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/seq_file.h>
27 #include <linux/mutex.h>
28 #include <net/net_namespace.h>
31 #include <net/ip6_route.h>
32 #include <linux/netfilter_ipv4/ip_queue.h>
33 #include <linux/netfilter_ipv4/ip_tables.h>
34 #include <linux/netfilter_ipv6/ip6_tables.h>
36 #define IPQ_QMAX_DEFAULT 1024
37 #define IPQ_PROC_FS_NAME "ip6_queue"
38 #define NET_IPQ_QMAX 2088
39 #define NET_IPQ_QMAX_NAME "ip6_queue_maxlen"
41 struct ipq_queue_entry
{
42 struct list_head list
;
47 typedef int (*ipq_cmpfn
)(struct ipq_queue_entry
*, unsigned long);
49 static unsigned char copy_mode __read_mostly
= IPQ_COPY_NONE
;
50 static unsigned int queue_maxlen __read_mostly
= IPQ_QMAX_DEFAULT
;
51 static DEFINE_RWLOCK(queue_lock
);
52 static int peer_pid __read_mostly
;
53 static unsigned int copy_range __read_mostly
;
54 static unsigned int queue_total
;
55 static unsigned int queue_dropped
= 0;
56 static unsigned int queue_user_dropped
= 0;
57 static struct sock
*ipqnl __read_mostly
;
58 static LIST_HEAD(queue_list
);
59 static DEFINE_MUTEX(ipqnl_mutex
);
62 ipq_issue_verdict(struct ipq_queue_entry
*entry
, int verdict
)
65 nf_reinject(entry
->skb
, entry
->info
, verdict
);
71 __ipq_enqueue_entry(struct ipq_queue_entry
*entry
)
73 list_add(&entry
->list
, &queue_list
);
78 * Find and return a queued entry matched by cmpfn, or return the last
79 * entry if cmpfn is NULL.
81 static inline struct ipq_queue_entry
*
82 __ipq_find_entry(ipq_cmpfn cmpfn
, unsigned long data
)
86 list_for_each_prev(p
, &queue_list
) {
87 struct ipq_queue_entry
*entry
= (struct ipq_queue_entry
*)p
;
89 if (!cmpfn
|| cmpfn(entry
, data
))
96 __ipq_dequeue_entry(struct ipq_queue_entry
*entry
)
98 list_del(&entry
->list
);
102 static inline struct ipq_queue_entry
*
103 __ipq_find_dequeue_entry(ipq_cmpfn cmpfn
, unsigned long data
)
105 struct ipq_queue_entry
*entry
;
107 entry
= __ipq_find_entry(cmpfn
, data
);
111 __ipq_dequeue_entry(entry
);
117 __ipq_flush(int verdict
)
119 struct ipq_queue_entry
*entry
;
121 while ((entry
= __ipq_find_dequeue_entry(NULL
, 0)))
122 ipq_issue_verdict(entry
, verdict
);
126 __ipq_set_mode(unsigned char mode
, unsigned int range
)
137 case IPQ_COPY_PACKET
:
140 if (copy_range
> 0xFFFF)
155 net_disable_timestamp();
156 __ipq_set_mode(IPQ_COPY_NONE
, 0);
157 __ipq_flush(NF_DROP
);
160 static struct ipq_queue_entry
*
161 ipq_find_dequeue_entry(ipq_cmpfn cmpfn
, unsigned long data
)
163 struct ipq_queue_entry
*entry
;
165 write_lock_bh(&queue_lock
);
166 entry
= __ipq_find_dequeue_entry(cmpfn
, data
);
167 write_unlock_bh(&queue_lock
);
172 ipq_flush(int verdict
)
174 write_lock_bh(&queue_lock
);
175 __ipq_flush(verdict
);
176 write_unlock_bh(&queue_lock
);
179 static struct sk_buff
*
180 ipq_build_packet_message(struct ipq_queue_entry
*entry
, int *errp
)
182 sk_buff_data_t old_tail
;
186 struct ipq_packet_msg
*pmsg
;
187 struct nlmsghdr
*nlh
;
190 read_lock_bh(&queue_lock
);
195 size
= NLMSG_SPACE(sizeof(*pmsg
));
199 case IPQ_COPY_PACKET
:
200 if ((entry
->skb
->ip_summed
== CHECKSUM_PARTIAL
||
201 entry
->skb
->ip_summed
== CHECKSUM_COMPLETE
) &&
202 (*errp
= skb_checksum_help(entry
->skb
))) {
203 read_unlock_bh(&queue_lock
);
206 if (copy_range
== 0 || copy_range
> entry
->skb
->len
)
207 data_len
= entry
->skb
->len
;
209 data_len
= copy_range
;
211 size
= NLMSG_SPACE(sizeof(*pmsg
) + data_len
);
216 read_unlock_bh(&queue_lock
);
220 read_unlock_bh(&queue_lock
);
222 skb
= alloc_skb(size
, GFP_ATOMIC
);
226 old_tail
= skb
->tail
;
227 nlh
= NLMSG_PUT(skb
, 0, 0, IPQM_PACKET
, size
- sizeof(*nlh
));
228 pmsg
= NLMSG_DATA(nlh
);
229 memset(pmsg
, 0, sizeof(*pmsg
));
231 pmsg
->packet_id
= (unsigned long )entry
;
232 pmsg
->data_len
= data_len
;
233 tv
= ktime_to_timeval(entry
->skb
->tstamp
);
234 pmsg
->timestamp_sec
= tv
.tv_sec
;
235 pmsg
->timestamp_usec
= tv
.tv_usec
;
236 pmsg
->mark
= entry
->skb
->mark
;
237 pmsg
->hook
= entry
->info
->hook
;
238 pmsg
->hw_protocol
= entry
->skb
->protocol
;
240 if (entry
->info
->indev
)
241 strcpy(pmsg
->indev_name
, entry
->info
->indev
->name
);
243 pmsg
->indev_name
[0] = '\0';
245 if (entry
->info
->outdev
)
246 strcpy(pmsg
->outdev_name
, entry
->info
->outdev
->name
);
248 pmsg
->outdev_name
[0] = '\0';
250 if (entry
->info
->indev
&& entry
->skb
->dev
) {
251 pmsg
->hw_type
= entry
->skb
->dev
->type
;
252 pmsg
->hw_addrlen
= dev_parse_header(entry
->skb
, pmsg
->hw_addr
);
256 if (skb_copy_bits(entry
->skb
, 0, pmsg
->payload
, data_len
))
259 nlh
->nlmsg_len
= skb
->tail
- old_tail
;
266 printk(KERN_ERR
"ip6_queue: error creating packet message\n");
271 ipq_enqueue_packet(struct sk_buff
*skb
, struct nf_info
*info
,
272 unsigned int queuenum
, void *data
)
274 int status
= -EINVAL
;
275 struct sk_buff
*nskb
;
276 struct ipq_queue_entry
*entry
;
278 if (copy_mode
== IPQ_COPY_NONE
)
281 entry
= kmalloc(sizeof(*entry
), GFP_ATOMIC
);
283 printk(KERN_ERR
"ip6_queue: OOM in ipq_enqueue_packet()\n");
290 nskb
= ipq_build_packet_message(entry
, &status
);
294 write_lock_bh(&queue_lock
);
297 goto err_out_free_nskb
;
299 if (queue_total
>= queue_maxlen
) {
303 printk (KERN_WARNING
"ip6_queue: fill at %d entries, "
304 "dropping packet(s). Dropped: %d\n", queue_total
,
306 goto err_out_free_nskb
;
309 /* netlink_unicast will either free the nskb or attach it to a socket */
310 status
= netlink_unicast(ipqnl
, nskb
, peer_pid
, MSG_DONTWAIT
);
312 queue_user_dropped
++;
316 __ipq_enqueue_entry(entry
);
318 write_unlock_bh(&queue_lock
);
325 write_unlock_bh(&queue_lock
);
333 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 err
= pskb_expand_head(e
->skb
, 0,
350 diff
- skb_tailroom(e
->skb
),
353 printk(KERN_WARNING
"ip6_queue: OOM "
354 "in mangle, dropping packet\n");
358 skb_put(e
->skb
, diff
);
360 if (!skb_make_writable(e
->skb
, v
->data_len
))
362 skb_copy_to_linear_data(e
->skb
, v
->payload
, v
->data_len
);
363 e
->skb
->ip_summed
= CHECKSUM_NONE
;
369 id_cmp(struct ipq_queue_entry
*e
, unsigned long id
)
371 return (id
== (unsigned long )e
);
375 ipq_set_verdict(struct ipq_verdict_msg
*vmsg
, unsigned int len
)
377 struct ipq_queue_entry
*entry
;
379 if (vmsg
->value
> NF_MAX_VERDICT
)
382 entry
= ipq_find_dequeue_entry(id_cmp
, vmsg
->id
);
386 int verdict
= vmsg
->value
;
388 if (vmsg
->data_len
&& vmsg
->data_len
== len
)
389 if (ipq_mangle_ipv6(vmsg
, entry
) < 0)
392 ipq_issue_verdict(entry
, verdict
);
398 ipq_set_mode(unsigned char mode
, unsigned int range
)
402 write_lock_bh(&queue_lock
);
403 status
= __ipq_set_mode(mode
, range
);
404 write_unlock_bh(&queue_lock
);
409 ipq_receive_peer(struct ipq_peer_msg
*pmsg
,
410 unsigned char type
, unsigned int len
)
414 if (len
< sizeof(*pmsg
))
419 status
= ipq_set_mode(pmsg
->msg
.mode
.value
,
420 pmsg
->msg
.mode
.range
);
424 if (pmsg
->msg
.verdict
.value
> NF_MAX_VERDICT
)
427 status
= ipq_set_verdict(&pmsg
->msg
.verdict
,
428 len
- sizeof(*pmsg
));
437 dev_cmp(struct ipq_queue_entry
*entry
, unsigned long ifindex
)
439 if (entry
->info
->indev
)
440 if (entry
->info
->indev
->ifindex
== ifindex
)
443 if (entry
->info
->outdev
)
444 if (entry
->info
->outdev
->ifindex
== ifindex
)
451 ipq_dev_drop(int ifindex
)
453 struct ipq_queue_entry
*entry
;
455 while ((entry
= ipq_find_dequeue_entry(dev_cmp
, ifindex
)) != NULL
)
456 ipq_issue_verdict(entry
, NF_DROP
);
459 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
462 __ipq_rcv_skb(struct sk_buff
*skb
)
464 int status
, type
, pid
, flags
, nlmsglen
, skblen
;
465 struct nlmsghdr
*nlh
;
468 if (skblen
< sizeof(*nlh
))
471 nlh
= nlmsg_hdr(skb
);
472 nlmsglen
= nlh
->nlmsg_len
;
473 if (nlmsglen
< sizeof(*nlh
) || skblen
< nlmsglen
)
476 pid
= nlh
->nlmsg_pid
;
477 flags
= nlh
->nlmsg_flags
;
479 if(pid
<= 0 || !(flags
& NLM_F_REQUEST
) || flags
& NLM_F_MULTI
)
480 RCV_SKB_FAIL(-EINVAL
);
482 if (flags
& MSG_TRUNC
)
483 RCV_SKB_FAIL(-ECOMM
);
485 type
= nlh
->nlmsg_type
;
486 if (type
< NLMSG_NOOP
|| type
>= IPQM_MAX
)
487 RCV_SKB_FAIL(-EINVAL
);
489 if (type
<= IPQM_BASE
)
492 if (security_netlink_recv(skb
, CAP_NET_ADMIN
))
493 RCV_SKB_FAIL(-EPERM
);
495 write_lock_bh(&queue_lock
);
498 if (peer_pid
!= pid
) {
499 write_unlock_bh(&queue_lock
);
500 RCV_SKB_FAIL(-EBUSY
);
503 net_enable_timestamp();
507 write_unlock_bh(&queue_lock
);
509 status
= ipq_receive_peer(NLMSG_DATA(nlh
), type
,
510 nlmsglen
- NLMSG_LENGTH(0));
512 RCV_SKB_FAIL(status
);
514 if (flags
& NLM_F_ACK
)
515 netlink_ack(skb
, nlh
, 0);
520 ipq_rcv_skb(struct sk_buff
*skb
)
522 mutex_lock(&ipqnl_mutex
);
524 mutex_unlock(&ipqnl_mutex
);
528 ipq_rcv_dev_event(struct notifier_block
*this,
529 unsigned long event
, void *ptr
)
531 struct net_device
*dev
= ptr
;
533 if (dev
->nd_net
!= &init_net
)
536 /* Drop any packets associated with the downed device */
537 if (event
== NETDEV_DOWN
)
538 ipq_dev_drop(dev
->ifindex
);
542 static struct notifier_block ipq_dev_notifier
= {
543 .notifier_call
= ipq_rcv_dev_event
,
547 ipq_rcv_nl_event(struct notifier_block
*this,
548 unsigned long event
, void *ptr
)
550 struct netlink_notify
*n
= ptr
;
552 if (event
== NETLINK_URELEASE
&&
553 n
->protocol
== NETLINK_IP6_FW
&& n
->pid
) {
554 write_lock_bh(&queue_lock
);
555 if ((n
->net
== &init_net
) && (n
->pid
== peer_pid
))
557 write_unlock_bh(&queue_lock
);
562 static struct notifier_block ipq_nl_notifier
= {
563 .notifier_call
= ipq_rcv_nl_event
,
566 static struct ctl_table_header
*ipq_sysctl_header
;
568 static ctl_table ipq_table
[] = {
570 .ctl_name
= NET_IPQ_QMAX
,
571 .procname
= NET_IPQ_QMAX_NAME
,
572 .data
= &queue_maxlen
,
573 .maxlen
= sizeof(queue_maxlen
),
575 .proc_handler
= proc_dointvec
580 static ctl_table ipq_dir_table
[] = {
582 .ctl_name
= NET_IPV6
,
590 static ctl_table ipq_root_table
[] = {
595 .child
= ipq_dir_table
600 static int ip6_queue_show(struct seq_file
*m
, void *v
)
602 read_lock_bh(&queue_lock
);
608 "Queue length : %u\n"
609 "Queue max. length : %u\n"
610 "Queue dropped : %u\n"
611 "Netfilter dropped : %u\n",
620 read_unlock_bh(&queue_lock
);
624 static int ip6_queue_open(struct inode
*inode
, struct file
*file
)
626 return single_open(file
, ip6_queue_show
, NULL
);
629 static const struct file_operations ip6_queue_proc_fops
= {
630 .open
= ip6_queue_open
,
633 .release
= single_release
,
634 .owner
= THIS_MODULE
,
637 static const struct nf_queue_handler nfqh
= {
639 .outfn
= &ipq_enqueue_packet
,
642 static int __init
ip6_queue_init(void)
644 int status
= -ENOMEM
;
645 struct proc_dir_entry
*proc
;
647 netlink_register_notifier(&ipq_nl_notifier
);
648 ipqnl
= netlink_kernel_create(&init_net
, NETLINK_IP6_FW
, 0,
649 ipq_rcv_skb
, NULL
, THIS_MODULE
);
651 printk(KERN_ERR
"ip6_queue: failed to create netlink socket\n");
652 goto cleanup_netlink_notifier
;
655 proc
= create_proc_entry(IPQ_PROC_FS_NAME
, 0, init_net
.proc_net
);
657 proc
->owner
= THIS_MODULE
;
658 proc
->proc_fops
= &ip6_queue_proc_fops
;
660 printk(KERN_ERR
"ip6_queue: failed to create proc entry\n");
664 register_netdevice_notifier(&ipq_dev_notifier
);
665 ipq_sysctl_header
= register_sysctl_table(ipq_root_table
);
667 status
= nf_register_queue_handler(PF_INET6
, &nfqh
);
669 printk(KERN_ERR
"ip6_queue: failed to register queue handler\n");
675 unregister_sysctl_table(ipq_sysctl_header
);
676 unregister_netdevice_notifier(&ipq_dev_notifier
);
677 proc_net_remove(&init_net
, IPQ_PROC_FS_NAME
);
680 sock_release(ipqnl
->sk_socket
);
681 mutex_lock(&ipqnl_mutex
);
682 mutex_unlock(&ipqnl_mutex
);
684 cleanup_netlink_notifier
:
685 netlink_unregister_notifier(&ipq_nl_notifier
);
689 static void __exit
ip6_queue_fini(void)
691 nf_unregister_queue_handlers(&nfqh
);
695 unregister_sysctl_table(ipq_sysctl_header
);
696 unregister_netdevice_notifier(&ipq_dev_notifier
);
697 proc_net_remove(&init_net
, IPQ_PROC_FS_NAME
);
699 sock_release(ipqnl
->sk_socket
);
700 mutex_lock(&ipqnl_mutex
);
701 mutex_unlock(&ipqnl_mutex
);
703 netlink_unregister_notifier(&ipq_nl_notifier
);
706 MODULE_DESCRIPTION("IPv6 packet queue handler");
707 MODULE_LICENSE("GPL");
709 module_init(ip6_queue_init
);
710 module_exit(ip6_queue_fini
);