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 struct iphdr
*user_iph
= (struct iphdr
*)v
->payload
;
287 struct sk_buff
*nskb
;
289 if (v
->data_len
< sizeof(*user_iph
))
291 diff
= v
->data_len
- e
->skb
->len
;
293 if (pskb_trim(e
->skb
, v
->data_len
))
295 } else if (diff
> 0) {
296 if (v
->data_len
> 0xFFFF)
298 if (diff
> skb_tailroom(e
->skb
)) {
299 nskb
= skb_copy_expand(e
->skb
, skb_headroom(e
->skb
),
302 printk(KERN_WARNING
"ip_queue: error "
303 "in mangle, dropping packet\n");
309 skb_put(e
->skb
, diff
);
311 if (!skb_make_writable(e
->skb
, v
->data_len
))
313 skb_copy_to_linear_data(e
->skb
, v
->payload
, v
->data_len
);
314 e
->skb
->ip_summed
= CHECKSUM_NONE
;
320 ipq_set_verdict(struct ipq_verdict_msg
*vmsg
, unsigned int len
)
322 struct nf_queue_entry
*entry
;
324 if (vmsg
->value
> NF_MAX_VERDICT
)
327 entry
= ipq_find_dequeue_entry(vmsg
->id
);
331 int verdict
= vmsg
->value
;
333 if (vmsg
->data_len
&& vmsg
->data_len
== len
)
334 if (ipq_mangle_ipv4(vmsg
, entry
) < 0)
337 nf_reinject(entry
, verdict
);
343 ipq_set_mode(unsigned char mode
, unsigned int range
)
347 write_lock_bh(&queue_lock
);
348 status
= __ipq_set_mode(mode
, range
);
349 write_unlock_bh(&queue_lock
);
354 ipq_receive_peer(struct ipq_peer_msg
*pmsg
,
355 unsigned char type
, unsigned int len
)
359 if (len
< sizeof(*pmsg
))
364 status
= ipq_set_mode(pmsg
->msg
.mode
.value
,
365 pmsg
->msg
.mode
.range
);
369 if (pmsg
->msg
.verdict
.value
> NF_MAX_VERDICT
)
372 status
= ipq_set_verdict(&pmsg
->msg
.verdict
,
373 len
- sizeof(*pmsg
));
382 dev_cmp(struct nf_queue_entry
*entry
, unsigned long ifindex
)
385 if (entry
->indev
->ifindex
== ifindex
)
388 if (entry
->outdev
->ifindex
== ifindex
)
390 #ifdef CONFIG_BRIDGE_NETFILTER
391 if (entry
->skb
->nf_bridge
) {
392 if (entry
->skb
->nf_bridge
->physindev
&&
393 entry
->skb
->nf_bridge
->physindev
->ifindex
== ifindex
)
395 if (entry
->skb
->nf_bridge
->physoutdev
&&
396 entry
->skb
->nf_bridge
->physoutdev
->ifindex
== ifindex
)
404 ipq_dev_drop(int ifindex
)
406 ipq_flush(dev_cmp
, ifindex
);
409 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
412 __ipq_rcv_skb(struct sk_buff
*skb
)
414 int status
, type
, pid
, flags
, nlmsglen
, skblen
;
415 struct nlmsghdr
*nlh
;
418 if (skblen
< sizeof(*nlh
))
421 nlh
= nlmsg_hdr(skb
);
422 nlmsglen
= nlh
->nlmsg_len
;
423 if (nlmsglen
< sizeof(*nlh
) || skblen
< nlmsglen
)
426 pid
= nlh
->nlmsg_pid
;
427 flags
= nlh
->nlmsg_flags
;
429 if(pid
<= 0 || !(flags
& NLM_F_REQUEST
) || flags
& NLM_F_MULTI
)
430 RCV_SKB_FAIL(-EINVAL
);
432 if (flags
& MSG_TRUNC
)
433 RCV_SKB_FAIL(-ECOMM
);
435 type
= nlh
->nlmsg_type
;
436 if (type
< NLMSG_NOOP
|| type
>= IPQM_MAX
)
437 RCV_SKB_FAIL(-EINVAL
);
439 if (type
<= IPQM_BASE
)
442 if (security_netlink_recv(skb
, CAP_NET_ADMIN
))
443 RCV_SKB_FAIL(-EPERM
);
445 write_lock_bh(&queue_lock
);
448 if (peer_pid
!= pid
) {
449 write_unlock_bh(&queue_lock
);
450 RCV_SKB_FAIL(-EBUSY
);
453 net_enable_timestamp();
457 write_unlock_bh(&queue_lock
);
459 status
= ipq_receive_peer(NLMSG_DATA(nlh
), type
,
460 nlmsglen
- NLMSG_LENGTH(0));
462 RCV_SKB_FAIL(status
);
464 if (flags
& NLM_F_ACK
)
465 netlink_ack(skb
, nlh
, 0);
470 ipq_rcv_skb(struct sk_buff
*skb
)
472 mutex_lock(&ipqnl_mutex
);
474 mutex_unlock(&ipqnl_mutex
);
478 ipq_rcv_dev_event(struct notifier_block
*this,
479 unsigned long event
, void *ptr
)
481 struct net_device
*dev
= ptr
;
483 if (dev_net(dev
) != &init_net
)
486 /* Drop any packets associated with the downed device */
487 if (event
== NETDEV_DOWN
)
488 ipq_dev_drop(dev
->ifindex
);
492 static struct notifier_block ipq_dev_notifier
= {
493 .notifier_call
= ipq_rcv_dev_event
,
497 ipq_rcv_nl_event(struct notifier_block
*this,
498 unsigned long event
, void *ptr
)
500 struct netlink_notify
*n
= ptr
;
502 if (event
== NETLINK_URELEASE
&&
503 n
->protocol
== NETLINK_FIREWALL
&& n
->pid
) {
504 write_lock_bh(&queue_lock
);
505 if ((n
->net
== &init_net
) && (n
->pid
== peer_pid
))
507 write_unlock_bh(&queue_lock
);
512 static struct notifier_block ipq_nl_notifier
= {
513 .notifier_call
= ipq_rcv_nl_event
,
517 static struct ctl_table_header
*ipq_sysctl_header
;
519 static ctl_table ipq_table
[] = {
521 .ctl_name
= NET_IPQ_QMAX
,
522 .procname
= NET_IPQ_QMAX_NAME
,
523 .data
= &queue_maxlen
,
524 .maxlen
= sizeof(queue_maxlen
),
526 .proc_handler
= proc_dointvec
532 #ifdef CONFIG_PROC_FS
533 static int ip_queue_show(struct seq_file
*m
, void *v
)
535 read_lock_bh(&queue_lock
);
541 "Queue length : %u\n"
542 "Queue max. length : %u\n"
543 "Queue dropped : %u\n"
544 "Netlink dropped : %u\n",
553 read_unlock_bh(&queue_lock
);
557 static int ip_queue_open(struct inode
*inode
, struct file
*file
)
559 return single_open(file
, ip_queue_show
, NULL
);
562 static const struct file_operations ip_queue_proc_fops
= {
563 .open
= ip_queue_open
,
566 .release
= single_release
,
567 .owner
= THIS_MODULE
,
571 static const struct nf_queue_handler nfqh
= {
573 .outfn
= &ipq_enqueue_packet
,
576 static int __init
ip_queue_init(void)
578 int status
= -ENOMEM
;
579 struct proc_dir_entry
*proc __maybe_unused
;
581 netlink_register_notifier(&ipq_nl_notifier
);
582 ipqnl
= netlink_kernel_create(&init_net
, NETLINK_FIREWALL
, 0,
583 ipq_rcv_skb
, NULL
, THIS_MODULE
);
585 printk(KERN_ERR
"ip_queue: failed to create netlink socket\n");
586 goto cleanup_netlink_notifier
;
589 #ifdef CONFIG_PROC_FS
590 proc
= proc_create(IPQ_PROC_FS_NAME
, 0, init_net
.proc_net
,
591 &ip_queue_proc_fops
);
593 printk(KERN_ERR
"ip_queue: failed to create proc entry\n");
597 register_netdevice_notifier(&ipq_dev_notifier
);
599 ipq_sysctl_header
= register_sysctl_paths(net_ipv4_ctl_path
, ipq_table
);
601 status
= nf_register_queue_handler(PF_INET
, &nfqh
);
603 printk(KERN_ERR
"ip_queue: failed to register queue handler\n");
610 unregister_sysctl_table(ipq_sysctl_header
);
612 unregister_netdevice_notifier(&ipq_dev_notifier
);
613 proc_net_remove(&init_net
, IPQ_PROC_FS_NAME
);
614 cleanup_ipqnl
: __maybe_unused
615 netlink_kernel_release(ipqnl
);
616 mutex_lock(&ipqnl_mutex
);
617 mutex_unlock(&ipqnl_mutex
);
619 cleanup_netlink_notifier
:
620 netlink_unregister_notifier(&ipq_nl_notifier
);
624 static void __exit
ip_queue_fini(void)
626 nf_unregister_queue_handlers(&nfqh
);
631 unregister_sysctl_table(ipq_sysctl_header
);
633 unregister_netdevice_notifier(&ipq_dev_notifier
);
634 proc_net_remove(&init_net
, IPQ_PROC_FS_NAME
);
636 netlink_kernel_release(ipqnl
);
637 mutex_lock(&ipqnl_mutex
);
638 mutex_unlock(&ipqnl_mutex
);
640 netlink_unregister_notifier(&ipq_nl_notifier
);
643 MODULE_DESCRIPTION("IPv4 packet queue handler");
644 MODULE_AUTHOR("James Morris <jmorris@intercode.com.au>");
645 MODULE_LICENSE("GPL");
647 module_init(ip_queue_init
);
648 module_exit(ip_queue_fini
);