1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * This is a module which is used for queueing packets and communicating with
4 * userspace via nfnetlink.
5 *
6 * (C) 2005 by Harald Welte <laforge@netfilter.org>
7 * (C) 2007 by Patrick McHardy <kaber@trash.net>
8 *
9 * Based on the old ipv4-only ip_queue.c:
10 * (C) 2000-2002 James Morris <jmorris@intercode.com.au>
11 * (C) 2003-2005 Netfilter Core Team <coreteam@netfilter.org>
12 */
13
14 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
15
16 #include <linux/module.h>
17 #include <linux/skbuff.h>
18 #include <linux/init.h>
19 #include <linux/spinlock.h>
20 #include <linux/slab.h>
21 #include <linux/notifier.h>
22 #include <linux/netdevice.h>
23 #include <linux/netfilter.h>
24 #include <linux/proc_fs.h>
25 #include <linux/netfilter_ipv4.h>
26 #include <linux/netfilter_ipv6.h>
27 #include <linux/netfilter_bridge.h>
28 #include <linux/netfilter/nfnetlink.h>
29 #include <linux/netfilter/nfnetlink_queue.h>
30 #include <linux/netfilter/nf_conntrack_common.h>
31 #include <linux/list.h>
32 #include <net/sock.h>
33 #include <net/tcp_states.h>
34 #include <net/netfilter/nf_queue.h>
35 #include <net/netns/generic.h>
36
37 #include <linux/atomic.h>
38
39 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
40 #include "../bridge/br_private.h"
41 #endif
42
43 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
44 #include <net/netfilter/nf_conntrack.h>
45 #endif
46
47 #define NFQNL_QMAX_DEFAULT 1024
48
49 /* We're using struct nlattr which has 16bit nla_len. Note that nla_len
50 * includes the header length. Thus, the maximum packet length that we
51 * support is 65531 bytes. We send truncated packets if the specified length
52 * is larger than that. Userspace can check for presence of NFQA_CAP_LEN
53 * attribute to detect truncation.
54 */
55 #define NFQNL_MAX_COPY_RANGE (0xffff - NLA_HDRLEN)
56
57 struct nfqnl_instance {
58 struct hlist_node hlist; /* global list of queues */
59 struct rcu_head rcu;
60
61 u32 peer_portid;
62 unsigned int queue_maxlen;
63 unsigned int copy_range;
64 unsigned int queue_dropped;
65 unsigned int queue_user_dropped;
66
67
68 u_int16_t queue_num; /* number of this queue */
69 u_int8_t copy_mode;
70 u_int32_t flags; /* Set using NFQA_CFG_FLAGS */
71 /*
72 * Following fields are dirtied for each queued packet,
73 * keep them in same cache line if possible.
74 */
75 spinlock_t lock ____cacheline_aligned_in_smp;
76 unsigned int queue_total;
77 unsigned int id_sequence; /* 'sequence' of pkt ids */
78 struct list_head queue_list; /* packets in queue */
79 };
80
81 typedef int (*nfqnl_cmpfn)(struct nf_queue_entry *, unsigned long);
82
83 static unsigned int nfnl_queue_net_id __read_mostly;
84
85 #define INSTANCE_BUCKETS 16
86 struct nfnl_queue_net {
87 spinlock_t instances_lock;
88 struct hlist_head instance_table[INSTANCE_BUCKETS];
89 };
90
nfnl_queue_pernet(struct net * net)91 static struct nfnl_queue_net *nfnl_queue_pernet(struct net *net)
92 {
93 return net_generic(net, nfnl_queue_net_id);
94 }
95
instance_hashfn(u_int16_t queue_num)96 static inline u_int8_t instance_hashfn(u_int16_t queue_num)
97 {
98 return ((queue_num >> 8) ^ queue_num) % INSTANCE_BUCKETS;
99 }
100
101 static struct nfqnl_instance *
instance_lookup(struct nfnl_queue_net * q,u_int16_t queue_num)102 instance_lookup(struct nfnl_queue_net *q, u_int16_t queue_num)
103 {
104 struct hlist_head *head;
105 struct nfqnl_instance *inst;
106
107 head = &q->instance_table[instance_hashfn(queue_num)];
108 hlist_for_each_entry_rcu(inst, head, hlist) {
109 if (inst->queue_num == queue_num)
110 return inst;
111 }
112 return NULL;
113 }
114
115 static struct nfqnl_instance *
instance_create(struct nfnl_queue_net * q,u_int16_t queue_num,u32 portid)116 instance_create(struct nfnl_queue_net *q, u_int16_t queue_num, u32 portid)
117 {
118 struct nfqnl_instance *inst;
119 unsigned int h;
120 int err;
121
122 spin_lock(&q->instances_lock);
123 if (instance_lookup(q, queue_num)) {
124 err = -EEXIST;
125 goto out_unlock;
126 }
127
128 inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
129 if (!inst) {
130 err = -ENOMEM;
131 goto out_unlock;
132 }
133
134 inst->queue_num = queue_num;
135 inst->peer_portid = portid;
136 inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
137 inst->copy_range = NFQNL_MAX_COPY_RANGE;
138 inst->copy_mode = NFQNL_COPY_NONE;
139 spin_lock_init(&inst->lock);
140 INIT_LIST_HEAD(&inst->queue_list);
141
142 if (!try_module_get(THIS_MODULE)) {
143 err = -EAGAIN;
144 goto out_free;
145 }
146
147 h = instance_hashfn(queue_num);
148 hlist_add_head_rcu(&inst->hlist, &q->instance_table[h]);
149
150 spin_unlock(&q->instances_lock);
151
152 return inst;
153
154 out_free:
155 kfree(inst);
156 out_unlock:
157 spin_unlock(&q->instances_lock);
158 return ERR_PTR(err);
159 }
160
161 static void nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn,
162 unsigned long data);
163
164 static void
instance_destroy_rcu(struct rcu_head * head)165 instance_destroy_rcu(struct rcu_head *head)
166 {
167 struct nfqnl_instance *inst = container_of(head, struct nfqnl_instance,
168 rcu);
169
170 nfqnl_flush(inst, NULL, 0);
171 kfree(inst);
172 module_put(THIS_MODULE);
173 }
174
175 static void
__instance_destroy(struct nfqnl_instance * inst)176 __instance_destroy(struct nfqnl_instance *inst)
177 {
178 hlist_del_rcu(&inst->hlist);
179 call_rcu(&inst->rcu, instance_destroy_rcu);
180 }
181
182 static void
instance_destroy(struct nfnl_queue_net * q,struct nfqnl_instance * inst)183 instance_destroy(struct nfnl_queue_net *q, struct nfqnl_instance *inst)
184 {
185 spin_lock(&q->instances_lock);
186 __instance_destroy(inst);
187 spin_unlock(&q->instances_lock);
188 }
189
190 static inline void
__enqueue_entry(struct nfqnl_instance * queue,struct nf_queue_entry * entry)191 __enqueue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
192 {
193 list_add_tail(&entry->list, &queue->queue_list);
194 queue->queue_total++;
195 }
196
197 static void
__dequeue_entry(struct nfqnl_instance * queue,struct nf_queue_entry * entry)198 __dequeue_entry(struct nfqnl_instance *queue, struct nf_queue_entry *entry)
199 {
200 list_del(&entry->list);
201 queue->queue_total--;
202 }
203
204 static struct nf_queue_entry *
find_dequeue_entry(struct nfqnl_instance * queue,unsigned int id)205 find_dequeue_entry(struct nfqnl_instance *queue, unsigned int id)
206 {
207 struct nf_queue_entry *entry = NULL, *i;
208
209 spin_lock_bh(&queue->lock);
210
211 list_for_each_entry(i, &queue->queue_list, list) {
212 if (i->id == id) {
213 entry = i;
214 break;
215 }
216 }
217
218 if (entry)
219 __dequeue_entry(queue, entry);
220
221 spin_unlock_bh(&queue->lock);
222
223 return entry;
224 }
225
nfqnl_reinject(struct nf_queue_entry * entry,unsigned int verdict)226 static void nfqnl_reinject(struct nf_queue_entry *entry, unsigned int verdict)
227 {
228 struct nf_ct_hook *ct_hook;
229 int err;
230
231 if (verdict == NF_ACCEPT ||
232 verdict == NF_REPEAT ||
233 verdict == NF_STOP) {
234 rcu_read_lock();
235 ct_hook = rcu_dereference(nf_ct_hook);
236 if (ct_hook) {
237 err = ct_hook->update(entry->state.net, entry->skb);
238 if (err < 0)
239 verdict = NF_DROP;
240 }
241 rcu_read_unlock();
242 }
243 nf_reinject(entry, verdict);
244 }
245
246 static void
nfqnl_flush(struct nfqnl_instance * queue,nfqnl_cmpfn cmpfn,unsigned long data)247 nfqnl_flush(struct nfqnl_instance *queue, nfqnl_cmpfn cmpfn, unsigned long data)
248 {
249 struct nf_queue_entry *entry, *next;
250
251 spin_lock_bh(&queue->lock);
252 list_for_each_entry_safe(entry, next, &queue->queue_list, list) {
253 if (!cmpfn || cmpfn(entry, data)) {
254 list_del(&entry->list);
255 queue->queue_total--;
256 nfqnl_reinject(entry, NF_DROP);
257 }
258 }
259 spin_unlock_bh(&queue->lock);
260 }
261
262 static int
nfqnl_put_packet_info(struct sk_buff * nlskb,struct sk_buff * packet,bool csum_verify)263 nfqnl_put_packet_info(struct sk_buff *nlskb, struct sk_buff *packet,
264 bool csum_verify)
265 {
266 __u32 flags = 0;
267
268 if (packet->ip_summed == CHECKSUM_PARTIAL)
269 flags = NFQA_SKB_CSUMNOTREADY;
270 else if (csum_verify)
271 flags = NFQA_SKB_CSUM_NOTVERIFIED;
272
273 if (skb_is_gso(packet))
274 flags |= NFQA_SKB_GSO;
275
276 return flags ? nla_put_be32(nlskb, NFQA_SKB_INFO, htonl(flags)) : 0;
277 }
278
nfqnl_put_sk_uidgid(struct sk_buff * skb,struct sock * sk)279 static int nfqnl_put_sk_uidgid(struct sk_buff *skb, struct sock *sk)
280 {
281 const struct cred *cred;
282
283 if (!sk_fullsock(sk))
284 return 0;
285
286 read_lock_bh(&sk->sk_callback_lock);
287 if (sk->sk_socket && sk->sk_socket->file) {
288 cred = sk->sk_socket->file->f_cred;
289 if (nla_put_be32(skb, NFQA_UID,
290 htonl(from_kuid_munged(&init_user_ns, cred->fsuid))))
291 goto nla_put_failure;
292 if (nla_put_be32(skb, NFQA_GID,
293 htonl(from_kgid_munged(&init_user_ns, cred->fsgid))))
294 goto nla_put_failure;
295 }
296 read_unlock_bh(&sk->sk_callback_lock);
297 return 0;
298
299 nla_put_failure:
300 read_unlock_bh(&sk->sk_callback_lock);
301 return -1;
302 }
303
nfqnl_get_sk_secctx(struct sk_buff * skb,char ** secdata)304 static u32 nfqnl_get_sk_secctx(struct sk_buff *skb, char **secdata)
305 {
306 u32 seclen = 0;
307 #if IS_ENABLED(CONFIG_NETWORK_SECMARK)
308 if (!skb || !sk_fullsock(skb->sk))
309 return 0;
310
311 read_lock_bh(&skb->sk->sk_callback_lock);
312
313 if (skb->secmark)
314 security_secid_to_secctx(skb->secmark, secdata, &seclen);
315
316 read_unlock_bh(&skb->sk->sk_callback_lock);
317 #endif
318 return seclen;
319 }
320
nfqnl_get_bridge_size(struct nf_queue_entry * entry)321 static u32 nfqnl_get_bridge_size(struct nf_queue_entry *entry)
322 {
323 struct sk_buff *entskb = entry->skb;
324 u32 nlalen = 0;
325
326 if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb))
327 return 0;
328
329 if (skb_vlan_tag_present(entskb))
330 nlalen += nla_total_size(nla_total_size(sizeof(__be16)) +
331 nla_total_size(sizeof(__be16)));
332
333 if (entskb->network_header > entskb->mac_header)
334 nlalen += nla_total_size((entskb->network_header -
335 entskb->mac_header));
336
337 return nlalen;
338 }
339
nfqnl_put_bridge(struct nf_queue_entry * entry,struct sk_buff * skb)340 static int nfqnl_put_bridge(struct nf_queue_entry *entry, struct sk_buff *skb)
341 {
342 struct sk_buff *entskb = entry->skb;
343
344 if (entry->state.pf != PF_BRIDGE || !skb_mac_header_was_set(entskb))
345 return 0;
346
347 if (skb_vlan_tag_present(entskb)) {
348 struct nlattr *nest;
349
350 nest = nla_nest_start(skb, NFQA_VLAN);
351 if (!nest)
352 goto nla_put_failure;
353
354 if (nla_put_be16(skb, NFQA_VLAN_TCI, htons(entskb->vlan_tci)) ||
355 nla_put_be16(skb, NFQA_VLAN_PROTO, entskb->vlan_proto))
356 goto nla_put_failure;
357
358 nla_nest_end(skb, nest);
359 }
360
361 if (entskb->mac_header < entskb->network_header) {
362 int len = (int)(entskb->network_header - entskb->mac_header);
363
364 if (nla_put(skb, NFQA_L2HDR, len, skb_mac_header(entskb)))
365 goto nla_put_failure;
366 }
367
368 return 0;
369
370 nla_put_failure:
371 return -1;
372 }
373
374 static struct sk_buff *
nfqnl_build_packet_message(struct net * net,struct nfqnl_instance * queue,struct nf_queue_entry * entry,__be32 ** packet_id_ptr)375 nfqnl_build_packet_message(struct net *net, struct nfqnl_instance *queue,
376 struct nf_queue_entry *entry,
377 __be32 **packet_id_ptr)
378 {
379 size_t size;
380 size_t data_len = 0, cap_len = 0;
381 unsigned int hlen = 0;
382 struct sk_buff *skb;
383 struct nlattr *nla;
384 struct nfqnl_msg_packet_hdr *pmsg;
385 struct nlmsghdr *nlh;
386 struct sk_buff *entskb = entry->skb;
387 struct net_device *indev;
388 struct net_device *outdev;
389 struct nf_conn *ct = NULL;
390 enum ip_conntrack_info ctinfo;
391 struct nfnl_ct_hook *nfnl_ct;
392 bool csum_verify;
393 char *secdata = NULL;
394 u32 seclen = 0;
395
396 size = nlmsg_total_size(sizeof(struct nfgenmsg))
397 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr))
398 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
399 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
400 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
401 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
402 + nla_total_size(sizeof(u_int32_t)) /* ifindex */
403 #endif
404 + nla_total_size(sizeof(u_int32_t)) /* mark */
405 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw))
406 + nla_total_size(sizeof(u_int32_t)) /* skbinfo */
407 + nla_total_size(sizeof(u_int32_t)); /* cap_len */
408
409 if (entskb->tstamp)
410 size += nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp));
411
412 size += nfqnl_get_bridge_size(entry);
413
414 if (entry->state.hook <= NF_INET_FORWARD ||
415 (entry->state.hook == NF_INET_POST_ROUTING && entskb->sk == NULL))
416 csum_verify = !skb_csum_unnecessary(entskb);
417 else
418 csum_verify = false;
419
420 outdev = entry->state.out;
421
422 switch ((enum nfqnl_config_mode)READ_ONCE(queue->copy_mode)) {
423 case NFQNL_COPY_META:
424 case NFQNL_COPY_NONE:
425 break;
426
427 case NFQNL_COPY_PACKET:
428 if (!(queue->flags & NFQA_CFG_F_GSO) &&
429 entskb->ip_summed == CHECKSUM_PARTIAL &&
430 skb_checksum_help(entskb))
431 return NULL;
432
433 data_len = READ_ONCE(queue->copy_range);
434 if (data_len > entskb->len)
435 data_len = entskb->len;
436
437 hlen = skb_zerocopy_headlen(entskb);
438 hlen = min_t(unsigned int, hlen, data_len);
439 size += sizeof(struct nlattr) + hlen;
440 cap_len = entskb->len;
441 break;
442 }
443
444 nfnl_ct = rcu_dereference(nfnl_ct_hook);
445
446 if (queue->flags & NFQA_CFG_F_CONNTRACK) {
447 if (nfnl_ct != NULL) {
448 ct = nfnl_ct->get_ct(entskb, &ctinfo);
449 if (ct != NULL)
450 size += nfnl_ct->build_size(ct);
451 }
452 }
453
454 if (queue->flags & NFQA_CFG_F_UID_GID) {
455 size += (nla_total_size(sizeof(u_int32_t)) /* uid */
456 + nla_total_size(sizeof(u_int32_t))); /* gid */
457 }
458
459 if ((queue->flags & NFQA_CFG_F_SECCTX) && entskb->sk) {
460 seclen = nfqnl_get_sk_secctx(entskb, &secdata);
461 if (seclen)
462 size += nla_total_size(seclen);
463 }
464
465 skb = alloc_skb(size, GFP_ATOMIC);
466 if (!skb) {
467 skb_tx_error(entskb);
468 goto nlmsg_failure;
469 }
470
471 nlh = nfnl_msg_put(skb, 0, 0,
472 nfnl_msg_type(NFNL_SUBSYS_QUEUE, NFQNL_MSG_PACKET),
473 0, entry->state.pf, NFNETLINK_V0,
474 htons(queue->queue_num));
475 if (!nlh) {
476 skb_tx_error(entskb);
477 kfree_skb(skb);
478 goto nlmsg_failure;
479 }
480
481 nla = __nla_reserve(skb, NFQA_PACKET_HDR, sizeof(*pmsg));
482 pmsg = nla_data(nla);
483 pmsg->hw_protocol = entskb->protocol;
484 pmsg->hook = entry->state.hook;
485 *packet_id_ptr = &pmsg->packet_id;
486
487 indev = entry->state.in;
488 if (indev) {
489 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
490 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV, htonl(indev->ifindex)))
491 goto nla_put_failure;
492 #else
493 if (entry->state.pf == PF_BRIDGE) {
494 /* Case 1: indev is physical input device, we need to
495 * look for bridge group (when called from
496 * netfilter_bridge) */
497 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
498 htonl(indev->ifindex)) ||
499 /* this is the bridge group "brX" */
500 /* rcu_read_lock()ed by __nf_queue */
501 nla_put_be32(skb, NFQA_IFINDEX_INDEV,
502 htonl(br_port_get_rcu(indev)->br->dev->ifindex)))
503 goto nla_put_failure;
504 } else {
505 int physinif;
506
507 /* Case 2: indev is bridge group, we need to look for
508 * physical device (when called from ipv4) */
509 if (nla_put_be32(skb, NFQA_IFINDEX_INDEV,
510 htonl(indev->ifindex)))
511 goto nla_put_failure;
512
513 physinif = nf_bridge_get_physinif(entskb);
514 if (physinif &&
515 nla_put_be32(skb, NFQA_IFINDEX_PHYSINDEV,
516 htonl(physinif)))
517 goto nla_put_failure;
518 }
519 #endif
520 }
521
522 if (outdev) {
523 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
524 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV, htonl(outdev->ifindex)))
525 goto nla_put_failure;
526 #else
527 if (entry->state.pf == PF_BRIDGE) {
528 /* Case 1: outdev is physical output device, we need to
529 * look for bridge group (when called from
530 * netfilter_bridge) */
531 if (nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
532 htonl(outdev->ifindex)) ||
533 /* this is the bridge group "brX" */
534 /* rcu_read_lock()ed by __nf_queue */
535 nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
536 htonl(br_port_get_rcu(outdev)->br->dev->ifindex)))
537 goto nla_put_failure;
538 } else {
539 int physoutif;
540
541 /* Case 2: outdev is bridge group, we need to look for
542 * physical output device (when called from ipv4) */
543 if (nla_put_be32(skb, NFQA_IFINDEX_OUTDEV,
544 htonl(outdev->ifindex)))
545 goto nla_put_failure;
546
547 physoutif = nf_bridge_get_physoutif(entskb);
548 if (physoutif &&
549 nla_put_be32(skb, NFQA_IFINDEX_PHYSOUTDEV,
550 htonl(physoutif)))
551 goto nla_put_failure;
552 }
553 #endif
554 }
555
556 if (entskb->mark &&
557 nla_put_be32(skb, NFQA_MARK, htonl(entskb->mark)))
558 goto nla_put_failure;
559
560 if (indev && entskb->dev &&
561 skb_mac_header_was_set(entskb) &&
562 skb_mac_header_len(entskb) != 0) {
563 struct nfqnl_msg_packet_hw phw;
564 int len;
565
566 memset(&phw, 0, sizeof(phw));
567 len = dev_parse_header(entskb, phw.hw_addr);
568 if (len) {
569 phw.hw_addrlen = htons(len);
570 if (nla_put(skb, NFQA_HWADDR, sizeof(phw), &phw))
571 goto nla_put_failure;
572 }
573 }
574
575 if (nfqnl_put_bridge(entry, skb) < 0)
576 goto nla_put_failure;
577
578 if (entry->state.hook <= NF_INET_FORWARD && entskb->tstamp) {
579 struct nfqnl_msg_packet_timestamp ts;
580 struct timespec64 kts = ktime_to_timespec64(entskb->tstamp);
581
582 ts.sec = cpu_to_be64(kts.tv_sec);
583 ts.usec = cpu_to_be64(kts.tv_nsec / NSEC_PER_USEC);
584
585 if (nla_put(skb, NFQA_TIMESTAMP, sizeof(ts), &ts))
586 goto nla_put_failure;
587 }
588
589 if ((queue->flags & NFQA_CFG_F_UID_GID) && entskb->sk &&
590 nfqnl_put_sk_uidgid(skb, entskb->sk) < 0)
591 goto nla_put_failure;
592
593 if (seclen && nla_put(skb, NFQA_SECCTX, seclen, secdata))
594 goto nla_put_failure;
595
596 if (ct && nfnl_ct->build(skb, ct, ctinfo, NFQA_CT, NFQA_CT_INFO) < 0)
597 goto nla_put_failure;
598
599 if (cap_len > data_len &&
600 nla_put_be32(skb, NFQA_CAP_LEN, htonl(cap_len)))
601 goto nla_put_failure;
602
603 if (nfqnl_put_packet_info(skb, entskb, csum_verify))
604 goto nla_put_failure;
605
606 if (data_len) {
607 struct nlattr *nla;
608
609 if (skb_tailroom(skb) < sizeof(*nla) + hlen)
610 goto nla_put_failure;
611
612 nla = skb_put(skb, sizeof(*nla));
613 nla->nla_type = NFQA_PAYLOAD;
614 nla->nla_len = nla_attr_size(data_len);
615
616 if (skb_zerocopy(skb, entskb, data_len, hlen))
617 goto nla_put_failure;
618 }
619
620 nlh->nlmsg_len = skb->len;
621 if (seclen)
622 security_release_secctx(secdata, seclen);
623 return skb;
624
625 nla_put_failure:
626 skb_tx_error(entskb);
627 kfree_skb(skb);
628 net_err_ratelimited("nf_queue: error creating packet message\n");
629 nlmsg_failure:
630 if (seclen)
631 security_release_secctx(secdata, seclen);
632 return NULL;
633 }
634
nf_ct_drop_unconfirmed(const struct nf_queue_entry * entry)635 static bool nf_ct_drop_unconfirmed(const struct nf_queue_entry *entry)
636 {
637 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
638 static const unsigned long flags = IPS_CONFIRMED | IPS_DYING;
639 const struct nf_conn *ct = (void *)skb_nfct(entry->skb);
640
641 if (ct && ((ct->status & flags) == IPS_DYING))
642 return true;
643 #endif
644 return false;
645 }
646
647 static int
__nfqnl_enqueue_packet(struct net * net,struct nfqnl_instance * queue,struct nf_queue_entry * entry)648 __nfqnl_enqueue_packet(struct net *net, struct nfqnl_instance *queue,
649 struct nf_queue_entry *entry)
650 {
651 struct sk_buff *nskb;
652 int err = -ENOBUFS;
653 __be32 *packet_id_ptr;
654 int failopen = 0;
655
656 nskb = nfqnl_build_packet_message(net, queue, entry, &packet_id_ptr);
657 if (nskb == NULL) {
658 err = -ENOMEM;
659 goto err_out;
660 }
661 spin_lock_bh(&queue->lock);
662
663 if (nf_ct_drop_unconfirmed(entry))
664 goto err_out_free_nskb;
665
666 if (queue->queue_total >= queue->queue_maxlen) {
667 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
668 failopen = 1;
669 err = 0;
670 } else {
671 queue->queue_dropped++;
672 net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
673 queue->queue_total);
674 }
675 goto err_out_free_nskb;
676 }
677 entry->id = ++queue->id_sequence;
678 *packet_id_ptr = htonl(entry->id);
679
680 /* nfnetlink_unicast will either free the nskb or add it to a socket */
681 err = nfnetlink_unicast(nskb, net, queue->peer_portid);
682 if (err < 0) {
683 if (queue->flags & NFQA_CFG_F_FAIL_OPEN) {
684 failopen = 1;
685 err = 0;
686 } else {
687 queue->queue_user_dropped++;
688 }
689 goto err_out_unlock;
690 }
691
692 __enqueue_entry(queue, entry);
693
694 spin_unlock_bh(&queue->lock);
695 return 0;
696
697 err_out_free_nskb:
698 kfree_skb(nskb);
699 err_out_unlock:
700 spin_unlock_bh(&queue->lock);
701 if (failopen)
702 nfqnl_reinject(entry, NF_ACCEPT);
703 err_out:
704 return err;
705 }
706
707 static struct nf_queue_entry *
nf_queue_entry_dup(struct nf_queue_entry * e)708 nf_queue_entry_dup(struct nf_queue_entry *e)
709 {
710 struct nf_queue_entry *entry = kmemdup(e, e->size, GFP_ATOMIC);
711
712 if (!entry)
713 return NULL;
714
715 if (nf_queue_entry_get_refs(entry))
716 return entry;
717
718 kfree(entry);
719 return NULL;
720 }
721
722 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
723 /* When called from bridge netfilter, skb->data must point to MAC header
724 * before calling skb_gso_segment(). Else, original MAC header is lost
725 * and segmented skbs will be sent to wrong destination.
726 */
nf_bridge_adjust_skb_data(struct sk_buff * skb)727 static void nf_bridge_adjust_skb_data(struct sk_buff *skb)
728 {
729 if (nf_bridge_info_get(skb))
730 __skb_push(skb, skb->network_header - skb->mac_header);
731 }
732
nf_bridge_adjust_segmented_data(struct sk_buff * skb)733 static void nf_bridge_adjust_segmented_data(struct sk_buff *skb)
734 {
735 if (nf_bridge_info_get(skb))
736 __skb_pull(skb, skb->network_header - skb->mac_header);
737 }
738 #else
739 #define nf_bridge_adjust_skb_data(s) do {} while (0)
740 #define nf_bridge_adjust_segmented_data(s) do {} while (0)
741 #endif
742
743 static int
__nfqnl_enqueue_packet_gso(struct net * net,struct nfqnl_instance * queue,struct sk_buff * skb,struct nf_queue_entry * entry)744 __nfqnl_enqueue_packet_gso(struct net *net, struct nfqnl_instance *queue,
745 struct sk_buff *skb, struct nf_queue_entry *entry)
746 {
747 int ret = -ENOMEM;
748 struct nf_queue_entry *entry_seg;
749
750 nf_bridge_adjust_segmented_data(skb);
751
752 if (skb->next == NULL) { /* last packet, no need to copy entry */
753 struct sk_buff *gso_skb = entry->skb;
754 entry->skb = skb;
755 ret = __nfqnl_enqueue_packet(net, queue, entry);
756 if (ret)
757 entry->skb = gso_skb;
758 return ret;
759 }
760
761 skb_mark_not_on_list(skb);
762
763 entry_seg = nf_queue_entry_dup(entry);
764 if (entry_seg) {
765 entry_seg->skb = skb;
766 ret = __nfqnl_enqueue_packet(net, queue, entry_seg);
767 if (ret)
768 nf_queue_entry_free(entry_seg);
769 }
770 return ret;
771 }
772
773 static int
nfqnl_enqueue_packet(struct nf_queue_entry * entry,unsigned int queuenum)774 nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
775 {
776 unsigned int queued;
777 struct nfqnl_instance *queue;
778 struct sk_buff *skb, *segs, *nskb;
779 int err = -ENOBUFS;
780 struct net *net = entry->state.net;
781 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
782
783 /* rcu_read_lock()ed by nf_hook_thresh */
784 queue = instance_lookup(q, queuenum);
785 if (!queue)
786 return -ESRCH;
787
788 if (queue->copy_mode == NFQNL_COPY_NONE)
789 return -EINVAL;
790
791 skb = entry->skb;
792
793 switch (entry->state.pf) {
794 case NFPROTO_IPV4:
795 skb->protocol = htons(ETH_P_IP);
796 break;
797 case NFPROTO_IPV6:
798 skb->protocol = htons(ETH_P_IPV6);
799 break;
800 }
801
802 if ((queue->flags & NFQA_CFG_F_GSO) || !skb_is_gso(skb))
803 return __nfqnl_enqueue_packet(net, queue, entry);
804
805 nf_bridge_adjust_skb_data(skb);
806 segs = skb_gso_segment(skb, 0);
807 /* Does not use PTR_ERR to limit the number of error codes that can be
808 * returned by nf_queue. For instance, callers rely on -ESRCH to
809 * mean 'ignore this hook'.
810 */
811 if (IS_ERR_OR_NULL(segs))
812 goto out_err;
813 queued = 0;
814 err = 0;
815 skb_list_walk_safe(segs, segs, nskb) {
816 if (err == 0)
817 err = __nfqnl_enqueue_packet_gso(net, queue,
818 segs, entry);
819 if (err == 0)
820 queued++;
821 else
822 kfree_skb(segs);
823 }
824
825 if (queued) {
826 if (err) /* some segments are already queued */
827 nf_queue_entry_free(entry);
828 kfree_skb(skb);
829 return 0;
830 }
831 out_err:
832 nf_bridge_adjust_segmented_data(skb);
833 return err;
834 }
835
836 static int
nfqnl_mangle(void * data,unsigned int data_len,struct nf_queue_entry * e,int diff)837 nfqnl_mangle(void *data, unsigned int data_len, struct nf_queue_entry *e, int diff)
838 {
839 struct sk_buff *nskb;
840
841 if (diff < 0) {
842 unsigned int min_len = skb_transport_offset(e->skb);
843
844 if (data_len < min_len)
845 return -EINVAL;
846
847 if (pskb_trim(e->skb, data_len))
848 return -ENOMEM;
849 } else if (diff > 0) {
850 if (data_len > 0xFFFF)
851 return -EINVAL;
852 if (diff > skb_tailroom(e->skb)) {
853 nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
854 diff, GFP_ATOMIC);
855 if (!nskb)
856 return -ENOMEM;
857 kfree_skb(e->skb);
858 e->skb = nskb;
859 }
860 skb_put(e->skb, diff);
861 }
862 if (skb_ensure_writable(e->skb, data_len))
863 return -ENOMEM;
864 skb_copy_to_linear_data(e->skb, data, data_len);
865 e->skb->ip_summed = CHECKSUM_NONE;
866 return 0;
867 }
868
869 static int
nfqnl_set_mode(struct nfqnl_instance * queue,unsigned char mode,unsigned int range)870 nfqnl_set_mode(struct nfqnl_instance *queue,
871 unsigned char mode, unsigned int range)
872 {
873 int status = 0;
874
875 spin_lock_bh(&queue->lock);
876 switch (mode) {
877 case NFQNL_COPY_NONE:
878 case NFQNL_COPY_META:
879 queue->copy_mode = mode;
880 queue->copy_range = 0;
881 break;
882
883 case NFQNL_COPY_PACKET:
884 queue->copy_mode = mode;
885 if (range == 0 || range > NFQNL_MAX_COPY_RANGE)
886 queue->copy_range = NFQNL_MAX_COPY_RANGE;
887 else
888 queue->copy_range = range;
889 break;
890
891 default:
892 status = -EINVAL;
893
894 }
895 spin_unlock_bh(&queue->lock);
896
897 return status;
898 }
899
900 static int
dev_cmp(struct nf_queue_entry * entry,unsigned long ifindex)901 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
902 {
903 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
904 int physinif, physoutif;
905
906 physinif = nf_bridge_get_physinif(entry->skb);
907 physoutif = nf_bridge_get_physoutif(entry->skb);
908
909 if (physinif == ifindex || physoutif == ifindex)
910 return 1;
911 #endif
912 if (entry->state.in)
913 if (entry->state.in->ifindex == ifindex)
914 return 1;
915 if (entry->state.out)
916 if (entry->state.out->ifindex == ifindex)
917 return 1;
918
919 return 0;
920 }
921
922 /* drop all packets with either indev or outdev == ifindex from all queue
923 * instances */
924 static void
nfqnl_dev_drop(struct net * net,int ifindex)925 nfqnl_dev_drop(struct net *net, int ifindex)
926 {
927 int i;
928 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
929
930 rcu_read_lock();
931
932 for (i = 0; i < INSTANCE_BUCKETS; i++) {
933 struct nfqnl_instance *inst;
934 struct hlist_head *head = &q->instance_table[i];
935
936 hlist_for_each_entry_rcu(inst, head, hlist)
937 nfqnl_flush(inst, dev_cmp, ifindex);
938 }
939
940 rcu_read_unlock();
941 }
942
943 static int
nfqnl_rcv_dev_event(struct notifier_block * this,unsigned long event,void * ptr)944 nfqnl_rcv_dev_event(struct notifier_block *this,
945 unsigned long event, void *ptr)
946 {
947 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
948
949 /* Drop any packets associated with the downed device */
950 if (event == NETDEV_DOWN)
951 nfqnl_dev_drop(dev_net(dev), dev->ifindex);
952 return NOTIFY_DONE;
953 }
954
955 static struct notifier_block nfqnl_dev_notifier = {
956 .notifier_call = nfqnl_rcv_dev_event,
957 };
958
nfqnl_nf_hook_drop(struct net * net)959 static void nfqnl_nf_hook_drop(struct net *net)
960 {
961 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
962 int i;
963
964 for (i = 0; i < INSTANCE_BUCKETS; i++) {
965 struct nfqnl_instance *inst;
966 struct hlist_head *head = &q->instance_table[i];
967
968 hlist_for_each_entry_rcu(inst, head, hlist)
969 nfqnl_flush(inst, NULL, 0);
970 }
971 }
972
973 static int
nfqnl_rcv_nl_event(struct notifier_block * this,unsigned long event,void * ptr)974 nfqnl_rcv_nl_event(struct notifier_block *this,
975 unsigned long event, void *ptr)
976 {
977 struct netlink_notify *n = ptr;
978 struct nfnl_queue_net *q = nfnl_queue_pernet(n->net);
979
980 if (event == NETLINK_URELEASE && n->protocol == NETLINK_NETFILTER) {
981 int i;
982
983 /* destroy all instances for this portid */
984 spin_lock(&q->instances_lock);
985 for (i = 0; i < INSTANCE_BUCKETS; i++) {
986 struct hlist_node *t2;
987 struct nfqnl_instance *inst;
988 struct hlist_head *head = &q->instance_table[i];
989
990 hlist_for_each_entry_safe(inst, t2, head, hlist) {
991 if (n->portid == inst->peer_portid)
992 __instance_destroy(inst);
993 }
994 }
995 spin_unlock(&q->instances_lock);
996 }
997 return NOTIFY_DONE;
998 }
999
1000 static struct notifier_block nfqnl_rtnl_notifier = {
1001 .notifier_call = nfqnl_rcv_nl_event,
1002 };
1003
1004 static const struct nla_policy nfqa_vlan_policy[NFQA_VLAN_MAX + 1] = {
1005 [NFQA_VLAN_TCI] = { .type = NLA_U16},
1006 [NFQA_VLAN_PROTO] = { .type = NLA_U16},
1007 };
1008
1009 static const struct nla_policy nfqa_verdict_policy[NFQA_MAX+1] = {
1010 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
1011 [NFQA_MARK] = { .type = NLA_U32 },
1012 [NFQA_PAYLOAD] = { .type = NLA_UNSPEC },
1013 [NFQA_CT] = { .type = NLA_UNSPEC },
1014 [NFQA_EXP] = { .type = NLA_UNSPEC },
1015 [NFQA_VLAN] = { .type = NLA_NESTED },
1016 };
1017
1018 static const struct nla_policy nfqa_verdict_batch_policy[NFQA_MAX+1] = {
1019 [NFQA_VERDICT_HDR] = { .len = sizeof(struct nfqnl_msg_verdict_hdr) },
1020 [NFQA_MARK] = { .type = NLA_U32 },
1021 };
1022
1023 static struct nfqnl_instance *
verdict_instance_lookup(struct nfnl_queue_net * q,u16 queue_num,u32 nlportid)1024 verdict_instance_lookup(struct nfnl_queue_net *q, u16 queue_num, u32 nlportid)
1025 {
1026 struct nfqnl_instance *queue;
1027
1028 queue = instance_lookup(q, queue_num);
1029 if (!queue)
1030 return ERR_PTR(-ENODEV);
1031
1032 if (queue->peer_portid != nlportid)
1033 return ERR_PTR(-EPERM);
1034
1035 return queue;
1036 }
1037
1038 static struct nfqnl_msg_verdict_hdr*
verdicthdr_get(const struct nlattr * const nfqa[])1039 verdicthdr_get(const struct nlattr * const nfqa[])
1040 {
1041 struct nfqnl_msg_verdict_hdr *vhdr;
1042 unsigned int verdict;
1043
1044 if (!nfqa[NFQA_VERDICT_HDR])
1045 return NULL;
1046
1047 vhdr = nla_data(nfqa[NFQA_VERDICT_HDR]);
1048 verdict = ntohl(vhdr->verdict) & NF_VERDICT_MASK;
1049 if (verdict > NF_MAX_VERDICT || verdict == NF_STOLEN)
1050 return NULL;
1051 return vhdr;
1052 }
1053
nfq_id_after(unsigned int id,unsigned int max)1054 static int nfq_id_after(unsigned int id, unsigned int max)
1055 {
1056 return (int)(id - max) > 0;
1057 }
1058
nfqnl_recv_verdict_batch(struct net * net,struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const nfqa[],struct netlink_ext_ack * extack)1059 static int nfqnl_recv_verdict_batch(struct net *net, struct sock *ctnl,
1060 struct sk_buff *skb,
1061 const struct nlmsghdr *nlh,
1062 const struct nlattr * const nfqa[],
1063 struct netlink_ext_ack *extack)
1064 {
1065 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1066 struct nf_queue_entry *entry, *tmp;
1067 unsigned int verdict, maxid;
1068 struct nfqnl_msg_verdict_hdr *vhdr;
1069 struct nfqnl_instance *queue;
1070 LIST_HEAD(batch_list);
1071 u16 queue_num = ntohs(nfmsg->res_id);
1072 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1073
1074 queue = verdict_instance_lookup(q, queue_num,
1075 NETLINK_CB(skb).portid);
1076 if (IS_ERR(queue))
1077 return PTR_ERR(queue);
1078
1079 vhdr = verdicthdr_get(nfqa);
1080 if (!vhdr)
1081 return -EINVAL;
1082
1083 verdict = ntohl(vhdr->verdict);
1084 maxid = ntohl(vhdr->id);
1085
1086 spin_lock_bh(&queue->lock);
1087
1088 list_for_each_entry_safe(entry, tmp, &queue->queue_list, list) {
1089 if (nfq_id_after(entry->id, maxid))
1090 break;
1091 __dequeue_entry(queue, entry);
1092 list_add_tail(&entry->list, &batch_list);
1093 }
1094
1095 spin_unlock_bh(&queue->lock);
1096
1097 if (list_empty(&batch_list))
1098 return -ENOENT;
1099
1100 list_for_each_entry_safe(entry, tmp, &batch_list, list) {
1101 if (nfqa[NFQA_MARK])
1102 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1103
1104 nfqnl_reinject(entry, verdict);
1105 }
1106 return 0;
1107 }
1108
nfqnl_ct_parse(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)1109 static struct nf_conn *nfqnl_ct_parse(struct nfnl_ct_hook *nfnl_ct,
1110 const struct nlmsghdr *nlh,
1111 const struct nlattr * const nfqa[],
1112 struct nf_queue_entry *entry,
1113 enum ip_conntrack_info *ctinfo)
1114 {
1115 struct nf_conn *ct;
1116
1117 ct = nfnl_ct->get_ct(entry->skb, ctinfo);
1118 if (ct == NULL)
1119 return NULL;
1120
1121 if (nfnl_ct->parse(nfqa[NFQA_CT], ct) < 0)
1122 return NULL;
1123
1124 if (nfqa[NFQA_EXP])
1125 nfnl_ct->attach_expect(nfqa[NFQA_EXP], ct,
1126 NETLINK_CB(entry->skb).portid,
1127 nlmsg_report(nlh));
1128 return ct;
1129 }
1130
nfqa_parse_bridge(struct nf_queue_entry * entry,const struct nlattr * const nfqa[])1131 static int nfqa_parse_bridge(struct nf_queue_entry *entry,
1132 const struct nlattr * const nfqa[])
1133 {
1134 if (nfqa[NFQA_VLAN]) {
1135 struct nlattr *tb[NFQA_VLAN_MAX + 1];
1136 int err;
1137
1138 err = nla_parse_nested_deprecated(tb, NFQA_VLAN_MAX,
1139 nfqa[NFQA_VLAN],
1140 nfqa_vlan_policy, NULL);
1141 if (err < 0)
1142 return err;
1143
1144 if (!tb[NFQA_VLAN_TCI] || !tb[NFQA_VLAN_PROTO])
1145 return -EINVAL;
1146
1147 __vlan_hwaccel_put_tag(entry->skb,
1148 nla_get_be16(tb[NFQA_VLAN_PROTO]),
1149 ntohs(nla_get_be16(tb[NFQA_VLAN_TCI])));
1150 }
1151
1152 if (nfqa[NFQA_L2HDR]) {
1153 int mac_header_len = entry->skb->network_header -
1154 entry->skb->mac_header;
1155
1156 if (mac_header_len != nla_len(nfqa[NFQA_L2HDR]))
1157 return -EINVAL;
1158 else if (mac_header_len > 0)
1159 memcpy(skb_mac_header(entry->skb),
1160 nla_data(nfqa[NFQA_L2HDR]),
1161 mac_header_len);
1162 }
1163
1164 return 0;
1165 }
1166
nfqnl_recv_verdict(struct net * net,struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const nfqa[],struct netlink_ext_ack * extack)1167 static int nfqnl_recv_verdict(struct net *net, struct sock *ctnl,
1168 struct sk_buff *skb,
1169 const struct nlmsghdr *nlh,
1170 const struct nlattr * const nfqa[],
1171 struct netlink_ext_ack *extack)
1172 {
1173 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1174 u_int16_t queue_num = ntohs(nfmsg->res_id);
1175 struct nfqnl_msg_verdict_hdr *vhdr;
1176 struct nfqnl_instance *queue;
1177 unsigned int verdict;
1178 struct nf_queue_entry *entry;
1179 enum ip_conntrack_info ctinfo;
1180 struct nfnl_ct_hook *nfnl_ct;
1181 struct nf_conn *ct = NULL;
1182 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1183 int err;
1184
1185 queue = verdict_instance_lookup(q, queue_num,
1186 NETLINK_CB(skb).portid);
1187 if (IS_ERR(queue))
1188 return PTR_ERR(queue);
1189
1190 vhdr = verdicthdr_get(nfqa);
1191 if (!vhdr)
1192 return -EINVAL;
1193
1194 verdict = ntohl(vhdr->verdict);
1195
1196 entry = find_dequeue_entry(queue, ntohl(vhdr->id));
1197 if (entry == NULL)
1198 return -ENOENT;
1199
1200 /* rcu lock already held from nfnl->call_rcu. */
1201 nfnl_ct = rcu_dereference(nfnl_ct_hook);
1202
1203 if (nfqa[NFQA_CT]) {
1204 if (nfnl_ct != NULL)
1205 ct = nfqnl_ct_parse(nfnl_ct, nlh, nfqa, entry, &ctinfo);
1206 }
1207
1208 if (entry->state.pf == PF_BRIDGE) {
1209 err = nfqa_parse_bridge(entry, nfqa);
1210 if (err < 0)
1211 return err;
1212 }
1213
1214 if (nfqa[NFQA_PAYLOAD]) {
1215 u16 payload_len = nla_len(nfqa[NFQA_PAYLOAD]);
1216 int diff = payload_len - entry->skb->len;
1217
1218 if (nfqnl_mangle(nla_data(nfqa[NFQA_PAYLOAD]),
1219 payload_len, entry, diff) < 0)
1220 verdict = NF_DROP;
1221
1222 if (ct && diff)
1223 nfnl_ct->seq_adjust(entry->skb, ct, ctinfo, diff);
1224 }
1225
1226 if (nfqa[NFQA_MARK])
1227 entry->skb->mark = ntohl(nla_get_be32(nfqa[NFQA_MARK]));
1228
1229 nfqnl_reinject(entry, verdict);
1230 return 0;
1231 }
1232
nfqnl_recv_unsupp(struct net * net,struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const nfqa[],struct netlink_ext_ack * extack)1233 static int nfqnl_recv_unsupp(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 return -ENOTSUPP;
1239 }
1240
1241 static const struct nla_policy nfqa_cfg_policy[NFQA_CFG_MAX+1] = {
1242 [NFQA_CFG_CMD] = { .len = sizeof(struct nfqnl_msg_config_cmd) },
1243 [NFQA_CFG_PARAMS] = { .len = sizeof(struct nfqnl_msg_config_params) },
1244 [NFQA_CFG_QUEUE_MAXLEN] = { .type = NLA_U32 },
1245 [NFQA_CFG_MASK] = { .type = NLA_U32 },
1246 [NFQA_CFG_FLAGS] = { .type = NLA_U32 },
1247 };
1248
1249 static const struct nf_queue_handler nfqh = {
1250 .outfn = nfqnl_enqueue_packet,
1251 .nf_hook_drop = nfqnl_nf_hook_drop,
1252 };
1253
nfqnl_recv_config(struct net * net,struct sock * ctnl,struct sk_buff * skb,const struct nlmsghdr * nlh,const struct nlattr * const nfqa[],struct netlink_ext_ack * extack)1254 static int nfqnl_recv_config(struct net *net, struct sock *ctnl,
1255 struct sk_buff *skb, const struct nlmsghdr *nlh,
1256 const struct nlattr * const nfqa[],
1257 struct netlink_ext_ack *extack)
1258 {
1259 struct nfgenmsg *nfmsg = nlmsg_data(nlh);
1260 u_int16_t queue_num = ntohs(nfmsg->res_id);
1261 struct nfqnl_instance *queue;
1262 struct nfqnl_msg_config_cmd *cmd = NULL;
1263 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1264 __u32 flags = 0, mask = 0;
1265 int ret = 0;
1266
1267 if (nfqa[NFQA_CFG_CMD]) {
1268 cmd = nla_data(nfqa[NFQA_CFG_CMD]);
1269
1270 /* Obsolete commands without queue context */
1271 switch (cmd->command) {
1272 case NFQNL_CFG_CMD_PF_BIND: return 0;
1273 case NFQNL_CFG_CMD_PF_UNBIND: return 0;
1274 }
1275 }
1276
1277 /* Check if we support these flags in first place, dependencies should
1278 * be there too not to break atomicity.
1279 */
1280 if (nfqa[NFQA_CFG_FLAGS]) {
1281 if (!nfqa[NFQA_CFG_MASK]) {
1282 /* A mask is needed to specify which flags are being
1283 * changed.
1284 */
1285 return -EINVAL;
1286 }
1287
1288 flags = ntohl(nla_get_be32(nfqa[NFQA_CFG_FLAGS]));
1289 mask = ntohl(nla_get_be32(nfqa[NFQA_CFG_MASK]));
1290
1291 if (flags >= NFQA_CFG_F_MAX)
1292 return -EOPNOTSUPP;
1293
1294 #if !IS_ENABLED(CONFIG_NETWORK_SECMARK)
1295 if (flags & mask & NFQA_CFG_F_SECCTX)
1296 return -EOPNOTSUPP;
1297 #endif
1298 if ((flags & mask & NFQA_CFG_F_CONNTRACK) &&
1299 !rcu_access_pointer(nfnl_ct_hook)) {
1300 #ifdef CONFIG_MODULES
1301 nfnl_unlock(NFNL_SUBSYS_QUEUE);
1302 request_module("ip_conntrack_netlink");
1303 nfnl_lock(NFNL_SUBSYS_QUEUE);
1304 if (rcu_access_pointer(nfnl_ct_hook))
1305 return -EAGAIN;
1306 #endif
1307 return -EOPNOTSUPP;
1308 }
1309 }
1310
1311 rcu_read_lock();
1312 queue = instance_lookup(q, queue_num);
1313 if (queue && queue->peer_portid != NETLINK_CB(skb).portid) {
1314 ret = -EPERM;
1315 goto err_out_unlock;
1316 }
1317
1318 if (cmd != NULL) {
1319 switch (cmd->command) {
1320 case NFQNL_CFG_CMD_BIND:
1321 if (queue) {
1322 ret = -EBUSY;
1323 goto err_out_unlock;
1324 }
1325 queue = instance_create(q, queue_num,
1326 NETLINK_CB(skb).portid);
1327 if (IS_ERR(queue)) {
1328 ret = PTR_ERR(queue);
1329 goto err_out_unlock;
1330 }
1331 break;
1332 case NFQNL_CFG_CMD_UNBIND:
1333 if (!queue) {
1334 ret = -ENODEV;
1335 goto err_out_unlock;
1336 }
1337 instance_destroy(q, queue);
1338 goto err_out_unlock;
1339 case NFQNL_CFG_CMD_PF_BIND:
1340 case NFQNL_CFG_CMD_PF_UNBIND:
1341 break;
1342 default:
1343 ret = -ENOTSUPP;
1344 goto err_out_unlock;
1345 }
1346 }
1347
1348 if (!queue) {
1349 ret = -ENODEV;
1350 goto err_out_unlock;
1351 }
1352
1353 if (nfqa[NFQA_CFG_PARAMS]) {
1354 struct nfqnl_msg_config_params *params =
1355 nla_data(nfqa[NFQA_CFG_PARAMS]);
1356
1357 nfqnl_set_mode(queue, params->copy_mode,
1358 ntohl(params->copy_range));
1359 }
1360
1361 if (nfqa[NFQA_CFG_QUEUE_MAXLEN]) {
1362 __be32 *queue_maxlen = nla_data(nfqa[NFQA_CFG_QUEUE_MAXLEN]);
1363
1364 spin_lock_bh(&queue->lock);
1365 queue->queue_maxlen = ntohl(*queue_maxlen);
1366 spin_unlock_bh(&queue->lock);
1367 }
1368
1369 if (nfqa[NFQA_CFG_FLAGS]) {
1370 spin_lock_bh(&queue->lock);
1371 queue->flags &= ~mask;
1372 queue->flags |= flags & mask;
1373 spin_unlock_bh(&queue->lock);
1374 }
1375
1376 err_out_unlock:
1377 rcu_read_unlock();
1378 return ret;
1379 }
1380
1381 static const struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
1382 [NFQNL_MSG_PACKET] = { .call_rcu = nfqnl_recv_unsupp,
1383 .attr_count = NFQA_MAX, },
1384 [NFQNL_MSG_VERDICT] = { .call_rcu = nfqnl_recv_verdict,
1385 .attr_count = NFQA_MAX,
1386 .policy = nfqa_verdict_policy },
1387 [NFQNL_MSG_CONFIG] = { .call = nfqnl_recv_config,
1388 .attr_count = NFQA_CFG_MAX,
1389 .policy = nfqa_cfg_policy },
1390 [NFQNL_MSG_VERDICT_BATCH]={ .call_rcu = nfqnl_recv_verdict_batch,
1391 .attr_count = NFQA_MAX,
1392 .policy = nfqa_verdict_batch_policy },
1393 };
1394
1395 static const struct nfnetlink_subsystem nfqnl_subsys = {
1396 .name = "nf_queue",
1397 .subsys_id = NFNL_SUBSYS_QUEUE,
1398 .cb_count = NFQNL_MSG_MAX,
1399 .cb = nfqnl_cb,
1400 };
1401
1402 #ifdef CONFIG_PROC_FS
1403 struct iter_state {
1404 struct seq_net_private p;
1405 unsigned int bucket;
1406 };
1407
get_first(struct seq_file * seq)1408 static struct hlist_node *get_first(struct seq_file *seq)
1409 {
1410 struct iter_state *st = seq->private;
1411 struct net *net;
1412 struct nfnl_queue_net *q;
1413
1414 if (!st)
1415 return NULL;
1416
1417 net = seq_file_net(seq);
1418 q = nfnl_queue_pernet(net);
1419 for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
1420 if (!hlist_empty(&q->instance_table[st->bucket]))
1421 return q->instance_table[st->bucket].first;
1422 }
1423 return NULL;
1424 }
1425
get_next(struct seq_file * seq,struct hlist_node * h)1426 static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
1427 {
1428 struct iter_state *st = seq->private;
1429 struct net *net = seq_file_net(seq);
1430
1431 h = h->next;
1432 while (!h) {
1433 struct nfnl_queue_net *q;
1434
1435 if (++st->bucket >= INSTANCE_BUCKETS)
1436 return NULL;
1437
1438 q = nfnl_queue_pernet(net);
1439 h = q->instance_table[st->bucket].first;
1440 }
1441 return h;
1442 }
1443
get_idx(struct seq_file * seq,loff_t pos)1444 static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
1445 {
1446 struct hlist_node *head;
1447 head = get_first(seq);
1448
1449 if (head)
1450 while (pos && (head = get_next(seq, head)))
1451 pos--;
1452 return pos ? NULL : head;
1453 }
1454
seq_start(struct seq_file * s,loff_t * pos)1455 static void *seq_start(struct seq_file *s, loff_t *pos)
1456 __acquires(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1457 {
1458 spin_lock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1459 return get_idx(s, *pos);
1460 }
1461
seq_next(struct seq_file * s,void * v,loff_t * pos)1462 static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
1463 {
1464 (*pos)++;
1465 return get_next(s, v);
1466 }
1467
seq_stop(struct seq_file * s,void * v)1468 static void seq_stop(struct seq_file *s, void *v)
1469 __releases(nfnl_queue_pernet(seq_file_net(s))->instances_lock)
1470 {
1471 spin_unlock(&nfnl_queue_pernet(seq_file_net(s))->instances_lock);
1472 }
1473
seq_show(struct seq_file * s,void * v)1474 static int seq_show(struct seq_file *s, void *v)
1475 {
1476 const struct nfqnl_instance *inst = v;
1477
1478 seq_printf(s, "%5u %6u %5u %1u %5u %5u %5u %8u %2d\n",
1479 inst->queue_num,
1480 inst->peer_portid, inst->queue_total,
1481 inst->copy_mode, inst->copy_range,
1482 inst->queue_dropped, inst->queue_user_dropped,
1483 inst->id_sequence, 1);
1484 return 0;
1485 }
1486
1487 static const struct seq_operations nfqnl_seq_ops = {
1488 .start = seq_start,
1489 .next = seq_next,
1490 .stop = seq_stop,
1491 .show = seq_show,
1492 };
1493 #endif /* PROC_FS */
1494
nfnl_queue_net_init(struct net * net)1495 static int __net_init nfnl_queue_net_init(struct net *net)
1496 {
1497 unsigned int i;
1498 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1499
1500 for (i = 0; i < INSTANCE_BUCKETS; i++)
1501 INIT_HLIST_HEAD(&q->instance_table[i]);
1502
1503 spin_lock_init(&q->instances_lock);
1504
1505 #ifdef CONFIG_PROC_FS
1506 if (!proc_create_net("nfnetlink_queue", 0440, net->nf.proc_netfilter,
1507 &nfqnl_seq_ops, sizeof(struct iter_state)))
1508 return -ENOMEM;
1509 #endif
1510 nf_register_queue_handler(net, &nfqh);
1511 return 0;
1512 }
1513
nfnl_queue_net_exit(struct net * net)1514 static void __net_exit nfnl_queue_net_exit(struct net *net)
1515 {
1516 struct nfnl_queue_net *q = nfnl_queue_pernet(net);
1517 unsigned int i;
1518
1519 nf_unregister_queue_handler(net);
1520 #ifdef CONFIG_PROC_FS
1521 remove_proc_entry("nfnetlink_queue", net->nf.proc_netfilter);
1522 #endif
1523 for (i = 0; i < INSTANCE_BUCKETS; i++)
1524 WARN_ON_ONCE(!hlist_empty(&q->instance_table[i]));
1525 }
1526
nfnl_queue_net_exit_batch(struct list_head * net_exit_list)1527 static void nfnl_queue_net_exit_batch(struct list_head *net_exit_list)
1528 {
1529 synchronize_rcu();
1530 }
1531
1532 static struct pernet_operations nfnl_queue_net_ops = {
1533 .init = nfnl_queue_net_init,
1534 .exit = nfnl_queue_net_exit,
1535 .exit_batch = nfnl_queue_net_exit_batch,
1536 .id = &nfnl_queue_net_id,
1537 .size = sizeof(struct nfnl_queue_net),
1538 };
1539
nfnetlink_queue_init(void)1540 static int __init nfnetlink_queue_init(void)
1541 {
1542 int status;
1543
1544 status = register_pernet_subsys(&nfnl_queue_net_ops);
1545 if (status < 0) {
1546 pr_err("failed to register pernet ops\n");
1547 goto out;
1548 }
1549
1550 netlink_register_notifier(&nfqnl_rtnl_notifier);
1551 status = nfnetlink_subsys_register(&nfqnl_subsys);
1552 if (status < 0) {
1553 pr_err("failed to create netlink socket\n");
1554 goto cleanup_netlink_notifier;
1555 }
1556
1557 status = register_netdevice_notifier(&nfqnl_dev_notifier);
1558 if (status < 0) {
1559 pr_err("failed to register netdevice notifier\n");
1560 goto cleanup_netlink_subsys;
1561 }
1562
1563 return status;
1564
1565 cleanup_netlink_subsys:
1566 nfnetlink_subsys_unregister(&nfqnl_subsys);
1567 cleanup_netlink_notifier:
1568 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1569 unregister_pernet_subsys(&nfnl_queue_net_ops);
1570 out:
1571 return status;
1572 }
1573
nfnetlink_queue_fini(void)1574 static void __exit nfnetlink_queue_fini(void)
1575 {
1576 unregister_netdevice_notifier(&nfqnl_dev_notifier);
1577 nfnetlink_subsys_unregister(&nfqnl_subsys);
1578 netlink_unregister_notifier(&nfqnl_rtnl_notifier);
1579 unregister_pernet_subsys(&nfnl_queue_net_ops);
1580
1581 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1582 }
1583
1584 MODULE_DESCRIPTION("netfilter packet queue handler");
1585 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1586 MODULE_LICENSE("GPL");
1587 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE);
1588
1589 module_init(nfnetlink_queue_init);
1590 module_exit(nfnetlink_queue_fini);
1591