2 * This is a module which is used for queueing IPv6 packets and
3 * communicating with userspace via netlink.
5 * (C) 2001 Fernando Anton, this code is GPL.
6 * IPv64 Project - Work based in IPv64 draft by Arturo Azcorra.
7 * Universidad Carlos III de Madrid - Leganes (Madrid) - Spain
8 * Universidad Politecnica de Alcala de Henares - Alcala de H. (Madrid) - Spain
9 * email: fanton@it.uc3m.es
11 * This program is free software; you can redistribute it and/or modify
12 * it under the terms of the GNU General Public License version 2 as
13 * published by the Free Software Foundation.
15 * 2001-11-06: First try. Working with ip_queue.c for IPv4 and trying
17 * HEAVILY based in ipqueue.c by James Morris. It's just
18 * a little modified version of it, so he's nearly the
20 * Few changes needed, mainly the hard_routing code and
21 * the netlink socket protocol (we're NETLINK_IP6_FW).
22 * 2002-06-25: Code cleanup. [JM: ported cleanup over from ip_queue.c]
23 * 2005-02-04: Added /proc counter for dropped packets; fixed so
24 * packets aren't delivered to user space if they're going
27 #include <linux/module.h>
28 #include <linux/skbuff.h>
29 #include <linux/init.h>
30 #include <linux/ipv6.h>
31 #include <linux/notifier.h>
32 #include <linux/netdevice.h>
33 #include <linux/netfilter.h>
34 #include <linux/netlink.h>
35 #include <linux/spinlock.h>
36 #include <linux/sysctl.h>
37 #include <linux/proc_fs.h>
38 #include <linux/mutex.h>
41 #include <net/ip6_route.h>
42 #include <linux/netfilter_ipv4/ip_queue.h>
43 #include <linux/netfilter_ipv4/ip_tables.h>
44 #include <linux/netfilter_ipv6/ip6_tables.h>
46 #define IPQ_QMAX_DEFAULT 1024
47 #define IPQ_PROC_FS_NAME "ip6_queue"
48 #define NET_IPQ_QMAX 2088
49 #define NET_IPQ_QMAX_NAME "ip6_queue_maxlen"
51 struct ipq_queue_entry {
52 struct list_head list;
57 typedef int (*ipq_cmpfn)(struct ipq_queue_entry *, unsigned long);
59 static unsigned char copy_mode __read_mostly = IPQ_COPY_NONE;
60 static unsigned int queue_maxlen __read_mostly = IPQ_QMAX_DEFAULT;
61 static DEFINE_RWLOCK(queue_lock);
62 static int peer_pid __read_mostly;
63 static unsigned int copy_range __read_mostly;
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 __read_mostly;
68 static LIST_HEAD(queue_list);
69 static DEFINE_MUTEX(ipqnl_mutex);
72 ipq_issue_verdict(struct ipq_queue_entry *entry, int verdict)
75 nf_reinject(entry->skb, entry->info, verdict);
81 __ipq_enqueue_entry(struct ipq_queue_entry *entry)
83 list_add(&entry->list, &queue_list);
88 * Find and return a queued entry matched by cmpfn, or return the last
89 * entry if cmpfn is NULL.
91 static inline struct ipq_queue_entry *
92 __ipq_find_entry(ipq_cmpfn cmpfn, unsigned long data)
96 list_for_each_prev(p, &queue_list) {
97 struct ipq_queue_entry *entry = (struct ipq_queue_entry *)p;
99 if (!cmpfn || cmpfn(entry, data))
106 __ipq_dequeue_entry(struct ipq_queue_entry *entry)
108 list_del(&entry->list);
112 static inline struct ipq_queue_entry *
113 __ipq_find_dequeue_entry(ipq_cmpfn cmpfn, unsigned long data)
115 struct ipq_queue_entry *entry;
117 entry = __ipq_find_entry(cmpfn, data);
121 __ipq_dequeue_entry(entry);
127 __ipq_flush(int verdict)
129 struct ipq_queue_entry *entry;
131 while ((entry = __ipq_find_dequeue_entry(NULL, 0)))
132 ipq_issue_verdict(entry, verdict);
136 __ipq_set_mode(unsigned char mode, unsigned int range)
147 case IPQ_COPY_PACKET:
150 if (copy_range > 0xFFFF)
165 net_disable_timestamp();
166 __ipq_set_mode(IPQ_COPY_NONE, 0);
167 __ipq_flush(NF_DROP);
170 static struct ipq_queue_entry *
171 ipq_find_dequeue_entry(ipq_cmpfn cmpfn, unsigned long data)
173 struct ipq_queue_entry *entry;
175 write_lock_bh(&queue_lock);
176 entry = __ipq_find_dequeue_entry(cmpfn, data);
177 write_unlock_bh(&queue_lock);
182 ipq_flush(int verdict)
184 write_lock_bh(&queue_lock);
185 __ipq_flush(verdict);
186 write_unlock_bh(&queue_lock);
189 static struct sk_buff *
190 ipq_build_packet_message(struct ipq_queue_entry *entry, int *errp)
192 unsigned char *old_tail;
196 struct ipq_packet_msg *pmsg;
197 struct nlmsghdr *nlh;
199 read_lock_bh(&queue_lock);
204 size = NLMSG_SPACE(sizeof(*pmsg));
208 case IPQ_COPY_PACKET:
209 if ((entry->skb->ip_summed == CHECKSUM_PARTIAL ||
210 entry->skb->ip_summed == CHECKSUM_COMPLETE) &&
211 (*errp = skb_checksum_help(entry->skb))) {
212 read_unlock_bh(&queue_lock);
215 if (copy_range == 0 || copy_range > entry->skb->len)
216 data_len = entry->skb->len;
218 data_len = copy_range;
220 size = NLMSG_SPACE(sizeof(*pmsg) + data_len);
225 read_unlock_bh(&queue_lock);
229 read_unlock_bh(&queue_lock);
231 skb = alloc_skb(size, GFP_ATOMIC);
236 nlh = NLMSG_PUT(skb, 0, 0, IPQM_PACKET, size - sizeof(*nlh));
237 pmsg = NLMSG_DATA(nlh);
238 memset(pmsg, 0, sizeof(*pmsg));
240 pmsg->packet_id = (unsigned long )entry;
241 pmsg->data_len = data_len;
242 pmsg->timestamp_sec = entry->skb->tstamp.off_sec;
243 pmsg->timestamp_usec = entry->skb->tstamp.off_usec;
244 pmsg->mark = entry->skb->mark;
245 pmsg->hook = entry->info->hook;
246 pmsg->hw_protocol = entry->skb->protocol;
248 if (entry->info->indev)
249 strcpy(pmsg->indev_name, entry->info->indev->name);
251 pmsg->indev_name[0] = '\0';
253 if (entry->info->outdev)
254 strcpy(pmsg->outdev_name, entry->info->outdev->name);
256 pmsg->outdev_name[0] = '\0';
258 if (entry->info->indev && entry->skb->dev) {
259 pmsg->hw_type = entry->skb->dev->type;
260 if (entry->skb->dev->hard_header_parse)
262 entry->skb->dev->hard_header_parse(entry->skb,
267 if (skb_copy_bits(entry->skb, 0, pmsg->payload, data_len))
270 nlh->nlmsg_len = skb->tail - old_tail;
277 printk(KERN_ERR "ip6_queue: error creating packet message\n");
282 ipq_enqueue_packet(struct sk_buff *skb, struct nf_info *info,
283 unsigned int queuenum, void *data)
285 int status = -EINVAL;
286 struct sk_buff *nskb;
287 struct ipq_queue_entry *entry;
289 if (copy_mode == IPQ_COPY_NONE)
292 entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
294 printk(KERN_ERR "ip6_queue: OOM in ipq_enqueue_packet()\n");
301 nskb = ipq_build_packet_message(entry, &status);
305 write_lock_bh(&queue_lock);
308 goto err_out_free_nskb;
310 if (queue_total >= queue_maxlen) {
314 printk (KERN_WARNING "ip6_queue: fill at %d entries, "
315 "dropping packet(s). Dropped: %d\n", queue_total,
317 goto err_out_free_nskb;
320 /* netlink_unicast will either free the nskb or attach it to a socket */
321 status = netlink_unicast(ipqnl, nskb, peer_pid, MSG_DONTWAIT);
323 queue_user_dropped++;
327 __ipq_enqueue_entry(entry);
329 write_unlock_bh(&queue_lock);
336 write_unlock_bh(&queue_lock);
344 ipq_mangle_ipv6(ipq_verdict_msg_t *v, struct ipq_queue_entry *e)
347 struct ipv6hdr *user_iph = (struct ipv6hdr *)v->payload;
349 if (v->data_len < sizeof(*user_iph))
351 diff = v->data_len - e->skb->len;
353 if (pskb_trim(e->skb, v->data_len))
355 } else if (diff > 0) {
356 if (v->data_len > 0xFFFF)
358 if (diff > skb_tailroom(e->skb)) {
359 struct sk_buff *newskb;
361 newskb = skb_copy_expand(e->skb,
362 skb_headroom(e->skb),
365 if (newskb == NULL) {
366 printk(KERN_WARNING "ip6_queue: OOM "
367 "in mangle, dropping packet\n");
371 skb_set_owner_w(newskb, e->skb->sk);
375 skb_put(e->skb, diff);
377 if (!skb_make_writable(&e->skb, v->data_len))
379 memcpy(e->skb->data, v->payload, v->data_len);
380 e->skb->ip_summed = CHECKSUM_NONE;
386 id_cmp(struct ipq_queue_entry *e, unsigned long id)
388 return (id == (unsigned long )e);
392 ipq_set_verdict(struct ipq_verdict_msg *vmsg, unsigned int len)
394 struct ipq_queue_entry *entry;
396 if (vmsg->value > NF_MAX_VERDICT)
399 entry = ipq_find_dequeue_entry(id_cmp, vmsg->id);
403 int verdict = vmsg->value;
405 if (vmsg->data_len && vmsg->data_len == len)
406 if (ipq_mangle_ipv6(vmsg, entry) < 0)
409 ipq_issue_verdict(entry, verdict);
415 ipq_set_mode(unsigned char mode, unsigned int range)
419 write_lock_bh(&queue_lock);
420 status = __ipq_set_mode(mode, range);
421 write_unlock_bh(&queue_lock);
426 ipq_receive_peer(struct ipq_peer_msg *pmsg,
427 unsigned char type, unsigned int len)
431 if (len < sizeof(*pmsg))
436 status = ipq_set_mode(pmsg->msg.mode.value,
437 pmsg->msg.mode.range);
441 if (pmsg->msg.verdict.value > NF_MAX_VERDICT)
444 status = ipq_set_verdict(&pmsg->msg.verdict,
445 len - sizeof(*pmsg));
454 dev_cmp(struct ipq_queue_entry *entry, unsigned long ifindex)
456 if (entry->info->indev)
457 if (entry->info->indev->ifindex == ifindex)
460 if (entry->info->outdev)
461 if (entry->info->outdev->ifindex == ifindex)
468 ipq_dev_drop(int ifindex)
470 struct ipq_queue_entry *entry;
472 while ((entry = ipq_find_dequeue_entry(dev_cmp, ifindex)) != NULL)
473 ipq_issue_verdict(entry, NF_DROP);
476 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
479 ipq_rcv_skb(struct sk_buff *skb)
481 int status, type, pid, flags, nlmsglen, skblen;
482 struct nlmsghdr *nlh;
485 if (skblen < sizeof(*nlh))
488 nlh = (struct nlmsghdr *)skb->data;
489 nlmsglen = nlh->nlmsg_len;
490 if (nlmsglen < sizeof(*nlh) || skblen < nlmsglen)
493 pid = nlh->nlmsg_pid;
494 flags = nlh->nlmsg_flags;
496 if(pid <= 0 || !(flags & NLM_F_REQUEST) || flags & NLM_F_MULTI)
497 RCV_SKB_FAIL(-EINVAL);
499 if (flags & MSG_TRUNC)
500 RCV_SKB_FAIL(-ECOMM);
502 type = nlh->nlmsg_type;
503 if (type < NLMSG_NOOP || type >= IPQM_MAX)
504 RCV_SKB_FAIL(-EINVAL);
506 if (type <= IPQM_BASE)
509 if (security_netlink_recv(skb, CAP_NET_ADMIN))
510 RCV_SKB_FAIL(-EPERM);
512 write_lock_bh(&queue_lock);
515 if (peer_pid != pid) {
516 write_unlock_bh(&queue_lock);
517 RCV_SKB_FAIL(-EBUSY);
520 net_enable_timestamp();
524 write_unlock_bh(&queue_lock);
526 status = ipq_receive_peer(NLMSG_DATA(nlh), type,
527 nlmsglen - NLMSG_LENGTH(0));
529 RCV_SKB_FAIL(status);
531 if (flags & NLM_F_ACK)
532 netlink_ack(skb, nlh, 0);
537 ipq_rcv_sk(struct sock *sk, int len)
542 mutex_lock(&ipqnl_mutex);
544 for (qlen = skb_queue_len(&sk->sk_receive_queue); qlen; qlen--) {
545 skb = skb_dequeue(&sk->sk_receive_queue);
550 mutex_unlock(&ipqnl_mutex);
554 ipq_rcv_dev_event(struct notifier_block *this,
555 unsigned long event, void *ptr)
557 struct net_device *dev = ptr;
559 /* Drop any packets associated with the downed device */
560 if (event == NETDEV_DOWN)
561 ipq_dev_drop(dev->ifindex);
565 static struct notifier_block ipq_dev_notifier = {
566 .notifier_call = ipq_rcv_dev_event,
570 ipq_rcv_nl_event(struct notifier_block *this,
571 unsigned long event, void *ptr)
573 struct netlink_notify *n = ptr;
575 if (event == NETLINK_URELEASE &&
576 n->protocol == NETLINK_IP6_FW && n->pid) {
577 write_lock_bh(&queue_lock);
578 if (n->pid == peer_pid)
580 write_unlock_bh(&queue_lock);
585 static struct notifier_block ipq_nl_notifier = {
586 .notifier_call = ipq_rcv_nl_event,
589 static struct ctl_table_header *ipq_sysctl_header;
591 static ctl_table ipq_table[] = {
593 .ctl_name = NET_IPQ_QMAX,
594 .procname = NET_IPQ_QMAX_NAME,
595 .data = &queue_maxlen,
596 .maxlen = sizeof(queue_maxlen),
598 .proc_handler = proc_dointvec
603 static ctl_table ipq_dir_table[] = {
605 .ctl_name = NET_IPV6,
613 static ctl_table ipq_root_table[] = {
618 .child = ipq_dir_table
623 #ifdef CONFIG_PROC_FS
625 ipq_get_info(char *buffer, char **start, off_t offset, int length)
629 read_lock_bh(&queue_lock);
631 len = sprintf(buffer,
635 "Queue length : %u\n"
636 "Queue max. length : %u\n"
637 "Queue dropped : %u\n"
638 "Netfilter dropped : %u\n",
647 read_unlock_bh(&queue_lock);
649 *start = buffer + offset;
657 #endif /* CONFIG_PROC_FS */
659 static struct nf_queue_handler nfqh = {
661 .outfn = &ipq_enqueue_packet,
664 static int __init ip6_queue_init(void)
666 int status = -ENOMEM;
667 struct proc_dir_entry *proc;
669 netlink_register_notifier(&ipq_nl_notifier);
670 ipqnl = netlink_kernel_create(NETLINK_IP6_FW, 0, ipq_rcv_sk,
673 printk(KERN_ERR "ip6_queue: failed to create netlink socket\n");
674 goto cleanup_netlink_notifier;
677 proc = proc_net_create(IPQ_PROC_FS_NAME, 0, ipq_get_info);
679 proc->owner = THIS_MODULE;
681 printk(KERN_ERR "ip6_queue: failed to create proc entry\n");
685 register_netdevice_notifier(&ipq_dev_notifier);
686 ipq_sysctl_header = register_sysctl_table(ipq_root_table, 0);
688 status = nf_register_queue_handler(PF_INET6, &nfqh);
690 printk(KERN_ERR "ip6_queue: failed to register queue handler\n");
696 unregister_sysctl_table(ipq_sysctl_header);
697 unregister_netdevice_notifier(&ipq_dev_notifier);
698 proc_net_remove(IPQ_PROC_FS_NAME);
701 sock_release(ipqnl->sk_socket);
702 mutex_lock(&ipqnl_mutex);
703 mutex_unlock(&ipqnl_mutex);
705 cleanup_netlink_notifier:
706 netlink_unregister_notifier(&ipq_nl_notifier);
710 static void __exit ip6_queue_fini(void)
712 nf_unregister_queue_handlers(&nfqh);
716 unregister_sysctl_table(ipq_sysctl_header);
717 unregister_netdevice_notifier(&ipq_dev_notifier);
718 proc_net_remove(IPQ_PROC_FS_NAME);
720 sock_release(ipqnl->sk_socket);
721 mutex_lock(&ipqnl_mutex);
722 mutex_unlock(&ipqnl_mutex);
724 netlink_unregister_notifier(&ipq_nl_notifier);
727 MODULE_DESCRIPTION("IPv6 packet queue handler");
728 MODULE_LICENSE("GPL");
730 module_init(ip6_queue_init);
731 module_exit(ip6_queue_fini);