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/net.h>
28 #include <linux/mutex.h>
29 #include <linux/slab.h>
30 #include <net/net_namespace.h>
32 #include <net/route.h>
33 #include <net/netfilter/nf_queue.h>
36 #define IPQ_QMAX_DEFAULT 1024
37 #define IPQ_PROC_FS_NAME "ip_queue"
38 #define NET_IPQ_QMAX 2088
39 #define NET_IPQ_QMAX_NAME "ip_queue_maxlen"
41 typedef int (*ipq_cmpfn
)(struct nf_queue_entry
*, unsigned long);
43 static unsigned char copy_mode __read_mostly
= IPQ_COPY_NONE
;
44 static unsigned int queue_maxlen __read_mostly
= IPQ_QMAX_DEFAULT
;
45 static DEFINE_RWLOCK(queue_lock
);
46 static int peer_pid __read_mostly
;
47 static unsigned int copy_range __read_mostly
;
48 static unsigned int queue_total
;
49 static unsigned int queue_dropped
= 0;
50 static unsigned int queue_user_dropped
= 0;
51 static struct sock
*ipqnl __read_mostly
;
52 static LIST_HEAD(queue_list
);
53 static DEFINE_MUTEX(ipqnl_mutex
);
56 __ipq_enqueue_entry(struct nf_queue_entry
*entry
)
58 list_add_tail(&entry
->list
, &queue_list
);
63 __ipq_set_mode(unsigned char mode
, unsigned int range
)
77 if (copy_range
> 0xFFFF)
88 static void __ipq_flush(ipq_cmpfn cmpfn
, unsigned long data
);
94 net_disable_timestamp();
95 __ipq_set_mode(IPQ_COPY_NONE
, 0);
99 static struct nf_queue_entry
*
100 ipq_find_dequeue_entry(unsigned long id
)
102 struct nf_queue_entry
*entry
= NULL
, *i
;
104 write_lock_bh(&queue_lock
);
106 list_for_each_entry(i
, &queue_list
, list
) {
107 if ((unsigned long)i
== id
) {
114 list_del(&entry
->list
);
118 write_unlock_bh(&queue_lock
);
123 __ipq_flush(ipq_cmpfn cmpfn
, unsigned long data
)
125 struct nf_queue_entry
*entry
, *next
;
127 list_for_each_entry_safe(entry
, next
, &queue_list
, list
) {
128 if (!cmpfn
|| cmpfn(entry
, data
)) {
129 list_del(&entry
->list
);
131 nf_reinject(entry
, NF_DROP
);
137 ipq_flush(ipq_cmpfn cmpfn
, unsigned long data
)
139 write_lock_bh(&queue_lock
);
140 __ipq_flush(cmpfn
, data
);
141 write_unlock_bh(&queue_lock
);
144 static struct sk_buff
*
145 ipq_build_packet_message(struct nf_queue_entry
*entry
, int *errp
)
147 sk_buff_data_t old_tail
;
151 struct ipq_packet_msg
*pmsg
;
152 struct nlmsghdr
*nlh
;
155 read_lock_bh(&queue_lock
);
160 size
= NLMSG_SPACE(sizeof(*pmsg
));
163 case IPQ_COPY_PACKET
:
164 if (entry
->skb
->ip_summed
== CHECKSUM_PARTIAL
&&
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
;
228 printk(KERN_ERR
"ip_queue: error creating packet message\n");
233 ipq_enqueue_packet(struct nf_queue_entry
*entry
, unsigned int queuenum
)
235 int status
= -EINVAL
;
236 struct sk_buff
*nskb
;
238 if (copy_mode
== IPQ_COPY_NONE
)
241 nskb
= ipq_build_packet_message(entry
, &status
);
245 write_lock_bh(&queue_lock
);
248 goto err_out_free_nskb
;
250 if (queue_total
>= queue_maxlen
) {
254 printk (KERN_WARNING
"ip_queue: full at %d entries, "
255 "dropping packets(s). Dropped: %d\n", queue_total
,
257 goto err_out_free_nskb
;
260 /* netlink_unicast will either free the nskb or attach it to a socket */
261 status
= netlink_unicast(ipqnl
, nskb
, peer_pid
, MSG_DONTWAIT
);
263 queue_user_dropped
++;
267 __ipq_enqueue_entry(entry
);
269 write_unlock_bh(&queue_lock
);
276 write_unlock_bh(&queue_lock
);
281 ipq_mangle_ipv4(ipq_verdict_msg_t
*v
, struct nf_queue_entry
*e
)
284 struct iphdr
*user_iph
= (struct iphdr
*)v
->payload
;
285 struct sk_buff
*nskb
;
287 if (v
->data_len
< sizeof(*user_iph
))
289 diff
= v
->data_len
- e
->skb
->len
;
291 if (pskb_trim(e
->skb
, v
->data_len
))
293 } else if (diff
> 0) {
294 if (v
->data_len
> 0xFFFF)
296 if (diff
> skb_tailroom(e
->skb
)) {
297 nskb
= skb_copy_expand(e
->skb
, skb_headroom(e
->skb
),
300 printk(KERN_WARNING
"ip_queue: error "
301 "in mangle, dropping packet\n");
307 skb_put(e
->skb
, diff
);
309 if (!skb_make_writable(e
->skb
, v
->data_len
))
311 skb_copy_to_linear_data(e
->skb
, v
->payload
, v
->data_len
);
312 e
->skb
->ip_summed
= CHECKSUM_NONE
;
318 ipq_set_verdict(struct ipq_verdict_msg
*vmsg
, unsigned int len
)
320 struct nf_queue_entry
*entry
;
322 if (vmsg
->value
> NF_MAX_VERDICT
)
325 entry
= ipq_find_dequeue_entry(vmsg
->id
);
329 int verdict
= vmsg
->value
;
331 if (vmsg
->data_len
&& vmsg
->data_len
== len
)
332 if (ipq_mangle_ipv4(vmsg
, entry
) < 0)
335 nf_reinject(entry
, verdict
);
341 ipq_set_mode(unsigned char mode
, unsigned int range
)
345 write_lock_bh(&queue_lock
);
346 status
= __ipq_set_mode(mode
, range
);
347 write_unlock_bh(&queue_lock
);
352 ipq_receive_peer(struct ipq_peer_msg
*pmsg
,
353 unsigned char type
, unsigned int len
)
357 if (len
< sizeof(*pmsg
))
362 status
= ipq_set_mode(pmsg
->msg
.mode
.value
,
363 pmsg
->msg
.mode
.range
);
367 if (pmsg
->msg
.verdict
.value
> NF_MAX_VERDICT
)
370 status
= ipq_set_verdict(&pmsg
->msg
.verdict
,
371 len
- sizeof(*pmsg
));
380 dev_cmp(struct nf_queue_entry
*entry
, unsigned long ifindex
)
383 if (entry
->indev
->ifindex
== ifindex
)
386 if (entry
->outdev
->ifindex
== ifindex
)
388 #ifdef CONFIG_BRIDGE_NETFILTER
389 if (entry
->skb
->nf_bridge
) {
390 if (entry
->skb
->nf_bridge
->physindev
&&
391 entry
->skb
->nf_bridge
->physindev
->ifindex
== ifindex
)
393 if (entry
->skb
->nf_bridge
->physoutdev
&&
394 entry
->skb
->nf_bridge
->physoutdev
->ifindex
== ifindex
)
402 ipq_dev_drop(int ifindex
)
404 ipq_flush(dev_cmp
, ifindex
);
407 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
410 __ipq_rcv_skb(struct sk_buff
*skb
)
412 int status
, type
, pid
, flags
, nlmsglen
, skblen
;
413 struct nlmsghdr
*nlh
;
416 if (skblen
< sizeof(*nlh
))
419 nlh
= nlmsg_hdr(skb
);
420 nlmsglen
= nlh
->nlmsg_len
;
421 if (nlmsglen
< sizeof(*nlh
) || skblen
< nlmsglen
)
424 pid
= nlh
->nlmsg_pid
;
425 flags
= nlh
->nlmsg_flags
;
427 if(pid
<= 0 || !(flags
& NLM_F_REQUEST
) || flags
& NLM_F_MULTI
)
428 RCV_SKB_FAIL(-EINVAL
);
430 if (flags
& MSG_TRUNC
)
431 RCV_SKB_FAIL(-ECOMM
);
433 type
= nlh
->nlmsg_type
;
434 if (type
< NLMSG_NOOP
|| type
>= IPQM_MAX
)
435 RCV_SKB_FAIL(-EINVAL
);
437 if (type
<= IPQM_BASE
)
440 if (security_netlink_recv(skb
, CAP_NET_ADMIN
))
441 RCV_SKB_FAIL(-EPERM
);
443 write_lock_bh(&queue_lock
);
446 if (peer_pid
!= pid
) {
447 write_unlock_bh(&queue_lock
);
448 RCV_SKB_FAIL(-EBUSY
);
451 net_enable_timestamp();
455 write_unlock_bh(&queue_lock
);
457 status
= ipq_receive_peer(NLMSG_DATA(nlh
), type
,
458 nlmsglen
- NLMSG_LENGTH(0));
460 RCV_SKB_FAIL(status
);
462 if (flags
& NLM_F_ACK
)
463 netlink_ack(skb
, nlh
, 0);
467 ipq_rcv_skb(struct sk_buff
*skb
)
469 mutex_lock(&ipqnl_mutex
);
471 mutex_unlock(&ipqnl_mutex
);
475 ipq_rcv_dev_event(struct notifier_block
*this,
476 unsigned long event
, void *ptr
)
478 struct net_device
*dev
= ptr
;
480 if (!net_eq(dev_net(dev
), &init_net
))
483 /* Drop any packets associated with the downed device */
484 if (event
== NETDEV_DOWN
)
485 ipq_dev_drop(dev
->ifindex
);
489 static struct notifier_block ipq_dev_notifier
= {
490 .notifier_call
= ipq_rcv_dev_event
,
494 ipq_rcv_nl_event(struct notifier_block
*this,
495 unsigned long event
, void *ptr
)
497 struct netlink_notify
*n
= ptr
;
499 if (event
== NETLINK_URELEASE
&& n
->protocol
== NETLINK_FIREWALL
) {
500 write_lock_bh(&queue_lock
);
501 if ((net_eq(n
->net
, &init_net
)) && (n
->pid
== peer_pid
))
503 write_unlock_bh(&queue_lock
);
508 static struct notifier_block ipq_nl_notifier
= {
509 .notifier_call
= ipq_rcv_nl_event
,
513 static struct ctl_table_header
*ipq_sysctl_header
;
515 static ctl_table ipq_table
[] = {
517 .procname
= NET_IPQ_QMAX_NAME
,
518 .data
= &queue_maxlen
,
519 .maxlen
= sizeof(queue_maxlen
),
521 .proc_handler
= proc_dointvec
527 #ifdef CONFIG_PROC_FS
528 static int ip_queue_show(struct seq_file
*m
, void *v
)
530 read_lock_bh(&queue_lock
);
536 "Queue length : %u\n"
537 "Queue max. length : %u\n"
538 "Queue dropped : %u\n"
539 "Netlink dropped : %u\n",
548 read_unlock_bh(&queue_lock
);
552 static int ip_queue_open(struct inode
*inode
, struct file
*file
)
554 return single_open(file
, ip_queue_show
, NULL
);
557 static const struct file_operations ip_queue_proc_fops
= {
558 .open
= ip_queue_open
,
561 .release
= single_release
,
562 .owner
= THIS_MODULE
,
566 static const struct nf_queue_handler nfqh
= {
568 .outfn
= &ipq_enqueue_packet
,
571 static int __init
ip_queue_init(void)
573 int status
= -ENOMEM
;
574 struct proc_dir_entry
*proc __maybe_unused
;
576 netlink_register_notifier(&ipq_nl_notifier
);
577 ipqnl
= netlink_kernel_create(&init_net
, NETLINK_FIREWALL
, 0,
578 ipq_rcv_skb
, NULL
, THIS_MODULE
);
580 printk(KERN_ERR
"ip_queue: failed to create netlink socket\n");
581 goto cleanup_netlink_notifier
;
584 #ifdef CONFIG_PROC_FS
585 proc
= proc_create(IPQ_PROC_FS_NAME
, 0, init_net
.proc_net
,
586 &ip_queue_proc_fops
);
588 printk(KERN_ERR
"ip_queue: failed to create proc entry\n");
592 register_netdevice_notifier(&ipq_dev_notifier
);
594 ipq_sysctl_header
= register_sysctl_paths(net_ipv4_ctl_path
, ipq_table
);
596 status
= nf_register_queue_handler(NFPROTO_IPV4
, &nfqh
);
598 printk(KERN_ERR
"ip_queue: failed to register queue handler\n");
605 unregister_sysctl_table(ipq_sysctl_header
);
607 unregister_netdevice_notifier(&ipq_dev_notifier
);
608 proc_net_remove(&init_net
, IPQ_PROC_FS_NAME
);
609 cleanup_ipqnl
: __maybe_unused
610 netlink_kernel_release(ipqnl
);
611 mutex_lock(&ipqnl_mutex
);
612 mutex_unlock(&ipqnl_mutex
);
614 cleanup_netlink_notifier
:
615 netlink_unregister_notifier(&ipq_nl_notifier
);
619 static void __exit
ip_queue_fini(void)
621 nf_unregister_queue_handlers(&nfqh
);
626 unregister_sysctl_table(ipq_sysctl_header
);
628 unregister_netdevice_notifier(&ipq_dev_notifier
);
629 proc_net_remove(&init_net
, IPQ_PROC_FS_NAME
);
631 netlink_kernel_release(ipqnl
);
632 mutex_lock(&ipqnl_mutex
);
633 mutex_unlock(&ipqnl_mutex
);
635 netlink_unregister_notifier(&ipq_nl_notifier
);
638 MODULE_DESCRIPTION("IPv4 packet queue handler");
639 MODULE_AUTHOR("James Morris <jmorris@intercode.com.au>");
640 MODULE_LICENSE("GPL");
641 MODULE_ALIAS_NET_PF_PROTO(PF_NETLINK
, NETLINK_FIREWALL
);
643 module_init(ip_queue_init
);
644 module_exit(ip_queue_fini
);