get_first(seq);
1427
1428 if (head)
1429 while (pos && (head = get_next(seq, head)))
1430 pos--;
1431 return pos ? NULL : head;
1432 }
1433
1434 static void *seq_start(struct seq_file *s, loff_t *pos)
1435 __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1436 {
1437 spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1438 return get_idx(s, *pos);
1439 }
1440
1441 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1442 {
1443 (*pos)++;
1444 return get_next(s, v);
1445 }
1446
1447 static void seq_stop(struct seq_file *s, void *v)
1448 __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1449 {
1450 spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1451 }
1452
1453 static int seq_show(struct seq_file *s, void *v)
1454 {
1455 const struct nfqnl_instance *inst = v;
1456
1457 seq_printf(s, "%5u %6u %5u %1u %5u %5u %5u %8u %2d\n",
1458 inst->queue_num,
1459 inst->peer_portid, inst->queue_total,
1460 inst->copy_mode, inst->copy_range,
1461 inst->queue_dropped, inst->queue_user_dropped,
1462 inst->id_sequence, 1);
1463 return 0;
1464 }
1465
1466 static const struct seq_operations nfqnl_seq_ops = {
1467 .start = seq_start,
1468 .next = seq_next,
1469 .stop = seq_stop,
1470 .show = seq_show,
1471 };
1472
1473 static int nfqnl_open(struct inode *inode, struct file *file)
1474 {
1475 return seq_open_net(inode, file, &nfqnl_seq_ops,
1476 sizeof(struct iter_state));
1477 }
1478
1479 static const struct file_operations nfqnl_file_ops = {
1480 .open = nfqnl_open,
1481 .read = seq_read,
1482 .llseek = seq_lseek,
1483 .release = seq_release_net,
1484 };
1485
1486 #endif /* PROC_FS */
1487
1488 static int __net_init nfnl_queue_net_init(struct net *net)
1489 {
1490 unsigned int i;
1491 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1492
1493 for (i = 0; i < INSTANCE_BUCKETS; i++)
1494 INIT_HLIST_HEAD(&q->instance_table[i]);
1495
1496 spin_lock_init(&q->instances_lock);
1497
1498 #ifdef CONFIG_PROC_FS
1499 if (!proc_create("nfnetlink_queue", 0440,
1500 net->nf.proc_netfilter, &nfqnl_file_ops))
1501 return -ENOMEM;
1502 #endif
1503 nf_register_queue_handler(net, &nfqh);
1504 return 0;
1505 }
1506
1507 static void __net_exit nfnl_queue_net_exit(struct net *net)
1508 {
1509 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1510 unsigned int i;
1511
1512 nf_unregister_queue_handler(net);
1513 #ifdef CONFIG_PROC_FS
1514 remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1515 #endif
1516 for (i = 0; i < INSTANCE_BUCKETS; i++)
1517 WARN_ON_ONCE(!hlist_empty(&q->instance_table[i]));
1518 }
1519
1520 static void nfnl_queue_net_exit_batch(struct list_head *net_exit_list)
1521 {
1522 synchronize_rcu();
1523 }
1524
1525 static struct pernet_operations nfnl_queue_net_ops = {
1526 .init = nfnl_queue_net_init,
1527 .exit = nfnl_queue_net_exit,
1528 .exit_batch = nfnl_queue_net_exit_batch,
1529 .id = &nfnl_queue_net_id,
1530 .size = sizeof(struct nfnl_queue_net),
1531 };
1532
1533 static int __init nfnetlink_queue_init(void)
1534 {
1535 int status;
1536
1537 status = register_pernet_subsys(&nfnl_queue_net_ops);
1538 if (status < 0) {
1539 pr_err("nf_queue: failed to register pernet ops\n");
1540 goto out;
1541 }
1542
1543 netlink_register_notifier(&nfqnl_rtnl_notifier);
1544 status = nfnetlink_subsys_register(&nfqnl_subsys);
1545 if (status < 0) {
1546 pr_err("nf_queue: failed to create netlink socket\n");
1547 goto cleanup_netlink_notifier;
1548 }
1549
1550 status = register_netdevice_notifier(&nfqnl_dev_notifier);
1551 if (status < 0) {
1552 pr_err("nf_queue: failed to register netdevice notifier\n");
1553 goto cleanup_netlink_subsys;
1554 }
1555
1556 return status;
1557
1558 cleanup_netlink_subsys:
1559 nfnetlink_subsys_unregister(&nfqnl_subsys);
1560 cleanup_netlink_notifier:
1561 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1562 unregister_pernet_subsys(&nfnl_queue_net_ops);
1563 out:
1564 return status;
1565 }
1566
1567 static void __exit nfnetlink_queue_fini(void)
1568 {
1569 unregister_netdevice_notifier(&nfqnl_dev_notifier);
1570 nfnetlink_subsys_unregister(&nfqnl_subsys);
1571 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1572 unregister_pernet_subsys(&nfnl_queue_net_ops);
1573
1574 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1575 }
1576
1577 MODULE_DESCRIPTION("netfilter packet queue handler");
1578 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1579 MODULE_LICENSE("GPL");
1580 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1581
1582 module_init(nfnetlink_queue_init);
1583 module_exit(nfnetlink_queue_fini);
CRIS linux kernel tree