2 * This is a module which is used for queueing IPv4 packets and
3 * communicating with userspace via netlink.
5 * (C) 2000-2002 James Morris <jmorris@intercode.com.au>
6 * (C) 2003-2005 Netfilter Core Team <coreteam@netfilter.org>
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 as
10 * published by the Free Software Foundation.
12 #include <linux/module.h>
13 #include <linux/skbuff.h>
14 #include <linux/init.h>
16 #include <linux/notifier.h>
17 #include <linux/netdevice.h>
18 #include <linux/netfilter.h>
19 #include <linux/netfilter_ipv4/ip_queue.h>
20 #include <linux/netfilter_ipv4/ip_tables.h>
21 #include <linux/netlink.h>
22 #include <linux/spinlock.h>
23 #include <linux/sysctl.h>
24 #include <linux/proc_fs.h>
25 #include <linux/security.h>
26 #include <linux/mutex.h>
27 #include <net/net_namespace.h>
29 #include <net/route.h>
31 #define IPQ_QMAX_DEFAULT 1024
32 #define IPQ_PROC_FS_NAME "ip_queue"
33 #define NET_IPQ_QMAX 2088
34 #define NET_IPQ_QMAX_NAME "ip_queue_maxlen"
36 struct ipq_queue_entry
{
37 struct list_head list
;
42 typedef int (*ipq_cmpfn
)(struct ipq_queue_entry
*, unsigned long);
44 static unsigned char copy_mode __read_mostly
= IPQ_COPY_NONE
;
45 static unsigned int queue_maxlen __read_mostly
= IPQ_QMAX_DEFAULT
;
46 static DEFINE_RWLOCK(queue_lock
);
47 static int peer_pid __read_mostly
;
48 static unsigned int copy_range __read_mostly
;
49 static unsigned int queue_total
;
50 static unsigned int queue_dropped
= 0;
51 static unsigned int queue_user_dropped
= 0;
52 static struct sock
*ipqnl __read_mostly
;
53 static LIST_HEAD(queue_list
);
54 static DEFINE_MUTEX(ipqnl_mutex
);
57 ipq_issue_verdict(struct ipq_queue_entry
*entry
, int verdict
)
59 /* TCP input path (and probably other bits) assume to be called
60 * from softirq context, not from syscall, like ipq_issue_verdict is
61 * called. TCP input path deadlocks with locks taken from timer
62 * softirq, e.g. We therefore emulate this by local_bh_disable() */
65 nf_reinject(entry
->skb
, entry
->info
, verdict
);
72 __ipq_enqueue_entry(struct ipq_queue_entry
*entry
)
74 list_add(&entry
->list
, &queue_list
);
79 * Find and return a queued entry matched by cmpfn, or return the last
80 * entry if cmpfn is NULL.
82 static inline struct ipq_queue_entry
*
83 __ipq_find_entry(ipq_cmpfn cmpfn
, unsigned long data
)
87 list_for_each_prev(p
, &queue_list
) {
88 struct ipq_queue_entry
*entry
= (struct ipq_queue_entry
*)p
;
90 if (!cmpfn
|| cmpfn(entry
, data
))
97 __ipq_dequeue_entry(struct ipq_queue_entry
*entry
)
99 list_del(&entry
->list
);
103 static inline struct ipq_queue_entry
*
104 __ipq_find_dequeue_entry(ipq_cmpfn cmpfn
, unsigned long data
)
106 struct ipq_queue_entry
*entry
;
108 entry
= __ipq_find_entry(cmpfn
, data
);
112 __ipq_dequeue_entry(entry
);
118 __ipq_flush(int verdict
)
120 struct ipq_queue_entry
*entry
;
122 while ((entry
= __ipq_find_dequeue_entry(NULL
, 0)))
123 ipq_issue_verdict(entry
, verdict
);
127 __ipq_set_mode(unsigned char mode
, unsigned int range
)
138 case IPQ_COPY_PACKET
:
141 if (copy_range
> 0xFFFF)
156 net_disable_timestamp();
157 __ipq_set_mode(IPQ_COPY_NONE
, 0);
158 __ipq_flush(NF_DROP
);
161 static struct ipq_queue_entry
*
162 ipq_find_dequeue_entry(ipq_cmpfn cmpfn
, unsigned long data
)
164 struct ipq_queue_entry
*entry
;
166 write_lock_bh(&queue_lock
);
167 entry
= __ipq_find_dequeue_entry(cmpfn
, data
);
168 write_unlock_bh(&queue_lock
);
173 ipq_flush(int verdict
)
175 write_lock_bh(&queue_lock
);
176 __ipq_flush(verdict
);
177 write_unlock_bh(&queue_lock
);
180 static struct sk_buff
*
181 ipq_build_packet_message(struct ipq_queue_entry
*entry
, int *errp
)
183 sk_buff_data_t old_tail
;
187 struct ipq_packet_msg
*pmsg
;
188 struct nlmsghdr
*nlh
;
191 read_lock_bh(&queue_lock
);
196 size
= NLMSG_SPACE(sizeof(*pmsg
));
200 case IPQ_COPY_PACKET
:
201 if ((entry
->skb
->ip_summed
== CHECKSUM_PARTIAL
||
202 entry
->skb
->ip_summed
== CHECKSUM_COMPLETE
) &&
203 (*errp
= skb_checksum_help(entry
->skb
))) {
204 read_unlock_bh(&queue_lock
);
207 if (copy_range
== 0 || copy_range
> entry
->skb
->len
)
208 data_len
= entry
->skb
->len
;
210 data_len
= copy_range
;
212 size
= NLMSG_SPACE(sizeof(*pmsg
) + data_len
);
217 read_unlock_bh(&queue_lock
);
221 read_unlock_bh(&queue_lock
);
223 skb
= alloc_skb(size
, GFP_ATOMIC
);
227 old_tail
= skb
->tail
;
228 nlh
= NLMSG_PUT(skb
, 0, 0, IPQM_PACKET
, size
- sizeof(*nlh
));
229 pmsg
= NLMSG_DATA(nlh
);
230 memset(pmsg
, 0, sizeof(*pmsg
));
232 pmsg
->packet_id
= (unsigned long )entry
;
233 pmsg
->data_len
= data_len
;
234 tv
= ktime_to_timeval(entry
->skb
->tstamp
);
235 pmsg
->timestamp_sec
= tv
.tv_sec
;
236 pmsg
->timestamp_usec
= tv
.tv_usec
;
237 pmsg
->mark
= entry
->skb
->mark
;
238 pmsg
->hook
= entry
->info
->hook
;
239 pmsg
->hw_protocol
= entry
->skb
->protocol
;
241 if (entry
->info
->indev
)
242 strcpy(pmsg
->indev_name
, entry
->info
->indev
->name
);
244 pmsg
->indev_name
[0] = '\0';
246 if (entry
->info
->outdev
)
247 strcpy(pmsg
->outdev_name
, entry
->info
->outdev
->name
);
249 pmsg
->outdev_name
[0] = '\0';
251 if (entry
->info
->indev
&& entry
->skb
->dev
) {
252 pmsg
->hw_type
= entry
->skb
->dev
->type
;
253 if (entry
->skb
->dev
->hard_header_parse
)
255 entry
->skb
->dev
->hard_header_parse(entry
->skb
,
260 if (skb_copy_bits(entry
->skb
, 0, pmsg
->payload
, data_len
))
263 nlh
->nlmsg_len
= skb
->tail
- old_tail
;
270 printk(KERN_ERR
"ip_queue: error creating packet message\n");
275 ipq_enqueue_packet(struct sk_buff
*skb
, struct nf_info
*info
,
276 unsigned int queuenum
, void *data
)
278 int status
= -EINVAL
;
279 struct sk_buff
*nskb
;
280 struct ipq_queue_entry
*entry
;
282 if (copy_mode
== IPQ_COPY_NONE
)
285 entry
= kmalloc(sizeof(*entry
), GFP_ATOMIC
);
287 printk(KERN_ERR
"ip_queue: OOM in ipq_enqueue_packet()\n");
294 nskb
= ipq_build_packet_message(entry
, &status
);
298 write_lock_bh(&queue_lock
);
301 goto err_out_free_nskb
;
303 if (queue_total
>= queue_maxlen
) {
307 printk (KERN_WARNING
"ip_queue: full at %d entries, "
308 "dropping packets(s). Dropped: %d\n", queue_total
,
310 goto err_out_free_nskb
;
313 /* netlink_unicast will either free the nskb or attach it to a socket */
314 status
= netlink_unicast(ipqnl
, nskb
, peer_pid
, MSG_DONTWAIT
);
316 queue_user_dropped
++;
320 __ipq_enqueue_entry(entry
);
322 write_unlock_bh(&queue_lock
);
329 write_unlock_bh(&queue_lock
);
337 ipq_mangle_ipv4(ipq_verdict_msg_t
*v
, struct ipq_queue_entry
*e
)
340 struct iphdr
*user_iph
= (struct iphdr
*)v
->payload
;
342 if (v
->data_len
< sizeof(*user_iph
))
344 diff
= v
->data_len
- e
->skb
->len
;
346 if (pskb_trim(e
->skb
, v
->data_len
))
348 } else if (diff
> 0) {
349 if (v
->data_len
> 0xFFFF)
351 if (diff
> skb_tailroom(e
->skb
)) {
352 struct sk_buff
*newskb
;
354 newskb
= skb_copy_expand(e
->skb
,
355 skb_headroom(e
->skb
),
358 if (newskb
== NULL
) {
359 printk(KERN_WARNING
"ip_queue: OOM "
360 "in mangle, dropping packet\n");
364 skb_set_owner_w(newskb
, e
->skb
->sk
);
368 skb_put(e
->skb
, diff
);
370 if (!skb_make_writable(&e
->skb
, v
->data_len
))
372 skb_copy_to_linear_data(e
->skb
, v
->payload
, v
->data_len
);
373 e
->skb
->ip_summed
= CHECKSUM_NONE
;
379 id_cmp(struct ipq_queue_entry
*e
, unsigned long id
)
381 return (id
== (unsigned long )e
);
385 ipq_set_verdict(struct ipq_verdict_msg
*vmsg
, unsigned int len
)
387 struct ipq_queue_entry
*entry
;
389 if (vmsg
->value
> NF_MAX_VERDICT
)
392 entry
= ipq_find_dequeue_entry(id_cmp
, vmsg
->id
);
396 int verdict
= vmsg
->value
;
398 if (vmsg
->data_len
&& vmsg
->data_len
== len
)
399 if (ipq_mangle_ipv4(vmsg
, entry
) < 0)
402 ipq_issue_verdict(entry
, verdict
);
408 ipq_set_mode(unsigned char mode
, unsigned int range
)
412 write_lock_bh(&queue_lock
);
413 status
= __ipq_set_mode(mode
, range
);
414 write_unlock_bh(&queue_lock
);
419 ipq_receive_peer(struct ipq_peer_msg
*pmsg
,
420 unsigned char type
, unsigned int len
)
424 if (len
< sizeof(*pmsg
))
429 status
= ipq_set_mode(pmsg
->msg
.mode
.value
,
430 pmsg
->msg
.mode
.range
);
434 if (pmsg
->msg
.verdict
.value
> NF_MAX_VERDICT
)
437 status
= ipq_set_verdict(&pmsg
->msg
.verdict
,
438 len
- sizeof(*pmsg
));
447 dev_cmp(struct ipq_queue_entry
*entry
, unsigned long ifindex
)
449 if (entry
->info
->indev
)
450 if (entry
->info
->indev
->ifindex
== ifindex
)
452 if (entry
->info
->outdev
)
453 if (entry
->info
->outdev
->ifindex
== ifindex
)
455 #ifdef CONFIG_BRIDGE_NETFILTER
456 if (entry
->skb
->nf_bridge
) {
457 if (entry
->skb
->nf_bridge
->physindev
&&
458 entry
->skb
->nf_bridge
->physindev
->ifindex
== ifindex
)
460 if (entry
->skb
->nf_bridge
->physoutdev
&&
461 entry
->skb
->nf_bridge
->physoutdev
->ifindex
== ifindex
)
469 ipq_dev_drop(int ifindex
)
471 struct ipq_queue_entry
*entry
;
473 while ((entry
= ipq_find_dequeue_entry(dev_cmp
, ifindex
)) != NULL
)
474 ipq_issue_verdict(entry
, NF_DROP
);
477 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
480 ipq_rcv_skb(struct sk_buff
*skb
)
482 int status
, type
, pid
, flags
, nlmsglen
, skblen
;
483 struct nlmsghdr
*nlh
;
486 if (skblen
< sizeof(*nlh
))
489 nlh
= nlmsg_hdr(skb
);
490 nlmsglen
= nlh
->nlmsg_len
;
491 if (nlmsglen
< sizeof(*nlh
) || skblen
< nlmsglen
)
494 pid
= nlh
->nlmsg_pid
;
495 flags
= nlh
->nlmsg_flags
;
497 if(pid
<= 0 || !(flags
& NLM_F_REQUEST
) || flags
& NLM_F_MULTI
)
498 RCV_SKB_FAIL(-EINVAL
);
500 if (flags
& MSG_TRUNC
)
501 RCV_SKB_FAIL(-ECOMM
);
503 type
= nlh
->nlmsg_type
;
504 if (type
< NLMSG_NOOP
|| type
>= IPQM_MAX
)
505 RCV_SKB_FAIL(-EINVAL
);
507 if (type
<= IPQM_BASE
)
510 if (security_netlink_recv(skb
, CAP_NET_ADMIN
))
511 RCV_SKB_FAIL(-EPERM
);
513 write_lock_bh(&queue_lock
);
516 if (peer_pid
!= pid
) {
517 write_unlock_bh(&queue_lock
);
518 RCV_SKB_FAIL(-EBUSY
);
521 net_enable_timestamp();
525 write_unlock_bh(&queue_lock
);
527 status
= ipq_receive_peer(NLMSG_DATA(nlh
), type
,
528 nlmsglen
- NLMSG_LENGTH(0));
530 RCV_SKB_FAIL(status
);
532 if (flags
& NLM_F_ACK
)
533 netlink_ack(skb
, nlh
, 0);
538 ipq_rcv_sk(struct sock
*sk
, int len
)
543 mutex_lock(&ipqnl_mutex
);
545 for (qlen
= skb_queue_len(&sk
->sk_receive_queue
); qlen
; qlen
--) {
546 skb
= skb_dequeue(&sk
->sk_receive_queue
);
551 mutex_unlock(&ipqnl_mutex
);
555 ipq_rcv_dev_event(struct notifier_block
*this,
556 unsigned long event
, void *ptr
)
558 struct net_device
*dev
= ptr
;
560 if (dev
->nd_net
!= &init_net
)
563 /* Drop any packets associated with the downed device */
564 if (event
== NETDEV_DOWN
)
565 ipq_dev_drop(dev
->ifindex
);
569 static struct notifier_block ipq_dev_notifier
= {
570 .notifier_call
= ipq_rcv_dev_event
,
574 ipq_rcv_nl_event(struct notifier_block
*this,
575 unsigned long event
, void *ptr
)
577 struct netlink_notify
*n
= ptr
;
579 if (event
== NETLINK_URELEASE
&&
580 n
->protocol
== NETLINK_FIREWALL
&& n
->pid
) {
581 write_lock_bh(&queue_lock
);
582 if ((n
->net
== &init_net
) && (n
->pid
== peer_pid
))
584 write_unlock_bh(&queue_lock
);
589 static struct notifier_block ipq_nl_notifier
= {
590 .notifier_call
= ipq_rcv_nl_event
,
593 static struct ctl_table_header
*ipq_sysctl_header
;
595 static ctl_table ipq_table
[] = {
597 .ctl_name
= NET_IPQ_QMAX
,
598 .procname
= NET_IPQ_QMAX_NAME
,
599 .data
= &queue_maxlen
,
600 .maxlen
= sizeof(queue_maxlen
),
602 .proc_handler
= proc_dointvec
607 static ctl_table ipq_dir_table
[] = {
609 .ctl_name
= NET_IPV4
,
617 static ctl_table ipq_root_table
[] = {
622 .child
= ipq_dir_table
627 #ifdef CONFIG_PROC_FS
629 ipq_get_info(char *buffer
, char **start
, off_t offset
, int length
)
633 read_lock_bh(&queue_lock
);
635 len
= sprintf(buffer
,
639 "Queue length : %u\n"
640 "Queue max. length : %u\n"
641 "Queue dropped : %u\n"
642 "Netlink dropped : %u\n",
651 read_unlock_bh(&queue_lock
);
653 *start
= buffer
+ offset
;
661 #endif /* CONFIG_PROC_FS */
663 static struct nf_queue_handler nfqh
= {
665 .outfn
= &ipq_enqueue_packet
,
668 static int __init
ip_queue_init(void)
670 int status
= -ENOMEM
;
671 struct proc_dir_entry
*proc
;
673 netlink_register_notifier(&ipq_nl_notifier
);
674 ipqnl
= netlink_kernel_create(&init_net
, NETLINK_FIREWALL
, 0,
675 ipq_rcv_sk
, NULL
, THIS_MODULE
);
677 printk(KERN_ERR
"ip_queue: failed to create netlink socket\n");
678 goto cleanup_netlink_notifier
;
681 proc
= proc_net_create(&init_net
, IPQ_PROC_FS_NAME
, 0, ipq_get_info
);
683 proc
->owner
= THIS_MODULE
;
685 printk(KERN_ERR
"ip_queue: failed to create proc entry\n");
689 register_netdevice_notifier(&ipq_dev_notifier
);
690 ipq_sysctl_header
= register_sysctl_table(ipq_root_table
);
692 status
= nf_register_queue_handler(PF_INET
, &nfqh
);
694 printk(KERN_ERR
"ip_queue: failed to register queue handler\n");
700 unregister_sysctl_table(ipq_sysctl_header
);
701 unregister_netdevice_notifier(&ipq_dev_notifier
);
702 proc_net_remove(&init_net
, IPQ_PROC_FS_NAME
);
704 sock_release(ipqnl
->sk_socket
);
705 mutex_lock(&ipqnl_mutex
);
706 mutex_unlock(&ipqnl_mutex
);
708 cleanup_netlink_notifier
:
709 netlink_unregister_notifier(&ipq_nl_notifier
);
713 static void __exit
ip_queue_fini(void)
715 nf_unregister_queue_handlers(&nfqh
);
719 unregister_sysctl_table(ipq_sysctl_header
);
720 unregister_netdevice_notifier(&ipq_dev_notifier
);
721 proc_net_remove(&init_net
, IPQ_PROC_FS_NAME
);
723 sock_release(ipqnl
->sk_socket
);
724 mutex_lock(&ipqnl_mutex
);
725 mutex_unlock(&ipqnl_mutex
);
727 netlink_unregister_notifier(&ipq_nl_notifier
);
730 MODULE_DESCRIPTION("IPv4 packet queue handler");
731 MODULE_AUTHOR("James Morris <jmorris@intercode.com.au>");
732 MODULE_LICENSE("GPL");
734 module_init(ip_queue_init
);
735 module_exit(ip_queue_fini
);