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