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 * 2001-11-06: First try. Working with ip_queue.c for IPv4 and trying
17 * HEAVILY based in ipqueue.c by James Morris. It's just
18 * a little modified version of it, so he's nearly the
20 * Few changes needed, mainly the hard_routing code and
21 * the netlink socket protocol (we're NETLINK_IP6_FW).
22 * 2002-06-25: Code cleanup. [JM: ported cleanup over from ip_queue.c]
23 * 2005-02-04: Added /proc counter for dropped packets; fixed so
24 * packets aren't delivered to user space if they're going
27 #include <linux/module.h>
28 #include <linux/skbuff.h>
29 #include <linux/init.h>
30 #include <linux/ipv6.h>
31 #include <linux/notifier.h>
32 #include <linux/netdevice.h>
33 #include <linux/netfilter.h>
34 #include <linux/netlink.h>
35 #include <linux/spinlock.h>
36 #include <linux/sysctl.h>
37 #include <linux/proc_fs.h>
40 #include <net/ip6_route.h>
41 #include <linux/netfilter_ipv4/ip_queue.h>
42 #include <linux/netfilter_ipv4/ip_tables.h>
43 #include <linux/netfilter_ipv6/ip6_tables.h>
45 #define IPQ_QMAX_DEFAULT 1024
46 #define IPQ_PROC_FS_NAME "ip6_queue"
47 #define NET_IPQ_QMAX 2088
48 #define NET_IPQ_QMAX_NAME "ip6_queue_maxlen"
50 struct ipq_queue_entry
{
51 struct list_head list
;
56 typedef int (*ipq_cmpfn
)(struct ipq_queue_entry
*, unsigned long);
58 static unsigned char copy_mode
= IPQ_COPY_NONE
;
59 static unsigned int queue_maxlen
= IPQ_QMAX_DEFAULT
;
60 static DEFINE_RWLOCK(queue_lock
);
62 static unsigned int copy_range
;
63 static unsigned int queue_total
;
64 static unsigned int queue_dropped
= 0;
65 static unsigned int queue_user_dropped
= 0;
66 static struct sock
*ipqnl
;
67 static LIST_HEAD(queue_list
);
68 static DECLARE_MUTEX(ipqnl_sem
);
71 ipq_issue_verdict(struct ipq_queue_entry
*entry
, int verdict
)
74 nf_reinject(entry
->skb
, entry
->info
, verdict
);
80 __ipq_enqueue_entry(struct ipq_queue_entry
*entry
)
82 list_add(&entry
->list
, &queue_list
);
87 * Find and return a queued entry matched by cmpfn, or return the last
88 * entry if cmpfn is NULL.
90 static inline struct ipq_queue_entry
*
91 __ipq_find_entry(ipq_cmpfn cmpfn
, unsigned long data
)
95 list_for_each_prev(p
, &queue_list
) {
96 struct ipq_queue_entry
*entry
= (struct ipq_queue_entry
*)p
;
98 if (!cmpfn
|| cmpfn(entry
, data
))
105 __ipq_dequeue_entry(struct ipq_queue_entry
*entry
)
107 list_del(&entry
->list
);
111 static inline struct ipq_queue_entry
*
112 __ipq_find_dequeue_entry(ipq_cmpfn cmpfn
, unsigned long data
)
114 struct ipq_queue_entry
*entry
;
116 entry
= __ipq_find_entry(cmpfn
, data
);
120 __ipq_dequeue_entry(entry
);
126 __ipq_flush(int verdict
)
128 struct ipq_queue_entry
*entry
;
130 while ((entry
= __ipq_find_dequeue_entry(NULL
, 0)))
131 ipq_issue_verdict(entry
, verdict
);
135 __ipq_set_mode(unsigned char mode
, unsigned int range
)
146 case IPQ_COPY_PACKET
:
149 if (copy_range
> 0xFFFF)
164 net_disable_timestamp();
165 __ipq_set_mode(IPQ_COPY_NONE
, 0);
166 __ipq_flush(NF_DROP
);
169 static struct ipq_queue_entry
*
170 ipq_find_dequeue_entry(ipq_cmpfn cmpfn
, unsigned long data
)
172 struct ipq_queue_entry
*entry
;
174 write_lock_bh(&queue_lock
);
175 entry
= __ipq_find_dequeue_entry(cmpfn
, data
);
176 write_unlock_bh(&queue_lock
);
181 ipq_flush(int verdict
)
183 write_lock_bh(&queue_lock
);
184 __ipq_flush(verdict
);
185 write_unlock_bh(&queue_lock
);
188 static struct sk_buff
*
189 ipq_build_packet_message(struct ipq_queue_entry
*entry
, int *errp
)
191 unsigned char *old_tail
;
195 struct ipq_packet_msg
*pmsg
;
196 struct nlmsghdr
*nlh
;
198 read_lock_bh(&queue_lock
);
203 size
= NLMSG_SPACE(sizeof(*pmsg
));
207 case IPQ_COPY_PACKET
:
208 if (entry
->skb
->ip_summed
== CHECKSUM_HW
&&
209 (*errp
= skb_checksum_help(entry
->skb
,
210 entry
->info
->outdev
== NULL
))) {
211 read_unlock_bh(&queue_lock
);
214 if (copy_range
== 0 || copy_range
> entry
->skb
->len
)
215 data_len
= entry
->skb
->len
;
217 data_len
= copy_range
;
219 size
= NLMSG_SPACE(sizeof(*pmsg
) + data_len
);
224 read_unlock_bh(&queue_lock
);
228 read_unlock_bh(&queue_lock
);
230 skb
= alloc_skb(size
, GFP_ATOMIC
);
235 nlh
= NLMSG_PUT(skb
, 0, 0, IPQM_PACKET
, size
- sizeof(*nlh
));
236 pmsg
= NLMSG_DATA(nlh
);
237 memset(pmsg
, 0, sizeof(*pmsg
));
239 pmsg
->packet_id
= (unsigned long )entry
;
240 pmsg
->data_len
= data_len
;
241 pmsg
->timestamp_sec
= entry
->skb
->tstamp
.off_sec
;
242 pmsg
->timestamp_usec
= entry
->skb
->tstamp
.off_usec
;
243 pmsg
->mark
= entry
->skb
->nfmark
;
244 pmsg
->hook
= entry
->info
->hook
;
245 pmsg
->hw_protocol
= entry
->skb
->protocol
;
247 if (entry
->info
->indev
)
248 strcpy(pmsg
->indev_name
, entry
->info
->indev
->name
);
250 pmsg
->indev_name
[0] = '\0';
252 if (entry
->info
->outdev
)
253 strcpy(pmsg
->outdev_name
, entry
->info
->outdev
->name
);
255 pmsg
->outdev_name
[0] = '\0';
257 if (entry
->info
->indev
&& entry
->skb
->dev
) {
258 pmsg
->hw_type
= entry
->skb
->dev
->type
;
259 if (entry
->skb
->dev
->hard_header_parse
)
261 entry
->skb
->dev
->hard_header_parse(entry
->skb
,
266 if (skb_copy_bits(entry
->skb
, 0, pmsg
->payload
, data_len
))
269 nlh
->nlmsg_len
= skb
->tail
- old_tail
;
276 printk(KERN_ERR
"ip6_queue: error creating packet message\n");
281 ipq_enqueue_packet(struct sk_buff
*skb
, struct nf_info
*info
,
282 unsigned int queuenum
, void *data
)
284 int status
= -EINVAL
;
285 struct sk_buff
*nskb
;
286 struct ipq_queue_entry
*entry
;
288 if (copy_mode
== IPQ_COPY_NONE
)
291 entry
= kmalloc(sizeof(*entry
), GFP_ATOMIC
);
293 printk(KERN_ERR
"ip6_queue: OOM in ipq_enqueue_packet()\n");
300 nskb
= ipq_build_packet_message(entry
, &status
);
304 write_lock_bh(&queue_lock
);
307 goto err_out_free_nskb
;
309 if (queue_total
>= queue_maxlen
) {
313 printk (KERN_WARNING
"ip6_queue: fill at %d entries, "
314 "dropping packet(s). Dropped: %d\n", queue_total
,
316 goto err_out_free_nskb
;
319 /* netlink_unicast will either free the nskb or attach it to a socket */
320 status
= netlink_unicast(ipqnl
, nskb
, peer_pid
, MSG_DONTWAIT
);
322 queue_user_dropped
++;
326 __ipq_enqueue_entry(entry
);
328 write_unlock_bh(&queue_lock
);
335 write_unlock_bh(&queue_lock
);
343 ipq_mangle_ipv6(ipq_verdict_msg_t
*v
, struct ipq_queue_entry
*e
)
346 struct ipv6hdr
*user_iph
= (struct ipv6hdr
*)v
->payload
;
348 if (v
->data_len
< sizeof(*user_iph
))
350 diff
= v
->data_len
- e
->skb
->len
;
352 skb_trim(e
->skb
, v
->data_len
);
354 if (v
->data_len
> 0xFFFF)
356 if (diff
> skb_tailroom(e
->skb
)) {
357 struct sk_buff
*newskb
;
359 newskb
= skb_copy_expand(e
->skb
,
360 skb_headroom(e
->skb
),
363 if (newskb
== NULL
) {
364 printk(KERN_WARNING
"ip6_queue: OOM "
365 "in mangle, dropping packet\n");
369 skb_set_owner_w(newskb
, e
->skb
->sk
);
373 skb_put(e
->skb
, diff
);
375 if (!skb_make_writable(&e
->skb
, v
->data_len
))
377 memcpy(e
->skb
->data
, v
->payload
, v
->data_len
);
378 e
->skb
->ip_summed
= CHECKSUM_NONE
;
384 id_cmp(struct ipq_queue_entry
*e
, unsigned long id
)
386 return (id
== (unsigned long )e
);
390 ipq_set_verdict(struct ipq_verdict_msg
*vmsg
, unsigned int len
)
392 struct ipq_queue_entry
*entry
;
394 if (vmsg
->value
> NF_MAX_VERDICT
)
397 entry
= ipq_find_dequeue_entry(id_cmp
, vmsg
->id
);
401 int verdict
= vmsg
->value
;
403 if (vmsg
->data_len
&& vmsg
->data_len
== len
)
404 if (ipq_mangle_ipv6(vmsg
, entry
) < 0)
407 ipq_issue_verdict(entry
, verdict
);
413 ipq_set_mode(unsigned char mode
, unsigned int range
)
417 write_lock_bh(&queue_lock
);
418 status
= __ipq_set_mode(mode
, range
);
419 write_unlock_bh(&queue_lock
);
424 ipq_receive_peer(struct ipq_peer_msg
*pmsg
,
425 unsigned char type
, unsigned int len
)
429 if (len
< sizeof(*pmsg
))
434 status
= ipq_set_mode(pmsg
->msg
.mode
.value
,
435 pmsg
->msg
.mode
.range
);
439 if (pmsg
->msg
.verdict
.value
> NF_MAX_VERDICT
)
442 status
= ipq_set_verdict(&pmsg
->msg
.verdict
,
443 len
- sizeof(*pmsg
));
452 dev_cmp(struct ipq_queue_entry
*entry
, unsigned long ifindex
)
454 if (entry
->info
->indev
)
455 if (entry
->info
->indev
->ifindex
== ifindex
)
458 if (entry
->info
->outdev
)
459 if (entry
->info
->outdev
->ifindex
== ifindex
)
466 ipq_dev_drop(int ifindex
)
468 struct ipq_queue_entry
*entry
;
470 while ((entry
= ipq_find_dequeue_entry(dev_cmp
, ifindex
)) != NULL
)
471 ipq_issue_verdict(entry
, NF_DROP
);
474 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
477 ipq_rcv_skb(struct sk_buff
*skb
)
479 int status
, type
, pid
, flags
, nlmsglen
, skblen
;
480 struct nlmsghdr
*nlh
;
483 if (skblen
< sizeof(*nlh
))
486 nlh
= (struct nlmsghdr
*)skb
->data
;
487 nlmsglen
= nlh
->nlmsg_len
;
488 if (nlmsglen
< sizeof(*nlh
) || skblen
< nlmsglen
)
491 pid
= nlh
->nlmsg_pid
;
492 flags
= nlh
->nlmsg_flags
;
494 if(pid
<= 0 || !(flags
& NLM_F_REQUEST
) || flags
& NLM_F_MULTI
)
495 RCV_SKB_FAIL(-EINVAL
);
497 if (flags
& MSG_TRUNC
)
498 RCV_SKB_FAIL(-ECOMM
);
500 type
= nlh
->nlmsg_type
;
501 if (type
< NLMSG_NOOP
|| type
>= IPQM_MAX
)
502 RCV_SKB_FAIL(-EINVAL
);
504 if (type
<= IPQM_BASE
)
507 if (security_netlink_recv(skb
))
508 RCV_SKB_FAIL(-EPERM
);
510 write_lock_bh(&queue_lock
);
513 if (peer_pid
!= pid
) {
514 write_unlock_bh(&queue_lock
);
515 RCV_SKB_FAIL(-EBUSY
);
518 net_enable_timestamp();
522 write_unlock_bh(&queue_lock
);
524 status
= ipq_receive_peer(NLMSG_DATA(nlh
), type
,
525 skblen
- NLMSG_LENGTH(0));
527 RCV_SKB_FAIL(status
);
529 if (flags
& NLM_F_ACK
)
530 netlink_ack(skb
, nlh
, 0);
535 ipq_rcv_sk(struct sock
*sk
, int len
)
542 for (qlen
= skb_queue_len(&sk
->sk_receive_queue
); qlen
; qlen
--) {
543 skb
= skb_dequeue(&sk
->sk_receive_queue
);
552 ipq_rcv_dev_event(struct notifier_block
*this,
553 unsigned long event
, void *ptr
)
555 struct net_device
*dev
= ptr
;
557 /* Drop any packets associated with the downed device */
558 if (event
== NETDEV_DOWN
)
559 ipq_dev_drop(dev
->ifindex
);
563 static struct notifier_block ipq_dev_notifier
= {
564 .notifier_call
= ipq_rcv_dev_event
,
568 ipq_rcv_nl_event(struct notifier_block
*this,
569 unsigned long event
, void *ptr
)
571 struct netlink_notify
*n
= ptr
;
573 if (event
== NETLINK_URELEASE
&&
574 n
->protocol
== NETLINK_IP6_FW
&& n
->pid
) {
575 write_lock_bh(&queue_lock
);
576 if (n
->pid
== peer_pid
)
578 write_unlock_bh(&queue_lock
);
583 static struct notifier_block ipq_nl_notifier
= {
584 .notifier_call
= ipq_rcv_nl_event
,
587 static struct ctl_table_header
*ipq_sysctl_header
;
589 static ctl_table ipq_table
[] = {
591 .ctl_name
= NET_IPQ_QMAX
,
592 .procname
= NET_IPQ_QMAX_NAME
,
593 .data
= &queue_maxlen
,
594 .maxlen
= sizeof(queue_maxlen
),
596 .proc_handler
= proc_dointvec
601 static ctl_table ipq_dir_table
[] = {
603 .ctl_name
= NET_IPV6
,
611 static ctl_table ipq_root_table
[] = {
616 .child
= ipq_dir_table
622 ipq_get_info(char *buffer
, char **start
, off_t offset
, int length
)
626 read_lock_bh(&queue_lock
);
628 len
= sprintf(buffer
,
632 "Queue length : %u\n"
633 "Queue max. length : %u\n"
634 "Queue dropped : %u\n"
635 "Netfilter dropped : %u\n",
644 read_unlock_bh(&queue_lock
);
646 *start
= buffer
+ offset
;
655 static struct nf_queue_handler nfqh
= {
657 .outfn
= &ipq_enqueue_packet
,
661 init_or_cleanup(int init
)
663 int status
= -ENOMEM
;
664 struct proc_dir_entry
*proc
;
669 netlink_register_notifier(&ipq_nl_notifier
);
670 ipqnl
= netlink_kernel_create(NETLINK_IP6_FW
, 0, ipq_rcv_sk
,
673 printk(KERN_ERR
"ip6_queue: failed to create netlink socket\n");
674 goto cleanup_netlink_notifier
;
677 proc
= proc_net_create(IPQ_PROC_FS_NAME
, 0, ipq_get_info
);
679 proc
->owner
= THIS_MODULE
;
681 printk(KERN_ERR
"ip6_queue: failed to create proc entry\n");
685 register_netdevice_notifier(&ipq_dev_notifier
);
686 ipq_sysctl_header
= register_sysctl_table(ipq_root_table
, 0);
688 status
= nf_register_queue_handler(PF_INET6
, &nfqh
);
690 printk(KERN_ERR
"ip6_queue: failed to register queue handler\n");
696 nf_unregister_queue_handlers(&nfqh
);
701 unregister_sysctl_table(ipq_sysctl_header
);
702 unregister_netdevice_notifier(&ipq_dev_notifier
);
703 proc_net_remove(IPQ_PROC_FS_NAME
);
706 sock_release(ipqnl
->sk_socket
);
710 cleanup_netlink_notifier
:
711 netlink_unregister_notifier(&ipq_nl_notifier
);
715 static int __init
init(void)
718 return init_or_cleanup(1);
721 static void __exit
fini(void)
726 MODULE_DESCRIPTION("IPv6 packet queue handler");
727 MODULE_LICENSE("GPL");