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/seq_file.h>
26 #include <linux/security.h>
27 #include <linux/mutex.h>
28 #include <net/net_namespace.h>
30 #include <net/route.h>
31 #include <net/netfilter/nf_queue.h>
34 #define IPQ_QMAX_DEFAULT 1024
35 #define IPQ_PROC_FS_NAME "ip_queue"
36 #define NET_IPQ_QMAX 2088
37 #define NET_IPQ_QMAX_NAME "ip_queue_maxlen"
39 typedef int (*ipq_cmpfn
)(struct nf_queue_entry
*, unsigned long);
41 static unsigned char copy_mode __read_mostly
= IPQ_COPY_NONE
;
42 static unsigned int queue_maxlen __read_mostly
= IPQ_QMAX_DEFAULT
;
43 static DEFINE_RWLOCK(queue_lock
);
44 static int peer_pid __read_mostly
;
45 static unsigned int copy_range __read_mostly
;
46 static unsigned int queue_total
;
47 static unsigned int queue_dropped
= 0;
48 static unsigned int queue_user_dropped
= 0;
49 static struct sock
*ipqnl __read_mostly
;
50 static LIST_HEAD(queue_list
);
51 static DEFINE_MUTEX(ipqnl_mutex
);
54 __ipq_enqueue_entry(struct nf_queue_entry
*entry
)
56 list_add_tail(&entry
->list
, &queue_list
);
61 __ipq_set_mode(unsigned char mode
, unsigned int range
)
75 if (copy_range
> 0xFFFF)
86 static void __ipq_flush(ipq_cmpfn cmpfn
, unsigned long data
);
92 net_disable_timestamp();
93 __ipq_set_mode(IPQ_COPY_NONE
, 0);
97 static struct nf_queue_entry
*
98 ipq_find_dequeue_entry(unsigned long id
)
100 struct nf_queue_entry
*entry
= NULL
, *i
;
102 write_lock_bh(&queue_lock
);
104 list_for_each_entry(i
, &queue_list
, list
) {
105 if ((unsigned long)i
== id
) {
112 list_del(&entry
->list
);
116 write_unlock_bh(&queue_lock
);
121 __ipq_flush(ipq_cmpfn cmpfn
, unsigned long data
)
123 struct nf_queue_entry
*entry
, *next
;
125 list_for_each_entry_safe(entry
, next
, &queue_list
, list
) {
126 if (!cmpfn
|| cmpfn(entry
, data
)) {
127 list_del(&entry
->list
);
129 nf_reinject(entry
, NF_DROP
);
135 ipq_flush(ipq_cmpfn cmpfn
, unsigned long data
)
137 write_lock_bh(&queue_lock
);
138 __ipq_flush(cmpfn
, data
);
139 write_unlock_bh(&queue_lock
);
142 static struct sk_buff
*
143 ipq_build_packet_message(struct nf_queue_entry
*entry
, int *errp
)
145 sk_buff_data_t old_tail
;
149 struct ipq_packet_msg
*pmsg
;
150 struct nlmsghdr
*nlh
;
153 read_lock_bh(&queue_lock
);
158 size
= NLMSG_SPACE(sizeof(*pmsg
));
162 case IPQ_COPY_PACKET
:
163 if ((entry
->skb
->ip_summed
== CHECKSUM_PARTIAL
||
164 entry
->skb
->ip_summed
== CHECKSUM_COMPLETE
) &&
165 (*errp
= skb_checksum_help(entry
->skb
))) {
166 read_unlock_bh(&queue_lock
);
169 if (copy_range
== 0 || copy_range
> entry
->skb
->len
)
170 data_len
= entry
->skb
->len
;
172 data_len
= copy_range
;
174 size
= NLMSG_SPACE(sizeof(*pmsg
) + data_len
);
179 read_unlock_bh(&queue_lock
);
183 read_unlock_bh(&queue_lock
);
185 skb
= alloc_skb(size
, GFP_ATOMIC
);
189 old_tail
= skb
->tail
;
190 nlh
= NLMSG_PUT(skb
, 0, 0, IPQM_PACKET
, size
- sizeof(*nlh
));
191 pmsg
= NLMSG_DATA(nlh
);
192 memset(pmsg
, 0, sizeof(*pmsg
));
194 pmsg
->packet_id
= (unsigned long )entry
;
195 pmsg
->data_len
= data_len
;
196 tv
= ktime_to_timeval(entry
->skb
->tstamp
);
197 pmsg
->timestamp_sec
= tv
.tv_sec
;
198 pmsg
->timestamp_usec
= tv
.tv_usec
;
199 pmsg
->mark
= entry
->skb
->mark
;
200 pmsg
->hook
= entry
->hook
;
201 pmsg
->hw_protocol
= entry
->skb
->protocol
;
204 strcpy(pmsg
->indev_name
, entry
->indev
->name
);
206 pmsg
->indev_name
[0] = '\0';
209 strcpy(pmsg
->outdev_name
, entry
->outdev
->name
);
211 pmsg
->outdev_name
[0] = '\0';
213 if (entry
->indev
&& entry
->skb
->dev
) {
214 pmsg
->hw_type
= entry
->skb
->dev
->type
;
215 pmsg
->hw_addrlen
= dev_parse_header(entry
->skb
,
220 if (skb_copy_bits(entry
->skb
, 0, pmsg
->payload
, data_len
))
223 nlh
->nlmsg_len
= skb
->tail
- old_tail
;
230 printk(KERN_ERR
"ip_queue: error creating packet message\n");
235 ipq_enqueue_packet(struct nf_queue_entry
*entry
, unsigned int queuenum
)
237 int status
= -EINVAL
;
238 struct sk_buff
*nskb
;
240 if (copy_mode
== IPQ_COPY_NONE
)
243 nskb
= ipq_build_packet_message(entry
, &status
);
247 write_lock_bh(&queue_lock
);
250 goto err_out_free_nskb
;
252 if (queue_total
>= queue_maxlen
) {
256 printk (KERN_WARNING
"ip_queue: full at %d entries, "
257 "dropping packets(s). Dropped: %d\n", queue_total
,
259 goto err_out_free_nskb
;
262 /* netlink_unicast will either free the nskb or attach it to a socket */
263 status
= netlink_unicast(ipqnl
, nskb
, peer_pid
, MSG_DONTWAIT
);
265 queue_user_dropped
++;
269 __ipq_enqueue_entry(entry
);
271 write_unlock_bh(&queue_lock
);
278 write_unlock_bh(&queue_lock
);
283 ipq_mangle_ipv4(ipq_verdict_msg_t
*v
, struct nf_queue_entry
*e
)
286 <<<<<<< HEAD
:net
/ipv4
/netfilter
/ip_queue
.c
289 >>>>>>> 264e3e889d86e552b4191d69bb60f4f3b383135a
:net
/ipv4
/netfilter
/ip_queue
.c
290 struct iphdr
*user_iph
= (struct iphdr
*)v
->payload
;
291 <<<<<<< HEAD
:net
/ipv4
/netfilter
/ip_queue
.c
293 struct sk_buff
*nskb
;
294 >>>>>>> 264e3e889d86e552b4191d69bb60f4f3b383135a
:net
/ipv4
/netfilter
/ip_queue
.c
296 if (v
->data_len
< sizeof(*user_iph
))
298 diff
= v
->data_len
- e
->skb
->len
;
300 if (pskb_trim(e
->skb
, v
->data_len
))
302 } else if (diff
> 0) {
303 if (v
->data_len
> 0xFFFF)
305 if (diff
> skb_tailroom(e
->skb
)) {
306 <<<<<<< HEAD
:net
/ipv4
/netfilter
/ip_queue
.c
307 err
= pskb_expand_head(e
->skb
, 0,
309 nskb
= skb_copy_expand(e
->skb
, 0,
310 >>>>>>> 264e3e889d86e552b4191d69bb60f4f3b383135a
:net
/ipv4
/netfilter
/ip_queue
.c
311 diff
- skb_tailroom(e
->skb
),
313 <<<<<<< HEAD
:net
/ipv4
/netfilter
/ip_queue
.c
317 >>>>>>> 264e3e889d86e552b4191d69bb60f4f3b383135a
:net
/ipv4
/netfilter
/ip_queue
.c
318 printk(KERN_WARNING
"ip_queue: error "
319 <<<<<<< HEAD
:net
/ipv4
/netfilter
/ip_queue
.c
320 "in mangle, dropping packet: %d\n", -err
);
323 "in mangle, dropping packet\n");
325 >>>>>>> 264e3e889d86e552b4191d69bb60f4f3b383135a
:net
/ipv4
/netfilter
/ip_queue
.c
327 <<<<<<< HEAD
:net
/ipv4
/netfilter
/ip_queue
.c
331 >>>>>>> 264e3e889d86e552b4191d69bb60f4f3b383135a
:net
/ipv4
/netfilter
/ip_queue
.c
333 skb_put(e
->skb
, diff
);
335 if (!skb_make_writable(e
->skb
, v
->data_len
))
337 skb_copy_to_linear_data(e
->skb
, v
->payload
, v
->data_len
);
338 e
->skb
->ip_summed
= CHECKSUM_NONE
;
344 ipq_set_verdict(struct ipq_verdict_msg
*vmsg
, unsigned int len
)
346 struct nf_queue_entry
*entry
;
348 if (vmsg
->value
> NF_MAX_VERDICT
)
351 entry
= ipq_find_dequeue_entry(vmsg
->id
);
355 int verdict
= vmsg
->value
;
357 if (vmsg
->data_len
&& vmsg
->data_len
== len
)
358 if (ipq_mangle_ipv4(vmsg
, entry
) < 0)
361 nf_reinject(entry
, verdict
);
367 ipq_set_mode(unsigned char mode
, unsigned int range
)
371 write_lock_bh(&queue_lock
);
372 status
= __ipq_set_mode(mode
, range
);
373 write_unlock_bh(&queue_lock
);
378 ipq_receive_peer(struct ipq_peer_msg
*pmsg
,
379 unsigned char type
, unsigned int len
)
383 if (len
< sizeof(*pmsg
))
388 status
= ipq_set_mode(pmsg
->msg
.mode
.value
,
389 pmsg
->msg
.mode
.range
);
393 if (pmsg
->msg
.verdict
.value
> NF_MAX_VERDICT
)
396 status
= ipq_set_verdict(&pmsg
->msg
.verdict
,
397 len
- sizeof(*pmsg
));
406 dev_cmp(struct nf_queue_entry
*entry
, unsigned long ifindex
)
409 if (entry
->indev
->ifindex
== ifindex
)
412 if (entry
->outdev
->ifindex
== ifindex
)
414 #ifdef CONFIG_BRIDGE_NETFILTER
415 if (entry
->skb
->nf_bridge
) {
416 if (entry
->skb
->nf_bridge
->physindev
&&
417 entry
->skb
->nf_bridge
->physindev
->ifindex
== ifindex
)
419 if (entry
->skb
->nf_bridge
->physoutdev
&&
420 entry
->skb
->nf_bridge
->physoutdev
->ifindex
== ifindex
)
428 ipq_dev_drop(int ifindex
)
430 ipq_flush(dev_cmp
, ifindex
);
433 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
436 __ipq_rcv_skb(struct sk_buff
*skb
)
438 int status
, type
, pid
, flags
, nlmsglen
, skblen
;
439 struct nlmsghdr
*nlh
;
442 if (skblen
< sizeof(*nlh
))
445 nlh
= nlmsg_hdr(skb
);
446 nlmsglen
= nlh
->nlmsg_len
;
447 if (nlmsglen
< sizeof(*nlh
) || skblen
< nlmsglen
)
450 pid
= nlh
->nlmsg_pid
;
451 flags
= nlh
->nlmsg_flags
;
453 if(pid
<= 0 || !(flags
& NLM_F_REQUEST
) || flags
& NLM_F_MULTI
)
454 RCV_SKB_FAIL(-EINVAL
);
456 if (flags
& MSG_TRUNC
)
457 RCV_SKB_FAIL(-ECOMM
);
459 type
= nlh
->nlmsg_type
;
460 if (type
< NLMSG_NOOP
|| type
>= IPQM_MAX
)
461 RCV_SKB_FAIL(-EINVAL
);
463 if (type
<= IPQM_BASE
)
466 if (security_netlink_recv(skb
, CAP_NET_ADMIN
))
467 RCV_SKB_FAIL(-EPERM
);
469 write_lock_bh(&queue_lock
);
472 if (peer_pid
!= pid
) {
473 write_unlock_bh(&queue_lock
);
474 RCV_SKB_FAIL(-EBUSY
);
477 net_enable_timestamp();
481 write_unlock_bh(&queue_lock
);
483 status
= ipq_receive_peer(NLMSG_DATA(nlh
), type
,
484 nlmsglen
- NLMSG_LENGTH(0));
486 RCV_SKB_FAIL(status
);
488 if (flags
& NLM_F_ACK
)
489 netlink_ack(skb
, nlh
, 0);
494 ipq_rcv_skb(struct sk_buff
*skb
)
496 mutex_lock(&ipqnl_mutex
);
498 mutex_unlock(&ipqnl_mutex
);
502 ipq_rcv_dev_event(struct notifier_block
*this,
503 unsigned long event
, void *ptr
)
505 struct net_device
*dev
= ptr
;
507 if (dev
->nd_net
!= &init_net
)
510 /* Drop any packets associated with the downed device */
511 if (event
== NETDEV_DOWN
)
512 ipq_dev_drop(dev
->ifindex
);
516 static struct notifier_block ipq_dev_notifier
= {
517 .notifier_call
= ipq_rcv_dev_event
,
521 ipq_rcv_nl_event(struct notifier_block
*this,
522 unsigned long event
, void *ptr
)
524 struct netlink_notify
*n
= ptr
;
526 if (event
== NETLINK_URELEASE
&&
527 n
->protocol
== NETLINK_FIREWALL
&& n
->pid
) {
528 write_lock_bh(&queue_lock
);
529 if ((n
->net
== &init_net
) && (n
->pid
== peer_pid
))
531 write_unlock_bh(&queue_lock
);
536 static struct notifier_block ipq_nl_notifier
= {
537 .notifier_call
= ipq_rcv_nl_event
,
541 static struct ctl_table_header
*ipq_sysctl_header
;
543 static ctl_table ipq_table
[] = {
545 .ctl_name
= NET_IPQ_QMAX
,
546 .procname
= NET_IPQ_QMAX_NAME
,
547 .data
= &queue_maxlen
,
548 .maxlen
= sizeof(queue_maxlen
),
550 .proc_handler
= proc_dointvec
556 #ifdef CONFIG_PROC_FS
557 static int ip_queue_show(struct seq_file
*m
, void *v
)
559 read_lock_bh(&queue_lock
);
565 "Queue length : %u\n"
566 "Queue max. length : %u\n"
567 "Queue dropped : %u\n"
568 "Netlink dropped : %u\n",
577 read_unlock_bh(&queue_lock
);
581 static int ip_queue_open(struct inode
*inode
, struct file
*file
)
583 return single_open(file
, ip_queue_show
, NULL
);
586 static const struct file_operations ip_queue_proc_fops
= {
587 .open
= ip_queue_open
,
590 .release
= single_release
,
591 .owner
= THIS_MODULE
,
595 static const struct nf_queue_handler nfqh
= {
597 .outfn
= &ipq_enqueue_packet
,
600 static int __init
ip_queue_init(void)
602 int status
= -ENOMEM
;
603 struct proc_dir_entry
*proc __maybe_unused
;
605 netlink_register_notifier(&ipq_nl_notifier
);
606 ipqnl
= netlink_kernel_create(&init_net
, NETLINK_FIREWALL
, 0,
607 ipq_rcv_skb
, NULL
, THIS_MODULE
);
609 printk(KERN_ERR
"ip_queue: failed to create netlink socket\n");
610 goto cleanup_netlink_notifier
;
613 #ifdef CONFIG_PROC_FS
614 proc
= create_proc_entry(IPQ_PROC_FS_NAME
, 0, init_net
.proc_net
);
616 proc
->owner
= THIS_MODULE
;
617 proc
->proc_fops
= &ip_queue_proc_fops
;
619 printk(KERN_ERR
"ip_queue: failed to create proc entry\n");
623 register_netdevice_notifier(&ipq_dev_notifier
);
625 ipq_sysctl_header
= register_sysctl_paths(net_ipv4_ctl_path
, ipq_table
);
627 status
= nf_register_queue_handler(PF_INET
, &nfqh
);
629 printk(KERN_ERR
"ip_queue: failed to register queue handler\n");
636 unregister_sysctl_table(ipq_sysctl_header
);
638 unregister_netdevice_notifier(&ipq_dev_notifier
);
639 proc_net_remove(&init_net
, IPQ_PROC_FS_NAME
);
640 cleanup_ipqnl
: __maybe_unused
641 netlink_kernel_release(ipqnl
);
642 mutex_lock(&ipqnl_mutex
);
643 mutex_unlock(&ipqnl_mutex
);
645 cleanup_netlink_notifier
:
646 netlink_unregister_notifier(&ipq_nl_notifier
);
650 static void __exit
ip_queue_fini(void)
652 nf_unregister_queue_handlers(&nfqh
);
657 unregister_sysctl_table(ipq_sysctl_header
);
659 unregister_netdevice_notifier(&ipq_dev_notifier
);
660 proc_net_remove(&init_net
, IPQ_PROC_FS_NAME
);
662 netlink_kernel_release(ipqnl
);
663 mutex_lock(&ipqnl_mutex
);
664 mutex_unlock(&ipqnl_mutex
);
666 netlink_unregister_notifier(&ipq_nl_notifier
);
669 MODULE_DESCRIPTION("IPv4 packet queue handler");
670 MODULE_AUTHOR("James Morris <jmorris@intercode.com.au>");
671 MODULE_LICENSE("GPL");
673 module_init(ip_queue_init
);
674 module_exit(ip_queue_fini
);