Merge git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6
[wrt350n-kernel.git] / net / ipv4 / netfilter / ip_queue.c
blobe1bf5d6f71891cf1bb89b7a312515a394fbfe78e
1 /*
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>
15 #include <linux/ip.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>
29 #include <net/sock.h>
30 #include <net/route.h>
31 #include <net/netfilter/nf_queue.h>
32 #include <net/ip.h>
34 #define IPQ_QMAX_DEFAULT 1024
35 #define IPQ_PROC_FS_NAME "ip_queue"
36 #define NET_IPQ_QMAX 2088
37 #define NET_IPQ_QMAX_NAME "ip_queue_maxlen"
39 typedef int (*ipq_cmpfn)(struct nf_queue_entry *, unsigned long);
41 static unsigned char copy_mode __read_mostly = IPQ_COPY_NONE;
42 static unsigned int queue_maxlen __read_mostly = IPQ_QMAX_DEFAULT;
43 static DEFINE_RWLOCK(queue_lock);
44 static int peer_pid __read_mostly;
45 static unsigned int copy_range __read_mostly;
46 static unsigned int queue_total;
47 static unsigned int queue_dropped = 0;
48 static unsigned int queue_user_dropped = 0;
49 static struct sock *ipqnl __read_mostly;
50 static LIST_HEAD(queue_list);
51 static DEFINE_MUTEX(ipqnl_mutex);
53 static inline void
54 __ipq_enqueue_entry(struct nf_queue_entry *entry)
56 list_add_tail(&entry->list, &queue_list);
57 queue_total++;
60 static inline int
61 __ipq_set_mode(unsigned char mode, unsigned int range)
63 int status = 0;
65 switch(mode) {
66 case IPQ_COPY_NONE:
67 case IPQ_COPY_META:
68 copy_mode = mode;
69 copy_range = 0;
70 break;
72 case IPQ_COPY_PACKET:
73 copy_mode = mode;
74 copy_range = range;
75 if (copy_range > 0xFFFF)
76 copy_range = 0xFFFF;
77 break;
79 default:
80 status = -EINVAL;
83 return status;
86 static void __ipq_flush(ipq_cmpfn cmpfn, unsigned long data);
88 static inline void
89 __ipq_reset(void)
91 peer_pid = 0;
92 net_disable_timestamp();
93 __ipq_set_mode(IPQ_COPY_NONE, 0);
94 __ipq_flush(NULL, 0);
97 static struct nf_queue_entry *
98 ipq_find_dequeue_entry(unsigned long id)
100 struct nf_queue_entry *entry = NULL, *i;
102 write_lock_bh(&queue_lock);
104 list_for_each_entry(i, &queue_list, list) {
105 if ((unsigned long)i == id) {
106 entry = i;
107 break;
111 if (entry) {
112 list_del(&entry->list);
113 queue_total--;
116 write_unlock_bh(&queue_lock);
117 return entry;
120 static void
121 __ipq_flush(ipq_cmpfn cmpfn, unsigned long data)
123 struct nf_queue_entry *entry, *next;
125 list_for_each_entry_safe(entry, next, &queue_list, list) {
126 if (!cmpfn || cmpfn(entry, data)) {
127 list_del(&entry->list);
128 queue_total--;
129 nf_reinject(entry, NF_DROP);
134 static void
135 ipq_flush(ipq_cmpfn cmpfn, unsigned long data)
137 write_lock_bh(&queue_lock);
138 __ipq_flush(cmpfn, data);
139 write_unlock_bh(&queue_lock);
142 static struct sk_buff *
143 ipq_build_packet_message(struct nf_queue_entry *entry, int *errp)
145 sk_buff_data_t old_tail;
146 size_t size = 0;
147 size_t data_len = 0;
148 struct sk_buff *skb;
149 struct ipq_packet_msg *pmsg;
150 struct nlmsghdr *nlh;
151 struct timeval tv;
153 read_lock_bh(&queue_lock);
155 switch (copy_mode) {
156 case IPQ_COPY_META:
157 case IPQ_COPY_NONE:
158 size = NLMSG_SPACE(sizeof(*pmsg));
159 data_len = 0;
160 break;
162 case IPQ_COPY_PACKET:
163 if ((entry->skb->ip_summed == CHECKSUM_PARTIAL ||
164 entry->skb->ip_summed == CHECKSUM_COMPLETE) &&
165 (*errp = skb_checksum_help(entry->skb))) {
166 read_unlock_bh(&queue_lock);
167 return NULL;
169 if (copy_range == 0 || copy_range > entry->skb->len)
170 data_len = entry->skb->len;
171 else
172 data_len = copy_range;
174 size = NLMSG_SPACE(sizeof(*pmsg) + data_len);
175 break;
177 default:
178 *errp = -EINVAL;
179 read_unlock_bh(&queue_lock);
180 return NULL;
183 read_unlock_bh(&queue_lock);
185 skb = alloc_skb(size, GFP_ATOMIC);
186 if (!skb)
187 goto nlmsg_failure;
189 old_tail = skb->tail;
190 nlh = NLMSG_PUT(skb, 0, 0, IPQM_PACKET, size - sizeof(*nlh));
191 pmsg = NLMSG_DATA(nlh);
192 memset(pmsg, 0, sizeof(*pmsg));
194 pmsg->packet_id = (unsigned long )entry;
195 pmsg->data_len = data_len;
196 tv = ktime_to_timeval(entry->skb->tstamp);
197 pmsg->timestamp_sec = tv.tv_sec;
198 pmsg->timestamp_usec = tv.tv_usec;
199 pmsg->mark = entry->skb->mark;
200 pmsg->hook = entry->hook;
201 pmsg->hw_protocol = entry->skb->protocol;
203 if (entry->indev)
204 strcpy(pmsg->indev_name, entry->indev->name);
205 else
206 pmsg->indev_name[0] = '\0';
208 if (entry->outdev)
209 strcpy(pmsg->outdev_name, entry->outdev->name);
210 else
211 pmsg->outdev_name[0] = '\0';
213 if (entry->indev && entry->skb->dev) {
214 pmsg->hw_type = entry->skb->dev->type;
215 pmsg->hw_addrlen = dev_parse_header(entry->skb,
216 pmsg->hw_addr);
219 if (data_len)
220 if (skb_copy_bits(entry->skb, 0, pmsg->payload, data_len))
221 BUG();
223 nlh->nlmsg_len = skb->tail - old_tail;
224 return skb;
226 nlmsg_failure:
227 if (skb)
228 kfree_skb(skb);
229 *errp = -EINVAL;
230 printk(KERN_ERR "ip_queue: error creating packet message\n");
231 return NULL;
234 static int
235 ipq_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
237 int status = -EINVAL;
238 struct sk_buff *nskb;
240 if (copy_mode == IPQ_COPY_NONE)
241 return -EAGAIN;
243 nskb = ipq_build_packet_message(entry, &status);
244 if (nskb == NULL)
245 return status;
247 write_lock_bh(&queue_lock);
249 if (!peer_pid)
250 goto err_out_free_nskb;
252 if (queue_total >= queue_maxlen) {
253 queue_dropped++;
254 status = -ENOSPC;
255 if (net_ratelimit())
256 printk (KERN_WARNING "ip_queue: full at %d entries, "
257 "dropping packets(s). Dropped: %d\n", queue_total,
258 queue_dropped);
259 goto err_out_free_nskb;
262 /* netlink_unicast will either free the nskb or attach it to a socket */
263 status = netlink_unicast(ipqnl, nskb, peer_pid, MSG_DONTWAIT);
264 if (status < 0) {
265 queue_user_dropped++;
266 goto err_out_unlock;
269 __ipq_enqueue_entry(entry);
271 write_unlock_bh(&queue_lock);
272 return status;
274 err_out_free_nskb:
275 kfree_skb(nskb);
277 err_out_unlock:
278 write_unlock_bh(&queue_lock);
279 return status;
282 static int
283 ipq_mangle_ipv4(ipq_verdict_msg_t *v, struct nf_queue_entry *e)
285 int diff;
286 <<<<<<< HEAD:net/ipv4/netfilter/ip_queue.c
287 int err;
288 =======
289 >>>>>>> 264e3e889d86e552b4191d69bb60f4f3b383135a:net/ipv4/netfilter/ip_queue.c
290 struct iphdr *user_iph = (struct iphdr *)v->payload;
291 <<<<<<< HEAD:net/ipv4/netfilter/ip_queue.c
292 =======
293 struct sk_buff *nskb;
294 >>>>>>> 264e3e889d86e552b4191d69bb60f4f3b383135a:net/ipv4/netfilter/ip_queue.c
296 if (v->data_len < sizeof(*user_iph))
297 return 0;
298 diff = v->data_len - e->skb->len;
299 if (diff < 0) {
300 if (pskb_trim(e->skb, v->data_len))
301 return -ENOMEM;
302 } else if (diff > 0) {
303 if (v->data_len > 0xFFFF)
304 return -EINVAL;
305 if (diff > skb_tailroom(e->skb)) {
306 <<<<<<< HEAD:net/ipv4/netfilter/ip_queue.c
307 err = pskb_expand_head(e->skb, 0,
308 =======
309 nskb = skb_copy_expand(e->skb, 0,
310 >>>>>>> 264e3e889d86e552b4191d69bb60f4f3b383135a:net/ipv4/netfilter/ip_queue.c
311 diff - skb_tailroom(e->skb),
312 GFP_ATOMIC);
313 <<<<<<< HEAD:net/ipv4/netfilter/ip_queue.c
314 if (err) {
315 =======
316 if (!nskb) {
317 >>>>>>> 264e3e889d86e552b4191d69bb60f4f3b383135a:net/ipv4/netfilter/ip_queue.c
318 printk(KERN_WARNING "ip_queue: error "
319 <<<<<<< HEAD:net/ipv4/netfilter/ip_queue.c
320 "in mangle, dropping packet: %d\n", -err);
321 return err;
322 =======
323 "in mangle, dropping packet\n");
324 return -ENOMEM;
325 >>>>>>> 264e3e889d86e552b4191d69bb60f4f3b383135a:net/ipv4/netfilter/ip_queue.c
327 <<<<<<< HEAD:net/ipv4/netfilter/ip_queue.c
328 =======
329 kfree_skb(e->skb);
330 e->skb = nskb;
331 >>>>>>> 264e3e889d86e552b4191d69bb60f4f3b383135a:net/ipv4/netfilter/ip_queue.c
333 skb_put(e->skb, diff);
335 if (!skb_make_writable(e->skb, v->data_len))
336 return -ENOMEM;
337 skb_copy_to_linear_data(e->skb, v->payload, v->data_len);
338 e->skb->ip_summed = CHECKSUM_NONE;
340 return 0;
343 static int
344 ipq_set_verdict(struct ipq_verdict_msg *vmsg, unsigned int len)
346 struct nf_queue_entry *entry;
348 if (vmsg->value > NF_MAX_VERDICT)
349 return -EINVAL;
351 entry = ipq_find_dequeue_entry(vmsg->id);
352 if (entry == NULL)
353 return -ENOENT;
354 else {
355 int verdict = vmsg->value;
357 if (vmsg->data_len && vmsg->data_len == len)
358 if (ipq_mangle_ipv4(vmsg, entry) < 0)
359 verdict = NF_DROP;
361 nf_reinject(entry, verdict);
362 return 0;
366 static int
367 ipq_set_mode(unsigned char mode, unsigned int range)
369 int status;
371 write_lock_bh(&queue_lock);
372 status = __ipq_set_mode(mode, range);
373 write_unlock_bh(&queue_lock);
374 return status;
377 static int
378 ipq_receive_peer(struct ipq_peer_msg *pmsg,
379 unsigned char type, unsigned int len)
381 int status = 0;
383 if (len < sizeof(*pmsg))
384 return -EINVAL;
386 switch (type) {
387 case IPQM_MODE:
388 status = ipq_set_mode(pmsg->msg.mode.value,
389 pmsg->msg.mode.range);
390 break;
392 case IPQM_VERDICT:
393 if (pmsg->msg.verdict.value > NF_MAX_VERDICT)
394 status = -EINVAL;
395 else
396 status = ipq_set_verdict(&pmsg->msg.verdict,
397 len - sizeof(*pmsg));
398 break;
399 default:
400 status = -EINVAL;
402 return status;
405 static int
406 dev_cmp(struct nf_queue_entry *entry, unsigned long ifindex)
408 if (entry->indev)
409 if (entry->indev->ifindex == ifindex)
410 return 1;
411 if (entry->outdev)
412 if (entry->outdev->ifindex == ifindex)
413 return 1;
414 #ifdef CONFIG_BRIDGE_NETFILTER
415 if (entry->skb->nf_bridge) {
416 if (entry->skb->nf_bridge->physindev &&
417 entry->skb->nf_bridge->physindev->ifindex == ifindex)
418 return 1;
419 if (entry->skb->nf_bridge->physoutdev &&
420 entry->skb->nf_bridge->physoutdev->ifindex == ifindex)
421 return 1;
423 #endif
424 return 0;
427 static void
428 ipq_dev_drop(int ifindex)
430 ipq_flush(dev_cmp, ifindex);
433 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
435 static inline void
436 __ipq_rcv_skb(struct sk_buff *skb)
438 int status, type, pid, flags, nlmsglen, skblen;
439 struct nlmsghdr *nlh;
441 skblen = skb->len;
442 if (skblen < sizeof(*nlh))
443 return;
445 nlh = nlmsg_hdr(skb);
446 nlmsglen = nlh->nlmsg_len;
447 if (nlmsglen < sizeof(*nlh) || skblen < nlmsglen)
448 return;
450 pid = nlh->nlmsg_pid;
451 flags = nlh->nlmsg_flags;
453 if(pid <= 0 || !(flags & NLM_F_REQUEST) || flags & NLM_F_MULTI)
454 RCV_SKB_FAIL(-EINVAL);
456 if (flags & MSG_TRUNC)
457 RCV_SKB_FAIL(-ECOMM);
459 type = nlh->nlmsg_type;
460 if (type < NLMSG_NOOP || type >= IPQM_MAX)
461 RCV_SKB_FAIL(-EINVAL);
463 if (type <= IPQM_BASE)
464 return;
466 if (security_netlink_recv(skb, CAP_NET_ADMIN))
467 RCV_SKB_FAIL(-EPERM);
469 write_lock_bh(&queue_lock);
471 if (peer_pid) {
472 if (peer_pid != pid) {
473 write_unlock_bh(&queue_lock);
474 RCV_SKB_FAIL(-EBUSY);
476 } else {
477 net_enable_timestamp();
478 peer_pid = pid;
481 write_unlock_bh(&queue_lock);
483 status = ipq_receive_peer(NLMSG_DATA(nlh), type,
484 nlmsglen - NLMSG_LENGTH(0));
485 if (status < 0)
486 RCV_SKB_FAIL(status);
488 if (flags & NLM_F_ACK)
489 netlink_ack(skb, nlh, 0);
490 return;
493 static void
494 ipq_rcv_skb(struct sk_buff *skb)
496 mutex_lock(&ipqnl_mutex);
497 __ipq_rcv_skb(skb);
498 mutex_unlock(&ipqnl_mutex);
501 static int
502 ipq_rcv_dev_event(struct notifier_block *this,
503 unsigned long event, void *ptr)
505 struct net_device *dev = ptr;
507 if (dev->nd_net != &init_net)
508 return NOTIFY_DONE;
510 /* Drop any packets associated with the downed device */
511 if (event == NETDEV_DOWN)
512 ipq_dev_drop(dev->ifindex);
513 return NOTIFY_DONE;
516 static struct notifier_block ipq_dev_notifier = {
517 .notifier_call = ipq_rcv_dev_event,
520 static int
521 ipq_rcv_nl_event(struct notifier_block *this,
522 unsigned long event, void *ptr)
524 struct netlink_notify *n = ptr;
526 if (event == NETLINK_URELEASE &&
527 n->protocol == NETLINK_FIREWALL && n->pid) {
528 write_lock_bh(&queue_lock);
529 if ((n->net == &init_net) && (n->pid == peer_pid))
530 __ipq_reset();
531 write_unlock_bh(&queue_lock);
533 return NOTIFY_DONE;
536 static struct notifier_block ipq_nl_notifier = {
537 .notifier_call = ipq_rcv_nl_event,
540 #ifdef CONFIG_SYSCTL
541 static struct ctl_table_header *ipq_sysctl_header;
543 static ctl_table ipq_table[] = {
545 .ctl_name = NET_IPQ_QMAX,
546 .procname = NET_IPQ_QMAX_NAME,
547 .data = &queue_maxlen,
548 .maxlen = sizeof(queue_maxlen),
549 .mode = 0644,
550 .proc_handler = proc_dointvec
552 { .ctl_name = 0 }
554 #endif
556 #ifdef CONFIG_PROC_FS
557 static int ip_queue_show(struct seq_file *m, void *v)
559 read_lock_bh(&queue_lock);
561 seq_printf(m,
562 "Peer PID : %d\n"
563 "Copy mode : %hu\n"
564 "Copy range : %u\n"
565 "Queue length : %u\n"
566 "Queue max. length : %u\n"
567 "Queue dropped : %u\n"
568 "Netlink dropped : %u\n",
569 peer_pid,
570 copy_mode,
571 copy_range,
572 queue_total,
573 queue_maxlen,
574 queue_dropped,
575 queue_user_dropped);
577 read_unlock_bh(&queue_lock);
578 return 0;
581 static int ip_queue_open(struct inode *inode, struct file *file)
583 return single_open(file, ip_queue_show, NULL);
586 static const struct file_operations ip_queue_proc_fops = {
587 .open = ip_queue_open,
588 .read = seq_read,
589 .llseek = seq_lseek,
590 .release = single_release,
591 .owner = THIS_MODULE,
593 #endif
595 static const struct nf_queue_handler nfqh = {
596 .name = "ip_queue",
597 .outfn = &ipq_enqueue_packet,
600 static int __init ip_queue_init(void)
602 int status = -ENOMEM;
603 struct proc_dir_entry *proc __maybe_unused;
605 netlink_register_notifier(&ipq_nl_notifier);
606 ipqnl = netlink_kernel_create(&init_net, NETLINK_FIREWALL, 0,
607 ipq_rcv_skb, NULL, THIS_MODULE);
608 if (ipqnl == NULL) {
609 printk(KERN_ERR "ip_queue: failed to create netlink socket\n");
610 goto cleanup_netlink_notifier;
613 #ifdef CONFIG_PROC_FS
614 proc = create_proc_entry(IPQ_PROC_FS_NAME, 0, init_net.proc_net);
615 if (proc) {
616 proc->owner = THIS_MODULE;
617 proc->proc_fops = &ip_queue_proc_fops;
618 } else {
619 printk(KERN_ERR "ip_queue: failed to create proc entry\n");
620 goto cleanup_ipqnl;
622 #endif
623 register_netdevice_notifier(&ipq_dev_notifier);
624 #ifdef CONFIG_SYSCTL
625 ipq_sysctl_header = register_sysctl_paths(net_ipv4_ctl_path, ipq_table);
626 #endif
627 status = nf_register_queue_handler(PF_INET, &nfqh);
628 if (status < 0) {
629 printk(KERN_ERR "ip_queue: failed to register queue handler\n");
630 goto cleanup_sysctl;
632 return status;
634 cleanup_sysctl:
635 #ifdef CONFIG_SYSCTL
636 unregister_sysctl_table(ipq_sysctl_header);
637 #endif
638 unregister_netdevice_notifier(&ipq_dev_notifier);
639 proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
640 cleanup_ipqnl: __maybe_unused
641 netlink_kernel_release(ipqnl);
642 mutex_lock(&ipqnl_mutex);
643 mutex_unlock(&ipqnl_mutex);
645 cleanup_netlink_notifier:
646 netlink_unregister_notifier(&ipq_nl_notifier);
647 return status;
650 static void __exit ip_queue_fini(void)
652 nf_unregister_queue_handlers(&nfqh);
653 synchronize_net();
654 ipq_flush(NULL, 0);
656 #ifdef CONFIG_SYSCTL
657 unregister_sysctl_table(ipq_sysctl_header);
658 #endif
659 unregister_netdevice_notifier(&ipq_dev_notifier);
660 proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
662 netlink_kernel_release(ipqnl);
663 mutex_lock(&ipqnl_mutex);
664 mutex_unlock(&ipqnl_mutex);
666 netlink_unregister_notifier(&ipq_nl_notifier);
669 MODULE_DESCRIPTION("IPv4 packet queue handler");
670 MODULE_AUTHOR("James Morris <jmorris@intercode.com.au>");
671 MODULE_LICENSE("GPL");
673 module_init(ip_queue_init);
674 module_exit(ip_queue_fini);