2 * This is a module which is used for queueing IPv6 packets and
3 * communicating with userspace via netlink.
5 * (C) 2001 Fernando Anton, this code is GPL.
6 * IPv64 Project - Work based in IPv64 draft by Arturo Azcorra.
7 * Universidad Carlos III de Madrid - Leganes (Madrid) - Spain
8 * Universidad Politecnica de Alcala de Henares - Alcala de H. (Madrid) - Spain
9 * email: fanton@it.uc3m.es
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.
15 #include <linux/module.h>
16 #include <linux/skbuff.h>
17 #include <linux/init.h>
18 #include <linux/ipv6.h>
19 #include <linux/notifier.h>
20 #include <linux/netdevice.h>
21 #include <linux/netfilter.h>
22 #include <linux/netlink.h>
23 #include <linux/spinlock.h>
24 #include <linux/sysctl.h>
25 #include <linux/proc_fs.h>
26 #include <linux/mutex.h>
29 #include <net/ip6_route.h>
30 #include <linux/netfilter_ipv4/ip_queue.h>
31 #include <linux/netfilter_ipv4/ip_tables.h>
32 #include <linux/netfilter_ipv6/ip6_tables.h>
34 #define IPQ_QMAX_DEFAULT 1024
35 #define IPQ_PROC_FS_NAME "ip6_queue"
36 #define NET_IPQ_QMAX 2088
37 #define NET_IPQ_QMAX_NAME "ip6_queue_maxlen"
39 struct ipq_queue_entry
{
40 struct list_head list
;
45 typedef int (*ipq_cmpfn
)(struct ipq_queue_entry
*, unsigned long);
47 static unsigned char copy_mode __read_mostly
= IPQ_COPY_NONE
;
48 static unsigned int queue_maxlen __read_mostly
= IPQ_QMAX_DEFAULT
;
49 static DEFINE_RWLOCK(queue_lock
);
50 static int peer_pid __read_mostly
;
51 static unsigned int copy_range __read_mostly
;
52 static unsigned int queue_total
;
53 static unsigned int queue_dropped
= 0;
54 static unsigned int queue_user_dropped
= 0;
55 static struct sock
*ipqnl __read_mostly
;
56 static LIST_HEAD(queue_list
);
57 static DEFINE_MUTEX(ipqnl_mutex
);
60 ipq_issue_verdict(struct ipq_queue_entry
*entry
, int verdict
)
63 nf_reinject(entry
->skb
, entry
->info
, verdict
);
69 __ipq_enqueue_entry(struct ipq_queue_entry
*entry
)
71 list_add(&entry
->list
, &queue_list
);
76 * Find and return a queued entry matched by cmpfn, or return the last
77 * entry if cmpfn is NULL.
79 static inline struct ipq_queue_entry
*
80 __ipq_find_entry(ipq_cmpfn cmpfn
, unsigned long data
)
84 list_for_each_prev(p
, &queue_list
) {
85 struct ipq_queue_entry
*entry
= (struct ipq_queue_entry
*)p
;
87 if (!cmpfn
|| cmpfn(entry
, data
))
94 __ipq_dequeue_entry(struct ipq_queue_entry
*entry
)
96 list_del(&entry
->list
);
100 static inline struct ipq_queue_entry
*
101 __ipq_find_dequeue_entry(ipq_cmpfn cmpfn
, unsigned long data
)
103 struct ipq_queue_entry
*entry
;
105 entry
= __ipq_find_entry(cmpfn
, data
);
109 __ipq_dequeue_entry(entry
);
115 __ipq_flush(int verdict
)
117 struct ipq_queue_entry
*entry
;
119 while ((entry
= __ipq_find_dequeue_entry(NULL
, 0)))
120 ipq_issue_verdict(entry
, verdict
);
124 __ipq_set_mode(unsigned char mode
, unsigned int range
)
135 case IPQ_COPY_PACKET
:
138 if (copy_range
> 0xFFFF)
153 net_disable_timestamp();
154 __ipq_set_mode(IPQ_COPY_NONE
, 0);
155 __ipq_flush(NF_DROP
);
158 static struct ipq_queue_entry
*
159 ipq_find_dequeue_entry(ipq_cmpfn cmpfn
, unsigned long data
)
161 struct ipq_queue_entry
*entry
;
163 write_lock_bh(&queue_lock
);
164 entry
= __ipq_find_dequeue_entry(cmpfn
, data
);
165 write_unlock_bh(&queue_lock
);
170 ipq_flush(int verdict
)
172 write_lock_bh(&queue_lock
);
173 __ipq_flush(verdict
);
174 write_unlock_bh(&queue_lock
);
177 static struct sk_buff
*
178 ipq_build_packet_message(struct ipq_queue_entry
*entry
, int *errp
)
180 sk_buff_data_t old_tail
;
184 struct ipq_packet_msg
*pmsg
;
185 struct nlmsghdr
*nlh
;
188 read_lock_bh(&queue_lock
);
193 size
= NLMSG_SPACE(sizeof(*pmsg
));
197 case IPQ_COPY_PACKET
:
198 if ((entry
->skb
->ip_summed
== CHECKSUM_PARTIAL
||
199 entry
->skb
->ip_summed
== CHECKSUM_COMPLETE
) &&
200 (*errp
= skb_checksum_help(entry
->skb
))) {
201 read_unlock_bh(&queue_lock
);
204 if (copy_range
== 0 || copy_range
> entry
->skb
->len
)
205 data_len
= entry
->skb
->len
;
207 data_len
= copy_range
;
209 size
= NLMSG_SPACE(sizeof(*pmsg
) + data_len
);
214 read_unlock_bh(&queue_lock
);
218 read_unlock_bh(&queue_lock
);
220 skb
= alloc_skb(size
, GFP_ATOMIC
);
224 old_tail
= skb
->tail
;
225 nlh
= NLMSG_PUT(skb
, 0, 0, IPQM_PACKET
, size
- sizeof(*nlh
));
226 pmsg
= NLMSG_DATA(nlh
);
227 memset(pmsg
, 0, sizeof(*pmsg
));
229 pmsg
->packet_id
= (unsigned long )entry
;
230 pmsg
->data_len
= data_len
;
231 tv
= ktime_to_timeval(entry
->skb
->tstamp
);
232 pmsg
->timestamp_sec
= tv
.tv_sec
;
233 pmsg
->timestamp_usec
= tv
.tv_usec
;
234 pmsg
->mark
= entry
->skb
->mark
;
235 pmsg
->hook
= entry
->info
->hook
;
236 pmsg
->hw_protocol
= entry
->skb
->protocol
;
238 if (entry
->info
->indev
)
239 strcpy(pmsg
->indev_name
, entry
->info
->indev
->name
);
241 pmsg
->indev_name
[0] = '\0';
243 if (entry
->info
->outdev
)
244 strcpy(pmsg
->outdev_name
, entry
->info
->outdev
->name
);
246 pmsg
->outdev_name
[0] = '\0';
248 if (entry
->info
->indev
&& entry
->skb
->dev
) {
249 pmsg
->hw_type
= entry
->skb
->dev
->type
;
250 if (entry
->skb
->dev
->hard_header_parse
)
252 entry
->skb
->dev
->hard_header_parse(entry
->skb
,
257 if (skb_copy_bits(entry
->skb
, 0, pmsg
->payload
, data_len
))
260 nlh
->nlmsg_len
= skb
->tail
- old_tail
;
267 printk(KERN_ERR
"ip6_queue: error creating packet message\n");
272 ipq_enqueue_packet(struct sk_buff
*skb
, struct nf_info
*info
,
273 unsigned int queuenum
, void *data
)
275 int status
= -EINVAL
;
276 struct sk_buff
*nskb
;
277 struct ipq_queue_entry
*entry
;
279 if (copy_mode
== IPQ_COPY_NONE
)
282 entry
= kmalloc(sizeof(*entry
), GFP_ATOMIC
);
284 printk(KERN_ERR
"ip6_queue: OOM in ipq_enqueue_packet()\n");
291 nskb
= ipq_build_packet_message(entry
, &status
);
295 write_lock_bh(&queue_lock
);
298 goto err_out_free_nskb
;
300 if (queue_total
>= queue_maxlen
) {
304 printk (KERN_WARNING
"ip6_queue: fill at %d entries, "
305 "dropping packet(s). Dropped: %d\n", queue_total
,
307 goto err_out_free_nskb
;
310 /* netlink_unicast will either free the nskb or attach it to a socket */
311 status
= netlink_unicast(ipqnl
, nskb
, peer_pid
, MSG_DONTWAIT
);
313 queue_user_dropped
++;
317 __ipq_enqueue_entry(entry
);
319 write_unlock_bh(&queue_lock
);
326 write_unlock_bh(&queue_lock
);
334 ipq_mangle_ipv6(ipq_verdict_msg_t
*v
, struct ipq_queue_entry
*e
)
337 struct ipv6hdr
*user_iph
= (struct ipv6hdr
*)v
->payload
;
339 if (v
->data_len
< sizeof(*user_iph
))
341 diff
= v
->data_len
- e
->skb
->len
;
343 if (pskb_trim(e
->skb
, v
->data_len
))
345 } else if (diff
> 0) {
346 if (v
->data_len
> 0xFFFF)
348 if (diff
> skb_tailroom(e
->skb
)) {
349 struct sk_buff
*newskb
;
351 newskb
= skb_copy_expand(e
->skb
,
352 skb_headroom(e
->skb
),
355 if (newskb
== NULL
) {
356 printk(KERN_WARNING
"ip6_queue: OOM "
357 "in mangle, dropping packet\n");
361 skb_set_owner_w(newskb
, e
->skb
->sk
);
365 skb_put(e
->skb
, diff
);
367 if (!skb_make_writable(&e
->skb
, v
->data_len
))
369 skb_copy_to_linear_data(e
->skb
, v
->payload
, v
->data_len
);
370 e
->skb
->ip_summed
= CHECKSUM_NONE
;
376 id_cmp(struct ipq_queue_entry
*e
, unsigned long id
)
378 return (id
== (unsigned long )e
);
382 ipq_set_verdict(struct ipq_verdict_msg
*vmsg
, unsigned int len
)
384 struct ipq_queue_entry
*entry
;
386 if (vmsg
->value
> NF_MAX_VERDICT
)
389 entry
= ipq_find_dequeue_entry(id_cmp
, vmsg
->id
);
393 int verdict
= vmsg
->value
;
395 if (vmsg
->data_len
&& vmsg
->data_len
== len
)
396 if (ipq_mangle_ipv6(vmsg
, entry
) < 0)
399 ipq_issue_verdict(entry
, verdict
);
405 ipq_set_mode(unsigned char mode
, unsigned int range
)
409 write_lock_bh(&queue_lock
);
410 status
= __ipq_set_mode(mode
, range
);
411 write_unlock_bh(&queue_lock
);
416 ipq_receive_peer(struct ipq_peer_msg
*pmsg
,
417 unsigned char type
, unsigned int len
)
421 if (len
< sizeof(*pmsg
))
426 status
= ipq_set_mode(pmsg
->msg
.mode
.value
,
427 pmsg
->msg
.mode
.range
);
431 if (pmsg
->msg
.verdict
.value
> NF_MAX_VERDICT
)
434 status
= ipq_set_verdict(&pmsg
->msg
.verdict
,
435 len
- sizeof(*pmsg
));
444 dev_cmp(struct ipq_queue_entry
*entry
, unsigned long ifindex
)
446 if (entry
->info
->indev
)
447 if (entry
->info
->indev
->ifindex
== ifindex
)
450 if (entry
->info
->outdev
)
451 if (entry
->info
->outdev
->ifindex
== ifindex
)
458 ipq_dev_drop(int ifindex
)
460 struct ipq_queue_entry
*entry
;
462 while ((entry
= ipq_find_dequeue_entry(dev_cmp
, ifindex
)) != NULL
)
463 ipq_issue_verdict(entry
, NF_DROP
);
466 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
469 ipq_rcv_skb(struct sk_buff
*skb
)
471 int status
, type
, pid
, flags
, nlmsglen
, skblen
;
472 struct nlmsghdr
*nlh
;
475 if (skblen
< sizeof(*nlh
))
478 nlh
= nlmsg_hdr(skb
);
479 nlmsglen
= nlh
->nlmsg_len
;
480 if (nlmsglen
< sizeof(*nlh
) || skblen
< nlmsglen
)
483 pid
= nlh
->nlmsg_pid
;
484 flags
= nlh
->nlmsg_flags
;
486 if(pid
<= 0 || !(flags
& NLM_F_REQUEST
) || flags
& NLM_F_MULTI
)
487 RCV_SKB_FAIL(-EINVAL
);
489 if (flags
& MSG_TRUNC
)
490 RCV_SKB_FAIL(-ECOMM
);
492 type
= nlh
->nlmsg_type
;
493 if (type
< NLMSG_NOOP
|| type
>= IPQM_MAX
)
494 RCV_SKB_FAIL(-EINVAL
);
496 if (type
<= IPQM_BASE
)
499 if (security_netlink_recv(skb
, CAP_NET_ADMIN
))
500 RCV_SKB_FAIL(-EPERM
);
502 write_lock_bh(&queue_lock
);
505 if (peer_pid
!= pid
) {
506 write_unlock_bh(&queue_lock
);
507 RCV_SKB_FAIL(-EBUSY
);
510 net_enable_timestamp();
514 write_unlock_bh(&queue_lock
);
516 status
= ipq_receive_peer(NLMSG_DATA(nlh
), type
,
517 nlmsglen
- NLMSG_LENGTH(0));
519 RCV_SKB_FAIL(status
);
521 if (flags
& NLM_F_ACK
)
522 netlink_ack(skb
, nlh
, 0);
527 ipq_rcv_sk(struct sock
*sk
, int len
)
532 mutex_lock(&ipqnl_mutex
);
534 for (qlen
= skb_queue_len(&sk
->sk_receive_queue
); qlen
; qlen
--) {
535 skb
= skb_dequeue(&sk
->sk_receive_queue
);
540 mutex_unlock(&ipqnl_mutex
);
544 ipq_rcv_dev_event(struct notifier_block
*this,
545 unsigned long event
, void *ptr
)
547 struct net_device
*dev
= ptr
;
549 /* Drop any packets associated with the downed device */
550 if (event
== NETDEV_DOWN
)
551 ipq_dev_drop(dev
->ifindex
);
555 static struct notifier_block ipq_dev_notifier
= {
556 .notifier_call
= ipq_rcv_dev_event
,
560 ipq_rcv_nl_event(struct notifier_block
*this,
561 unsigned long event
, void *ptr
)
563 struct netlink_notify
*n
= ptr
;
565 if (event
== NETLINK_URELEASE
&&
566 n
->protocol
== NETLINK_IP6_FW
&& n
->pid
) {
567 write_lock_bh(&queue_lock
);
568 if (n
->pid
== peer_pid
)
570 write_unlock_bh(&queue_lock
);
575 static struct notifier_block ipq_nl_notifier
= {
576 .notifier_call
= ipq_rcv_nl_event
,
579 static struct ctl_table_header
*ipq_sysctl_header
;
581 static ctl_table ipq_table
[] = {
583 .ctl_name
= NET_IPQ_QMAX
,
584 .procname
= NET_IPQ_QMAX_NAME
,
585 .data
= &queue_maxlen
,
586 .maxlen
= sizeof(queue_maxlen
),
588 .proc_handler
= proc_dointvec
593 static ctl_table ipq_dir_table
[] = {
595 .ctl_name
= NET_IPV6
,
603 static ctl_table ipq_root_table
[] = {
608 .child
= ipq_dir_table
613 #ifdef CONFIG_PROC_FS
615 ipq_get_info(char *buffer
, char **start
, off_t offset
, int length
)
619 read_lock_bh(&queue_lock
);
621 len
= sprintf(buffer
,
625 "Queue length : %u\n"
626 "Queue max. length : %u\n"
627 "Queue dropped : %u\n"
628 "Netfilter dropped : %u\n",
637 read_unlock_bh(&queue_lock
);
639 *start
= buffer
+ offset
;
647 #endif /* CONFIG_PROC_FS */
649 static struct nf_queue_handler nfqh
= {
651 .outfn
= &ipq_enqueue_packet
,
654 static int __init
ip6_queue_init(void)
656 int status
= -ENOMEM
;
657 struct proc_dir_entry
*proc
;
659 netlink_register_notifier(&ipq_nl_notifier
);
660 ipqnl
= netlink_kernel_create(NETLINK_IP6_FW
, 0, ipq_rcv_sk
, NULL
,
663 printk(KERN_ERR
"ip6_queue: failed to create netlink socket\n");
664 goto cleanup_netlink_notifier
;
667 proc
= proc_net_create(IPQ_PROC_FS_NAME
, 0, ipq_get_info
);
669 proc
->owner
= THIS_MODULE
;
671 printk(KERN_ERR
"ip6_queue: failed to create proc entry\n");
675 register_netdevice_notifier(&ipq_dev_notifier
);
676 ipq_sysctl_header
= register_sysctl_table(ipq_root_table
);
678 status
= nf_register_queue_handler(PF_INET6
, &nfqh
);
680 printk(KERN_ERR
"ip6_queue: failed to register queue handler\n");
686 unregister_sysctl_table(ipq_sysctl_header
);
687 unregister_netdevice_notifier(&ipq_dev_notifier
);
688 proc_net_remove(IPQ_PROC_FS_NAME
);
691 sock_release(ipqnl
->sk_socket
);
692 mutex_lock(&ipqnl_mutex
);
693 mutex_unlock(&ipqnl_mutex
);
695 cleanup_netlink_notifier
:
696 netlink_unregister_notifier(&ipq_nl_notifier
);
700 static void __exit
ip6_queue_fini(void)
702 nf_unregister_queue_handlers(&nfqh
);
706 unregister_sysctl_table(ipq_sysctl_header
);
707 unregister_netdevice_notifier(&ipq_dev_notifier
);
708 proc_net_remove(IPQ_PROC_FS_NAME
);
710 sock_release(ipqnl
->sk_socket
);
711 mutex_lock(&ipqnl_mutex
);
712 mutex_unlock(&ipqnl_mutex
);
714 netlink_unregister_notifier(&ipq_nl_notifier
);
717 MODULE_DESCRIPTION("IPv6 packet queue handler");
718 MODULE_LICENSE("GPL");
720 module_init(ip6_queue_init
);
721 module_exit(ip6_queue_fini
);