]> pilppa.org Git - linux-2.6-omap-h63xx.git/blob - net/ipv6/netfilter/ip6_queue.c
[NETFILTER]: nf_queue: move queueing related functions/struct to seperate header
[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/seq_file.h>
27 #include <linux/mutex.h>
28 #include <net/net_namespace.h>
29 #include <net/sock.h>
30 #include <net/ipv6.h>
31 #include <net/ip6_route.h>
32 #include <net/netfilter/nf_queue.h>
33 #include <linux/netfilter_ipv4/ip_queue.h>
34 #include <linux/netfilter_ipv4/ip_tables.h>
35 #include <linux/netfilter_ipv6/ip6_tables.h>
36
37 #define IPQ_QMAX_DEFAULT 1024
38 #define IPQ_PROC_FS_NAME "ip6_queue"
39 #define NET_IPQ_QMAX 2088
40 #define NET_IPQ_QMAX_NAME "ip6_queue_maxlen"
41
42 struct ipq_queue_entry {
43         struct list_head list;
44         struct nf_info *info;
45         struct sk_buff *skb;
46 };
47
48 typedef int (*ipq_cmpfn)(struct ipq_queue_entry *, unsigned long);
49
50 static unsigned char copy_mode __read_mostly = IPQ_COPY_NONE;
51 static unsigned int queue_maxlen __read_mostly = IPQ_QMAX_DEFAULT;
52 static DEFINE_RWLOCK(queue_lock);
53 static int peer_pid __read_mostly;
54 static unsigned int copy_range __read_mostly;
55 static unsigned int queue_total;
56 static unsigned int queue_dropped = 0;
57 static unsigned int queue_user_dropped = 0;
58 static struct sock *ipqnl __read_mostly;
59 static LIST_HEAD(queue_list);
60 static DEFINE_MUTEX(ipqnl_mutex);
61
62 static void
63 ipq_issue_verdict(struct ipq_queue_entry *entry, int verdict)
64 {
65         local_bh_disable();
66         nf_reinject(entry->skb, entry->info, verdict);
67         local_bh_enable();
68         kfree(entry);
69 }
70
71 static inline void
72 __ipq_enqueue_entry(struct ipq_queue_entry *entry)
73 {
74        list_add(&entry->list, &queue_list);
75        queue_total++;
76 }
77
78 /*
79  * Find and return a queued entry matched by cmpfn, or return the last
80  * entry if cmpfn is NULL.
81  */
82 static inline struct ipq_queue_entry *
83 __ipq_find_entry(ipq_cmpfn cmpfn, unsigned long data)
84 {
85         struct list_head *p;
86
87         list_for_each_prev(p, &queue_list) {
88                 struct ipq_queue_entry *entry = (struct ipq_queue_entry *)p;
89
90                 if (!cmpfn || cmpfn(entry, data))
91                         return entry;
92         }
93         return NULL;
94 }
95
96 static inline void
97 __ipq_dequeue_entry(struct ipq_queue_entry *entry)
98 {
99         list_del(&entry->list);
100         queue_total--;
101 }
102
103 static inline struct ipq_queue_entry *
104 __ipq_find_dequeue_entry(ipq_cmpfn cmpfn, unsigned long data)
105 {
106         struct ipq_queue_entry *entry;
107
108         entry = __ipq_find_entry(cmpfn, data);
109         if (entry == NULL)
110                 return NULL;
111
112         __ipq_dequeue_entry(entry);
113         return entry;
114 }
115
116
117 static inline void
118 __ipq_flush(int verdict)
119 {
120         struct ipq_queue_entry *entry;
121
122         while ((entry = __ipq_find_dequeue_entry(NULL, 0)))
123                 ipq_issue_verdict(entry, verdict);
124 }
125
126 static inline int
127 __ipq_set_mode(unsigned char mode, unsigned int range)
128 {
129         int status = 0;
130
131         switch(mode) {
132         case IPQ_COPY_NONE:
133         case IPQ_COPY_META:
134                 copy_mode = mode;
135                 copy_range = 0;
136                 break;
137
138         case IPQ_COPY_PACKET:
139                 copy_mode = mode;
140                 copy_range = range;
141                 if (copy_range > 0xFFFF)
142                         copy_range = 0xFFFF;
143                 break;
144
145         default:
146                 status = -EINVAL;
147
148         }
149         return status;
150 }
151
152 static inline void
153 __ipq_reset(void)
154 {
155         peer_pid = 0;
156         net_disable_timestamp();
157         __ipq_set_mode(IPQ_COPY_NONE, 0);
158         __ipq_flush(NF_DROP);
159 }
160
161 static struct ipq_queue_entry *
162 ipq_find_dequeue_entry(ipq_cmpfn cmpfn, unsigned long data)
163 {
164         struct ipq_queue_entry *entry;
165
166         write_lock_bh(&queue_lock);
167         entry = __ipq_find_dequeue_entry(cmpfn, data);
168         write_unlock_bh(&queue_lock);
169         return entry;
170 }
171
172 static void
173 ipq_flush(int verdict)
174 {
175         write_lock_bh(&queue_lock);
176         __ipq_flush(verdict);
177         write_unlock_bh(&queue_lock);
178 }
179
180 static struct sk_buff *
181 ipq_build_packet_message(struct ipq_queue_entry *entry, int *errp)
182 {
183         sk_buff_data_t old_tail;
184         size_t size = 0;
185         size_t data_len = 0;
186         struct sk_buff *skb;
187         struct ipq_packet_msg *pmsg;
188         struct nlmsghdr *nlh;
189         struct timeval tv;
190
191         read_lock_bh(&queue_lock);
192
193         switch (copy_mode) {
194         case IPQ_COPY_META:
195         case IPQ_COPY_NONE:
196                 size = NLMSG_SPACE(sizeof(*pmsg));
197                 data_len = 0;
198                 break;
199
200         case IPQ_COPY_PACKET:
201                 if ((entry->skb->ip_summed == CHECKSUM_PARTIAL ||
202                      entry->skb->ip_summed == CHECKSUM_COMPLETE) &&
203                     (*errp = skb_checksum_help(entry->skb))) {
204                         read_unlock_bh(&queue_lock);
205                         return NULL;
206                 }
207                 if (copy_range == 0 || copy_range > entry->skb->len)
208                         data_len = entry->skb->len;
209                 else
210                         data_len = copy_range;
211
212                 size = NLMSG_SPACE(sizeof(*pmsg) + data_len);
213                 break;
214
215         default:
216                 *errp = -EINVAL;
217                 read_unlock_bh(&queue_lock);
218                 return NULL;
219         }
220
221         read_unlock_bh(&queue_lock);
222
223         skb = alloc_skb(size, GFP_ATOMIC);
224         if (!skb)
225                 goto nlmsg_failure;
226
227         old_tail = skb->tail;
228         nlh = NLMSG_PUT(skb, 0, 0, IPQM_PACKET, size - sizeof(*nlh));
229         pmsg = NLMSG_DATA(nlh);
230         memset(pmsg, 0, sizeof(*pmsg));
231
232         pmsg->packet_id       = (unsigned long )entry;
233         pmsg->data_len        = data_len;
234         tv = ktime_to_timeval(entry->skb->tstamp);
235         pmsg->timestamp_sec   = tv.tv_sec;
236         pmsg->timestamp_usec  = tv.tv_usec;
237         pmsg->mark            = entry->skb->mark;
238         pmsg->hook            = entry->info->hook;
239         pmsg->hw_protocol     = entry->skb->protocol;
240
241         if (entry->info->indev)
242                 strcpy(pmsg->indev_name, entry->info->indev->name);
243         else
244                 pmsg->indev_name[0] = '\0';
245
246         if (entry->info->outdev)
247                 strcpy(pmsg->outdev_name, entry->info->outdev->name);
248         else
249                 pmsg->outdev_name[0] = '\0';
250
251         if (entry->info->indev && entry->skb->dev) {
252                 pmsg->hw_type = entry->skb->dev->type;
253                 pmsg->hw_addrlen = dev_parse_header(entry->skb, pmsg->hw_addr);
254         }
255
256         if (data_len)
257                 if (skb_copy_bits(entry->skb, 0, pmsg->payload, data_len))
258                         BUG();
259
260         nlh->nlmsg_len = skb->tail - old_tail;
261         return skb;
262
263 nlmsg_failure:
264         if (skb)
265                 kfree_skb(skb);
266         *errp = -EINVAL;
267         printk(KERN_ERR "ip6_queue: error creating packet message\n");
268         return NULL;
269 }
270
271 static int
272 ipq_enqueue_packet(struct sk_buff *skb, struct nf_info *info,
273                    unsigned int queuenum)
274 {
275         int status = -EINVAL;
276         struct sk_buff *nskb;
277         struct ipq_queue_entry *entry;
278
279         if (copy_mode == IPQ_COPY_NONE)
280                 return -EAGAIN;
281
282         entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
283         if (entry == NULL) {
284                 printk(KERN_ERR "ip6_queue: OOM in ipq_enqueue_packet()\n");
285                 return -ENOMEM;
286         }
287
288         entry->info = info;
289         entry->skb = skb;
290
291         nskb = ipq_build_packet_message(entry, &status);
292         if (nskb == NULL)
293                 goto err_out_free;
294
295         write_lock_bh(&queue_lock);
296
297         if (!peer_pid)
298                 goto err_out_free_nskb;
299
300         if (queue_total >= queue_maxlen) {
301                 queue_dropped++;
302                 status = -ENOSPC;
303                 if (net_ratelimit())
304                         printk (KERN_WARNING "ip6_queue: fill at %d entries, "
305                                 "dropping packet(s).  Dropped: %d\n", queue_total,
306                                 queue_dropped);
307                 goto err_out_free_nskb;
308         }
309
310         /* netlink_unicast will either free the nskb or attach it to a socket */
311         status = netlink_unicast(ipqnl, nskb, peer_pid, MSG_DONTWAIT);
312         if (status < 0) {
313                 queue_user_dropped++;
314                 goto err_out_unlock;
315         }
316
317         __ipq_enqueue_entry(entry);
318
319         write_unlock_bh(&queue_lock);
320         return status;
321
322 err_out_free_nskb:
323         kfree_skb(nskb);
324
325 err_out_unlock:
326         write_unlock_bh(&queue_lock);
327
328 err_out_free:
329         kfree(entry);
330         return status;
331 }
332
333 static int
334 ipq_mangle_ipv6(ipq_verdict_msg_t *v, struct ipq_queue_entry *e)
335 {
336         int diff;
337         int err;
338         struct ipv6hdr *user_iph = (struct ipv6hdr *)v->payload;
339
340         if (v->data_len < sizeof(*user_iph))
341                 return 0;
342         diff = v->data_len - e->skb->len;
343         if (diff < 0) {
344                 if (pskb_trim(e->skb, v->data_len))
345                         return -ENOMEM;
346         } else if (diff > 0) {
347                 if (v->data_len > 0xFFFF)
348                         return -EINVAL;
349                 if (diff > skb_tailroom(e->skb)) {
350                         err = pskb_expand_head(e->skb, 0,
351                                                diff - skb_tailroom(e->skb),
352                                                GFP_ATOMIC);
353                         if (err) {
354                                 printk(KERN_WARNING "ip6_queue: OOM "
355                                       "in mangle, dropping packet\n");
356                                 return err;
357                         }
358                 }
359                 skb_put(e->skb, diff);
360         }
361         if (!skb_make_writable(e->skb, v->data_len))
362                 return -ENOMEM;
363         skb_copy_to_linear_data(e->skb, v->payload, v->data_len);
364         e->skb->ip_summed = CHECKSUM_NONE;
365
366         return 0;
367 }
368
369 static inline int
370 id_cmp(struct ipq_queue_entry *e, unsigned long id)
371 {
372         return (id == (unsigned long )e);
373 }
374
375 static int
376 ipq_set_verdict(struct ipq_verdict_msg *vmsg, unsigned int len)
377 {
378         struct ipq_queue_entry *entry;
379
380         if (vmsg->value > NF_MAX_VERDICT)
381                 return -EINVAL;
382
383         entry = ipq_find_dequeue_entry(id_cmp, vmsg->id);
384         if (entry == NULL)
385                 return -ENOENT;
386         else {
387                 int verdict = vmsg->value;
388
389                 if (vmsg->data_len && vmsg->data_len == len)
390                         if (ipq_mangle_ipv6(vmsg, entry) < 0)
391                                 verdict = NF_DROP;
392
393                 ipq_issue_verdict(entry, verdict);
394                 return 0;
395         }
396 }
397
398 static int
399 ipq_set_mode(unsigned char mode, unsigned int range)
400 {
401         int status;
402
403         write_lock_bh(&queue_lock);
404         status = __ipq_set_mode(mode, range);
405         write_unlock_bh(&queue_lock);
406         return status;
407 }
408
409 static int
410 ipq_receive_peer(struct ipq_peer_msg *pmsg,
411                  unsigned char type, unsigned int len)
412 {
413         int status = 0;
414
415         if (len < sizeof(*pmsg))
416                 return -EINVAL;
417
418         switch (type) {
419         case IPQM_MODE:
420                 status = ipq_set_mode(pmsg->msg.mode.value,
421                                       pmsg->msg.mode.range);
422                 break;
423
424         case IPQM_VERDICT:
425                 if (pmsg->msg.verdict.value > NF_MAX_VERDICT)
426                         status = -EINVAL;
427                 else
428                         status = ipq_set_verdict(&pmsg->msg.verdict,
429                                                  len - sizeof(*pmsg));
430                         break;
431         default:
432                 status = -EINVAL;
433         }
434         return status;
435 }
436
437 static int
438 dev_cmp(struct ipq_queue_entry *entry, unsigned long ifindex)
439 {
440         if (entry->info->indev)
441                 if (entry->info->indev->ifindex == ifindex)
442                         return 1;
443
444         if (entry->info->outdev)
445                 if (entry->info->outdev->ifindex == ifindex)
446                         return 1;
447
448         return 0;
449 }
450
451 static void
452 ipq_dev_drop(int ifindex)
453 {
454         struct ipq_queue_entry *entry;
455
456         while ((entry = ipq_find_dequeue_entry(dev_cmp, ifindex)) != NULL)
457                 ipq_issue_verdict(entry, NF_DROP);
458 }
459
460 #define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
461
462 static inline void
463 __ipq_rcv_skb(struct sk_buff *skb)
464 {
465         int status, type, pid, flags, nlmsglen, skblen;
466         struct nlmsghdr *nlh;
467
468         skblen = skb->len;
469         if (skblen < sizeof(*nlh))
470                 return;
471
472         nlh = nlmsg_hdr(skb);
473         nlmsglen = nlh->nlmsg_len;
474         if (nlmsglen < sizeof(*nlh) || skblen < nlmsglen)
475                 return;
476
477         pid = nlh->nlmsg_pid;
478         flags = nlh->nlmsg_flags;
479
480         if(pid <= 0 || !(flags & NLM_F_REQUEST) || flags & NLM_F_MULTI)
481                 RCV_SKB_FAIL(-EINVAL);
482
483         if (flags & MSG_TRUNC)
484                 RCV_SKB_FAIL(-ECOMM);
485
486         type = nlh->nlmsg_type;
487         if (type < NLMSG_NOOP || type >= IPQM_MAX)
488                 RCV_SKB_FAIL(-EINVAL);
489
490         if (type <= IPQM_BASE)
491                 return;
492
493         if (security_netlink_recv(skb, CAP_NET_ADMIN))
494                 RCV_SKB_FAIL(-EPERM);
495
496         write_lock_bh(&queue_lock);
497
498         if (peer_pid) {
499                 if (peer_pid != pid) {
500                         write_unlock_bh(&queue_lock);
501                         RCV_SKB_FAIL(-EBUSY);
502                 }
503         } else {
504                 net_enable_timestamp();
505                 peer_pid = pid;
506         }
507
508         write_unlock_bh(&queue_lock);
509
510         status = ipq_receive_peer(NLMSG_DATA(nlh), type,
511                                   nlmsglen - NLMSG_LENGTH(0));
512         if (status < 0)
513                 RCV_SKB_FAIL(status);
514
515         if (flags & NLM_F_ACK)
516                 netlink_ack(skb, nlh, 0);
517         return;
518 }
519
520 static void
521 ipq_rcv_skb(struct sk_buff *skb)
522 {
523         mutex_lock(&ipqnl_mutex);
524         __ipq_rcv_skb(skb);
525         mutex_unlock(&ipqnl_mutex);
526 }
527
528 static int
529 ipq_rcv_dev_event(struct notifier_block *this,
530                   unsigned long event, void *ptr)
531 {
532         struct net_device *dev = ptr;
533
534         if (dev->nd_net != &init_net)
535                 return NOTIFY_DONE;
536
537         /* Drop any packets associated with the downed device */
538         if (event == NETDEV_DOWN)
539                 ipq_dev_drop(dev->ifindex);
540         return NOTIFY_DONE;
541 }
542
543 static struct notifier_block ipq_dev_notifier = {
544         .notifier_call  = ipq_rcv_dev_event,
545 };
546
547 static int
548 ipq_rcv_nl_event(struct notifier_block *this,
549                  unsigned long event, void *ptr)
550 {
551         struct netlink_notify *n = ptr;
552
553         if (event == NETLINK_URELEASE &&
554             n->protocol == NETLINK_IP6_FW && n->pid) {
555                 write_lock_bh(&queue_lock);
556                 if ((n->net == &init_net) && (n->pid == peer_pid))
557                         __ipq_reset();
558                 write_unlock_bh(&queue_lock);
559         }
560         return NOTIFY_DONE;
561 }
562
563 static struct notifier_block ipq_nl_notifier = {
564         .notifier_call  = ipq_rcv_nl_event,
565 };
566
567 static struct ctl_table_header *ipq_sysctl_header;
568
569 static ctl_table ipq_table[] = {
570         {
571                 .ctl_name       = NET_IPQ_QMAX,
572                 .procname       = NET_IPQ_QMAX_NAME,
573                 .data           = &queue_maxlen,
574                 .maxlen         = sizeof(queue_maxlen),
575                 .mode           = 0644,
576                 .proc_handler   = proc_dointvec
577         },
578         { .ctl_name = 0 }
579 };
580
581 static ctl_table ipq_dir_table[] = {
582         {
583                 .ctl_name       = NET_IPV6,
584                 .procname       = "ipv6",
585                 .mode           = 0555,
586                 .child          = ipq_table
587         },
588         { .ctl_name = 0 }
589 };
590
591 static ctl_table ipq_root_table[] = {
592         {
593                 .ctl_name       = CTL_NET,
594                 .procname       = "net",
595                 .mode           = 0555,
596                 .child          = ipq_dir_table
597         },
598         { .ctl_name = 0 }
599 };
600
601 static int ip6_queue_show(struct seq_file *m, void *v)
602 {
603         read_lock_bh(&queue_lock);
604
605         seq_printf(m,
606                       "Peer PID          : %d\n"
607                       "Copy mode         : %hu\n"
608                       "Copy range        : %u\n"
609                       "Queue length      : %u\n"
610                       "Queue max. length : %u\n"
611                       "Queue dropped     : %u\n"
612                       "Netfilter dropped : %u\n",
613                       peer_pid,
614                       copy_mode,
615                       copy_range,
616                       queue_total,
617                       queue_maxlen,
618                       queue_dropped,
619                       queue_user_dropped);
620
621         read_unlock_bh(&queue_lock);
622         return 0;
623 }
624
625 static int ip6_queue_open(struct inode *inode, struct file *file)
626 {
627         return single_open(file, ip6_queue_show, NULL);
628 }
629
630 static const struct file_operations ip6_queue_proc_fops = {
631         .open           = ip6_queue_open,
632         .read           = seq_read,
633         .llseek         = seq_lseek,
634         .release        = single_release,
635         .owner          = THIS_MODULE,
636 };
637
638 static const struct nf_queue_handler nfqh = {
639         .name   = "ip6_queue",
640         .outfn  = &ipq_enqueue_packet,
641 };
642
643 static int __init ip6_queue_init(void)
644 {
645         int status = -ENOMEM;
646         struct proc_dir_entry *proc;
647
648         netlink_register_notifier(&ipq_nl_notifier);
649         ipqnl = netlink_kernel_create(&init_net, NETLINK_IP6_FW, 0,
650                                       ipq_rcv_skb, NULL, THIS_MODULE);
651         if (ipqnl == NULL) {
652                 printk(KERN_ERR "ip6_queue: failed to create netlink socket\n");
653                 goto cleanup_netlink_notifier;
654         }
655
656         proc = create_proc_entry(IPQ_PROC_FS_NAME, 0, init_net.proc_net);
657         if (proc) {
658                 proc->owner = THIS_MODULE;
659                 proc->proc_fops = &ip6_queue_proc_fops;
660         } else {
661                 printk(KERN_ERR "ip6_queue: failed to create proc entry\n");
662                 goto cleanup_ipqnl;
663         }
664
665         register_netdevice_notifier(&ipq_dev_notifier);
666         ipq_sysctl_header = register_sysctl_table(ipq_root_table);
667
668         status = nf_register_queue_handler(PF_INET6, &nfqh);
669         if (status < 0) {
670                 printk(KERN_ERR "ip6_queue: failed to register queue handler\n");
671                 goto cleanup_sysctl;
672         }
673         return status;
674
675 cleanup_sysctl:
676         unregister_sysctl_table(ipq_sysctl_header);
677         unregister_netdevice_notifier(&ipq_dev_notifier);
678         proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
679
680 cleanup_ipqnl:
681         sock_release(ipqnl->sk_socket);
682         mutex_lock(&ipqnl_mutex);
683         mutex_unlock(&ipqnl_mutex);
684
685 cleanup_netlink_notifier:
686         netlink_unregister_notifier(&ipq_nl_notifier);
687         return status;
688 }
689
690 static void __exit ip6_queue_fini(void)
691 {
692         nf_unregister_queue_handlers(&nfqh);
693         synchronize_net();
694         ipq_flush(NF_DROP);
695
696         unregister_sysctl_table(ipq_sysctl_header);
697         unregister_netdevice_notifier(&ipq_dev_notifier);
698         proc_net_remove(&init_net, IPQ_PROC_FS_NAME);
699
700         sock_release(ipqnl->sk_socket);
701         mutex_lock(&ipqnl_mutex);
702         mutex_unlock(&ipqnl_mutex);
703
704         netlink_unregister_notifier(&ipq_nl_notifier);
705 }
706
707 MODULE_DESCRIPTION("IPv6 packet queue handler");
708 MODULE_LICENSE("GPL");
709
710 module_init(ip6_queue_init);
711 module_exit(ip6_queue_fini);