2 * This is a module which is used for queueing packets and communicating with
3 * userspace via nfnetlink.
5 * (C) 2005 by Harald Welte <laforge@netfilter.org>
6 * (C) 2007 by Patrick McHardy <kaber@trash.net>
8 * Based on the old ipv4-only ip_queue.c:
9 * (C) 2000-2002 James Morris <jmorris@intercode.com.au>
10 * (C) 2003-2005 Netfilter Core Team <coreteam@netfilter.org>
12 * This program is free software; you can redistribute it and/or modify
13 * it under the terms of the GNU General Public License version 2 as
14 * published by the Free Software Foundation.
17 #include <linux/module.h>
18 #include <linux/skbuff.h>
19 #include <linux/init.h>
20 #include <linux/spinlock.h>
21 #include <linux/slab.h>
22 #include <linux/notifier.h>
23 #include <linux/netdevice.h>
24 #include <linux/netfilter.h>
25 #include <linux/proc_fs.h>
26 #include <linux/netfilter_ipv4.h>
27 #include <linux/netfilter_ipv6.h>
28 #include <linux/netfilter_bridge.h>
29 #include <linux/netfilter/nfnetlink.h>
30 #include <linux/netfilter/nfnetlink_queue.h>
31 #include <linux/netfilter/nf_conntrack_common.h>
32 #include <linux/list.h>
34 #include <net/tcp_states.h>
35 #include <net/netfilter/nf_queue.h>
36 #include <net/netns/generic.h>
38 #include <linux/atomic.h>
40 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
41 #include "../bridge/br_private.h"
44 #define NFQNL_QMAX_DEFAULT 1024
46 /* We're using struct nlattr which has 16bit nla_len. Note that nla_len
47 * includes the header length. Thus, the maximum packet length that we
48 * support is 65531 bytes. We send truncated packets if the specified length
49 * is larger than that. Userspace can check for presence of NFQA_CAP_LEN
50 * attribute to detect truncation.
52 #define NFQNL_MAX_COPY_RANGE (0xffff - NLA_HDRLEN)
54 struct nfqnl_instance
{
55 struct hlist_node hlist
; /* global list of queues */
59 unsigned int queue_maxlen
;
60 unsigned int copy_range
;
61 unsigned int queue_dropped
;
62 unsigned int queue_user_dropped
;
65 u_int16_t queue_num
; /* number of this queue */
67 u_int32_t flags
; /* Set using NFQA_CFG_FLAGS */
69 * Following fields are dirtied for each queued packet,
70 * keep them in same cache line if possible.
73 unsigned int queue_total
;
74 unsigned int id_sequence
; /* 'sequence' of pkt ids */
75 struct list_head queue_list
; /* packets in queue */
78 typedef int (*nfqnl_cmpfn
)(struct nf_queue_entry
*, unsigned long);
80 static int nfnl_queue_net_id __read_mostly
;
82 #define INSTANCE_BUCKETS 16
83 struct nfnl_queue_net
{
84 spinlock_t instances_lock
;
85 struct hlist_head instance_table
[INSTANCE_BUCKETS
];
88 static struct nfnl_queue_net
*nfnl_queue_pernet(struct net
*net
)
90 return net_generic(net
, nfnl_queue_net_id
);
93 static inline u_int8_t
instance_hashfn(u_int16_t queue_num
)
95 return ((queue_num
>> 8) ^ queue_num
) % INSTANCE_BUCKETS
;
98 static struct nfqnl_instance
*
99 instance_lookup(struct nfnl_queue_net
*q
, u_int16_t queue_num
)
101 struct hlist_head
*head
;
102 struct nfqnl_instance
*inst
;
104 head
= &q
->instance_table
[instance_hashfn(queue_num
)];
105 hlist_for_each_entry_rcu(inst
, head
, hlist
) {
106 if (inst
->queue_num
== queue_num
)
112 static struct nfqnl_instance
*
113 instance_create(struct nfnl_queue_net
*q
, u_int16_t queue_num
, u32 portid
)
115 struct nfqnl_instance
*inst
;
119 spin_lock(&q
->instances_lock
);
120 if (instance_lookup(q
, queue_num
)) {
125 inst
= kzalloc(sizeof(*inst
), GFP_ATOMIC
);
131 inst
->queue_num
= queue_num
;
132 inst
->peer_portid
= portid
;
133 inst
->queue_maxlen
= NFQNL_QMAX_DEFAULT
;
134 inst
->copy_range
= NFQNL_MAX_COPY_RANGE
;
135 inst
->copy_mode
= NFQNL_COPY_NONE
;
136 spin_lock_init(&inst
->lock
);
137 INIT_LIST_HEAD(&inst
->queue_list
);
139 if (!try_module_get(THIS_MODULE
)) {
144 h
= instance_hashfn(queue_num
);
145 hlist_add_head_rcu(&inst
->hlist
, &q
->instance_table
[h
]);
147 spin_unlock(&q
->instances_lock
);
154 spin_unlock(&q
->instances_lock
);
158 static void nfqnl_flush(struct nfqnl_instance
*queue
, nfqnl_cmpfn cmpfn
,
162 instance_destroy_rcu(struct rcu_head
*head
)
164 struct nfqnl_instance
*inst
= container_of(head
, struct nfqnl_instance
,
167 nfqnl_flush(inst
, NULL
, 0);
169 module_put(THIS_MODULE
);
173 __instance_destroy(struct nfqnl_instance
*inst
)
175 hlist_del_rcu(&inst
->hlist
);
176 call_rcu(&inst
->rcu
, instance_destroy_rcu
);
180 instance_destroy(struct nfnl_queue_net
*q
, struct nfqnl_instance
*inst
)
182 spin_lock(&q
->instances_lock
);
183 __instance_destroy(inst
);
184 spin_unlock(&q
->instances_lock
);
188 __enqueue_entry(struct nfqnl_instance
*queue
, struct nf_queue_entry
*entry
)
190 list_add_tail(&entry
->list
, &queue
->queue_list
);
191 queue
->queue_total
++;
195 __dequeue_entry(struct nfqnl_instance
*queue
, struct nf_queue_entry
*entry
)
197 list_del(&entry
->list
);
198 queue
->queue_total
--;
201 static struct nf_queue_entry
*
202 find_dequeue_entry(struct nfqnl_instance
*queue
, unsigned int id
)
204 struct nf_queue_entry
*entry
= NULL
, *i
;
206 spin_lock_bh(&queue
->lock
);
208 list_for_each_entry(i
, &queue
->queue_list
, list
) {
216 __dequeue_entry(queue
, entry
);
218 spin_unlock_bh(&queue
->lock
);
224 nfqnl_flush(struct nfqnl_instance
*queue
, nfqnl_cmpfn cmpfn
, unsigned long data
)
226 struct nf_queue_entry
*entry
, *next
;
228 spin_lock_bh(&queue
->lock
);
229 list_for_each_entry_safe(entry
, next
, &queue
->queue_list
, list
) {
230 if (!cmpfn
|| cmpfn(entry
, data
)) {
231 list_del(&entry
->list
);
232 queue
->queue_total
--;
233 nf_reinject(entry
, NF_DROP
);
236 spin_unlock_bh(&queue
->lock
);
240 nfqnl_put_packet_info(struct sk_buff
*nlskb
, struct sk_buff
*packet
,
245 if (packet
->ip_summed
== CHECKSUM_PARTIAL
)
246 flags
= NFQA_SKB_CSUMNOTREADY
;
247 else if (csum_verify
)
248 flags
= NFQA_SKB_CSUM_NOTVERIFIED
;
250 if (skb_is_gso(packet
))
251 flags
|= NFQA_SKB_GSO
;
253 return flags
? nla_put_be32(nlskb
, NFQA_SKB_INFO
, htonl(flags
)) : 0;
256 static int nfqnl_put_sk_uidgid(struct sk_buff
*skb
, struct sock
*sk
)
258 const struct cred
*cred
;
260 if (!sk_fullsock(sk
))
263 read_lock_bh(&sk
->sk_callback_lock
);
264 if (sk
->sk_socket
&& sk
->sk_socket
->file
) {
265 cred
= sk
->sk_socket
->file
->f_cred
;
266 if (nla_put_be32(skb
, NFQA_UID
,
267 htonl(from_kuid_munged(&init_user_ns
, cred
->fsuid
))))
268 goto nla_put_failure
;
269 if (nla_put_be32(skb
, NFQA_GID
,
270 htonl(from_kgid_munged(&init_user_ns
, cred
->fsgid
))))
271 goto nla_put_failure
;
273 read_unlock_bh(&sk
->sk_callback_lock
);
277 read_unlock_bh(&sk
->sk_callback_lock
);
281 static u32
nfqnl_get_sk_secctx(struct sk_buff
*skb
, char **secdata
)
284 #if IS_ENABLED(CONFIG_NETWORK_SECMARK)
285 if (!skb
|| !sk_fullsock(skb
->sk
))
288 read_lock_bh(&skb
->sk
->sk_callback_lock
);
291 security_secid_to_secctx(skb
->secmark
, secdata
, &seclen
);
293 read_unlock_bh(&skb
->sk
->sk_callback_lock
);
298 static struct sk_buff
*
299 nfqnl_build_packet_message(struct net
*net
, struct nfqnl_instance
*queue
,
300 struct nf_queue_entry
*entry
,
301 __be32
**packet_id_ptr
)
304 size_t data_len
= 0, cap_len
= 0, rem_len
= 0;
305 unsigned int hlen
= 0;
308 struct nfqnl_msg_packet_hdr
*pmsg
;
309 struct nlmsghdr
*nlh
;
310 struct nfgenmsg
*nfmsg
;
311 struct sk_buff
*entskb
= entry
->skb
;
312 struct net_device
*indev
;
313 struct net_device
*outdev
;
314 struct nf_conn
*ct
= NULL
;
315 enum ip_conntrack_info
uninitialized_var(ctinfo
);
316 struct nfnl_ct_hook
*nfnl_ct
;
318 char *secdata
= NULL
;
321 size
= nlmsg_total_size(sizeof(struct nfgenmsg
))
322 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr
))
323 + nla_total_size(sizeof(u_int32_t
)) /* ifindex */
324 + nla_total_size(sizeof(u_int32_t
)) /* ifindex */
325 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
326 + nla_total_size(sizeof(u_int32_t
)) /* ifindex */
327 + nla_total_size(sizeof(u_int32_t
)) /* ifindex */
329 + nla_total_size(sizeof(u_int32_t
)) /* mark */
330 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw
))
331 + nla_total_size(sizeof(u_int32_t
)) /* skbinfo */
332 + nla_total_size(sizeof(u_int32_t
)); /* cap_len */
334 if (entskb
->tstamp
.tv64
)
335 size
+= nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp
));
337 if (entry
->state
.hook
<= NF_INET_FORWARD
||
338 (entry
->state
.hook
== NF_INET_POST_ROUTING
&& entskb
->sk
== NULL
))
339 csum_verify
= !skb_csum_unnecessary(entskb
);
343 outdev
= entry
->state
.out
;
345 switch ((enum nfqnl_config_mode
)ACCESS_ONCE(queue
->copy_mode
)) {
346 case NFQNL_COPY_META
:
347 case NFQNL_COPY_NONE
:
350 case NFQNL_COPY_PACKET
:
351 if (!(queue
->flags
& NFQA_CFG_F_GSO
) &&
352 entskb
->ip_summed
== CHECKSUM_PARTIAL
&&
353 skb_checksum_help(entskb
))
356 data_len
= ACCESS_ONCE(queue
->copy_range
);
357 if (data_len
> entskb
->len
)
358 data_len
= entskb
->len
;
360 hlen
= skb_zerocopy_headlen(entskb
);
361 hlen
= min_t(unsigned int, hlen
, data_len
);
362 size
+= sizeof(struct nlattr
) + hlen
;
363 cap_len
= entskb
->len
;
364 rem_len
= data_len
- hlen
;
368 if (queue
->flags
& NFQA_CFG_F_CONNTRACK
) {
369 nfnl_ct
= rcu_dereference(nfnl_ct_hook
);
370 if (nfnl_ct
!= NULL
) {
371 ct
= nfnl_ct
->get_ct(entskb
, &ctinfo
);
373 size
+= nfnl_ct
->build_size(ct
);
377 if (queue
->flags
& NFQA_CFG_F_UID_GID
) {
378 size
+= (nla_total_size(sizeof(u_int32_t
)) /* uid */
379 + nla_total_size(sizeof(u_int32_t
))); /* gid */
382 if ((queue
->flags
& NFQA_CFG_F_SECCTX
) && entskb
->sk
) {
383 seclen
= nfqnl_get_sk_secctx(entskb
, &secdata
);
385 size
+= nla_total_size(seclen
);
388 skb
= __netlink_alloc_skb(net
->nfnl
, size
, rem_len
, queue
->peer_portid
,
391 skb_tx_error(entskb
);
395 nlh
= nlmsg_put(skb
, 0, 0,
396 NFNL_SUBSYS_QUEUE
<< 8 | NFQNL_MSG_PACKET
,
397 sizeof(struct nfgenmsg
), 0);
399 skb_tx_error(entskb
);
403 nfmsg
= nlmsg_data(nlh
);
404 nfmsg
->nfgen_family
= entry
->state
.pf
;
405 nfmsg
->version
= NFNETLINK_V0
;
406 nfmsg
->res_id
= htons(queue
->queue_num
);
408 nla
= __nla_reserve(skb
, NFQA_PACKET_HDR
, sizeof(*pmsg
));
409 pmsg
= nla_data(nla
);
410 pmsg
->hw_protocol
= entskb
->protocol
;
411 pmsg
->hook
= entry
->state
.hook
;
412 *packet_id_ptr
= &pmsg
->packet_id
;
414 indev
= entry
->state
.in
;
416 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
417 if (nla_put_be32(skb
, NFQA_IFINDEX_INDEV
, htonl(indev
->ifindex
)))
418 goto nla_put_failure
;
420 if (entry
->state
.pf
== PF_BRIDGE
) {
421 /* Case 1: indev is physical input device, we need to
422 * look for bridge group (when called from
423 * netfilter_bridge) */
424 if (nla_put_be32(skb
, NFQA_IFINDEX_PHYSINDEV
,
425 htonl(indev
->ifindex
)) ||
426 /* this is the bridge group "brX" */
427 /* rcu_read_lock()ed by __nf_queue */
428 nla_put_be32(skb
, NFQA_IFINDEX_INDEV
,
429 htonl(br_port_get_rcu(indev
)->br
->dev
->ifindex
)))
430 goto nla_put_failure
;
434 /* Case 2: indev is bridge group, we need to look for
435 * physical device (when called from ipv4) */
436 if (nla_put_be32(skb
, NFQA_IFINDEX_INDEV
,
437 htonl(indev
->ifindex
)))
438 goto nla_put_failure
;
440 physinif
= nf_bridge_get_physinif(entskb
);
442 nla_put_be32(skb
, NFQA_IFINDEX_PHYSINDEV
,
444 goto nla_put_failure
;
450 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
451 if (nla_put_be32(skb
, NFQA_IFINDEX_OUTDEV
, htonl(outdev
->ifindex
)))
452 goto nla_put_failure
;
454 if (entry
->state
.pf
== PF_BRIDGE
) {
455 /* Case 1: outdev is physical output device, we need to
456 * look for bridge group (when called from
457 * netfilter_bridge) */
458 if (nla_put_be32(skb
, NFQA_IFINDEX_PHYSOUTDEV
,
459 htonl(outdev
->ifindex
)) ||
460 /* this is the bridge group "brX" */
461 /* rcu_read_lock()ed by __nf_queue */
462 nla_put_be32(skb
, NFQA_IFINDEX_OUTDEV
,
463 htonl(br_port_get_rcu(outdev
)->br
->dev
->ifindex
)))
464 goto nla_put_failure
;
468 /* Case 2: outdev is bridge group, we need to look for
469 * physical output device (when called from ipv4) */
470 if (nla_put_be32(skb
, NFQA_IFINDEX_OUTDEV
,
471 htonl(outdev
->ifindex
)))
472 goto nla_put_failure
;
474 physoutif
= nf_bridge_get_physoutif(entskb
);
476 nla_put_be32(skb
, NFQA_IFINDEX_PHYSOUTDEV
,
478 goto nla_put_failure
;
484 nla_put_be32(skb
, NFQA_MARK
, htonl(entskb
->mark
)))
485 goto nla_put_failure
;
487 if (indev
&& entskb
->dev
&&
488 entskb
->mac_header
!= entskb
->network_header
) {
489 struct nfqnl_msg_packet_hw phw
;
492 memset(&phw
, 0, sizeof(phw
));
493 len
= dev_parse_header(entskb
, phw
.hw_addr
);
495 phw
.hw_addrlen
= htons(len
);
496 if (nla_put(skb
, NFQA_HWADDR
, sizeof(phw
), &phw
))
497 goto nla_put_failure
;
501 if (entskb
->tstamp
.tv64
) {
502 struct nfqnl_msg_packet_timestamp ts
;
503 struct timespec64 kts
= ktime_to_timespec64(skb
->tstamp
);
505 ts
.sec
= cpu_to_be64(kts
.tv_sec
);
506 ts
.usec
= cpu_to_be64(kts
.tv_nsec
/ NSEC_PER_USEC
);
508 if (nla_put(skb
, NFQA_TIMESTAMP
, sizeof(ts
), &ts
))
509 goto nla_put_failure
;
512 if ((queue
->flags
& NFQA_CFG_F_UID_GID
) && entskb
->sk
&&
513 nfqnl_put_sk_uidgid(skb
, entskb
->sk
) < 0)
514 goto nla_put_failure
;
516 if (seclen
&& nla_put(skb
, NFQA_SECCTX
, seclen
, secdata
))
517 goto nla_put_failure
;
519 if (ct
&& nfnl_ct
->build(skb
, ct
, ctinfo
, NFQA_CT
, NFQA_CT_INFO
) < 0)
520 goto nla_put_failure
;
522 if (cap_len
> data_len
&&
523 nla_put_be32(skb
, NFQA_CAP_LEN
, htonl(cap_len
)))
524 goto nla_put_failure
;
526 if (nfqnl_put_packet_info(skb
, entskb
, csum_verify
))
527 goto nla_put_failure
;
532 if (skb_tailroom(skb
) < sizeof(*nla
) + hlen
)
533 goto nla_put_failure
;
535 nla
= (struct nlattr
*)skb_put(skb
, sizeof(*nla
));
536 nla
->nla_type
= NFQA_PAYLOAD
;
537 nla
->nla_len
= nla_attr_size(data_len
);
539 if (skb_zerocopy(skb
, entskb
, data_len
, hlen
))
540 goto nla_put_failure
;
543 nlh
->nlmsg_len
= skb
->len
;
547 skb_tx_error(entskb
);
549 net_err_ratelimited("nf_queue: error creating packet message\n");
554 __nfqnl_enqueue_packet(struct net
*net
, struct nfqnl_instance
*queue
,
555 struct nf_queue_entry
*entry
)
557 struct sk_buff
*nskb
;
559 __be32
*packet_id_ptr
;
562 nskb
= nfqnl_build_packet_message(net
, queue
, entry
, &packet_id_ptr
);
567 spin_lock_bh(&queue
->lock
);
569 if (queue
->queue_total
>= queue
->queue_maxlen
) {
570 if (queue
->flags
& NFQA_CFG_F_FAIL_OPEN
) {
574 queue
->queue_dropped
++;
575 net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
578 goto err_out_free_nskb
;
580 entry
->id
= ++queue
->id_sequence
;
581 *packet_id_ptr
= htonl(entry
->id
);
583 /* nfnetlink_unicast will either free the nskb or add it to a socket */
584 err
= nfnetlink_unicast(nskb
, net
, queue
->peer_portid
, MSG_DONTWAIT
);
586 queue
->queue_user_dropped
++;
590 __enqueue_entry(queue
, entry
);
592 spin_unlock_bh(&queue
->lock
);
598 spin_unlock_bh(&queue
->lock
);
600 nf_reinject(entry
, NF_ACCEPT
);
605 static struct nf_queue_entry
*
606 nf_queue_entry_dup(struct nf_queue_entry
*e
)
608 struct nf_queue_entry
*entry
= kmemdup(e
, e
->size
, GFP_ATOMIC
);
610 nf_queue_entry_get_refs(entry
);
614 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
615 /* When called from bridge netfilter, skb->data must point to MAC header
616 * before calling skb_gso_segment(). Else, original MAC header is lost
617 * and segmented skbs will be sent to wrong destination.
619 static void nf_bridge_adjust_skb_data(struct sk_buff
*skb
)
622 __skb_push(skb
, skb
->network_header
- skb
->mac_header
);
625 static void nf_bridge_adjust_segmented_data(struct sk_buff
*skb
)
628 __skb_pull(skb
, skb
->network_header
- skb
->mac_header
);
631 #define nf_bridge_adjust_skb_data(s) do {} while (0)
632 #define nf_bridge_adjust_segmented_data(s) do {} while (0)
635 static void free_entry(struct nf_queue_entry
*entry
)
637 nf_queue_entry_release_refs(entry
);
642 __nfqnl_enqueue_packet_gso(struct net
*net
, struct nfqnl_instance
*queue
,
643 struct sk_buff
*skb
, struct nf_queue_entry
*entry
)
646 struct nf_queue_entry
*entry_seg
;
648 nf_bridge_adjust_segmented_data(skb
);
650 if (skb
->next
== NULL
) { /* last packet, no need to copy entry */
651 struct sk_buff
*gso_skb
= entry
->skb
;
653 ret
= __nfqnl_enqueue_packet(net
, queue
, entry
);
655 entry
->skb
= gso_skb
;
661 entry_seg
= nf_queue_entry_dup(entry
);
663 entry_seg
->skb
= skb
;
664 ret
= __nfqnl_enqueue_packet(net
, queue
, entry_seg
);
666 free_entry(entry_seg
);
672 nfqnl_enqueue_packet(struct nf_queue_entry
*entry
, unsigned int queuenum
)
675 struct nfqnl_instance
*queue
;
676 struct sk_buff
*skb
, *segs
;
678 struct net
*net
= entry
->state
.net
;
679 struct nfnl_queue_net
*q
= nfnl_queue_pernet(net
);
681 /* rcu_read_lock()ed by nf_hook_slow() */
682 queue
= instance_lookup(q
, queuenum
);
686 if (queue
->copy_mode
== NFQNL_COPY_NONE
)
691 switch (entry
->state
.pf
) {
693 skb
->protocol
= htons(ETH_P_IP
);
696 skb
->protocol
= htons(ETH_P_IPV6
);
700 if ((queue
->flags
& NFQA_CFG_F_GSO
) || !skb_is_gso(skb
))
701 return __nfqnl_enqueue_packet(net
, queue
, entry
);
703 nf_bridge_adjust_skb_data(skb
);
704 segs
= skb_gso_segment(skb
, 0);
705 /* Does not use PTR_ERR to limit the number of error codes that can be
706 * returned by nf_queue. For instance, callers rely on -ESRCH to
707 * mean 'ignore this hook'.
709 if (IS_ERR_OR_NULL(segs
))
714 struct sk_buff
*nskb
= segs
->next
;
716 err
= __nfqnl_enqueue_packet_gso(net
, queue
,
726 if (err
) /* some segments are already queued */
732 nf_bridge_adjust_segmented_data(skb
);
737 nfqnl_mangle(void *data
, int data_len
, struct nf_queue_entry
*e
, int diff
)
739 struct sk_buff
*nskb
;
742 if (pskb_trim(e
->skb
, data_len
))
744 } else if (diff
> 0) {
745 if (data_len
> 0xFFFF)
747 if (diff
> skb_tailroom(e
->skb
)) {
748 nskb
= skb_copy_expand(e
->skb
, skb_headroom(e
->skb
),
751 printk(KERN_WARNING
"nf_queue: OOM "
752 "in mangle, dropping packet\n");
758 skb_put(e
->skb
, diff
);
760 if (!skb_make_writable(e
->skb
, data_len
))
762 skb_copy_to_linear_data(e
->skb
, data
, data_len
);
763 e
->skb
->ip_summed
= CHECKSUM_NONE
;
768 nfqnl_set_mode(struct nfqnl_instance
*queue
,
769 unsigned char mode
, unsigned int range
)
773 spin_lock_bh(&queue
->lock
);
775 case NFQNL_COPY_NONE
:
776 case NFQNL_COPY_META
:
777 queue
->copy_mode
= mode
;
778 queue
->copy_range
= 0;
781 case NFQNL_COPY_PACKET
:
782 queue
->copy_mode
= mode
;
783 if (range
== 0 || range
> NFQNL_MAX_COPY_RANGE
)
784 queue
->copy_range
= NFQNL_MAX_COPY_RANGE
;
786 queue
->copy_range
= range
;
793 spin_unlock_bh(&queue
->lock
);
799 dev_cmp(struct nf_queue_entry
*entry
, unsigned long ifindex
)
802 if (entry
->state
.in
->ifindex
== ifindex
)
804 if (entry
->state
.out
)
805 if (entry
->state
.out
->ifindex
== ifindex
)
807 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
808 if (entry
->skb
->nf_bridge
) {
809 int physinif
, physoutif
;
811 physinif
= nf_bridge_get_physinif(entry
->skb
);
812 physoutif
= nf_bridge_get_physoutif(entry
->skb
);
814 if (physinif
== ifindex
|| physoutif
== ifindex
)
821 /* drop all packets with either indev or outdev == ifindex from all queue
824 nfqnl_dev_drop(struct net
*net
, int ifindex
)
827 struct nfnl_queue_net
*q
= nfnl_queue_pernet(net
);
831 for (i
= 0; i
< INSTANCE_BUCKETS
; i
++) {
832 struct nfqnl_instance
*inst
;
833 struct hlist_head
*head
= &q
->instance_table
[i
];
835 hlist_for_each_entry_rcu(inst
, head
, hlist
)
836 nfqnl_flush(inst
, dev_cmp
, ifindex
);
843 nfqnl_rcv_dev_event(struct notifier_block
*this,
844 unsigned long event
, void *ptr
)
846 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
848 /* Drop any packets associated with the downed device */
849 if (event
== NETDEV_DOWN
)
850 nfqnl_dev_drop(dev_net(dev
), dev
->ifindex
);
854 static struct notifier_block nfqnl_dev_notifier
= {
855 .notifier_call
= nfqnl_rcv_dev_event
,
858 static int nf_hook_cmp(struct nf_queue_entry
*entry
, unsigned long ops_ptr
)
860 return entry
->elem
== (struct nf_hook_ops
*)ops_ptr
;
863 static void nfqnl_nf_hook_drop(struct net
*net
, struct nf_hook_ops
*hook
)
865 struct nfnl_queue_net
*q
= nfnl_queue_pernet(net
);
869 for (i
= 0; i
< INSTANCE_BUCKETS
; i
++) {
870 struct nfqnl_instance
*inst
;
871 struct hlist_head
*head
= &q
->instance_table
[i
];
873 hlist_for_each_entry_rcu(inst
, head
, hlist
)
874 nfqnl_flush(inst
, nf_hook_cmp
, (unsigned long)hook
);
880 nfqnl_rcv_nl_event(struct notifier_block
*this,
881 unsigned long event
, void *ptr
)
883 struct netlink_notify
*n
= ptr
;
884 struct nfnl_queue_net
*q
= nfnl_queue_pernet(n
->net
);
886 if (event
== NETLINK_URELEASE
&& n
->protocol
== NETLINK_NETFILTER
) {
889 /* destroy all instances for this portid */
890 spin_lock(&q
->instances_lock
);
891 for (i
= 0; i
< INSTANCE_BUCKETS
; i
++) {
892 struct hlist_node
*t2
;
893 struct nfqnl_instance
*inst
;
894 struct hlist_head
*head
= &q
->instance_table
[i
];
896 hlist_for_each_entry_safe(inst
, t2
, head
, hlist
) {
897 if (n
->portid
== inst
->peer_portid
)
898 __instance_destroy(inst
);
901 spin_unlock(&q
->instances_lock
);
906 static struct notifier_block nfqnl_rtnl_notifier
= {
907 .notifier_call
= nfqnl_rcv_nl_event
,
910 static const struct nla_policy nfqa_verdict_policy
[NFQA_MAX
+1] = {
911 [NFQA_VERDICT_HDR
] = { .len
= sizeof(struct nfqnl_msg_verdict_hdr
) },
912 [NFQA_MARK
] = { .type
= NLA_U32
},
913 [NFQA_PAYLOAD
] = { .type
= NLA_UNSPEC
},
914 [NFQA_CT
] = { .type
= NLA_UNSPEC
},
915 [NFQA_EXP
] = { .type
= NLA_UNSPEC
},
918 static const struct nla_policy nfqa_verdict_batch_policy
[NFQA_MAX
+1] = {
919 [NFQA_VERDICT_HDR
] = { .len
= sizeof(struct nfqnl_msg_verdict_hdr
) },
920 [NFQA_MARK
] = { .type
= NLA_U32
},
923 static struct nfqnl_instance
*
924 verdict_instance_lookup(struct nfnl_queue_net
*q
, u16 queue_num
, u32 nlportid
)
926 struct nfqnl_instance
*queue
;
928 queue
= instance_lookup(q
, queue_num
);
930 return ERR_PTR(-ENODEV
);
932 if (queue
->peer_portid
!= nlportid
)
933 return ERR_PTR(-EPERM
);
938 static struct nfqnl_msg_verdict_hdr
*
939 verdicthdr_get(const struct nlattr
* const nfqa
[])
941 struct nfqnl_msg_verdict_hdr
*vhdr
;
942 unsigned int verdict
;
944 if (!nfqa
[NFQA_VERDICT_HDR
])
947 vhdr
= nla_data(nfqa
[NFQA_VERDICT_HDR
]);
948 verdict
= ntohl(vhdr
->verdict
) & NF_VERDICT_MASK
;
949 if (verdict
> NF_MAX_VERDICT
|| verdict
== NF_STOLEN
)
954 static int nfq_id_after(unsigned int id
, unsigned int max
)
956 return (int)(id
- max
) > 0;
960 nfqnl_recv_verdict_batch(struct sock
*ctnl
, struct sk_buff
*skb
,
961 const struct nlmsghdr
*nlh
,
962 const struct nlattr
* const nfqa
[])
964 struct nfgenmsg
*nfmsg
= nlmsg_data(nlh
);
965 struct nf_queue_entry
*entry
, *tmp
;
966 unsigned int verdict
, maxid
;
967 struct nfqnl_msg_verdict_hdr
*vhdr
;
968 struct nfqnl_instance
*queue
;
969 LIST_HEAD(batch_list
);
970 u16 queue_num
= ntohs(nfmsg
->res_id
);
972 struct net
*net
= sock_net(ctnl
);
973 struct nfnl_queue_net
*q
= nfnl_queue_pernet(net
);
975 queue
= verdict_instance_lookup(q
, queue_num
,
976 NETLINK_CB(skb
).portid
);
978 return PTR_ERR(queue
);
980 vhdr
= verdicthdr_get(nfqa
);
984 verdict
= ntohl(vhdr
->verdict
);
985 maxid
= ntohl(vhdr
->id
);
987 spin_lock_bh(&queue
->lock
);
989 list_for_each_entry_safe(entry
, tmp
, &queue
->queue_list
, list
) {
990 if (nfq_id_after(entry
->id
, maxid
))
992 __dequeue_entry(queue
, entry
);
993 list_add_tail(&entry
->list
, &batch_list
);
996 spin_unlock_bh(&queue
->lock
);
998 if (list_empty(&batch_list
))
1001 list_for_each_entry_safe(entry
, tmp
, &batch_list
, list
) {
1002 if (nfqa
[NFQA_MARK
])
1003 entry
->skb
->mark
= ntohl(nla_get_be32(nfqa
[NFQA_MARK
]));
1004 nf_reinject(entry
, verdict
);
1009 static struct nf_conn
*nfqnl_ct_parse(struct nfnl_ct_hook
*nfnl_ct
,
1010 const struct nlmsghdr
*nlh
,
1011 const struct nlattr
* const nfqa
[],
1012 struct nf_queue_entry
*entry
,
1013 enum ip_conntrack_info
*ctinfo
)
1017 ct
= nfnl_ct
->get_ct(entry
->skb
, ctinfo
);
1021 if (nfnl_ct
->parse(nfqa
[NFQA_CT
], ct
) < 0)
1025 nfnl_ct
->attach_expect(nfqa
[NFQA_EXP
], ct
,
1026 NETLINK_CB(entry
->skb
).portid
,
1032 nfqnl_recv_verdict(struct sock
*ctnl
, struct sk_buff
*skb
,
1033 const struct nlmsghdr
*nlh
,
1034 const struct nlattr
* const nfqa
[])
1036 struct nfgenmsg
*nfmsg
= nlmsg_data(nlh
);
1037 u_int16_t queue_num
= ntohs(nfmsg
->res_id
);
1039 struct nfqnl_msg_verdict_hdr
*vhdr
;
1040 struct nfqnl_instance
*queue
;
1041 unsigned int verdict
;
1042 struct nf_queue_entry
*entry
;
1043 enum ip_conntrack_info
uninitialized_var(ctinfo
);
1044 struct nfnl_ct_hook
*nfnl_ct
;
1045 struct nf_conn
*ct
= NULL
;
1047 struct net
*net
= sock_net(ctnl
);
1048 struct nfnl_queue_net
*q
= nfnl_queue_pernet(net
);
1050 queue
= instance_lookup(q
, queue_num
);
1052 queue
= verdict_instance_lookup(q
, queue_num
,
1053 NETLINK_CB(skb
).portid
);
1055 return PTR_ERR(queue
);
1057 vhdr
= verdicthdr_get(nfqa
);
1061 verdict
= ntohl(vhdr
->verdict
);
1063 entry
= find_dequeue_entry(queue
, ntohl(vhdr
->id
));
1067 if (nfqa
[NFQA_CT
]) {
1068 /* rcu lock already held from nfnl->call_rcu. */
1069 nfnl_ct
= rcu_dereference(nfnl_ct_hook
);
1070 if (nfnl_ct
!= NULL
)
1071 ct
= nfqnl_ct_parse(nfnl_ct
, nlh
, nfqa
, entry
, &ctinfo
);
1074 if (nfqa
[NFQA_PAYLOAD
]) {
1075 u16 payload_len
= nla_len(nfqa
[NFQA_PAYLOAD
]);
1076 int diff
= payload_len
- entry
->skb
->len
;
1078 if (nfqnl_mangle(nla_data(nfqa
[NFQA_PAYLOAD
]),
1079 payload_len
, entry
, diff
) < 0)
1083 nfnl_ct
->seq_adjust(entry
->skb
, ct
, ctinfo
, diff
);
1086 if (nfqa
[NFQA_MARK
])
1087 entry
->skb
->mark
= ntohl(nla_get_be32(nfqa
[NFQA_MARK
]));
1089 nf_reinject(entry
, verdict
);
1094 nfqnl_recv_unsupp(struct sock
*ctnl
, struct sk_buff
*skb
,
1095 const struct nlmsghdr
*nlh
,
1096 const struct nlattr
* const nfqa
[])
1101 static const struct nla_policy nfqa_cfg_policy
[NFQA_CFG_MAX
+1] = {
1102 [NFQA_CFG_CMD
] = { .len
= sizeof(struct nfqnl_msg_config_cmd
) },
1103 [NFQA_CFG_PARAMS
] = { .len
= sizeof(struct nfqnl_msg_config_params
) },
1106 static const struct nf_queue_handler nfqh
= {
1107 .outfn
= &nfqnl_enqueue_packet
,
1108 .nf_hook_drop
= &nfqnl_nf_hook_drop
,
1112 nfqnl_recv_config(struct sock
*ctnl
, struct sk_buff
*skb
,
1113 const struct nlmsghdr
*nlh
,
1114 const struct nlattr
* const nfqa
[])
1116 struct nfgenmsg
*nfmsg
= nlmsg_data(nlh
);
1117 u_int16_t queue_num
= ntohs(nfmsg
->res_id
);
1118 struct nfqnl_instance
*queue
;
1119 struct nfqnl_msg_config_cmd
*cmd
= NULL
;
1120 struct net
*net
= sock_net(ctnl
);
1121 struct nfnl_queue_net
*q
= nfnl_queue_pernet(net
);
1124 if (nfqa
[NFQA_CFG_CMD
]) {
1125 cmd
= nla_data(nfqa
[NFQA_CFG_CMD
]);
1127 /* Obsolete commands without queue context */
1128 switch (cmd
->command
) {
1129 case NFQNL_CFG_CMD_PF_BIND
: return 0;
1130 case NFQNL_CFG_CMD_PF_UNBIND
: return 0;
1135 queue
= instance_lookup(q
, queue_num
);
1136 if (queue
&& queue
->peer_portid
!= NETLINK_CB(skb
).portid
) {
1138 goto err_out_unlock
;
1142 switch (cmd
->command
) {
1143 case NFQNL_CFG_CMD_BIND
:
1146 goto err_out_unlock
;
1148 queue
= instance_create(q
, queue_num
,
1149 NETLINK_CB(skb
).portid
);
1150 if (IS_ERR(queue
)) {
1151 ret
= PTR_ERR(queue
);
1152 goto err_out_unlock
;
1155 case NFQNL_CFG_CMD_UNBIND
:
1158 goto err_out_unlock
;
1160 instance_destroy(q
, queue
);
1162 case NFQNL_CFG_CMD_PF_BIND
:
1163 case NFQNL_CFG_CMD_PF_UNBIND
:
1171 if (nfqa
[NFQA_CFG_PARAMS
]) {
1172 struct nfqnl_msg_config_params
*params
;
1176 goto err_out_unlock
;
1178 params
= nla_data(nfqa
[NFQA_CFG_PARAMS
]);
1179 nfqnl_set_mode(queue
, params
->copy_mode
,
1180 ntohl(params
->copy_range
));
1183 if (nfqa
[NFQA_CFG_QUEUE_MAXLEN
]) {
1184 __be32
*queue_maxlen
;
1188 goto err_out_unlock
;
1190 queue_maxlen
= nla_data(nfqa
[NFQA_CFG_QUEUE_MAXLEN
]);
1191 spin_lock_bh(&queue
->lock
);
1192 queue
->queue_maxlen
= ntohl(*queue_maxlen
);
1193 spin_unlock_bh(&queue
->lock
);
1196 if (nfqa
[NFQA_CFG_FLAGS
]) {
1201 goto err_out_unlock
;
1204 if (!nfqa
[NFQA_CFG_MASK
]) {
1205 /* A mask is needed to specify which flags are being
1209 goto err_out_unlock
;
1212 flags
= ntohl(nla_get_be32(nfqa
[NFQA_CFG_FLAGS
]));
1213 mask
= ntohl(nla_get_be32(nfqa
[NFQA_CFG_MASK
]));
1215 if (flags
>= NFQA_CFG_F_MAX
) {
1217 goto err_out_unlock
;
1219 #if !IS_ENABLED(CONFIG_NETWORK_SECMARK)
1220 if (flags
& mask
& NFQA_CFG_F_SECCTX
) {
1222 goto err_out_unlock
;
1225 spin_lock_bh(&queue
->lock
);
1226 queue
->flags
&= ~mask
;
1227 queue
->flags
|= flags
& mask
;
1228 spin_unlock_bh(&queue
->lock
);
1236 static const struct nfnl_callback nfqnl_cb
[NFQNL_MSG_MAX
] = {
1237 [NFQNL_MSG_PACKET
] = { .call_rcu
= nfqnl_recv_unsupp
,
1238 .attr_count
= NFQA_MAX
, },
1239 [NFQNL_MSG_VERDICT
] = { .call_rcu
= nfqnl_recv_verdict
,
1240 .attr_count
= NFQA_MAX
,
1241 .policy
= nfqa_verdict_policy
},
1242 [NFQNL_MSG_CONFIG
] = { .call
= nfqnl_recv_config
,
1243 .attr_count
= NFQA_CFG_MAX
,
1244 .policy
= nfqa_cfg_policy
},
1245 [NFQNL_MSG_VERDICT_BATCH
]={ .call_rcu
= nfqnl_recv_verdict_batch
,
1246 .attr_count
= NFQA_MAX
,
1247 .policy
= nfqa_verdict_batch_policy
},
1250 static const struct nfnetlink_subsystem nfqnl_subsys
= {
1252 .subsys_id
= NFNL_SUBSYS_QUEUE
,
1253 .cb_count
= NFQNL_MSG_MAX
,
1257 #ifdef CONFIG_PROC_FS
1259 struct seq_net_private p
;
1260 unsigned int bucket
;
1263 static struct hlist_node
*get_first(struct seq_file
*seq
)
1265 struct iter_state
*st
= seq
->private;
1267 struct nfnl_queue_net
*q
;
1272 net
= seq_file_net(seq
);
1273 q
= nfnl_queue_pernet(net
);
1274 for (st
->bucket
= 0; st
->bucket
< INSTANCE_BUCKETS
; st
->bucket
++) {
1275 if (!hlist_empty(&q
->instance_table
[st
->bucket
]))
1276 return q
->instance_table
[st
->bucket
].first
;
1281 static struct hlist_node
*get_next(struct seq_file
*seq
, struct hlist_node
*h
)
1283 struct iter_state
*st
= seq
->private;
1284 struct net
*net
= seq_file_net(seq
);
1288 struct nfnl_queue_net
*q
;
1290 if (++st
->bucket
>= INSTANCE_BUCKETS
)
1293 q
= nfnl_queue_pernet(net
);
1294 h
= q
->instance_table
[st
->bucket
].first
;
1299 static struct hlist_node
*get_idx(struct seq_file
*seq
, loff_t pos
)
1301 struct hlist_node
*head
;
1302 head
= get_first(seq
);
1305 while (pos
&& (head
= get_next(seq
, head
)))
1307 return pos
? NULL
: head
;
1310 static void *seq_start(struct seq_file
*s
, loff_t
*pos
)
1311 __acquires(nfnl_queue_pernet(seq_file_net(s
))->instances_lock
)
1313 spin_lock(&nfnl_queue_pernet(seq_file_net(s
))->instances_lock
);
1314 return get_idx(s
, *pos
);
1317 static void *seq_next(struct seq_file
*s
, void *v
, loff_t
*pos
)
1320 return get_next(s
, v
);
1323 static void seq_stop(struct seq_file
*s
, void *v
)
1324 __releases(nfnl_queue_pernet(seq_file_net(s
))->instances_lock
)
1326 spin_unlock(&nfnl_queue_pernet(seq_file_net(s
))->instances_lock
);
1329 static int seq_show(struct seq_file
*s
, void *v
)
1331 const struct nfqnl_instance
*inst
= v
;
1333 seq_printf(s
, "%5u %6u %5u %1u %5u %5u %5u %8u %2d\n",
1335 inst
->peer_portid
, inst
->queue_total
,
1336 inst
->copy_mode
, inst
->copy_range
,
1337 inst
->queue_dropped
, inst
->queue_user_dropped
,
1338 inst
->id_sequence
, 1);
1342 static const struct seq_operations nfqnl_seq_ops
= {
1349 static int nfqnl_open(struct inode
*inode
, struct file
*file
)
1351 return seq_open_net(inode
, file
, &nfqnl_seq_ops
,
1352 sizeof(struct iter_state
));
1355 static const struct file_operations nfqnl_file_ops
= {
1356 .owner
= THIS_MODULE
,
1359 .llseek
= seq_lseek
,
1360 .release
= seq_release_net
,
1363 #endif /* PROC_FS */
1365 static int __net_init
nfnl_queue_net_init(struct net
*net
)
1368 struct nfnl_queue_net
*q
= nfnl_queue_pernet(net
);
1370 for (i
= 0; i
< INSTANCE_BUCKETS
; i
++)
1371 INIT_HLIST_HEAD(&q
->instance_table
[i
]);
1373 spin_lock_init(&q
->instances_lock
);
1375 #ifdef CONFIG_PROC_FS
1376 if (!proc_create("nfnetlink_queue", 0440,
1377 net
->nf
.proc_netfilter
, &nfqnl_file_ops
))
1383 static void __net_exit
nfnl_queue_net_exit(struct net
*net
)
1385 #ifdef CONFIG_PROC_FS
1386 remove_proc_entry("nfnetlink_queue", net
->nf
.proc_netfilter
);
1390 static struct pernet_operations nfnl_queue_net_ops
= {
1391 .init
= nfnl_queue_net_init
,
1392 .exit
= nfnl_queue_net_exit
,
1393 .id
= &nfnl_queue_net_id
,
1394 .size
= sizeof(struct nfnl_queue_net
),
1397 static int __init
nfnetlink_queue_init(void)
1401 status
= register_pernet_subsys(&nfnl_queue_net_ops
);
1403 pr_err("nf_queue: failed to register pernet ops\n");
1407 netlink_register_notifier(&nfqnl_rtnl_notifier
);
1408 status
= nfnetlink_subsys_register(&nfqnl_subsys
);
1410 pr_err("nf_queue: failed to create netlink socket\n");
1411 goto cleanup_netlink_notifier
;
1414 register_netdevice_notifier(&nfqnl_dev_notifier
);
1415 nf_register_queue_handler(&nfqh
);
1418 cleanup_netlink_notifier
:
1419 netlink_unregister_notifier(&nfqnl_rtnl_notifier
);
1424 static void __exit
nfnetlink_queue_fini(void)
1426 nf_unregister_queue_handler();
1427 unregister_netdevice_notifier(&nfqnl_dev_notifier
);
1428 nfnetlink_subsys_unregister(&nfqnl_subsys
);
1429 netlink_unregister_notifier(&nfqnl_rtnl_notifier
);
1430 unregister_pernet_subsys(&nfnl_queue_net_ops
);
1432 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1435 MODULE_DESCRIPTION("netfilter packet queue handler");
1436 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1437 MODULE_LICENSE("GPL");
1438 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE
);
1440 module_init(nfnetlink_queue_init
);
1441 module_exit(nfnetlink_queue_fini
);