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>
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
11 * 2000-03-27: Simplified code (thanks to Andi Kleen for clues).
12 * 2000-05-20: Fixed notifier problems (following Miguel Freitas' report).
13 * 2000-06-19: Fixed so nfmark is copied to metadata (reported by Sebastian
15 * 2000-08-01: Added Nick Williams' MAC support.
16 * 2002-06-25: Code cleanup.
19 #include <linux/module.h>
20 #include <linux/skbuff.h>
21 #include <linux/init.h>
23 #include <linux/notifier.h>
24 #include <linux/netdevice.h>
25 #include <linux/netfilter.h>
26 #include <linux/netfilter_ipv4/ip_queue.h>
27 #include <linux/netfilter_ipv4/ip_tables.h>
28 #include <linux/netlink.h>
29 #include <linux/spinlock.h>
30 #include <linux/sysctl.h>
31 #include <linux/proc_fs.h>
32 #include <linux/security.h>
34 #include <net/route.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"
47 struct ipq_queue_entry {
48 struct list_head list;
51 struct ipq_rt_info rt_info;
54 typedef int (*ipq_cmpfn)(struct ipq_queue_entry *, unsigned long);
56 static unsigned char copy_mode = IPQ_COPY_NONE;
57 static unsigned int queue_maxlen = IPQ_QMAX_DEFAULT;
58 static rwlock_t queue_lock = RW_LOCK_UNLOCKED;
60 static unsigned int copy_range;
61 static unsigned int queue_total;
62 static struct sock *ipqnl;
63 static LIST_HEAD(queue_list);
64 static DECLARE_MUTEX(ipqnl_sem);
67 ipq_issue_verdict(struct ipq_queue_entry *entry, int verdict)
69 nf_reinject(entry->skb, entry->info, verdict);
74 __ipq_enqueue_entry(struct ipq_queue_entry *entry)
76 if (queue_total >= queue_maxlen) {
78 printk(KERN_WARNING "ip_queue: full at %d entries, "
79 "dropping packet(s).\n", queue_total);
82 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 (copy_range == 0 || copy_range > entry->skb->len)
210 data_len = entry->skb->len;
212 data_len = copy_range;
214 size = NLMSG_SPACE(sizeof(*pmsg) + data_len);
219 read_unlock_bh(&queue_lock);
223 read_unlock_bh(&queue_lock);
225 skb = alloc_skb(size, GFP_ATOMIC);
230 nlh = NLMSG_PUT(skb, 0, 0, IPQM_PACKET, size - sizeof(*nlh));
231 pmsg = NLMSG_DATA(nlh);
232 memset(pmsg, 0, sizeof(*pmsg));
234 pmsg->packet_id = (unsigned long )entry;
235 pmsg->data_len = data_len;
236 pmsg->timestamp_sec = entry->skb->stamp.tv_sec;
237 pmsg->timestamp_usec = entry->skb->stamp.tv_usec;
238 pmsg->mark = entry->skb->nfmark;
239 pmsg->hook = entry->info->hook;
240 pmsg->hw_protocol = entry->skb->protocol;
242 if (entry->info->indev)
243 strcpy(pmsg->indev_name, entry->info->indev->name);
245 pmsg->indev_name[0] = '\0';
247 if (entry->info->outdev)
248 strcpy(pmsg->outdev_name, entry->info->outdev->name);
250 pmsg->outdev_name[0] = '\0';
252 if (entry->info->indev && entry->skb->dev) {
253 pmsg->hw_type = entry->skb->dev->type;
254 if (entry->skb->dev->hard_header_parse)
256 entry->skb->dev->hard_header_parse(entry->skb,
261 if (skb_copy_bits(entry->skb, 0, pmsg->payload, data_len))
264 nlh->nlmsg_len = skb->tail - old_tail;
271 printk(KERN_ERR "ip_queue: error creating packet message\n");
276 ipq_enqueue_packet(struct sk_buff *skb, struct nf_info *info, void *data)
278 int status = -EINVAL;
279 struct sk_buff *nskb;
280 struct ipq_queue_entry *entry;
282 if (copy_mode == IPQ_COPY_NONE)
285 entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
287 printk(KERN_ERR "ip_queue: OOM in ipq_enqueue_packet()\n");
294 if (entry->info->hook == NF_IP_LOCAL_OUT) {
295 struct iphdr *iph = skb->nh.iph;
297 entry->rt_info.tos = iph->tos;
298 entry->rt_info.daddr = iph->daddr;
299 entry->rt_info.saddr = iph->saddr;
302 nskb = ipq_build_packet_message(entry, &status);
306 write_lock_bh(&queue_lock);
309 goto err_out_free_nskb;
311 /* netlink_unicast will either free the nskb or attach it to a socket */
312 status = netlink_unicast(ipqnl, nskb, peer_pid, MSG_DONTWAIT);
316 status = __ipq_enqueue_entry(entry);
320 write_unlock_bh(&queue_lock);
327 write_unlock_bh(&queue_lock);
335 ipq_mangle_ipv4(ipq_verdict_msg_t *v, struct ipq_queue_entry *e)
338 struct iphdr *user_iph = (struct iphdr *)v->payload;
340 if (v->data_len < sizeof(*user_iph))
342 diff = v->data_len - e->skb->len;
344 skb_trim(e->skb, v->data_len);
346 if (v->data_len > 0xFFFF)
348 if (diff > skb_tailroom(e->skb)) {
349 struct sk_buff *newskb;
351 newskb = skb_copy_expand(e->skb,
352 skb_headroom(e->skb),
355 if (newskb == NULL) {
356 printk(KERN_WARNING "ip_queue: OOM "
357 "in mangle, dropping packet\n");
361 skb_set_owner_w(newskb, e->skb->sk);
365 skb_put(e->skb, diff);
367 if (!skb_ip_make_writable(&e->skb, v->data_len))
369 memcpy(e->skb->data, v->payload, v->data_len);
370 e->skb->nfcache |= NFC_ALTERED;
373 * Extra routing may needed on local out, as the QUEUE target never
374 * returns control to the table.
376 if (e->info->hook == NF_IP_LOCAL_OUT) {
377 struct iphdr *iph = e->skb->nh.iph;
379 if (!(iph->tos == e->rt_info.tos
380 && iph->daddr == e->rt_info.daddr
381 && iph->saddr == e->rt_info.saddr))
382 return ip_route_me_harder(&e->skb);
388 id_cmp(struct ipq_queue_entry *e, unsigned long id)
390 return (id == (unsigned long )e);
394 ipq_set_verdict(struct ipq_verdict_msg *vmsg, unsigned int len)
396 struct ipq_queue_entry *entry;
398 if (vmsg->value > NF_MAX_VERDICT)
401 entry = ipq_find_dequeue_entry(id_cmp, vmsg->id);
405 int verdict = vmsg->value;
407 if (vmsg->data_len && vmsg->data_len == len)
408 if (ipq_mangle_ipv4(vmsg, entry) < 0)
411 ipq_issue_verdict(entry, verdict);
417 ipq_set_mode(unsigned char mode, unsigned int range)
421 write_lock_bh(&queue_lock);
422 status = __ipq_set_mode(mode, range);
423 write_unlock_bh(&queue_lock);
428 ipq_receive_peer(struct ipq_peer_msg *pmsg,
429 unsigned char type, unsigned int len)
433 if (len < sizeof(*pmsg))
438 status = ipq_set_mode(pmsg->msg.mode.value,
439 pmsg->msg.mode.range);
443 if (pmsg->msg.verdict.value > NF_MAX_VERDICT)
446 status = ipq_set_verdict(&pmsg->msg.verdict,
447 len - sizeof(*pmsg));
456 dev_cmp(struct ipq_queue_entry *entry, unsigned long ifindex)
458 if (entry->info->indev)
459 if (entry->info->indev->ifindex == ifindex)
462 if (entry->info->outdev)
463 if (entry->info->outdev->ifindex == ifindex)
470 ipq_dev_drop(int ifindex)
472 struct ipq_queue_entry *entry;
474 while ((entry = ipq_find_dequeue_entry(dev_cmp, ifindex)) != NULL)
475 ipq_issue_verdict(entry, NF_DROP);
478 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
481 ipq_rcv_skb(struct sk_buff *skb)
483 int status, type, pid, flags, nlmsglen, skblen;
484 struct nlmsghdr *nlh;
487 if (skblen < sizeof(*nlh))
490 nlh = (struct nlmsghdr *)skb->data;
491 nlmsglen = nlh->nlmsg_len;
492 if (nlmsglen < sizeof(*nlh) || skblen < nlmsglen)
495 pid = nlh->nlmsg_pid;
496 flags = nlh->nlmsg_flags;
498 if(pid <= 0 || !(flags & NLM_F_REQUEST) || flags & NLM_F_MULTI)
499 RCV_SKB_FAIL(-EINVAL);
501 if (flags & MSG_TRUNC)
502 RCV_SKB_FAIL(-ECOMM);
504 type = nlh->nlmsg_type;
505 if (type < NLMSG_NOOP || type >= IPQM_MAX)
506 RCV_SKB_FAIL(-EINVAL);
508 if (type <= IPQM_BASE)
511 if (security_netlink_recv(skb))
512 RCV_SKB_FAIL(-EPERM);
514 write_lock_bh(&queue_lock);
517 if (peer_pid != pid) {
518 write_unlock_bh(&queue_lock);
519 RCV_SKB_FAIL(-EBUSY);
522 net_enable_timestamp();
526 write_unlock_bh(&queue_lock);
528 status = ipq_receive_peer(NLMSG_DATA(nlh), type,
529 skblen - NLMSG_LENGTH(0));
531 RCV_SKB_FAIL(status);
533 if (flags & NLM_F_ACK)
534 netlink_ack(skb, nlh, 0);
539 ipq_rcv_sk(struct sock *sk, int len)
544 if (down_trylock(&ipqnl_sem))
547 while ((skb = skb_dequeue(&sk->sk_receive_queue)) != NULL) {
554 } while (ipqnl && ipqnl->sk_receive_queue.qlen);
558 ipq_rcv_dev_event(struct notifier_block *this,
559 unsigned long event, void *ptr)
561 struct net_device *dev = ptr;
563 /* Drop any packets associated with the downed device */
564 if (event == NETDEV_DOWN)
565 ipq_dev_drop(dev->ifindex);
569 static struct notifier_block ipq_dev_notifier = {
570 .notifier_call = ipq_rcv_dev_event,
574 ipq_rcv_nl_event(struct notifier_block *this,
575 unsigned long event, void *ptr)
577 struct netlink_notify *n = ptr;
579 if (event == NETLINK_URELEASE &&
580 n->protocol == NETLINK_FIREWALL && n->pid) {
581 write_lock_bh(&queue_lock);
582 if (n->pid == peer_pid)
584 write_unlock_bh(&queue_lock);
589 static struct notifier_block ipq_nl_notifier = {
590 .notifier_call = ipq_rcv_nl_event,
593 static struct ctl_table_header *ipq_sysctl_header;
595 static ctl_table ipq_table[] = {
597 .ctl_name = NET_IPQ_QMAX,
598 .procname = NET_IPQ_QMAX_NAME,
599 .data = &queue_maxlen,
600 .maxlen = sizeof(queue_maxlen),
602 .proc_handler = proc_dointvec
607 static ctl_table ipq_dir_table[] = {
609 .ctl_name = NET_IPV4,
617 static ctl_table ipq_root_table[] = {
622 .child = ipq_dir_table
627 #ifdef CONFIG_PROC_FS
629 ipq_get_info(char *buffer, char **start, off_t offset, int length)
633 read_lock_bh(&queue_lock);
635 len = sprintf(buffer,
639 "Queue length : %u\n"
640 "Queue max. length : %u\n",
647 read_unlock_bh(&queue_lock);
649 *start = buffer + offset;
657 #endif /* CONFIG_PROC_FS */
660 init_or_cleanup(int init)
662 int status = -ENOMEM;
663 struct proc_dir_entry *proc;
668 netlink_register_notifier(&ipq_nl_notifier);
669 ipqnl = netlink_kernel_create(NETLINK_FIREWALL, ipq_rcv_sk);
671 printk(KERN_ERR "ip_queue: failed to create netlink socket\n");
672 goto cleanup_netlink_notifier;
675 proc = proc_net_create(IPQ_PROC_FS_NAME, 0, ipq_get_info);
677 proc->owner = THIS_MODULE;
679 printk(KERN_ERR "ip_queue: failed to create proc entry\n");
683 register_netdevice_notifier(&ipq_dev_notifier);
684 ipq_sysctl_header = register_sysctl_table(ipq_root_table, 0);
686 status = nf_register_queue_handler(PF_INET, ipq_enqueue_packet, NULL);
688 printk(KERN_ERR "ip_queue: failed to register queue handler\n");
694 nf_unregister_queue_handler(PF_INET);
699 unregister_sysctl_table(ipq_sysctl_header);
700 unregister_netdevice_notifier(&ipq_dev_notifier);
701 proc_net_remove(IPQ_PROC_FS_NAME);
704 sock_release(ipqnl->sk_socket);
708 cleanup_netlink_notifier:
709 netlink_unregister_notifier(&ipq_nl_notifier);
713 static int __init init(void)
716 return init_or_cleanup(1);
719 static void __exit fini(void)
724 MODULE_DESCRIPTION("IPv4 packet queue handler");
725 MODULE_AUTHOR("James Morris <jmorris@intercode.com.au>");
726 MODULE_LICENSE("GPL");