2 * This is a module which is used for queueing packets and communicating with
3 * userspace via nfetlink.
5 * (C) 2005 by Harald Welte <laforge@netfilter.org>
7 * Based on the old ipv4-only ip_queue.c:
8 * (C) 2000-2002 James Morris <jmorris@intercode.com.au>
9 * (C) 2003-2005 Netfilter Core Team <coreteam@netfilter.org>
11 * This program is free software; you can redistribute it and/or modify
12 * it under the terms of the GNU General Public License version 2 as
13 * published by the Free Software Foundation.
16 #include <linux/module.h>
17 #include <linux/skbuff.h>
18 #include <linux/init.h>
19 #include <linux/spinlock.h>
20 #include <linux/notifier.h>
21 #include <linux/netdevice.h>
22 #include <linux/netfilter.h>
23 #include <linux/proc_fs.h>
24 #include <linux/netfilter_ipv4.h>
25 #include <linux/netfilter_ipv6.h>
26 #include <linux/netfilter/nfnetlink.h>
27 #include <linux/netfilter/nfnetlink_queue.h>
28 #include <linux/list.h>
31 #include <asm/atomic.h>
33 #ifdef CONFIG_BRIDGE_NETFILTER
34 #include "../bridge/br_private.h"
37 #define NFQNL_QMAX_DEFAULT 1024
40 #define QDEBUG(x, args ...) printk(KERN_DEBUG "%s(%d):%s(): " x, \
41 __FILE__, __LINE__, __FUNCTION__, \
44 #define QDEBUG(x, ...)
47 struct nfqnl_queue_entry
{
48 struct list_head list
;
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_total
;
62 unsigned int queue_dropped
;
63 unsigned int queue_user_dropped
;
65 atomic_t id_sequence
; /* 'sequence' of pkt ids */
67 u_int16_t queue_num
; /* number of this queue */
72 struct list_head queue_list
; /* packets in queue */
75 typedef int (*nfqnl_cmpfn
)(struct nfqnl_queue_entry
*, unsigned long);
77 static DEFINE_RWLOCK(instances_lock
);
79 #define INSTANCE_BUCKETS 16
80 static struct hlist_head instance_table
[INSTANCE_BUCKETS
];
82 static inline u_int8_t
instance_hashfn(u_int16_t queue_num
)
84 return ((queue_num
>> 8) | queue_num
) % INSTANCE_BUCKETS
;
87 static struct nfqnl_instance
*
88 __instance_lookup(u_int16_t queue_num
)
90 struct hlist_head
*head
;
91 struct hlist_node
*pos
;
92 struct nfqnl_instance
*inst
;
94 head
= &instance_table
[instance_hashfn(queue_num
)];
95 hlist_for_each_entry(inst
, pos
, head
, hlist
) {
96 if (inst
->queue_num
== queue_num
)
102 static struct nfqnl_instance
*
103 instance_lookup_get(u_int16_t queue_num
)
105 struct nfqnl_instance
*inst
;
107 read_lock_bh(&instances_lock
);
108 inst
= __instance_lookup(queue_num
);
110 atomic_inc(&inst
->use
);
111 read_unlock_bh(&instances_lock
);
117 instance_put(struct nfqnl_instance
*inst
)
119 if (inst
&& atomic_dec_and_test(&inst
->use
)) {
120 QDEBUG("kfree(inst=%p)\n", inst
);
125 static struct nfqnl_instance
*
126 instance_create(u_int16_t queue_num
, int pid
)
128 struct nfqnl_instance
*inst
;
130 QDEBUG("entering for queue_num=%u, pid=%d\n", queue_num
, pid
);
132 write_lock_bh(&instances_lock
);
133 if (__instance_lookup(queue_num
)) {
135 QDEBUG("aborting, instance already exists\n");
139 inst
= kmalloc(sizeof(*inst
), GFP_ATOMIC
);
143 memset(inst
, 0, sizeof(*inst
));
144 inst
->queue_num
= queue_num
;
145 inst
->peer_pid
= pid
;
146 inst
->queue_maxlen
= NFQNL_QMAX_DEFAULT
;
147 inst
->copy_range
= 0xfffff;
148 inst
->copy_mode
= NFQNL_COPY_NONE
;
149 atomic_set(&inst
->id_sequence
, 0);
150 /* needs to be two, since we _put() after creation */
151 atomic_set(&inst
->use
, 2);
152 inst
->lock
= SPIN_LOCK_UNLOCKED
;
153 INIT_LIST_HEAD(&inst
->queue_list
);
155 if (!try_module_get(THIS_MODULE
))
158 hlist_add_head(&inst
->hlist
,
159 &instance_table
[instance_hashfn(queue_num
)]);
161 write_unlock_bh(&instances_lock
);
163 QDEBUG("successfully created new instance\n");
170 write_unlock_bh(&instances_lock
);
174 static void nfqnl_flush(struct nfqnl_instance
*queue
, int verdict
);
177 _instance_destroy2(struct nfqnl_instance
*inst
, int lock
)
179 /* first pull it out of the global list */
181 write_lock_bh(&instances_lock
);
183 QDEBUG("removing instance %p (queuenum=%u) from hash\n",
184 inst
, inst
->queue_num
);
185 hlist_del(&inst
->hlist
);
188 write_unlock_bh(&instances_lock
);
190 /* then flush all pending skbs from the queue */
191 nfqnl_flush(inst
, NF_DROP
);
193 /* and finally put the refcount */
196 module_put(THIS_MODULE
);
200 __instance_destroy(struct nfqnl_instance
*inst
)
202 _instance_destroy2(inst
, 0);
206 instance_destroy(struct nfqnl_instance
*inst
)
208 _instance_destroy2(inst
, 1);
214 issue_verdict(struct nfqnl_queue_entry
*entry
, int verdict
)
216 QDEBUG("entering for entry %p, verdict %u\n", entry
, verdict
);
218 /* TCP input path (and probably other bits) assume to be called
219 * from softirq context, not from syscall, like issue_verdict is
220 * called. TCP input path deadlocks with locks taken from timer
221 * softirq, e.g. We therefore emulate this by local_bh_disable() */
224 nf_reinject(entry
->skb
, entry
->info
, verdict
);
231 __enqueue_entry(struct nfqnl_instance
*queue
,
232 struct nfqnl_queue_entry
*entry
)
234 list_add(&entry
->list
, &queue
->queue_list
);
235 queue
->queue_total
++;
239 * Find and return a queued entry matched by cmpfn, or return the last
240 * entry if cmpfn is NULL.
242 static inline struct nfqnl_queue_entry
*
243 __find_entry(struct nfqnl_instance
*queue
, nfqnl_cmpfn cmpfn
,
248 list_for_each_prev(p
, &queue
->queue_list
) {
249 struct nfqnl_queue_entry
*entry
= (struct nfqnl_queue_entry
*)p
;
251 if (!cmpfn
|| cmpfn(entry
, data
))
258 __dequeue_entry(struct nfqnl_instance
*q
, struct nfqnl_queue_entry
*entry
)
260 list_del(&entry
->list
);
264 static inline struct nfqnl_queue_entry
*
265 __find_dequeue_entry(struct nfqnl_instance
*queue
,
266 nfqnl_cmpfn cmpfn
, unsigned long data
)
268 struct nfqnl_queue_entry
*entry
;
270 entry
= __find_entry(queue
, cmpfn
, data
);
274 __dequeue_entry(queue
, entry
);
280 __nfqnl_flush(struct nfqnl_instance
*queue
, int verdict
)
282 struct nfqnl_queue_entry
*entry
;
284 while ((entry
= __find_dequeue_entry(queue
, NULL
, 0)))
285 issue_verdict(entry
, verdict
);
289 __nfqnl_set_mode(struct nfqnl_instance
*queue
,
290 unsigned char mode
, unsigned int range
)
295 case NFQNL_COPY_NONE
:
296 case NFQNL_COPY_META
:
297 queue
->copy_mode
= mode
;
298 queue
->copy_range
= 0;
301 case NFQNL_COPY_PACKET
:
302 queue
->copy_mode
= mode
;
303 /* we're using struct nfattr which has 16bit nfa_len */
305 queue
->copy_range
= 0xffff;
307 queue
->copy_range
= range
;
317 static struct nfqnl_queue_entry
*
318 find_dequeue_entry(struct nfqnl_instance
*queue
,
319 nfqnl_cmpfn cmpfn
, unsigned long data
)
321 struct nfqnl_queue_entry
*entry
;
323 spin_lock_bh(&queue
->lock
);
324 entry
= __find_dequeue_entry(queue
, cmpfn
, data
);
325 spin_unlock_bh(&queue
->lock
);
331 nfqnl_flush(struct nfqnl_instance
*queue
, int verdict
)
333 spin_lock_bh(&queue
->lock
);
334 __nfqnl_flush(queue
, verdict
);
335 spin_unlock_bh(&queue
->lock
);
338 static struct sk_buff
*
339 nfqnl_build_packet_message(struct nfqnl_instance
*queue
,
340 struct nfqnl_queue_entry
*entry
, int *errp
)
342 unsigned char *old_tail
;
346 struct nfqnl_msg_packet_hdr pmsg
;
347 struct nlmsghdr
*nlh
;
348 struct nfgenmsg
*nfmsg
;
349 unsigned int tmp_uint
;
353 /* all macros expand to constant values at compile time */
354 size
= NLMSG_SPACE(sizeof(struct nfqnl_msg_packet_hdr
))
355 + NLMSG_SPACE(sizeof(u_int32_t
)) /* ifindex */
356 + NLMSG_SPACE(sizeof(u_int32_t
)) /* ifindex */
357 #ifdef CONFIG_BRIDGE_NETFILTER
358 + NLMSG_SPACE(sizeof(u_int32_t
)) /* ifindex */
359 + NLMSG_SPACE(sizeof(u_int32_t
)) /* ifindex */
361 + NLMSG_SPACE(sizeof(u_int32_t
)) /* mark */
362 + NLMSG_SPACE(sizeof(struct nfqnl_msg_packet_hw
))
363 + NLMSG_SPACE(sizeof(struct nfqnl_msg_packet_timestamp
));
365 spin_lock_bh(&queue
->lock
);
367 switch (queue
->copy_mode
) {
368 case NFQNL_COPY_META
:
369 case NFQNL_COPY_NONE
:
373 case NFQNL_COPY_PACKET
:
374 if (entry
->skb
->ip_summed
== CHECKSUM_HW
&&
375 (*errp
= skb_checksum_help(entry
->skb
,
376 entry
->info
->outdev
== NULL
))) {
377 spin_unlock_bh(&queue
->lock
);
380 if (queue
->copy_range
== 0
381 || queue
->copy_range
> entry
->skb
->len
)
382 data_len
= entry
->skb
->len
;
384 data_len
= queue
->copy_range
;
386 size
+= NLMSG_SPACE(data_len
);
391 spin_unlock_bh(&queue
->lock
);
395 spin_unlock_bh(&queue
->lock
);
397 skb
= alloc_skb(size
, GFP_ATOMIC
);
402 nlh
= NLMSG_PUT(skb
, 0, 0,
403 NFNL_SUBSYS_QUEUE
<< 8 | NFQNL_MSG_PACKET
,
404 sizeof(struct nfgenmsg
));
405 nfmsg
= NLMSG_DATA(nlh
);
406 nfmsg
->nfgen_family
= entry
->info
->pf
;
407 nfmsg
->version
= NFNETLINK_V0
;
408 nfmsg
->res_id
= htons(queue
->queue_num
);
410 pmsg
.packet_id
= htonl(entry
->id
);
411 pmsg
.hw_protocol
= htons(entry
->skb
->protocol
);
412 pmsg
.hook
= entry
->info
->hook
;
414 NFA_PUT(skb
, NFQA_PACKET_HDR
, sizeof(pmsg
), &pmsg
);
416 if (entry
->info
->indev
) {
417 tmp_uint
= htonl(entry
->info
->indev
->ifindex
);
418 #ifndef CONFIG_BRIDGE_NETFILTER
419 NFA_PUT(skb
, NFQA_IFINDEX_INDEV
, sizeof(tmp_uint
), &tmp_uint
);
421 if (entry
->info
->pf
== PF_BRIDGE
) {
422 /* Case 1: indev is physical input device, we need to
423 * look for bridge group (when called from
424 * netfilter_bridge) */
425 NFA_PUT(skb
, NFQA_IFINDEX_PHYSINDEV
, sizeof(tmp_uint
),
427 /* this is the bridge group "brX" */
428 tmp_uint
= htonl(entry
->info
->indev
->br_port
->br
->dev
->ifindex
);
429 NFA_PUT(skb
, NFQA_IFINDEX_INDEV
, sizeof(tmp_uint
),
432 /* Case 2: indev is bridge group, we need to look for
433 * physical device (when called from ipv4) */
434 NFA_PUT(skb
, NFQA_IFINDEX_INDEV
, sizeof(tmp_uint
),
436 if (entry
->skb
->nf_bridge
437 && entry
->skb
->nf_bridge
->physindev
) {
438 tmp_uint
= htonl(entry
->skb
->nf_bridge
->physindev
->ifindex
);
439 NFA_PUT(skb
, NFQA_IFINDEX_PHYSINDEV
,
440 sizeof(tmp_uint
), &tmp_uint
);
446 if (entry
->info
->outdev
) {
447 tmp_uint
= htonl(entry
->info
->outdev
->ifindex
);
448 #ifndef CONFIG_BRIDGE_NETFILTER
449 NFA_PUT(skb
, NFQA_IFINDEX_OUTDEV
, sizeof(tmp_uint
), &tmp_uint
);
451 if (entry
->info
->pf
== PF_BRIDGE
) {
452 /* Case 1: outdev is physical output device, we need to
453 * look for bridge group (when called from
454 * netfilter_bridge) */
455 NFA_PUT(skb
, NFQA_IFINDEX_PHYSOUTDEV
, sizeof(tmp_uint
),
457 /* this is the bridge group "brX" */
458 tmp_uint
= htonl(entry
->info
->outdev
->br_port
->br
->dev
->ifindex
);
459 NFA_PUT(skb
, NFQA_IFINDEX_OUTDEV
, sizeof(tmp_uint
),
462 /* Case 2: outdev is bridge group, we need to look for
463 * physical output device (when called from ipv4) */
464 NFA_PUT(skb
, NFQA_IFINDEX_OUTDEV
, sizeof(tmp_uint
),
466 if (entry
->skb
->nf_bridge
467 && entry
->skb
->nf_bridge
->physoutdev
) {
468 tmp_uint
= htonl(entry
->skb
->nf_bridge
->physoutdev
->ifindex
);
469 NFA_PUT(skb
, NFQA_IFINDEX_PHYSOUTDEV
,
470 sizeof(tmp_uint
), &tmp_uint
);
476 if (entry
->skb
->nfmark
) {
477 tmp_uint
= htonl(entry
->skb
->nfmark
);
478 NFA_PUT(skb
, NFQA_MARK
, sizeof(u_int32_t
), &tmp_uint
);
481 if (entry
->info
->indev
&& entry
->skb
->dev
482 && entry
->skb
->dev
->hard_header_parse
) {
483 struct nfqnl_msg_packet_hw phw
;
486 entry
->skb
->dev
->hard_header_parse(entry
->skb
,
488 phw
.hw_addrlen
= htons(phw
.hw_addrlen
);
489 NFA_PUT(skb
, NFQA_HWADDR
, sizeof(phw
), &phw
);
492 if (entry
->skb
->tstamp
.off_sec
) {
493 struct nfqnl_msg_packet_timestamp ts
;
495 ts
.sec
= cpu_to_be64(skb_tv_base
.tv_sec
+ entry
->skb
->tstamp
.off_sec
);
496 ts
.usec
= cpu_to_be64(skb_tv_base
.tv_usec
+ entry
->skb
->tstamp
.off_usec
);
498 NFA_PUT(skb
, NFQA_TIMESTAMP
, sizeof(ts
), &ts
);
503 int size
= NFA_LENGTH(data_len
);
505 if (skb_tailroom(skb
) < (int)NFA_SPACE(data_len
)) {
506 printk(KERN_WARNING
"nf_queue: no tailroom!\n");
510 nfa
= (struct nfattr
*)skb_put(skb
, NFA_ALIGN(size
));
511 nfa
->nfa_type
= NFQA_PAYLOAD
;
514 if (skb_copy_bits(entry
->skb
, 0, NFA_DATA(nfa
), data_len
))
518 nlh
->nlmsg_len
= skb
->tail
- old_tail
;
527 printk(KERN_ERR
"nf_queue: error creating packet message\n");
532 nfqnl_enqueue_packet(struct sk_buff
*skb
, struct nf_info
*info
,
533 unsigned int queuenum
, void *data
)
535 int status
= -EINVAL
;
536 struct sk_buff
*nskb
;
537 struct nfqnl_instance
*queue
;
538 struct nfqnl_queue_entry
*entry
;
542 queue
= instance_lookup_get(queuenum
);
544 QDEBUG("no queue instance matching\n");
548 if (queue
->copy_mode
== NFQNL_COPY_NONE
) {
549 QDEBUG("mode COPY_NONE, aborting\n");
554 entry
= kmalloc(sizeof(*entry
), GFP_ATOMIC
);
558 "nf_queue: OOM in nfqnl_enqueue_packet()\n");
565 entry
->id
= atomic_inc_return(&queue
->id_sequence
);
567 nskb
= nfqnl_build_packet_message(queue
, entry
, &status
);
571 spin_lock_bh(&queue
->lock
);
573 if (!queue
->peer_pid
)
574 goto err_out_free_nskb
;
576 if (queue
->queue_total
>= queue
->queue_maxlen
) {
577 queue
->queue_dropped
++;
580 printk(KERN_WARNING
"ip_queue: full at %d entries, "
581 "dropping packets(s). Dropped: %d\n",
582 queue
->queue_total
, queue
->queue_dropped
);
583 goto err_out_free_nskb
;
586 /* nfnetlink_unicast will either free the nskb or add it to a socket */
587 status
= nfnetlink_unicast(nskb
, queue
->peer_pid
, MSG_DONTWAIT
);
589 queue
->queue_user_dropped
++;
593 __enqueue_entry(queue
, entry
);
595 spin_unlock_bh(&queue
->lock
);
603 spin_unlock_bh(&queue
->lock
);
613 nfqnl_mangle(void *data
, int data_len
, struct nfqnl_queue_entry
*e
)
617 diff
= data_len
- e
->skb
->len
;
619 skb_trim(e
->skb
, data_len
);
621 if (data_len
> 0xFFFF)
623 if (diff
> skb_tailroom(e
->skb
)) {
624 struct sk_buff
*newskb
;
626 newskb
= skb_copy_expand(e
->skb
,
627 skb_headroom(e
->skb
),
630 if (newskb
== NULL
) {
631 printk(KERN_WARNING
"ip_queue: OOM "
632 "in mangle, dropping packet\n");
636 skb_set_owner_w(newskb
, e
->skb
->sk
);
640 skb_put(e
->skb
, diff
);
642 if (!skb_make_writable(&e
->skb
, data_len
))
644 memcpy(e
->skb
->data
, data
, data_len
);
645 e
->skb
->ip_summed
= CHECKSUM_NONE
;
650 id_cmp(struct nfqnl_queue_entry
*e
, unsigned long id
)
652 return (id
== e
->id
);
656 nfqnl_set_mode(struct nfqnl_instance
*queue
,
657 unsigned char mode
, unsigned int range
)
661 spin_lock_bh(&queue
->lock
);
662 status
= __nfqnl_set_mode(queue
, mode
, range
);
663 spin_unlock_bh(&queue
->lock
);
669 dev_cmp(struct nfqnl_queue_entry
*entry
, unsigned long ifindex
)
671 if (entry
->info
->indev
)
672 if (entry
->info
->indev
->ifindex
== ifindex
)
675 if (entry
->info
->outdev
)
676 if (entry
->info
->outdev
->ifindex
== ifindex
)
682 /* drop all packets with either indev or outdev == ifindex from all queue
685 nfqnl_dev_drop(int ifindex
)
689 QDEBUG("entering for ifindex %u\n", ifindex
);
691 /* this only looks like we have to hold the readlock for a way too long
692 * time, issue_verdict(), nf_reinject(), ... - but we always only
693 * issue NF_DROP, which is processed directly in nf_reinject() */
694 read_lock_bh(&instances_lock
);
696 for (i
= 0; i
< INSTANCE_BUCKETS
; i
++) {
697 struct hlist_node
*tmp
;
698 struct nfqnl_instance
*inst
;
699 struct hlist_head
*head
= &instance_table
[i
];
701 hlist_for_each_entry(inst
, tmp
, head
, hlist
) {
702 struct nfqnl_queue_entry
*entry
;
703 while ((entry
= find_dequeue_entry(inst
, dev_cmp
,
705 issue_verdict(entry
, NF_DROP
);
709 read_unlock_bh(&instances_lock
);
712 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
715 nfqnl_rcv_dev_event(struct notifier_block
*this,
716 unsigned long event
, void *ptr
)
718 struct net_device
*dev
= ptr
;
720 /* Drop any packets associated with the downed device */
721 if (event
== NETDEV_DOWN
)
722 nfqnl_dev_drop(dev
->ifindex
);
726 static struct notifier_block nfqnl_dev_notifier
= {
727 .notifier_call
= nfqnl_rcv_dev_event
,
731 nfqnl_rcv_nl_event(struct notifier_block
*this,
732 unsigned long event
, void *ptr
)
734 struct netlink_notify
*n
= ptr
;
736 if (event
== NETLINK_URELEASE
&&
737 n
->protocol
== NETLINK_NETFILTER
&& n
->pid
) {
740 /* destroy all instances for this pid */
741 write_lock_bh(&instances_lock
);
742 for (i
= 0; i
< INSTANCE_BUCKETS
; i
++) {
743 struct hlist_node
*tmp
, *t2
;
744 struct nfqnl_instance
*inst
;
745 struct hlist_head
*head
= &instance_table
[i
];
747 hlist_for_each_entry_safe(inst
, tmp
, t2
, head
, hlist
) {
748 if (n
->pid
== inst
->peer_pid
)
749 __instance_destroy(inst
);
752 write_unlock_bh(&instances_lock
);
757 static struct notifier_block nfqnl_rtnl_notifier
= {
758 .notifier_call
= nfqnl_rcv_nl_event
,
761 static const int nfqa_verdict_min
[NFQA_MAX
] = {
762 [NFQA_VERDICT_HDR
-1] = sizeof(struct nfqnl_msg_verdict_hdr
),
763 [NFQA_MARK
-1] = sizeof(u_int32_t
),
764 [NFQA_PAYLOAD
-1] = 0,
768 nfqnl_recv_verdict(struct sock
*ctnl
, struct sk_buff
*skb
,
769 struct nlmsghdr
*nlh
, struct nfattr
*nfqa
[], int *errp
)
771 struct nfgenmsg
*nfmsg
= NLMSG_DATA(nlh
);
772 u_int16_t queue_num
= ntohs(nfmsg
->res_id
);
774 struct nfqnl_msg_verdict_hdr
*vhdr
;
775 struct nfqnl_instance
*queue
;
776 unsigned int verdict
;
777 struct nfqnl_queue_entry
*entry
;
780 if (nfattr_bad_size(nfqa
, NFQA_MAX
, nfqa_verdict_min
)) {
781 QDEBUG("bad attribute size\n");
785 queue
= instance_lookup_get(queue_num
);
789 if (queue
->peer_pid
!= NETLINK_CB(skb
).pid
) {
794 if (!nfqa
[NFQA_VERDICT_HDR
-1]) {
799 vhdr
= NFA_DATA(nfqa
[NFQA_VERDICT_HDR
-1]);
800 verdict
= ntohl(vhdr
->verdict
);
802 if ((verdict
& NF_VERDICT_MASK
) > NF_MAX_VERDICT
) {
807 entry
= find_dequeue_entry(queue
, id_cmp
, ntohl(vhdr
->id
));
813 if (nfqa
[NFQA_PAYLOAD
-1]) {
814 if (nfqnl_mangle(NFA_DATA(nfqa
[NFQA_PAYLOAD
-1]),
815 NFA_PAYLOAD(nfqa
[NFQA_PAYLOAD
-1]), entry
) < 0)
819 if (nfqa
[NFQA_MARK
-1])
820 skb
->nfmark
= ntohl(*(u_int32_t
*)NFA_DATA(nfqa
[NFQA_MARK
-1]));
822 issue_verdict(entry
, verdict
);
832 nfqnl_recv_unsupp(struct sock
*ctnl
, struct sk_buff
*skb
,
833 struct nlmsghdr
*nlh
, struct nfattr
*nfqa
[], int *errp
)
838 static const int nfqa_cfg_min
[NFQA_CFG_MAX
] = {
839 [NFQA_CFG_CMD
-1] = sizeof(struct nfqnl_msg_config_cmd
),
840 [NFQA_CFG_PARAMS
-1] = sizeof(struct nfqnl_msg_config_params
),
843 static struct nf_queue_handler nfqh
= {
845 .outfn
= &nfqnl_enqueue_packet
,
849 nfqnl_recv_config(struct sock
*ctnl
, struct sk_buff
*skb
,
850 struct nlmsghdr
*nlh
, struct nfattr
*nfqa
[], int *errp
)
852 struct nfgenmsg
*nfmsg
= NLMSG_DATA(nlh
);
853 u_int16_t queue_num
= ntohs(nfmsg
->res_id
);
854 struct nfqnl_instance
*queue
;
857 QDEBUG("entering for msg %u\n", NFNL_MSG_TYPE(nlh
->nlmsg_type
));
859 if (nfattr_bad_size(nfqa
, NFQA_CFG_MAX
, nfqa_cfg_min
)) {
860 QDEBUG("bad attribute size\n");
864 queue
= instance_lookup_get(queue_num
);
865 if (nfqa
[NFQA_CFG_CMD
-1]) {
866 struct nfqnl_msg_config_cmd
*cmd
;
867 cmd
= NFA_DATA(nfqa
[NFQA_CFG_CMD
-1]);
868 QDEBUG("found CFG_CMD\n");
870 switch (cmd
->command
) {
871 case NFQNL_CFG_CMD_BIND
:
875 queue
= instance_create(queue_num
, NETLINK_CB(skb
).pid
);
879 case NFQNL_CFG_CMD_UNBIND
:
883 if (queue
->peer_pid
!= NETLINK_CB(skb
).pid
) {
888 instance_destroy(queue
);
890 case NFQNL_CFG_CMD_PF_BIND
:
891 QDEBUG("registering queue handler for pf=%u\n",
893 ret
= nf_register_queue_handler(ntohs(cmd
->pf
), &nfqh
);
895 case NFQNL_CFG_CMD_PF_UNBIND
:
896 QDEBUG("unregistering queue handler for pf=%u\n",
898 /* This is a bug and a feature. We can unregister
899 * other handlers(!) */
900 ret
= nf_unregister_queue_handler(ntohs(cmd
->pf
));
908 QDEBUG("no config command, and no instance ENOENT\n");
913 if (queue
->peer_pid
!= NETLINK_CB(skb
).pid
) {
914 QDEBUG("no config command, and wrong pid\n");
920 if (nfqa
[NFQA_CFG_PARAMS
-1]) {
921 struct nfqnl_msg_config_params
*params
;
922 params
= NFA_DATA(nfqa
[NFQA_CFG_PARAMS
-1]);
924 nfqnl_set_mode(queue
, params
->copy_mode
,
925 ntohl(params
->copy_range
));
933 static struct nfnl_callback nfqnl_cb
[NFQNL_MSG_MAX
] = {
934 [NFQNL_MSG_PACKET
] = { .call
= nfqnl_recv_unsupp
,
935 .attr_count
= NFQA_MAX
,
936 .cap_required
= CAP_NET_ADMIN
},
937 [NFQNL_MSG_VERDICT
] = { .call
= nfqnl_recv_verdict
,
938 .attr_count
= NFQA_MAX
,
939 .cap_required
= CAP_NET_ADMIN
},
940 [NFQNL_MSG_CONFIG
] = { .call
= nfqnl_recv_config
,
941 .attr_count
= NFQA_CFG_MAX
,
942 .cap_required
= CAP_NET_ADMIN
},
945 static struct nfnetlink_subsystem nfqnl_subsys
= {
947 .subsys_id
= NFNL_SUBSYS_QUEUE
,
948 .cb_count
= NFQNL_MSG_MAX
,
952 #ifdef CONFIG_PROC_FS
957 static struct hlist_node
*get_first(struct seq_file
*seq
)
959 struct iter_state
*st
= seq
->private;
964 for (st
->bucket
= 0; st
->bucket
< INSTANCE_BUCKETS
; st
->bucket
++) {
965 if (!hlist_empty(&instance_table
[st
->bucket
]))
966 return instance_table
[st
->bucket
].first
;
971 static struct hlist_node
*get_next(struct seq_file
*seq
, struct hlist_node
*h
)
973 struct iter_state
*st
= seq
->private;
977 if (++st
->bucket
>= INSTANCE_BUCKETS
)
980 h
= instance_table
[st
->bucket
].first
;
985 static struct hlist_node
*get_idx(struct seq_file
*seq
, loff_t pos
)
987 struct hlist_node
*head
;
988 head
= get_first(seq
);
991 while (pos
&& (head
= get_next(seq
, head
)))
993 return pos
? NULL
: head
;
996 static void *seq_start(struct seq_file
*seq
, loff_t
*pos
)
998 read_lock_bh(&instances_lock
);
999 return get_idx(seq
, *pos
);
1002 static void *seq_next(struct seq_file
*s
, void *v
, loff_t
*pos
)
1005 return get_next(s
, v
);
1008 static void seq_stop(struct seq_file
*s
, void *v
)
1010 read_unlock_bh(&instances_lock
);
1013 static int seq_show(struct seq_file
*s
, void *v
)
1015 const struct nfqnl_instance
*inst
= v
;
1017 return seq_printf(s
, "%5d %6d %5d %1d %5d %5d %5d %8d %2d\n",
1019 inst
->peer_pid
, inst
->queue_total
,
1020 inst
->copy_mode
, inst
->copy_range
,
1021 inst
->queue_dropped
, inst
->queue_user_dropped
,
1022 atomic_read(&inst
->id_sequence
),
1023 atomic_read(&inst
->use
));
1026 static struct seq_operations nfqnl_seq_ops
= {
1033 static int nfqnl_open(struct inode
*inode
, struct file
*file
)
1035 struct seq_file
*seq
;
1036 struct iter_state
*is
;
1039 is
= kmalloc(sizeof(*is
), GFP_KERNEL
);
1042 memset(is
, 0, sizeof(*is
));
1043 ret
= seq_open(file
, &nfqnl_seq_ops
);
1046 seq
= file
->private_data
;
1054 static struct file_operations nfqnl_file_ops
= {
1055 .owner
= THIS_MODULE
,
1058 .llseek
= seq_lseek
,
1059 .release
= seq_release_private
,
1062 #endif /* PROC_FS */
1065 init_or_cleanup(int init
)
1067 int i
, status
= -ENOMEM
;
1068 #ifdef CONFIG_PROC_FS
1069 struct proc_dir_entry
*proc_nfqueue
;
1075 for (i
= 0; i
< INSTANCE_BUCKETS
; i
++)
1076 INIT_HLIST_HEAD(&instance_table
[i
]);
1078 netlink_register_notifier(&nfqnl_rtnl_notifier
);
1079 status
= nfnetlink_subsys_register(&nfqnl_subsys
);
1081 printk(KERN_ERR
"nf_queue: failed to create netlink socket\n");
1082 goto cleanup_netlink_notifier
;
1085 #ifdef CONFIG_PROC_FS
1086 proc_nfqueue
= create_proc_entry("nfnetlink_queue", 0440,
1087 proc_net_netfilter
);
1089 goto cleanup_subsys
;
1090 proc_nfqueue
->proc_fops
= &nfqnl_file_ops
;
1093 register_netdevice_notifier(&nfqnl_dev_notifier
);
1098 nf_unregister_queue_handlers(&nfqh
);
1099 unregister_netdevice_notifier(&nfqnl_dev_notifier
);
1100 #ifdef CONFIG_PROC_FS
1101 remove_proc_entry("nfnetlink_queue", proc_net_netfilter
);
1104 nfnetlink_subsys_unregister(&nfqnl_subsys
);
1105 cleanup_netlink_notifier
:
1106 netlink_unregister_notifier(&nfqnl_rtnl_notifier
);
1110 static int __init
init(void)
1113 return init_or_cleanup(1);
1116 static void __exit
fini(void)
1121 MODULE_DESCRIPTION("netfilter packet queue handler");
1122 MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
1123 MODULE_LICENSE("GPL");
1124 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_QUEUE
);