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/nfnetlink.h>
29 #include <linux/netfilter/nfnetlink_queue.h>
30 #include <linux/list.h>
32 #include <net/tcp_states.h>
33 #include <net/netfilter/nf_queue.h>
34 #include <net/netns/generic.h>
35 #include <net/netfilter/nfnetlink_queue.h>
37 #include <linux/atomic.h>
39 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
40 #include "../bridge/br_private.h"
43 #define NFQNL_QMAX_DEFAULT 1024
45 /* We're using struct nlattr which has 16bit nla_len. Note that nla_len
46 * includes the header length. Thus, the maximum packet length that we
47 * support is 65531 bytes. We send truncated packets if the specified length
48 * is larger than that. Userspace can check for presence of NFQA_CAP_LEN
49 * attribute to detect truncation.
51 #define NFQNL_MAX_COPY_RANGE (0xffff - NLA_HDRLEN)
53 struct nfqnl_instance
{
54 struct hlist_node hlist
; /* global list of queues */
58 unsigned int queue_maxlen
;
59 unsigned int copy_range
;
60 unsigned int queue_dropped
;
61 unsigned int queue_user_dropped
;
64 u_int16_t queue_num
; /* number of this queue */
66 u_int32_t flags
; /* Set using NFQA_CFG_FLAGS */
68 * Following fields are dirtied for each queued packet,
69 * keep them in same cache line if possible.
72 unsigned int queue_total
;
73 unsigned int id_sequence
; /* 'sequence' of pkt ids */
74 struct list_head queue_list
; /* packets in queue */
77 typedef int (*nfqnl_cmpfn
)(struct nf_queue_entry
*, unsigned long);
79 static int nfnl_queue_net_id __read_mostly
;
81 #define INSTANCE_BUCKETS 16
82 struct nfnl_queue_net
{
83 spinlock_t instances_lock
;
84 struct hlist_head instance_table
[INSTANCE_BUCKETS
];
87 static struct nfnl_queue_net
*nfnl_queue_pernet(struct net
*net
)
89 return net_generic(net
, nfnl_queue_net_id
);
92 static inline u_int8_t
instance_hashfn(u_int16_t queue_num
)
94 return ((queue_num
>> 8) ^ queue_num
) % INSTANCE_BUCKETS
;
97 static struct nfqnl_instance
*
98 instance_lookup(struct nfnl_queue_net
*q
, u_int16_t queue_num
)
100 struct hlist_head
*head
;
101 struct nfqnl_instance
*inst
;
103 head
= &q
->instance_table
[instance_hashfn(queue_num
)];
104 hlist_for_each_entry_rcu(inst
, head
, hlist
) {
105 if (inst
->queue_num
== queue_num
)
111 static struct nfqnl_instance
*
112 instance_create(struct nfnl_queue_net
*q
, u_int16_t queue_num
,
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
->sk_state
== TCP_TIME_WAIT
)
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 struct sk_buff
*
282 nfqnl_build_packet_message(struct net
*net
, struct nfqnl_instance
*queue
,
283 struct nf_queue_entry
*entry
,
284 __be32
**packet_id_ptr
)
287 size_t data_len
= 0, cap_len
= 0;
288 unsigned int hlen
= 0;
291 struct nfqnl_msg_packet_hdr
*pmsg
;
292 struct nlmsghdr
*nlh
;
293 struct nfgenmsg
*nfmsg
;
294 struct sk_buff
*entskb
= entry
->skb
;
295 struct net_device
*indev
;
296 struct net_device
*outdev
;
297 struct nf_conn
*ct
= NULL
;
298 enum ip_conntrack_info
uninitialized_var(ctinfo
);
301 size
= nlmsg_total_size(sizeof(struct nfgenmsg
))
302 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr
))
303 + nla_total_size(sizeof(u_int32_t
)) /* ifindex */
304 + nla_total_size(sizeof(u_int32_t
)) /* ifindex */
305 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
306 + nla_total_size(sizeof(u_int32_t
)) /* ifindex */
307 + nla_total_size(sizeof(u_int32_t
)) /* ifindex */
309 + nla_total_size(sizeof(u_int32_t
)) /* mark */
310 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw
))
311 + nla_total_size(sizeof(u_int32_t
)) /* skbinfo */
312 + nla_total_size(sizeof(u_int32_t
)); /* cap_len */
314 if (entskb
->tstamp
.tv64
)
315 size
+= nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp
));
317 if (entry
->hook
<= NF_INET_FORWARD
||
318 (entry
->hook
== NF_INET_POST_ROUTING
&& entskb
->sk
== NULL
))
319 csum_verify
= !skb_csum_unnecessary(entskb
);
323 outdev
= entry
->outdev
;
325 switch ((enum nfqnl_config_mode
)ACCESS_ONCE(queue
->copy_mode
)) {
326 case NFQNL_COPY_META
:
327 case NFQNL_COPY_NONE
:
330 case NFQNL_COPY_PACKET
:
331 if (!(queue
->flags
& NFQA_CFG_F_GSO
) &&
332 entskb
->ip_summed
== CHECKSUM_PARTIAL
&&
333 skb_checksum_help(entskb
))
336 data_len
= ACCESS_ONCE(queue
->copy_range
);
337 if (data_len
> entskb
->len
)
338 data_len
= entskb
->len
;
340 hlen
= skb_zerocopy_headlen(entskb
);
341 hlen
= min_t(unsigned int, hlen
, data_len
);
342 size
+= sizeof(struct nlattr
) + hlen
;
343 cap_len
= entskb
->len
;
347 if (queue
->flags
& NFQA_CFG_F_CONNTRACK
)
348 ct
= nfqnl_ct_get(entskb
, &size
, &ctinfo
);
350 if (queue
->flags
& NFQA_CFG_F_UID_GID
) {
351 size
+= (nla_total_size(sizeof(u_int32_t
)) /* uid */
352 + nla_total_size(sizeof(u_int32_t
))); /* gid */
355 skb
= nfnetlink_alloc_skb(net
, size
, queue
->peer_portid
,
358 skb_tx_error(entskb
);
362 nlh
= nlmsg_put(skb
, 0, 0,
363 NFNL_SUBSYS_QUEUE
<< 8 | NFQNL_MSG_PACKET
,
364 sizeof(struct nfgenmsg
), 0);
366 skb_tx_error(entskb
);
370 nfmsg
= nlmsg_data(nlh
);
371 nfmsg
->nfgen_family
= entry
->pf
;
372 nfmsg
->version
= NFNETLINK_V0
;
373 nfmsg
->res_id
= htons(queue
->queue_num
);
375 nla
= __nla_reserve(skb
, NFQA_PACKET_HDR
, sizeof(*pmsg
));
376 pmsg
= nla_data(nla
);
377 pmsg
->hw_protocol
= entskb
->protocol
;
378 pmsg
->hook
= entry
->hook
;
379 *packet_id_ptr
= &pmsg
->packet_id
;
381 indev
= entry
->indev
;
383 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
384 if (nla_put_be32(skb
, NFQA_IFINDEX_INDEV
, htonl(indev
->ifindex
)))
385 goto nla_put_failure
;
387 if (entry
->pf
== PF_BRIDGE
) {
388 /* Case 1: indev is physical input device, we need to
389 * look for bridge group (when called from
390 * netfilter_bridge) */
391 if (nla_put_be32(skb
, NFQA_IFINDEX_PHYSINDEV
,
392 htonl(indev
->ifindex
)) ||
393 /* this is the bridge group "brX" */
394 /* rcu_read_lock()ed by __nf_queue */
395 nla_put_be32(skb
, NFQA_IFINDEX_INDEV
,
396 htonl(br_port_get_rcu(indev
)->br
->dev
->ifindex
)))
397 goto nla_put_failure
;
399 /* Case 2: indev is bridge group, we need to look for
400 * physical device (when called from ipv4) */
401 if (nla_put_be32(skb
, NFQA_IFINDEX_INDEV
,
402 htonl(indev
->ifindex
)))
403 goto nla_put_failure
;
404 if (entskb
->nf_bridge
&& entskb
->nf_bridge
->physindev
&&
405 nla_put_be32(skb
, NFQA_IFINDEX_PHYSINDEV
,
406 htonl(entskb
->nf_bridge
->physindev
->ifindex
)))
407 goto nla_put_failure
;
413 #if !IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
414 if (nla_put_be32(skb
, NFQA_IFINDEX_OUTDEV
, htonl(outdev
->ifindex
)))
415 goto nla_put_failure
;
417 if (entry
->pf
== PF_BRIDGE
) {
418 /* Case 1: outdev is physical output device, we need to
419 * look for bridge group (when called from
420 * netfilter_bridge) */
421 if (nla_put_be32(skb
, NFQA_IFINDEX_PHYSOUTDEV
,
422 htonl(outdev
->ifindex
)) ||
423 /* this is the bridge group "brX" */
424 /* rcu_read_lock()ed by __nf_queue */
425 nla_put_be32(skb
, NFQA_IFINDEX_OUTDEV
,
426 htonl(br_port_get_rcu(outdev
)->br
->dev
->ifindex
)))
427 goto nla_put_failure
;
429 /* Case 2: outdev is bridge group, we need to look for
430 * physical output device (when called from ipv4) */
431 if (nla_put_be32(skb
, NFQA_IFINDEX_OUTDEV
,
432 htonl(outdev
->ifindex
)))
433 goto nla_put_failure
;
434 if (entskb
->nf_bridge
&& entskb
->nf_bridge
->physoutdev
&&
435 nla_put_be32(skb
, NFQA_IFINDEX_PHYSOUTDEV
,
436 htonl(entskb
->nf_bridge
->physoutdev
->ifindex
)))
437 goto nla_put_failure
;
443 nla_put_be32(skb
, NFQA_MARK
, htonl(entskb
->mark
)))
444 goto nla_put_failure
;
446 if (indev
&& entskb
->dev
&&
447 entskb
->mac_header
!= entskb
->network_header
) {
448 struct nfqnl_msg_packet_hw phw
;
451 memset(&phw
, 0, sizeof(phw
));
452 len
= dev_parse_header(entskb
, phw
.hw_addr
);
454 phw
.hw_addrlen
= htons(len
);
455 if (nla_put(skb
, NFQA_HWADDR
, sizeof(phw
), &phw
))
456 goto nla_put_failure
;
460 if (entskb
->tstamp
.tv64
) {
461 struct nfqnl_msg_packet_timestamp ts
;
462 struct timeval tv
= ktime_to_timeval(entskb
->tstamp
);
463 ts
.sec
= cpu_to_be64(tv
.tv_sec
);
464 ts
.usec
= cpu_to_be64(tv
.tv_usec
);
466 if (nla_put(skb
, NFQA_TIMESTAMP
, sizeof(ts
), &ts
))
467 goto nla_put_failure
;
470 if ((queue
->flags
& NFQA_CFG_F_UID_GID
) && entskb
->sk
&&
471 nfqnl_put_sk_uidgid(skb
, entskb
->sk
) < 0)
472 goto nla_put_failure
;
474 if (ct
&& nfqnl_ct_put(skb
, ct
, ctinfo
) < 0)
475 goto nla_put_failure
;
477 if (cap_len
> data_len
&&
478 nla_put_be32(skb
, NFQA_CAP_LEN
, htonl(cap_len
)))
479 goto nla_put_failure
;
481 if (nfqnl_put_packet_info(skb
, entskb
, csum_verify
))
482 goto nla_put_failure
;
487 if (skb_tailroom(skb
) < sizeof(*nla
) + hlen
)
488 goto nla_put_failure
;
490 nla
= (struct nlattr
*)skb_put(skb
, sizeof(*nla
));
491 nla
->nla_type
= NFQA_PAYLOAD
;
492 nla
->nla_len
= nla_attr_size(data_len
);
494 if (skb_zerocopy(skb
, entskb
, data_len
, hlen
))
495 goto nla_put_failure
;
498 nlh
->nlmsg_len
= skb
->len
;
502 skb_tx_error(entskb
);
504 net_err_ratelimited("nf_queue: error creating packet message\n");
509 __nfqnl_enqueue_packet(struct net
*net
, struct nfqnl_instance
*queue
,
510 struct nf_queue_entry
*entry
)
512 struct sk_buff
*nskb
;
514 __be32
*packet_id_ptr
;
517 nskb
= nfqnl_build_packet_message(net
, queue
, entry
, &packet_id_ptr
);
522 spin_lock_bh(&queue
->lock
);
524 if (queue
->queue_total
>= queue
->queue_maxlen
) {
525 if (queue
->flags
& NFQA_CFG_F_FAIL_OPEN
) {
529 queue
->queue_dropped
++;
530 net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
533 goto err_out_free_nskb
;
535 entry
->id
= ++queue
->id_sequence
;
536 *packet_id_ptr
= htonl(entry
->id
);
538 /* nfnetlink_unicast will either free the nskb or add it to a socket */
539 err
= nfnetlink_unicast(nskb
, net
, queue
->peer_portid
, MSG_DONTWAIT
);
541 queue
->queue_user_dropped
++;
545 __enqueue_entry(queue
, entry
);
547 spin_unlock_bh(&queue
->lock
);
553 spin_unlock_bh(&queue
->lock
);
555 nf_reinject(entry
, NF_ACCEPT
);
560 static struct nf_queue_entry
*
561 nf_queue_entry_dup(struct nf_queue_entry
*e
)
563 struct nf_queue_entry
*entry
= kmemdup(e
, e
->size
, GFP_ATOMIC
);
565 if (nf_queue_entry_get_refs(entry
))
572 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
573 /* When called from bridge netfilter, skb->data must point to MAC header
574 * before calling skb_gso_segment(). Else, original MAC header is lost
575 * and segmented skbs will be sent to wrong destination.
577 static void nf_bridge_adjust_skb_data(struct sk_buff
*skb
)
580 __skb_push(skb
, skb
->network_header
- skb
->mac_header
);
583 static void nf_bridge_adjust_segmented_data(struct sk_buff
*skb
)
586 __skb_pull(skb
, skb
->network_header
- skb
->mac_header
);
589 #define nf_bridge_adjust_skb_data(s) do {} while (0)
590 #define nf_bridge_adjust_segmented_data(s) do {} while (0)
593 static void free_entry(struct nf_queue_entry
*entry
)
595 nf_queue_entry_release_refs(entry
);
600 __nfqnl_enqueue_packet_gso(struct net
*net
, struct nfqnl_instance
*queue
,
601 struct sk_buff
*skb
, struct nf_queue_entry
*entry
)
604 struct nf_queue_entry
*entry_seg
;
606 nf_bridge_adjust_segmented_data(skb
);
608 if (skb
->next
== NULL
) { /* last packet, no need to copy entry */
609 struct sk_buff
*gso_skb
= entry
->skb
;
611 ret
= __nfqnl_enqueue_packet(net
, queue
, entry
);
613 entry
->skb
= gso_skb
;
619 entry_seg
= nf_queue_entry_dup(entry
);
621 entry_seg
->skb
= skb
;
622 ret
= __nfqnl_enqueue_packet(net
, queue
, entry_seg
);
624 free_entry(entry_seg
);
630 nfqnl_enqueue_packet(struct nf_queue_entry
*entry
, unsigned int queuenum
)
633 struct nfqnl_instance
*queue
;
634 struct sk_buff
*skb
, *segs
;
636 struct net
*net
= dev_net(entry
->indev
?
637 entry
->indev
: entry
->outdev
);
638 struct nfnl_queue_net
*q
= nfnl_queue_pernet(net
);
640 /* rcu_read_lock()ed by nf_hook_slow() */
641 queue
= instance_lookup(q
, queuenum
);
645 if (queue
->copy_mode
== NFQNL_COPY_NONE
)
652 skb
->protocol
= htons(ETH_P_IP
);
655 skb
->protocol
= htons(ETH_P_IPV6
);
659 if ((queue
->flags
& NFQA_CFG_F_GSO
) || !skb_is_gso(skb
))
660 return __nfqnl_enqueue_packet(net
, queue
, entry
);
662 nf_bridge_adjust_skb_data(skb
);
663 segs
= skb_gso_segment(skb
, 0);
664 /* Does not use PTR_ERR to limit the number of error codes that can be
665 * returned by nf_queue. For instance, callers rely on -ECANCELED to
666 * mean 'ignore this hook'.
668 if (IS_ERR_OR_NULL(segs
))
673 struct sk_buff
*nskb
= segs
->next
;
675 err
= __nfqnl_enqueue_packet_gso(net
, queue
,
685 if (err
) /* some segments are already queued */
691 nf_bridge_adjust_segmented_data(skb
);
696 nfqnl_mangle(void *data
, int data_len
, struct nf_queue_entry
*e
, int diff
)
698 struct sk_buff
*nskb
;
701 if (pskb_trim(e
->skb
, data_len
))
703 } else if (diff
> 0) {
704 if (data_len
> 0xFFFF)
706 if (diff
> skb_tailroom(e
->skb
)) {
707 nskb
= skb_copy_expand(e
->skb
, skb_headroom(e
->skb
),
710 printk(KERN_WARNING
"nf_queue: OOM "
711 "in mangle, dropping packet\n");
717 skb_put(e
->skb
, diff
);
719 if (!skb_make_writable(e
->skb
, data_len
))
721 skb_copy_to_linear_data(e
->skb
, data
, data_len
);
722 e
->skb
->ip_summed
= CHECKSUM_NONE
;
727 nfqnl_set_mode(struct nfqnl_instance
*queue
,
728 unsigned char mode
, unsigned int range
)
732 spin_lock_bh(&queue
->lock
);
734 case NFQNL_COPY_NONE
:
735 case NFQNL_COPY_META
:
736 queue
->copy_mode
= mode
;
737 queue
->copy_range
= 0;
740 case NFQNL_COPY_PACKET
:
741 queue
->copy_mode
= mode
;
742 if (range
== 0 || range
> NFQNL_MAX_COPY_RANGE
)
743 queue
->copy_range
= NFQNL_MAX_COPY_RANGE
;
745 queue
->copy_range
= range
;
752 spin_unlock_bh(&queue
->lock
);
758 dev_cmp(struct nf_queue_entry
*entry
, unsigned long ifindex
)
761 if (entry
->indev
->ifindex
== ifindex
)
764 if (entry
->outdev
->ifindex
== ifindex
)
766 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
767 if (entry
->skb
->nf_bridge
) {
768 if (entry
->skb
->nf_bridge
->physindev
&&
769 entry
->skb
->nf_bridge
->physindev
->ifindex
== ifindex
)
771 if (entry
->skb
->nf_bridge
->physoutdev
&&
772 entry
->skb
->nf_bridge
->physoutdev
->ifindex
== ifindex
)
779 /* drop all packets with either indev or outdev == ifindex from all queue
782 nfqnl_dev_drop(struct net
*net
, int ifindex
)
785 struct nfnl_queue_net
*q
= nfnl_queue_pernet(net
);
789 for (i
= 0; i
< INSTANCE_BUCKETS
; i
++) {
790 struct nfqnl_instance
*inst
;
791 struct hlist_head
*head
= &q
->instance_table
[i
];
793 hlist_for_each_entry_rcu(inst
, head
, hlist
)
794 nfqnl_flush(inst
, dev_cmp
, ifindex
);
800 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
803 nfqnl_rcv_dev_event(struct notifier_block
*this,
804 unsigned long event
, void *ptr
)
806 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
808 /* Drop any packets associated with the downed device */
809 if (event
== NETDEV_DOWN
)
810 nfqnl_dev_drop(dev_net(dev
), dev
->ifindex
);
814 static struct notifier_block nfqnl_dev_notifier
= {
815 .notifier_call
= nfqnl_rcv_dev_event
,
819 nfqnl_rcv_nl_event(struct notifier_block
*this,
820 unsigned long event
, void *ptr
)
822 struct netlink_notify
*n
= ptr
;
823 struct nfnl_queue_net
*q
= nfnl_queue_pernet(n
->net
);
825 if (event
== NETLINK_URELEASE
&& n
->protocol
== NETLINK_NETFILTER
) {
828 /* destroy all instances for this portid */
829 spin_lock(&q
->instances_lock
);
830 for (i
= 0; i
< INSTANCE_BUCKETS
; i
++) {
831 struct hlist_node
*t2
;
832 struct nfqnl_instance
*inst
;
833 struct hlist_head
*head
= &q
->instance_table
[i
];
835 hlist_for_each_entry_safe(inst
, t2
, head
, hlist
) {
836 if (n
->portid
== inst
->peer_portid
)
837 __instance_destroy(inst
);
840 spin_unlock(&q
->instances_lock
);
845 static struct notifier_block nfqnl_rtnl_notifier
= {
846 .notifier_call
= nfqnl_rcv_nl_event
,
849 static const struct nla_policy nfqa_verdict_policy
[NFQA_MAX
+1] = {
850 [NFQA_VERDICT_HDR
] = { .len
= sizeof(struct nfqnl_msg_verdict_hdr
) },
851 [NFQA_MARK
] = { .type
= NLA_U32
},
852 [NFQA_PAYLOAD
] = { .type
= NLA_UNSPEC
},
853 [NFQA_CT
] = { .type
= NLA_UNSPEC
},
854 [NFQA_EXP
] = { .type
= NLA_UNSPEC
},
857 static const struct nla_policy nfqa_verdict_batch_policy
[NFQA_MAX
+1] = {
858 [NFQA_VERDICT_HDR
] = { .len
= sizeof(struct nfqnl_msg_verdict_hdr
) },
859 [NFQA_MARK
] = { .type
= NLA_U32
},
862 static struct nfqnl_instance
*
863 verdict_instance_lookup(struct nfnl_queue_net
*q
, u16 queue_num
, int nlportid
)
865 struct nfqnl_instance
*queue
;
867 queue
= instance_lookup(q
, queue_num
);
869 return ERR_PTR(-ENODEV
);
871 if (queue
->peer_portid
!= nlportid
)
872 return ERR_PTR(-EPERM
);
877 static struct nfqnl_msg_verdict_hdr
*
878 verdicthdr_get(const struct nlattr
* const nfqa
[])
880 struct nfqnl_msg_verdict_hdr
*vhdr
;
881 unsigned int verdict
;
883 if (!nfqa
[NFQA_VERDICT_HDR
])
886 vhdr
= nla_data(nfqa
[NFQA_VERDICT_HDR
]);
887 verdict
= ntohl(vhdr
->verdict
) & NF_VERDICT_MASK
;
888 if (verdict
> NF_MAX_VERDICT
|| verdict
== NF_STOLEN
)
893 static int nfq_id_after(unsigned int id
, unsigned int max
)
895 return (int)(id
- max
) > 0;
899 nfqnl_recv_verdict_batch(struct sock
*ctnl
, struct sk_buff
*skb
,
900 const struct nlmsghdr
*nlh
,
901 const struct nlattr
* const nfqa
[])
903 struct nfgenmsg
*nfmsg
= nlmsg_data(nlh
);
904 struct nf_queue_entry
*entry
, *tmp
;
905 unsigned int verdict
, maxid
;
906 struct nfqnl_msg_verdict_hdr
*vhdr
;
907 struct nfqnl_instance
*queue
;
908 LIST_HEAD(batch_list
);
909 u16 queue_num
= ntohs(nfmsg
->res_id
);
911 struct net
*net
= sock_net(ctnl
);
912 struct nfnl_queue_net
*q
= nfnl_queue_pernet(net
);
914 queue
= verdict_instance_lookup(q
, queue_num
,
915 NETLINK_CB(skb
).portid
);
917 return PTR_ERR(queue
);
919 vhdr
= verdicthdr_get(nfqa
);
923 verdict
= ntohl(vhdr
->verdict
);
924 maxid
= ntohl(vhdr
->id
);
926 spin_lock_bh(&queue
->lock
);
928 list_for_each_entry_safe(entry
, tmp
, &queue
->queue_list
, list
) {
929 if (nfq_id_after(entry
->id
, maxid
))
931 __dequeue_entry(queue
, entry
);
932 list_add_tail(&entry
->list
, &batch_list
);
935 spin_unlock_bh(&queue
->lock
);
937 if (list_empty(&batch_list
))
940 list_for_each_entry_safe(entry
, tmp
, &batch_list
, list
) {
942 entry
->skb
->mark
= ntohl(nla_get_be32(nfqa
[NFQA_MARK
]));
943 nf_reinject(entry
, verdict
);
949 nfqnl_recv_verdict(struct sock
*ctnl
, struct sk_buff
*skb
,
950 const struct nlmsghdr
*nlh
,
951 const struct nlattr
* const nfqa
[])
953 struct nfgenmsg
*nfmsg
= nlmsg_data(nlh
);
954 u_int16_t queue_num
= ntohs(nfmsg
->res_id
);
956 struct nfqnl_msg_verdict_hdr
*vhdr
;
957 struct nfqnl_instance
*queue
;
958 unsigned int verdict
;
959 struct nf_queue_entry
*entry
;
960 enum ip_conntrack_info
uninitialized_var(ctinfo
);
961 struct nf_conn
*ct
= NULL
;
963 struct net
*net
= sock_net(ctnl
);
964 struct nfnl_queue_net
*q
= nfnl_queue_pernet(net
);
966 queue
= instance_lookup(q
, queue_num
);
968 queue
= verdict_instance_lookup(q
, queue_num
,
969 NETLINK_CB(skb
).portid
);
971 return PTR_ERR(queue
);
973 vhdr
= verdicthdr_get(nfqa
);
977 verdict
= ntohl(vhdr
->verdict
);
979 entry
= find_dequeue_entry(queue
, ntohl(vhdr
->id
));
984 ct
= nfqnl_ct_parse(entry
->skb
, nfqa
[NFQA_CT
], &ctinfo
);
985 if (ct
&& nfqa
[NFQA_EXP
]) {
986 nfqnl_attach_expect(ct
, nfqa
[NFQA_EXP
],
987 NETLINK_CB(skb
).portid
,
992 if (nfqa
[NFQA_PAYLOAD
]) {
993 u16 payload_len
= nla_len(nfqa
[NFQA_PAYLOAD
]);
994 int diff
= payload_len
- entry
->skb
->len
;
996 if (nfqnl_mangle(nla_data(nfqa
[NFQA_PAYLOAD
]),
997 payload_len
, entry
, diff
) < 0)
1001 nfqnl_ct_seq_adjust(entry
->skb
, ct
, ctinfo
, diff
);
1004 if (nfqa
[NFQA_MARK
])
1005 entry
->skb
->mark
= ntohl(nla_get_be32(nfqa
[NFQA_MARK
]));
1007 nf_reinject(entry
, verdict
);
1012 nfqnl_recv_unsupp(struct sock
*ctnl
, struct sk_buff
*skb
,
1013 const struct nlmsghdr
*nlh
,
1014 const struct nlattr
* const nfqa
[])
1019 static const struct nla_policy nfqa_cfg_policy
[NFQA_CFG_MAX
+1] = {
1020 [NFQA_CFG_CMD
] = { .len
= sizeof(struct nfqnl_msg_config_cmd
) },
1021 [NFQA_CFG_PARAMS
] = { .len
= sizeof(struct nfqnl_msg_config_params
) },
1024 static const struct nf_queue_handler nfqh
= {
1025 .outfn
= &nfqnl_enqueue_packet
,
1029 nfqnl_recv_config(struct sock
*ctnl
, struct sk_buff
*skb
,
1030 const struct nlmsghdr
*nlh
,
1031 const struct nlattr
* const nfqa
[])
1033 struct nfgenmsg
*nfmsg
= nlmsg_data(nlh
);
1034 u_int16_t queue_num
= ntohs(nfmsg
->res_id
);
1035 struct nfqnl_instance
*queue
;
1036 struct nfqnl_msg_config_cmd
*cmd
= NULL
;
1037 struct net
*net
= sock_net(ctnl
);
1038 struct nfnl_queue_net
*q
= nfnl_queue_pernet(net
);
1041 if (nfqa
[NFQA_CFG_CMD
]) {
1042 cmd
= nla_data(nfqa
[NFQA_CFG_CMD
]);
1044 /* Obsolete commands without queue context */
1045 switch (cmd
->command
) {
1046 case NFQNL_CFG_CMD_PF_BIND
: return 0;
1047 case NFQNL_CFG_CMD_PF_UNBIND
: return 0;
1052 queue
= instance_lookup(q
, queue_num
);
1053 if (queue
&& queue
->peer_portid
!= NETLINK_CB(skb
).portid
) {
1055 goto err_out_unlock
;
1059 switch (cmd
->command
) {
1060 case NFQNL_CFG_CMD_BIND
:
1063 goto err_out_unlock
;
1065 queue
= instance_create(q
, queue_num
,
1066 NETLINK_CB(skb
).portid
);
1067 if (IS_ERR(queue
)) {
1068 ret
= PTR_ERR(queue
);
1069 goto err_out_unlock
;
1072 case NFQNL_CFG_CMD_UNBIND
:
1075 goto err_out_unlock
;
1077 instance_destroy(q
, queue
);
1079 case NFQNL_CFG_CMD_PF_BIND
:
1080 case NFQNL_CFG_CMD_PF_UNBIND
:
1088 if (nfqa
[NFQA_CFG_PARAMS
]) {
1089 struct nfqnl_msg_config_params
*params
;
1093 goto err_out_unlock
;
1095 params
= nla_data(nfqa
[NFQA_CFG_PARAMS
]);
1096 nfqnl_set_mode(queue
, params
->copy_mode
,
1097 ntohl(params
->copy_range
));
1100 if (nfqa
[NFQA_CFG_QUEUE_MAXLEN
]) {
1101 __be32
*queue_maxlen
;
1105 goto err_out_unlock
;
1107 queue_maxlen
= nla_data(nfqa
[NFQA_CFG_QUEUE_MAXLEN
]);
1108 spin_lock_bh(&queue
->lock
);
1109 queue
->queue_maxlen
= ntohl(*queue_maxlen
);
1110 spin_unlock_bh(&queue
->lock
);
1113 if (nfqa
[NFQA_CFG_FLAGS
]) {
1118 goto err_out_unlock
;
1121 if (!nfqa
[NFQA_CFG_MASK
]) {
1122 /* A mask is needed to specify which flags are being
1126 goto err_out_unlock
;
1129 flags
= ntohl(nla_get_be32(nfqa
[NFQA_CFG_FLAGS
]));
1130 mask
= ntohl(nla_get_be32(nfqa
[NFQA_CFG_MASK
]));
1132 if (flags
>= NFQA_CFG_F_MAX
) {
1134 goto err_out_unlock
;
1137 spin_lock_bh(&queue
->lock
);
1138 queue
->flags
&= ~mask
;
1139 queue
->flags
|= flags
& mask
;
1140 spin_unlock_bh(&queue
->lock
);
1148 static const struct nfnl_callback nfqnl_cb
[NFQNL_MSG_MAX
] = {
1149 [NFQNL_MSG_PACKET
] = { .call_rcu
= nfqnl_recv_unsupp
,
1150 .attr_count
= NFQA_MAX
, },
1151 [NFQNL_MSG_VERDICT
] = { .call_rcu
= nfqnl_recv_verdict
,
1152 .attr_count
= NFQA_MAX
,
1153 .policy
= nfqa_verdict_policy
},
1154 [NFQNL_MSG_CONFIG
] = { .call
= nfqnl_recv_config
,
1155 .attr_count
= NFQA_CFG_MAX
,
1156 .policy
= nfqa_cfg_policy
},
1157 [NFQNL_MSG_VERDICT_BATCH
]={ .call_rcu
= nfqnl_recv_verdict_batch
,
1158 .attr_count
= NFQA_MAX
,
1159 .policy
= nfqa_verdict_batch_policy
},
1162 static const struct nfnetlink_subsystem nfqnl_subsys
= {
1164 .subsys_id
= NFNL_SUBSYS_QUEUE
,
1165 .cb_count
= NFQNL_MSG_MAX
,
1169 #ifdef CONFIG_PROC_FS
1171 struct seq_net_private p
;
1172 unsigned int bucket
;
1175 static struct hlist_node
*get_first(struct seq_file
*seq
)
1177 struct iter_state
*st
= seq
->private;
1179 struct nfnl_queue_net
*q
;
1184 net
= seq_file_net(seq
);
1185 q
= nfnl_queue_pernet(net
);
1186 for (st
->bucket
= 0; st
->bucket
< INSTANCE_BUCKETS
; st
->bucket
++) {
1187 if (!hlist_empty(&q
->instance_table
[st
->bucket
]))
1188 return q
->instance_table
[st
->bucket
].first
;
1193 static struct hlist_node
*get_next(struct seq_file
*seq
, struct hlist_node
*h
)
1195 struct iter_state
*st
= seq
->private;
1196 struct net
*net
= seq_file_net(seq
);
1200 struct nfnl_queue_net
*q
;
1202 if (++st
->bucket
>= INSTANCE_BUCKETS
)
1205 q
= nfnl_queue_pernet(net
);
1206 h
= q
->instance_table
[st
->bucket
].first
;
1211 static struct hlist_node
*get_idx(struct seq_file
*seq
, loff_t pos
)
1213 struct hlist_node
*head
;
1214 head
= get_first(seq
);
1217 while (pos
&& (head
= get_next(seq
, head
)))
1219 return pos
? NULL
: head
;
1222 static void *seq_start(struct seq_file
*s
, loff_t
*pos
)
1223 __acquires(nfnl_queue_pernet(seq_file_net(s
))->instances_lock
)
1225 spin_lock(&nfnl_queue_pernet(seq_file_net(s
))->instances_lock
);
1226 return get_idx(s
, *pos
);
1229 static void *seq_next(struct seq_file
*s
, void *v
, loff_t
*pos
)
1232 return get_next(s
, v
);
1235 static void seq_stop(struct seq_file
*s
, void *v
)
1236 __releases(nfnl_queue_pernet(seq_file_net(s
))->instances_lock
)
1238 spin_unlock(&nfnl_queue_pernet(seq_file_net(s
))->instances_lock
);
1241 static int seq_show(struct seq_file
*s
, void *v
)
1243 const struct nfqnl_instance
*inst
= v
;
1245 return seq_printf(s
, "%5d %6d %5d %1d %5d %5d %5d %8d %2d\n",
1247 inst
->peer_portid
, inst
->queue_total
,
1248 inst
->copy_mode
, inst
->copy_range
,
1249 inst
->queue_dropped
, inst
->queue_user_dropped
,
1250 inst
->id_sequence
, 1);
1253 static const struct seq_operations nfqnl_seq_ops
= {
1260 static int nfqnl_open(struct inode
*inode
, struct file
*file
)
1262 return seq_open_net(inode
, file
, &nfqnl_seq_ops
,
1263 sizeof(struct iter_state
));
1266 static const struct file_operations nfqnl_file_ops
= {
1267 .owner
= THIS_MODULE
,
1270 .llseek
= seq_lseek
,
1271 .release
= seq_release_net
,
1274 #endif /* PROC_FS */
1276 static int __net_init
nfnl_queue_net_init(struct net
*net
)
1279 struct nfnl_queue_net
*q
= nfnl_queue_pernet(net
);
1281 for (i
= 0; i
< INSTANCE_BUCKETS
; i
++)
1282 INIT_HLIST_HEAD(&q
->instance_table
[i
]);
1284 spin_lock_init(&q
->instances_lock
);
1286 #ifdef CONFIG_PROC_FS
1287 if (!proc_create("nfnetlink_queue", 0440,
1288 net
->nf
.proc_netfilter
, &nfqnl_file_ops
))
1294 static void __net_exit
nfnl_queue_net_exit(struct net
*net
)
1296 #ifdef CONFIG_PROC_FS
1297 remove_proc_entry("nfnetlink_queue", net
->nf
.proc_netfilter
);
1301 static struct pernet_operations nfnl_queue_net_ops
= {
1302 .init
= nfnl_queue_net_init
,
1303 .exit
= nfnl_queue_net_exit
,
1304 .id
= &nfnl_queue_net_id
,
1305 .size
= sizeof(struct nfnl_queue_net
),
1308 static int __init
nfnetlink_queue_init(void)
1310 int status
= -ENOMEM
;
1312 netlink_register_notifier(&nfqnl_rtnl_notifier
);
1313 status
= nfnetlink_subsys_register(&nfqnl_subsys
);
1315 pr_err("nf_queue: failed to create netlink socket\n");
1316 goto cleanup_netlink_notifier
;
1319 status
= register_pernet_subsys(&nfnl_queue_net_ops
);
1321 pr_err("nf_queue: failed to register pernet ops\n");
1322 goto cleanup_subsys
;
1324 register_netdevice_notifier(&nfqnl_dev_notifier
);
1325 nf_register_queue_handler(&nfqh
);
1329 nfnetlink_subsys_unregister(&nfqnl_subsys
);
1330 cleanup_netlink_notifier
:
1331 netlink_unregister_notifier(&nfqnl_rtnl_notifier
);
1335 static void __exit
nfnetlink_queue_fini(void)
1337 nf_unregister_queue_handler();
1338 unregister_netdevice_notifier(&nfqnl_dev_notifier
);
1339 unregister_pernet_subsys(&nfnl_queue_net_ops
);
1340 nfnetlink_subsys_unregister(&nfqnl_subsys
);
1341 netlink_unregister_notifier(&nfqnl_rtnl_notifier
);
1343 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1346 MODULE_DESCRIPTION("netfilter packet queue handler");
1347 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1348 MODULE_LICENSE("GPL");
1349 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE
);
1351 module_init(nfnetlink_queue_init
);
1352 module_exit(nfnetlink_queue_fini
);