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/seq_file.h>
27 #include <linux/mutex.h>
28 #include <net/net_namespace.h>
31 #include <net/ip6_route.h>
32 #include <net/netfilter/nf_queue.h>
33 #include <linux/netfilter_ipv4/ip_queue.h>
34 #include <linux/netfilter_ipv4/ip_tables.h>
35 #include <linux/netfilter_ipv6/ip6_tables.h>
37 #define IPQ_QMAX_DEFAULT 1024
38 #define IPQ_PROC_FS_NAME "ip6_queue"
39 #define NET_IPQ_QMAX 2088
40 #define NET_IPQ_QMAX_NAME "ip6_queue_maxlen"
42 typedef int (*ipq_cmpfn)(struct nf_queue_entry *, unsigned long);
44 static unsigned char copy_mode __read_mostly = IPQ_COPY_NONE;
45 static unsigned int queue_maxlen __read_mostly = IPQ_QMAX_DEFAULT;
46 static DEFINE_RWLOCK(queue_lock);
47 static int peer_pid __read_mostly;
48 static unsigned int copy_range __read_mostly;
49 static unsigned int queue_total;
50 static unsigned int queue_dropped = 0;
51 static unsigned int queue_user_dropped = 0;
52 static struct sock *ipqnl __read_mostly;
53 static LIST_HEAD(queue_list);
54 static DEFINE_MUTEX(ipqnl_mutex);
57 ipq_issue_verdict(struct nf_queue_entry *entry, int verdict)
60 nf_reinject(entry, verdict);
65 __ipq_enqueue_entry(struct nf_queue_entry *entry)
67 list_add_tail(&entry->list, &queue_list);
72 __ipq_set_mode(unsigned char mode, unsigned int range)
86 if (copy_range > 0xFFFF)
97 static void __ipq_flush(ipq_cmpfn cmpfn, unsigned long data);
103 net_disable_timestamp();
104 __ipq_set_mode(IPQ_COPY_NONE, 0);
105 __ipq_flush(NULL, 0);
108 static struct nf_queue_entry *
109 ipq_find_dequeue_entry(unsigned long id)
111 struct nf_queue_entry *entry = NULL, *i;
113 write_lock_bh(&queue_lock);
115 list_for_each_entry(i, &queue_list, list) {
116 if ((unsigned long)i == id) {
123 list_del(&entry->list);
127 write_unlock_bh(&queue_lock);
132 __ipq_flush(ipq_cmpfn cmpfn, unsigned long data)
134 struct nf_queue_entry *entry, *next;
136 list_for_each_entry_safe(entry, next, &queue_list, list) {
137 if (!cmpfn || cmpfn(entry, data)) {
138 list_del(&entry->list);
140 ipq_issue_verdict(entry, NF_DROP);
146 ipq_flush(ipq_cmpfn cmpfn, unsigned long data)
148 write_lock_bh(&queue_lock);
149 __ipq_flush(cmpfn, data);
150 write_unlock_bh(&queue_lock);
153 static struct sk_buff *
154 ipq_build_packet_message(struct nf_queue_entry *entry, int *errp)
156 sk_buff_data_t old_tail;
160 struct ipq_packet_msg *pmsg;
161 struct nlmsghdr *nlh;
164 read_lock_bh(&queue_lock);
169 size = NLMSG_SPACE(sizeof(*pmsg));
173 case IPQ_COPY_PACKET:
174 if ((entry->skb->ip_summed == CHECKSUM_PARTIAL ||
175 entry->skb->ip_summed == CHECKSUM_COMPLETE) &&
176 (*errp = skb_checksum_help(entry->skb))) {
177 read_unlock_bh(&queue_lock);
180 if (copy_range == 0 || copy_range > entry->skb->len)
181 data_len = entry->skb->len;
183 data_len = copy_range;
185 size = NLMSG_SPACE(sizeof(*pmsg) + data_len);
190 read_unlock_bh(&queue_lock);
194 read_unlock_bh(&queue_lock);
196 skb = alloc_skb(size, GFP_ATOMIC);
200 old_tail = skb->tail;
201 nlh = NLMSG_PUT(skb, 0, 0, IPQM_PACKET, size - sizeof(*nlh));
202 pmsg = NLMSG_DATA(nlh);
203 memset(pmsg, 0, sizeof(*pmsg));
205 pmsg->packet_id = (unsigned long )entry;
206 pmsg->data_len = data_len;
207 tv = ktime_to_timeval(entry->skb->tstamp);
208 pmsg->timestamp_sec = tv.tv_sec;
209 pmsg->timestamp_usec = tv.tv_usec;
210 pmsg->mark = entry->skb->mark;
211 pmsg->hook = entry->hook;
212 pmsg->hw_protocol = entry->skb->protocol;
215 strcpy(pmsg->indev_name, entry->indev->name);
217 pmsg->indev_name[0] = '\0';
220 strcpy(pmsg->outdev_name, entry->outdev->name);
222 pmsg->outdev_name[0] = '\0';
224 if (entry->indev && entry->skb->dev) {
225 pmsg->hw_type = entry->skb->dev->type;
226 pmsg->hw_addrlen = dev_parse_header(entry->skb, pmsg->hw_addr);
230 if (skb_copy_bits(entry->skb, 0, pmsg->payload, data_len))
233 nlh->nlmsg_len = skb->tail - old_tail;
240 printk(KERN_ERR "ip6_queue: error creating packet message\n");
245 ipq_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
247 int status = -EINVAL;
248 struct sk_buff *nskb;
250 if (copy_mode == IPQ_COPY_NONE)
253 nskb = ipq_build_packet_message(entry, &status);
257 write_lock_bh(&queue_lock);
260 goto err_out_free_nskb;
262 if (queue_total >= queue_maxlen) {
266 printk (KERN_WARNING "ip6_queue: fill at %d entries, "
267 "dropping packet(s). Dropped: %d\n", queue_total,
269 goto err_out_free_nskb;
272 /* netlink_unicast will either free the nskb or attach it to a socket */
273 status = netlink_unicast(ipqnl, nskb, peer_pid, MSG_DONTWAIT);
275 queue_user_dropped++;
279 __ipq_enqueue_entry(entry);
281 write_unlock_bh(&queue_lock);
288 write_unlock_bh(&queue_lock);
293 ipq_mangle_ipv6(ipq_verdict_msg_t *v, struct nf_queue_entry *e)
297 struct ipv6hdr *user_iph = (struct ipv6hdr *)v->payload;
299 if (v->data_len < sizeof(*user_iph))
301 diff = v->data_len - e->skb->len;
303 if (pskb_trim(e->skb, v->data_len))
305 } else if (diff > 0) {
306 if (v->data_len > 0xFFFF)
308 if (diff > skb_tailroom(e->skb)) {
309 err = pskb_expand_head(e->skb, 0,
310 diff - skb_tailroom(e->skb),
313 printk(KERN_WARNING "ip6_queue: OOM "
314 "in mangle, dropping packet\n");
318 skb_put(e->skb, diff);
320 if (!skb_make_writable(e->skb, v->data_len))
322 skb_copy_to_linear_data(e->skb, v->payload, v->data_len);
323 e->skb->ip_summed = CHECKSUM_NONE;
329 ipq_set_verdict(struct ipq_verdict_msg *vmsg, unsigned int len)
331 struct nf_queue_entry *entry;
333 if (vmsg->value > NF_MAX_VERDICT)
336 entry = ipq_find_dequeue_entry(vmsg->id);
340 int verdict = vmsg->value;
342 if (vmsg->data_len && vmsg->data_len == len)
343 if (ipq_mangle_ipv6(vmsg, entry) < 0)
346 ipq_issue_verdict(entry, verdict);
352 ipq_set_mode(unsigned char mode, unsigned int range)
356 write_lock_bh(&queue_lock);
357 status = __ipq_set_mode(mode, range);
358 write_unlock_bh(&queue_lock);
363 ipq_receive_peer(struct ipq_peer_msg *pmsg,
364 unsigned char type, unsigned int len)
368 if (len < sizeof(*pmsg))
373 status = ipq_set_mode(pmsg->msg.mode.value,
374 pmsg->msg.mode.range);
378 if (pmsg->msg.verdict.value > NF_MAX_VERDICT)
381 status = ipq_set_verdict(&pmsg->msg.verdict,
382 len - sizeof(*pmsg));
391 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
394 if (entry->indev->ifindex == ifindex)
398 if (entry->outdev->ifindex == ifindex)
400 #ifdef CONFIG_BRIDGE_NETFILTER
401 if (entry->skb->nf_bridge) {
402 if (entry->skb->nf_bridge->physindev &&
403 entry->skb->nf_bridge->physindev->ifindex == ifindex)
405 if (entry->skb->nf_bridge->physoutdev &&
406 entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
414 ipq_dev_drop(int ifindex)
416 ipq_flush(dev_cmp, ifindex);
419 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
422 __ipq_rcv_skb(struct sk_buff *skb)
424 int status, type, pid, flags, nlmsglen, skblen;
425 struct nlmsghdr *nlh;
428 if (skblen < sizeof(*nlh))
431 nlh = nlmsg_hdr(skb);
432 nlmsglen = nlh->nlmsg_len;
433 if (nlmsglen < sizeof(*nlh) || skblen < nlmsglen)
436 pid = nlh->nlmsg_pid;
437 flags = nlh->nlmsg_flags;
439 if(pid <= 0 || !(flags & NLM_F_REQUEST) || flags & NLM_F_MULTI)
440 RCV_SKB_FAIL(-EINVAL);
442 if (flags & MSG_TRUNC)
443 RCV_SKB_FAIL(-ECOMM);
445 type = nlh->nlmsg_type;
446 if (type < NLMSG_NOOP || type >= IPQM_MAX)
447 RCV_SKB_FAIL(-EINVAL);
449 if (type <= IPQM_BASE)
452 if (security_netlink_recv(skb, CAP_NET_ADMIN))
453 RCV_SKB_FAIL(-EPERM);
455 write_lock_bh(&queue_lock);
458 if (peer_pid != pid) {
459 write_unlock_bh(&queue_lock);
460 RCV_SKB_FAIL(-EBUSY);
463 net_enable_timestamp();
467 write_unlock_bh(&queue_lock);
469 status = ipq_receive_peer(NLMSG_DATA(nlh), type,
470 nlmsglen - NLMSG_LENGTH(0));
472 RCV_SKB_FAIL(status);
474 if (flags & NLM_F_ACK)
475 netlink_ack(skb, nlh, 0);
480 ipq_rcv_skb(struct sk_buff *skb)
482 mutex_lock(&ipqnl_mutex);
484 mutex_unlock(&ipqnl_mutex);
488 ipq_rcv_dev_event(struct notifier_block *this,
489 unsigned long event, void *ptr)
491 struct net_device *dev = ptr;
493 if (dev->nd_net != &init_net)
496 /* Drop any packets associated with the downed device */
497 if (event == NETDEV_DOWN)
498 ipq_dev_drop(dev->ifindex);
502 static struct notifier_block ipq_dev_notifier = {
503 .notifier_call = ipq_rcv_dev_event,
507 ipq_rcv_nl_event(struct notifier_block *this,
508 unsigned long event, void *ptr)
510 struct netlink_notify *n = ptr;
512 if (event == NETLINK_URELEASE &&
513 n->protocol == NETLINK_IP6_FW && n->pid) {
514 write_lock_bh(&queue_lock);
515 if ((n->net == &init_net) && (n->pid == peer_pid))
517 write_unlock_bh(&queue_lock);
522 static struct notifier_block ipq_nl_notifier = {
523 .notifier_call = ipq_rcv_nl_event,
526 static struct ctl_table_header *ipq_sysctl_header;
528 static ctl_table ipq_table[] = {
530 .ctl_name = NET_IPQ_QMAX,
531 .procname = NET_IPQ_QMAX_NAME,
532 .data = &queue_maxlen,
533 .maxlen = sizeof(queue_maxlen),
535 .proc_handler = proc_dointvec
540 static ctl_table ipq_dir_table[] = {
542 .ctl_name = NET_IPV6,
550 static ctl_table ipq_root_table[] = {
555 .child = ipq_dir_table
560 static int ip6_queue_show(struct seq_file *m, void *v)
562 read_lock_bh(&queue_lock);
568 "Queue length : %u\n"
569 "Queue max. length : %u\n"
570 "Queue dropped : %u\n"
571 "Netfilter dropped : %u\n",
580 read_unlock_bh(&queue_lock);
584 static int ip6_queue_open(struct inode *inode, struct file *file)
586 return single_open(file, ip6_queue_show, NULL);
589 static const struct file_operations ip6_queue_proc_fops = {
590 .open = ip6_queue_open,
593 .release = single_release,
594 .owner = THIS_MODULE,
597 static const struct nf_queue_handler nfqh = {
599 .outfn = &ipq_enqueue_packet,
602 static int __init ip6_queue_init(void)
604 int status = -ENOMEM;
605 struct proc_dir_entry *proc;
607 netlink_register_notifier(&ipq_nl_notifier);
608 ipqnl = netlink_kernel_create(&init_net, NETLINK_IP6_FW, 0,
609 ipq_rcv_skb, NULL, THIS_MODULE);
611 printk(KERN_ERR "ip6_queue: failed to create netlink socket\n");
612 goto cleanup_netlink_notifier;
615 proc = create_proc_entry(IPQ_PROC_FS_NAME, 0, init_net.proc_net);
617 proc->owner = THIS_MODULE;
618 proc->proc_fops = &ip6_queue_proc_fops;
620 printk(KERN_ERR "ip6_queue: failed to create proc entry\n");
624 register_netdevice_notifier(&ipq_dev_notifier);
625 ipq_sysctl_header = register_sysctl_table(ipq_root_table);
627 status = nf_register_queue_handler(PF_INET6, &nfqh);
629 printk(KERN_ERR "ip6_queue: failed to register queue handler\n");
635 unregister_sysctl_table(ipq_sysctl_header);
636 unregister_netdevice_notifier(&ipq_dev_notifier);
637 proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
640 sock_release(ipqnl->sk_socket);
641 mutex_lock(&ipqnl_mutex);
642 mutex_unlock(&ipqnl_mutex);
644 cleanup_netlink_notifier:
645 netlink_unregister_notifier(&ipq_nl_notifier);
649 static void __exit ip6_queue_fini(void)
651 nf_unregister_queue_handlers(&nfqh);
655 unregister_sysctl_table(ipq_sysctl_header);
656 unregister_netdevice_notifier(&ipq_dev_notifier);
657 proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
659 sock_release(ipqnl->sk_socket);
660 mutex_lock(&ipqnl_mutex);
661 mutex_unlock(&ipqnl_mutex);
663 netlink_unregister_notifier(&ipq_nl_notifier);
666 MODULE_DESCRIPTION("IPv6 packet queue handler");
667 MODULE_LICENSE("GPL");
669 module_init(ip6_queue_init);
670 module_exit(ip6_queue_fini);