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/net.h>
28 #include <linux/mutex.h>
29 #include <linux/slab.h>
30 #include <net/net_namespace.h>
32 #include <net/route.h>
33 #include <net/netfilter/nf_queue.h>
36 #define IPQ_QMAX_DEFAULT 1024
37 #define IPQ_PROC_FS_NAME "ip_queue"
38 #define NET_IPQ_QMAX 2088
39 #define NET_IPQ_QMAX_NAME "ip_queue_maxlen"
41 typedef int (*ipq_cmpfn)(struct nf_queue_entry *, unsigned long);
43 static unsigned char copy_mode __read_mostly = IPQ_COPY_NONE;
44 static unsigned int queue_maxlen __read_mostly = IPQ_QMAX_DEFAULT;
45 static DEFINE_SPINLOCK(queue_lock);
46 static int peer_pid __read_mostly;
47 static unsigned int copy_range __read_mostly;
48 static unsigned int queue_total;
49 static unsigned int queue_dropped = 0;
50 static unsigned int queue_user_dropped = 0;
51 static struct sock *ipqnl __read_mostly;
52 static LIST_HEAD(queue_list);
53 static DEFINE_MUTEX(ipqnl_mutex);
56 __ipq_enqueue_entry(struct nf_queue_entry *entry)
58 list_add_tail(&entry->list, &queue_list);
63 __ipq_set_mode(unsigned char mode, unsigned int range)
88 static void __ipq_flush(ipq_cmpfn cmpfn, unsigned long data);
94 net_disable_timestamp();
95 __ipq_set_mode(IPQ_COPY_NONE, 0);
99 static struct nf_queue_entry *
100 ipq_find_dequeue_entry(unsigned long id)
102 struct nf_queue_entry *entry = NULL, *i;
104 spin_lock_bh(&queue_lock);
106 list_for_each_entry(i, &queue_list, list) {
107 if ((unsigned long)i == id) {
114 list_del(&entry->list);
118 spin_unlock_bh(&queue_lock);
123 __ipq_flush(ipq_cmpfn cmpfn, unsigned long data)
125 struct nf_queue_entry *entry, *next;
127 list_for_each_entry_safe(entry, next, &queue_list, list) {
128 if (!cmpfn || cmpfn(entry, data)) {
129 list_del(&entry->list);
131 nf_reinject(entry, NF_DROP);
137 ipq_flush(ipq_cmpfn cmpfn, unsigned long data)
139 spin_lock_bh(&queue_lock);
140 __ipq_flush(cmpfn, data);
141 spin_unlock_bh(&queue_lock);
144 static struct sk_buff *
145 ipq_build_packet_message(struct nf_queue_entry *entry, int *errp)
147 sk_buff_data_t old_tail;
151 struct ipq_packet_msg *pmsg;
152 struct nlmsghdr *nlh;
155 switch (ACCESS_ONCE(copy_mode)) {
158 size = NLMSG_SPACE(sizeof(*pmsg));
161 case IPQ_COPY_PACKET:
162 if (entry->skb->ip_summed == CHECKSUM_PARTIAL &&
163 (*errp = skb_checksum_help(entry->skb)))
166 data_len = ACCESS_ONCE(copy_range);
167 if (data_len == 0 || data_len > entry->skb->len)
168 data_len = entry->skb->len;
170 size = NLMSG_SPACE(sizeof(*pmsg) + data_len);
178 skb = alloc_skb(size, GFP_ATOMIC);
182 old_tail = skb->tail;
183 nlh = NLMSG_PUT(skb, 0, 0, IPQM_PACKET, size - sizeof(*nlh));
184 pmsg = NLMSG_DATA(nlh);
185 memset(pmsg, 0, sizeof(*pmsg));
187 pmsg->packet_id = (unsigned long )entry;
188 pmsg->data_len = data_len;
189 tv = ktime_to_timeval(entry->skb->tstamp);
190 pmsg->timestamp_sec = tv.tv_sec;
191 pmsg->timestamp_usec = tv.tv_usec;
192 pmsg->mark = entry->skb->mark;
193 pmsg->hook = entry->hook;
194 pmsg->hw_protocol = entry->skb->protocol;
197 strcpy(pmsg->indev_name, entry->indev->name);
199 pmsg->indev_name[0] = '\0';
202 strcpy(pmsg->outdev_name, entry->outdev->name);
204 pmsg->outdev_name[0] = '\0';
206 if (entry->indev && entry->skb->dev &&
207 entry->skb->mac_header != entry->skb->network_header) {
208 pmsg->hw_type = entry->skb->dev->type;
209 pmsg->hw_addrlen = dev_parse_header(entry->skb,
214 if (skb_copy_bits(entry->skb, 0, pmsg->payload, data_len))
217 nlh->nlmsg_len = skb->tail - old_tail;
222 printk(KERN_ERR "ip_queue: error creating packet message\n");
227 ipq_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
229 int status = -EINVAL;
230 struct sk_buff *nskb;
232 if (copy_mode == IPQ_COPY_NONE)
235 nskb = ipq_build_packet_message(entry, &status);
239 spin_lock_bh(&queue_lock);
242 goto err_out_free_nskb;
244 if (queue_total >= queue_maxlen) {
248 printk (KERN_WARNING "ip_queue: full at %d entries, "
249 "dropping packets(s). Dropped: %d\n", queue_total,
251 goto err_out_free_nskb;
254 /* netlink_unicast will either free the nskb or attach it to a socket */
255 status = netlink_unicast(ipqnl, nskb, peer_pid, MSG_DONTWAIT);
257 queue_user_dropped++;
261 __ipq_enqueue_entry(entry);
263 spin_unlock_bh(&queue_lock);
270 spin_unlock_bh(&queue_lock);
275 ipq_mangle_ipv4(ipq_verdict_msg_t *v, struct nf_queue_entry *e)
278 struct iphdr *user_iph = (struct iphdr *)v->payload;
279 struct sk_buff *nskb;
281 if (v->data_len < sizeof(*user_iph))
283 diff = v->data_len - e->skb->len;
285 if (pskb_trim(e->skb, v->data_len))
287 } else if (diff > 0) {
288 if (v->data_len > 0xFFFF)
290 if (diff > skb_tailroom(e->skb)) {
291 nskb = skb_copy_expand(e->skb, skb_headroom(e->skb),
294 printk(KERN_WARNING "ip_queue: error "
295 "in mangle, dropping packet\n");
301 skb_put(e->skb, diff);
303 if (!skb_make_writable(e->skb, v->data_len))
305 skb_copy_to_linear_data(e->skb, v->payload, v->data_len);
306 e->skb->ip_summed = CHECKSUM_NONE;
312 ipq_set_verdict(struct ipq_verdict_msg *vmsg, unsigned int len)
314 struct nf_queue_entry *entry;
316 if (vmsg->value > NF_MAX_VERDICT)
319 entry = ipq_find_dequeue_entry(vmsg->id);
323 int verdict = vmsg->value;
325 if (vmsg->data_len && vmsg->data_len == len)
326 if (ipq_mangle_ipv4(vmsg, entry) < 0)
329 nf_reinject(entry, verdict);
335 ipq_set_mode(unsigned char mode, unsigned int range)
339 spin_lock_bh(&queue_lock);
340 status = __ipq_set_mode(mode, range);
341 spin_unlock_bh(&queue_lock);
346 ipq_receive_peer(struct ipq_peer_msg *pmsg,
347 unsigned char type, unsigned int len)
351 if (len < sizeof(*pmsg))
356 status = ipq_set_mode(pmsg->msg.mode.value,
357 pmsg->msg.mode.range);
361 if (pmsg->msg.verdict.value > NF_MAX_VERDICT)
364 status = ipq_set_verdict(&pmsg->msg.verdict,
365 len - sizeof(*pmsg));
374 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
377 if (entry->indev->ifindex == ifindex)
380 if (entry->outdev->ifindex == ifindex)
382 #ifdef CONFIG_BRIDGE_NETFILTER
383 if (entry->skb->nf_bridge) {
384 if (entry->skb->nf_bridge->physindev &&
385 entry->skb->nf_bridge->physindev->ifindex == ifindex)
387 if (entry->skb->nf_bridge->physoutdev &&
388 entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
396 ipq_dev_drop(int ifindex)
398 ipq_flush(dev_cmp, ifindex);
401 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
404 __ipq_rcv_skb(struct sk_buff *skb)
406 int status, type, pid, flags;
407 unsigned int nlmsglen, skblen;
408 struct nlmsghdr *nlh;
411 if (skblen < sizeof(*nlh))
414 nlh = nlmsg_hdr(skb);
415 nlmsglen = nlh->nlmsg_len;
416 if (nlmsglen < sizeof(*nlh) || skblen < nlmsglen)
419 pid = nlh->nlmsg_pid;
420 flags = nlh->nlmsg_flags;
422 if(pid <= 0 || !(flags & NLM_F_REQUEST) || flags & NLM_F_MULTI)
423 RCV_SKB_FAIL(-EINVAL);
425 if (flags & MSG_TRUNC)
426 RCV_SKB_FAIL(-ECOMM);
428 type = nlh->nlmsg_type;
429 if (type < NLMSG_NOOP || type >= IPQM_MAX)
430 RCV_SKB_FAIL(-EINVAL);
432 if (type <= IPQM_BASE)
435 if (security_netlink_recv(skb, CAP_NET_ADMIN))
436 RCV_SKB_FAIL(-EPERM);
438 spin_lock_bh(&queue_lock);
441 if (peer_pid != pid) {
442 spin_unlock_bh(&queue_lock);
443 RCV_SKB_FAIL(-EBUSY);
446 net_enable_timestamp();
450 spin_unlock_bh(&queue_lock);
452 status = ipq_receive_peer(NLMSG_DATA(nlh), type,
453 nlmsglen - NLMSG_LENGTH(0));
455 RCV_SKB_FAIL(status);
457 if (flags & NLM_F_ACK)
458 netlink_ack(skb, nlh, 0);
462 ipq_rcv_skb(struct sk_buff *skb)
464 mutex_lock(&ipqnl_mutex);
466 mutex_unlock(&ipqnl_mutex);
470 ipq_rcv_dev_event(struct notifier_block *this,
471 unsigned long event, void *ptr)
473 struct net_device *dev = ptr;
475 if (!net_eq(dev_net(dev), &init_net))
478 /* Drop any packets associated with the downed device */
479 if (event == NETDEV_DOWN)
480 ipq_dev_drop(dev->ifindex);
484 static struct notifier_block ipq_dev_notifier = {
485 .notifier_call = ipq_rcv_dev_event,
489 ipq_rcv_nl_event(struct notifier_block *this,
490 unsigned long event, void *ptr)
492 struct netlink_notify *n = ptr;
494 if (event == NETLINK_URELEASE && n->protocol == NETLINK_FIREWALL) {
495 spin_lock_bh(&queue_lock);
496 if ((net_eq(n->net, &init_net)) && (n->pid == peer_pid))
498 spin_unlock_bh(&queue_lock);
503 static struct notifier_block ipq_nl_notifier = {
504 .notifier_call = ipq_rcv_nl_event,
508 static struct ctl_table_header *ipq_sysctl_header;
510 static ctl_table ipq_table[] = {
512 .procname = NET_IPQ_QMAX_NAME,
513 .data = &queue_maxlen,
514 .maxlen = sizeof(queue_maxlen),
516 .proc_handler = proc_dointvec
522 #ifdef CONFIG_PROC_FS
523 static int ip_queue_show(struct seq_file *m, void *v)
525 spin_lock_bh(&queue_lock);
531 "Queue length : %u\n"
532 "Queue max. length : %u\n"
533 "Queue dropped : %u\n"
534 "Netlink dropped : %u\n",
543 spin_unlock_bh(&queue_lock);
547 static int ip_queue_open(struct inode *inode, struct file *file)
549 return single_open(file, ip_queue_show, NULL);
552 static const struct file_operations ip_queue_proc_fops = {
553 .open = ip_queue_open,
556 .release = single_release,
557 .owner = THIS_MODULE,
561 static const struct nf_queue_handler nfqh = {
563 .outfn = &ipq_enqueue_packet,
566 static int __init ip_queue_init(void)
568 int status = -ENOMEM;
569 struct proc_dir_entry *proc __maybe_unused;
571 netlink_register_notifier(&ipq_nl_notifier);
572 ipqnl = netlink_kernel_create(&init_net, NETLINK_FIREWALL, 0,
573 ipq_rcv_skb, NULL, THIS_MODULE);
575 printk(KERN_ERR "ip_queue: failed to create netlink socket\n");
576 goto cleanup_netlink_notifier;
579 #ifdef CONFIG_PROC_FS
580 proc = proc_create(IPQ_PROC_FS_NAME, 0, init_net.proc_net,
581 &ip_queue_proc_fops);
583 printk(KERN_ERR "ip_queue: failed to create proc entry\n");
587 register_netdevice_notifier(&ipq_dev_notifier);
589 ipq_sysctl_header = register_sysctl_paths(net_ipv4_ctl_path, ipq_table);
591 status = nf_register_queue_handler(NFPROTO_IPV4, &nfqh);
593 printk(KERN_ERR "ip_queue: failed to register queue handler\n");
600 unregister_sysctl_table(ipq_sysctl_header);
602 unregister_netdevice_notifier(&ipq_dev_notifier);
603 proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
604 cleanup_ipqnl: __maybe_unused
605 netlink_kernel_release(ipqnl);
606 mutex_lock(&ipqnl_mutex);
607 mutex_unlock(&ipqnl_mutex);
609 cleanup_netlink_notifier:
610 netlink_unregister_notifier(&ipq_nl_notifier);
614 static void __exit ip_queue_fini(void)
616 nf_unregister_queue_handlers(&nfqh);
621 unregister_sysctl_table(ipq_sysctl_header);
623 unregister_netdevice_notifier(&ipq_dev_notifier);
624 proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
626 netlink_kernel_release(ipqnl);
627 mutex_lock(&ipqnl_mutex);
628 mutex_unlock(&ipqnl_mutex);
630 netlink_unregister_notifier(&ipq_nl_notifier);
633 MODULE_DESCRIPTION("IPv4 packet queue handler");
634 MODULE_AUTHOR("James Morris <jmorris@intercode.com.au>");
635 MODULE_LICENSE("GPL");
636 MODULE_ALIAS_NET_PF_PROTO(PF_NETLINK, NETLINK_FIREWALL);
638 module_init(ip_queue_init);
639 module_exit(ip_queue_fini);