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 #include <linux/module.h>
16 #include <linux/skbuff.h>
17 #include <linux/init.h>
18 #include <linux/ipv6.h>
19 #include <linux/notifier.h>
20 #include <linux/netdevice.h>
21 #include <linux/netfilter.h>
22 #include <linux/netlink.h>
23 #include <linux/spinlock.h>
24 #include <linux/sysctl.h>
25 #include <linux/proc_fs.h>
26 #include <linux/mutex.h>
27 #include <net/net_namespace.h>
30 #include <net/ip6_route.h>
31 #include <linux/netfilter_ipv4/ip_queue.h>
32 #include <linux/netfilter_ipv4/ip_tables.h>
33 #include <linux/netfilter_ipv6/ip6_tables.h>
35 #define IPQ_QMAX_DEFAULT 1024
36 #define IPQ_PROC_FS_NAME "ip6_queue"
37 #define NET_IPQ_QMAX 2088
38 #define NET_IPQ_QMAX_NAME "ip6_queue_maxlen"
40 struct ipq_queue_entry {
41 struct list_head list;
46 typedef int (*ipq_cmpfn)(struct ipq_queue_entry *, unsigned long);
48 static unsigned char copy_mode __read_mostly = IPQ_COPY_NONE;
49 static unsigned int queue_maxlen __read_mostly = IPQ_QMAX_DEFAULT;
50 static DEFINE_RWLOCK(queue_lock);
51 static int peer_pid __read_mostly;
52 static unsigned int copy_range __read_mostly;
53 static unsigned int queue_total;
54 static unsigned int queue_dropped = 0;
55 static unsigned int queue_user_dropped = 0;
56 static struct sock *ipqnl __read_mostly;
57 static LIST_HEAD(queue_list);
58 static DEFINE_MUTEX(ipqnl_mutex);
61 ipq_issue_verdict(struct ipq_queue_entry *entry, int verdict)
64 nf_reinject(entry->skb, entry->info, verdict);
70 __ipq_enqueue_entry(struct ipq_queue_entry *entry)
72 list_add(&entry->list, &queue_list);
77 * Find and return a queued entry matched by cmpfn, or return the last
78 * entry if cmpfn is NULL.
80 static inline struct ipq_queue_entry *
81 __ipq_find_entry(ipq_cmpfn cmpfn, unsigned long data)
85 list_for_each_prev(p, &queue_list) {
86 struct ipq_queue_entry *entry = (struct ipq_queue_entry *)p;
88 if (!cmpfn || cmpfn(entry, data))
95 __ipq_dequeue_entry(struct ipq_queue_entry *entry)
97 list_del(&entry->list);
101 static inline struct ipq_queue_entry *
102 __ipq_find_dequeue_entry(ipq_cmpfn cmpfn, unsigned long data)
104 struct ipq_queue_entry *entry;
106 entry = __ipq_find_entry(cmpfn, data);
110 __ipq_dequeue_entry(entry);
116 __ipq_flush(int verdict)
118 struct ipq_queue_entry *entry;
120 while ((entry = __ipq_find_dequeue_entry(NULL, 0)))
121 ipq_issue_verdict(entry, verdict);
125 __ipq_set_mode(unsigned char mode, unsigned int range)
136 case IPQ_COPY_PACKET:
139 if (copy_range > 0xFFFF)
154 net_disable_timestamp();
155 __ipq_set_mode(IPQ_COPY_NONE, 0);
156 __ipq_flush(NF_DROP);
159 static struct ipq_queue_entry *
160 ipq_find_dequeue_entry(ipq_cmpfn cmpfn, unsigned long data)
162 struct ipq_queue_entry *entry;
164 write_lock_bh(&queue_lock);
165 entry = __ipq_find_dequeue_entry(cmpfn, data);
166 write_unlock_bh(&queue_lock);
171 ipq_flush(int verdict)
173 write_lock_bh(&queue_lock);
174 __ipq_flush(verdict);
175 write_unlock_bh(&queue_lock);
178 static struct sk_buff *
179 ipq_build_packet_message(struct ipq_queue_entry *entry, int *errp)
181 sk_buff_data_t old_tail;
185 struct ipq_packet_msg *pmsg;
186 struct nlmsghdr *nlh;
189 read_lock_bh(&queue_lock);
194 size = NLMSG_SPACE(sizeof(*pmsg));
198 case IPQ_COPY_PACKET:
199 if ((entry->skb->ip_summed == CHECKSUM_PARTIAL ||
200 entry->skb->ip_summed == CHECKSUM_COMPLETE) &&
201 (*errp = skb_checksum_help(entry->skb))) {
202 read_unlock_bh(&queue_lock);
205 if (copy_range == 0 || copy_range > entry->skb->len)
206 data_len = entry->skb->len;
208 data_len = copy_range;
210 size = NLMSG_SPACE(sizeof(*pmsg) + data_len);
215 read_unlock_bh(&queue_lock);
219 read_unlock_bh(&queue_lock);
221 skb = alloc_skb(size, GFP_ATOMIC);
225 old_tail = skb->tail;
226 nlh = NLMSG_PUT(skb, 0, 0, IPQM_PACKET, size - sizeof(*nlh));
227 pmsg = NLMSG_DATA(nlh);
228 memset(pmsg, 0, sizeof(*pmsg));
230 pmsg->packet_id = (unsigned long )entry;
231 pmsg->data_len = data_len;
232 tv = ktime_to_timeval(entry->skb->tstamp);
233 pmsg->timestamp_sec = tv.tv_sec;
234 pmsg->timestamp_usec = tv.tv_usec;
235 pmsg->mark = entry->skb->mark;
236 pmsg->hook = entry->info->hook;
237 pmsg->hw_protocol = entry->skb->protocol;
239 if (entry->info->indev)
240 strcpy(pmsg->indev_name, entry->info->indev->name);
242 pmsg->indev_name[0] = '\0';
244 if (entry->info->outdev)
245 strcpy(pmsg->outdev_name, entry->info->outdev->name);
247 pmsg->outdev_name[0] = '\0';
249 if (entry->info->indev && entry->skb->dev) {
250 pmsg->hw_type = entry->skb->dev->type;
251 pmsg->hw_addrlen = dev_parse_header(entry->skb, pmsg->hw_addr);
255 if (skb_copy_bits(entry->skb, 0, pmsg->payload, data_len))
258 nlh->nlmsg_len = skb->tail - old_tail;
265 printk(KERN_ERR "ip6_queue: error creating packet message\n");
270 ipq_enqueue_packet(struct sk_buff *skb, struct nf_info *info,
271 unsigned int queuenum, void *data)
273 int status = -EINVAL;
274 struct sk_buff *nskb;
275 struct ipq_queue_entry *entry;
277 if (copy_mode == IPQ_COPY_NONE)
280 entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
282 printk(KERN_ERR "ip6_queue: OOM in ipq_enqueue_packet()\n");
289 nskb = ipq_build_packet_message(entry, &status);
293 write_lock_bh(&queue_lock);
296 goto err_out_free_nskb;
298 if (queue_total >= queue_maxlen) {
302 printk (KERN_WARNING "ip6_queue: fill at %d entries, "
303 "dropping packet(s). Dropped: %d\n", queue_total,
305 goto err_out_free_nskb;
308 /* netlink_unicast will either free the nskb or attach it to a socket */
309 status = netlink_unicast(ipqnl, nskb, peer_pid, MSG_DONTWAIT);
311 queue_user_dropped++;
315 __ipq_enqueue_entry(entry);
317 write_unlock_bh(&queue_lock);
324 write_unlock_bh(&queue_lock);
332 ipq_mangle_ipv6(ipq_verdict_msg_t *v, struct ipq_queue_entry *e)
336 struct ipv6hdr *user_iph = (struct ipv6hdr *)v->payload;
338 if (v->data_len < sizeof(*user_iph))
340 diff = v->data_len - e->skb->len;
342 if (pskb_trim(e->skb, v->data_len))
344 } else if (diff > 0) {
345 if (v->data_len > 0xFFFF)
347 if (diff > skb_tailroom(e->skb)) {
348 err = pskb_expand_head(e->skb, 0,
349 diff - skb_tailroom(e->skb),
352 printk(KERN_WARNING "ip6_queue: OOM "
353 "in mangle, dropping packet\n");
357 skb_put(e->skb, diff);
359 if (!skb_make_writable(e->skb, v->data_len))
361 skb_copy_to_linear_data(e->skb, v->payload, v->data_len);
362 e->skb->ip_summed = CHECKSUM_NONE;
368 id_cmp(struct ipq_queue_entry *e, unsigned long id)
370 return (id == (unsigned long )e);
374 ipq_set_verdict(struct ipq_verdict_msg *vmsg, unsigned int len)
376 struct ipq_queue_entry *entry;
378 if (vmsg->value > NF_MAX_VERDICT)
381 entry = ipq_find_dequeue_entry(id_cmp, vmsg->id);
385 int verdict = vmsg->value;
387 if (vmsg->data_len && vmsg->data_len == len)
388 if (ipq_mangle_ipv6(vmsg, entry) < 0)
391 ipq_issue_verdict(entry, verdict);
397 ipq_set_mode(unsigned char mode, unsigned int range)
401 write_lock_bh(&queue_lock);
402 status = __ipq_set_mode(mode, range);
403 write_unlock_bh(&queue_lock);
408 ipq_receive_peer(struct ipq_peer_msg *pmsg,
409 unsigned char type, unsigned int len)
413 if (len < sizeof(*pmsg))
418 status = ipq_set_mode(pmsg->msg.mode.value,
419 pmsg->msg.mode.range);
423 if (pmsg->msg.verdict.value > NF_MAX_VERDICT)
426 status = ipq_set_verdict(&pmsg->msg.verdict,
427 len - sizeof(*pmsg));
436 dev_cmp(struct ipq_queue_entry *entry, unsigned long ifindex)
438 if (entry->info->indev)
439 if (entry->info->indev->ifindex == ifindex)
442 if (entry->info->outdev)
443 if (entry->info->outdev->ifindex == ifindex)
450 ipq_dev_drop(int ifindex)
452 struct ipq_queue_entry *entry;
454 while ((entry = ipq_find_dequeue_entry(dev_cmp, ifindex)) != NULL)
455 ipq_issue_verdict(entry, NF_DROP);
458 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
461 __ipq_rcv_skb(struct sk_buff *skb)
463 int status, type, pid, flags, nlmsglen, skblen;
464 struct nlmsghdr *nlh;
467 if (skblen < sizeof(*nlh))
470 nlh = nlmsg_hdr(skb);
471 nlmsglen = nlh->nlmsg_len;
472 if (nlmsglen < sizeof(*nlh) || skblen < nlmsglen)
475 pid = nlh->nlmsg_pid;
476 flags = nlh->nlmsg_flags;
478 if(pid <= 0 || !(flags & NLM_F_REQUEST) || flags & NLM_F_MULTI)
479 RCV_SKB_FAIL(-EINVAL);
481 if (flags & MSG_TRUNC)
482 RCV_SKB_FAIL(-ECOMM);
484 type = nlh->nlmsg_type;
485 if (type < NLMSG_NOOP || type >= IPQM_MAX)
486 RCV_SKB_FAIL(-EINVAL);
488 if (type <= IPQM_BASE)
491 if (security_netlink_recv(skb, CAP_NET_ADMIN))
492 RCV_SKB_FAIL(-EPERM);
494 write_lock_bh(&queue_lock);
497 if (peer_pid != pid) {
498 write_unlock_bh(&queue_lock);
499 RCV_SKB_FAIL(-EBUSY);
502 net_enable_timestamp();
506 write_unlock_bh(&queue_lock);
508 status = ipq_receive_peer(NLMSG_DATA(nlh), type,
509 nlmsglen - NLMSG_LENGTH(0));
511 RCV_SKB_FAIL(status);
513 if (flags & NLM_F_ACK)
514 netlink_ack(skb, nlh, 0);
519 ipq_rcv_skb(struct sk_buff *skb)
521 mutex_lock(&ipqnl_mutex);
523 mutex_unlock(&ipqnl_mutex);
527 ipq_rcv_dev_event(struct notifier_block *this,
528 unsigned long event, void *ptr)
530 struct net_device *dev = ptr;
532 if (dev->nd_net != &init_net)
535 /* Drop any packets associated with the downed device */
536 if (event == NETDEV_DOWN)
537 ipq_dev_drop(dev->ifindex);
541 static struct notifier_block ipq_dev_notifier = {
542 .notifier_call = ipq_rcv_dev_event,
546 ipq_rcv_nl_event(struct notifier_block *this,
547 unsigned long event, void *ptr)
549 struct netlink_notify *n = ptr;
551 if (event == NETLINK_URELEASE &&
552 n->protocol == NETLINK_IP6_FW && n->pid) {
553 write_lock_bh(&queue_lock);
554 if ((n->net == &init_net) && (n->pid == peer_pid))
556 write_unlock_bh(&queue_lock);
561 static struct notifier_block ipq_nl_notifier = {
562 .notifier_call = ipq_rcv_nl_event,
565 static struct ctl_table_header *ipq_sysctl_header;
567 static ctl_table ipq_table[] = {
569 .ctl_name = NET_IPQ_QMAX,
570 .procname = NET_IPQ_QMAX_NAME,
571 .data = &queue_maxlen,
572 .maxlen = sizeof(queue_maxlen),
574 .proc_handler = proc_dointvec
579 static ctl_table ipq_dir_table[] = {
581 .ctl_name = NET_IPV6,
589 static ctl_table ipq_root_table[] = {
594 .child = ipq_dir_table
599 #ifdef CONFIG_PROC_FS
601 ipq_get_info(char *buffer, char **start, off_t offset, int length)
605 read_lock_bh(&queue_lock);
607 len = sprintf(buffer,
611 "Queue length : %u\n"
612 "Queue max. length : %u\n"
613 "Queue dropped : %u\n"
614 "Netfilter dropped : %u\n",
623 read_unlock_bh(&queue_lock);
625 *start = buffer + offset;
633 #endif /* CONFIG_PROC_FS */
635 static struct nf_queue_handler nfqh = {
637 .outfn = &ipq_enqueue_packet,
640 static int __init ip6_queue_init(void)
642 int status = -ENOMEM;
643 struct proc_dir_entry *proc;
645 netlink_register_notifier(&ipq_nl_notifier);
646 ipqnl = netlink_kernel_create(&init_net, NETLINK_IP6_FW, 0,
647 ipq_rcv_skb, NULL, THIS_MODULE);
649 printk(KERN_ERR "ip6_queue: failed to create netlink socket\n");
650 goto cleanup_netlink_notifier;
653 proc = proc_net_create(&init_net, IPQ_PROC_FS_NAME, 0, ipq_get_info);
655 proc->owner = THIS_MODULE;
657 printk(KERN_ERR "ip6_queue: failed to create proc entry\n");
661 register_netdevice_notifier(&ipq_dev_notifier);
662 ipq_sysctl_header = register_sysctl_table(ipq_root_table);
664 status = nf_register_queue_handler(PF_INET6, &nfqh);
666 printk(KERN_ERR "ip6_queue: failed to register queue handler\n");
672 unregister_sysctl_table(ipq_sysctl_header);
673 unregister_netdevice_notifier(&ipq_dev_notifier);
674 proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
677 sock_release(ipqnl->sk_socket);
678 mutex_lock(&ipqnl_mutex);
679 mutex_unlock(&ipqnl_mutex);
681 cleanup_netlink_notifier:
682 netlink_unregister_notifier(&ipq_nl_notifier);
686 static void __exit ip6_queue_fini(void)
688 nf_unregister_queue_handlers(&nfqh);
692 unregister_sysctl_table(ipq_sysctl_header);
693 unregister_netdevice_notifier(&ipq_dev_notifier);
694 proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
696 sock_release(ipqnl->sk_socket);
697 mutex_lock(&ipqnl_mutex);
698 mutex_unlock(&ipqnl_mutex);
700 netlink_unregister_notifier(&ipq_nl_notifier);
703 MODULE_DESCRIPTION("IPv6 packet queue handler");
704 MODULE_LICENSE("GPL");
706 module_init(ip6_queue_init);
707 module_exit(ip6_queue_fini);