]> pilppa.org Git - linux-2.6-omap-h63xx.git/blob - net/ipv6/netfilter/ip6_queue.c
Merge branch 'locks' of git://linux-nfs.org/~bfields/linux
[linux-2.6-omap-h63xx.git] / net / ipv6 / netfilter / ip6_queue.c
1 /*
2  * This is a module which is used for queueing IPv6 packets and
3  * communicating with userspace via netlink.
4  *
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
10  *
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.
14  */
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/mutex.h>
27 #include <net/net_namespace.h>
28 #include <net/sock.h>
29 #include <net/ipv6.h>
30 #include <net/ip6_route.h>
31 #include <linux/netfilter_ipv4/ip_queue.h>
32 #include <linux/netfilter_ipv4/ip_tables.h>
33 #include <linux/netfilter_ipv6/ip6_tables.h>
34
35 #define IPQ_QMAX_DEFAULT 1024
36 #define IPQ_PROC_FS_NAME "ip6_queue"
37 #define NET_IPQ_QMAX 2088
38 #define NET_IPQ_QMAX_NAME "ip6_queue_maxlen"
39
40 struct ipq_queue_entry {
41         struct list_head list;
42         struct nf_info *info;
43         struct sk_buff *skb;
44 };
45
46 typedef int (*ipq_cmpfn)(struct ipq_queue_entry *, unsigned long);
47
48 static unsigned char copy_mode __read_mostly = IPQ_COPY_NONE;
49 static unsigned int queue_maxlen __read_mostly = IPQ_QMAX_DEFAULT;
50 static DEFINE_RWLOCK(queue_lock);
51 static int peer_pid __read_mostly;
52 static unsigned int copy_range __read_mostly;
53 static unsigned int queue_total;
54 static unsigned int queue_dropped = 0;
55 static unsigned int queue_user_dropped = 0;
56 static struct sock *ipqnl __read_mostly;
57 static LIST_HEAD(queue_list);
58 static DEFINE_MUTEX(ipqnl_mutex);
59
60 static void
61 ipq_issue_verdict(struct ipq_queue_entry *entry, int verdict)
62 {
63         local_bh_disable();
64         nf_reinject(entry->skb, entry->info, verdict);
65         local_bh_enable();
66         kfree(entry);
67 }
68
69 static inline void
70 __ipq_enqueue_entry(struct ipq_queue_entry *entry)
71 {
72        list_add(&entry->list, &queue_list);
73        queue_total++;
74 }
75
76 /*
77  * Find and return a queued entry matched by cmpfn, or return the last
78  * entry if cmpfn is NULL.
79  */
80 static inline struct ipq_queue_entry *
81 __ipq_find_entry(ipq_cmpfn cmpfn, unsigned long data)
82 {
83         struct list_head *p;
84
85         list_for_each_prev(p, &queue_list) {
86                 struct ipq_queue_entry *entry = (struct ipq_queue_entry *)p;
87
88                 if (!cmpfn || cmpfn(entry, data))
89                         return entry;
90         }
91         return NULL;
92 }
93
94 static inline void
95 __ipq_dequeue_entry(struct ipq_queue_entry *entry)
96 {
97         list_del(&entry->list);
98         queue_total--;
99 }
100
101 static inline struct ipq_queue_entry *
102 __ipq_find_dequeue_entry(ipq_cmpfn cmpfn, unsigned long data)
103 {
104         struct ipq_queue_entry *entry;
105
106         entry = __ipq_find_entry(cmpfn, data);
107         if (entry == NULL)
108                 return NULL;
109
110         __ipq_dequeue_entry(entry);
111         return entry;
112 }
113
114
115 static inline void
116 __ipq_flush(int verdict)
117 {
118         struct ipq_queue_entry *entry;
119
120         while ((entry = __ipq_find_dequeue_entry(NULL, 0)))
121                 ipq_issue_verdict(entry, verdict);
122 }
123
124 static inline int
125 __ipq_set_mode(unsigned char mode, unsigned int range)
126 {
127         int status = 0;
128
129         switch(mode) {
130         case IPQ_COPY_NONE:
131         case IPQ_COPY_META:
132                 copy_mode = mode;
133                 copy_range = 0;
134                 break;
135
136         case IPQ_COPY_PACKET:
137                 copy_mode = mode;
138                 copy_range = range;
139                 if (copy_range > 0xFFFF)
140                         copy_range = 0xFFFF;
141                 break;
142
143         default:
144                 status = -EINVAL;
145
146         }
147         return status;
148 }
149
150 static inline void
151 __ipq_reset(void)
152 {
153         peer_pid = 0;
154         net_disable_timestamp();
155         __ipq_set_mode(IPQ_COPY_NONE, 0);
156         __ipq_flush(NF_DROP);
157 }
158
159 static struct ipq_queue_entry *
160 ipq_find_dequeue_entry(ipq_cmpfn cmpfn, unsigned long data)
161 {
162         struct ipq_queue_entry *entry;
163
164         write_lock_bh(&queue_lock);
165         entry = __ipq_find_dequeue_entry(cmpfn, data);
166         write_unlock_bh(&queue_lock);
167         return entry;
168 }
169
170 static void
171 ipq_flush(int verdict)
172 {
173         write_lock_bh(&queue_lock);
174         __ipq_flush(verdict);
175         write_unlock_bh(&queue_lock);
176 }
177
178 static struct sk_buff *
179 ipq_build_packet_message(struct ipq_queue_entry *entry, int *errp)
180 {
181         sk_buff_data_t old_tail;
182         size_t size = 0;
183         size_t data_len = 0;
184         struct sk_buff *skb;
185         struct ipq_packet_msg *pmsg;
186         struct nlmsghdr *nlh;
187         struct timeval tv;
188
189         read_lock_bh(&queue_lock);
190
191         switch (copy_mode) {
192         case IPQ_COPY_META:
193         case IPQ_COPY_NONE:
194                 size = NLMSG_SPACE(sizeof(*pmsg));
195                 data_len = 0;
196                 break;
197
198         case IPQ_COPY_PACKET:
199                 if ((entry->skb->ip_summed == CHECKSUM_PARTIAL ||
200                      entry->skb->ip_summed == CHECKSUM_COMPLETE) &&
201                     (*errp = skb_checksum_help(entry->skb))) {
202                         read_unlock_bh(&queue_lock);
203                         return NULL;
204                 }
205                 if (copy_range == 0 || copy_range > entry->skb->len)
206                         data_len = entry->skb->len;
207                 else
208                         data_len = copy_range;
209
210                 size = NLMSG_SPACE(sizeof(*pmsg) + data_len);
211                 break;
212
213         default:
214                 *errp = -EINVAL;
215                 read_unlock_bh(&queue_lock);
216                 return NULL;
217         }
218
219         read_unlock_bh(&queue_lock);
220
221         skb = alloc_skb(size, GFP_ATOMIC);
222         if (!skb)
223                 goto nlmsg_failure;
224
225         old_tail = skb->tail;
226         nlh = NLMSG_PUT(skb, 0, 0, IPQM_PACKET, size - sizeof(*nlh));
227         pmsg = NLMSG_DATA(nlh);
228         memset(pmsg, 0, sizeof(*pmsg));
229
230         pmsg->packet_id       = (unsigned long )entry;
231         pmsg->data_len        = data_len;
232         tv = ktime_to_timeval(entry->skb->tstamp);
233         pmsg->timestamp_sec   = tv.tv_sec;
234         pmsg->timestamp_usec  = tv.tv_usec;
235         pmsg->mark            = entry->skb->mark;
236         pmsg->hook            = entry->info->hook;
237         pmsg->hw_protocol     = entry->skb->protocol;
238
239         if (entry->info->indev)
240                 strcpy(pmsg->indev_name, entry->info->indev->name);
241         else
242                 pmsg->indev_name[0] = '\0';
243
244         if (entry->info->outdev)
245                 strcpy(pmsg->outdev_name, entry->info->outdev->name);
246         else
247                 pmsg->outdev_name[0] = '\0';
248
249         if (entry->info->indev && entry->skb->dev) {
250                 pmsg->hw_type = entry->skb->dev->type;
251                 pmsg->hw_addrlen = dev_parse_header(entry->skb, pmsg->hw_addr);
252         }
253
254         if (data_len)
255                 if (skb_copy_bits(entry->skb, 0, pmsg->payload, data_len))
256                         BUG();
257
258         nlh->nlmsg_len = skb->tail - old_tail;
259         return skb;
260
261 nlmsg_failure:
262         if (skb)
263                 kfree_skb(skb);
264         *errp = -EINVAL;
265         printk(KERN_ERR "ip6_queue: error creating packet message\n");
266         return NULL;
267 }
268
269 static int
270 ipq_enqueue_packet(struct sk_buff *skb, struct nf_info *info,
271                    unsigned int queuenum, void *data)
272 {
273         int status = -EINVAL;
274         struct sk_buff *nskb;
275         struct ipq_queue_entry *entry;
276
277         if (copy_mode == IPQ_COPY_NONE)
278                 return -EAGAIN;
279
280         entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
281         if (entry == NULL) {
282                 printk(KERN_ERR "ip6_queue: OOM in ipq_enqueue_packet()\n");
283                 return -ENOMEM;
284         }
285
286         entry->info = info;
287         entry->skb = skb;
288
289         nskb = ipq_build_packet_message(entry, &status);
290         if (nskb == NULL)
291                 goto err_out_free;
292
293         write_lock_bh(&queue_lock);
294
295         if (!peer_pid)
296                 goto err_out_free_nskb;
297
298         if (queue_total >= queue_maxlen) {
299                 queue_dropped++;
300                 status = -ENOSPC;
301                 if (net_ratelimit())
302                         printk (KERN_WARNING "ip6_queue: fill at %d entries, "
303                                 "dropping packet(s).  Dropped: %d\n", queue_total,
304                                 queue_dropped);
305                 goto err_out_free_nskb;
306         }
307
308         /* netlink_unicast will either free the nskb or attach it to a socket */
309         status = netlink_unicast(ipqnl, nskb, peer_pid, MSG_DONTWAIT);
310         if (status < 0) {
311                 queue_user_dropped++;
312                 goto err_out_unlock;
313         }
314
315         __ipq_enqueue_entry(entry);
316
317         write_unlock_bh(&queue_lock);
318         return status;
319
320 err_out_free_nskb:
321         kfree_skb(nskb);
322
323 err_out_unlock:
324         write_unlock_bh(&queue_lock);
325
326 err_out_free:
327         kfree(entry);
328         return status;
329 }
330
331 static int
332 ipq_mangle_ipv6(ipq_verdict_msg_t *v, struct ipq_queue_entry *e)
333 {
334         int diff;
335         int err;
336         struct ipv6hdr *user_iph = (struct ipv6hdr *)v->payload;
337
338         if (v->data_len < sizeof(*user_iph))
339                 return 0;
340         diff = v->data_len - e->skb->len;
341         if (diff < 0) {
342                 if (pskb_trim(e->skb, v->data_len))
343                         return -ENOMEM;
344         } else if (diff > 0) {
345                 if (v->data_len > 0xFFFF)
346                         return -EINVAL;
347                 if (diff > skb_tailroom(e->skb)) {
348                         err = pskb_expand_head(e->skb, 0,
349                                                diff - skb_tailroom(e->skb),
350                                                GFP_ATOMIC);
351                         if (err) {
352                                 printk(KERN_WARNING "ip6_queue: OOM "
353                                       "in mangle, dropping packet\n");
354                                 return err;
355                         }
356                 }
357                 skb_put(e->skb, diff);
358         }
359         if (!skb_make_writable(e->skb, v->data_len))
360                 return -ENOMEM;
361         skb_copy_to_linear_data(e->skb, v->payload, v->data_len);
362         e->skb->ip_summed = CHECKSUM_NONE;
363
364         return 0;
365 }
366
367 static inline int
368 id_cmp(struct ipq_queue_entry *e, unsigned long id)
369 {
370         return (id == (unsigned long )e);
371 }
372
373 static int
374 ipq_set_verdict(struct ipq_verdict_msg *vmsg, unsigned int len)
375 {
376         struct ipq_queue_entry *entry;
377
378         if (vmsg->value > NF_MAX_VERDICT)
379                 return -EINVAL;
380
381         entry = ipq_find_dequeue_entry(id_cmp, vmsg->id);
382         if (entry == NULL)
383                 return -ENOENT;
384         else {
385                 int verdict = vmsg->value;
386
387                 if (vmsg->data_len && vmsg->data_len == len)
388                         if (ipq_mangle_ipv6(vmsg, entry) < 0)
389                                 verdict = NF_DROP;
390
391                 ipq_issue_verdict(entry, verdict);
392                 return 0;
393         }
394 }
395
396 static int
397 ipq_set_mode(unsigned char mode, unsigned int range)
398 {
399         int status;
400
401         write_lock_bh(&queue_lock);
402         status = __ipq_set_mode(mode, range);
403         write_unlock_bh(&queue_lock);
404         return status;
405 }
406
407 static int
408 ipq_receive_peer(struct ipq_peer_msg *pmsg,
409                  unsigned char type, unsigned int len)
410 {
411         int status = 0;
412
413         if (len < sizeof(*pmsg))
414                 return -EINVAL;
415
416         switch (type) {
417         case IPQM_MODE:
418                 status = ipq_set_mode(pmsg->msg.mode.value,
419                                       pmsg->msg.mode.range);
420                 break;
421
422         case IPQM_VERDICT:
423                 if (pmsg->msg.verdict.value > NF_MAX_VERDICT)
424                         status = -EINVAL;
425                 else
426                         status = ipq_set_verdict(&pmsg->msg.verdict,
427                                                  len - sizeof(*pmsg));
428                         break;
429         default:
430                 status = -EINVAL;
431         }
432         return status;
433 }
434
435 static int
436 dev_cmp(struct ipq_queue_entry *entry, unsigned long ifindex)
437 {
438         if (entry->info->indev)
439                 if (entry->info->indev->ifindex == ifindex)
440                         return 1;
441
442         if (entry->info->outdev)
443                 if (entry->info->outdev->ifindex == ifindex)
444                         return 1;
445
446         return 0;
447 }
448
449 static void
450 ipq_dev_drop(int ifindex)
451 {
452         struct ipq_queue_entry *entry;
453
454         while ((entry = ipq_find_dequeue_entry(dev_cmp, ifindex)) != NULL)
455                 ipq_issue_verdict(entry, NF_DROP);
456 }
457
458 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
459
460 static inline void
461 __ipq_rcv_skb(struct sk_buff *skb)
462 {
463         int status, type, pid, flags, nlmsglen, skblen;
464         struct nlmsghdr *nlh;
465
466         skblen = skb->len;
467         if (skblen < sizeof(*nlh))
468                 return;
469
470         nlh = nlmsg_hdr(skb);
471         nlmsglen = nlh->nlmsg_len;
472         if (nlmsglen < sizeof(*nlh) || skblen < nlmsglen)
473                 return;
474
475         pid = nlh->nlmsg_pid;
476         flags = nlh->nlmsg_flags;
477
478         if(pid <= 0 || !(flags & NLM_F_REQUEST) || flags & NLM_F_MULTI)
479                 RCV_SKB_FAIL(-EINVAL);
480
481         if (flags & MSG_TRUNC)
482                 RCV_SKB_FAIL(-ECOMM);
483
484         type = nlh->nlmsg_type;
485         if (type < NLMSG_NOOP || type >= IPQM_MAX)
486                 RCV_SKB_FAIL(-EINVAL);
487
488         if (type <= IPQM_BASE)
489                 return;
490
491         if (security_netlink_recv(skb, CAP_NET_ADMIN))
492                 RCV_SKB_FAIL(-EPERM);
493
494         write_lock_bh(&queue_lock);
495
496         if (peer_pid) {
497                 if (peer_pid != pid) {
498                         write_unlock_bh(&queue_lock);
499                         RCV_SKB_FAIL(-EBUSY);
500                 }
501         } else {
502                 net_enable_timestamp();
503                 peer_pid = pid;
504         }
505
506         write_unlock_bh(&queue_lock);
507
508         status = ipq_receive_peer(NLMSG_DATA(nlh), type,
509                                   nlmsglen - NLMSG_LENGTH(0));
510         if (status < 0)
511                 RCV_SKB_FAIL(status);
512
513         if (flags & NLM_F_ACK)
514                 netlink_ack(skb, nlh, 0);
515         return;
516 }
517
518 static void
519 ipq_rcv_skb(struct sk_buff *skb)
520 {
521         mutex_lock(&ipqnl_mutex);
522         __ipq_rcv_skb(skb);
523         mutex_unlock(&ipqnl_mutex);
524 }
525
526 static int
527 ipq_rcv_dev_event(struct notifier_block *this,
528                   unsigned long event, void *ptr)
529 {
530         struct net_device *dev = ptr;
531
532         if (dev->nd_net != &init_net)
533                 return NOTIFY_DONE;
534
535         /* Drop any packets associated with the downed device */
536         if (event == NETDEV_DOWN)
537                 ipq_dev_drop(dev->ifindex);
538         return NOTIFY_DONE;
539 }
540
541 static struct notifier_block ipq_dev_notifier = {
542         .notifier_call  = ipq_rcv_dev_event,
543 };
544
545 static int
546 ipq_rcv_nl_event(struct notifier_block *this,
547                  unsigned long event, void *ptr)
548 {
549         struct netlink_notify *n = ptr;
550
551         if (event == NETLINK_URELEASE &&
552             n->protocol == NETLINK_IP6_FW && n->pid) {
553                 write_lock_bh(&queue_lock);
554                 if ((n->net == &init_net) && (n->pid == peer_pid))
555                         __ipq_reset();
556                 write_unlock_bh(&queue_lock);
557         }
558         return NOTIFY_DONE;
559 }
560
561 static struct notifier_block ipq_nl_notifier = {
562         .notifier_call  = ipq_rcv_nl_event,
563 };
564
565 static struct ctl_table_header *ipq_sysctl_header;
566
567 static ctl_table ipq_table[] = {
568         {
569                 .ctl_name       = NET_IPQ_QMAX,
570                 .procname       = NET_IPQ_QMAX_NAME,
571                 .data           = &queue_maxlen,
572                 .maxlen         = sizeof(queue_maxlen),
573                 .mode           = 0644,
574                 .proc_handler   = proc_dointvec
575         },
576         { .ctl_name = 0 }
577 };
578
579 static ctl_table ipq_dir_table[] = {
580         {
581                 .ctl_name       = NET_IPV6,
582                 .procname       = "ipv6",
583                 .mode           = 0555,
584                 .child          = ipq_table
585         },
586         { .ctl_name = 0 }
587 };
588
589 static ctl_table ipq_root_table[] = {
590         {
591                 .ctl_name       = CTL_NET,
592                 .procname       = "net",
593                 .mode           = 0555,
594                 .child          = ipq_dir_table
595         },
596         { .ctl_name = 0 }
597 };
598
599 #ifdef CONFIG_PROC_FS
600 static int
601 ipq_get_info(char *buffer, char **start, off_t offset, int length)
602 {
603         int len;
604
605         read_lock_bh(&queue_lock);
606
607         len = sprintf(buffer,
608                       "Peer PID          : %d\n"
609                       "Copy mode         : %hu\n"
610                       "Copy range        : %u\n"
611                       "Queue length      : %u\n"
612                       "Queue max. length : %u\n"
613                       "Queue dropped     : %u\n"
614                       "Netfilter dropped : %u\n",
615                       peer_pid,
616                       copy_mode,
617                       copy_range,
618                       queue_total,
619                       queue_maxlen,
620                       queue_dropped,
621                       queue_user_dropped);
622
623         read_unlock_bh(&queue_lock);
624
625         *start = buffer + offset;
626         len -= offset;
627         if (len > length)
628                 len = length;
629         else if (len < 0)
630                 len = 0;
631         return len;
632 }
633 #endif /* CONFIG_PROC_FS */
634
635 static struct nf_queue_handler nfqh = {
636         .name   = "ip6_queue",
637         .outfn  = &ipq_enqueue_packet,
638 };
639
640 static int __init ip6_queue_init(void)
641 {
642         int status = -ENOMEM;
643         struct proc_dir_entry *proc;
644
645         netlink_register_notifier(&ipq_nl_notifier);
646         ipqnl = netlink_kernel_create(&init_net, NETLINK_IP6_FW, 0,
647                                       ipq_rcv_skb, NULL, THIS_MODULE);
648         if (ipqnl == NULL) {
649                 printk(KERN_ERR "ip6_queue: failed to create netlink socket\n");
650                 goto cleanup_netlink_notifier;
651         }
652
653         proc = proc_net_create(&init_net, IPQ_PROC_FS_NAME, 0, ipq_get_info);
654         if (proc)
655                 proc->owner = THIS_MODULE;
656         else {
657                 printk(KERN_ERR "ip6_queue: failed to create proc entry\n");
658                 goto cleanup_ipqnl;
659         }
660
661         register_netdevice_notifier(&ipq_dev_notifier);
662         ipq_sysctl_header = register_sysctl_table(ipq_root_table);
663
664         status = nf_register_queue_handler(PF_INET6, &nfqh);
665         if (status < 0) {
666                 printk(KERN_ERR "ip6_queue: failed to register queue handler\n");
667                 goto cleanup_sysctl;
668         }
669         return status;
670
671 cleanup_sysctl:
672         unregister_sysctl_table(ipq_sysctl_header);
673         unregister_netdevice_notifier(&ipq_dev_notifier);
674         proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
675
676 cleanup_ipqnl:
677         sock_release(ipqnl->sk_socket);
678         mutex_lock(&ipqnl_mutex);
679         mutex_unlock(&ipqnl_mutex);
680
681 cleanup_netlink_notifier:
682         netlink_unregister_notifier(&ipq_nl_notifier);
683         return status;
684 }
685
686 static void __exit ip6_queue_fini(void)
687 {
688         nf_unregister_queue_handlers(&nfqh);
689         synchronize_net();
690         ipq_flush(NF_DROP);
691
692         unregister_sysctl_table(ipq_sysctl_header);
693         unregister_netdevice_notifier(&ipq_dev_notifier);
694         proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
695
696         sock_release(ipqnl->sk_socket);
697         mutex_lock(&ipqnl_mutex);
698         mutex_unlock(&ipqnl_mutex);
699
700         netlink_unregister_notifier(&ipq_nl_notifier);
701 }
702
703 MODULE_DESCRIPTION("IPv6 packet queue handler");
704 MODULE_LICENSE("GPL");
705
706 module_init(ip6_queue_init);
707 module_exit(ip6_queue_fini);