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>
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
11 * 2000-03-27: Simplified code (thanks to Andi Kleen for clues).
12 * 2000-05-20: Fixed notifier problems (following Miguel Freitas' report).
13 * 2000-06-19: Fixed so nfmark is copied to metadata (reported by Sebastian
15 * 2000-08-01: Added Nick Williams' MAC support.
16 * 2002-06-25: Code cleanup.
17 * 2005-01-10: Added /proc counter for dropped packets; fixed so
18 * packets aren't delivered to user space if they're going
22 #include <linux/module.h>
23 #include <linux/skbuff.h>
24 #include <linux/init.h>
26 #include <linux/notifier.h>
27 #include <linux/netdevice.h>
28 #include <linux/netfilter.h>
29 #include <linux/netfilter_ipv4/ip_queue.h>
30 #include <linux/netfilter_ipv4/ip_tables.h>
31 #include <linux/netlink.h>
32 #include <linux/spinlock.h>
33 #include <linux/sysctl.h>
34 #include <linux/proc_fs.h>
35 #include <linux/security.h>
37 #include <net/route.h>
39 #define IPQ_QMAX_DEFAULT 1024
40 #define IPQ_PROC_FS_NAME "ip_queue"
41 #define NET_IPQ_QMAX 2088
42 #define NET_IPQ_QMAX_NAME "ip_queue_maxlen"
50 struct ipq_queue_entry
{
51 struct list_head list
;
54 struct ipq_rt_info rt_info
;
57 typedef int (*ipq_cmpfn
)(struct ipq_queue_entry
*, unsigned long);
59 static unsigned char copy_mode
= IPQ_COPY_NONE
;
60 static unsigned int queue_maxlen
= IPQ_QMAX_DEFAULT
;
61 static DEFINE_RWLOCK(queue_lock
);
63 static unsigned int copy_range
;
64 static unsigned int queue_total
;
65 static unsigned int queue_dropped
= 0;
66 static unsigned int queue_user_dropped
= 0;
67 static struct sock
*ipqnl
;
68 static LIST_HEAD(queue_list
);
69 static DECLARE_MUTEX(ipqnl_sem
);
72 ipq_issue_verdict(struct ipq_queue_entry
*entry
, int verdict
)
74 nf_reinject(entry
->skb
, entry
->info
, verdict
);
79 __ipq_enqueue_entry(struct ipq_queue_entry
*entry
)
81 list_add(&entry
->list
, &queue_list
);
86 * Find and return a queued entry matched by cmpfn, or return the last
87 * entry if cmpfn is NULL.
89 static inline struct ipq_queue_entry
*
90 __ipq_find_entry(ipq_cmpfn cmpfn
, unsigned long data
)
94 list_for_each_prev(p
, &queue_list
) {
95 struct ipq_queue_entry
*entry
= (struct ipq_queue_entry
*)p
;
97 if (!cmpfn
|| cmpfn(entry
, data
))
104 __ipq_dequeue_entry(struct ipq_queue_entry
*entry
)
106 list_del(&entry
->list
);
110 static inline struct ipq_queue_entry
*
111 __ipq_find_dequeue_entry(ipq_cmpfn cmpfn
, unsigned long data
)
113 struct ipq_queue_entry
*entry
;
115 entry
= __ipq_find_entry(cmpfn
, data
);
119 __ipq_dequeue_entry(entry
);
125 __ipq_flush(int verdict
)
127 struct ipq_queue_entry
*entry
;
129 while ((entry
= __ipq_find_dequeue_entry(NULL
, 0)))
130 ipq_issue_verdict(entry
, verdict
);
134 __ipq_set_mode(unsigned char mode
, unsigned int range
)
145 case IPQ_COPY_PACKET
:
148 if (copy_range
> 0xFFFF)
163 net_disable_timestamp();
164 __ipq_set_mode(IPQ_COPY_NONE
, 0);
165 __ipq_flush(NF_DROP
);
168 static struct ipq_queue_entry
*
169 ipq_find_dequeue_entry(ipq_cmpfn cmpfn
, unsigned long data
)
171 struct ipq_queue_entry
*entry
;
173 write_lock_bh(&queue_lock
);
174 entry
= __ipq_find_dequeue_entry(cmpfn
, data
);
175 write_unlock_bh(&queue_lock
);
180 ipq_flush(int verdict
)
182 write_lock_bh(&queue_lock
);
183 __ipq_flush(verdict
);
184 write_unlock_bh(&queue_lock
);
187 static struct sk_buff
*
188 ipq_build_packet_message(struct ipq_queue_entry
*entry
, int *errp
)
190 unsigned char *old_tail
;
194 struct ipq_packet_msg
*pmsg
;
195 struct nlmsghdr
*nlh
;
197 read_lock_bh(&queue_lock
);
202 size
= NLMSG_SPACE(sizeof(*pmsg
));
206 case IPQ_COPY_PACKET
:
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
);
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 pmsg
->timestamp_sec
= entry
->skb
->stamp
.tv_sec
;
235 pmsg
->timestamp_usec
= entry
->skb
->stamp
.tv_usec
;
236 pmsg
->mark
= entry
->skb
->nfmark
;
237 pmsg
->hook
= entry
->info
->hook
;
238 pmsg
->hw_protocol
= entry
->skb
->protocol
;
240 if (entry
->info
->indev
)
241 strcpy(pmsg
->indev_name
, entry
->info
->indev
->name
);
243 pmsg
->indev_name
[0] = '\0';
245 if (entry
->info
->outdev
)
246 strcpy(pmsg
->outdev_name
, entry
->info
->outdev
->name
);
248 pmsg
->outdev_name
[0] = '\0';
250 if (entry
->info
->indev
&& entry
->skb
->dev
) {
251 pmsg
->hw_type
= entry
->skb
->dev
->type
;
252 if (entry
->skb
->dev
->hard_header_parse
)
254 entry
->skb
->dev
->hard_header_parse(entry
->skb
,
259 if (skb_copy_bits(entry
->skb
, 0, pmsg
->payload
, data_len
))
262 nlh
->nlmsg_len
= skb
->tail
- old_tail
;
269 printk(KERN_ERR
"ip_queue: error creating packet message\n");
274 ipq_enqueue_packet(struct sk_buff
*skb
, struct nf_info
*info
, void *data
)
276 int status
= -EINVAL
;
277 struct sk_buff
*nskb
;
278 struct ipq_queue_entry
*entry
;
280 if (copy_mode
== IPQ_COPY_NONE
)
283 entry
= kmalloc(sizeof(*entry
), GFP_ATOMIC
);
285 printk(KERN_ERR
"ip_queue: OOM in ipq_enqueue_packet()\n");
292 if (entry
->info
->hook
== NF_IP_LOCAL_OUT
) {
293 struct iphdr
*iph
= skb
->nh
.iph
;
295 entry
->rt_info
.tos
= iph
->tos
;
296 entry
->rt_info
.daddr
= iph
->daddr
;
297 entry
->rt_info
.saddr
= iph
->saddr
;
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
"ip_queue: full at %d entries, "
314 "dropping packets(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_ipv4(ipq_verdict_msg_t
*v
, struct ipq_queue_entry
*e
)
346 struct iphdr
*user_iph
= (struct iphdr
*)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
"ip_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_ip_make_writable(&e
->skb
, v
->data_len
))
377 memcpy(e
->skb
->data
, v
->payload
, v
->data_len
);
378 e
->skb
->nfcache
|= NFC_ALTERED
;
381 * Extra routing may needed on local out, as the QUEUE target never
382 * returns control to the table.
384 if (e
->info
->hook
== NF_IP_LOCAL_OUT
) {
385 struct iphdr
*iph
= e
->skb
->nh
.iph
;
387 if (!(iph
->tos
== e
->rt_info
.tos
388 && iph
->daddr
== e
->rt_info
.daddr
389 && iph
->saddr
== e
->rt_info
.saddr
))
390 return ip_route_me_harder(&e
->skb
);
396 id_cmp(struct ipq_queue_entry
*e
, unsigned long id
)
398 return (id
== (unsigned long )e
);
402 ipq_set_verdict(struct ipq_verdict_msg
*vmsg
, unsigned int len
)
404 struct ipq_queue_entry
*entry
;
406 if (vmsg
->value
> NF_MAX_VERDICT
)
409 entry
= ipq_find_dequeue_entry(id_cmp
, vmsg
->id
);
413 int verdict
= vmsg
->value
;
415 if (vmsg
->data_len
&& vmsg
->data_len
== len
)
416 if (ipq_mangle_ipv4(vmsg
, entry
) < 0)
419 ipq_issue_verdict(entry
, verdict
);
425 ipq_set_mode(unsigned char mode
, unsigned int range
)
429 write_lock_bh(&queue_lock
);
430 status
= __ipq_set_mode(mode
, range
);
431 write_unlock_bh(&queue_lock
);
436 ipq_receive_peer(struct ipq_peer_msg
*pmsg
,
437 unsigned char type
, unsigned int len
)
441 if (len
< sizeof(*pmsg
))
446 status
= ipq_set_mode(pmsg
->msg
.mode
.value
,
447 pmsg
->msg
.mode
.range
);
451 if (pmsg
->msg
.verdict
.value
> NF_MAX_VERDICT
)
454 status
= ipq_set_verdict(&pmsg
->msg
.verdict
,
455 len
- sizeof(*pmsg
));
464 dev_cmp(struct ipq_queue_entry
*entry
, unsigned long ifindex
)
466 if (entry
->info
->indev
)
467 if (entry
->info
->indev
->ifindex
== ifindex
)
470 if (entry
->info
->outdev
)
471 if (entry
->info
->outdev
->ifindex
== ifindex
)
478 ipq_dev_drop(int ifindex
)
480 struct ipq_queue_entry
*entry
;
482 while ((entry
= ipq_find_dequeue_entry(dev_cmp
, ifindex
)) != NULL
)
483 ipq_issue_verdict(entry
, NF_DROP
);
486 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
489 ipq_rcv_skb(struct sk_buff
*skb
)
491 int status
, type
, pid
, flags
, nlmsglen
, skblen
;
492 struct nlmsghdr
*nlh
;
495 if (skblen
< sizeof(*nlh
))
498 nlh
= (struct nlmsghdr
*)skb
->data
;
499 nlmsglen
= nlh
->nlmsg_len
;
500 if (nlmsglen
< sizeof(*nlh
) || skblen
< nlmsglen
)
503 pid
= nlh
->nlmsg_pid
;
504 flags
= nlh
->nlmsg_flags
;
506 if(pid
<= 0 || !(flags
& NLM_F_REQUEST
) || flags
& NLM_F_MULTI
)
507 RCV_SKB_FAIL(-EINVAL
);
509 if (flags
& MSG_TRUNC
)
510 RCV_SKB_FAIL(-ECOMM
);
512 type
= nlh
->nlmsg_type
;
513 if (type
< NLMSG_NOOP
|| type
>= IPQM_MAX
)
514 RCV_SKB_FAIL(-EINVAL
);
516 if (type
<= IPQM_BASE
)
519 if (security_netlink_recv(skb
))
520 RCV_SKB_FAIL(-EPERM
);
522 write_lock_bh(&queue_lock
);
525 if (peer_pid
!= pid
) {
526 write_unlock_bh(&queue_lock
);
527 RCV_SKB_FAIL(-EBUSY
);
530 net_enable_timestamp();
534 write_unlock_bh(&queue_lock
);
536 status
= ipq_receive_peer(NLMSG_DATA(nlh
), type
,
537 skblen
- NLMSG_LENGTH(0));
539 RCV_SKB_FAIL(status
);
541 if (flags
& NLM_F_ACK
)
542 netlink_ack(skb
, nlh
, 0);
547 ipq_rcv_sk(struct sock
*sk
, int len
)
552 if (down_trylock(&ipqnl_sem
))
555 while ((skb
= skb_dequeue(&sk
->sk_receive_queue
)) != NULL
) {
562 } while (ipqnl
&& ipqnl
->sk_receive_queue
.qlen
);
566 ipq_rcv_dev_event(struct notifier_block
*this,
567 unsigned long event
, void *ptr
)
569 struct net_device
*dev
= ptr
;
571 /* Drop any packets associated with the downed device */
572 if (event
== NETDEV_DOWN
)
573 ipq_dev_drop(dev
->ifindex
);
577 static struct notifier_block ipq_dev_notifier
= {
578 .notifier_call
= ipq_rcv_dev_event
,
582 ipq_rcv_nl_event(struct notifier_block
*this,
583 unsigned long event
, void *ptr
)
585 struct netlink_notify
*n
= ptr
;
587 if (event
== NETLINK_URELEASE
&&
588 n
->protocol
== NETLINK_FIREWALL
&& n
->pid
) {
589 write_lock_bh(&queue_lock
);
590 if (n
->pid
== peer_pid
)
592 write_unlock_bh(&queue_lock
);
597 static struct notifier_block ipq_nl_notifier
= {
598 .notifier_call
= ipq_rcv_nl_event
,
601 static struct ctl_table_header
*ipq_sysctl_header
;
603 static ctl_table ipq_table
[] = {
605 .ctl_name
= NET_IPQ_QMAX
,
606 .procname
= NET_IPQ_QMAX_NAME
,
607 .data
= &queue_maxlen
,
608 .maxlen
= sizeof(queue_maxlen
),
610 .proc_handler
= proc_dointvec
615 static ctl_table ipq_dir_table
[] = {
617 .ctl_name
= NET_IPV4
,
625 static ctl_table ipq_root_table
[] = {
630 .child
= ipq_dir_table
635 #ifdef CONFIG_PROC_FS
637 ipq_get_info(char *buffer
, char **start
, off_t offset
, int length
)
641 read_lock_bh(&queue_lock
);
643 len
= sprintf(buffer
,
647 "Queue length : %u\n"
648 "Queue max. length : %u\n"
649 "Queue dropped : %u\n"
650 "Netlink dropped : %u\n",
659 read_unlock_bh(&queue_lock
);
661 *start
= buffer
+ offset
;
669 #endif /* CONFIG_PROC_FS */
672 init_or_cleanup(int init
)
674 int status
= -ENOMEM
;
675 struct proc_dir_entry
*proc
;
680 netlink_register_notifier(&ipq_nl_notifier
);
681 ipqnl
= netlink_kernel_create(NETLINK_FIREWALL
, ipq_rcv_sk
);
683 printk(KERN_ERR
"ip_queue: failed to create netlink socket\n");
684 goto cleanup_netlink_notifier
;
687 proc
= proc_net_create(IPQ_PROC_FS_NAME
, 0, ipq_get_info
);
689 proc
->owner
= THIS_MODULE
;
691 printk(KERN_ERR
"ip_queue: failed to create proc entry\n");
695 register_netdevice_notifier(&ipq_dev_notifier
);
696 ipq_sysctl_header
= register_sysctl_table(ipq_root_table
, 0);
698 status
= nf_register_queue_handler(PF_INET
, ipq_enqueue_packet
, NULL
);
700 printk(KERN_ERR
"ip_queue: failed to register queue handler\n");
706 nf_unregister_queue_handler(PF_INET
);
711 unregister_sysctl_table(ipq_sysctl_header
);
712 unregister_netdevice_notifier(&ipq_dev_notifier
);
713 proc_net_remove(IPQ_PROC_FS_NAME
);
716 sock_release(ipqnl
->sk_socket
);
720 cleanup_netlink_notifier
:
721 netlink_unregister_notifier(&ipq_nl_notifier
);
725 static int __init
init(void)
728 return init_or_cleanup(1);
731 static void __exit
fini(void)
736 MODULE_DESCRIPTION("IPv4 packet queue handler");
737 MODULE_AUTHOR("James Morris <jmorris@intercode.com.au>");
738 MODULE_LICENSE("GPL");