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/mutex.h>
28 #include <net/net_namespace.h>
30 #include <net/route.h>
31 #include <net/netfilter/nf_queue.h>
33 #define IPQ_QMAX_DEFAULT 1024
34 #define IPQ_PROC_FS_NAME "ip_queue"
35 #define NET_IPQ_QMAX 2088
36 #define NET_IPQ_QMAX_NAME "ip_queue_maxlen"
38 typedef int (*ipq_cmpfn)(struct nf_queue_entry *, unsigned long);
40 static unsigned char copy_mode __read_mostly = IPQ_COPY_NONE;
41 static unsigned int queue_maxlen __read_mostly = IPQ_QMAX_DEFAULT;
42 static DEFINE_RWLOCK(queue_lock);
43 static int peer_pid __read_mostly;
44 static unsigned int copy_range __read_mostly;
45 static unsigned int queue_total;
46 static unsigned int queue_dropped = 0;
47 static unsigned int queue_user_dropped = 0;
48 static struct sock *ipqnl __read_mostly;
49 static LIST_HEAD(queue_list);
50 static DEFINE_MUTEX(ipqnl_mutex);
53 __ipq_enqueue_entry(struct nf_queue_entry *entry)
55 list_add_tail(&entry->list, &queue_list);
60 __ipq_set_mode(unsigned char mode, unsigned int range)
74 if (copy_range > 0xFFFF)
85 static void __ipq_flush(ipq_cmpfn cmpfn, unsigned long data);
91 net_disable_timestamp();
92 __ipq_set_mode(IPQ_COPY_NONE, 0);
96 static struct nf_queue_entry *
97 ipq_find_dequeue_entry(unsigned long id)
99 struct nf_queue_entry *entry = NULL, *i;
101 write_lock_bh(&queue_lock);
103 list_for_each_entry(i, &queue_list, list) {
104 if ((unsigned long)i == id) {
111 list_del(&entry->list);
115 write_unlock_bh(&queue_lock);
120 __ipq_flush(ipq_cmpfn cmpfn, unsigned long data)
122 struct nf_queue_entry *entry, *next;
124 list_for_each_entry_safe(entry, next, &queue_list, list) {
125 if (!cmpfn || cmpfn(entry, data)) {
126 list_del(&entry->list);
128 nf_reinject(entry, NF_DROP);
134 ipq_flush(ipq_cmpfn cmpfn, unsigned long data)
136 write_lock_bh(&queue_lock);
137 __ipq_flush(cmpfn, data);
138 write_unlock_bh(&queue_lock);
141 static struct sk_buff *
142 ipq_build_packet_message(struct nf_queue_entry *entry, int *errp)
144 sk_buff_data_t old_tail;
148 struct ipq_packet_msg *pmsg;
149 struct nlmsghdr *nlh;
152 read_lock_bh(&queue_lock);
157 size = NLMSG_SPACE(sizeof(*pmsg));
161 case IPQ_COPY_PACKET:
162 if ((entry->skb->ip_summed == CHECKSUM_PARTIAL ||
163 entry->skb->ip_summed == CHECKSUM_COMPLETE) &&
164 (*errp = skb_checksum_help(entry->skb))) {
165 read_unlock_bh(&queue_lock);
168 if (copy_range == 0 || copy_range > entry->skb->len)
169 data_len = entry->skb->len;
171 data_len = copy_range;
173 size = NLMSG_SPACE(sizeof(*pmsg) + data_len);
178 read_unlock_bh(&queue_lock);
182 read_unlock_bh(&queue_lock);
184 skb = alloc_skb(size, GFP_ATOMIC);
188 old_tail = skb->tail;
189 nlh = NLMSG_PUT(skb, 0, 0, IPQM_PACKET, size - sizeof(*nlh));
190 pmsg = NLMSG_DATA(nlh);
191 memset(pmsg, 0, sizeof(*pmsg));
193 pmsg->packet_id = (unsigned long )entry;
194 pmsg->data_len = data_len;
195 tv = ktime_to_timeval(entry->skb->tstamp);
196 pmsg->timestamp_sec = tv.tv_sec;
197 pmsg->timestamp_usec = tv.tv_usec;
198 pmsg->mark = entry->skb->mark;
199 pmsg->hook = entry->hook;
200 pmsg->hw_protocol = entry->skb->protocol;
203 strcpy(pmsg->indev_name, entry->indev->name);
205 pmsg->indev_name[0] = '\0';
208 strcpy(pmsg->outdev_name, entry->outdev->name);
210 pmsg->outdev_name[0] = '\0';
212 if (entry->indev && entry->skb->dev) {
213 pmsg->hw_type = entry->skb->dev->type;
214 pmsg->hw_addrlen = dev_parse_header(entry->skb,
219 if (skb_copy_bits(entry->skb, 0, pmsg->payload, data_len))
222 nlh->nlmsg_len = skb->tail - old_tail;
229 printk(KERN_ERR "ip_queue: error creating packet message\n");
234 ipq_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
236 int status = -EINVAL;
237 struct sk_buff *nskb;
239 if (copy_mode == IPQ_COPY_NONE)
242 nskb = ipq_build_packet_message(entry, &status);
246 write_lock_bh(&queue_lock);
249 goto err_out_free_nskb;
251 if (queue_total >= queue_maxlen) {
255 printk (KERN_WARNING "ip_queue: full at %d entries, "
256 "dropping packets(s). Dropped: %d\n", queue_total,
258 goto err_out_free_nskb;
261 /* netlink_unicast will either free the nskb or attach it to a socket */
262 status = netlink_unicast(ipqnl, nskb, peer_pid, MSG_DONTWAIT);
264 queue_user_dropped++;
268 __ipq_enqueue_entry(entry);
270 write_unlock_bh(&queue_lock);
277 write_unlock_bh(&queue_lock);
282 ipq_mangle_ipv4(ipq_verdict_msg_t *v, struct nf_queue_entry *e)
286 struct iphdr *user_iph = (struct iphdr *)v->payload;
288 if (v->data_len < sizeof(*user_iph))
290 diff = v->data_len - e->skb->len;
292 if (pskb_trim(e->skb, v->data_len))
294 } else if (diff > 0) {
295 if (v->data_len > 0xFFFF)
297 if (diff > skb_tailroom(e->skb)) {
298 err = pskb_expand_head(e->skb, 0,
299 diff - skb_tailroom(e->skb),
302 printk(KERN_WARNING "ip_queue: error "
303 "in mangle, dropping packet: %d\n", -err);
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);
468 ipq_rcv_skb(struct sk_buff *skb)
470 mutex_lock(&ipqnl_mutex);
472 mutex_unlock(&ipqnl_mutex);
476 ipq_rcv_dev_event(struct notifier_block *this,
477 unsigned long event, void *ptr)
479 struct net_device *dev = ptr;
481 if (dev->nd_net != &init_net)
484 /* Drop any packets associated with the downed device */
485 if (event == NETDEV_DOWN)
486 ipq_dev_drop(dev->ifindex);
490 static struct notifier_block ipq_dev_notifier = {
491 .notifier_call = ipq_rcv_dev_event,
495 ipq_rcv_nl_event(struct notifier_block *this,
496 unsigned long event, void *ptr)
498 struct netlink_notify *n = ptr;
500 if (event == NETLINK_URELEASE &&
501 n->protocol == NETLINK_FIREWALL && n->pid) {
502 write_lock_bh(&queue_lock);
503 if ((n->net == &init_net) && (n->pid == peer_pid))
505 write_unlock_bh(&queue_lock);
510 static struct notifier_block ipq_nl_notifier = {
511 .notifier_call = ipq_rcv_nl_event,
514 static struct ctl_table_header *ipq_sysctl_header;
516 static ctl_table ipq_table[] = {
518 .ctl_name = NET_IPQ_QMAX,
519 .procname = NET_IPQ_QMAX_NAME,
520 .data = &queue_maxlen,
521 .maxlen = sizeof(queue_maxlen),
523 .proc_handler = proc_dointvec
528 static ctl_table ipq_dir_table[] = {
530 .ctl_name = NET_IPV4,
538 static ctl_table ipq_root_table[] = {
543 .child = ipq_dir_table
548 static int ip_queue_show(struct seq_file *m, void *v)
550 read_lock_bh(&queue_lock);
556 "Queue length : %u\n"
557 "Queue max. length : %u\n"
558 "Queue dropped : %u\n"
559 "Netlink dropped : %u\n",
568 read_unlock_bh(&queue_lock);
572 static int ip_queue_open(struct inode *inode, struct file *file)
574 return single_open(file, ip_queue_show, NULL);
577 static const struct file_operations ip_queue_proc_fops = {
578 .open = ip_queue_open,
581 .release = single_release,
582 .owner = THIS_MODULE,
585 static const struct nf_queue_handler nfqh = {
587 .outfn = &ipq_enqueue_packet,
590 static int __init ip_queue_init(void)
592 int status = -ENOMEM;
593 struct proc_dir_entry *proc;
595 netlink_register_notifier(&ipq_nl_notifier);
596 ipqnl = netlink_kernel_create(&init_net, NETLINK_FIREWALL, 0,
597 ipq_rcv_skb, NULL, THIS_MODULE);
599 printk(KERN_ERR "ip_queue: failed to create netlink socket\n");
600 goto cleanup_netlink_notifier;
603 proc = create_proc_entry(IPQ_PROC_FS_NAME, 0, init_net.proc_net);
605 proc->owner = THIS_MODULE;
606 proc->proc_fops = &ip_queue_proc_fops;
608 printk(KERN_ERR "ip_queue: failed to create proc entry\n");
612 register_netdevice_notifier(&ipq_dev_notifier);
613 ipq_sysctl_header = register_sysctl_table(ipq_root_table);
615 status = nf_register_queue_handler(PF_INET, &nfqh);
617 printk(KERN_ERR "ip_queue: failed to register queue handler\n");
623 unregister_sysctl_table(ipq_sysctl_header);
624 unregister_netdevice_notifier(&ipq_dev_notifier);
625 proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
627 sock_release(ipqnl->sk_socket);
628 mutex_lock(&ipqnl_mutex);
629 mutex_unlock(&ipqnl_mutex);
631 cleanup_netlink_notifier:
632 netlink_unregister_notifier(&ipq_nl_notifier);
636 static void __exit ip_queue_fini(void)
638 nf_unregister_queue_handlers(&nfqh);
642 unregister_sysctl_table(ipq_sysctl_header);
643 unregister_netdevice_notifier(&ipq_dev_notifier);
644 proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
646 sock_release(ipqnl->sk_socket);
647 mutex_lock(&ipqnl_mutex);
648 mutex_unlock(&ipqnl_mutex);
650 netlink_unregister_notifier(&ipq_nl_notifier);
653 MODULE_DESCRIPTION("IPv4 packet queue handler");
654 MODULE_AUTHOR("James Morris <jmorris@intercode.com.au>");
655 MODULE_LICENSE("GPL");
657 module_init(ip_queue_init);
658 module_exit(ip_queue_fini);