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/netfilter/nf_queue.h>
33 #include <net/netns/generic.h>
34 #include <net/netfilter/nfnetlink_queue.h>
36 #include <linux/atomic.h>
38 #ifdef CONFIG_BRIDGE_NETFILTER
39 #include "../bridge/br_private.h"
42 #define NFQNL_QMAX_DEFAULT 1024
44 /* We're using struct nlattr which has 16bit nla_len. Note that nla_len
45 * includes the header length. Thus, the maximum packet length that we
46 * support is 65531 bytes. We send truncated packets if the specified length
47 * is larger than that. Userspace can check for presence of NFQA_CAP_LEN
48 * attribute to detect truncation.
50 #define NFQNL_MAX_COPY_RANGE (0xffff - NLA_HDRLEN)
52 struct nfqnl_instance
{
53 struct hlist_node hlist
; /* global list of queues */
57 unsigned int queue_maxlen
;
58 unsigned int copy_range
;
59 unsigned int queue_dropped
;
60 unsigned int queue_user_dropped
;
63 u_int16_t queue_num
; /* number of this queue */
65 u_int32_t flags
; /* Set using NFQA_CFG_FLAGS */
67 * Following fields are dirtied for each queued packet,
68 * keep them in same cache line if possible.
71 unsigned int queue_total
;
72 unsigned int id_sequence
; /* 'sequence' of pkt ids */
73 struct list_head queue_list
; /* packets in queue */
76 typedef int (*nfqnl_cmpfn
)(struct nf_queue_entry
*, unsigned long);
78 static int nfnl_queue_net_id __read_mostly
;
80 #define INSTANCE_BUCKETS 16
81 struct nfnl_queue_net
{
82 spinlock_t instances_lock
;
83 struct hlist_head instance_table
[INSTANCE_BUCKETS
];
86 static struct nfnl_queue_net
*nfnl_queue_pernet(struct net
*net
)
88 return net_generic(net
, nfnl_queue_net_id
);
91 static inline u_int8_t
instance_hashfn(u_int16_t queue_num
)
93 return ((queue_num
>> 8) ^ queue_num
) % INSTANCE_BUCKETS
;
96 static struct nfqnl_instance
*
97 instance_lookup(struct nfnl_queue_net
*q
, u_int16_t queue_num
)
99 struct hlist_head
*head
;
100 struct nfqnl_instance
*inst
;
102 head
= &q
->instance_table
[instance_hashfn(queue_num
)];
103 hlist_for_each_entry_rcu(inst
, head
, hlist
) {
104 if (inst
->queue_num
== queue_num
)
110 static struct nfqnl_instance
*
111 instance_create(struct nfnl_queue_net
*q
, u_int16_t queue_num
,
114 struct nfqnl_instance
*inst
;
118 spin_lock(&q
->instances_lock
);
119 if (instance_lookup(q
, queue_num
)) {
124 inst
= kzalloc(sizeof(*inst
), GFP_ATOMIC
);
130 inst
->queue_num
= queue_num
;
131 inst
->peer_portid
= portid
;
132 inst
->queue_maxlen
= NFQNL_QMAX_DEFAULT
;
133 inst
->copy_range
= NFQNL_MAX_COPY_RANGE
;
134 inst
->copy_mode
= NFQNL_COPY_NONE
;
135 spin_lock_init(&inst
->lock
);
136 INIT_LIST_HEAD(&inst
->queue_list
);
138 if (!try_module_get(THIS_MODULE
)) {
143 h
= instance_hashfn(queue_num
);
144 hlist_add_head_rcu(&inst
->hlist
, &q
->instance_table
[h
]);
146 spin_unlock(&q
->instances_lock
);
153 spin_unlock(&q
->instances_lock
);
157 static void nfqnl_flush(struct nfqnl_instance
*queue
, nfqnl_cmpfn cmpfn
,
161 instance_destroy_rcu(struct rcu_head
*head
)
163 struct nfqnl_instance
*inst
= container_of(head
, struct nfqnl_instance
,
166 nfqnl_flush(inst
, NULL
, 0);
168 module_put(THIS_MODULE
);
172 __instance_destroy(struct nfqnl_instance
*inst
)
174 hlist_del_rcu(&inst
->hlist
);
175 call_rcu(&inst
->rcu
, instance_destroy_rcu
);
179 instance_destroy(struct nfnl_queue_net
*q
, struct nfqnl_instance
*inst
)
181 spin_lock(&q
->instances_lock
);
182 __instance_destroy(inst
);
183 spin_unlock(&q
->instances_lock
);
187 __enqueue_entry(struct nfqnl_instance
*queue
, struct nf_queue_entry
*entry
)
189 list_add_tail(&entry
->list
, &queue
->queue_list
);
190 queue
->queue_total
++;
194 __dequeue_entry(struct nfqnl_instance
*queue
, struct nf_queue_entry
*entry
)
196 list_del(&entry
->list
);
197 queue
->queue_total
--;
200 static struct nf_queue_entry
*
201 find_dequeue_entry(struct nfqnl_instance
*queue
, unsigned int id
)
203 struct nf_queue_entry
*entry
= NULL
, *i
;
205 spin_lock_bh(&queue
->lock
);
207 list_for_each_entry(i
, &queue
->queue_list
, list
) {
215 __dequeue_entry(queue
, entry
);
217 spin_unlock_bh(&queue
->lock
);
223 nfqnl_flush(struct nfqnl_instance
*queue
, nfqnl_cmpfn cmpfn
, unsigned long data
)
225 struct nf_queue_entry
*entry
, *next
;
227 spin_lock_bh(&queue
->lock
);
228 list_for_each_entry_safe(entry
, next
, &queue
->queue_list
, list
) {
229 if (!cmpfn
|| cmpfn(entry
, data
)) {
230 list_del(&entry
->list
);
231 queue
->queue_total
--;
232 nf_reinject(entry
, NF_DROP
);
235 spin_unlock_bh(&queue
->lock
);
239 nfqnl_zcopy(struct sk_buff
*to
, const struct sk_buff
*from
, int len
, int hlen
)
242 int plen
= 0; /* length of skb->head fragment */
246 /* dont bother with small payloads */
247 if (len
<= skb_tailroom(to
)) {
248 skb_copy_bits(from
, 0, skb_put(to
, len
), len
);
253 skb_copy_bits(from
, 0, skb_put(to
, hlen
), hlen
);
256 plen
= min_t(int, skb_headlen(from
), len
);
258 page
= virt_to_head_page(from
->head
);
259 offset
= from
->data
- (unsigned char *)page_address(page
);
260 __skb_fill_page_desc(to
, 0, page
, offset
, plen
);
267 to
->truesize
+= len
+ plen
;
268 to
->len
+= len
+ plen
;
269 to
->data_len
+= len
+ plen
;
271 for (i
= 0; i
< skb_shinfo(from
)->nr_frags
; i
++) {
274 skb_shinfo(to
)->frags
[j
] = skb_shinfo(from
)->frags
[i
];
275 skb_shinfo(to
)->frags
[j
].size
= min_t(int, skb_shinfo(to
)->frags
[j
].size
, len
);
276 len
-= skb_shinfo(to
)->frags
[j
].size
;
280 skb_shinfo(to
)->nr_frags
= j
;
284 nfqnl_put_packet_info(struct sk_buff
*nlskb
, struct sk_buff
*packet
,
289 if (packet
->ip_summed
== CHECKSUM_PARTIAL
)
290 flags
= NFQA_SKB_CSUMNOTREADY
;
291 else if (csum_verify
)
292 flags
= NFQA_SKB_CSUM_NOTVERIFIED
;
294 if (skb_is_gso(packet
))
295 flags
|= NFQA_SKB_GSO
;
297 return flags
? nla_put_be32(nlskb
, NFQA_SKB_INFO
, htonl(flags
)) : 0;
300 static struct sk_buff
*
301 nfqnl_build_packet_message(struct nfqnl_instance
*queue
,
302 struct nf_queue_entry
*entry
,
303 __be32
**packet_id_ptr
)
306 size_t data_len
= 0, cap_len
= 0;
310 struct nfqnl_msg_packet_hdr
*pmsg
;
311 struct nlmsghdr
*nlh
;
312 struct nfgenmsg
*nfmsg
;
313 struct sk_buff
*entskb
= entry
->skb
;
314 struct net_device
*indev
;
315 struct net_device
*outdev
;
316 struct nf_conn
*ct
= NULL
;
317 enum ip_conntrack_info
uninitialized_var(ctinfo
);
320 size
= nlmsg_total_size(sizeof(struct nfgenmsg
))
321 + nla_total_size(sizeof(struct nfqnl_msg_packet_hdr
))
322 + nla_total_size(sizeof(u_int32_t
)) /* ifindex */
323 + nla_total_size(sizeof(u_int32_t
)) /* ifindex */
324 #ifdef CONFIG_BRIDGE_NETFILTER
325 + nla_total_size(sizeof(u_int32_t
)) /* ifindex */
326 + nla_total_size(sizeof(u_int32_t
)) /* ifindex */
328 + nla_total_size(sizeof(u_int32_t
)) /* mark */
329 + nla_total_size(sizeof(struct nfqnl_msg_packet_hw
))
330 + nla_total_size(sizeof(u_int32_t
)) /* skbinfo */
331 + nla_total_size(sizeof(u_int32_t
)); /* cap_len */
333 if (entskb
->tstamp
.tv64
)
334 size
+= nla_total_size(sizeof(struct nfqnl_msg_packet_timestamp
));
336 if (entry
->hook
<= NF_INET_FORWARD
||
337 (entry
->hook
== NF_INET_POST_ROUTING
&& entskb
->sk
== NULL
))
338 csum_verify
= !skb_csum_unnecessary(entskb
);
342 outdev
= entry
->outdev
;
344 switch ((enum nfqnl_config_mode
)ACCESS_ONCE(queue
->copy_mode
)) {
345 case NFQNL_COPY_META
:
346 case NFQNL_COPY_NONE
:
349 case NFQNL_COPY_PACKET
:
350 if (!(queue
->flags
& NFQA_CFG_F_GSO
) &&
351 entskb
->ip_summed
== CHECKSUM_PARTIAL
&&
352 skb_checksum_help(entskb
))
355 data_len
= ACCESS_ONCE(queue
->copy_range
);
356 if (data_len
> entskb
->len
)
357 data_len
= entskb
->len
;
359 if (!entskb
->head_frag
||
360 skb_headlen(entskb
) < L1_CACHE_BYTES
||
361 skb_shinfo(entskb
)->nr_frags
>= MAX_SKB_FRAGS
)
362 hlen
= skb_headlen(entskb
);
364 if (skb_has_frag_list(entskb
))
366 hlen
= min_t(int, data_len
, hlen
);
367 size
+= sizeof(struct nlattr
) + hlen
;
368 cap_len
= entskb
->len
;
372 if (queue
->flags
& NFQA_CFG_F_CONNTRACK
)
373 ct
= nfqnl_ct_get(entskb
, &size
, &ctinfo
);
375 skb
= nfnetlink_alloc_skb(&init_net
, size
, queue
->peer_portid
,
380 nlh
= nlmsg_put(skb
, 0, 0,
381 NFNL_SUBSYS_QUEUE
<< 8 | NFQNL_MSG_PACKET
,
382 sizeof(struct nfgenmsg
), 0);
387 nfmsg
= nlmsg_data(nlh
);
388 nfmsg
->nfgen_family
= entry
->pf
;
389 nfmsg
->version
= NFNETLINK_V0
;
390 nfmsg
->res_id
= htons(queue
->queue_num
);
392 nla
= __nla_reserve(skb
, NFQA_PACKET_HDR
, sizeof(*pmsg
));
393 pmsg
= nla_data(nla
);
394 pmsg
->hw_protocol
= entskb
->protocol
;
395 pmsg
->hook
= entry
->hook
;
396 *packet_id_ptr
= &pmsg
->packet_id
;
398 indev
= entry
->indev
;
400 #ifndef CONFIG_BRIDGE_NETFILTER
401 if (nla_put_be32(skb
, NFQA_IFINDEX_INDEV
, htonl(indev
->ifindex
)))
402 goto nla_put_failure
;
404 if (entry
->pf
== PF_BRIDGE
) {
405 /* Case 1: indev is physical input device, we need to
406 * look for bridge group (when called from
407 * netfilter_bridge) */
408 if (nla_put_be32(skb
, NFQA_IFINDEX_PHYSINDEV
,
409 htonl(indev
->ifindex
)) ||
410 /* this is the bridge group "brX" */
411 /* rcu_read_lock()ed by __nf_queue */
412 nla_put_be32(skb
, NFQA_IFINDEX_INDEV
,
413 htonl(br_port_get_rcu(indev
)->br
->dev
->ifindex
)))
414 goto nla_put_failure
;
416 /* Case 2: indev is bridge group, we need to look for
417 * physical device (when called from ipv4) */
418 if (nla_put_be32(skb
, NFQA_IFINDEX_INDEV
,
419 htonl(indev
->ifindex
)))
420 goto nla_put_failure
;
421 if (entskb
->nf_bridge
&& entskb
->nf_bridge
->physindev
&&
422 nla_put_be32(skb
, NFQA_IFINDEX_PHYSINDEV
,
423 htonl(entskb
->nf_bridge
->physindev
->ifindex
)))
424 goto nla_put_failure
;
430 #ifndef CONFIG_BRIDGE_NETFILTER
431 if (nla_put_be32(skb
, NFQA_IFINDEX_OUTDEV
, htonl(outdev
->ifindex
)))
432 goto nla_put_failure
;
434 if (entry
->pf
== PF_BRIDGE
) {
435 /* Case 1: outdev is physical output device, we need to
436 * look for bridge group (when called from
437 * netfilter_bridge) */
438 if (nla_put_be32(skb
, NFQA_IFINDEX_PHYSOUTDEV
,
439 htonl(outdev
->ifindex
)) ||
440 /* this is the bridge group "brX" */
441 /* rcu_read_lock()ed by __nf_queue */
442 nla_put_be32(skb
, NFQA_IFINDEX_OUTDEV
,
443 htonl(br_port_get_rcu(outdev
)->br
->dev
->ifindex
)))
444 goto nla_put_failure
;
446 /* Case 2: outdev is bridge group, we need to look for
447 * physical output device (when called from ipv4) */
448 if (nla_put_be32(skb
, NFQA_IFINDEX_OUTDEV
,
449 htonl(outdev
->ifindex
)))
450 goto nla_put_failure
;
451 if (entskb
->nf_bridge
&& entskb
->nf_bridge
->physoutdev
&&
452 nla_put_be32(skb
, NFQA_IFINDEX_PHYSOUTDEV
,
453 htonl(entskb
->nf_bridge
->physoutdev
->ifindex
)))
454 goto nla_put_failure
;
460 nla_put_be32(skb
, NFQA_MARK
, htonl(entskb
->mark
)))
461 goto nla_put_failure
;
463 if (indev
&& entskb
->dev
&&
464 entskb
->mac_header
!= entskb
->network_header
) {
465 struct nfqnl_msg_packet_hw phw
;
468 memset(&phw
, 0, sizeof(phw
));
469 len
= dev_parse_header(entskb
, phw
.hw_addr
);
471 phw
.hw_addrlen
= htons(len
);
472 if (nla_put(skb
, NFQA_HWADDR
, sizeof(phw
), &phw
))
473 goto nla_put_failure
;
477 if (entskb
->tstamp
.tv64
) {
478 struct nfqnl_msg_packet_timestamp ts
;
479 struct timeval tv
= ktime_to_timeval(entskb
->tstamp
);
480 ts
.sec
= cpu_to_be64(tv
.tv_sec
);
481 ts
.usec
= cpu_to_be64(tv
.tv_usec
);
483 if (nla_put(skb
, NFQA_TIMESTAMP
, sizeof(ts
), &ts
))
484 goto nla_put_failure
;
487 if (ct
&& nfqnl_ct_put(skb
, ct
, ctinfo
) < 0)
488 goto nla_put_failure
;
490 if (cap_len
> data_len
&&
491 nla_put_be32(skb
, NFQA_CAP_LEN
, htonl(cap_len
)))
492 goto nla_put_failure
;
494 if (nfqnl_put_packet_info(skb
, entskb
, csum_verify
))
495 goto nla_put_failure
;
500 if (skb_tailroom(skb
) < sizeof(*nla
) + hlen
)
501 goto nla_put_failure
;
503 nla
= (struct nlattr
*)skb_put(skb
, sizeof(*nla
));
504 nla
->nla_type
= NFQA_PAYLOAD
;
505 nla
->nla_len
= nla_attr_size(data_len
);
507 nfqnl_zcopy(skb
, entskb
, data_len
, hlen
);
510 nlh
->nlmsg_len
= skb
->len
;
515 net_err_ratelimited("nf_queue: error creating packet message\n");
520 __nfqnl_enqueue_packet(struct net
*net
, struct nfqnl_instance
*queue
,
521 struct nf_queue_entry
*entry
)
523 struct sk_buff
*nskb
;
525 __be32
*packet_id_ptr
;
528 nskb
= nfqnl_build_packet_message(queue
, entry
, &packet_id_ptr
);
533 spin_lock_bh(&queue
->lock
);
535 if (queue
->queue_total
>= queue
->queue_maxlen
) {
536 if (queue
->flags
& NFQA_CFG_F_FAIL_OPEN
) {
540 queue
->queue_dropped
++;
541 net_warn_ratelimited("nf_queue: full at %d entries, dropping packets(s)\n",
544 goto err_out_free_nskb
;
546 entry
->id
= ++queue
->id_sequence
;
547 *packet_id_ptr
= htonl(entry
->id
);
549 /* nfnetlink_unicast will either free the nskb or add it to a socket */
550 err
= nfnetlink_unicast(nskb
, net
, queue
->peer_portid
, MSG_DONTWAIT
);
552 queue
->queue_user_dropped
++;
556 __enqueue_entry(queue
, entry
);
558 spin_unlock_bh(&queue
->lock
);
564 spin_unlock_bh(&queue
->lock
);
566 nf_reinject(entry
, NF_ACCEPT
);
571 static struct nf_queue_entry
*
572 nf_queue_entry_dup(struct nf_queue_entry
*e
)
574 struct nf_queue_entry
*entry
= kmemdup(e
, e
->size
, GFP_ATOMIC
);
576 if (nf_queue_entry_get_refs(entry
))
583 #ifdef CONFIG_BRIDGE_NETFILTER
584 /* When called from bridge netfilter, skb->data must point to MAC header
585 * before calling skb_gso_segment(). Else, original MAC header is lost
586 * and segmented skbs will be sent to wrong destination.
588 static void nf_bridge_adjust_skb_data(struct sk_buff
*skb
)
591 __skb_push(skb
, skb
->network_header
- skb
->mac_header
);
594 static void nf_bridge_adjust_segmented_data(struct sk_buff
*skb
)
597 __skb_pull(skb
, skb
->network_header
- skb
->mac_header
);
600 #define nf_bridge_adjust_skb_data(s) do {} while (0)
601 #define nf_bridge_adjust_segmented_data(s) do {} while (0)
604 static void free_entry(struct nf_queue_entry
*entry
)
606 nf_queue_entry_release_refs(entry
);
611 __nfqnl_enqueue_packet_gso(struct net
*net
, struct nfqnl_instance
*queue
,
612 struct sk_buff
*skb
, struct nf_queue_entry
*entry
)
615 struct nf_queue_entry
*entry_seg
;
617 nf_bridge_adjust_segmented_data(skb
);
619 if (skb
->next
== NULL
) { /* last packet, no need to copy entry */
620 struct sk_buff
*gso_skb
= entry
->skb
;
622 ret
= __nfqnl_enqueue_packet(net
, queue
, entry
);
624 entry
->skb
= gso_skb
;
630 entry_seg
= nf_queue_entry_dup(entry
);
632 entry_seg
->skb
= skb
;
633 ret
= __nfqnl_enqueue_packet(net
, queue
, entry_seg
);
635 free_entry(entry_seg
);
641 nfqnl_enqueue_packet(struct nf_queue_entry
*entry
, unsigned int queuenum
)
644 struct nfqnl_instance
*queue
;
645 struct sk_buff
*skb
, *segs
;
647 struct net
*net
= dev_net(entry
->indev
?
648 entry
->indev
: entry
->outdev
);
649 struct nfnl_queue_net
*q
= nfnl_queue_pernet(net
);
651 /* rcu_read_lock()ed by nf_hook_slow() */
652 queue
= instance_lookup(q
, queuenum
);
656 if (queue
->copy_mode
== NFQNL_COPY_NONE
)
663 skb
->protocol
= htons(ETH_P_IP
);
666 skb
->protocol
= htons(ETH_P_IPV6
);
670 if ((queue
->flags
& NFQA_CFG_F_GSO
) || !skb_is_gso(skb
))
671 return __nfqnl_enqueue_packet(net
, queue
, entry
);
673 nf_bridge_adjust_skb_data(skb
);
674 segs
= skb_gso_segment(skb
, 0);
675 /* Does not use PTR_ERR to limit the number of error codes that can be
676 * returned by nf_queue. For instance, callers rely on -ECANCELED to
677 * mean 'ignore this hook'.
684 struct sk_buff
*nskb
= segs
->next
;
686 err
= __nfqnl_enqueue_packet_gso(net
, queue
,
696 if (err
) /* some segments are already queued */
702 nf_bridge_adjust_segmented_data(skb
);
707 nfqnl_mangle(void *data
, int data_len
, struct nf_queue_entry
*e
, int diff
)
709 struct sk_buff
*nskb
;
712 if (pskb_trim(e
->skb
, data_len
))
714 } else if (diff
> 0) {
715 if (data_len
> 0xFFFF)
717 if (diff
> skb_tailroom(e
->skb
)) {
718 nskb
= skb_copy_expand(e
->skb
, skb_headroom(e
->skb
),
721 printk(KERN_WARNING
"nf_queue: OOM "
722 "in mangle, dropping packet\n");
728 skb_put(e
->skb
, diff
);
730 if (!skb_make_writable(e
->skb
, data_len
))
732 skb_copy_to_linear_data(e
->skb
, data
, data_len
);
733 e
->skb
->ip_summed
= CHECKSUM_NONE
;
738 nfqnl_set_mode(struct nfqnl_instance
*queue
,
739 unsigned char mode
, unsigned int range
)
743 spin_lock_bh(&queue
->lock
);
745 case NFQNL_COPY_NONE
:
746 case NFQNL_COPY_META
:
747 queue
->copy_mode
= mode
;
748 queue
->copy_range
= 0;
751 case NFQNL_COPY_PACKET
:
752 queue
->copy_mode
= mode
;
753 if (range
== 0 || range
> NFQNL_MAX_COPY_RANGE
)
754 queue
->copy_range
= NFQNL_MAX_COPY_RANGE
;
756 queue
->copy_range
= range
;
763 spin_unlock_bh(&queue
->lock
);
769 dev_cmp(struct nf_queue_entry
*entry
, unsigned long ifindex
)
772 if (entry
->indev
->ifindex
== ifindex
)
775 if (entry
->outdev
->ifindex
== ifindex
)
777 #ifdef CONFIG_BRIDGE_NETFILTER
778 if (entry
->skb
->nf_bridge
) {
779 if (entry
->skb
->nf_bridge
->physindev
&&
780 entry
->skb
->nf_bridge
->physindev
->ifindex
== ifindex
)
782 if (entry
->skb
->nf_bridge
->physoutdev
&&
783 entry
->skb
->nf_bridge
->physoutdev
->ifindex
== ifindex
)
790 /* drop all packets with either indev or outdev == ifindex from all queue
793 nfqnl_dev_drop(struct net
*net
, int ifindex
)
796 struct nfnl_queue_net
*q
= nfnl_queue_pernet(net
);
800 for (i
= 0; i
< INSTANCE_BUCKETS
; i
++) {
801 struct nfqnl_instance
*inst
;
802 struct hlist_head
*head
= &q
->instance_table
[i
];
804 hlist_for_each_entry_rcu(inst
, head
, hlist
)
805 nfqnl_flush(inst
, dev_cmp
, ifindex
);
811 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
814 nfqnl_rcv_dev_event(struct notifier_block
*this,
815 unsigned long event
, void *ptr
)
817 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
819 /* Drop any packets associated with the downed device */
820 if (event
== NETDEV_DOWN
)
821 nfqnl_dev_drop(dev_net(dev
), dev
->ifindex
);
825 static struct notifier_block nfqnl_dev_notifier
= {
826 .notifier_call
= nfqnl_rcv_dev_event
,
830 nfqnl_rcv_nl_event(struct notifier_block
*this,
831 unsigned long event
, void *ptr
)
833 struct netlink_notify
*n
= ptr
;
834 struct nfnl_queue_net
*q
= nfnl_queue_pernet(n
->net
);
836 if (event
== NETLINK_URELEASE
&& n
->protocol
== NETLINK_NETFILTER
) {
839 /* destroy all instances for this portid */
840 spin_lock(&q
->instances_lock
);
841 for (i
= 0; i
< INSTANCE_BUCKETS
; i
++) {
842 struct hlist_node
*t2
;
843 struct nfqnl_instance
*inst
;
844 struct hlist_head
*head
= &q
->instance_table
[i
];
846 hlist_for_each_entry_safe(inst
, t2
, head
, hlist
) {
847 if (n
->portid
== inst
->peer_portid
)
848 __instance_destroy(inst
);
851 spin_unlock(&q
->instances_lock
);
856 static struct notifier_block nfqnl_rtnl_notifier
= {
857 .notifier_call
= nfqnl_rcv_nl_event
,
860 static const struct nla_policy nfqa_verdict_policy
[NFQA_MAX
+1] = {
861 [NFQA_VERDICT_HDR
] = { .len
= sizeof(struct nfqnl_msg_verdict_hdr
) },
862 [NFQA_MARK
] = { .type
= NLA_U32
},
863 [NFQA_PAYLOAD
] = { .type
= NLA_UNSPEC
},
864 [NFQA_CT
] = { .type
= NLA_UNSPEC
},
865 [NFQA_EXP
] = { .type
= NLA_UNSPEC
},
868 static const struct nla_policy nfqa_verdict_batch_policy
[NFQA_MAX
+1] = {
869 [NFQA_VERDICT_HDR
] = { .len
= sizeof(struct nfqnl_msg_verdict_hdr
) },
870 [NFQA_MARK
] = { .type
= NLA_U32
},
873 static struct nfqnl_instance
*
874 verdict_instance_lookup(struct nfnl_queue_net
*q
, u16 queue_num
, int nlportid
)
876 struct nfqnl_instance
*queue
;
878 queue
= instance_lookup(q
, queue_num
);
880 return ERR_PTR(-ENODEV
);
882 if (queue
->peer_portid
!= nlportid
)
883 return ERR_PTR(-EPERM
);
888 static struct nfqnl_msg_verdict_hdr
*
889 verdicthdr_get(const struct nlattr
* const nfqa
[])
891 struct nfqnl_msg_verdict_hdr
*vhdr
;
892 unsigned int verdict
;
894 if (!nfqa
[NFQA_VERDICT_HDR
])
897 vhdr
= nla_data(nfqa
[NFQA_VERDICT_HDR
]);
898 verdict
= ntohl(vhdr
->verdict
) & NF_VERDICT_MASK
;
899 if (verdict
> NF_MAX_VERDICT
|| verdict
== NF_STOLEN
)
904 static int nfq_id_after(unsigned int id
, unsigned int max
)
906 return (int)(id
- max
) > 0;
910 nfqnl_recv_verdict_batch(struct sock
*ctnl
, struct sk_buff
*skb
,
911 const struct nlmsghdr
*nlh
,
912 const struct nlattr
* const nfqa
[])
914 struct nfgenmsg
*nfmsg
= nlmsg_data(nlh
);
915 struct nf_queue_entry
*entry
, *tmp
;
916 unsigned int verdict
, maxid
;
917 struct nfqnl_msg_verdict_hdr
*vhdr
;
918 struct nfqnl_instance
*queue
;
919 LIST_HEAD(batch_list
);
920 u16 queue_num
= ntohs(nfmsg
->res_id
);
922 struct net
*net
= sock_net(ctnl
);
923 struct nfnl_queue_net
*q
= nfnl_queue_pernet(net
);
925 queue
= verdict_instance_lookup(q
, queue_num
,
926 NETLINK_CB(skb
).portid
);
928 return PTR_ERR(queue
);
930 vhdr
= verdicthdr_get(nfqa
);
934 verdict
= ntohl(vhdr
->verdict
);
935 maxid
= ntohl(vhdr
->id
);
937 spin_lock_bh(&queue
->lock
);
939 list_for_each_entry_safe(entry
, tmp
, &queue
->queue_list
, list
) {
940 if (nfq_id_after(entry
->id
, maxid
))
942 __dequeue_entry(queue
, entry
);
943 list_add_tail(&entry
->list
, &batch_list
);
946 spin_unlock_bh(&queue
->lock
);
948 if (list_empty(&batch_list
))
951 list_for_each_entry_safe(entry
, tmp
, &batch_list
, list
) {
953 entry
->skb
->mark
= ntohl(nla_get_be32(nfqa
[NFQA_MARK
]));
954 nf_reinject(entry
, verdict
);
960 nfqnl_recv_verdict(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 u_int16_t queue_num
= ntohs(nfmsg
->res_id
);
967 struct nfqnl_msg_verdict_hdr
*vhdr
;
968 struct nfqnl_instance
*queue
;
969 unsigned int verdict
;
970 struct nf_queue_entry
*entry
;
971 enum ip_conntrack_info
uninitialized_var(ctinfo
);
972 struct nf_conn
*ct
= NULL
;
974 struct net
*net
= sock_net(ctnl
);
975 struct nfnl_queue_net
*q
= nfnl_queue_pernet(net
);
977 queue
= instance_lookup(q
, queue_num
);
979 queue
= verdict_instance_lookup(q
, queue_num
,
980 NETLINK_CB(skb
).portid
);
982 return PTR_ERR(queue
);
984 vhdr
= verdicthdr_get(nfqa
);
988 verdict
= ntohl(vhdr
->verdict
);
990 entry
= find_dequeue_entry(queue
, ntohl(vhdr
->id
));
995 ct
= nfqnl_ct_parse(entry
->skb
, nfqa
[NFQA_CT
], &ctinfo
);
996 if (ct
&& nfqa
[NFQA_EXP
]) {
997 nfqnl_attach_expect(ct
, nfqa
[NFQA_EXP
],
998 NETLINK_CB(skb
).portid
,
1003 if (nfqa
[NFQA_PAYLOAD
]) {
1004 u16 payload_len
= nla_len(nfqa
[NFQA_PAYLOAD
]);
1005 int diff
= payload_len
- entry
->skb
->len
;
1007 if (nfqnl_mangle(nla_data(nfqa
[NFQA_PAYLOAD
]),
1008 payload_len
, entry
, diff
) < 0)
1012 nfqnl_ct_seq_adjust(entry
->skb
, ct
, ctinfo
, diff
);
1015 if (nfqa
[NFQA_MARK
])
1016 entry
->skb
->mark
= ntohl(nla_get_be32(nfqa
[NFQA_MARK
]));
1018 nf_reinject(entry
, verdict
);
1023 nfqnl_recv_unsupp(struct sock
*ctnl
, struct sk_buff
*skb
,
1024 const struct nlmsghdr
*nlh
,
1025 const struct nlattr
* const nfqa
[])
1030 static const struct nla_policy nfqa_cfg_policy
[NFQA_CFG_MAX
+1] = {
1031 [NFQA_CFG_CMD
] = { .len
= sizeof(struct nfqnl_msg_config_cmd
) },
1032 [NFQA_CFG_PARAMS
] = { .len
= sizeof(struct nfqnl_msg_config_params
) },
1035 static const struct nf_queue_handler nfqh
= {
1036 .outfn
= &nfqnl_enqueue_packet
,
1040 nfqnl_recv_config(struct sock
*ctnl
, struct sk_buff
*skb
,
1041 const struct nlmsghdr
*nlh
,
1042 const struct nlattr
* const nfqa
[])
1044 struct nfgenmsg
*nfmsg
= nlmsg_data(nlh
);
1045 u_int16_t queue_num
= ntohs(nfmsg
->res_id
);
1046 struct nfqnl_instance
*queue
;
1047 struct nfqnl_msg_config_cmd
*cmd
= NULL
;
1048 struct net
*net
= sock_net(ctnl
);
1049 struct nfnl_queue_net
*q
= nfnl_queue_pernet(net
);
1052 if (nfqa
[NFQA_CFG_CMD
]) {
1053 cmd
= nla_data(nfqa
[NFQA_CFG_CMD
]);
1055 /* Obsolete commands without queue context */
1056 switch (cmd
->command
) {
1057 case NFQNL_CFG_CMD_PF_BIND
: return 0;
1058 case NFQNL_CFG_CMD_PF_UNBIND
: return 0;
1063 queue
= instance_lookup(q
, queue_num
);
1064 if (queue
&& queue
->peer_portid
!= NETLINK_CB(skb
).portid
) {
1066 goto err_out_unlock
;
1070 switch (cmd
->command
) {
1071 case NFQNL_CFG_CMD_BIND
:
1074 goto err_out_unlock
;
1076 queue
= instance_create(q
, queue_num
,
1077 NETLINK_CB(skb
).portid
);
1078 if (IS_ERR(queue
)) {
1079 ret
= PTR_ERR(queue
);
1080 goto err_out_unlock
;
1083 case NFQNL_CFG_CMD_UNBIND
:
1086 goto err_out_unlock
;
1088 instance_destroy(q
, queue
);
1090 case NFQNL_CFG_CMD_PF_BIND
:
1091 case NFQNL_CFG_CMD_PF_UNBIND
:
1099 if (nfqa
[NFQA_CFG_PARAMS
]) {
1100 struct nfqnl_msg_config_params
*params
;
1104 goto err_out_unlock
;
1106 params
= nla_data(nfqa
[NFQA_CFG_PARAMS
]);
1107 nfqnl_set_mode(queue
, params
->copy_mode
,
1108 ntohl(params
->copy_range
));
1111 if (nfqa
[NFQA_CFG_QUEUE_MAXLEN
]) {
1112 __be32
*queue_maxlen
;
1116 goto err_out_unlock
;
1118 queue_maxlen
= nla_data(nfqa
[NFQA_CFG_QUEUE_MAXLEN
]);
1119 spin_lock_bh(&queue
->lock
);
1120 queue
->queue_maxlen
= ntohl(*queue_maxlen
);
1121 spin_unlock_bh(&queue
->lock
);
1124 if (nfqa
[NFQA_CFG_FLAGS
]) {
1129 goto err_out_unlock
;
1132 if (!nfqa
[NFQA_CFG_MASK
]) {
1133 /* A mask is needed to specify which flags are being
1137 goto err_out_unlock
;
1140 flags
= ntohl(nla_get_be32(nfqa
[NFQA_CFG_FLAGS
]));
1141 mask
= ntohl(nla_get_be32(nfqa
[NFQA_CFG_MASK
]));
1143 if (flags
>= NFQA_CFG_F_MAX
) {
1145 goto err_out_unlock
;
1148 spin_lock_bh(&queue
->lock
);
1149 queue
->flags
&= ~mask
;
1150 queue
->flags
|= flags
& mask
;
1151 spin_unlock_bh(&queue
->lock
);
1159 static const struct nfnl_callback nfqnl_cb
[NFQNL_MSG_MAX
] = {
1160 [NFQNL_MSG_PACKET
] = { .call_rcu
= nfqnl_recv_unsupp
,
1161 .attr_count
= NFQA_MAX
, },
1162 [NFQNL_MSG_VERDICT
] = { .call_rcu
= nfqnl_recv_verdict
,
1163 .attr_count
= NFQA_MAX
,
1164 .policy
= nfqa_verdict_policy
},
1165 [NFQNL_MSG_CONFIG
] = { .call
= nfqnl_recv_config
,
1166 .attr_count
= NFQA_CFG_MAX
,
1167 .policy
= nfqa_cfg_policy
},
1168 [NFQNL_MSG_VERDICT_BATCH
]={ .call_rcu
= nfqnl_recv_verdict_batch
,
1169 .attr_count
= NFQA_MAX
,
1170 .policy
= nfqa_verdict_batch_policy
},
1173 static const struct nfnetlink_subsystem nfqnl_subsys
= {
1175 .subsys_id
= NFNL_SUBSYS_QUEUE
,
1176 .cb_count
= NFQNL_MSG_MAX
,
1180 #ifdef CONFIG_PROC_FS
1182 struct seq_net_private p
;
1183 unsigned int bucket
;
1186 static struct hlist_node
*get_first(struct seq_file
*seq
)
1188 struct iter_state
*st
= seq
->private;
1190 struct nfnl_queue_net
*q
;
1195 net
= seq_file_net(seq
);
1196 q
= nfnl_queue_pernet(net
);
1197 for (st
->bucket
= 0; st
->bucket
< INSTANCE_BUCKETS
; st
->bucket
++) {
1198 if (!hlist_empty(&q
->instance_table
[st
->bucket
]))
1199 return q
->instance_table
[st
->bucket
].first
;
1204 static struct hlist_node
*get_next(struct seq_file
*seq
, struct hlist_node
*h
)
1206 struct iter_state
*st
= seq
->private;
1207 struct net
*net
= seq_file_net(seq
);
1211 struct nfnl_queue_net
*q
;
1213 if (++st
->bucket
>= INSTANCE_BUCKETS
)
1216 q
= nfnl_queue_pernet(net
);
1217 h
= q
->instance_table
[st
->bucket
].first
;
1222 static struct hlist_node
*get_idx(struct seq_file
*seq
, loff_t pos
)
1224 struct hlist_node
*head
;
1225 head
= get_first(seq
);
1228 while (pos
&& (head
= get_next(seq
, head
)))
1230 return pos
? NULL
: head
;
1233 static void *seq_start(struct seq_file
*s
, loff_t
*pos
)
1234 __acquires(nfnl_queue_pernet(seq_file_net(s
))->instances_lock
)
1236 spin_lock(&nfnl_queue_pernet(seq_file_net(s
))->instances_lock
);
1237 return get_idx(s
, *pos
);
1240 static void *seq_next(struct seq_file
*s
, void *v
, loff_t
*pos
)
1243 return get_next(s
, v
);
1246 static void seq_stop(struct seq_file
*s
, void *v
)
1247 __releases(nfnl_queue_pernet(seq_file_net(s
))->instances_lock
)
1249 spin_unlock(&nfnl_queue_pernet(seq_file_net(s
))->instances_lock
);
1252 static int seq_show(struct seq_file
*s
, void *v
)
1254 const struct nfqnl_instance
*inst
= v
;
1256 return seq_printf(s
, "%5d %6d %5d %1d %5d %5d %5d %8d %2d\n",
1258 inst
->peer_portid
, inst
->queue_total
,
1259 inst
->copy_mode
, inst
->copy_range
,
1260 inst
->queue_dropped
, inst
->queue_user_dropped
,
1261 inst
->id_sequence
, 1);
1264 static const struct seq_operations nfqnl_seq_ops
= {
1271 static int nfqnl_open(struct inode
*inode
, struct file
*file
)
1273 return seq_open_net(inode
, file
, &nfqnl_seq_ops
,
1274 sizeof(struct iter_state
));
1277 static const struct file_operations nfqnl_file_ops
= {
1278 .owner
= THIS_MODULE
,
1281 .llseek
= seq_lseek
,
1282 .release
= seq_release_net
,
1285 #endif /* PROC_FS */
1287 static int __net_init
nfnl_queue_net_init(struct net
*net
)
1290 struct nfnl_queue_net
*q
= nfnl_queue_pernet(net
);
1292 for (i
= 0; i
< INSTANCE_BUCKETS
; i
++)
1293 INIT_HLIST_HEAD(&q
->instance_table
[i
]);
1295 spin_lock_init(&q
->instances_lock
);
1297 #ifdef CONFIG_PROC_FS
1298 if (!proc_create("nfnetlink_queue", 0440,
1299 net
->nf
.proc_netfilter
, &nfqnl_file_ops
))
1305 static void __net_exit
nfnl_queue_net_exit(struct net
*net
)
1307 #ifdef CONFIG_PROC_FS
1308 remove_proc_entry("nfnetlink_queue", net
->nf
.proc_netfilter
);
1312 static struct pernet_operations nfnl_queue_net_ops
= {
1313 .init
= nfnl_queue_net_init
,
1314 .exit
= nfnl_queue_net_exit
,
1315 .id
= &nfnl_queue_net_id
,
1316 .size
= sizeof(struct nfnl_queue_net
),
1319 static int __init
nfnetlink_queue_init(void)
1321 int status
= -ENOMEM
;
1323 netlink_register_notifier(&nfqnl_rtnl_notifier
);
1324 status
= nfnetlink_subsys_register(&nfqnl_subsys
);
1326 pr_err("nf_queue: failed to create netlink socket\n");
1327 goto cleanup_netlink_notifier
;
1330 status
= register_pernet_subsys(&nfnl_queue_net_ops
);
1332 pr_err("nf_queue: failed to register pernet ops\n");
1333 goto cleanup_subsys
;
1335 register_netdevice_notifier(&nfqnl_dev_notifier
);
1336 nf_register_queue_handler(&nfqh
);
1340 nfnetlink_subsys_unregister(&nfqnl_subsys
);
1341 cleanup_netlink_notifier
:
1342 netlink_unregister_notifier(&nfqnl_rtnl_notifier
);
1346 static void __exit
nfnetlink_queue_fini(void)
1348 nf_unregister_queue_handler();
1349 unregister_netdevice_notifier(&nfqnl_dev_notifier
);
1350 unregister_pernet_subsys(&nfnl_queue_net_ops
);
1351 nfnetlink_subsys_unregister(&nfqnl_subsys
);
1352 netlink_unregister_notifier(&nfqnl_rtnl_notifier
);
1354 rcu_barrier(); /* Wait for completion of call_rcu()'s */
1357 MODULE_DESCRIPTION("netfilter packet queue handler");
1358 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1359 MODULE_LICENSE("GPL");
1360 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE
);
1362 module_init(nfnetlink_queue_init
);
1363 module_exit(nfnetlink_queue_fini
);